The Subcommittee on Standards and Security of the National Committee on Vital and Health Statistics (NCVHS) held hearings on August 28-29, 2002, at the Hubert H. Humphrey Building in Washington, D.C. The meeting was open to the public. Present:
Subcommittee members
Staff and Liaisons
Others
On August 28-29 the Subcommittee on Standards and Security held the first of several hearings planned for 2002 and the first-half of 2003 on Patient Medical Record Information (PMRI) terminologies. The Subcommittee's intent is to make recommendations next year to the Secretary for next steps related to national standards in this area. The first day's focus was on the scope of these efforts and criteria; the second day included discussions with public, private, and federal consortia on issues of interoperability, PMRI standards, and terminologies as well as further discussions about issues related to replacement of ICD-9-CM with ICD-10-CM and ICD-10-PCS, recommendations for next steps, and a white paper the Subcommittee prepared to the Secretary concerning Health Insurance Portability and Accountability Act (HIPAA) issues and Diagnostic and Statistical Manual, version 4 (DSM-4). During the first day, the Subcommittee heard 14 presentations and held discussions on terminology standards. The second day, eight presenters testified on coordination of ongoing national healthcare standards initiatives, standards implementation in the United Kingdom (U.K.), and transitioning to new code sets
Agenda Item: Overview
Mr. Blair noted the testifiers would provide feedback on the Subcommittee's proposed scope and whether the criteria for selection were appropriate, pragmatic, realistic and sufficient. Both were derived from the Committee's July 2000 report to the Secretary that set forth priorities for going forward with PMRI standard recommendations and a set of 10 recommendations including the selection of standards. The report also addressed accelerating development of PMRI standards and guiding principles for selection as well as other guidances the Subcommittee applied to PMRI terminology. In February 2002, the Committee applied these guiding principles to the selection of PMRI message format standards and made recommendations.
Panel 1: Terminology
Dr. Chute applauded the preliminary criteria and validatedthe Committee's efforts on creating multi-purpose roles, recognizing terminologies, outlining reasonable starting domains, and collating common criteria. He reviewed motivational issues associated with development of terminology in patient medical records supporting his recommendations for augmenting the criteria and scope. Noting a major purpose of a terminology was to capture symptoms, findings, events and interventions with sufficient detail and specificity to use them in decision support, information retrieval and recognition, Dr. Chute emphasized keeping these use cases separate from their format and structure. In parallel with detailed specific nomenclatures, he noted the need to aggregate this information characteristically by way of high-level classifications, thus enabling accurate, reproducible groupings of information for various purposes, including fiscal reimbursement management issues. Dr. Chute proposed rethinking the notion of a continuum of terminology from a detailed nomenclature to a classification into a more broadly based one-size-no-longer-fit-all approach.
Noting the idea that an underlying nomenclature could have detailed specific patient information and extend along a common continuum to higher level classifications or groups had been perceived as a fairly linear, uniform and single-dimensional problem, Dr. Chute proposed a more complex use case with needs in decision support and error detection, public health and surveillance, reimbursement and management, outcome and epidemiology. He suggested publishing proposed aggregation logics as a component of their attached classifications. Dr. Chute concluded that, whatever they did, they'd confront significant challenges. He said the impact on vendors would be considerable and everyone would have to select from what was available today or supporting development along lines of content, authoring environments, and newly described aggregation and description logics.
American Nurses Association (ANA) recommended renaming the "Procedures" category "Interventions" and categorizing terminologies by multiple dimensions accounting for domain coverage, primary purpose and the terminologies' underlying semantic structure. Dr. Bakken identified an expanded list of data nursing codes and advised that the criteria for selection of PMRI terminologies be tailored to the terminologies and include documented testing of reliability, validity and clinical utility. She emphasized clarifying criteria related to each type of semantic structure, especially clinically specific (concept-oriented) terminologies. Dr. Bakken stressed the importance of identifying terminology characteristics or abilities from those provided by models or supporting software. And she noted that a reference terminology model (ISO, 2002) that integrated nursing concepts and enabled ANA to meet the notion of finely detailed clinical specifications and rules engines that support the description logic was proposed as a draft International Standard.
Dr. Price said the U.K.'s National Health Service (NHS) had taken more centralist control of implementation to achieve aggressive targets for 2005. NHS had remained committed to a single comprehensive scheme (i.e., SNOMED CT) since 1992 and invested $3,000,000 in 2002 to lay the foundation for implementation. Dr. Price offered suggestions for classification of terminologies that fit NHS' worldview and noted NHS believed the provision of extension, subset and mapping mechanisms addressed counter-arguments to comprehensive schemes. Dr. Price said a global business model needed to be open, contributory, non-predatory and underpinned by central (government) support. He emphasized that terminology wasn't an end in itself, but was needed to add value.
Dr. Nelson addressed a number of areas that needed strengthening (terminologies, atomic concepts, relationships, priorities, pricing and the maintenance process) and encouraged the Subcommittee to recognize distinct uses of the terminology rather than focus on a continuum. He encouraged them to consider terminologies in terms of the level of discourse and areas covered. Observing that the level of discourse determined atomicity, Dr. Nelson noted the difficulty of defining atomic concepts when the level kept changing and suggested what they sought was a post-coordinated expression for anything greater than their current level of discourse. He emphasized that terminologies had to have representational integrity and that the biggest priority was patient safety.
Medicomp's biggest challenge was the human interface and concept processing at a clinical, rather than strictly terminological level that accommodated the clinical thought process without interfering with it. He said a hierarchy of pre-coordinated concepts enabled assignment of predictable properties and that the combination of fully expressed clinical concepts and predictable properties gave software developers the means to build applications for clinical documentation and decision support at the point of care. Mr. Lareau urged the Subcommittee to accommodate terminology needs of healthcare providers at the point of care. He emphasized the importance in defining the scope of PMRI terminologies to consider the role of the patient. Considering priorities for the selection process, Mr. Lareau noted much of the healthcare reimbursement structure was based on diagnosis, procedure, drug and laboratory codes, and changes in terminology requirements for these systems might be significant. He said terminologies directly affecting current, widely implemented processes should have highest priority to avoid push back from vendors. Mr. Lareau suggested the Subcommittee consider that terminologies might require fully expressed combinatorial clinical concepts in order to be actionable at the point of care and shouldn't be limited to purely atomic terms.
Panel 2: Terminology
Considering PMRI terminologies, Dr. Berglund noted convergence could imply combining multiple systems (doubtful at this time) or enabling interworking between systems and interoperation. Noting code sets weren't the same as data sets and message formats differed from codes, he advised considering them separately, based on their criteria. Noting HL7 and X12N weren't code sets, but specifications for messaging formats, he suggested distinguishing them from codes in terminologies. Dr. Berglund identified other groups and terminologies for the Subcommittee to consider. Noting standards for messaging, diagnosis, and procedure codes were specified and widely used, he advised giving them priority for PMRI selection. He urged expeditiously specifying the move to ICD-10-CM. Citing the need for clinically specific codes to represent more detailed information, he recommended giving them first priority among groups without specified standards. Drug and nursing codes were also needed. A single standard might not be feasible and continued use of different systems might be reasonable. Discussing changes to preliminary criteria for selection of PMRI terminologies, Dr. Berglund said different criteria were needed for different groups or subsets; selection criteria for messaging formats and data sets should differ from those for terminologies, and classification systems intended to aggregate should have different criteria than clinically specific code systems.
Dr. Brown said the Department of Veterans Affairs (DVA) found the organization of some of the PMRI terminologies subsets variable and suggested a domain orientation for a schema. He also questioned the "convergence" subset. DVA believed a single terminology wouldn't serve all needs and convergence was more likely to occur as incremental steps among interlocking terminologies with complementary content. Noting the administration's need for DVA-specific codes, Dr. Brown said DVA and Department of Defense (DoD) had standards that could be integrated into the proposed interlocking set of complimentary terminologies and he recommended that resources be devoted to developing those code sets as needed. DVA believed interlocking terminology sets relating to the domains of medications, laboratory and problems should have highest priority. Dr. Brown identified additional criteria for selecting PMRI terminologies. DVA didn't believe market place acceptance equated to the types of aggregation or quality needed at the point of contact, and new developments should be encouraged if they supported decision-making, patient care quality and safety. Discussing the relationship to other terminologies, Dr. Brown said semi- and automated mapping were the goal and explicit semantic models were a step toward achieving that mapping. He noted aggregation's importance in managing assets and doing population studies. Dr. Mike Lincoln added that mapping was critically important for the DVA's HDR program.
Capt. Kelly said DoD could live with the grouping of the subsets of PMRI terminologies, but he added that grouping weren't as critical as the need to play well together. He noted that DoD used CDT-3 as its dental standard and would prefer CDT-3 to CDP-2. And he pointed out to the Subcommittee that the lack of an accepted standard for identifying individuals and their key role (patient, provider) impeded progress toward implementing a PMRI. While DoD had difficulty understanding what was meant by the convergence code set, Capt. Kelly agreed that message and clinically specific code sets were most important. DoD encouraged use of repeatable model-based processes and tools to create and define terminologies, concepts, and relationships between concepts. Capt. Kelly said relationships among concepts should not only be explicitly defined, internally consistent, and non-redundant--but, also, model based. DoD encouraged future development of complex terminology using model-based development. While DoD didn't believe the ANSI umbrella of organizations was essential, Capt. Kelly said it added credibility and structure and DoD supported the ANSI consensus process for voting/balloting on the acceptance of new versions of a standard.
Observing that the categorizations of the terminologies the testifiers were asked to talk about seemed self-expressed by the terminology developers, Dr. Elkin suggested a more appropriate way. He noted five speakers alluded to similar issues and their need for these levels of details for underlying clinical knowledge, aggregations for DRGs, clinical codes, ICD and a network view. Dr. Elkin said he saw that network view as the decision support that indicated where on the linear scale one wished to be. He noted a tradeoff between complexity and expressivity with compositional systems. Dr. Elkin recommended adding several new terminologies. He noted that clinical terminologies were becoming robust enough for clinical use. And he pointed out that there was an ANSI standard, ASTM E2087, for creating the right sets of quality indicators for determining PMRI terminologies.
Dr. Elkin emphasized that terminologies should be evaluated within their purpose and scope. He noted the importance of knowing what terminologies were pre-mapped to, could and couldn't map to. He said normalization of content assured that the same terms or concepts within the terminology would be identified. Normalization of semantics assured there wasn't semantic overlap between semantic operators, preventing semantic drift. With multiple hierarchies, it was important to determine if consistency of view supported explicit uncertainty. How evolving definitions were supported over time also was important to meaning; so was ensuring that representational form didn't influence the usefulness of the terminology or hierarchies. Dr. Elkin also posed a number of questions around purpose and scope.
Dr. Spackman concurred that, from a user perspective, there were two main ways to categorize PMRI terminologies: comprehensive, integrated clinical terminologies and other clinical terminologies of limited scope. Users needed a comprehensive, integrated whole. Their requirements should be prioritized and terminologies should be grouped according to that perspective. Recalling testifiers didn't understand "convergence," but advocated interlocking, Dr. Spackman suggested they were the same: work that had to be done to meet the user requirement for interoperability. He outlined five main categories of user requirements for PMRI terminology and said they couldn't be met by merely a collection of complementary PMRI terminologies but required an integrated whole. He identified a number of reasons why integration was necessary. Considering cost as a criterion, Dr. Spackman noted that without an integrated terminology, the opportunity and incremental costs of system development and maintenance were orders of magnitude greater. The quality of health care was linked to an ability to mobilize technology and an assumption of interoperability that they didn't have without an integrated standard PMRI terminology.(Vendor opinion -- delete)
Dr. Spackman illustrated with a diagram how SNOMED CT served as a convergent terminology and others either fully or partially integrated or mapped from it. He acknowledged that SNOMED needed to be integrated with the core reference terminology. He agreed on the need to map core clinical concepts to classifications and proposed building subsets and extensions to support linkage or modeling of domain- and use-limited terminologies. He urged the Committee to focus on continuous improvement and deployment of a core. Dr. Spackman told the Committee it could make major steps forward if it required a proven maintenance infrastructure, prevented analysis paralysis and disruption of a forward trend with industry and providers. The marketplace was moving to deploy SNOMED CT and he said clinicians, researchers and the public could reap the benefits.
Panel 3: Terminology
Dr. Campbell recounted how 10 years ago the Institute of Medicine portrayed the computer-based patient record, a vision resting upon a model with deliverables requiring coded or control terminology for functionality. Considerable work was done on messaging interoperability, and associated reference terminologies and technologies emerged. Dr. Campbell hoped they'd discuss today the path to achieving that vision, by creating what the Subcommittee termed semantic interoperability. Dr. Campbell suggested the path began by looking not just at available codes, but the architectures employing them. He encouraged considering adopting a model comprised of three layers, the core being the clinical and the outer layer the external or public/ government layer. The core was a reference terminology and its deliverables were decision support that could be shared. The second layer was comprised of modeled code, legacy term sets, and other systems necessary to the clinical function of the CPR, modeled on a negotiated basis with a clinical reference core, within the construction of an information model. The third layer was administrative and governmental layer, with interfaces and a variety of users with important but primarily non-clinical needs for CPR data.
The core convergence reference terminology was designed to be comprehensive across everything needed for multidisciplinary clinical care. Its primary editorial policies were clinical. Preferably, it was maintained by an SDO with clinical professional societies overseeing its sophisticated clinical content. Layer two's features were modeled codes that integrated important clinical terminology and served departmental, special purpose, messaging systems, professional and interface terminologies. Features were constructed using the reference core, but modeled into the context and behavior required based upon the original legacy owner's meaning values. A dialogue between the reference terminology and modeled code sets was essential for a successful model. Layer three was code sets required for administration, finances, billing and epidemiology. It followed editing rules promulgated by organizations needing them, was maintained in a different way, and held to other standards. Dr. Campbell proposed three major priorities: (1) a collaborative strategy to define a core reference terminology to serve clinical medicine, (2) establishing knowledge-based solid mapping from the reference core to the third layer, and (3) instituting changes in clinical content coverage within layer two. Dr. Campbell offered observations about criteria for code or terminology selection in each layer. And he noted that, from an administrative perspective, completeness was at the third layer, while clinically, completeness was in the inner two layers holding the reference terminology and its cooperating model systems.
Dr. Campbell suggested that core reference terminology developers must have cooperative agreements with layer two developers so content and editorial evolution were orderly and cohesive. He suggested that the Subcommittee consider development of reference information for definition of complex and content-laden concepts and identifying an SDO to maintain the model. Considering concept orientation, Dr. Campbell noted the Committee had done much to address reference terminology issues. Noting many of those criteria couldn't be applied to layer three, because they had no control over those terminologies, he emphasized that reference terminology relationships within the core must be sufficient in definition and scope to support clinical guideline implementation. Dr. Campbell believed core terminologies provided domain content to support complete clinical reference terminology and layer two terminologies modeled, defined, and enhanced important legacy and departmental classifications and terminologies. Both mapped to layer three terminologies to support reimbursement and statistical classifications.
Ms. Haber remarked on a historical disconnect of data collection priorities and methods. She noted NCI had an inheritance of legacy systems. Achieving connectivity between established mission critical systems involved complex issues of data conversion, storage, and retrieval. Health information models changed with cancer and healthcare shifting emphasis from treatment to prevention, rapid transformation of disease models to genetic and molecular classifications, and the pace of change in science, technology and methodology. Vocabulary content and structure had to quickly reflect these advances while permitting identifications and retrievals from earlier systems and vocabulary versions.
Noting the challenge of addressing knowledge systems at the broader level of standardized PMRI, Ms. Haber said linking related concepts in the fields of population, genetics, developmental therapies, and disease was essential to NCI's mission and this ongoing transformation meant any proposed information model not incorporating the assumption of frequent change would fail. Vocabulary standards had to insure adaptability of a terminology or information model to incorporate intellectual content changes. The capability, commitment, and funding to support frequent and regular cycles and the flexibility of a terminology structure to map readily to other major component vocabularies were perhaps even more important indicators of the ultimate success of a standard vocabulary than its contents.
Ms. Haber recommended that the PMRI selection process accord priority to terminology development efforts linked to and integrated with related clinical document architecture, format, and messaging standards (e.g., HL7 vocabulary standards development process). She added that vocabulary sources should have explicit, defined, and regular processes for technical and content quality review and assurance.
Dr. Overhage found the categorization of codes the Committee proposed in its August 2000 report to the Secretary difficult to work with in terms of segregating out queries to aid in analyzing these domains. Dr. Overhage advised it was more useful to take an object-oriented or entity approach. Whatever they did with these terminologies, both the clinical processes and other uses of the data had to be pulled together. In discussing terminologies, Dr. Overhage noted they often talked about multiple things. Typically, at least three levels or subdomains of terminologies were conceptually needed for any domain. Commenting on criteria for viewing different terminologies, Dr. Overhage noted they'd heard one criterion was the terminology be relatively inexpensive to acquire and implement. He emphasized the importance of acceptance in the marketplace. Some ways to prioritize were by: the quantity of electronically available data, relative global value of that domain or data, and terminologies that were "good enough places to start." He said the Logical Object Identifier Numerical Coding System (LOINC) was a potential candidate terminology, at least for the laboratory domain, and reviewed ways it succeeded. Dr. Overhage emphasized the importance of keeping terminology domains conceptually pure and clinical processes connected. Mapping could be a partial, always risky and difficult-to-maintain answer. Differentiating concepts and structures was another way.
Two years ago AMA commissioned a study identifying a comprehensive list of core clinical data elements for electronic medical record systems. Mr. Beebe recommended the Committee prioritize the PMRI terminologies based on current use, practical applications and marketplace acceptance, as well as each system's benefits. AMA believed the criteria were acceptable, however Mr. Beebe said the criteria should be divided into essential and desirable, and different domains should be considered. He noted the four criteria derived from the PMRI guiding principles applied to PMRI terminologies. AMA believed that PMRI terminology maintainers should meet the guiding principles expected of current maintainers of the selected standards. Mr. Beebe suggested development of PMRI standards must focus on market acceptance and clinical needs. He noted the Committee's responsibilities were fundamentally different than for HIPAA administrative transactions: no federal legislation required implementation of PMRI standards or code sets that applied to such standards.
Discussion on Terminology
Dr. Cohn noted the Subcommittee had a day in October and a day in December for more input and will schedule time in 2003. They intended to have a letter for the Secretary by the June meeting. Whether they had a draft by their March-April deadline depended on the complexity of the information gathering.
Mr. Blair noted they'd heard testimony on the scope, including suggestions for a different perspective, and feedback on criteria for selection, priorities, and whether everything should be ANSI-accredited. The testimony didn't diverge much from the criteria for selection, though it might be divided into different categories and Mr. Beebe considered it an editorial of desirables. Mr. Blair will shape the hard-copy testimony into a matrix and clarify Dr. Overhage's and Dr. Campbell's models and information. The Subcommittee will review the input in October. They'd heard testimony on the scope and criteria for selection. Mr. Blair noted they had to decide how and when to gather information from terminology developers so they could match it against criteria. He noted many issues were more complex than with the message format standards and they'd need assistance from a consultant with medical terminology expertise. In October they'd identify criteria and further questions. November-December, they'd hear from terminology developers. January-March, users will testify. Hopefully, in March-April they'd look at a draft.
Dr. Cohn addressed the business case. Themes included lab terminology and drugs. The next hearing might provide an update on the DVA-NLM-FDA effort. Other pieces were what payers, providers and vendors considered high-value and an IOM activity looking at patient safety data standards they'd hear about this fall. Dr. McDonald suggested sending the mailing to developers of the codes as well as message developers. Mr. Blair will compile a derivative of the testimony and a draft of the questionnaire to look at in October.
Given the testimony, Dr. Cohn suggested they'd end up with more insight on potential criteria and approach than the issues of scope and high value. Dr. Yasnoff agreed that business case issues for vocabulary were more difficult; whatever their recommendations, expenditures would be much larger than for message format standards. He suggested a business consultant do a formal analysis of the business case for development and maintenance of a PMRI terminology and user cost of implementation. Dr. Yasnoff noted several business cases.
Though long-term benefits were hard to quantify, Mr. Augustine predicted that standard terminology would be increasingly important with health care inflation in double digits and baby boomers eligible for Medicare. Dr. Cohn asked Mr. Augustine to work on this model. Mr. Blair noted the business case wasn't just the cost in expenses and maintenance, but also value. Dr. Yasnoff suggested recounting how much DVA and DOD spent developing vocabularies; costs that standardized terminology could avoid. Dr. Steindel noted they'd review a lot of this data in the CHI initiative. He predicted that much of the business case's justification would be the non-quantifiable opportunity costs of not doing this. Mr. Augustine said another way to bolster the business case, though it too wasn't quantifiable, was with examples of what could be done: instead of referencing ICD-9-CM and Current Procedural Terminology (CPT) codes (an inexact science), one could see physician practice patterns in the aggregate across the health system, bolstering opportunities technologically on the research side and for fine-tuning reimbursement systems.
Dr. McDonald revisited the question of the Subcommittee's end game. He believed they were talking about turning text into codes, which assumed either extensive labor activity or progress in natural language processing, and proposed considering both in this analysis. Dr. Cohn suggested considering the issue of short-term versus long-term and dealing with Dr. McDonald's point about things ambiguous and uncertain. Dr. McDonald observed finding slots for what moved around was less of a leap. Noting people were making large investments in longer-range solutions, Dr. Cohn said a case could be made for cost-and-time effective ways to provide more decision support or data capture capability. Dr. McDonald concurred, but rather than entail huge changes in process or spend substantial funds before anyone succeeded, he encouraged using vocabularies richly. Mr. Blair recalled that during the decision process on message format standards, an extremely compelling point was how vendors and end users had invested heavily in HL7. Dr. McDonald agreed the Subcommittee should hear their testimony.
Dr. Steindel addressed the scope of the letter. NHS spent over 13 years developing the U.K.'s terminology model and the system's first round of testing was still a year away. The Subcommittee was told only one terminology was "out there"; Dr. Steindel asked where the good-use cases were. He emphasized the need to determine what to concretely propose in the letter and models, noting CHI grappled with that question and that they'd heard about low hanging fruit.
Dr. Steindel noted Dr. Campbell had just introduced his way of looking at terminology and the informatics community hadn't had time to discuss how it fit in with everything else. He suggested the Subcommittee do as they'd intended and define criteria for a questionnaire aimed at people developing terminologies, and ask vendors how to meet selection criteria. Dr. Steindel said they needed to determine the extent of overlap and confirm these groupings made sense. He said the Subcommittee might want to devote time to discussing all they'd heard. Dr. Cohn proposed that they needed to hear from vendors and potential users, ensuring everyone thought the same and shared priorities. He agreed that, based on what they'd heard, a strong business case had to be associated with everything. He suggested they might decide tomorrow to leverage what they learned CHI might be doing.
Dr. Yasnoff emphasized that a more formal approach to the business case was an essential underpinning to making the right recommendations and provided a rationale and substantial backing for those choices. Dr. Fitzmaurice summarized what he'd heard: overall the criteria were right, most candidate codes had usage and, depending on scope, there'd be different criteria for various aspects, everybody said start either with labs or drugs, they'd heard much about economics of terminologies, a vision of interoperability was built on terminology adding value to databases and value added to legacy systems as one mapped to research, syntheses or clinical practice guidelines, nursing codes could be used to improving health care processes, determining the value of mapping, many criteria dealt with quality of codes and terminologies, terminologies weren't independent in the long run, timeliness was an issue, and the value (relative to cost) of a piecemeal versus whole terminology approach.
Mr. Blair noted that Dr. Campbell and others proposed, even with a stair-step approach, giving priority to the clinically specific core. He'd heard Dr. Campbell propose an alternate way of organizing the scope model. Dr. Fitzmaurice depicted it as having everything that helped care for a patient in the middle; beyond that were departments organized around taking care of patients. Outside that was administration and infrastructure. Mr. Blair agreed. He added that Dr. Campbell had looked at the scope the Subcommittee offered as a starting point and responded that his representation was a better way to view, organize and prioritize. Dr. Fitzmaurice saw many relationships in Dr. Campbell's inner core. Dr. Fitzmaurice reflected that the vision around the table of what they wanted to do with terminology and codes was constant. But he suggested the time it would take to describe and show relationships would be best spent following a piecemeal approach choosing standards and building the core. Dr. McDonald noted they were about to pick ICD-10- PCS, a clinical code set; operationally, this next year they'd be piecemealing. Dr. Fitzmaurice encouraged using Dr. Campbell's model as a concept with a piecemeal approach to individual standards, and not worry about it fitting into Dr. Campbell's center core.
Dr. Steindel recalled Dr. Overhage's comments that Dr. Campbell's model could be built in pieces and looked at from a structural point of view of how select parts fit together in the various layers. Dr. Steindel contended that Dr. Campbell's model required a way of thinking they weren't prepared for. It had many places to hang concepts, but he didn't know where the hooks were, only that there were things he could put on them. Ms. Greenberg concurred it would be premature to embrace. She questioned the underlying assumption that components in the external layer lacked clinical meaning or utility; Dr. McDonald agreed that many administrative components were rich with clinical data. Ms. Greenberg said, right now, there was only one reference terminology, and many could embrace consistency and some reference terminology with either mapping or models. Dr. Steindel agreed, adding he wasn't sure there was a third layer. Dr. Yasnoff depicted the model with reference terminology at the core and a layer around it of compound concepts and terms from the reference terminology. The third layer was aggregations of simple and compound concepts into higher-level terms. He suggested the Subcommittee ask others to look at the model and convey their view of its utility, significance and value at the next meeting. Dr. Cohn agreed that other testifiers' comments might be useful.
Dr. McDonald cautioned that political/social decisions might not create what they wanted. He suggested some things (e.g., ICD-9-CM) were ripest and used everywhere. Expressing concern that the Committee had said in a previous letter that the HL7 RIM should be the core, Dr. McDonald urged they resolve where the first modeling occurred and bless level three components already making progress.
Dr. Cohn said the Subcommittee had to contemplate the core model and whether, as Dr. McDonald queried, it was the HL7 RIM or this conceptual model. They'd talked a lot about terminology versus information models, but still needed a way to describe the core model that helped everyone. While the HL7 RIM might be the information model, Mr. Blair noted Dr. Campbell proposed considering a semantic interoperability model. He recalled that the Subcommittee indicated in its message format recommendations that all other message format standards had to coordinate or harmonize with the HL7 RIM, because interoperability was driven by the RIM for message format securities. He noted they had to discuss whether all terminology should be derivatives of the HL7 RIM.
Dr. McDonald said the question was whether they decided the vocabulary fit together in some abstract sense or chose a process identifying what was sent and where and ensured there were places in the vocabulary model describing what went there. Mr. Blair noted Dr. Overhage felt, pragmatically, they'd move faster if they determined the terminologies needed to be clinically specific and drove down from the RIM; Dr. Campbell was saying other health care information objectives besides messaging benefited from his model. Dr. Elkin suggested there were two options for merging the information and terminology models that allowed semantic interoperability: either they agreed to be non-overlapping and coordinated or, wherever they overlapped, they had to be identical. Dr. McDonald cautioned about assuming success with the model building on its own with vocabularists, noting they had to think about object structures and what made messages interoperable.
Dr. McDonald noted a database was needed to connect the vocabularies together. Members discussed gaining input from HL7's vocabulary SIG, Jim Cimino, and RIM. The next step was to standardize vocabulary, which could get a lot of "bang for the buck." They still had to decide who provided that standard, but scope should be decided based on the law, what was already done, and what came next. Mr. Blair agreed, but expressed concern about a sense of urgency to move parallel with semantic interoperability. Dr. Cohn asked if the answer was to at least do one and as much of the other as possible. Dr. McDonald queried whether the Committee should standardize this or encourage the industry to provide it. He noted messages needed common vocabulary notes and fields, so all machines could use them. Once they got down to the right terminology, Dr. Steindel noted there was agreement. He recalled Dr. Elkin saying the information and terminology models overlapped. Part of the terminology model fit into, but wasn't covered, in the information model. They needed a "grammar" on how to combine those bits. Dr. McDonald noted grammar could be inserted, but most places sought bigger code.
Dr. Steindel encouraged suspending the information-versus-terminology-model mindset, emphasizing a continuum and the reality that some things (e.g., Dr. Campbell's medical guidelines project and CDC's public health case definitions) had to be combined to realize all the expressions. Noting they'd hear in October about an interface piece and opportunities, but adding that the profession had talked about this a long time and that the next day a coalescence of groups would be looking to them to get something on paper, members discussed listening to all the ideas and then, drawing upon abstract principles and "a boot strap," propose a scope. Members discussed the need for a consultant to compile and analyze feedback from terminology developers and provide an appropriate business analysis. Mr. Blair asked for names of people who met the criteria.
Observing that people worked on standards development and other healthcare initiatives from a broad range of perspectives and might not be familiar with other initiatives out there, Mr. Blair and Dr. Steindel provided a quick glimpse at organizations that promoted, developed and coordinated healthcare information standards. Everyone encountered different terms (electronic health records, electronic patient records, computer-based patient records) and some definitions were based on usage and experience, rather than academic specific definitions easily referenced. HIPAA used patient medical record information. Some private sector major SDOs were noted, along with initiatives involved in facilitating, promoting or coordinating standards in the private sector, as well as participating government agencies. Mr. Blair presented a time line and posed questions, including whether new standards or healthcare initiatives addressed: (1) new problems and approaches, (2) known problems with a new perspective, (3) problems with new methods or technologies, (4) known problems with greater industry support or energy. Other major questions were how these standards initiatives related to others and how they fit into the National Health Information Infrastructure (NHII) that the Committee and other organizations proposed and worked on in similar initiatives.
Panel 4: Coordination of Ongoing National Healthcare Standards Initiatives
Dr. Hammond said the public-private collaborative, Connecting for Health's, purpose was to catalyze specific actions that rapidly cleared the way for an interconnected, electronic health information infrastructure. He identified three basic components: accelerating adoption of clinical data standards, sharing best practices for secure and private transmission of medical information, and understanding what health care consumers need and expect from an interconnected, electronic health information infrastructure. Through Connecting for Health, people from previously independent, vertical silos of activity were talking and engaging as they defined the process of producing, collecting and using data for healthcare in a way usable for multiple purposes. Recognizing this was a huge, difficult task requiring priorities, they crosswalked priority data needs to the operable standards while initiating work on strategies to move other standards into play to deliver required data to all stakeholders.
Next steps for the Data Standards Working Group include: accelerating adoption of required data standards under development, articulating the need for standards not underway, making recommendations about terminology, developing strategies to clear barriers and accelerate adoption of a priority "operable set" of standards, and articulating the road to interoperability. Connecting for Health's interactions with PMRI include providing a forum for data users, systems developers, and clinical care providers; engaging in a dialogue about possible/doable/desirable and developing consensus; and facilitating adoption and rapid implementation of six data standards.
Identifying three interlocking HL7 standards he considered mature enough to use: the Reference Information Model (RIM); the Data Interchange Model; and level 1 of Clinical Documents, Dr. Hammond discussed how standards fit together and that understanding how they merged made it possible to produce standards more powerful than any viewed independently. Dr. Hammond noted plans for addressing gaps. He said standards groups and experts actively collaborated through overlapping involvement and communication with other initiatives. Many people realized the problem needed to be dealt with and most were moving in the same direction. Dr. Hammond suggested the Committee had a responsibility to help them work together and avoid redundancy.
The Consolidated Health Informatics Initiative (CHI), a government-wide initiative focused on health information interoperability standards in the federal healthcare arena, originated in GCPR, which the Subcommittee serves as an advisor. Some 20 partnering agencies, including agencies within HHS as well as DVA, DoD, the Department of State and Social Security, participate. About 100 medical, technical, policy and management experts are involved. CHI's goal was to establish federal health information interoperability standards as the basis for electronic health data transfer in every activity and project inside all health-related federal agencies. CHI sought to work with SDOs and industry to identify and get sign-off on mature standards. Dr. Kolodner expressed he hoped the federal government's adoption would create a tipping point and the private sector, judging those standards will be around and even more widely accepted, will voluntarily adopt them. He discussed a complementary activity, HealthePeople federal, and how DVA and DoD committed to activities that, ultimately, will result in their repositories of data for beneficiaries being based on the same standards and data models and the ability to exchange information as appropriate.
Ms. Love observed that health data systems formed the physical foundation of NHII and that often data wasn't a line item of high priority in state budgets. She noted limitations in the Uniform Hospital Discharge Dataset (UHDDS) upon which many states base their reporting systems. States had 50 versions of the UB-92 format adapted to their individual needs. National Association of Health Data Organizations (NAHDO) identified high priority data elements for the states and worked with the Public Health Data Standards Consortium and states to integrate a national standard that reflected what states collected. Noting states worked to supplement weaknesses of the UB-92 and that the claims attachment was mandated by HIPAA, Ms. Love suggested it offered an opportunity, near term, to understand how clinical data elements could be automated, captured and sent to a government agency.
Ms. Love discussed lessons NAHDO learned with administrative data that applied to mining and extracting data from the medical record, noting that the magnitude now faced would be greater: (1) state laws varied and members creatively addressed privacy and confidentiality concerns, (2) data wasn't perfect, but with concerted effort improved over time, (3) data collection priorities weren't without cost, the business case had to be made and everyone had to work together, (4) defining how data would be collected and used built trust, (5) analytic infrastructure was critical, (6) who paid was a huge concern for data agencies, and (7) comparability across provider systems and states had to be addressed.
Ms. Love expressed confidence in technical solutions and bringing everything together, saying it boiled down to leadership, building trust and money. Ms. Love encouraged the Committee to continue defining minimum core data needs for external reporting. She advised it was critical that the Committee frame a message at the state level about the societal good of clinical information to the public. And she recommended linking the Public Health Data Standards Consortium's recommendations with the Committee. Ms. Love pleaded to make administrative outpatient data development and standardization a priority across HHS.
Mr. Reese said the founders of National Alliance for Health Information Technology (NAHIT) believed it was time for the field to engage in the standards discussion. CEOs and COOs of provider organizations realized patient safety couldn't be solved without the integration of complex computer systems and that employee recruiting and retention required computerization and workflow management in order to improve daily work life. Capacity management was another factor: more efficiency and precision were needed to gain throughput and better integrate hospitals, rehab facilities, long-term and home care. Mr. Reese identified issues driving healthcare leadership in the field to pay attention to standards issues and provide AHA with support to spawn the national alliance.
Mr. Reese described experiences, which enabled Sentara to better utilize labor distribution and provide standardized nomenclature within the health system, and led it to become a founding member of NAHIT. Mr. Reese observed that the work being done by the various committees and everything listed that day wasn't on the CIO's radar screens. He suggested the Subcommittee "follow the money." When a standard was developed, it was the CIO who purchased the multi-million dollar system and told the provider which standard to deliver. Once CIOs understood what to buy, they'd enforce the standard at the local level. Mr. Reese said having the standards (and CIOs who understood them) would result in a major cost reduction and increase efficiency at the operational level. NAHIT was interested in standards related to bar coding, patient identification, communication and application systems. Its mission was to mobilize the field and assist in the implementation of standards; NAHIT convened the payers, providers, IT manufacturers and supply chain; took out small projects that could succeed; and led the field to recognize its role and responsibility in the adoption of standards. Mr. Reese said NAHIT was committed to a consensus approach and its strength lay in bringing the field together where they could educate people like himself about the standards and what had to be done. He reiterated that healthcare leadership had to resolve issues of patient safety, retention and recruiting and capacity of management.
Dr. Overhage discussed the e-Health (eHI) initiative and its foundation to bring together public and private sector stakeholders, promote and coordinate national adoption of health data standards and system interoperability, and meet clinical and public health needs. Dr. Overhage described the Public Private Collaboration for Public Health, an eHI initiative that develops and communicates strategies to efficiently and securely capture and transmit standardized electronic data of public health importance. Dr. Overhage said critical success factors include: public and private sectors working closely together, leveraging expertise and competencies; focus on specific, well-defined, highly visible needs for public health data; data largely already in electronic formats and likely to be converted and transmitted quickly; aligning efforts and needs with existing funding; leveraging the work of experts and "those in the trenches" and building upon lessons learned and best practices; involving key stakeholders; driving usage of standards and strategies through implementation and demonstration projects; and aligning efforts with national incentives,
Dr. Overhage noted key areas requiring additional development or implementation and lessons the initiative learned. eHI's model of rapid short-term intensive efforts focused on specific problems demonstrating significant results in a modest time scale and that the industry was ready for early implementation steps. eHI demonstrated the value of bringing together multiple stakeholders in a public-private collaboration committed to working with others through overlapping involvement that leveraged complementary skills, knowledge and expertise. Dr. Overhage emphasized ongoing, open communication and concrete alignment of activities. He said the Committee could assist by: continuing to create a clear, consistent message about "operable" standards and terminologies to adopt near term, encouraging and supporting demonstration projects, and developing a model for migration to a PMRI standards-based environment. He noted it was vital that migration strategy become a critical factor in how everyone thought about things.
Agenda Item: Comments about standards Implementation in the United Kingdom
Dr. Price summarized NHS's decade worth of national health information technology strategies based on: the patient as center of operational clinical systems, IT supporting clinical care while secondarily deriving information to optimize cost effectiveness, and managing expectations about IT's outputs and benefits. NHS's current focus is on implementing key targets with centralized management and standardization, while recognizing industry's key role. NHS's separate standards approval mechanism relies on evidence of live implementation, plans for managing human behavioral change management aspects of the new standard, and evidence of stakeholder engagement. Dr. Price emphasized that standards integration remained a significant challenge.
Dr. Price emphasized that standards needed to be interoperable and noted the challenge that presented. NHS considered CT implementation a bridge between yesterday and today. Dr. Price said the foundation had set business objectives for implementing the standards and intended to make this product suitable for key initiatives, particularly short-term strategic requirements. NHS aimed to manage the supply chain around this standard (engaging with end users, political groups, managers using secondary information, and suppliers), integrate with other standards and complete the standards approval process. Dr. Price noted business issues had to be addressed around distributing the standard and terminology, and producing and costing a formal, long-term implementation plan. He stressed the importance of determining who was responsible for performance managing implementation. Dr. Price estimated the cost of upgrading the U.S. clinical system to SNOMED would account for only 20 percent of the implementation costs; amending legacy data and primary care records would require the rest. In planning, NHS assessed the value of its primary care data, the extent it would formally be migrated and archived, and the implications. Dr. Price described Evaluating and Rollout of Terminology in Clinical Applications (EARTICA), which oversees rolling out SNOMED. He said implementating a standard across NHS took a year and cost about $3 million.
Agenda Item: Comments about Healthcare Information and Management Systems Society
Mr. Roberts outlined HIMSS' ongoing national initiatives (e.g., Integrating the Healthcare Enterprise, a coordination with the Radiological Society of North America, promotes coordination of established communication standards to improve systems interoperability). In August, HIMSS merged with CPRI Host and will sponsor a field hearing February 11 in Seneca, California made up of industry experts and congressional staff on whether the federal government should mandate an electronic health record. In June, HIMSS unanimously adopted NHII as the society's long-term public policy initiative. Its NHII task force evaluated initial projects including creating an inventory of existing technologies and practices, developing a gap analysis on what needs to be created and a HIMSS version of NHII.
Panel 5: Issues in Transitioning to New Code Sets
Mr. Daley said diagnosis and procedure codes were the cornerstones of health care administrative systems and wholesale change had to be approached cautiously, so potential impact could be evaluated and code changes didn't result in unintended, adverse consequences. A recent report from the General Accounting Office (GAO) raised concerns about ICD-10-PCS implementation challenges and the associated financial burden on the industry. Clinical codes underpinned virtually everything in the system and change to ICD-10 would impact payment, medical policies, and quality improvement programs while benefiting design, fraud, and abuse detection. Anticipated code sets costs approached Y2K costs for transactions. Neither the costs nor benefits were fully demonstrated. Mr. Daley cautioned that in-house and packaged software would be impacted. Reimbursement and contracting procedures, care management policies, and training forms could require change. Payer software including interfaces would need carefully coordinated updates, and logic built into many applications that directly interrogated specific code values would add to the complexity. A fully functioning crosswalk had to be available and a change of this magnitude required extensive testing.
Modifying diagnosis and procedure coding would impact the heart of the reimbursement process. AMA projected 8.4 million physicians' contracts and fee schedules would require renegotiation. Mr. Daley estimated hospitals would have to renegotiate 140,000 contracts. Procedures would require review and potential modification because of new diagnosis and procedure coding. Extensive training would be required. Even with a suitable crosswalk, maintaining dual-coding structures would be, at best, cumbersome, and ability to capture historical data or review utilization information would be difficult. Many concerns had to be resolved. BCBSA recommended there should be no migration until initial HIPAA mandates were successfully implemented and a thorough impact analysis, including cost/benefits of migrating or not migrating as well as transitional considerations were analyzed and alternatives assessed.
While use of ICD-10 codes for medical diagnosis and procedure coding had been debated for years, Dr. Cross said what hadn't been assessed was the potential cost and administrative burden of moving to a new, complicated coding structure. He urged the Subcommittee to thoroughly consider associated costs and complexities, cautioning that changing to ICD-10 for diagnosis and inpatient procedure codes would result in a tremendous burden, especially for physicians. The GAO report analyzing whether a single coding system should be adopted for all inpatient and outpatient medical procedures noted difficulties switching to ICD-10-PCS and stated "the design and logic of 10-PCS raised concerns about potential challenges in its implementation." Aetna concurred with the GAO report's conclusions on the potential administrative and financial burdens of changing to ICD-10. Dr. Cross urged the Committee to conduct a detailed analysis of the cost of this change before deciding and noted the ultimate consideration was that all costs would be passed on to consumers already experiencing a significant rise in health care costs.
Discussion: Standards and Security Issues - Draft Recommendations for Code Sets
Noting the Subcommittee held a number of hearings on this complex, contentious issue and it was hard to tease apart diagnosis and procedures, Dr. Cohn outlined options: a letter recommending putting the two together, moving to an NPRM or NOI, doing nothing now, or further deliberation (which wouldn't be for months). They'd heard the industry preferred implementing together, but also that staggering made conceptual sense, and that additional studies of implementation costs and benefits were needed before an NPRM or NOI.
Mr. Daley said ICD-10's increased specificity could identify conditions to help expedite payment, care management, fraud and abuse protection. BCBSA was concerned about the complexity and degree of impact. AAHP worried about administrative and financial costs of dealing with additional complexity. Noting Dr. Cohn asked for a cost/benefit study, Mr. Blair emphasized including providers from the standpoint of improving patient outcomes, clinical research, and information hospital administrators needed to improve the cost structure. He noted they'd heard that the cost/benefit for the payer sector wasn't enough.
Dr. McDonald reported that the written comments (two-and-a-half for; three-and-a-half against) left him cynical. AHA's studies boiled down to 100 charts evaluated by CMS's contractors. Dr. McDonald called for an independent reviewer's detailed numerics. Nellie Leon-Chisen said AHA believed ICD-10-PCS was sufficient for inpatient hospital reporting. Recalling Dr. McDonald's observation at the June meeting that physicians weren't clamoring for more codes but wanted guidance, Kathryn Serkes, speaking for the Association of American Physicians and Surgeons (AAPS), noted physicians wanted to address issues of audits, prosecutions, and fraud. Mr. Beebe reported that AMA contracted with PricewaterhouseCoopers for a cost/benefit analysis of different scenarios for the inpatient setting or achieving a single procedure coding system. Mr. Augustine pointed out that managed care and payers were good at collecting and analyzing data ahead of the curve. Noting their progress utilizing medical and disease management, he said he'd like to see more information or put what they had into a better format.
Dr. McDonald said he wouldn't be opposed to an NOI with these caveats, but expressed concern about acting on the basis of 100 sample records. Dr. Cohn proposed a letter to the Secretary, noting the need for a cost/benefit analysis within six months. Meanwhile the Subcommittee could collect similar data. Ms. Brooks predicted that, if the Committee did an NOI or NPRM, hospitals would do their own independent studies. An NPRM would force government to analyze the cost and calculate a projection the industry would read carefully and respond to through several iterations. Ms. Brooks also anticipated that, with ICD-10-PCS, hospitals would continue to code what they did today under DRGs, even though more codes could be used for internal purposes.
Ms. Trudel explained that a NOI provided a view of potential issues, laid out options, and, being in the Federal Register, drew wide public comment. After assessing that input, an NPRM could present options and solicit further comment. An NOI wouldn't require a full-fledged impact analysis. Ms. Greenberg cautioned that an NPRM had to follow an NOI before doing a final rule and would delay the process. They'd heard concerns about how much longer organizations could continue with codes 25 years old that couldn't be updated and reflected antiquated terminology discarded by the rest of the world. She noted Ms. Trudel had pointed out that they could accomplish everything in an NPRM that they might with an NOI and force the Department to do the impact or cost/benefit analysis.
Dr. McDonald said, with the knowledge they had now, he'd go with an NPRM. Mr. Augustine reiterated a preference for an NPRM with options for an impact analysis and adopting both, one, or a future approach. Noting an NOI could tie the General Counsel's hands, Ms. Trudel proposed recommending specific concerns and issues, including cost/benefit analysis and documented testing of actual coding using real-life charts. Mr. Blair concurred with Dr. McDonald's concerns about broader, more comprehensive testing. Testifiers reiterated the need for the business case and Mr. Blair said he wanted to ensure that case was across the board and broader than cost/benefit for payers. An NPRM would allow for the cost/benefit issues and additional testing. He expressed concern over further delay. Dr. McDonald said making the wrong decision fast wouldn't help. Mr. Blair pointed out that they'd be able to draw upon negative feedback. Dr. McDonald said he'd looked: there were 7,000 more codes available in ICD-9-CM, Volume 3 and ways to code for another year. Ms. Brooks noted a write up on the issue of extending existing codes. She reported that the way the tree added alphanumeric digits didn't get at the root of the structure's problem. She recalled public meetings 10 years ago with the ICD-9-CM Coordination Maintenance Committee exploring ways to expand the four digits. People had been adamantly against it.
Dr. Cohn noted Mr. Blair proposed an NPRM. With the right caveats, Dr. McDonald opted for an NOI. Dr. Cohn said key issues were differences between mechanisms (e.g., NPRMs and NOIs), open issues needing resolution, and uncertainty over a solution. They were talking multiple things. He asked if members wanted a letter expressing concerns and next steps or to recommend an NOI or NPRM. Mr. Augustine reiterated the possibility of an NPRM, with the option of implementing separately or both internally. Mr. Blair noted they'd discussed caveats and testing. Dr. Fitzmaurice cautioned against stretching this beyond when many members remained at the table. Though an NPRM might solve some problems, he understood Dr. McDonald's concern about ensuring the right decision and he acknowledged an NOI was appealing. But an NPRM would force the government to do an impact or benefit/cost analysis before making comments and framed issues so people could comment and trust government to follow that NPRM and turn out a good final rule. Dr. Fitzmaurice said he was sensitive to the time an NPRM took, but it had worked with other HIPAA standards and everybody agreed they needed information. He concurred with Ms. Trudel who'd urged the Subcommittee not to tie the hands of the Department, but encouraged it to provide enough information to make a better recommendation at another time.
Ms. Trudel reiterated that the Subcommittee had four alternatives: do nothing, recommend a specific code set, hold more hearings, or do additional analysis using the Administrative Procedural Act process to gain input and involvement--CMS would decide on an NOI, NPRM, open town hall meeting, or industry forum announced in the Federal Register. Dr. McDonald said he'd stop short of an NPRM. Ms. Greenberg said they'd heard, "Go with diagnosis first." At the last hearing, they'd been told, "Go with 10-PCS and 10-CM together." Dr. McDonald said he didn't consider ICD-10-CM much different, but expressed concern about PCS. Ms. Greenberg noted some European countries had seven years experience with the diagnosis codes and many had implemented ICD-10 and clinical modifications; 10-PCS hadn't been available. Mr. Blair noted an NOI didn't set the process in a bound time line. With an NPRM, people weren't so vulnerable to open-ended delay and might determine if their direction was misguided, incomplete or incorrect. Dr. McDonald noted CMS's enthusiasm for moving with ICD and PCS, but emphasized that a hundred cases didn't indicate anything. Dr. McDonald said he was interested in comparisons, not continuous improvement, and wouldn't make decisions on any process change in his hospital based on that data.
Dr. Zubeldia considered issues about ICD-10-CM diagnosis codes, ICD-10-PCS institutional procedures, and ICD-10-PCS for professional (not recommended) quite different and advised separate NOIs or NPRMs. He favored the NPRM because it forced government to make an impact analysis, and he stressed there had to be separate analyses for procedure and diagnosis codes, as well as procedure codes for professional services. Another issue was whether adoption should be synchronized. Observing that even today's testifiers weren't as concerned about implementing diagnosis codes as procedure codes, Dr. Zubeldia voted for the NPRM with the caveat that components be addressed separately. Mr. Blair endorsed an NPRM with Dr. McDonald's caveat about testing. Mr. Blair observed that Dr. Zubeldia's suggestion to separate sections in the NPRM for ICD-10-PCS, ICD-10-CM, and other pieces was similar to their first draft with the deadline of no sooner than October 2005. Noting the NPRM process failed to identify problems with the NDC code, Theresa Dalmer, BCBSA, cautioned that an operational date of 2005 was ambitious. She doubted that in 60 days they'd get the information needed to accurately anticipate impact. Noting comments about impact on hospitals emphasized evaluating across the industry, Ms. Dalmer noted many players didn't look at the Federal Register or NPRMs and would have to be reached in other ways.
Mr. Blair suggested extending the 60-day response period to 90 or 120 days. He envisioned the NPRM establishing a mechanism that formally solicited comments and drew in the business case. Dr. Zubeldia expressed concern that an impact analysis might not reflect private industry's view. Recalling that AMA might do its own impact analysis, he asked how the process could reflect private industry analyses. Dr. Fitzmaurice noted perspectives submitted as comments to the NPRM had to be considered. Dr. Zubeldia pointed out that comments didn't have to be published in the final rule; he suggested the government could informally publish a white paper before the NPRM depicting its impact analysis alongside private industry's analysis, and putting both in the NPRM. Other analyses could be referenced and points of agreement stated. Ms. Greenberg clarified that the rule-making process went through numerous segments of government; publishing prior to the NPRM didn't make sense because one couldn't go through that vetting process.
Mr. Augustine agreed with Dr. Zubeldia's original proposal of an NPRM, but separated out aspects. Dr. Zubeldia emphasized that the industry needed to see the government's and other impact analyses before evaluating the NPRM. He supported an NOI. Dr. Zubeldia proposed that the government do a full impact analysis, compare it with private analyses, and move expediently to rule making. He advocated an NPRM, if required for an impact analysis, and pre-publishing preparatory work for the NPRM. The staff said that wasn't possible. Mr. Blair suggested extending the public comment period. Dr. Cohn reminded everyone that they were a subcommittee of an advisory body to HHS and limited in what they could do. Members discussed whether the Committee could contract for an impact analysis. Mr. Augustine said he'd prefer doing their own impact analysis before taking the next step. Dr. Fitzmaurice proposed a study with hard coders double coding.
Ms. Greenberg said the Subcommittee could commission a parallel study; it wouldn't accomplish what the NPRM would, but they could hold hearings about other studies. But she cautioned that was no different than delaying a few more years and didn't inform the field. Dr. McDonald advised that, given the time frame, a NOI was more sensible because it generated studies. Dr. Cohn recorded a vote for an NPRM, with one negative. Members discussed asking the Secretary to inform the public of their intent and urging industry to perform its own impact and cost/benefit studies before any NPRM. Reflecting on Mr. Blair's desire for full industry appraisal, Ms. Greenberg suggested commissioning a study that "quantified" benefits. Mr. Augustine cautioned that was extremely difficult. Members approved the direction of the original document. A revision will be distributed before the next meeting. Dr. McDonald was assured that he'd have an opportunity to discuss his minority opinion to the full Committee. Dr. Cohn explained he'd abstained because this dealt with procedure coding, as well as diagnosis codes; he said he wouldn't give an opinion to the Committee. Noting the implementation date recommended in the document seemed impossible, Dr. Cohn suggested thinking about it as they moved forward.
Discussion: DSM-IV
Dr. Cohn noted the Secretary received letters from the American Psychiatric Association (APA) requesting that DSM be named a code standard for use in psychiatry and mental health and that APA become the designated standards maintenance organization (DSMO). The Subcommittee's initial view was that these activities should be separate. Typically, DSMOs maintained transactions, not diagnostic criteria. And the Department received no comments recommending adoption of DSM-IV, though they'd heard concern that, once ICD-9-CM was the standard for diagnosis reporting, mental health couldn't use the diagnostic criteria in DSM. Dr. Cohn sought members' and APA's perspective. If there was agreement, the Subcommittee would draft a letter to the Secretary for review at the next meeting stating continued use of DSM-IV diagnostic criteria that crosswalked to ICD-9-CM codes didn't pose a conflict with code sets adopted under HIPAA and that adoption of diagnostic criteria used to establish diagnosis was outside HIPAA's scope. APA expressed concern that, without DSM-IV encoded in the HIPAA statute, the default descriptor would be the ICD-9-CM glossary and definitions. The Subcommittee will ask the Department's General Counsel if the concept of diagnostic criteria was, indeed, separate and distinct from the HIPAA code sets.
Ms. Trudel recalled that Dr. Cohn had indicated the DSMOs were specific to the transaction standards and not clinical code set maintainers. Dr. Darrel Regier, Director, Research, APA, agreed the issue was the code set and mechanism necessary for a publicly approved code maintenance organization. APA's counsel, Robert Charrow, clarified that there were two issues: the scope of the regulatory provision and the legal impact of a single official code set that had no diagnostic criteria. Eliminating DSM-IV as an official code set had profound implications on 675 state laws and, once HIPAA was effective, the informal crosswalk wouldn't be viable. Dr. Cohn said the Subcommittee's main concern was that the crosswalk between DSM and ICD was appropriate and people could continue to use diagnostic criteria. Members and staff were searching for a reasonable, speedily implementable solution. Dr. Cohn cautioned against hopes that this could be done with an NPRM before HIPAA was implemented, but he assured everyone that the Subcommittee believed there was a solution. The next step was obtaining CMS's legal counsel's review. Hopefully, lawyer-to-lawyer discussions would resolve this by the Subcommittee's next breakout session. Ms. Trudel will work within CMS and the Subcommittee will continue the conversation.
Day One
Agenda Item: Overview
Mr. Blair thanked everyone who'd come to help the Committee get the right start on identifying and selecting PMRI terminologies and making recommendations to HHS. He noted the testifiers previously received questions focusing on: (1) feedback on the proposal and whether it ought to encompass different areas or was too large or small, and (2) whether the criteria for selection were appropriate, pragmatic, realistic and sufficient. Mr. Blair said the scope and criteria for selection were derived from the Committee's July 2000 report to the Secretary that set forth priorities for going forward with PMRI standard recommendations and a set of ten recommendations, including the selection of standards. Mr. Blair pointed out that the report also addressed accelerating development of PMRI standards and guiding principles for selection in general terms and other guidances the Subcommittee sought to apply to PMRI terminology. In February 2002, the Committee applied these guiding principles to the selection of PMRI message format standards and set forth recommendations. The July 2000 report is available on the NCVHS Web site.
Panel 1: Terminology
Dr. Chute applauded what had been done in terms of preliminary criteria and acknowledged that the Committee had created multi-purpose roles and recognition for terminologies, outlined reasonable starting domains, and faithfully collated widely acknowledged common criteria. Dr. Chute reviewed motivational issues associated with development of terminology in patient medical records to support his recommendations for augmenting the criteria and scope. He noted that a major purpose of a terminology was to capture symptoms, findings, events and interventions with sufficient detail and specificity to enable their use in decision support, information retrieval, recognition and other familiar applications. He emphasized the importance of keeping use cases separate from their format and structure. He added it was important to remember that, in parallel with these detailed specific nomenclatures they might accommodate in patient medical records, a need would persist to aggregate this information characteristically by way of high-level classifications. Thus they had to enable accurate and reproducible grouping of this information for a variety of purposes, including but not limited to fiscal reimbursement management types of activities: public health, surveillance and statistical issues. Dr. Chute said they could imagine this continuum through various and increasing levels of aggregation or groupings, and cluster like conditions or events into small groupings or accommodate broad groups of related concerns in the context of these higher level classifications or aggregations.
Dr. Chute indicated that view-and-use cases had implications on the notion of grouping an organization. He proposed that they rethink the notion of a historical model of this continuum of terminology from a detailed nomenclature to a classification in a more broadly based context: a one-size-no-longer-fit-all approach. Historically (since 1996 national terminology symposia work), the idea that an underlying nomenclature could have detailed specific patient information and extend along a common continuum to higher level classifications or groups was perceived as a fairly linear, uniform and single dimensional problem. Examples were the evolving detailed clinical nomenclatures, the intermediate classification of ICD and its relatives, and higher-level groupings of DRGs. Dr. Chute proposed confronting a more complex use case with needs in decision support and error detection, public health and surveillance, reimbursement and management, outcome and epidemiology. Regarding this field of findings, events and interventions as an instance of a patient record, Dr. Chute said it was probable and plausible to reaggregate this patient in a variety of contexts, using diverse rules to achieve distinct classification contexts.
Noting it was improbable that a single classification or set of rule bases could accommodate the spectrum of use cases encountered in classification, Dr. Chute suggested imagining rules for a cascading set of groups or criteria and levels of aggregation within a common rule base. He noted these use cases and their spectrum of context had implications for criteria used for specifying and developing terminologies internationally that could be simplistically accommodated into two notions of description logics he termed their sibling aggregation (the old rule-based mechanisms by which clinical events could be intelligently aggregated into higher-level classifications) logics. He proposed publishing these aggregation logics that Dr. Spackman and others proposed as components of their attached classifications; e.g., if ICD-10-CM were the acknowledged content for public health and/or reimbursement criteria, the rules and logic by which specific detailed clinical observations, events and interventions could be aggregated into higher level ICD rubrics would be a component of a published ICD content, not a characteristic of a mapping element associated with an underlying terminology.
Whatever they did, Dr. Chute concluded they'd confront significant challenges (e.g., the human interface problem--how physicians, nurses and other health providers interacted with the terminology remained a considerable challenge and burden, particularly in face of time costs and the existing provider burden). Similarly, the impact on vendors developing health information systems would be considerable as they confronted the challenge of incorporating relatively complex and sophisticated representations of patient information into underlying products. And everyone would be confronted with selecting from what was available today or supporting development along the lines of content, authoring environments and newly described aggregation and description logics. Dr. Chute added that the Committee might consider an additional criterion for the organization of terms in the context of their various aggregation levels and these families of associated classifications and logics. He suggested incorporating as criteria for consideration of well-formed terminologies their behavior and opportunity to participate in description logics as definitions of their underlying content, as well as supporting aggregation logics that enabled their higher-level aggregation into appropriate classifications and groups among the discriminating criteria considered for terminology.
Panel 1: Terminology
Speaking on behalf of ANA, Dr. Bakken noted that some subsets of PMRI terminologies were designated by domain coverage (e.g., nursing) while others were categorized by area of the healthcare process (e.g., diagnoses and procedure) or purpose (e.g., SNOMED CT was classified convergence). ANA recommended renaming the "Procedures" category "Interventions," reflecting the diversity in the coded activities, and categorizing terminologies by multiple dimensions that took into account their domain coverage, primary purpose and the underlying semantic structure of the terminologies. Dr. Bakken noted her written statement included an expanded list of data nursing codes and that the Omaha System and Perioperative Nursing Data Set (PNDS) were incorporated into the Unified Medical Language System (UMLS). Noting that DEEDS was included in the PMRI terminologies, Dr. Bakken suggested the Subcommittee consider including the Nursing Minimum Data Set and International Classification of Nursing Practice, an international product recognized by ANA.
Addressing priorities, Dr. Bakken remarked that terminologies supporting PMRI were essential in promoting patient safety and improving healthcare quality, multidisciplinary care and communication; she recommended that these terminologies include codes related to pharmacy, laboratory, diagnoses, interventions, and outcomes. Dr. Bakken advised that the criteria for selection include documented testing of reliability, validity and clinical utility. She noted a number of terminologies the Subcommittee referenced referred to those notions. She said the criteria required modification so that they were appropriate for the different types of terminologies. ANA originally used a single set of criteria for its recognition process, but as terminological science evolved they found that wasn't sufficient to recognize the types of terminology products submitted and developed criteria to address data sets, classifications and nomenclatures. Dr. Bakken emphasized the importance of clarifying the criteria that related to each type of semantic structure, particularly in relationship to the notion of clinically specific terminologies, which implied these were concept-oriented terminologies.
Dr. Bakken supported Dr. Chute's observations about the relationships among terminologies, noting it was important to clarify what were terminology characteristics or abilities rather than roles or abilities potentially provided by models or supporting software. Dr. Bakken noted that the International Council of Nurses (which includes ANA) and the International Medical Informatics Association's (IMIA) Nursing Informatics Special Interest Group sponsored a work item within the International Standards Organization (ISO) on developing a reference terminology model that integrated nursing concepts. The model, which Dr. Bakken said enabled ANA to meet the notion of finely detailed clinical specifications and rules engines that support the description logic, is proposed as a draft International Standard (ISO, 2002). ANA believed that PMRI terminology developers didn't need to be ANSI accredited, but should be affiliated with an ANSI-accredited organization (e.g., terminology registration with HL7). Meetings should be open, with timely public notification, and the voting/balloting process should be published and accommodate public comment and appeal. Criteria should incorporate existing ISO, ANSI standards processes and definitions.
Panel 1: Terminology
Dr. Price characterized U.K.'s NHS as a diverse and currently unstable environment. The government had a huge commitment to NHS IT and faced a general election in about three years. In order to achieve aggressive targets for 2005, NHS was taking a more centralist control of implementation, "ruthlessly" applying standards and a more controlled approach to industry procurement.
In the 1980s, multiple commercial terminology development schemes appeared within primary care in the NHS. In 1988, the clinical professions endorsed the READ Codes. NHS purchased the READ Codes in 1990 and produced Clinical Terms (CT), version 3 in 1995. Public inquiries into the READ Codes led to a review of terminology options. Five options were considered: (1) don't use a standard terminology, (2) continue using READ Codes, (3) an off the shelf alternative (e.g., SNOMED), (4) collaborate with SNOMED, or (5) utilize another collaboration. In 1999 NHS agreed with CAP to develop SNOMED CT. CT was released in 2002 and NHS invests heavily in evaluating and implementing foundation programs.
Dr. Price noted five high-level terminology implementation goals: nation-wide roll out, multi-purpose use (direct and indirect care), multi-professional uptake (doctors, nurses, allied health), multiple environments (primary care, hospitals, community), and extensibility for patient access and social care. Clinical information systems begin using SNOMED CT in April 2003.
In describing NHS's experience in classifying terminologies, using a comprehensive terminology and mapping it to formal classifications, Dr. Price reiterated Dr. Chute's comments about a continuum from direct to indirect patient care purposes. He noted that SNOMED CT, through its reference properties, picked up more of the indirect care functions.
Dr. Price said arguments for comprehensive schemes included: (1) the advantages of the single provider organization (unambiguous problem-owner, single focus for change requests and user support), (2) inherent interoperability (avoids overlapping--duplicated--content), (3) supports uniformity of file and code structures (strong support for vendor community) and (4) economies of scale in support and education, cross mapping to classifications, maintenance (especially QA) and distribution. However Dr. Price noted monopoly (or oligopoly) concerns, potential unresponsiveness to changing requirements, the large size structurally might require subsets, specialist requirements might be constrained by the uniform structure, and the difficulty of establishing wide editorial input. Arguments for interoperability included: (1) schemes closer to user purposes and expertise, (2) a strong sense of professional ownership and commitment, (3) rapid adoption of suitable existing schemes, and (4) increased responsiveness to change requests. Arguments against interoperability included: the financial overhead in maintaining interoperability (structural properties, release synchronization), a tendency for scope creep towards comprehensiveness that compounds interoperability issues (e.g., LOINC in moving from LUD to clinical LOINC picked up contents of other schemes). Dr. Price reported that a survey NHS conducted this year indicated that 50 percent of terminology experts anticipate multiple interoperable schemes within 10 years; 25 percent predict a single global scheme by then.
Discussing the criteria for selection, Dr. Price said the technical properties weren't normally controversial, but well-rehearsed desiderata backed up by academic and standards work. He noted that end users considered many advanced technical features to be fairly superfluous for short-and-medium-term requirements. Dr. Price added that market forces (supply and demand for terminology), financial structure, and long-term viability of provider were important considerations in NHS's development and review of the business case for investment in terminology. Dr. Price noted NHS end-users expressed little real demand for terminology, a fact reflected in limited commercial supply of large comprehensive schemes. He also commented on a lack of clarity about scope, purpose and benefits and a perception that any benefits from adopting a national terminology would be governmental and central rather than local. Commercially aggressive business models were viewed as non-viable; intellectual property terminologies were only considered valuable if people were willing to pay for them.
In terms of financial structuring, Dr. Price advised that the Subcommittee's general criteria (timely updates, flexible and adaptive, relatively inexpensive) were mutually exclusive. Based on NHS's experience developing READ Codes version 3 and its investment in SNOMED, Dr. Price predicted any new endeavor required 20-30 million dollars of start-up capital. A guaranteed income stream (hard money) of about $l5 million annually would be needed to maintain a global comprehensive terminology. Dr. Price said an aggressively commercial venture probably couldn't sustain the financial requirements; a non-commercial structure was needed.
Dr. Price emphasized the importance of the long-term viability of providers. The costs of switching terminology (migrating databases, reporting arrangements and protocols) were potentially high and any low-viability model was high risk. Open contributorship models with subscriptions to cover costs, a national support level, a core product as "public good" within the subscribing community, extensions to core and subsets for local needs offered the best advantages.
Dr. Price reiterated that NHS had remained committed to using a single comprehensive scheme (now SNOMED CT) since 1992, and were investing $3,000,000 in 2002 to lay the foundations for implementation of the NHS. NHS believed that the provision of extension, subset and mapping mechanisms addressed many of the counter-arguments to comprehensive schemes. A global business model needed to be open, contributory, non-predatory and underpinned by central (government) support. Dr. Price emphasized that terminology wasn't an end in itself, but was needed to add value.
Panel 1: Terminology
Dr. Nelson concurred with much of Dr. Chute's validation of the Subcommittee's effort, while addressing areas that needed strengthening: terminologies, atomic concepts, relationships, priorities, pricing and the maintenance process. Building upon Dr. Chute's description of a continuum of information terminologies ranging from high-level aggregation of information and groupings for reimbursement down to the patient level, Dr. Nelson expressed concern about another level below clinical, particularly in light of the increased information in genetics. Rather than focusing on a continuum, Dr. Nelson encouraged the Subcommittee to recognize distinct uses of the terminology.
Dr. Nelson agreed that the best way to organize the Subcommittee's consideration of terminologies was to view them both in terms of the level of discourse and areas covered. Addressing the notion of an atomic concept, he pointed out that the level of discourse determined atomicity and the granularity of discussion. He agreed with Dr. Bakken that the areas of coverage in terminologies considered shouldn't be split out according to the professions engaged. Although he termed them procedures, rather than interventions, he acknowledged these were different areas that could be looked at as coverage.
He proposed a number of additions of vocabularies including the RxNorm project, two other pharmacy knowledge-based vendors (MediSpan and Micromdex), International Classification of Diseases for Oncology (ICD-O), International Classification of Primary Care (ICPC), new efforts by dermatologists to organize patient observations, taxonomy of organisms and anatomy.
Noting that the level of discourse determined atomicity, Dr. Nelson commented on the difficulty of defining atomic concepts when the level of discourse kept changing. Observing that one could describe things in fine detail, yet never completely define them, he suggested atomic concept was a myth. What one could convey was the level of discourse with which something was expressed, and whether it was a pre- or post-coordinated expression. Dr. Nelson suggested that what they were trying to say was that they wanted a post-coordinated expression for anything at a greater level of detail than their current level of discourse. He contrasted human tolerance for ambiguity in terminology with computational needs, and noted Zipf's Law stating that the names of things talked about frequently get shorter, becoming atoms. At the level of the Centers for Medicare and Medicaid Services (CMS), names were shorter than on a clinical level.
Dr. Nelson said Dr. Chute's comments about descriptive and aggregative logics clearly described relationships with other terminologies. A major difficulty for developers was that many terminology systems had sets of rules that were humanly understandable, but weren't expressed in a way computationally achievable. He emphasized the need to make sure that mapping wasn't vulnerable to obfuscation for remuneration or "gaming." Terminologies had to have representational integrity, so things could only be said one way. Dr. Nelson cautioned that when considering reference terminologies one had to be aware of the implications of changes in the reference terminology on other terminologies that referred to them and how such changes were propagated.
Dr. Nelson observed that, while it was a nice first order approximation to say terminology was independent of the system and interface, in the long run it was dependent. Noting the problem of character sets and special characters reserved for the operating system, he suggested there had to be some set of open standards when using terminologies with multiple vendors that defined exactly what the terminology would and would not do, and what the systems were and weren't allowed to do.
Remarking that the biggest priority was patient safety, Dr. Nelson said the data suggested that, with computerized physician order entry, errors could be halved on laboratory and drug orders. Another priority was a problem list, describing procedures and interventions on a successful array. He emphasized that other problems had to be solved before achieving a complete patient description.
Dr. Nelson said maintenance had to be quick, timely, responsive to user needs, and actively seek input with rapid turnaround minimizing local update penalty. He called for a set timetable for revisions and an organized review process. Remarking that he didn't think it possible to do vocabulary development on a consensus basis, Dr. Nelson advocated picking the "quintessential middle child," who listened to everybody, understood the engineering, and let him or her struggle it out. Perfection wasn't possible, but if you let responsible people do the job and respond to the users, you could judge whether they did a good job. He emphasized the need for an open, organized review process with input from all interested parties.
Dr. Nelson also observed that in the competitive environment price was marginal cost and the marginal cost of providing terminology for one more provider wasn't high. He noted pricing could also be based on the value added to the system by the terminology (a hard thing to measure) or the alternative cost of somebody else doing it.
Panel 1: Terminology
Mr. Lareau said the biggest challenge for Medicomp (a developer of tools for documentation and clinical decision support at the point of care for 24 years) has been what Dr. Chute called the human interface challenge. Working with clinicians, Medicomp has been continually challenged to address dynamics of the clinician-patient encounter. A major part of that challenge was to determine how to accommodate the clinical thought process without interfering or changing it. Medicomp found that clinicians process "clinical concepts" versus individual or "atomic" terms, and the relationships between them, in order to determine a diagnosis and what Dr. Bakken called intervention. Mr. Lareau said his responses to the Subcommittee's questions came from this experience and the way these concepts were processed.
To illustrate the importance of concept processing at a clinical, versus strictly terminological level, Mr. Lareau gave the example of a patient with chest pain radiating to the left arm, as opposed to the back. The clinician might reasonably suspect the diagnosis of angina and, absent other findings, order an electrocardiogram or cardiac stress test. Even though terminologically quite similar (many of the same words were used), the concept of chest pain radiating to the back might reasonably lead a clinician to suspect a diagnosis of aortic dissection. Questioning about recent trauma or an order for imaging studies might be appropriate, but a cardiac stress test wouldn't be in order. Mr. Lareau presented another example to further illustrate the significance of fully expressed clinical concepts at the point of care. Although many of the same terms and some of the same secondary concepts were used, neck pain radiating to the left arm might indicate not angina or aortic dissection but a nerve problem originating in the neck, for which cervical spine imaging studies might be appropriate. Two examples included the concept of radiating chest pain; another pair included pain radiating to the left arm, but Mr. Lareau pointed out that all three indicated very different medical problems and required the fully expressed concept in order to be actionable at the point of care.
To enable efficient point of care use of these concepts, Medicomp organized them into a hierarchical structure. This hierarchy of pre-coordinated concepts enabled assignment of predictable properties to each concept (e.g., its relevance to various diagnoses). With chest pain radiating to the back, this concept had higher relevance to aortic dissection than with chest pain radiating to the left arm, although both inherited the properties of their hierarchical grandparent, chest pain.
The combination of fully expressed clinical concepts and predictable properties for use by software applications gave software developers the means to build applications for clinical documentation and decision support usable and actionable at the point of care. Medicomp recognized that more granular terminologies might be more appropriate for other uses (e.g., research) and acknowledged that mapping between terminologies must be a priority for all terminology standards developers.
Mr. Lareau suggested the Subcommittee give careful consideration to accommodating terminology needs of healthcare providers at the point of care. He noted that the organization of the subsets of PMRI terminologies could be reflected by either establishing an additional subset for point-of-care or human-interface terminologies or, in a more radical change, including point-of-care terminologies as a separate type of convergence terminology.
Medicomp believed that, for action, information from the encounter had to converge at the data storage layer and point of care. The value of a very granular atomic convergence terminology for research outcome studies and other analyses was clear. However, even in a large population, outcomes were improved one patient at a time, a process in which the patient/clinician encounter was a primary component. At the moment, information had to converge in a way that was actionable and didn't unduly interfere with the clinical thought process or lead to decreases in clinician productivity.
Mr. Lareau emphasized that it was important in defining the scope of PMRI terminologies to consider the role of the patient. As use of the Internet expanded, patients increasingly became involved in the management of their own health care (this was already seen in the vendor base of Medicomp's licensees) and terminology standards had to allow for appropriate levels of participation by patients, as well as clinicians.
In terms of priorities for the selection process, Mr. Lareau noted that existing investments in billing, laboratory and pharmacy systems total in the billions of dollars for equipment, software, and training processes already in place. Much of the healthcare reimbursement structure was based on the use of diagnosis, procedure, drug and laboratory codes. The effect of any change in terminology requirements for these systems might be significant, requiring extensive modifications to existing applications and reimbursement criteria. Payers, providers, healthcare enterprises and application developers had to be given sufficient lead time to implement changes.
From Medicomp's perspective, highest priority should be given to terminologies directly affecting current, widely implemented processes in order to avoid push back from vendors: diagnosis and procedure, drug and laboratory codes. Message-specific codes (e.g., HL7) should be next. Dr. Lareau emphasized that developers needed a firm understanding of how their systems will be required to communicate within NHII as well as sufficient lead time to implement the new standards. Although critical for patient safety, drugs and laboratory orders, Mr. Lareau noted convergence and point-of-care terminologies were neither as widely implemented nor critical to processes currently in place. And he suggested it could be premature to select convergent standards without having in place the points from which they converge.
Medicomp found the criteria for terminologies comprehensive and appropriate. But Mr. Lareau suggested the Subcommittee consider that terminologies might require fully expressed combinatorial clinical concepts in order to be actionable at the point of care and shouldn't be limited to purely atomic terms.
Mr. Lareau cautioned that the balloting process required of ANSI standard developers and the consensus process could inhibit the ability to distribute terminology updates to end users in a timely fashion and advised that it not be an absolute requirement. But he said terminology developers should be encouraged to maintain a practice of timely and public notification of standards development meetings. He agreed with Dr. Bakken that developers should be affiliated with an ANSI standard group, such as HL7. Medicomp believed it was important that all terminology standards developers actively participate in mapping efforts and provide tools to other terminology developers for cross mapping.
Discussion
Dr. Price clarified that in a survey NHS undertook last spring 123 terminology experts described their global pictures of terminology deployment in advanced health economies (e.g., the U.S. or European Community). Over half predicted in five years there will still be multiple, impossible schemes. In 10 years, over half expect to have multiple, interoperable schemes; 25 percent believed there will be a single global scheme.
Noting he sensed from Dr. Price's discussion that, in the U.K., the business case was perceived as primarily helping the government administer the health care system, whereas Mr. Lareau said it had to help the physician at the point of care, Dr. Cohn asked about the business case. Dr. Price said NHS invested tens of millions of dollars over the last decade in terminology efforts and, for probative reasons, had to put together business cases on structured, formal grounds including strategic, economic, commercial, financial and project management cases. Tomorrow he'd tell how NHS combined central initiatives that had structured business cases with local implementation initiatives that resulted in business cases tailored to local requirements.
Mr. Lareau remarked that the publication of the 1997 Evaluation and Management Physical Exam Guidelines provided the impetus for providers to avoid down coding and the clinical business case became a reimbursement business case. Nurses and PAs using point-of-care tools at the front end of the intake process to manage patients with a problem-oriented view of the record got better information with less clinician time. Individual providers looked on additional requirements for data collection as a drain on their time (the one non-renewable resource in the patient care environment) and were protective of it. There was tremendous cultural resistance toward the patient chart becoming open to anyone and standardization of codes at the point of care. The premise that this would help them provide better clinical care had to be proven over time with other methods of payback.
Mr. Augustine commented that his organization, Gambro Healthcare, was implementing a pseudo EMR with electronic signature and that the behavioral component was the most difficult part of the change. He asked what methods Mr. Lareau found useful. Observing that the change was radical and providers needed peer support, Mr. Lareau said Medicomp defined an appropriate number of clinical champions (people who could say, "I did this, you could do it.") in each enterprise who identified low hanging fruit. Implementation of standardized protocols for routine events previously done by the physician or a PA (e.g., systems reviews, gathering a history, and seeding the record for the clinician) led to early paybacks.
Noting one criterion was timeliness (how quickly a developer identified new terminologies, corrections, additions or extensions and incorporated and made them available to the industry), Mr. Blair asked about the response time terminology developers needed to update terminologies and get new terms into their terminology sets. Mr. Lareau said Medicomp had to publish at least twice a year a full, updated, quality-assurance tested terminology set. Medicomp had published quarterly, but vendors said they didn't have resources to go to sites every 90 days. So Medicomp went to a six-month model, and reserved sections of its terminology cross linked to other terminologies updated monthly. New terms (not changes of existing terms or synonymy) were available as down loads, controlled by the vendor, in the six months between release dates. Additions to vocabularies were easy, but modifications were tough and took longer because one had to go back and reflect those changes in previous records. The current model for the RxNorm project strove to support physician order entry and eliminate the local update penalty by having FDA approved products in the RxNorm form within a week. Dr. Nelson suggested asking vendors and system developers about a reasonable time frame for modification of terminologies on a national basis, but agreed with Mr. Lareau that it would be six months to a year.
Dr. Price said the READ Code release cycle changed from every three months to bi-annually four years ago. Occasionally a specific code was published between releases (e.g., a meningitis C vaccination campaign was unexpectedly introduced for undergraduate students). Drug releases were monthly and Dr. Price remarked that people might not be able to prescribe when a drug came on the market between releases without advance notice. But he noted local codes could be added and converted to formal READ Codes with the next cycle. Ms. Bakken observed that the time frame depended upon how rapidly domain content changed. Implicit in everyone's remarks was the realization that technology-based terminology services had to be available to local users between official updates.
Dr. Chute noted any answer was a function of the use case. Any time somebody couldn't act for lack of a terminology, as Dr. Price alluded, was one character of use case. But he said that was a modestly small category of dependency; typically, one had the option of free text opt-outs or local terminology additions. Another dimension to consider was the complexity of update. Soliciting terms from users, reviewing, vetting and accommodating them in a probationary context and incorporating them into a terminology that, in turn, was distributed and incorporated into systems was a long process. And putting a singular date or time frame on any aspect could be misleading.
Recalling that Dr. Price said start up took, $20-to-$30 million and possibly $15 million a year in maintenance costs, Dr. Fitzmaurice said he inferred from the U.K.'s decision to join with SNOMED that the cost of supplying vocabulary to one more doctor was slight compared to costs for the initial doctors, but that it took a large market to amortize that fixed cost at the beginning and spread maintenance costs. He asked if vocabulary might cost so much to build and maintain over time that no one segment of the market could support it and government would have to take collective action--or if the value still had to be determined over time.
Dr. Price said the U.K.'s approach recognized start-up costs were huge. He cautioned the Subcommittee not to underestimate marginal costs of bringing on new users, particularly when bringing on a new-user group specialty. Adding additional users required distribution and training, educational and health tech support, and increased the number of requests processed for modifications as well as the terminology. If one didn't have a centralist, collectivist approach to managing the national terminology problem, whichever approach was adopted had to manage the interoperability issues between the multiple terminologies in a way that enabled capitalizing on them for central information-sharing requirements.
Dr. Nelson said his own estimate of start-up costs for a vocabulary concurred with Dr. Price's, but the National Library of Medicine (NLM) had experience in maintaining vocabulary and disagreed. While he didn't see his front-end private practice beating down the doors for vocabularies, Dr. Nelson noted the potential for the common good and recommended, as a citizen, that this was an appropriate area for government to act. He said the first big win in computer applications in medicine was when people realized they'd save 20 minutes on the phone by getting lab results on their computer. He suggested they were looking at the next big win and, as Mr. Lareau advised, they should keep asking how this helped practitioners. Conscientious practitioners would respond to physician order entry because it would improve patient safety, but the longer it took and the less tangible the results, the harder it was for individuals or institutions to justify a major investment in terminology.
Asked about the role of government and the large fixed and maintenance costs, Mr. Lareau estimated fixed costs to start a terminology at 10-to-15 million dollars; maintenance costs possibly were lower. But, as the terminology became widely used, the responsibility for quality assurance and cross mapping and the distribution and deployment costs rose, becoming more of a burden than maintenance. Noting providers had been burned by computer systems that they had to be replaced every three-or-four years, Mr. Lareau suggested that one of the most important things government could do was provide standards. From the perspective of someone who developed a terminology and distributed it through vendors, DoD's announcement that it would use CHCS2 made the code set sustainable. When the Committee published the recommendations and the Secretary adopted the terminology standards, some of the providers' reluctance would dissipate.
Dr. McDonald questioned whether starting a terminology set was even plausible today. The American Chemical Abstract had all the chemicals, and other places covered a lot of these contents. He asked if it would be easier if they started with them. Dr. Nelson said the number he'd agreed with already accounted for them.
Dr. Cohn reflected on the issue of scope. They'd seen a variety of different terminologies covering different domains and he questioned the word "atomic." In recognizing this variety of needs, the Committee could nibble off domain by domain or look for larger things that might cover more domains. He asked what advice the panelists would give the Committee.
Dr. Bakken said she wouldn't encourage the Subcommittee to look at single purpose terminologies that didn't cover the greater threats from the perspective of discipline or across the care studies. Considering patient safety and health care policy led to a multidisciplinary perspective. Dr. Chute observed that one of the notions of a web of aggregating concepts was that the source from which these high-level classifications were aggregated would be multidisciplinary and, ideally, interoperable and interlocking. The question became what domain one chose for the most targets. Dr. Chute reiterated that they were dealing with a continuum interoperability issue and that these domains blend and merge to enumerate almost silo notions of domain sources. He suggested that might be a reasonable starting point, but doubted it was where they wanted to end up. Dr. Nelson agreed with Dr. Chute that it wasn't unreasonable to say there were low hanging fruit and one could establish single domain vocabularies. But he also noted that other standards that were high-level codes on procedures (e.g., ICD) couldn't be approached without human intervention for mapping and that rules needed to be established.
Mr. Lareau commented that, from a practical standpoint, it would be difficult to define that overarching, multidisciplinary home everything will map to in a short time period. While that was the goal, from a real world distribution perspective one had to start with domains widely implemented and used so people had lead time to implement domain changes.
Panel 2: Terminology
In considering the scope of the PMRI terminologies, Dr. Berglund noted the frame of reference the Subcommittee provided the testifiers was based on a diagram showing different areas connecting contiguous to convergence, which was represented as SNOMED CT. He noted one issue was what was meant by convergence and how this differed from clinically specific codes. Convergence could imply combining multiple systems into one. Dr. Berglund doubted that it was feasible at this time to use just one system, even SNOMED CT that had a lot of content. On the other hand, convergence could refer to enabling interworking between different systems and interoperation. It could be things such as the capability to use a reference terminology as a basis for creating formal definitions for code sets, with support for mapping to other systems. For current purposes, Dr. Berglund suggested that SNOMED CT be considered with the clinically specific group, rather than separated into its own group.
Dr. Berglund observed that message specific and other codes might be out of scope based on the definition of terminologies. Code sets weren't the same as data sets and message formats were different than codes. He suggested considering them separately, based on different criteria. And he asked about the extent to which message specific codes were already covered in other NCVHS recommendations. Other new groups and terminologies Dr. Berglund identified for the Subcommittee to consider included medical literature classification and the Medical Subject Headings (MeSH), though he added it wasn't clear the latter was in scope for PMRI representation. He noted the UMLS Metathesaurus was widely used, mainly in research settings, for mapping between different systems that represent PMRI and might meet some needs described under convergence, though it was based on mapping between terminologies rather then being a terminology itself. Dr. Berglund pointed that the Generalized Architecture for Languages Encyclopaedias and Nomenclatures (GALEN) could be considered either with clinically specific codes or convergence. Released under an open license that allowed free use through the non-profit organization OpenGALEN, one of its advantages was a sophisticated ability to use Galen Representation And Integration Language (GRAIL) language to combine terms. Dr. Berglund also noted that HL7 and X12N weren't really code sets but specifications for messaging formats and it was helpful to distinguish them from codes in terminologies.
Noting that standards for messaging and diagnosis and procedure codes were already specified and widely used, Dr. Berglund advised acting on these first for PMRI selection. He emphasized that it was important that these standards remain current and urged expeditiously specifying the move to ICD-10-CM. Citing the need for clinically specific codes to represent more detailed information, he recommended giving them first priority among the groups without specified standards. Drug codes were needed and fast changes in this domain were a potential issue. Considering that it might not be feasible to have a single standard, Dr. Berglund said continued use of different systems might be reasonable. He noted nursing codes were also needed for more detailed tracking of patient care.
In discussing changes to the preliminary criteria for selection of PMRI terminologies, Dr. Berglund noted different criteria were needed for different groups or subsets of terminologies. Selection criteria for messaging formats and data sets should be different from those for terminologies. And classification systems intended to aggregate should have different criteria than clinically specific code systems. Noting that many selection criteria were based on Cimino's desiderata, Dr. Berglund pointed out that those desiderata were intended for controlled terminologies with multiple purposes and that many of these criteria were best applied only to clinically specific group of codes.
Dr. Berglund advised that the compositional approach or non-semantic identifiers, while important for clinically specific codes, weren't necessary for other categories. He said characteristics of the group of terminologies had to be considered in deciding appropriate criteria. And he noted criteria specifying relationship with existing and proposed HIPAA standards in working with message formats and mapping with administrative standards was important for the consistency needed for subsequent standards specified to interoperate with the current standards. Convergence was also an issue if it was considered to be related to interoperability.
Dr. Berglund said the criteria should not encourage terminology developers to be ANSI accredited. However, timely, public notification of open meetings for standards development activities were important. Dr. Berglund didn't believe it was necessary to use an ANSI consensus process for voting/balloting of new standards: consensus could be difficult to reach, thwarting timely maintenance, and conflicting interested parties could push for opposing changes, stalling the update process, and the technical nature of the standards would limit the usefulness and the benefit of voting. Established processes for updates for established standards should be continued.
Dr. Berglund noted that CAP was an ANSI-accredited standard developer and the SNOMED CT structure was a proposed ANSI standard. Even if SNOMED CT was made an ANSI standard, Dr. Berglund said he didn't see a reason to require that all the terminologies being considered be distributed using this structure. He said licensing for proprietary terminologies should ideally be reasonable and nondiscriminatory, which might require that the costs for licensing be transparent. Dr. Berglund said another consideration was whether it might be reasonable in some areas to use one standard for some purposes and another for other purposes (e.g., drugs). He pointed out that it was preferable to recognize such issues before mandating a standard, rather than rescinding it.
Panel 2: Terminology
Dr. Brown said DVA found the organization of some PMRI terminologies subsets (e.g., clinically specific codes) quite variable in how they were composed and suggested a better schema would be a domain orientation: pharmacy, laboratory, diagnosis coding, history and physical exam terms. Dr. Brown questioned the "convergence" subset. DVA believed a single terminology wouldn't serve all needs and considered convergence more likely to occur as a series of incremental steps with interlocking terminologies and complementary content, as suggested by Dr. Chute.
Dr. Brown noted the administration's need for DVA-specific codes (e.g., eligibility, military periods of service, disability codes). DVA and DoD had standards to meet these needs that could be integrated in the proposed interlocking set of complimentary terminologies. DVA recommended that resources be devoted to developing these code sets as needed.
The DVA believed that interlocking terminology sets relating to the domains of medications, laboratory and problems should have highest priority. Dr. Brown said there was an immediate need for this core medical data for decision-making that, in a structured format, could directly impact patient safety and care. Other high priority items included smaller, targeted vocabulary sets critical for the DVA Electronic Medical Record (e.g., DNR/DNI and advanced directives).
Dr. Brown identified several criteria for selecting PMRI terminologies: (1) timely update process responsive to field requests (there was a large local update penalty and DVA had to supply terminology responsive to field need in order to have standardization), (2) relatively inexpensive to acquire and maintain (open source when possible), (3) centralized, standardized terminology services, (4) flexibility to adapt to change processes and technologies, (5) relative vendor independence; address intellectual property issues forthrightly, (6) meet informational and computational needs for a healthcare terminology--terminology should have clearly defined purpose and scope and fit into an interlocking set of complimentary terminologies required for convergence. DVA did not believe market place acceptance equated to the types of aggregation or quality needed to provide care at the point of contact; Dr. Brown said new developments should be encouraged if they support decision-making, patient care quality and safety. He echoed Dr. Nelson's view of atomic concepts, emphasizing the domain of discourse and that the issue was pre- and post- coordination. Each surface form should have a unique meaning, but one meaning per concept was implied.
Discussing the relationship to other terminologies, Dr. Brown said automated and semi-automated mapping was the goal and explicit semantic models were an important step towards achieving that mapping. He echoed Dr. Chute's thoughts on aggregation, noting its importance in managing assets and doing population studies. Dr. Lincoln remarked that mapping was critically important for the DVA's HDR program. He noted that DoD spent $11.7 million to map 31 sites at the CHDS-1 facilities to the 3M Healthcare Data Dictionary. Terminologies that could support automated mapping would be a big plus.
In terms of implementation practices, Dr. Brown said the tools were added value, but not a necessary criterion for selection. He suggested adding explicit semantic models capable of formal representation and best terminology practices (e.g., ASTM 2087). DVA believed that non-proprietary import and export formats should be supported. Intellectual property issues needed to be addressed. Additions or deletions from PMRI terminologies should never put patient data at risk. There should be explicit and scalable models of cost and opportunities for future migration. DVA was highly supportive of collaboration between government, academic, and industry partners regarding ANSI. Dr. Brown said that while it wasn't necessary for all terminology developers to be ANSI standard developers, he agreed with other panelists that the open process was critically important. He said an ANSI consensus process would be good.
Panel 2: Terminology
Capt. Kelly, a practicing neurology critical care doctor, said DoD could live with the grouping of the subsets of PMRI terminologies. He agreed that these were interlocking, interconnected standards and terminology sets and that what others said about the domains made sense from a clinical standpoint. However, he said grouping wasn't as critical as the fact that they all needed to play well together. He noted that DoD used CDT-3 as its dental standard and preferred that to CDP-2. And he pointed out that the lack of an accepted standard for identifying a person in his or her key role (patient, provider) impeded progress toward implementing a PMRI.
Capt. Kelly said that while DoD had difficulty understanding what was meant by the convergence code set, he agreed that the message and clinically specific code sets were most important. DoD frequently used the drug, diagnosis, and procedure code sets. He said the selected categories were most in need of consensus, would result in timelier and more effective information sharing, and were the codes most frequently used. DoD ranked the nursing and other code sets lower only because these code sets were used less frequently in their facilities and mostly on the inpatient side.
DoD encouraged the use of repeatable model-based processes and tools to create and define terminologies, concepts, and relationships between concepts, particularly for the more complex groups. Selected criteria could be focused and emphasized for these specific groups (e.g., the use of model-based processes and tools should be considered for the message-specific codes and convergence groups). Capt. Kelly emphasized that his staff felt strongly that this was key to developing the more complex code sets and maintaining appropriate linkages between them. He added that the relationships among concepts should not only be explicitly defined, internally consistent, non-redundant, but also model based.
DoD encouraged future development of complex terminology using model-based development. Capt. Kelly said that while DoD didn't believe it was absolutely essential, the ANSI umbrella of organizations added credibility and structure to the process. He concurred that open meetings were a good process. DoD didn't feel that ANSI accreditation for well-established terminology and simple code sets was critical, but in general supported the ANSI consensus process for voting/balloting on the acceptance of new versions of a standard.
Capt. Kelly encouraged the Committee to closely follow the other federal and private sector health activities in the standards arena being done by all the standards development organizations (SDOs), the Consolidated Health Informatics initiative, the eHI initiative, American Hospital Association (AHA) and AMA.
Panel 2: Terminology
Dr. Elkin shared a perspective gained from chairing the ASTM Standard Committee on Controlled Health Terminologies, serving as vice chair of efforts to standardize health informatics within ASTM, and working within HL7 as co-chair of the Templates Group. He remarked that the categorizations of the terminologies they were asked to talk about seemed self-expressed by the terminology developers, and he suggested there was a more appropriate way. He noted five speakers alluded to similar issues and their need for these levels of details for underlying clinical knowledge, aggregations for DRGs, clinical codes, ICD and a network view. Dr. Elkin saw that network view as the decision support that indicated where on the linear scale one wished to be. For instance, if one was designing ICD from scratch, one would pick the purposes for using ICD, note different factors putting pressure on the system, thus entering into the decision rules for selecting the level of granularity from the underlying detailed nomenclature. He noted a tradeoff between complexity and expressivity with compositional systems.
Dr. Elkin recommended adding several new terminologies: the Foundation Model of Anatomy from the University of Washington, the Dermatology Lexicon Project sponsored through NIAMS at the University of Rochester, the Patient Safety Coding System contracted for development by the Agency for Healthcare Research and Quality (AHRQ), and the Gene Ontology created by the Gene Ontology Consortium.
Responding to the charge of creating the right sets of quality indicators for determining PMRI terminologies, Dr. Elkin noted ANSI standard, ASTM E2087. He advised that clinical terminologies were becoming robust enough for clinical use. Examples had already been cited, and he said he was confident there'd be more. He noted competing philosophies on how these should be done; usually there was a cost for purchasing terminologies and always a user cost. Dr. Elkin emphasized that the quality of the terminology directly affected its usefulness.
Dr. Elkin said terminology developers cared about the standards because they stipulated the features associated with quality and, hopefully, gave direction for performing high quality evaluations of their terminologies. He added that terminology users and purchasers also should care because the general philosophy, now, was caveat emptor. Terminology standards assisted users in evaluating both terminologies and the strength of the evidence presented to support their use.
The general characteristics clearly indicated that terminology should be concept oriented. Dr. Elkin agree with Dr. Brown that this was probable and meant they should be non-redundant, non-ambiguous, non-vague and have internal consistency. Dr. Elkin emphasized that all these terminologies should be evaluated within their purpose and scope. He noted that the coverage of the terminology wasn't in the criteria, but was central to its usefulness.
Dr. Elkin emphasized the importance of knowing what these terminologies pre-mapped to and to what they were intended to--and couldn't be--mapped. Systematic or human-readable definitions were needed along with formal computer-readable definitions. Relationships needed to be explicit and the rules of subsumption clearly supported. One needed to know what was intended as a reference terminology and if the atomic portion was identifiable. Colloquial (e.g., interface, status or local terminologies) had to be sustained. Dr. Elkin observed that, with compositional systems that had the expressivity to represent the full breadth of medicine, it was important when immersed in compositional terminologies to realize the need for robust, structured underpinnings in order to avoid unrecognized ambiguity. Dr. Elkin emphasized the importance, compositionality, of: how that was done, how one made atomic and composite concepts, specified pre- and post-coordinated terms and the rule base and types of atomic and pre-coordinated subsets allowed and supported within the terminology.
He said normalization of content assured that exactly the same terms or concepts within the terminology were specified in a way that they'd be seen as the same. Normalization of semantics assured there was no semantic overlap between different semantic operators within the terminology, preventing semantic drift. With multiple hierarchies, it was also important to determine if consistency of view supported explicit uncertainty. How evolving definitions were supported over time was essential to meaning. So was making sure that representational form (e.g., the identifiers) didn't influence the usefulness of the terminology or hierarchies.
Dr. Elkin noted a number of points about maintenance were well covered within the site of the criteria, but he advocated adding more context-free identifiers and persistence of identifiers. He reiterated the need for historical tables: confirming the meaning and semantics of terms at the time they were stored in the record clarified the meaning for longitudinal record clarity and identity purposes.
He said there needed to be clearer version control and obsolescent marking. Redundancy needed to be recognized and people had to be encouraged to work toward language independence. Terminology developers wanted to update on an annual basis, while users invariably wanted overnight updates. They settled on updating the standard at least every three months. Remarking on the need for criteria for evaluation, Dr. Elkin noted many criteria were basically good methods for gaining generalizable results.
Dr. Elkin posed a number of questions around purpose and scope. What was the clinical area of intended and primary usage? Was persistence use expected? Did the terminology have longitudinality? What degree of automatic inferencing was intended to facilitate evaluations and mapping? What was the user/developer extensibility? How was natural language (the user interface most commonly used) supported? Were other functions (specific physician support or linkage to post marketing surveillance systems) intended? What was the current status--was it finished work or a product release? Was it ongoing or something to be evaluated at this point and at what level?
Dr. Elkin noted there were a number of measures of quality in terms of how people utilized tools available within the terminology when the value of the terminology was interlinked with those tools (e.g., the precision and recall of the terminology within its context). Measures included whether it was influenced by a standard search engine with the mapping process, did it have usability, and had it been tested. Was it feasible: had it been tested in real world environment and was that feasibility demonstratable. How closely did it align to the clinical environment? And the validity and generalizability of any studies were linked to their gold standard, relevance, blinding, randomization, test location, sample size, and personnel.
Mayo Clinic believed that following ANSI standards gave higher quality terminologies and promise of improved clinical care through a better understanding of practice and increased availability of decision support at the point of care. Currently a technical specification in ISO, terminology developers in about 20 countries in ISO utilized the ANSI standard. It was the HL7 criteria for use and used by NIH through the Office of the Director for selection of their drug terminology.
Panel 2: Terminology
Reinforcing what Dr. Price said earlier, Dr. Spackman reflected that, from a user perspective, there were only two main ways to categorize or group PMRI terminologies: comprehensive, integrated clinical terminologies and other clinical terminologies of limited scope. Limited scope clinical terminologies might be domain limited (drugs, nursing, devices, emergency departments) or use-limited (messaging, bar codes, reimbursement, user-interface). Fundamentally, users of terminologies needed a comprehensive, integrated role.
Dr. Spackman advised that the way to categorize or group these terminologies should be derived from the users' perspective. User needs and requirements should be listed and prioritized. And in selecting PMRI terminologies the Committee should recommend the development of an integrated whole, tested against the requirements with a revision/updating to meet them.
Recalling many testifiers said they didn't understand what "convergence" meant, but emphasized a need for interlocking, Dr. Spackman suggested they were talking about the same thing. Fundamentally, work had to be done to meet the user requirement for interoperability. Dr. Spackman termed that work convergence.
Dr. Spackman urged the Subcommittee to consider the requirements of software developers and vendors, IT support staff, clinicians and other end-users. He noted many users said the terminology wasn't useful in and of itself. He explained that SNOMED CT was developed and tested with user requirements as the guiding criteria, with input from an industry advisory group; convergent terminology groups in nursing, pharmacy and pathology; alpha testing of the structure and content by 42 organizations in six countries; Kaiser, HCA, U.K's NHS and other large enterprises; and publication of methods for feedback on the Web and in peer reviewed forums. Dr. Spackman asserted another criterion that hadn't been emphasized was evidence that the software vendor community acknowledged the terminology's value by its widespread licensing and intent to use it.
He pointed out that this wasn't a one-way passive process. As vendors developed systems using SNOMED CT, the International Editorial Board actively solicited and responded to their feedback. Requirements and priorities were closely followed; a direct result of SNOMED's recognition of the importance of vendor and user requirements.
Dr. Spackman outlined five main categories of user requirements for PMRI terminology: (1) requirements for patient records and software applications, (2) general requirements for a terminology, (3) software implementation, support and maintenance requirements, (4) requirements specific to certain user communities, and (5) requirements determined by realm-specific strategic imperatives. User requirements for patient records and software applications included: the general requirements for a terminology--the desiderata, software implementation support and maintenance requirements, requirements specific to particular user communities, and requirements determined by realm specific strategic imperatives. Recalling Dr. Price's references to the NHS's strategic imperatives for cancer and heart disease, Dr. Spackman anticipated the need within the U.S. to be responsive to realm specific strategic imperatives. He noted that all the user requirements for PMRI terminology were posted on the Web.
He said PMRI terminology developers in general should be accredited SDOs and required, if not formally accredited, to at least provide an open consensus process for input to revisions and update. An ANSI process (e.g., the canvass or accredited standards committee process) met this need. CAP was an ANSI accredited SDO and the structure of SNOMED CT was proposed and validated as an ANSI standard by the canvass process. SNOMED was in the process of revising the standard based on feedback and were following the open process for that structure standard.
Dr. Spackman focused the Subcommittee's attention on the pitfalls of approaching this as only a process of selection that presumed that picking some grouping or subgroup of the PMRI terminologies listed would somehow meet users' needs. Dr. Spackman said he saw no evidence that users' needs could be met by merely a collection of PMRI terminologies. Rather, users' needs required an integrated whole. He said selection of a small group of complementary terminologies appeared to "satisfice" (a term he said was coined by Herb Simon that meant obtaining an outcome that wasn't optimal, but good enough). But he noted this left significant needs unmet and, from the user perspective, wasn't enough.
Dr. Spackman identified a number of reasons why integration was necessary. Many different types of concepts had to be linked to each other (e.g., procedures use particular devices requiring linkage or interlocking, terminologies dealing with allergies had to link to drugs and substances one could be allergic to). Common concepts occurred across clinical and professional domains. And many different uses and users relied on the same information (e.g., documentation, decision support, reimbursement, research and quality related activities).
Considering cost a criterion, Dr. Spackman said, from a national perspective, it was far cheaper to develop an integrated terminology. Without an integrated terminology, the opportunity and incremental costs of system development and maintenance were orders of magnitude greater. He noted that public policy makers had a sense of urgency about this that echoed the experience and perspective of colleagues in the U.K. He noted Representative Nancy Johnson told NAHIT that if the industry didn't standardize terminology, government would. Dr. Spackman said the quality of health care at this moment linked to an ability to mobilize technology and an assumption of interoperability.
In the absence of an integrated standard PMRI terminology, Dr. Spackman said they didn't have interoperability. He recommended choosing the "best of breed" immediately. He noted SNOMED was consistently rated No. 1 by independent third parties. NLM solicitation cited SNOMED as "the only source capable of providing the necessary services." He advised that the failure of a good faith negotiation effort between NLM and CAP shouldn't deter NCVHS from recommending this choice that met users' needs and was in the best interest of health care.
Dr. Spackman emphasized that CAP stood ready to work collaboratively to achieve an integrated, requirements-driven terminology solution. Major collaborations included Kaiser Permanente, which used SNOMED CT to roll out an electronic health record system throughout its program. Other Kaiser clinical systems already used SNOMED. Noting Dr. Price mentioned had reported that, subject to successful development and testing, after April 1, 2003 any NHS computerized information system being developed to support any clinical information system would use SNOMED CT, Dr. Spackman said CAP considered SNOMED to have the best track record in developing and delivering what users needed in a clinical terminology.
Dr. Spackman illustrated with a diagram how SNOMED CT served as a convergent terminology and others either fully or partially integrated or mapped from SNOMED CT (e.g., it was possible to take NIC, NOC, NANDA and other nursing terminologies and integrate the concepts within the structure of SNOMED CT). Formal agreements with those nursing terminologies explicitly dealt with intellectual property issues and enhanced this converged integrated whole. In terms of relative size compared to other terminologies, ICD-9-CM, CPT for procedures, ICDO within the morphology section, or some of the nursing terminologies, each represented a small piece in the mosaic of interlocking pieces within SNOMED.
Dr. Spackman acknowledged that SNOMED needed to be integrated with the core reference terminology: clinically specific codes (e.g., LOINC), drug codes (e.g., virtual clinical drugs, U.S. proprietary/clinical drugs in FDA/VA), and nursing codes (e.g., PNDS, NIC and NOC). He also agreed on the need to map the core clinical concepts to classifications: CPT-4, ICD-9-CM, CDT-2 and -3, and ICIDH-2. Dr. Spackman proposed building subsets and extensions to support linkage or modeling of domain- and use-limited terminologies and urged the Committee to focus its energy and resources on continuous improvement and deployment of a core versus fragmented development.
Dr. Spackman advised the Committee that it could make major steps forward if it required a proven maintenance infrastructure, prevented analysis paralysis and disruption of a forward trend with industry and providers for "the perfect" or "the promise" of something unproven. He said the marketplace was moving to deploy SNOMED CT now so clinicians, researchers and the public could reap the benefits (interoperability, reduced medical errors, and improved quality of the nation's health) sooner rather than later.
Discussion
Dr. McDonald asked how many vendors had made license agreements, how much they spent on fees for SNOMED, and how many incorporated it. Dr. Spackman said they all licensed it for both development and deployment. Some had licenses since January; many didn't. An initial license was $3,000. There was a cost for deployment in various medical centers. Dr. Spackman said those costs were low and reasonable. Dr. McDonald countered that some costs he'd received from SNOMED weren't low. Dr. Spackman replied that some costs mentioned publicly in this forum were orders of magnitude out of scale.
Noting the first panel said developing terminology cost $15,000,000-$20,000,000 and maintenance was $15,000,000 or less per year, Dr. Steindel asked if the panelists considered that reasonable. Capt. Kelly (used Capt thoughout before, should be consistant) said DoD tried hard not to develop DoD-specific medical terminologies because it was so expensive, but he couldn't confirm the cost. Dr. Berglund also couldn't verify terminology development costs, but he acknowledged they could be significant and expressed hope development could be done without that level of expense. Dr. Elkin said honest benchmarking needed to be done, but two things could make the process much less expensive. Automated tools could generate description logic. Once they had atomic content, many, if not the lion's share, of the description logic could be generated algorithmically, decreasing costs exponentially. Mayo Clinic also favored a model originally developed in the Government Computer-based Patient Record (GCPR) process that used teams of modelers to initiate creation of terminologies. Mayo created editorial boards, giving credit (e.g., being editor of a journal) to junior faculty in academic units that were experts in the field. and costs were defrayed. Dr. Elkin noted such strategies could advance careers while decreasing costs of terminological development for any organization or group.
DVA believed having good tools that employed common technologies (e.g., description logics) could reduce costs. Dr. Lincoln recalled Dr. Elkin recommended benchmarking terminology development costs. DVA began preliminary efforts with the GCPR. Dr. Brown asked if they were trying to fit everything into the terminology. He noted they'd said that wasn't a likely approach, and he suggested they might agree on a way to decide (e.g., periods of service, PNR codes). But he didn't think one figure could be applied, because the jobs were different.
Dr. Spackman understood Dr. Price's estimate to be based on a global integrated terminology that met the needs of U.K., Canada, U.S., Australia and New Zealand, with translations to other languages. Dr. Spackman advised that about $10,000,000 per year was the order of magnitude for maintenance. Anything beyond that was fine-tuning what one did and when, needs met and priorities.
Asked to comment on other testifiers' suggestion that tools for description logic could lower costs, Dr. Spackman said CAP had used description logic tools for development of terminologies since 1996. CAP also used those tools with Kaiser and in training modelers; the focus was integration and interoperability. If that focus wasn't needed, he said a description logic foundation for the terminology might not be required and it would be easier to add things. Committing to an integrated clinical terminology that had a formal logical foundation and used description logic increased the cost because, given the complexity, it took about six months for a reasonably intelligent, well-informed clinician to use description logic in a clinical terminology and become a productive modeler. Improving the tools could make the job easier, but committing to using them engendered much of the cost of building an integrated solution.
Mr. Blair reflected that one new thought from this panel was that, while the cost of developing a terminology was important to consider and reflected in licensing, royalty agreements or federal payments, all that could be blindsided by the fact that a far larger cost to vendors and users was implementation. Mr. Blair asked whether the panelists' found the criteria for selection of standards incomplete (aside from license fees, they didn't indicate implementation costs) and about their perception of those relative costs and what factors, tools and aids a terminology developer might provide to reduce them.
Dr. Spackman noted they'd focused a lot of attention on building predictable structures vendors could import into their systems. Structures wouldn't change release to release. Vendors were given a well-tested set of specifications for how the terminology would be delivered and the functions of the different pieces. They'd also focused on how to make the terminology implementable. Special subsets of the terms and descriptions focused on a particular realm had to be added for primary care physicians. Dr. Spackman said the vendor community tested SNOMED and fed back that they'd really appreciate additional features (e.g., a compendium of the different concepts terms could refer to and the preferred concept in any search application). SNOMED was developing indirect response to feedback from the software implementers to ease transition from just a set of codes and terms to an implemented interface. SNOMED also drew on experiences of colleagues in the U.K. Dr. Price had mentioned that CT, version 3 didn't receive widespread uptake and that they'd learned their lesson and SNOMED CT overcame those barriers. SNOMED focused on and considered part of the integrated terminology whether: one implemented as a big bang or a-piece-at-a-time, exactly how implementation took place, and how one migrated users' legacy data. Dr. Spackman said any one who thought this was about putting a code next to a term had a narrow view of the integrated terminology task.
Capt. Kelly replied that the important thing was to pick the best standard and move quickly. Getting to common standards would enable interoperability of all information systems among federal and civilian health care agencies more than anything else and decrease total life-cycle costs. Capt. Kelly clarified that he wasn't saying that they shouldn't strive for the lowest possible licensing fees, but that those fees would be a small percentage of the implementation costs. He advised the Subcommittee to look at the qualitative benefit of standard data, which he said was at the epicenter of all patient safety initiatives and qualitative ways for improving health care. He cautioned that it was far more important to pick the best standard and shortsighted to not pick one because of cost.
Dr. Elkin said he took the question to be about the human factors that might influence the cost of implementation. He suggested that sound application of human factors engineering principles, and user center design--both in the terminology and tool intended to use it--could greatly affect the cost of implementation, which the group agreed was the bulk of the terminology cost. Dr. Elkin added a downstream cost he considered most important: inaccurate or ambiguous data harmed systems, decision-support efforts, and individuals entrusted to their care.
Dr. Lincoln said DVA also saw tangible benefit in standards and migrating its legacy systems to new terminology standards was a significant issue. With even a free standard, DVA faced substantial migration costs. He remarked that Dr. Brown wrote a paper about the cost of a free computerized PMRI implementing CPRS at Nashville. Pointing out that much of the cost was for mapping and converting a legacy system sensibly to the target information system environment, he advised having the terminology provide tools and methods. Noting the mapping issue DVA faced with its health data repository and DoD faced with its CHCS2 clinical data repository were significant, he emphasized that a way to ameliorate them would be good. Dr. Brown said he'd submit the paper, adding that DVA had experience mapping to LOINC in its laboratory environments; implementation costs varied depending upon where one was and was going. DVA was given a directive to map LOINC across 128 consolidated Vista sites and lab technicians spent a couple weeks at each site. Dr. Brown said DVA already had a fairly compatible lab system, so it wasn't a terribly painful implementation. He estimated it added up to about five FTE.
Dr. McDonald remarked that the implementation they'd heard described covered a broad spectrum. It cost about a nickel to implement the new version of CPT (insert the disk and it loaded): others they discussed (e.g., using an extremely detailed set of codes instead of English) were radical, expensive, and uncertain developments. Getting all the narrative text and coding meant care providers had to make many distinctions and more code levels. Dr. McDonald suggested that if they came up with codes the industry might embellish tools as they had elsewhere (e.g., an industry sprang up and created hierarchies and linkages around NDC). He proposed determining the code and letting the industry construct other knowledge bases on it.
Dr. Elkin noted Dr. Mc Donald's comment could be construed to convey that clinical specific terminologies weren't a good idea. Although it was unlikely Pickless would solve the interspace problem for defining concepts within the detailed health record, Dr. Elkin believed that natural language processing and post processing techniques could do that without a radical change in practice. Dr. Mc Donald assured them that he hadn't meant to imply that. Dr. Spackman agreed with Dr. Elkin that there was a nearer end of the spectrum directly connected to the benefits expected from the additional information, but he added that even that level of implementation would be extremely costly.
Noting both DoD and DVA had a history of developing components for electronic patient records, Dr. Fitzmaurice asked (1) about any collaboration between DVA and DOD in the development of vocabulary and applications, (2) if they'd considered using SNOMED to collaborate for sharing physician's notes and lab and drug information data so a DVA physician treating a former military patient could see information from best medical episodes, and (3) whether they'd thought about using SNOMED to collaborate. Dr. Lincoln said most of the DVA/DOD collaboration in data interchange involved Federal Health Information Exchange. DVA didn't use SNOMED as part of its electronic medical record. Most of DVA's current files for drugs, labs, and person demographics were developed in house, similar to CHCS files. Others (CPT, ICD, and Lexicon Utility files) were based on national standards. DVA was mapping files not based on national standards (e.g., document titles) to emerging standards to allow interoperability between DVA facilities, vista sites and, hopefully DoD. DVA hadn't looked at SNOMED for that. Capt. Kelly reported that DoD had worked with DVA on an initial mapping of their data standards and there was fairly good agreement on many common standards. Both actively participated in CHI (which would be discussed the next day) and DoD hoped to be a "poster child" in terms of sequential adoption of various standards. DoD hadn't made a commitment about SNOMED CT; CHCS-2 used a 3M health data dictionary.
Dr. Berglund reiterated that he'd like to see licensing be reasonable and non-discriminatory for proprietary standards and transparent licensing costs. He asked if anyone could estimate the cost for licensing SNOMED for as broad a use as collaboration between DVA and the DoD. Dr. Spackman said he didn't know how big DVA and DoD were or have licensing fee figures available, but that Diane Ashman would know. Dr. Berglund said he believed CAP was moving to be more transparent in its licensing.
Mr. Blair said that when they talked about using the cost of a terminology as criteria for evaluation and eventual recommendation, he thought he'd heard that they should look at implementation costs in addition to licensing. He asked if the criteria for selection should be expanded. Dr. Spackman advised that the implementation cost of a good standard would be influenced most by what came with the terminology and assisted implementation, begging Dr. Mc Donald's question. Were they talking about the Subcommittee making a recommendation about codes with a name attached and leaving the rest to the industry--or were they talking about doing what Dr. Spackman viewed necessary and accompanying those codes and strings with assistance for implementers, so implementation costs were lowered by a central effort that met common needs? Mr. Blair asked if he was saying to add implementation costs to the criteria. Dr. Spackman replied that implementation costs weren't directly connected to only the selection of a set of codes; potential savings were connected to other things a terminology could provide. He encouraged the Subcommittee to consider the ability to provide them as a criterion.
Dr. Berglund noted implementation costs could be significant and varied, depending upon the users and what they already had. It was difficult to get an accurate across-the-board estimate. Though abstract and hard to sort out, how much the benefits of implementation mitigated the cost was another factor to take into account. Dr. Lincoln reported that DVA looked at four categories of costs: acquisition, implementation, maintenance and migration (federal statute required that data be retained 75 years).
Dr. Elkin agreed with Mr. Beebe: one could grasp this better by dividing it into practical examples of what implementation costs might be. He suggested studying the cost of adding colloquial terms to the terminology, changes in reference terminology, and integrating terminology into the medical record along with training. Noting they'd talked a lot about cost and that health care consumed 11 percent of the gross domestic product (17 percent by 2010, given current health care inflation), Dr. Cohn said they had to be concerned about the value proposition for anything they recommended. Dr. Mc Donald remarked that the challenge of cost in this domain was that, for this standard to be workable, everybody had to use the same code: those who provided the code had a monopoly. Cost was important because there would be political kickback if they did that without competing vendors, where costs worked out in a wash. But they sought common or interlocking codes and so, probably, didn't want competing vendors. If he was a vendor and told he had to use a given code system and could be charged anything, Dr. Mc Donald said he'd object. Dr. Cohn assured everyone that he wasn't saying they didn't need to be sensitive.
Panel 3: Terminology
Dr. Campbell explained that he practiced medicine at the University of Nebraska, was an editorial board member of SNOMED, participated in HL7 and LOINC, and vended terminology services to some 10 enterprises within Last Word Software. Ten years ago the Institute of Medicine painted their vision of the computer-based patient record. That vision rested upon a model with a number of deliverables that clearly required coded or control terminology to create functionality. At the time, most coding systems were primarily billing, administrative, and epidemiological systems; since then, several coding systems had grown to a large enough mass that they could be considered clinically comprehensive. A great deal of work had been done on messaging interoperability and associated reference terminologies and technologies had emerged.
Dr. Campbell said he hoped they'd propose today to discuss the path whereby they achieved the institute's vision, preferably by creating a scenario where cohesive, comprehensive clinical reference terminology could be available to share knowledge between computers--what the Subcommittee called semantic interoperability. Noting a number of publications documented substantial investments in reference terminology creation and development, Dr. Campbell said the question was how to go from listing the terminologies the Subcommittee presented to creating what delivered that vision.
Dr. Campbell suggested that path began by looking, not just at the number of codes available, but also at the architecture for employing them. He encouraged the Subcommittee to consider adopting a slightly different model of how to organize and maintain codes for the computer-based patient record. He depicted a model comprised of three layers. Going from the center to the outside, one went from the clinical to the external (or public/ government) layer. Deliverables for the core layer were shared decision support. The core was a reference terminology brought together from resources available today from American standards developers.
However, Dr. Campbell noted reference terminologies couldn't provide complete clinical coverage for computer-based patient record systems. Other special interests and departmental systems had to be included in the overarching clinical structure (e.g., many nursing concepts and associated care processes historically had been poorly represented in reference terminology systems and had to be brought out). So the second layer in this model comprised of modeled code or legacy term sets, or other systems necessary to the clinical function of the CPR, was modeled on a negotiated basis with a clinical reference core, within the construction of an information model. Dr. Campbell defined an information model as an agreed upon set of definitions that took core reference concepts and modeled more complex or context-rich concepts. An example from that morning was the question of an order. An order might be modeled as an ordered set of a number of reference terminology codes, but the agreement about an order's basic structure was modeling that took the reference terminology into the broader arena of the entire clinical record. The third layer was the administrative and governmental layer, with interfaces and a variety of users who had important needs for the CPR data that weren't primarily clinical.
The core convergence reference terminology was designed to be comprehensive across all that was needed for multidisciplinary clinical care. Its primary editorial policies were clinical. That was its goal and how it was maintained and developed. Dr. Campbell suggested that, over time, it had to be edited (based upon feedback from guideline organizations that used the material to implement knowledge in running clinical systems) to ensure it delivered knowledge. Preferably, it was maintained by an SDO with a hand off to clinical professional societies who took over its increasingly sophisticated clinical content.
The features of the layer two set were modeled codes that integrate important clinical terminology. These terminologies weren't designed to be comprehensive (e.g., nursing term sets were never designed as an entire medical record, but were an important part of the CPR). This layer served departmental, special purpose, messaging systems, professional and perhaps interface terminologies that made the CPR easier to use. Everything within this layer was constructed using the reference core, but modeled into the required context and behavior based upon meaning values from the original legacy owners. These two layers had to be maintained in negotiation; a dialogue between the reference terminology and modeled code sets was essential.
Layer three was code sets required for administration, finances, billing and epidemiology. Fundamentally not patient care oriented, they follow editing rules promulgated by organizations that need them. (The purposes of ICD-9-CM were not precisely the purposes of clinical care.) The reference agency that managed these code sets basically controlled their content. Dr. Campbell suggested, relative to what others suggested that day, that this layer probably was maintained in a different way and held to other standards than the internal clinical layers, which followed basic concepts of how a reference terminology was maintained.
Dr. Campbell proposed three major priorities: (1) setting up a collaborative strategy to define a core reference terminology that could serve clinical medicine, (2) establishing knowledge-based solid mapping from the reference core to the third layer (if clinical data didn't yield necessary administrative and billing data in a relatively seamless way it wouldn't be acceptable in the American marketplace and vendors needed solid mapping to install and maintain computer systems), and (3) instituting changes in clinical content coverage that found within layer two. Dr. Campbell disagreed with previous speakers and believed it was important to converge nursing terminology with the rest of clinical care, creating a universal multidisciplinary clinical care model. Based on efforts initiated by the Institute of Medicine's projects on error reduction, he suggested that the drug information sets (e.g., Multum, First Data Bank) be given priority for early instantiation into the model layer.
Though it had gone unrecognized in recent years, Dr. Campbell observed that when SNOMED RT was released, LOINC and SNOMED basically converged, at least in the laboratory area. He suggested that convergence be fostered so that SNOMED CT, LOINC, and NLM's Rx Norm project could be considered as the logical candidates for realizing a core reference terminology in a short time frame. Noting that because of negotiation with NHS and its interest in maintaining its clinical project resource system, SNOMED planned to maintain a national clinical formulary for UK, Dr. Campbell suggested that considered a part of the strategy in this core reference terminology.
Dr. Campbell pulled the ICNP codes for nursing into the second layer, because a number of nursing colleagues referenced their importance in terms of convergence. He also mentioned ICBO and ICPC under medicine. He noted ICBO was subsumed within SNOMED and shouldn't represent much of a problem, though the primary care area was poorly represented in many systems. ICPC was mentioned most often in reference to expanded codes for primary care.
Referencing the three-layered model and his suggestions about which codes should go into each layer, Dr. Campbell offered observations about criteria for code or terminology selection for each layer. While agreeing that it would be great if this were free, he pointed out that often free didn't cut it in terms of function. He balanced that statement with the added criterion that core reference terminology developers must have a business model that organized and funded timely evolution, distribution, and perpetual maintenance of the terminology.
Considering clinical scope and organization, Dr. Campbell said the question of completeness only made sense looking at two layers. From the administrative perspective, completeness was in the third layer. In the clinical sense, completeness was in the inner two layers that held the reference terminology and its cooperating model systems. Approaching it that way, Dr. Campbell suggested that core reference terminology developers must cooperate within their negotiated domains to produce complete clinical concept definitions and a cohesive semantic network defining all central domains of the clinical practice record. He said the most pragmatic thing was to direct SNOMED, LOINC, and the Clinical Drugs Project to cooperate, decide where to develop and how to interact, and collaborate in an orderly way, because that was the fastest way to reach their goal.
Considering next interaction between the core and second layer, Dr. Campbell suggested that core reference terminology developers must have cooperative agreements with layer two developers so that content and editorial evolution were orderly and cohesive. Remarking that all of this implied an overarching information model for what they were doing existed within the inner two layers, Dr. Campbell suggested that the Subcommittee consider as an added item to their work task that a reference information for definition of complex and content-laden concepts be developed, and an SDO be identified to maintain this model, which was necessary for proper interaction between the inner two clinical layers.
Considering concept orientation, Dr. Campbell noted that much of what the Committee had done addressed issues of the reference terminologies. He suggested that many of those criteria couldn't be applied to layer three, because they had no control over those terminologies. But he encouraged the Subcommittee to consider what relationships were supported within the core, because these ultimately became part of the knowledge base used to deliver clinical decision support. Therefore, he emphasized that reference terminology relationships within the core must be sufficient in definition and scope to support clinical guideline implementation. They probably didn't need the most complicated models in order to be successful with the decision support they could implement. But they obviously had to have a good subsumption hierarchy and probably needed a modest number of additional relationships. .
Dr. Campbell identified his assumptions for each layer. He believed core terminologies provided domain content to support complete clinical reference terminology. Layer two terminologies modeled, defined, and enhanced important legacy and departmental classifications and terminologies. Both mapped to layer three terminologies to support reimbursement and statistical classifications. Dr. Campbell noted he'd agreed that maintenance releases must be timely, but he added that varied with different portions of the terminology. And he emphasized the need for a sound business model. He also suggested that core reference terminology developers collaborate to have domain content reviewed and regularly edited by clinical professional societies. And he advised that the core reference terminologies should be maintained by an open editorial process in conjunction with professional societies, preferably within an ANSI accredited SDO. Referencing three terminologies he'd recommended that were international in scope, Dr. Campbell also suggested considering proper and useful relationships with international standards such as were happening in SNOMED CT.
Panel 3: Terminology
Ms. Haber served as the coordinator and project officer for efforts related to both division- and institute-level vocabulary development, including the disease and drug terminology for NCI Thesaurus, a controlled reference terminology to sort NCI's vocabulary needs from basic and translational research to clinical care. CIPS Division was responsible for the production of Physician Data Query, which was NCI's database of evidence-based review summaries on cancer treatments, screening, prevention, genetics, and supportive care. As providers of information at the institute level, NCI also worked cooperatively with the NCI Center for Bioinformatics on NCI Thesaurus and NCI Metathesaurus, institute-level and meta vocabulary environments that were core components of the Center's Enterprise Vocabulary Services.
Ms. Haber said the issue of specifications and standards for PMRI and related terminologies acutely concerned NCI, which was both a user and provider of vocabularies related to NCI's diverse missions. Ms. Haber remarked on a historical disconnect of data collection priorities and methods, with information housed in discrete systems, and coded using different vocabularies for portfolio management, clinical care, research, public and provider information, and epidemiology. She noted there also were significant differences in the meanings attached to particular terms in the vocabularies.
Ms. Haber reported that NCI had an inheritance of legacy systems. Achieving connectivity between established mission critical systems involved complex issues of data conversion, storage, and retrieval, especially for retrospective research. Barriers included not only disparate electronic systems, but also, at times, the lack of an electronic system. Ms. Haber noted that health information models were also changing with cancer and healthcare shifting from an emphasis on treatment to prevention. Even more fundamental was the rapid transformation of disease models from traditional to the new genetic and molecular classifications. She noted the pace of change in science, technology, and methods. Vocabulary content and structure had to quickly reflect these advances in order to remain useful, while still permitting identification and retrieval of artifacts tagged and coded using earlier systems and vocabulary versions.
Ms. Haber said NCI approached these issues with Enterprise-level initiatives to integrate both intra and extramural data resources in collaboration with outside partners such as the clinical trials cooperative groups. The Center for Bioinformatics was developing tools and technologies to support these efforts, including a cancer data standards repository for standard meta data for clinical cancer research. The Center also collaborated with CIPS OC to develop NCI Thesaurus. Ms. Haber said that effort was spurred by the realization that linking related concepts in the fields of population, genetics, developmental therapies, and disease was essential to NCI's mission.
Ms. Haber said the rapid evolution of information models in cancer mirrored the challenge of addressing knowledge systems at the broader level of standardized PMRI. For the NCI, this fundamental and ongoing transformation meant that any proposed information model not incorporating the assumption of frequent change would fail. Rapid changes in information models and technological capabilities meant that standards development had to anticipate the future evolution of PMRI. Vocabulary standards had to insure the adaptability of a terminology or information model to incorporate intellectual content changes. Ms. Haber emphasized that the capability, commitment, and funding to support frequent and regular cycles and the flexibility of a terminology structure to map readily to other major component vocabularies were perhaps even more important indicators of the ultimate success of a standard vocabulary than an examination of its contents.
HIPAA requires portability, accessibility, and security of data, and hence, calls for standardized interchange of information across systems. Stable, reliable, and consistent meanings of vocabularies and codes are a fundamental component of this requirement. Integrated coding vocabularies are also essential to building and expanding the NHII, public, private, cooperative, and federal eHealth, and other related efforts to provide structures and information resources for personal, provider, and public health data needs.
In examining vocabularies being considered as standards, Ms. Haber advised it would be instructive to have a table showing coverage of the various vocabulary and code sets by content domain, as several (e.g., ICD, LOINC and SNOMED) had crossover content. She said providers should be asked to indicate depth and breadth of coverage, as well as other quality indicators for each domain. She noted the omission of some vocabularies necessary to provide coverage in special domain areas (e.g., ICD-0-3 for oncology reporting).
Ms. Haber recommended that the PMRI selection process accord priority to terminology development efforts linked to and integrated with related clinical document architecture, format, and messaging standards such as the HL7 vocabulary standards development process. She noted representatives from major divisions of NCI would more actively participant in the standards activities of HL7 as they realized that an inclusive, collaborative process offered the best opportunity to leverage the power of multiple developers. Cooperative endeavors at the federal level (e.g., NLM's Rx Norm and the DVA's National Drug File Reference Terminology developed in cooperation with FDA, NLM and NCI) might also serve as models for collaborations.
Under the category of General Criterion, Ms. Haber added that vocabulary sources should have explicit, defined, and regular processes for technical and content quality review and assurance. She emphasized again under maintenance, the necessity of a consistent, reliable, rapid update process sensitive to the changing needs of the national health community. Content control and regular review by domain experts was essential for usability and validity and should be a hallmark of every specialized vocabulary or vocabulary domain. She said PMRI terminology developers should be encouraged to be accredited as ANSI standard developers. While market acceptance of well principled terminologies could create de facto standards, she said standards that purported to serve the needs of a diverse community ought to be open to community involvement.
Ms. Haber said PMRI terminologies couldn't be examined without looking at tools available to create and maintain them, and the integrity of the mappings between them. She noted accurate mapping often was a task as complex and as much an intellectual product as the vocabulary. She added that ownership, maintenance, and integrity of mappings between major required recording vocabularies for PMRI should be as explicit as requirements for maintaining terminologies. And she noted a lack of sophisticated tools for these tasks was another barrier to integration.
Panel 3: Terminology
Dr. Overhage's testimony drew on his experiences over the last 10 years building a regional electronic medical records system, a microcosm of the tasks the Committee was charged with undertaking. He said he found the categorizations of codes proposed in the Committee's Scope of PMRI Terminology Table in its August 2000 report to the Secretary difficult to work with in terms of segregating out questions to be asked in order to appropriately analyze these domains. Dr. Overhage advised it was more useful to take an object-oriented or entity approach on whether these correspond to specific terminologies individually or portions of larger terminologies. He suggested it was more useful to think about units, orders, medications, allergies, clinical observations or results than aggregated or nursing codes and that this approach kept concepts more closely aligned. One often encountered the same concept labeled differently in other terminologies (e.g., adverse drug events). He also said this approach promoted reuse of the data. Whatever they were going to do with these terminologies, both the clinical processes and other uses of the data had to be pulled together. Whether that happened through mappings or other approaches, it was critical that the concepts be carried forward from the clinical processes in a faithful way.
Dr. Overhage emphasized that in talking about terminologies they often talked about more than one thing. He pointed out that for any particular domain, typically at least three levels or subdomains of terminologies were conceptually needed. First were questions or variables: e.g., glucose concentrations measured in serum, a Glasgow coma score assessed by an emergency room physician, or a discharge diagnosis encoded by a hospital discharge clerk. Second were answers, concepts or results. Sometimes numeric (e.g., the value of the glucose concentration) or free text (e.g., portions of a document); often these were coded as well; e.g., names of organisms or simply "positive" or "negative." In the third subdomain were batteries of codes bundled for convenience (e.g., electrolytes, test elements belonging to study members 1325, or vital signs).
Commenting on the criteria for looking at these different terminologies, Dr. Overhage noted they'd heard a lot that day about cost being a critical barrier and he noted one criterion was that the terminology be relatively inexpensive to acquire and implement. He also emphasized the importance of acceptance in the marketplace. Free didn't necessarily insure quality or value; acceptance and use were a barometer of how good or useful a terminology might be. Additional criteria to consider included: (1) that these terminologies reflect clinical realities (e.g., combination drugs), rather than being conceptually pure, (2) codes should be context-free, as several presenters noted, (3) compositional strategies needed to be consistent across the adopted terminologies and (4) consistency with an overall information model such as the HL7 RIM.
Dr. Overhage suggested that one way to prioritize was to look at the quantity of electronically available data: what could they take advantage of today? Another factor was the relative value of that particular domain or type of data on a global basis. For an individual patient or need, particular data was critical, but certain things were valuable across many domains and uses. A third factor was the availability of terminologies that were "good enough places to start."
Laboratory results, which already reside in electronic databases in most institutes, were targets for using those criteria and widely applicable for patient care. One potential candidate terminology was LOINC, designed to provide a universal identifier for observations, so information from different sources could be combined in one electronic medical record system or research management database. LOINC addressed all three subdomains, dealing primarily with batteries and questions (variables or observations--serum glucose or blood culture, batteries such as the SF36 set of questions for patient status assessment, a comprehensive metabolic panel or a collection of laboratory tests to assess the patient's overall metabolic state). LOINC started out in 1995 with 4,000 entries, today there are over 31,000 entries.
Dr. Overhage reviewed ways LOINC had succeeded. Government and other organizations had adopted it. Large pharmaceutical organizations and consortia considered LOINC an appropriate tool, and CDISC, a collaboration of large pharmaceutical groups for reporting clinical trials data, had incorporated LOINC as the key standard. Large reference laboratories utilized LOINC both as a tool for sharing data with clients and for resolving internal mapping problems as they purchased and consolidated services from regional labs. Kaiser, Partners of Boston, Columbia Presbyterian and other large health care systems and managed care organizations used LOINC to aggregate data from many sources. LOINC was successful with instrument vendors and test developers, labeling at the source as tests were defined and developed. And LOINC had been translated into multiple languages and adopted as the standard for a variety of initiatives throughout the world.
Dr. Overhage emphasized the importance of keeping terminology domains conceptually pure and clinical processes connected. Mapping could be a partial (but always risky and difficult to maintain) answer. Another way to do that was to differentiate concepts and structures: e.g., an adverse drug event as an occurrence of some clinical finding associated with a drug wasn't new or different, but a relationship rather than a concept. Dr. Overhage reiterated that LOINC satisfied many of the criteria, at least for the laboratory domain.
Panel 3: Terminology
Mr. Beebe said AMA believed that a patient's medical record should include sufficient information for physicians and other appropriate health care professionals to assess previous treatment, ensure continuity of care, decide upon further treatments and clinical activities, and avoid unnecessary or inappropriate tests or therapy. The medical record was the primary source of information for a patient's overall health care, meeting all clinical, legal, and administrative requirements. PMRI was medical and health care data about an individual, including facts, observations, interpretations, plans, actions, and outcomes. This description applied to either paper or electronic formats and the strictest protections of patient privacy must apply to such information. Comparable and accurate PMRI would assist in realizing the clinical utility of such information and enhance its value for clinical research and epidemiological purposes. At the same time, Mr. Beebe noted efforts to enhance comparability and computer operability must not detract from the fundamental clinical purpose of such information that focus, first and foremost, on the needs of patients, their physicians and other health care professionals in the institutions and facilities where they receive their care.
AMA believed that terminology supporting PMRI should be comprehensive and include all the clinical terms used by members of the health care team involved in record writing. Despite the need for sufficient terminology breadth to accomplish all the varied purposes of PMRI, AMA asserted that terminology selected be limited to discrete, non-overlapping functional areas, with one terminology per area. AMA was also concerned about accurate, consistent implementation of terminologies. Standard implementation guidelines for terminologies were essential for uniform national application of the code sets. Mr. Beebe said the purpose of administrative simplification wouldn't be achieved if health plans and providers implemented and interpreted medical data code sets as they saw fit. AMA recognized the limitations on the human element of code application and was concerned about the need for education on all the code sets or the development of computerized tools to achieve semi-automation in applying the codes. AMA believed that terminology supporting PMRI should be able to be cross-referenced with other terminology now and in the future.
Two years ago AMA commissioned a study conducted by Medical Systems Development identifying a comprehensive list of core clinical data elements for electronic medical record systems. The data elements identified were derived from a large number of resources including uniform data sets, accrediting and licensing agency requirements, industry standards, selected electronic medical records literature, and electronic medical records vendor system specifications. Mr. Beebe said the components of an electronic medical record that most concerned practicing physicians were the core clinical data elements and he noted that practice management software vendors continued to fall short in the development of their products. AMA believed the core clinical data elements (e.g., patient identification and demographic data; special patient health conditions; allergies; immunizations; health promotion/disease prevention; past medical history; family and social history; laboratory and other diagnostic orders and results; therapeutic services and procedure orders and results; medications prescribed and results) were similar to the object-oriented model put forward by Dr. Overhage. Mr. Beebe recommended the Committee prioritize the PMRI terminologies based on current use, practical applications and acceptance in the marketplace, as well as benefits of the system.
AMA believed that the criteria identified were acceptable towards selecting appropriate PMRI terminologies, however Mr. Beebe said it should be divided into essential and desirable criteria, as well as having consideration given to domains. He said the four criteria derived from the PMRI guiding principles (market acceptance; interoperability; comparability of data; and ability to support data quality, accountability, and integrity) should also apply to PMRI terminologies. As the Committee stated, market acceptance was primary because it identified standards that were implementable, cost justified, and flexible enough to meet the needs of most relevant marketplaces. Mr. Beebe pointed out that CPT procedure codes: (1) were maintained and updated by the AMA since 1966 and had yearly updates; (2) many products that incorporate CPT had been licensed, involving over 350 vendors; (3) physicians and many other health care professionals use CPT almost exclusively for reporting all health care claims; (4) One hundred percent of health care institutions use CPT; (5) all healthcare third-party payers that process medical claims use CPT, as well as CMS, the Department of Justice, DoD, the Indian Health Service (IHS), 42 state government agencies and Puerto Rico; (6) CPT codes had been used by the Medicare Program since 1983 and associated with RBRBS since 1992; (7) CPT was used internationally in South Africa, Mexico, England, New Zealand, Hong Kong, Belgium, Canada, Italy, Latvia, and Venezuela--translations for Japan and Israel were underway; and (8) CPT was working on achieving greater interoperability by developing a formal data model using description logic and subsumption hierarchies that Mr. Beebe noted were mentioned often that day.
AMA believed that all coding systems adopted as HIPAA standards should have an open updating process and any interested party should be able to submit proposals for additions and modifications. In addition, a responsible panel or committee of experts representative of a broad cross-section of relevant stakeholders should maintain the terminology. AMA didn't believe it was necessary for PMRI terminology developers to be ANSI accredited, however, the organization maintaining the code set should ensure continuity and efficient updating of the standard over time. AMA believed that PMRI terminology maintainers should meet the guiding principles expected of current maintainers of the selected standards. These standards include: (1) improve the efficiency and effectiveness of the health care system by leading to cost reductions or quality improvements; (2) meet needs of the health data standards user community, particularly health care providers, health plans and health care clearinghouses; (3) be as consistent and uniform (e.g., data element names, definitions, codes and privacy/security requirements) as possible; (4) have low additional development and implementation costs relative to their benefits; (5) be supported by an ANSI-accredited standard setting organization or other public or private organization that ensures continuity and efficient updating; (6) have timely development, testing, implementation and updating procedures to achieve administrative simplifications faster; (7) be technologically independent of computer platforms and transmission protocols except when they are explicitly part of the standard; (8) be precise, unambiguous, and simple as possible; (9) keep data collection and paperwork burdens as low as feasible; (10) and incorporate flexibility to adapt more easily to changes in the health care infrastructure (new services, organizations and provider types) and IT.
A recent study commissioned by AMA found that seven-out-of-ten physicians indicated that they and their staff needed more information about transaction, privacy and security standards. And 72 percent of responding physicians didn't believe HIPAA would save their practice money. Although AMA had long held that increased use of electronic financial and administrative transactions could increase the efficiency of physicians' practices, as well as the health care system overall, Mr. Beebe said the survey indicated physicians were frustrated with impending HIPAA regulations. He suggested development of PMRI standards build on this lesson by focusing on market acceptance and clinical needs. He said AMA understood that the responsibilities of NCVHS were to evaluate and recommend PMRI and applicable terminology. However, he emphasized that NCVHS's responsibilities on this issue were fundamentally different than those for HIPAA administrative transactions: no federal legislation required implementation of PMRI standards or the code sets that would apply to such standards.
Mr. Beebe advised keeping in mind that if used uniformly, PMRI standards must first and foremost meet the needs of the clinicians where care was provided. Standardized terminology could be extremely beneficial to the primary users, the physicians providing medical care to patients. Other uses (e.g., research and fraud detection) were secondary. Therefore, he said HIPAA standards for transactions in privacy and security should be implemented and demonstrated to be effective before PMRI standards were established.
Discussion
Dr. Campbell clarified that the core was to be a reference terminology that supported all basic types of clinical care (e.g., diagnoses, assessments, findings, symptoms, procedures). Layer two would bring in departmental or special needs and converge with the core. He noted these concepts could end up forming a fairly complex model: it might take eight different attributes to specify an order (i.e., the procedure or result code specified the action intended and when it started, stopped and its frequency). That information model defined how the reference core related to the definition of the order in the second layer. But Dr. Campbell emphasized that the entire inner two layers would be consistent; internally they'd share many attributes.
Asked if it having that core in layer one was necessary for a good PMRI, Dr. Campbell replied that pragmatically they were at a point where they'd developed a critical mass. It would be imprudent to abandon that investment in American SDO development. Dr. Campbell encouraged moving forward.
Mr. Blair said he was excited by Dr. Campbell's and Dr. Overhage's presentations and that the Committee needed to look to both to see how to go forward. He asked about conflicts and compatibilities and if they thought their visions were mutually exclusive or whether Dr. Campbell's layered model might be one dimension complemented by another that embodied Dr. Overhage's model. Dr. Overhage said the two visions weren't inconsistent and could be complementary. He pictured Dr. Campbell as having the world in the inner layer of his core model. Dr. Overhage said he broke that down into different domains that could be sorted, categorized and analyzed, asking where were they today and identifying migration issues (e.g., drugs were almost always represented in NDC codes, but every laboratory system had different codes for laboratory results and there were few commonalities.
Dr. Campbell said he didn't think Dr. Overhage and he differed in basic ways. He'd detected in Dr. Overhage's presentation flavors of editorial perception that grew out of the HL7 RIM, which he said was a legitimate way to look at how to structure the information system. Relative to the information model he'd referenced, HL7 RIM was designed to determine the appropriateness of the construction of HL7 message packages and Dr. Campbell pointed out that it had to change. He suggested that the information model he described had to occur in negotiation with the standards developers and focus on developing this broader set of definitions of a wide variety of constructions where HL7 hadn't much interest. HL7 had been responsible for a lot of message interoperability, but hadn't looked at much semantic interoperability. Dr. Campbell personally believed they had to head towards semantic interoperability if they were to achieve intelligent clinical information systems.
Ms. Haber noted that HL7 had extended beyond simply messaging. HL7 was looking at comprehensive vocabulary building and clinical document architecture efforts to collaborate on clinical trials information. She said HL7 was a place to look for integrated vocabulary concern and development.
Noting the discussion about the different contracts, Dr. McDonald clarified that HL7 was very interested in vocabulary, but that was still undefined. He emphasized that considering vocabularies in terms of a particular message structure was a strong idea, because it kept one from wandering into spaces that didn't exist or no one cared about. Dr. McDonald said he wasn't disagreeing with Dr. Campbell's model per se: dealing with things people really did kept one honest. One could test and see if they worked.
To emphasize that difference, Dr. Campbell referenced a project underway with IDX aimed at creating an environment where they could transport guidelines from one institution and computer system to others. Semantic interoperability required these features and they were trying to do them today. Dr. McDonald said he didn't disagree with the theory. Semantic interoperability was what people meant when they said they had messages, but they needed standards codes in them. Coming back to a model and seeing how something fit in a field and if it worked in real life was a good way to test ideas.
Dr. Cohn remarked that level two seemed to be a heterogeneous grouping of code and data sets. Dr. Campbell explained that, as Dr. McDonald often insisted, the computer-based patient record had a far broader range of components than they were sometimes willing to account for. They shared many elements and the question was how to bring them together in a model that unified. With a computer information system that knew the relationship between what one ordered and the result, order sets could become customized, problem-oriented data displays. The information model was created in the second layer and, with the various terminologies cooperating, one ended up with a modeled environment entirely consistent throughout the clinical layer. Dr. Cohn said he understood the model, but not why many other elements were there. Acknowledging there were "funny" neighbors, Dr. Campbell gave an example. He served as a middleware vendor of interface terminology for other IDX sites using SNOMED, maintaining entrance terminology in order to do physician problem list recording which was simply provided by creating pre-coordinated concepts. SNOMED's goal wasn't to create every concept known to man; compositional environments focused on that.
Dr. Cohn said he was confused by the way Dr. Overhage described LOINC; he'd thought the identifier accompanied the results coming back; usually one didn't know analities when ordering. He asked if they were saying this wasn't the answer, but the question. Dr. McDonald replied that one thought of a question as a synonym for a variable or a parameter (e.g., the OBX segment of HL7, the variable, could be sugar glucose, diastolic blood pressure, the Glasgow coma score, or a survey instrument question). The question was connected to an answer, whereas a variable had a value. Calling the result a question with an answer returned was synonymous and perhaps easier for non-technical people to grasp.
Referencing Mr. Beebe's written testimony ("Therefore, the AMA believes the government should limit its focus to broad recommendations for medical terminology framework with the specific code sets maintained by the private sector"), Dr. Fitzmaurice said he didn't think anyone had a problem with the private sector maintaining code sets. But he wondered if Mr. Beebe meant the government should limit its focus to broad recommendations, rather than mandate a particular terminology, code set or functions when it interacted with the private sector. Mr. Beebe said he meant the Committee should make recommendations for criteria, rather than specific code sets. Dr. Fitzmaurice contended that the Secretary would be better served if one of the Committee's functions were to make specific recommendations about codes sets or terminologies. Mr. Beebe replied that, if the criteria were specific enough, code sets that went with the criteria would be implied. Dr. Fitzmaurice said that was like wiring it to a particular terminology or code set. Mr. Beebe agreed.
Recalling that Mr. Beebe had singled out criteria for selecting PMRI message format standards that the Subcommittee hadn't thought could quantify, Mr. Blair asked which ones Mr. Beebe wanted brought back and how they might quantify or measure them, noting uncertainty about that was why the Subcommittee left them out. Mr. Beebe said he believed many of the more practical elements on the list of HIPAA guiding principles should be relevant for PMRI standards, because they dealt with basic issues of market acceptability and clinical context used by physicians and other health care providers on a day-to-day basis for patient care. He agreed with Dr. McDonald that they shouldn't go into more ephemeral notions outside of what was currently done, but needed to pay attention to practical, rubber-meets-the-road issues. Mr. Beebe clarified that providers and payers weren't charged for transactions from hospitals. AMA charged a flat rate of $50 for the data file and $10 per user for licensees. Information about volume discounts was on the Internet.
Dr. Steindel said one of the most fascinating things he'd heard was Dr. Campbell's idea about looking at things. He asked what was contained in the core; how it related to layer two; how much of layer two was subsumed in the core; and how much of the relationship between the core and layer two was contained in the information model. Dr. Campbell recounted recent discussions with the nursing associations where they defined a goal as a finding they wanted to observe at some future point, possibly related to a specific intervention. From that, one could construct a semantic model detailing attributes, evaluation dates, procedures, and responsibilities. Dr. McDonald noted HL7 models explicitly defined goals that way. Dr. Campbell said standards organizations had discussed this widely. It was an example of how a concept from a nursing domain negotiated its definition with a reference core. The legacy terminology (goals that were part of past nursing terminology efforts) could be modeled and incorporated with legacy terminology as part of the convergent terminology systems. Dr. Campbell said they had a way to build goals for new care plans--ones that didn't require as many new elements, but created the context that linked those elements together.
Dr. Steindel remarked that this went along with what Ms. Greenberg had said about the linking between clinical document architecture and the structure in which the message was sent, and the way in which context could make terminology less ambiguous. Dr. Campbell noted a constant tension between SDOs and these schemes, which were really context-ridden situations. All could be served, but they had to come to some agreement about their definitions.
Dr. Fitzmaurice asked Dr. Campbell if he saw the core layer as being equal to RIM. He also recommended that, when an SDO developed a standard, it map itself to the core and then back out to the function in level two. Dr. Campbell said he'd deleted that from the requirements list, because the reference terminology model was created as part of the process. The most mature candidate reference terminologies were enabled and encouraged to collaborate.
Asked why some classifications were in the second layer and others in the third, Dr. Campbell explained that both Dr. Overhage and he were pragmatic. A pragmatic that separated some classifications was whether they had any realistic likelihood to manage their content. ICD obviously had its own existence and purposes. ICPC was meant to be classifications of activities for primary care. Primary care doctors reported that some things in ICPC weren't well represented in SNOMED. But Dr. Campbell suggested that, if there was a dialogue between SNOMED and ICPC and definitions, they could come within the core and better serve primary care physicians. Ms. Greenberg noted there was also a dialogue between ICD and ICPC and comparability between DSM and ICD. Dr. Campbell said another reason for a line between the second and third layers was that some things were primarily necessary for the clinical environment. One could argue whether ICPC was necessary; Dr. Campbell believed it seemed logical.
Mr. Blair said he had the impression that there was a conceptual difference. He suggested that HL7 had developed a RIM that enabled it not only to develop new messages, but also improve interoperability for messages. That was its purpose and, in doing that, one drilled down to the semantic level for greater specificity. But Dr. Campbell started with a core purpose of semantic interoperability. Dr. Campbell indicated a need for harmonization or mapping between RIM and the core of his semantic interoperability construct. Mr. Blair pointed out that might be different than drilling down from the HL7 RIM.
Dr. Campbell agreed: from the standpoint of semantic interoperability to support guidelines, a disease or disorder should have attributes of what caused it and what organ it affected. Those were common inferences and one might want to reference all streptococcal cases, whether impetigo or strep pneumonia, pneumonia, or endocarditis. If the core reference model's definitions included an etiology attribute, those categories were supported readily. Vendors didn't have to add that knowledge. Dr. Campbell noted HL7's interest was different. HL7's driving interest in their model of assessments, data records they connected to, and fields was to ensure that one could look at an HL7 message for an order and know when any element wasn't included. RIM clearly indicated what should be part of that construction.
Noting skepticism, Dr. Campbell said HIPAA requirements could create more work than they could do. But he added that the Committee had the ability to nudge that next step forward, because many in the vendor community were waiting for clarity about the tools they would use.
Dr. Overhage differed with Dr. Campbell's timing, believing they had an opportunity to nudge interoperability near term by agreeing on how to code at least conceptual elements. He said trying to get to the shared knowledge representation in five years was ambitious, and he pointed out that they didn't have to be intimately tied together. He suggested thinking of spoke radiating out from the center of Dr. Campbell's three-layered model, the core convergent reference terminology, which conceptually segregated the data. Problems, symptoms and diagnoses might be one of those wedges represented in existing established reference terminologies (e.g., nursing, ICPC, ICD-9). Dr. Overhage agreed that, to the extent possible, they wanted to separate clinical and administrative uses and not be unnecessarily bound by administrative uses, though sometimes these might be prudent. He said he wasn't convinced that the most important thing, near term, was to build the knowledge base that tied everything together. He'd be happy when that came, but he didn't believe progress had to be held up to include it.
Dr. Campbell concurred. In 1992, they had semantic networks and 30,000-40,000 concept nodes. Semantic networks within this model now had 400,000 nodes. Content had substantially changed.
Dr. McDonald commented on two distinctions he heard in what Dr. Campbell said. He'd talked about how with real knowledge independent of the patient data one concluded connections. Dr. McDonald applauded that, noting they could do it. The other distinction was in the semantic node model: by making connections, one could create anything without a database. Dr. McDonald cautioned that the codes in specific fields had to be interoperable or remappable in order for a message to be interoperable. He said they had to pay attention to overlap in the models, because there was a model within the semantic and they didn't necessarily fit unless developed tightly.
Dr. McDonald used the goal in HL7, which was modeled as a test result or question, as an example. He cautioned that, if they weren't careful, they could end up with a great vocabulary system and database model akin to the message model.
Contending that Dr. McDonald argued at the fringes of the compositional network, Dr. Campbell reiterated that the network already had expanded. Dr. McDonald said they couldn't assume that with either theory, but had to get things rubbing together. He contended that was why Dr. Overhage's idea about concentrating on a particular reference model and how this network worked within that context got there faster.
Dr. Elkin observed that most the discussions were galvanized by a concentration on existing artifacts (e.g., LOINC, SNOMED, ICD) and positioning them to reach a conclusion. But he asserted they needed to think about the abstract pieces they had to put together to make this work and see if existing artifacts helped solve the problem. With an atomic reference terminology, they'd think about core concepts that couldn't be divided further within the base terminology that represented core concepts within health. Pre-coordinations within that terminology were above that. A unique identifier meant a particular concept, but could be made up of two or more other concepts within that terminology. Medical knowledge hung on some of these facts: e.g., colon cancer, which could be a malignant neoplasm of the colon, came together and meant exactly the same thing. However, things needed to be said about colon cancer (e.g., a diagnosis by finding a polyp on endoscopy) that couldn't be said separately about other things. The LSCT trial indicated they couldn't clearly pre-coordinate a terminology large enough to represent all they might need to say. Post-coordination had the same format and formalism as pre-coordination, but didn't exist within the terminology and had to be made up with multiple codes.
Dr. Elkin said this system of interaction had perfect places to hang interface or colloquial terminologies they'd discussed (e.g., the one Dr. Campbell used in IDX) that might not be totally mappable to a reference terminology. Dr. Elkin emphasized that Dr. Campbell's terminology was mappable. Those codes functioned as interface terminology and could be within the framework without changing anything about the model of the system it worked under. As one built semantic interoperability with this model, and looked at the information model in the world and coded it by the exact same terminology, one found a seamless transition between the world one lived in (the information model) and what it needed to represent (the terminological model). Dr. Elkin said this formed a continuum and basis for going forward. He suggested that every one of today's products had a piece of what they needed--and in other ways was flawed. Understanding that provided the best opportunity to move forward.
Dr. Carol Bickford, speaking for the American Nurses Association, remarked on the need for a code set for the public health environment (particularly in light of the bioterrorism initiatives) and discussions about supporting pieces. She pointed out that a concept of wellness and health was missing. Noting the focus on pathology, she encouraged the Subcommittee to be more expansive and not necessarily describe today's world, but push the envelope so whatever structure was put in place supported new knowledge and relationships.
Dr. Spackman mentioned that the Context Working Group within SNOMED recognized interactions between what could be defined with the core reference terminology model and other architectures that could make assertions about a patient in a record: e.g., HL7 RIM, CEN EV13606 standard, or Open GAR. SNOMED terminology could indicate knee pain, but couldn't convey knee pain present in this patient, at this time, in this record. That had been considered the responsibility of the record or message architecture. The group recognized a responsibility on the part of the terminology standard and these messaging standards to identify overlapping areas, have a formal structure both could translate representations to, and coordinate with all these models.
Discussion on Terminology
Dr. Cohn noted the Subcommittee had one day of the October hearings and a day in December for more input. They'd schedule more time in 2003. The intent was to have a letter for the Secretary by the June meeting. Members noted it took at least two meetings to develop a letter and there would be two-three sets of hearings before June; they had until March or April for a draft. Whether they had a draft by then depended on the complexity of the information gathering. The Subcommittee will check schedules for the first-half of 2003.
Mr. Blair noted they'd heard testimony on the scope, including suggestions for a different perspective, and feedback on criteria for selection, priorities, and whether things should be ANSI-accredited. He recalled that the testimony didn't diverge much from the criteria for selection, though it might be divided into criteria for different categories and Mr. Beebe considered it an editorial of desirables. Mr. Blair will shape the hard-copy testimony into a matrix and dig deeper into Dr. Overhage's and Dr. Campbell's models and clarify information with both. The Subcommittee will review the input in October. They'd heard testimony on the scope and criteria for selection, and Mr. Blair noted they had to decide how and when to gather information from terminology developers so they could match it against criteria. He remarked this would be more demanding than with the message format standards; many issues were more complex and they'd need assistance from a consultant with medical terminology expertise. In October they'd identify criteria and further questions. But Mr. Blair asserted that the most important questions might not be answered without having someone with terminology expertise follow-up with personal or telephone interviews and pull that input together in a digestible way. In November-December, they'd hear from terminology developers; January-March, they'd hear from users. Hopefully, in March-April, they'd begin looking at a draft.
Acknowledging he wasn't clear yet that they had scope or focus, Dr. Cohn addressed the business case. They'd heard several themes, including lab terminology and drugs. The next hearing might provide an update on the DVA-NLM-FDA effort. Other pieces were what payers, providers and vendors considered the high-value and an IOM activity looking at patient safety data standards they'd probably hear about this fall.
Dr. McDonald suggested sending the mailing to developers of the codes as well as message developers; they already had a template and could wrestle with criteria between meetings. Mr. Blair agreed that could get much of the information and the consultant could follow-up. Noting they'd hear about price if they made this a required standard, Dr. McDonald suggested looking at the cost for a 300-bed hospital with 500 patients. Mr. Blair will compile a derivative of the testimony and a draft of the questionnaire to look at in October, e-mailing them in advance if possible.
Given the testimony, Dr. Cohn suggested they'd have more insight on potential criteria and approach than the issue of scope and high value. Dr. Yasnoff agreed that business case issues for vocabulary were more difficult; whatever their recommendations, expenditures would be much larger than for message format standards. He suggested a business consultant do a formal analysis of the business case for development and maintenance of a PMRI terminology and user cost of implementation. Dr. Yasnoff noted several business cases (e.g., adoption by health plans or providers). Much of the testimony indicated a single vocabulary would be better than if everyone developed their own, but Dr. Yasnoff said quantitation could bolster the recommendations.
Though hard to quantify its long-term benefits, Mr. Augustine said standard terminology would be increasingly important with health care inflation in double digits and baby boomers eligible for Medicare. Dr. Cohn asked Mr. Augustine to work on this model. Mr. Blair noted the business case wasn't just the cost in expenses and maintenance, but also value. Dr. Yasnoff suggested recounting how much DVA and DoD spent developing vocabularies--costs that could be avoided with a standardized terminology. Dr. Steindel noted they'd be reviewing a lot of this data assisting CHI. He predicted that much of the business case's justification would be the non-quantifiable opportunity costs of not doing this. Mr. Augustine said another way to bolster the business case, though it also wasn't quantifiable, was with examples of what could be done (e.g., physician guidelines): instead of referencing ICD-9 and CPT codes (an inexact science), one could see physician practice patterns in the aggregate across the health system, bolstering opportunities both technologically on the research side and for fine-tuning reimbursement systems.
Dr. McDonald revisited the question of Subcommittee's end game. He believed they were talking about turning text into codes, which assumed either extensive labor activity (not likely to be welcomed by a health care industry trying to reduce costs) or progress in natural language processing, and he proposed considering both in this analysis. Dr. Cohn suggested considering the issue of short-term versus long-term and dealing with Dr. McDonald's point about things ambiguous and uncertain. Dr. McDonald observed finding slots for what moved around was less of a leap. Noting people were making large investments in longer-range solutions, Dr. Cohn said a case could be made for cost-and-time effective ways to provide more decision support or data capture capability. Dr. McDonald concurred, but instead of entailing huge changes in process or spending substantial funds before anyone succeeded, he encouraged using vocabularies richly.
Mr. Blair recalled that during the decision process on message format standards, an extremely compelling point heard from vendors and end users was how they'd invested heavily in HL7. Dr. McDonald agreed the Subcommittee should hear their testimony.
Dr. Steindel addressed the scope of the letter. Dr. Price had said NHS spent over 13 years developing the U.K.'s terminology model and the system's first round of testing was still a year away. The Subcommittee was told there was only one terminology "out there"; Dr. Steindel asked where the good-use cases were. He emphasized the need to determine what to concretely propose in the letter and models, noting CHI grappled with that question and that they'd heard about low hanging fruit that might lead to recommendations.
Dr. Steindel noted Dr. Campbell had just introduced his way of looking at terminology and the informatics community hadn't had time to discuss how it fit in with everything else. He suggested the Subcommittee do as they'd intended and define criteria for a questionnaire for people developing or thinking about terminologies, and ask vendors how to meet these selection criteria. Dr. Steindel said they needed to determine the extent of overlap and ask if these groupings made sense. He said the Subcommittee might want to devote time to discussing all they'd heard. Dr. Cohn proposed that they needed to hear from vendors and potential users, ensuring everyone thought the same and shared priorities. He agreed that, based on what they'd heard, a strong business case had to be associated with everything. And he suggested that they might decide tomorrow to leverage what they learned CHI might be doing.
Dr. Yasnoff emphasized that a more formal approach to the business case was an essential underpinning to making the right recommendations and provided a rationale and substantial backing for those choices. Dr. Fitzmaurice summarized what he'd heard: overall the criteria were right, most candidate codes had usage and, depending on scope, there'd be different criteria for various aspects, everybody said start either with labs or drugs, they'd heard much about economics of terminologies, a vision of interoperability was built on terminology adding value to databases and value added to legacy systems as one mapped to research, syntheses or clinical practice guidelines, nursing codes could be used to improving health care processes, determining the value of mapping, many criteria dealt with quality of codes and terminologies, terminologies weren't independent in the long run, timeliness was an issue, and what was the value (relative to cost) of a piecemeal versus whole terminology approach?
Mr. Blair noted Dr. Campbell and others proposed, even with a stair-step approach, giving priority to the clinically specific core. His understanding was that Dr. Campbell proposed an alternate way of organizing their scope model. Dr. Fitzmaurice depicted it as having everything that helped care for a patient in the middle; beyond that were departments organized around taking care of patients. On the outside was administration and infrastructure. Mr. Blair agreed. He added that Dr. Campbell had looked at the scope the Subcommittee offered as a starting point and responded that his representation was a better way to view, organize and prioritize. Dr. Fitzmaurice said he saw a lot of relationships in Dr. Campbell's inner core. Mr. Blair concurred, but he questioned how they'd evaluate whether other terminologies mapped or related to it if they wound up their evaluation of standards and still didn't know what the core was.
Dr. Fitzmaurice said he saw many relationships in Dr. Campbell's inner core. Dr. Fitzmaurice reflected that the vision around the table of what they wanted to do with terminology and codes was constant. But he suggested the time it would take to describe and show relationships could be best spent following a piecemeal approach choosing standards, until they'd built up the core. Dr. McDonald noted they were about to pick ICD PCS, a clinical code set; operationally, this next year they'd be piecemealing. Dr. Fitzmaurice said they should feel comfortable using Dr. Campbell's model as a concept with a piecemeal approach to individual standards and not worry it fitting into Dr. Campbell's center core. Mr. Blair said he'd be uncomfortable making that decision until they pulled everything together and examined it carefully.
Dr. Steindel recalled Dr. Overhage's comments that Dr. Campbell's model could be built in pieces and looked at from a structural point of view of how select parts fit together in the various layers. He contended that Dr. Campbell's model required a new set of thoughts they weren't prepared for. It had many places to hang concepts, but he didn't know where the hooks were, yet he knew there were things he could put on them. Ms. Greenberg concurred it would be premature to embrace. She questioned an underlying assumption that components in the external layer didn't have clinical meaning or utility; She said that wasn't necessarily true for ICF. Dr. McDonald agreed that many administrative things were rich with clinical data. Ms. Greenberg said essentially, right now, there was only one reference terminology, and many could embrace consistency and some reference terminology with either mapping or models. Dr. Steindel agreed, adding that he wasn't sure there was a third layer. The distinction between what Dr. Campbell put there as administrative code sets or components in billing blurred, as Ms. Greenberg had noted, with what went in the second layer. Rephrasing Dr. Campbell's, "Does it map to a reference terminology?," Dr. Steindel asked, "Could it?"
Dr. Yasnoff depicted the model with reference terminology at the core and a layer around it formed of terminologies consisting of compound concepts and terms from the reference terminology. The third layer consisted of aggregations of simple and compound concepts into higher-level terms. He suggested the Subcommittee ask other vocabulary people to look at the model and convey their view of its utility, significance and value at the next meeting. Dr. Cohn agreed that asking others that testified to comment might be useful.
Dr. McDonald clarified that the compositional/non-compositional in Dr. Yasnoff's comment weren't accurate: the core had RxNorm as the clinical drug as well as SNOMED and LOINC and each had things compositional. Dr. McDonald cautioned that political/social decisions might not create what they wanted to move forward. He suggested some things (e.g., ICD-9-CM) were ripest and used everywhere. Expressing concern that the Committee already said in its previous letter that HL-7 RIM should be the core, Dr. McDonald urged they resolve where the first modeling occurred. Vocabulary modeling was independent; working piecemeal or stepwise was the only way to make progress with other modeling. He encouraged blessing things in level three that already were making progress.
Dr. Cohn said the Subcommittee had to contemplate the core model and whether, as Dr. McDonald queried, it was RIM or this conceptual model. They'd talked a lot about terminology versus information models, but needed a way to describe the core model that helped everyone. Mr. Blair said Dr. Campbell seemed to propose, while RIM might be the information model, considering a semantic interoperability model. He recalled that the Subcommittee indicated in its message format recommendations that all other message format standards needed to coordinate or harmonize with RIM, because they felt interoperability was driven by RIM for message format securities. He said they needed to discuss whether that meant all terminology should be derivatives of RIM.
Dr. McDonald suggested the question was whether they decided the vocabulary fit together in some abstract sense or chose a process identifying what was sent and where and ensured there were places in the vocabulary model describing what went there. Mr. Blair noted Dr. Overhage felt, pragmatically, they'd move faster if they determined the terminologies needed to be clinically specific and drove down from RIM; Dr. Campbell was saying other health care information objectives besides messaging benefited from his model. Dr. Elkin suggested there were only two options for merging the information and terminology models that allowed semantic interoperability: either they agreed to be non-overlapping and coordinated or they overlapped and, at those points, had to be identical. Dr. McDonald cautioned about assuming success with the model building on its own and vocabularists, noting they had to think in terms of object structures and what made messages interoperable. Vocabularies had to be standardized, focusing on key pivotal vocabulary fields and others like patient ID. Some could be dealt with fairly rapidly. Additional vocabulary would be needed as they got into details.
Dr. McDonald noted a database was needed to help connect the vocabularies together. Members discussed gaining input from HL7's Vocabulary SIG, Jim Cimino, and RIM. The next step was to standardize vocabulary, which could get a lot of "bang for the buck." They still had to decide who provided that standard, but scope should be decided based on the law, what was already done, and what came next.
Mr. Blair agreed but expressed concern that there was also a sense of urgency to move in parallel with semantic interoperability. Dr. McDonald clarified that in his view the messages didn't work because they didn't have standard codes and so lacked semantic interoperability. That required a lot of mapping, making it hard to communicate outside one's organization. Dr. McDonald said a perfect analogy for this knowledge tree was what the community pharmacy, First Data Bank, did with a code that hadn't been very useful until First Data let people aggregate it differently (e.g., "Was it Medicare approved?" "Was it generic?") He predicted that, if they got these base codes done, there'd be a flowering of opportunities to aggregate, classify and identify specific connections and relationships. Mr. Blair noted the need for a balance. Dr. Cohn asked if the answer was to at least do one, and as much of the other as possible. Dr. McDonald queried whether to standardize this or encourage the industry to provide it. And he noted messages needed common vocabulary notes and fields, so all machines could use them.
Dr. Cohn added that, preferably, this meant there was some agreement on how all these things fit together, so everyone's retrievals matched. Dr. McDonald concurred that, in different worlds, this external information (e.g., glucose is a carbohydrate) would be tremendously useful, because one could do logic on it. Dr. Cohn said they were in agreement.
Once they got down to the right terminology, Dr. Steindel said there was agreement. He recalled Dr. Elkin saying the information and terminology models overlapped. Part of the terminology model fit into, but weren't covered, in the information model. They needed a "grammar" on how to combine those bits so it fit. Dr. McDonald noted grammar could be inserted, but most places sought bigger code.
Dr. Steindel encouraged suspending the information-versus-terminology-model mindset, emphasizing a continuum and the reality that some things (e.g., Dr. Campbell's medical guidelines project and CDC's public health case definitions) had to be combined to realize all the expressions. Noting they'd been told that in October they should hear about an interface piece and opportunities, but the profession had talked about this a long time and the next day a coalescence of groups would look to them to get something on paper, members discussed listening to all the ideas the next day and then, drawing upon abstract principles and "a boot strap," propose a scope. Ms. Burke-Bebee read the law stating HIPAA directed NCVHS to study the issues and report no later than July 6, 2000. Members discussed the need for a consultant to compile and analyze feedback from terminology developers and provide an appropriate business analysis. They considered possibly having two people: one with expertise in medical terminologies who could render an objective, equitable analysis and another (who didn't need the expertise) to write the recommendations. Mr. Blair asked for names of people who met their criteria.
Day Two
Observing that people worked on standards development and other healthcare initiatives from a broad range of perspectives and might not be familiar with other initiatives, Mr. Blair and Dr. Steindel provided a quick glimpse at organizations that had promoted, developed and coordinated healthcare information standards. Everyone encountered different terms (electronic health records, electronic patient records, computer-based patient records) and some definitions were based on usage and experience, rather than easily referenced academic specific definitions. HIPAA used patient medical record information. Some private sector major SDOs were noted, along with initiatives involved in facilitating, promoting or coordinating standards in the private sector, as well as participating government agencies. Mr. Blair presented a time line and posed questions, including whether new standards or healthcare initiatives addressed: (1) new problems and approaches, (2) known problems with a new perspective, (3) problems with new methods or technologies, (4) known problems with greater industry support or energy. Other major questions were how these standards initiatives related to others and how they fit into the NHII that the Committee and many other organizations had proposed and worked on in similar initiatives.
Panel 4: Coordination of Ongoing National Healthcare Standards Initiatives
Dr. Hammond said the vision of the public-private collaborative, Connecting for Health, was that providers, those responsible for population health, researchers, reimbursers, auditors and, most importantly, consumers, would have ready access to timely, relevant, reliable and secure healthcare information through an interconnected, electronic health information infrastructure to drive healthcare at reduced cost. Connecting for Health consisted of practicing clinicians, hospitals and health systems, employers and third-party payers, federal government agencies, healthcare IT suppliers, academic and research institutions, national standards groups, manufacturers, consumer and patient groups.
Dr. Hammond said the purpose of Connecting for Health was to catalyze specific actions that rapidly clear the way for an interconnected, electronic health information infrastructure. He identified three basic components: (1) accelerating the rate of adoption of clinical data standards, (2) sharing best practices for secure and private transmission of medical information, and (3) actively working to understand what health care consumers will need and expect from an interconnected, electronic health information infrastructure.
Connecting for Health was organized into three working groups: a Data Standards Working Group chaired by Dr. Hammond, a Privacy and Security Working Group, and a Personal Health Working Group. The latter two were still being defined. A steering group will help players from each group access others and ensure they were on track and that the pieces fit. The Data Standards Working Group was organized into: a Consensus Committee that dealt with the standards per se; an Implementation Committee, chaired by Dr. Overhage, that looked at getting these standards into the world and in use (recognizing the prominence of legacy systems, that pathways had to be developed, and that creating standards was a continuous process and a national pathway had to be built in); and the Policy Committee that interacted with the consumers of the data--government organizations, third-party payers and others that would use this information.
Dr. Hammond explained that the Data Standards Working Group developed and applied criteria, identifying "operable" standards sets. Some standards already had reached a level of maturity and were usable. In other cases work was underway. Other groups had to consider who'd develop those standards. An adoption process looked at the levers for adoption. How and who would pay for implementation and recognition of required maintenance were other considerations. Dr. Hammond said all these efforts were surrounded by barriers that Connecting for Health sought to identify and muster mobilization to overcome.
Connecting for Health held its first meeting July 31-August 1, 2002. Criteria were developed and applied to a standards set, identifying standards that had reached the level of maturity usable in today's systems. Overlapping data needs of key "data consumers" (e.g., practicing clinicians and providers, patients, payers, regulators, and public health--priority data classes") were identified. Dr. Hammond said the most exciting thing about Connecting for Health was that people were talking to each other from what had been independent, vertical silos of activity, engaging together as they defined the process of producing, collecting and using that data for healthcare in a way that was usable for multiple purposes. Recognizing that this was a huge, difficult task requiring priorities, they crosswalked these priority data needs to the operable standards so they could move immediately. Work was also initiated on strategies to move standards into play to deliver required data to all stakeholders.
Dr. Hammond outlined next steps for the Data Standards Working Group: (1) vet results of the kick-off meeting with other Connecting for Health working groups, the steering group, and other key stakeholders, (2) identify next steps to move the process, (3) accelerate adoption of required data standards whose development was underway, (4) articulate the need for required standards not yet underway, (5) articulate recommendations related to terminology, (6) develop actionable, practical strategies to clear barriers and accelerate adoption of a priority "operable set" of standards, and (7) articulate the road to interoperability. Dr. Hammond noted this involved identifying whose responsibility it was to do the next step, recognizing several groups of players: the standards producers, vendor community, and healthcare producers as well as healthcare consumers. The goal was to articulate the road to interoperability.
Connecting for Health's interactions with PMRI include: (1) providing a forum where data users, systems developers, and clinical care providers become informed, engage in a dialogue about the art of the possible, doable and desirable, and develop a consensus on standards and strategies to move adoption to a broad range of stakeholders and (2) facilitating broad scale adoption and rapid implementation of six data standards: terminology, business rules, clinical templates, clinical documents (beyond the header level where there was more detail specification, dealing marginally with templates contained in the body of the document), guidelines /decision support, and interconnected health information infrastructure/EHR. Noting that throughout his career he'd believed that, if they had a terminology everybody used, they could do wondrous things, Dr. Hammond said they were approaching a point where the necessity for a single terminology might drive a solution to many problems. Observing that, even with standards, people did what they wanted, he noted the need for business rules that defined, clearly, the data collected, how it was collected, and how this was put into the standard. Remarking that text streams were beautiful for flexibility, but terrible because one could do whatever one wanted, he said Connecting for Health aimed to address those gaps by identifying the criteria.
Dr. Hammond said clinical documents were at three levels: the first level being simply the document. Level 2 held more detail specification dealing marginally with templates contained in the body of the document. Level 3 was a fully structured document in which each component was retrievable. Dr. Hammond mentioned three standards he considered mature enough to use: RIM, the data interchange model, and level 1 of clinical documents. He pointed out that none of these standards were isolated; all were interlocked (e.g., terminology standards had to be carefully linked to a data model). In discussing standards requiring accelerated attention, he emphasized how standards fit together and that it was understanding how they merged that made it possible to produce a standard more powerful than any viewed independently.
Dr. Hammond said plans for addressing gaps included: (1) establishing criteria for assessing options in each gap area, (2) developing recommendations where options exist, (3) recommending and encouraging the undertaking of strategies for developing standards where none exist, and (4) identifying resources to accelerate the development of standards. Noting that most the standards they were considering came out of volunteer efforts and "catch-when-catch-can" time, Dr. Hammond said they needed to resolve that if they were going to move ahead.
Dr. Hammond said the relationship to other standards initiatives was also an issue. Standards groups and experts were represented across all of Connecting for Health's groups and existing standards were the basis for their recommendations. They actively collaborated through overlapping involvement and communication with other initiatives. Many people had discovered a problem needed to be dealt with and most were moving in the same direction. Dr. Hammond suggested that the Committee had a responsibility to help those groups work together and not be redundant in their activities. He said Connecting for Health's and the Committee's major contribution was their willingness to make decisions and endorse what they believed were the solutions to these problems.
Panel 4: Coordination of Ongoing National Healthcare Standards Initiatives
Dr. Kolodner provided information about CHI, a government-wide initiative focused on health information interoperability standards in the federal healthcare arena and originated in GCPR, to which the Subcommittee served as an advisor. CHI's vision was for federal government healthcare agencies to achieve the health information interoperability needed through IT truly leveraged with information sharing policies, standards and practices established as the basis for electronic data transfer and all activities and projects among federal agencies and departments. Some 20 partnering agencies including agencies within HHS as well as DVA, DoD, the Department of State and Social Security participated. About 100 medical, technical, policy and management experts were involved.
Dr. Kolodner noted CHI had four work elements: (1) adopt health information interoperability standards (a vocabulary that included specific health data models and communication standards), (2) assure alignment with HIPAA administrative transaction records and data sets, (3) assure alignment with HIPAA security and privacy solutions (while these solutions weren't in its scope, CHI would pick up and use them, and (4) foster change in the medical sharing culture so information could be moved appropriately and handled in more aggregate form and ways that protected privacy.
GCPR focused on moving information between three agencies: DVA, DoD and IHS and work elements that were the focus of CHI. CHI's goal was to establish federal health information interoperability standards as the basis for electronic health data transfer in every activity and project inside all health-related federal agencies. Beyond these federal agencies, the effort was voluntary. CHI looked to work with the SDOs and industry to identify mature standards and get sign-off, hopefully soon, by all the agencies. He noted that Dr. Hammond remarked that some standards were almost ready while others needed help getting started. Dr. Kolodner said CHI was particularly interested in starting with those that related to information already available that could be used immediately and following up as health information systems in the provider arena for DVA, DoD, HHS and others matured and could use them. CHI was also interested in information reported to FDA, Social Security or CMS.
Dr. Kolodner said standards identified for future acquisitions in government would be required for future acquisitions for all agencies. He said the hope was that the federal government's adoption would create a tipping point and the private sector, judging those standards would be around a while and likely to be even more widely accepted, would voluntarily adopt them, He said at some point CHI would ask the Subcommittee to consider whether this also became a practice of HIPAA. He emphasized that what they really wanted to do outside the federal government was to look to a positive, attractive model, rather than anything mandatory.
Noting that the Committee already indicated it would be advisory to CHI, Dr. Kolodner mentioned other ways members could assist the initiative: (1) providing periodic guidance and feedback regarding CHI's directions and plans, (2) helping CHI focus on areas needing new standards, (3) helping "market" the need for adopted standards to the public and private sector, and (4) partnering with and/or supporting CHI and participating agencies.
Dr. Kolodner discussed a complementary activity to CHI: HealthePeople, noting that this strategy that DVA had proposed was about health and individuals, not just medical and agencies. He said the initiative received a boost in October of 2001 in a meeting sponsored by the Institute of Medicine and Kaiser Permanente that identified the need for a robust set of communication standards and a model health information system that was demonstratably doable in order to realize IOM's goal of being paperless by 2010. Dr. Kolodner emphasized that the strategy was federal and national and aimed at achieving a variety of high-performance health information standards (including architectural, security and technical) that supported and had an impact on the health systems used by providers across all places of care, including at least one with public ownership for entities serving the poor. Collaboratively developed by both public and private sectors, it was a broad spectrum involving consumers and providers, payers and regulators and met all those entities' and sectors' needs.
While HealthePeople federal, which encompassed federal agencies that were healthcare providers, and the broader, next-generation, public-private Healthy People concept weren't the same systems, Dr. Kolodner noted their designs, architecture and capabilities were remarkably similar. By converging over time, the same standards could be incorporated and used for interoperability. Within and, in some cases, beyond the federal sector, activities would merge, leading to convergence toward national standards in high-performance information systems. In HealthePeople federal, DVA and DoD committed to activities that, ultimately, will result in their repositories of data for beneficiaries being based on the same standards and data models and the ability to exchange information as appropriate. In addition to complementarity at the data or software level, they aimed to identify all areas where converging efforts and developing common software was possible, reasonable and met their needs.
Panel 4: Coordination of Ongoing National Healthcare Standards Initiatives
Ms. Love noted that health data systems were the physical foundation of NHII and that, in many cases, data wasn't a line item of high priority in state budgets. Rather than work to expand and improve state healthcare data, often NAHDO had to work to assure the survival of state health care data collection. Over 40 in-patient hospital discharge data systems were maintained by state health data agencies. Thirty-nine states had legislative mandates; about a dozen state systems collected voluntarily. A state health data agency or private sector organization might implement a discharge data system and statewide data collection. The patient's healthcare encounter triggered the event and the abstract from hospital financial systems included patient demographic, diagnostic, procedure and reimbursement information reported to the state agencies.
Many states based their reporting systems on UHDDS, finding these data a source of low-cost, population-based health care utilization data that provide large numbers of observations. These data were used increasingly as sources for policy and market evaluation, quality improvement, research, and public health purposes. Ms. Love noted limitations: (1) lack of clinical data elements from the medical record, (2) diagnostic and procedure coding inconsistencies, and (3) 50 state formats of a standard. But she added that, with the help of many in the room over the last 20 years, the science was catching up and they better understood and could overcome limitations. Discharge data support national initiatives such as the Healthcare Cost and Utilization Project (HCUP) led by AHRQ. Twenty-nine states currently participate in that initiative.
Ms. Love explained that NAHDO was a non-profit, membership and educational organization created in the mid-1980s to help states improve the uniformity and comparability of that data and make it publicly available and timely for more cost, quality and access purposes. NAHDO brought together federal, state and private sector interests to strengthen and expand health care data reporting interstate. NAHDO provided technical assistance to states and advocacy to assure the survival and improvement of their healthcare systems.
Ms. Love emphasized that NAHDO members weren't solely focused on data, but were all in the information business. She noted there were no federal funding sources for healthcare utilization data and states relied on general appropriations, which could affect how that data system was implemented. One of NAHDO's priorities was to expand emergency department data reporting. Ms. Love said that for marginal costs, states with in-patient data had the capacity to increase non-in-patient data collection without imposing undue burdens on the providers. Noting that HIPAA brought people together, Ms. Love said NAHDO was part of e-health initiatives, Markle Foundation, and other initiatives they'd discussed.
She reported that NAHDO worked actively to fill data gaps and promote uniformity. States had to negotiate with providers and priorities of provider communities differed: the states had 50 versions of the UB-92 format adapted to their individual needs. NAHDO worked with states to document if a claim didn't serve those needs, the data priorities, data elements collected that they could live without, and how best to bring everyone into the structure of a national standard. NAHDO identified data elements that were high priority for the states and worked with the Public Health Data Standards Consortium, the New York State Department of Health and other states to integrate a national standard for public health and research reporting that resembled a claim, but more truly reflected what states collected.
Noting another gap, Ms. Love said states worked hard to supplement weaknesses of the UB-92. Some states manually abstracted from the medical record into the discharge data abstract. NAHDO worked with the HL-7 claims attachment to bridge EDI and X12-N with HL-7 and the medical record. Noting that the claims attachment was mandated by HIPAA, she suggested it could provide an opportunity near term to understand how clinical data elements could be automated, captured and sent to a government agency. A NAHDO representative participated in the HL-7 claims attachment work group.
NAHDO was actively involved with the Public Health Data Standards Consortium in identifying priority data elements and data gaps for inclusion in the legally defined standards process. NAHDO worked with organizations across the public health and research community to determine how the Health Care Service: Data Reporting Guide would identify data gaps and build the business case for public reporting of emerging data fields and elements.
Ms. Love emphasized that the critical data gap at the state level was outpatient data. Most healthcare was delivered in non-patient settings, but physician data systems weren't capable of reporting all the data needed. Ms. Love said NAHDO continued to help states expand their data-reporting continuum to include ambulatory surgery and ED data.
Ms. Love cautioned that taking data that 20 years ago was considered proprietary and turning it into a public good wasn't an easy process. She discussed lessons NAHDO had learned with administrative data that applied to mining and extracting data from the medical record, noting that the magnitude they faced would be greater.
Privacy and confidentiality were a major issue in the states. State laws varied and NAHDO members creatively addressed those concerns at their state level.
There was no perfect data, but, with concerted effort, data improved over time. Members worked closely with others to gain a sense of "how good was good enough to move forward."
Comparability across provider systems and states had to be addressed.
Noting she was impressed, as always, with Dr. Hammond and the work many in the room had done, Ms. Love expressed confidence that the technical solutions were there and that this was an opportunity to bring everything together. She emphasized that it all boiled down to leadership, building trust, and money. Leadership was needed to prioritize data development activities to achieve early success in PMRI implementation and encourage standardized terminology and data structure priorities in order to develop a key set of clinical data elements easily abstracted from any electronic medical record system. Ms. Love encouraged the Committee to continue defining minimum core data needs for external reporting. She offered her cadre of state health data agency experts to help NCVHS work groups and subcommittees assure that public health had a voice in the national standards process and that state data agencies, which face unique and technical-political-social challenges, were brought into the conversation. She advised it was critical that the Committee frame a message at the state level about the societal good of clinical information to the public. She recommended linking the Public Health Data Standards Consortium's recommendations with the Committee. Ms. Love validated the work of a group of unfunded volunteers that built national business cases for X12-N priority data development. Reiterating that NAHDO was committed to helping states expand their current data collection efforts, she pleaded to make administrative outpatient data development and standardization a priority across HHS.
Panel 4: Coordination of Ongoing National Healthcare Standards Initiatives
Mr. Reese noted NAHIT was still in its infancy. Its bylaws were still in draft form; the Board hadn't been appointed, and no subcommittees were formed. The founders (all members of AHA) were providers, payers, IT technical services companies, and vendors who believed it was time for the field to engage in the standards discussion. Mr. Reese said he'd been a CIO for 30 years, 21 of them with Sentara Healthcare, a $1.6- billion-per-year company that included 6 hospitals, 12 nursing homes, and 175 employed physicians. Also a payer and a durable medical equipment company, Sentara had 14,000 employees, home health and air ambulance services. Mr. Reese said he came from the field and this was his first exposure to any discussions about standards and it was enlightening. He said he was starting to understand the difficulty and complexity of healthcare transactions, adding that he thought the reluctance of healthcare leadership in the field to fix the standards' problem was changing. Healthcare challenges drove the provider/payer leadership to go after the standards issues. CEOs and COOs of provider organizations realized patient safety couldn't be solved without the integration of complex computer systems and that employee recruiting and retention (a lack of clinical job skills had led to an epidemic) required computerization and workflow management in order to improve the daily work life of clinicians, physicians, nurses and other key personnel. Mr. Reese said another factor was capacity management: there weren't enough physical plants on-line to treat all the boomers. More efficiency and precision were needed to gain throughput with hospitals and better integration between them, rehab facilities, long-term and home care. Mr. Reese identified issues that drove healthcare leadership in the field to pay attention to standards issues and provide AHA with support to spawn the national alliance.
Mr. Reese described Sentara's experience that led it to become a founding member of NAHIT. Sentara had six hospitals, but only one set of non-negotiable screens at the hospital information system level done collaboratively by the medical staff. Nurses floated between organizations where the processes represented by the computer system were exactly the same, enabling Sentara to better utilize labor distribution. Sentara also standardized the charge of Description Master, providing standardized nomenclature within the health system. The data could be pulled from the actual clinical transactions that deduce disease management. Once physicians agreed on an appropriate protocol of care and development guidelines, it was entered in the computer system and ordering pathway and was immediately available to all hospitals and caregivers.
Mr. Reese noted that Sentara presented a challenge in that, although it considered HL-7 and HIPAA good starting points, it didn't see them as standards, but as guidelines. He reported there was too much flexibility in interpretation within those standards, but acknowledged it was a great step and that the Committee was moving forward. Mr. Reese observed that all the work being done by the various committees and everything listed that day wasn't on the CIOs's radar screens. CIOs discussed doing business between computer systems, but never discussed the standards because they didn't know them. He suggested the Subcommittee "follow the money." Once a standard was developed, it was the CIO who purchased the multi-million dollar system and told the provider which standard to deliver. Once CIOs understood what to buy, they'd enforce that standard at the local level. And the marketplace would respond to the market.
Mr. Reese said about 400-computer systems drove Sentara's health system. He said having the standards (and CIOs who understood them) would result in a major cost reduction and increase efficiency at the operational level. That was why Sentara and NAHIT's other founders were impassioned about participating and learning to partner with the community.
Noting NAHIT was interested in standards related to bar coding, patient identification, communication and application systems, Mr. Reese suggested that bar coding the name tag of the person administering drugs, the patient's ID, and the med could enable someone who wasn't an R.N. to administer an oral solid (e.g., Tylenol) if the computer system acted as a safety net--freeing the nurse to do what she was licensed to do and easing the recruiting/retention issue.
Mr. Reese noted the primary business of a whole set of companies was to step in the void where there were no standards (e.g., offering Enterprise-wide Master Patient Index) providing patient identification. Similarly, Sentara was spending $600,000 on interoperability issues to use an interface engine from a third party whose primary piece of technology served as a broker between computer systems. Mr. Reese noted that standards could eradicate these issues.
NAHIT's mission was to mobilize the field and assist in the implementation of standards. Mr. Reese said NAHIT's direction was straightforward: convene the payers, providers, IT manufacturers and supply chain and then take out small projects that could be successful and see if they could get the field to recognize it had a role and responsibility in the adoption of standards.
He said NAHIT didn't consider themselves legislative, but an augmentation to other processes, "although they probably didn't understand all of them." He said NAHIT was committed to a consensus approach and that strength lay in bringing the field together in a particular environment where they could educate folks like him about the standards and what had to be done. He reiterated that healthcare leadership had to resolve issues of patient safety, retention and recruiting and capacity of management as they moved forward. NAHIT would work with the Committee and everyone designated.
Panel 4: Coordination of Ongoing National Healthcare Standards Initiatives
Dr. Overhage discussed the eHI initiative and its foundation for eHI's efforts to bring together public and private sector stakeholders, promote and coordinate the national adoption of health data standards and system interoperability, and meet clinical and public health needs. EHI's membership represented healthcare providers, delivery systems, healthcare information technology (IT) suppliers, payers, non-profit organizations and professional societies, pharmaceutical and medical device manufacturers, physician organizations, research and academic institutions. Its primary mission was to drive improvement in the quality, safety and cost- effectiveness of healthcare through IT. Its strategic priorities were to: increase awareness of the role of IT in driving greater quality, safety and cost effectiveness, lay the foundation for an "interconnected, electronic, health information infrastructure" by promoting the adoption of clinical data standards and enhanced connectivity, and build the case for public and private sector economic incentives for better quality healthcare enabled by IT.
As an example of initiatives eHI undertook to accomplish its aims, Dr. Overhage described the Public Private Collaboration for Public Health. The collaboration's purpose was to: develop and communicate strategies to efficiently and securely capture and transmit standardized electronic data of public health importance, facilitate that transmission by building upon and integrating recommendations within CDC's National Electronic Disease Surveillance System (NEDSS), leverage existing provider healthcare information systems and data sources to enhance public health surveillance processes and achieve results applicable in six to 12 months.
Dr. Overhage said eHI was undertaking projects that yielded results that were usable, visible and instructive in short periods of time, then repositioning and taking the next step. In doing this, they hoped to lay the foundation for further development of NHII.
Dr. Overhage identified critical success factors: the public and private sectors worked closely together, leveraging complementary expertise and competencies; a focus on supporting specific, well-defined, highly visible needs for data in public health; a focus on data that, for the most part, was already in electronic formats and more likely to be converted and transmitted quickly; aligning efforts and data needs with existing funding sources; and leveraging the work of experts and "those in the trenches" already performing this work and building upon lessons learned and best practices; involving key stakeholders including those that produce and use data and data enablers in the middle transferring the data; driving usage of standards and strategies through implementation and demonstration projects; and aligning efforts with national incentives (e.g., NCVHS).
The broad set of collaborators include federal entities, state and local public health agencies and partner organizations, healthcare technology suppliers, healthcare providers and provider groups, holders of large sources of healthcare data, academic and research institutions, standards organizations, and non-profit associations. The Public-Private Sector Collaboration included members and non-members of eHI, but volunteers within seven working groups focused on architecture, chief complaint data, orders, security, lab and microbiology results performed considerable work. The Joseph H. Cantor Foundation, CDC and the eHI initiative provided financial support.
Dr. Overhage reported that, since its inception in November of 2001, eHI: (1) proposed additions to CDC's Implementation Guide for Transmission of Laboratory Based Reporting Information, including standard formats and specification of data types that hadn't been addressed; (2) looked at alternative methods and responsibility models for integration brokering functions; (3) considered a conceptual framework for specifying components of microbiology results reporting; (4) summarized the scope, methods and findings of sites currently performing syndromic surveillance; and (5) identified focus areas for demonstration projects. Dr. Overhage observed that the industry/private sector collaboration was incredibly valuable: organizations freely shared data and results, enabling a highly leveraged activity.
Dr. Overhage said that over the next 30-60 days eHI would release implementation guides for lab results, microbiology results, orders and chief complaint data. Other next steps include promoting widespread adoption of the standards and strategies through communications, education, outreach targeted to providers, labs and public health agencies. Noting the huge investment, last year, by state and local health departments to address these problems, Dr. Overhage emphasized that it was critically important these weren't dispersed into diverse projects that didn't supplement and build on each other, but were focused into leveraging standards and health information in a consistent way and that success stories were showcased to build and sustain momentum.
eHI supported use of the PMRI standards and terminology as the standards for implementation and its use was consistent with the Committee's recommendations: e.g., HL-7.2x as the format for orders, chief complaint data, laboratory and microbiology results.
Dr. Overhage noted key areas that required additional development or implementation: (1) existing terminologies didn't support concepts important for public health surveillance (e.g., chief complaint data) and some available and well-defined terminologies had limited implementation (e.g., LOINC), (2) as Mr. Reese had noted, implementation guides were needed, (3) standardized approaches to defining and communicating rules associated with data usage (e.g., Dwyer Tables that list laboratory test result identifiers using LOINC codes associated with diseases or conditions of public health interest--a positive chlamydia DNA probe might indicate an infection with chlamydia), and (4) while full adoption of PMRI standards and terminologies might take significant time, current data could be leveraged near term through integration strategies (e.g., format converters and code translators).
He noted the initiative had learned a number of key lessons. eHI's model of rapid short-term intensive efforts focused on specific problems in public health had demonstrated that it could produce significant results in a modest time scale and that the healthcare industry was ready to move to early implementation steps. Noting they'd already heard that day about Connecting for Health and other activities underway, Dr. Overhage said eHI also demonstrated the value of bringing together multiple stakeholders in a public-private collaboration committed to working with other organizations through overlapping involvement that leveraged complementary skills, knowledge and expertise. Dr. Overhage emphasized ongoing, open communication and concrete alignment of activities. He said the Committee could assist the initiative by: (1) continuing to create a clear, consistent message that there was a set of "operable" standards and terminologies that could and should be adopted near term, (2) encouraging and supporting demonstration projects (eHI considered this particularly important in order to identify critical issues that might not otherwise be anticipated, demonstrate that the venture could work and value could be achieved, apply lessons learned to shaping future recommendations), and (3) developing a model for migration to a PMRI standards-based environment. Dr. Overhage emphasized that, as they worked with healthcare providers and information system vendors, it was particularly important that migration strategy emerged as a critical factor in how everyone thought about things.
Agenda Item: Comments about standards Implementation in the United Kingdom
Dr. Price explained that over the last 10 years the U.K.'s NHS had at least two major strategies and two revisions of the latest strategy for implementing an IT for healthcare for England. The first major IT strategy for NHS dated from 1992 and ran about six years. It was based, as all NHS strategies have been, on (1) the patient as the center of operational clinical systems and (2) a focus on using IT to support clinical care rather than epidemiology, while recognizing that information captured to support care should be secondarily derived into trying to optimize cost effectiveness in an environment that then perceived health IT as a fairly expensive luxury and (3) managing expectations about the outputs and benefits of IT. A new national strategy published in 1998, Information for Health (IFH), was introduced as an information strategy for the modern NHS to be implemented locally. Many of the standards would be developed nationally, but responsibility for implementation, performance managing, and achieving targets would be met at the local level where family practice primary care was delivered. A seven-year strategy designed to run through 2005, it was updated at the beginning of last year with an augmented document, Building the Information Core for Implementing the NHS Plan. Concern early in 2002 that many targets within the original strategy weren't adequately met led to rethinking implementation and another document, Delivering 21st Century IT in the NHS. All documents are available on NHS's website.
Dr. Price said IFH strategy (originally envisaged as a longitudinal electronic health record cradle-to-grave, with episodic electronic patient records for particular episodes of care in acute hospitals) still formed the basis of most of NHS's intentions and aimed for life-long electronic health records for the entire population. Collections would have round-the-clock on-line access to records and information about best practices to enable decision making and improve clinical care. Seamless care was to be ensured as patients moved through general practitioners, hospitals, and community services including chiropractic, podiatry, physical therapy, etc. NHS also sought to provide health planners and managers with the information they needed and fast, convenient public access to information, including accessing information resources for particular conditions, performance of individual hospitals, provider units and practitioners and one's own records. Dr. Price said he didn't believe there was any difference between these principles and those emerging from deliberations around U.S. intentions for health IT.
Sir John Patterson, Head of Research, Analysis and Information, U.K. Department of Health, reported directly to the Junior Minister and oversaw a number of bodies including two particularly concerned with information management: Information Authority (IA), a quasi-independent special health authority established by an act of Parliament had about 850 employees and an annual budget of some $160 million, excluding local implementation costs of the health IT strategy. The Information Policy Unit (IPU) commissioned items of work from IA. Both determined a budget and project plan leading to development of a robust business case. Dr. Price noted both bodies were part of a broader framework for modernizing NHService, the NHS Plan, which included e-government and a probability framework. He noted that John Patterson had expressed interest in the DVA system as an example of what a national electronic record might look like.
Dr. Price explained that NHS was a diverse organization and that, following the original game plan, some 300-400 local organizations produced implementation strategies to suit their priorities. Given the variety of legacy infrastructures and systems, this local strategy approach was meant to ensure that NHS wasn't too prescriptive centrally, so people could capitalize on what was appropriate and already in place. People developed plans to show how IFH targets would be met and integrate different components of the health delivery environment, acute hospital sector, and community. Dr. Price said that over the last few years, clinical governance, an umbrella term for performance management within the clinical environment, was seen as crucial to many of these initiatives. He noted NHS always sought assurance that local steps were taken to enhance the quality of locally held data from which centrally required information was derived.
In April, 300 new primary care trusts pulled together the 9,000 general practices in England and their 30,000 care physicians. Twenty-eight new strategic health authorities replaced eight previous regional health authorities. Family doctors received new contractual arrangements. Dr. Price said uncertainty and turf warring created an interesting and difficult environment in which to implement a national IT strategy. A general election in three-four years and the government's heavy commitment to NHS IT and many targets around clinical messaging, electronic records and booking in 2005 led NHS to review its implementation approach and consider providing local communities with, at the least, stronger guidance to ensure that local strategy development conformed to the central template.
Dr. Price said there was also a need for NHS to shift to a more centralist, aggressive approach to standardization. He noted the U.S. health IT environment seemed to engage better with industry and suppliers. NHS chose an approach where a small number of prime service providers subcontract with other suppliers to deliver, in various domains, the components of the strategy and wanted to specify the standards that prime service suppliers imposed on subcontractors. Delivering 21st Century IT set out the standards for the life-long health record service, which was the high-level macro architecture for the national strategic program. A Director General would oversee procurement, the application portfolio, some of the changed management and new working practices. Dr. Price also noted the need to integrate with e-government and the IT industry, deliver compliant open systems and political applications, phase the approach and--a first priority and the basis of Delivering 21st Century IT--specify the standards for an electronic patient record system.
Three key targets for 2005 were electronic prescribing, a patient's ability to electronically book an appointment with a specialist in his primary care physician's hospital, and the new personification of the life-long electronic health record--Integrated Core Record Service. Essential foundation services underneath this included consent, confidentiality and authentication. Dr. Price identified consent, security and confidentiality as the biggest barriers to making progress in electronic record demonstration evaluation work (e.g., identifying, centrally, who was the owner of a particular problem). Technical things needed to be put in place (e.g., re-procuring extra bandwidth in NHS's intranet). And data and data interchange standards based on open technology were needed.
Noting that IFH acknowledged that in the mid-1990s there wasn't a disciplined approach for agreeing on national clinical data standards for NHS, Dr. Price said their solution had been to create committees. The main decision making body was the Information Standards Board (ISB). Three subsidiary Boards oversaw clinical, technical, and management information standards. Industry, through a computer suppliers association, was represented on all four Boards. A fourth component dealt with knowledge management and making details of the standards available. Dr. Price said NCVHS most paralleled the main ISB.
Dr. Price said that in approving any standard the Board wanted to see a user requirements document detailing its particular purpose. The Board also needed a specification. Standards approval was an independent process undertaken outside development; ISB wasn't part of either the IPU or IA. There had to be at least one successful implementation before a standard could be considered as a draft; the Board wanted to see at least three other implementations before approving a full standard. Dr. Price emphasized that becoming a standard was an ongoing process of conformance testing and monitoring. As Dr. Hammond had said, standards changed and needed to be updated.
He commented on the range of international standards available to choose from when adopting an NHS standard (e.g., the European committee, CEN , ISO, and HL-7). He said HL-7 was generally seen as the strategic direction for standard setting within NHS. Noting there were other involvements (IEEE, Corbamed), Dr. Price said NHS considered industry's commitment to HL-7 important and was keen to continue working with definitive ISOs.
Dr. Price said a major headache for NHS was integrating standards and systems. Implementing electronic booking, messaging, and electronic records was underway. Standards were being developed in parallel. He envisioned that, in many cases, there'd be a point when new sets of standards converged with emerging systems. While acknowledging the logic of putting systems development on hold until the standards were in place, Dr. Price pointed out that there were reasons not to do so.
He emphasized that standards needed to be interoperable and noted the challenge that presented. The only way to approve a draft standard was to have it developed and tested in a live setting. Re-engineering the standards or systems around them downstream was perceived as potentially costly and might also invalidate any evaluation work done around the performance of the application using the standard or the validity of the standard.
NHS considered CT implementation as a bridge between yesterday and today. Dr. Price said the foundation had set business objectives for implementing the standards for the financial year and intended to refine this product, as necessary, making it suitable for key initiatives, particularly short-term strategic requirements. NHS wanted to manage the supply chain around this new standard, engaging with end users, political groups, managers who'd use secondary information, and engaging with suppliers (including prime service providers) NHS also wanted to integrate it with other standards, getting it through the standards approval process and on to a full standard. Being a public sector body and because standards change and the world moves on, Dr. Price said they had to establish a fairly minimalist, ongoing evaluation mechanism to check that, in using the standard, benefits continued to be delivered. He noted business issues had to be addressed around distributing the standard and terminology and producing and costing a formal, long-term implementation plan. Because the primary care community used an existing de facto standard that was the early version of the systems, migration support was another issue embodied in NHS's business objectives.
Noting the range of choices in looking at the organizational focus for implementing the standard for family doctors, Dr. Price stressed the importance of determining who was responsible for performance managing implementation and who'd "carry the cannon." Currently, NHS considered that responsibility rested with the 28 new strategic health authorities. Remarking on what to do in moving from existing systems to new ones, Dr. Price speculated that the cost of upgrading the U.S. clinical system to run with SNOMED, rather than READ Codes would only account for 20 percent of the implementation costs. Doing something with 15-years-worth of legacy data and 50,000,000 primary care records would account for the other 80 percent. As part of their planning process, NHS assessed the value of all of its primary care data and worked out the extent it should formally be migrated and archived and the implications.
Dr. Price discussed the program that oversaw rolling out SNOMED. EARTICA had two streams. An evaluation program looked at migration issues and had a multi-professional quality assurance component, ADVENT, that kicked off implementation in primary care. Two small projects dealt with allied health and social terms. Another foundation program picked up business issues and a business case around SNOMED to augment the development case: getting it through formal, government gateway processes for large scale IT projects, developing a long term implementation plan, putting in place a long-term evaluation mechanism, and a stream to use in implementing and integrating this new terminology, making sure it cross mapped to the 55 data sets currently under development in the NHS, to ICD-10, and to the procedure classification, OBCS-4 so they could derive secondary data. Dr. Price said implementing a new standard across NHS that put in place extensions for U.K.-specific drugs and administration terms would take a year and cost about $3,000,000.
Agenda Item: Comments about Healthcare Information and Management Systems Society
Mr. Roberts outlined HIMSS' five ongoing national initiatives. In its fourth year, Integrating the Healthcare Enterprise was a five-year project done in coordination with the Radiological Society of North America to improve systems interoperability for sharing patient data by promoting coordination of established communication standards (e.g., HL-7 and DICOM) to address specific integration needs.
HIMSS also worked closely with other associations including Coalition for Health Information Policy (comprised of HIMSS, American Medical Informatics Association, and American Health Information Management Association) and Electronic Health Initiative. HIMSS was also a founding member and on the Coordinating Committee of National Alliance of Health Information Technology.
In August, HIMSS merged with CPRI-Host with the goal of making the electronic medical record ubiquitous in healthcare. HIMSS will also sponsor a field hearing February 11 in Seneca, California made up of industry experts and congressional staff on whether the federal government should mandate an electronic health record. At its June Board of Directors meeting, HIMSS unanimously adopted NHII as the society's long-term public policy initiative. Its NHII task force evaluated initial projects including creating an inventory of existing technologies and practices, developing a gap analysis on what needs to be created and a HIMSS version of NHII. Mr. Roberts said HIMSS was available to assist in standardizing the terminology used throughout this area and identifying or reaffirming the relative attributes of terms.
Discussion
Dr. Kolodner explained that the HealthePeople strategy built on activities in DVA. DVA was involved with and maintained ongoing dialogues with many public entities mentioned today and strove to implement what was ready. He suggested they'd heard a web of connections across these efforts. He also noted they'd heard that a group underrepresented at the table was the individual practitioners who held most the key information about individuals.
Dr. Hammond said he thought the level of cooperation between groups was higher than ever. Many people worked together in four or five different settings. Ideally, he reflected, they'd all work together in one setting. But each organization started off with its constituency funding for its own purpose and met up with others similarly inclined. He proposed finding a neutral way to satisfy each group's needs, so the work was done only once.
Recalling they'd seen a bullet highlighting the need to inform, Dr. Hammond said most people didn't know what was happening in this world. Even those involved weren't aware of everything, which was why it seemed necessary to participate in each activity. He noted that side communications at this meeting already promoted a level of working together beyond today. The intent was there; only the process was needed. Pointing out that the U.K. solved this problem by having an "information authority," Dr. Hammond suggested a similar authority might help put everything together.
Dr. Overhage agreed that Dr. Kolodner's description of a web of communication was how things worked today. He noted that the e-health initiative work group had struggled with what to use for a drug vocabulary; once a high need was identified, a viable solution appeared. He said e-health avoided duplicating what others did, but took advantage of leverage to enhance communications. He depicted e-health as focusing on leveraging what was happening in industry that added value to what was done. Industry groups working with e-health routinely said they'd like to help, but didn't know what to do. E-health tried to leverage assets by building collaborations and actions that otherwise might not be possible.
Dr. McDonald remarked that e-health had a strong public health volunteerism slant; there was a lot of crossover for people in the governmental agencies and things Dr. Kolodner described and HL-7 served as a mixer. Noting he approved of how HIMSS described their role, Dr. McDonald said he'd always tried to merge organizations. One could only go to so many meetings, and many weren't necessary. He expressed uncertainty about the alliance, which seemed focused on bar codes and appeared to not overlap or offer crossover.
Dr. Cohn said he, too, tried to put order into this. Dr. Overhage had talked about basic elements for moving forward. Dr. Hammond had identified things to accelerate and implement. CHI offered a set of issues about making things real. Dr. Cohn said he was struck by Mr. Reese's comments as a CIO and reminded that, in some ways, they were successful. Using an analogy about electrical plugs and sockets, Dr. Cohn said he wasn't sure it was Mr. Reese's responsibility to know that he needed version 2.X of HL-7 and specific implementation and mechanisms. But he believed Mr. Reese did need assurance that, when he bought something, it did all he needed and was interoperable. Dr. Cohn said, personally, he'd demand that of the marketplace.
Mr. Reese suggested that the fault lay at the feet of innovation. Innovation occurred in clinical areas where computerization was used to help out the patient safety agenda and Mr. Reese pointed out that innovative phases of development always involved a lack of standards and trial and error. The marketplace hadn't prevailed because innovation had reined healthcare (e.g., the clinical diagnostic computing equipment: MRIs and CT scans) into the administrative support systems. Mr. Reese cautioned that, standards they might curb innovation. He said it was a matter of balance: at some point, someone had to say, "Let's make that electrical socket this way."
Dr. McDonald responded that he could get everything he wanted clinically in an HL-7 message at all five major hospitals in his city. The only problem was on the coding side. Dr. McDonald suggested that Sentara was more HL-7 standardized than Mr. Reese thought. Dr. Cohn reflected that they could have all the data in the world and, without a plan for using it, people to help and people willing to share the data, they'd only create a great edifice without value for either public health or policy.
Ms. Love remarked that for 20 years NAHDO believed the trickle-down theory of standards solved problems. Recalling what they'd heard about healthcare authority in the U.K., she noted the states had about 39 healthcare authorities or commissions that made data policy. She questioned how they'd: get everyone on board, get their message to those helping to make data policy, and inform the Department of Health about new data sets collected.
Dr. Kolodner suggested the question was how they got a messaging standard that stabilized (e.g., the plug and socket). The problem with terminology was that they were changing the electricity or forms transmitted. He predicted that, as medical knowledge advanced, this would be the nature of medicine far into the future. As the explanation of disease changed, terms wouldn't be stable and had to be updated (e.g., from peptic ulcer as hyperacidic to an infectious condition). But much of what was communicated didn't change as rapidly and could be put in place. Dr. Cohn commented that, while they talked in the HIPAA world about not changing the message standards often, code sets changed frequently. He suggested this might be a similar distinction.
Dr. Fitzmaurice complimented the testifiers on doing a remarkable job educating their constituencies and partners. Noting he'd seen a lot of interlocking directorates, he asked if the organizations planned to: participate in SDOs and work groups to get the standards their constituents could use, join in the standards coordinating efforts, fund jointly among constituents and agencies specific standards development and implementation activities, and demand standards of interoperability in purchases that nudged the market where they wanted it to be.
Ms. Adair clarified that members of the CHI council and staff would continue to participate in SDOs, which was an effective way of making their needs known at the threshold. Coordination between groups was one reason CMS came to today's hearing; they hoped to come away with a sense of how, as representatives of the federal healthcare sector, to hold forums with other groups. Ms. Adair clarified that CHI believed its function was to adopt the standards for this federal healthcare sector. Dr. Kolodner suggested that this was where the allusion to a related federal activity came in. If there was federal adoption of a standard (e.g., LOINC) then DVA, DoD, and any other health service in the provider organizations could work together at that systems application level, HealthePeople federal, to collaborate and leverage the investment each made to move it forward. Ms. Adair observed that, in talking about sharing applications, one had to consider that an application was the personification of one's business rules and see where those rules aligned. She noted standards, vocabulary and messaging weren't necessarily business rules, but how one moved.
Dr. Fitzmaurice said CHI had given good answers to questions others would ask as they measured each organization's successes. Remarking that there'd never been a time when there was so much common vision, he said they had a good base of education to move on. Ms. Adair said CHI's intent was to look at what NCVHS brought forward and not recreate any wheels, but build off what was out there.
Mr. Reese said NAHIT's view was more finite. He described NAHIT as the "traction" point in the process representing the market. Noting it had the necessary environments and, more importantly, the providers, payers, technology companies and supply chain "all in the room," Mr. Reese said NAHIT sought to participate in standard development and testing. Dr. Fitzmaurice concurred that producers, purchasers and vendors would drive the standards. SDOs wanted demand for what they produced. Speaking personally, Mr. Reese observed that sometimes "adult supervision" was also required. He predicted that as purchasers recognized various organizations had developed standards that were generally acceptable, they'd push providers to meet them.
Dr. Hammond said Connecting for Health's intent was to cover that spectrum and identify, influence and endorse in specific terms what they meant to accomplish, recognizing that, when a link was missing, one always couldn't do what one wanted. The other part of the problem, Dr. Hammond said, was that most people misunderstood how to create a standard. There were multiple ways. Microsoft provided as many standards as one wanted quickly, they'd seen what one got with that process. The government could create standards fast. In the open, consensus world of the consumer, a different process involved users of systems, providers of care, and vendors vested in what they had, but willing to change when connected with the user community's demands. Dr. Hammond emphasized that had to be staged; there had to be motivation. When something worked, people didn't change it just because there was a new technology. Dr. Hammond said interface engines permitted living in a bracket, instead of a line, of time. He said this was a piece of what Connecting for Health did.
Dr. Hammond said he firmly believed that standards used by the public (either by requirement or encouragement) had to be free, available and used for the marketplace. But he cautioned that nobody fully understood how much it cost to create standards. Repeatedly, people discovered a need and, if the current standard didn't support it, moved to create a new one, rather than fix the old. Dr. Hammond said they had to change that mind set.
He noted that people also misinterpreted what the standard meant. Dr. Hammond remarked that Dr. McDonald and he had used HL-7 for a number of years to transmit entire patient records, in both in- and out-patient care settings. Dr. Hammond acknowledged that he controlled both sides and knew how to interpret them, which was critical in making this work. He noted this went beyond the messaging and involved a conformance agreement among the players. Creating the perfect messaging standard didn't guarantee interoperability; other pieces had to be there. Dr. Hammond emphasized that they couldn't do only one thing and expect it to resolve everything. A suite of standards was needed to make all of this work. He said they were getting there.
Mr. Scanlon remarked that, to the extent each initiative focused on the same standards, the same people would be involved and, hopefully, have agreement on the basic suite. He noted a couple of items focused on technology, rather than standards. Mr. Reese's bar coding and others were IT applications that, presumably, reflected the hospital's priorities for technology investments.
Asked where he saw the capital coming for hospitals to make these transitions and changes, Mr. Reese said he wasn't sure about the capital for standard development, but the capital for a vendor to adopt a standard would come from the purchase agreements. He said one of the responsibilities of the alliance and of HIMSS was to ensure that they understood the standard and conveyed it to their membership. Once there was consensus about what was good for their communities, money would flow from purchases. Mr. Reese noted that when he bought software or a PAC system he collaborated with 10-15 CIOs, determining which version of what interface worked best. He spent as much time evaluating systems interoperability as he did the actual application. HIMSS, CHIMES, HI-C and other professional organizations coordinated within the alliance. CHIMES and HIMSS had other agenda items; Mr. Reese expressed hope the alliance only had one: the field's adoption of standards.
Mr. Blair reflected that his experience over the last five years on NCVHS evaluating, selecting and recommending standards to become part of HIPAA was sobering; while they'd held to their sense of urgency, it took them much longer than they'd thought. Message format standards were done; vocabulary standards only begun. While trying to pull together a direction for NHII, NCVHS was still selecting HIPAA PMRI standards. Responding to how long it took to get through the cycle with a notice of proposed rule making (NPRM) and final rules, Mr. Blair noted the Committee had recommended that HHS indicate guidelines along with incentives to accelerate standards. Mr. Blair said he sensed in the day's presentations (and the fact that there was a bill in the House of Representatives and at least one more about to be introduced in the Senate along with efforts in the Executive Branch) an urgency to move healthcare into the information age and reap the benefits of the paradigm shift. He asked if this urgency translated into a desire that the federal government be more aggressive, as England had become: not only accelerating things faster, but taking more of a leadership role and going towards mandates rather than guidelines--or if acceleration should be left to the private sector.
Dr. Hammond said his answer was different today than several years ago. The government produced standards, but on its own. To some extent, the government had a free ride, watching the private community develop standards and commenting on the slowness without engaging. Suggesting they'd reached the point of public-private collaboration and engagement, Dr. Hammond called for a partnership, not only in funding, but also in expertise and defining needs. He noted that creating standards was a multi-staged task. The first step was gaining necessary technical expertise. The next steps were broadening the community, cutting off sharp edges, polishing the process and closing gaps. With funding, this could happen rapidly. He also emphasized the importance of the community-at-large's buy-in and ownership, noting this included vendors, government and regulators, providers of care and consumers. Dr. Hammond asserted that they knew enough to accelerate this process and still retain openness: it was one thing to hand down a standard as a mandate and quite another when that standard was created through an open process. Observing that healthcare was a partnership, Dr. Hammond said the private community welcomed the public community's involvement. Once people began working together, they forgot who they represented and where they came from. They were a group solving a problem.
Dr. Overhage recalled Dr. Kolodner's use of the phrase "tipping point." He said they were on the precipice and people were poised: Mr. Reese had said vendors would implement whatever the market requested; people paying the bills said they were ready to request what was needed, but didn't know what to ask for; federal agencies said they'd love to get everything in a standardized way. The challenge was to provide confidence and a nudge.
Ms. Love remarked that the "hammer" of a mandate got people to the table but, when the bottom line linked to savings and it clearly was in everyone's best self-interest, people voluntarily and rapidly standardized around a specific transaction. Mandates were a tool, so long as they had incrementalism. Ms. Love cautioned it wasn't clear whether they were mandating standardizing certain transactions, business case information needs, or everything in the medical record. Mr. Blair explained that Congress directed that certain HIPAA standards be mandated. Executing that mandate was a powerful tool, but getting to the final rule was a lengthy process. Noting that, while one might go faster with guidelines, one had less authority, Mr. Blair asked for the testifiers sense of a balance. Ms. Love advised that, so long as there was a dialogue between users and collectors, a mandate was a viable tool. Dr. Cohn clarified that, with HIPAA, they dealt with capabilities as well as standards: mandating that entry systems throughout the country use a standard was different than anything they'd thought about.
Dr. Price observed that most the benefits came from the fact that they were adopting a standard. Noting an instance where people pressed ahead with a project using edifact technology they knew to be effectively obsolete, because a change management and benefit realization program used those standards and they imagined migrating downstream, Dr. Price cautioned about a perception that potential benefits of standardization had been lost due to lead time.
Dr. Yasnoff said the need to coordinate the standards initiatives was clear from the testimony. He asked what the government's role should be: what should it do to facilitate coordination either in general or among these organizations. Ms. Adair said she believed it was on a voluntary basis. CHI's pan was to have the federal agencies stipulate standards they would build into their target architectures so everyone knew what they were moving towards. CHI hadn't envisioned the implication to legacy systems (cost and questionable benefit) that Dr. Price indicated. Ms. Adair clarified that CHI's approach was more in line with Dr. Kolodner's tipping approach. Dr. Yasnoff explained that he was referencing Dr. Hammond's point about various initiatives all, in their own way, attempting to coordinate. He asked if, as Dr. Hammond phrased it, making sure all the energy was focused in one direction was a role for the government. Ms. Adair noted that many federal participants involved in a number of other projects sat on the CHI council. CHI hoped that, through this "cross-fertilization," messages would get out about what was going on and there'd be a feedback loop and a realization that people were working in a similar direction to reflect their business needs. While they might not make all the same choices, Ms. Adair expressed confidence they'd get to the interoperability that was their goal.
Mr. Reese said what he'd look for as an operator was that someone had an inventory of all the standards they strove to do and ensured that groups were formed to work on each, knew their status and who worked on them, and participated in breaking any bottlenecks. If that was the role of the government, so be it. Ms. Love said she'd like the government to solve the unique identifier problem.
Anticipating tremendous impact from this project, Dr. McDonald agreed that the government stating what it would use would be a tipping force and a rallying cry. He said such an approach would be especially brilliant because they couldn't mandate in America. Interface engines were a $300,000,000-$400,000,000-a-year business because everybody used standards, but not correctly. Reflecting on how hard it was to standardize, Dr. McDonald noted that, although mandated by law, 40 years later the metric system still wasn't standard in America. And it took 50 years to get color TVs in half the homes or build an interface system.
Noting that the U.K. was a small environment with a different way to manage that America could take lessons from, Dr. McDonald asked if physicians in the U.K. knew READ code numbers by heart, as they did CPT numbers. Dr. Price said physicians knew some five character alphanumeric READ Codes, but he doubted they'd remember the SNOWMED-64-bit, unsigned integer.
Mr. Scanlon said they assumed in the U.S. that the suite of HIPAA standards, including privacy and security, was more or less a platform for the next wave of standards. The mandate was kind of a double-edged sword, because everyone wasn't motivated to implement some of the standards, even though they were presumably industry standards. The HIPAA experience had been instructive, albeit it was a mandated, regulatory kind of approach for what was being envisioned here in terms of strategies. They'd thought they were adopting their own standards, but upon reflection they might not be as standard as they'd thought. They'd received a lot of push back from colleagues. Mr. Scanlon asked how one made this assessment about where they really were and what the likelihood was.
Mr. Reese said officially he thought HIPAA was a good thing, if it worked. Unfortunately, it got ground down and some of the advantages in the original legislation were being watered down as it went through the iterations. From an operator's perspective, a whole industry had sprung up around clearinghouses that worked like a big interface engine and pushed the data out to the target payer. When they'd started down the road with HIPAA, Mr. Reese said they'd counted on financial advantages. He said he'd constantly been in a battle to learn and wasn't aware of how the standards were developed. Then he'd gone into a do-nothing mode of because there was so much public uproar that he'd have wasted millions of dollars if he'd remediated the Sentara environment to reflect the original version of the HIPAA legislation. Mr. Reese applauded HIPAA, saying he'd thought it was a good idea that would make a difference and reduce costs, but now he questioned whether they'd get the benefit they originally thought.
Noting Dr. Overhage mentioned standardizing what Dr. Yasnoff interpreted as business rules (e.g., the Dwyer Tables), Dr. Yasnoff asked if he felt the HL-7 guidelines group addressed this issue sufficiently. Dr. Overhage said he hadn't seen much of the guidelines group's work, but found in the public-private collaborative business rules or business data maintenance without a home. He suggested the Committee create a base for them. One challenge was determining whether there might be a standard for representing them. Dr. Overhage said this wasn't HL-7's usual modus operandi and he doubted it could populate the standard with content. He noted someone had to determine who owned content that was of broad interest and national value. Dr. Yasnoff suggested the Committee might want to investigate what they'd heard further.
Dr. Hammond commented that there were all kinds of business rules. He recommended that the Committee encourage CMS to develop a machine-readable standard format for business rules as part of its care plan and reimbursement strategy, so programs could be built to interact with them. He remarked that what made standards so important was that the healthcare system had been pushed about as far as it could: there was a shortage of human resources and a need to replace them--the next advantage was the cost containment gained from automation. Dr. Hammond emphasized that business rules might be one of the most important things done with the standard.
Dr. Hammond pointed out that the standards community hadn't engaged the clinical specialty groups. HL-7, clinical trials, and pediatrics had some activities. But the community needed to identify and bring together these people and engage them in providing content.
Dr. Cohn said that, unlike Dr. Hammond's description of coordination, he'd heard groups trying to deal with acceleration and implementation (i.e., how to get things out of an HL-7 committee and used). He noted commonality of interest and agreement about areas needing work and decisions and said he believed government could serve as the tipping point, providing direction and guidance. And he noted the need for industry input to ensure government didn't wind up having one standard and private industry another. Dr. Cohn remarked that saying NHII was only as good as its weakest link might be close to the truth. He noted Mr. Blair extended the question of identifying standards to how to encourage the healthcare industry, already in relatively dire financial straits, to implement functionality with the standards.
Ms. Greenberg recalled that at the June meeting the full Committee asked the groups to return in September and then looked to the Subcommittee to accelerate that, which led to today's panel. While she didn't feel it necessary to repeat this in a month for the full Committee, Ms. Greenberg noted the need to report this discussion and partner with these activities throughout next year. Dr. Cohn said he'd heard about good activities and suggested a progress report at the following full Committee meeting in November. Ms. Adair said she felt she'd received an assignment to return in September and would report that their conversation brought people together interested in moving in the same direction.
Panel 5: Issues in Transitioning to New Code Sets
Mr. Daley noted the new national standards for health care transactions and code sets were adopted to simplify benefits administration and introduce greater savings and fewer hassles to consumers, health plans and health care providers. BCBSA fully supported those goals. He said the cornerstones of health care administrative systems were the diagnosis and procedure codes. While current Transactions and Code Sets Rule could be viewed as the equivalent of resurfacing a building's exterior, Mr. Daley said moving to ICD-10 diagnosis and procedure coding was comparable to replacing the building's internal framework. Wholesale change had to be approached cautiously, so potential impact could be evaluated and code changes didn't result in unintended, adverse consequences.
A report aired in March 2000 described migration to ICD-10 as the "most significant overhaul of the medical coding system since the advent of computers." A recent report from GAO raised concerns about ICD-10-PCS implementation challenges and the associated financial burden on the industry. The report referenced an example of the complexity of migrating to ICD-10: under ICD-9 a single code (code 36.91) designated coronary vessel aneurysm repair while ICD-10-PCS generated 180 corresponding codes. Diagnosis and procedure codes were used throughout the industry to classify symptoms, treatments or services; perform reimbursement functions; and conduct statistical analysis for health care training or rating. Clinical codes were the underpinning of virtually everything in the health care system and change to ICD-10 would impact payment, medical policies, and quality improvement programs while benefiting design, fraud, and abuse detection.
Given the vast differences between ICD-10-CM, ICD-10-PCS and currently used code sets, Mr. Daley said implementation of ICD-10 code sets would present tremendous challenges to the health care community, including the direct impact associated with being able to use the codes, and impact associated with transitioning between coding schemes.
The anticipated cost of the code sets approached Y2K costs for transactions. Neither the costs nor benefits were fully demonstrated. Noting significant transitional issues came amid initial HIPAA standards implementation, Mr. Daley cautioned that in-house and packaged software would be impacted. Reimbursement and contracting procedures, care management policies, and training forms might require change. The payer software, including interfaces between software built in house and vendor packages, would need carefully coordinated updates, and logic built into many applications that directly interrogated specific code values would add to the complexity of accommodating ICD-10. A fully functioning crosswalk had to be available and changes of this magnitude required extensive testing. Optical character recognition programs and voice response units would need adjustment. Provider software used for scheduling, billing, claim submission, financial performance and intensive care ER activity would all be impacted by ICD-10. Packaged software would require changes to screens, logic reports, and databases. Software vendors would need sufficient lead time to update applications, negotiate with clients and roll out new releases to support its sites.
Mr. Daley noted that modifying diagnosis and procedure coding would impact the heart of the reimbursement process based on knowing the patient's condition and how it was treated. DRGs, APCs and line pricing would require attention. AMA projected 8.4 million physicians' contracts and fee schedules would require renegotiation. Mr. Daley estimated hospitals would have to renegotiate 140,000 contracts. The impact on providers would include fee schedules, contracts, new software, learning to use new coding changes, and documentation to determine appropriate codes to use with the increased specificity. Mr. Daley said employers and members might experience an impact, and if the new ICD-10 codes impacted the amount billed for a particular visit, consumers' coinsurance payments could change. Administrative services, which only contracted ASO when the customer was self-funded, could see reduced payments shrink their bottom line.
Mr. Daley added that procedures would require review and potential modification because of new diagnosis and procedure coding. Extensive training would be required, resulting in potential for incorrect coding during this period. Even though HIPAA didn't apply to paper, provider visit sheets listing codes and paper claims and requisition forms for tests within hospitals would need adjustment. Trend analysis utilization management and other areas would have to resolve how to relate statistics coded under ICD-9 to those coded under ICD-10. Payers and providers would need systems and processes to manage both ICD-9 and -10 code structures during the transition. Even with a suitable crosswalk, maintaining dual-coding structures would be, at best, cumbersome, and ability to capture historical data or review utilization information would be difficult. Ability to perform retrospective audits and other reviews relied on comparison between historical and current claims. Payers needed accurate utilization data in order to determine rates for their customer accounts.
Mr. Daley identified concerns that needed to be resolved: (1) would information systems be required to support dual standards during transition or would a hard cut-over date be acceptable, (2) how would archived data that might be needed for legal purposes and audits be treated, (3) if HIPAA didn't apply to paper claims, could providers continue to use ICD-9 coding for an indefinite period, meeting the potential ongoing need to accommodate dual coding standards. Mr. Daley noted this could also impact certain electronic transactions (e.g., what would be the status of paper claims with ICD-9 coding if the transaction standards called for ICD-10 code). Other considerations included a potential for increased fraud during this cut-over period as everyone was still unfamiliar with the codes. Noting that, even if everything were honest, it would be difficult to determine if fraud occurred for another reason, Mr. Daley said patient treatment hopefully wouldn't be hindered, but any confusion about the coding structure could result in potential implications for care management. Also noting that costs and benefits hadn't been fully studied, Mr. Daley said an industry-wide evaluation of the potential downstream impacts had to be done.
BCBSA recommended that there should be no migration until the initial HIPAA mandates were successfully implemented and a thorough impact analysis, including cost/benefits of migrating or not migrating as well as transitional considerations were analyzed and alternatives assessed. Mr. Daley emphasized that this was a big change with high impact, risks, and costs. Implications hadn't been fully defined. Mr. Daley urged the Committee to approach with caution.
Panel 5: Issues in Transitioning to New Code Sets
Dr. Cross, testifying on behalf of Aetna and the American Association of Health Plans, noted the use of ICD-10 codes for medical diagnosis and procedure coding was debated for years. The primary reasons given for changing to the new coding system were concerns that ICD-9-CM codes might be inadequate and didn't provide sufficiently detailed IFH care research and statistical analysis. What hadn't been fully assessed was the potential cost and administrative burden of moving to a new, complicated coding structure. Dr. Cross urged the Subcommittee to thoroughly consider costs and complexities associated with the proposed migration to ICD-10.
Currently, codes contained in Volumes 1 and 2 of ICD-9-CM were used for diagnosis coding in most inpatient and outpatient settings. Volume 3 included procedure codes used in inpatient institutional settings (i.e., hospitals). There were separate procedure coding systems for other health care settings, primarily the Current Procedural Terminology, Fourth Edition (CPT-4) developed by the AMA and used in physicians' offices and other outpatient institutional settings (e.g., ambulatory surgical centers).
While CPT-4 procedure codes continued to be used in many outpatient settings, the proposed migration to ICD-10 would require use of ICD-10-PCS for procedure coding for inpatient institutional settings. Dr. Cross noted that the Committee was considering recommending use of ICD-10-CM for all diagnosis coding. However, he cautioned that a change to ICD-10 for diagnosis and inpatient procedure codes would result in a tremendous burden, especially for physicians, because of costs and complexity associated with such a major change.
Migration to ICD-10 diagnosis and procedure codes would result in a substantial increase in the number of fields used for the coding process and a significant change to a system requiring both numbers and letters. For example, current diagnosis codes were a numeric system (with supplementary letter codes) using a minimum of three and a maximum of five digits. ICD-10 diagnosis codes combined letters and numbers and used a minimum of three and a maximum of six digits. Current ICD-9-CM procedure codes were numeric with a minimum of three digits and a maximum of four digits, while ICD-10-PCS procedure codes were alphanumeric with seven required digits. As a result, the number of possible codes increased dramatically. As CMS explained during earlier hearings, ICD-9-CM, Volume 3 contained fewer than 4000 codes, but the current draft of ICD-10-PCS contained 197,769 codes. And CMS reported that number of procedure codes could be expanded with development of new medical technologies.
Changes to ICD-10 also would increase the complexity of the coding process. Migration would require substantial physician and hospital staff training, significantly adding to administrative costs. Switching to ICD-10 also would increase the amount of time physicians and administrative support staff must devote to the coding process. Currently, physicians and administrative support staff utilized a four-digit numeric code to designate procedures. With a change to ICD-10-PCS, there would be more fields in the code; all of them would have to be completed for an accurate, useful code. Physicians would have to provide additional and more specific information in the medical record on procedures used and were likely to be more "hands-on" in the coding processes. Administrative staffs would spend more time coding and might require more frequent consultation with medical staff to determine exactly which procedures were performed.
A coding change would also have significant impact on health care payers (e.g., insurers, managed care plans, employers and government programs as well as business partners--billing and claims service vendors and health care clearinghouses). Migration to ICD-10 would impact these entities' multiple systems. Existing hardware systems (e.g., computer mainframes and optical scanning systems) would have to be updated or replaced in order to process additional fields. New claims and other transactions forms would have to be developed and extensive software revisions required to incorporate the ICD-10 codes into payment and claims systems. Payment methodologies would have to be modified to account for new codes. Dr. Cross said that ICD-10-PCS, a tool for reporting procedures, wasn't a reimbursement methodology because it didn't include "global" codes necessary for payment adjudication systems. While ICD-10-PCS provided detailed and separate codes for every step of a medical procedure (e.g., heart surgery), it had no mechanism allowing the combination of codes related to reimbursement. New payment related codes had to be created and agreed upon using a combination of the individual ICD-10-PCS codes.
A recent GAO report analyzing whether a single coding system should be adopted for all inpatient and outpatient medical procedures included a discussion of the merits and disadvantages of changing to ICD-10-PCS. While indicating that ICD-10-PCS provided greater coding specificity, the report noted difficulties switching to ICD-10-PCS and stated "the design and logic of 10-PCS raised concerns about potential challenges in its implementation." In some cases, ICD-10-PCS's specificity created a significantly greater number of codes for certain sets of similar procedures that might reduce coding accuracy. Dr. Cross noted that ICD-10-PCS was also a departure from the design and logic of ICD-9-CM, Volume 3; the existing health care administration system would need to be changed significantly to accommodate the new code set, imposing additional financial costs and administrative burden on a health care industry already undertaking changes to comply with the adopted HIPAA standard code sets.
Aetna supported the conclusions outlined in the GAO report on the potential administrative and financial burdens of changing to ICD-10-PCS. Dr. Cross said it was crucial to conduct a study of the potential cost of this change before making a decision to migrate to ICD-10 for diagnosis, procedure coding or both. The financial impact on government programs such as Medicare and Medicaid and on private sector payers was unclear. The ultimate consideration must be that all costs would be passed on to consumers already experiencing a significant rise in health care costs. Dr. Cross urged NCVHS to make a more detailed analysis.
Discussion: Standards and Security Issues - Draft Recommendations for Code Sets
Asked if any organization in the US used ICD-10-PCS or had experience with it, Dr. Cross said he didn't believe ICD-10-PCS was in practical use within at least the commercial industry, and ICD-10 international was a totally different system. Pat Brooks said versions of draft ICD-10-PCS were available on the CMS home page since 1998. Another draft would be available this year. Germany might make a decision to implement ICD-10-PCS and other countries had translated it. Ms. Brooks pointed out that, although ICD-10-PCS was in the public domain, it was still in development and shouldn't yet be used. Dr. Cross commented that before anyone used it agreement was needed; otherwise it would be a one-way street. Ms. Brooks noted a conversion table between ICD-9-CM and ICD-10-PCS was on the home page.
Nellie Leon-Chisen, representing AHA, said that with the shortage of coders and more work she doubted that many hospitals double coded, but when ICD-10-PCS was made available for testing a number of hospitals volunteered to double code for a short period of time and submitted their comments to 3M and CMS. The majority said it was easy to learn and use and they liked it. Ms. Trudel said that the previous testimony appeared to fly in the face of what the Subcommittee heard about ICD-10 and ICD-9-CM over the past months: a number of people testified that ICD-9-CM was "broken" and that there was a finite amount of time before it ceased to be specific enough to meet some needs just mentioned.
Mr. Daley said for reimbursement purposes BCBSA didn't go down to the granular level of detail one might for research and other statistics and it reimbursed based on ICD-9-CM and had done okay with that system.
Dr. Cross said the challenge with ICD-9-CM was that it hadn't been utilized by practicing providers (e.g., physicians without the coding expertise or time to code at a higher level) as much as it could in the detail it already provided. With ICD-9-CM one could use a relatively generic valid code, using three digits and not necessarily have to increase the specificity by going down to the last two digits. Many ICD-9-CM submissions weren't as specific as they could be. Often, particularly with laboratories, just getting the diagnosis on the claims submission was troublesome. But Dr. Cross said Aetna had a wealth of data; if the practicing community appropriately used ICD-9-CM, it gave more than enough specificity. The challenge was to use ICD-9-CM in information that could change administration and impact health care outcomes.
Dr. Cross said at some point they'd run out of code combinations they'd like to have in ICD-9-CM. His major concern wasn't ICD-10-CM ( diagnosis), but ICD-10-PCS (procedure) coding. He advised that if they implemented ICD-10 they implement it in phases. He recommended doing ICD-10-CM for diagnosis first, rather than more complicated procedure coding linked to reimbursement logic, because then the industry and practicing physicians would get used to the methodology. Dr. Cross said the challenge was how to move from a CPT-4 environment into the ICD-10-PCS reimbursement and coding logic at odds with the goal of the statistical reporting environment.
Recalling that Dr. Cross mentioned physicians had difficulty with ICD-10-PCS, Ms. Brooks noted ICD-10-PCS wasn't developed to replace CPT for physicians' office use, but was meant to replace ICD-9-CM procedures in hospitals. Dr. Cross said that depended on the HIPAA standards picked in terms of the coding logic; they could use ICD-10-PCS and allow CPT-4 coding to take place. Mr. Daley noted another concern was that the original proposal was to move to a single procedure coding system. The reimbursement and payer community lacked knowledge of what ICD-10-PCS was all about and didn't know what the impact might be. He said further study might find it wasn't so bad; the GAO report seemed to support that concept, but they needed to study it. Mr. Daley expressed fear of a mandate and emphasized that they needed to know what they were doing if they made a recommendation.
Ms. Pickett clarified that the ICD-9-CM diagnosis codes were up to five digits and mandatory; Medicare and other payers required that a three- or four-digit code couldn't be used when a five-digit code existed. Mr. Daley added that a payer organization needed to interrogate all of them, but could give a general idea of the condition and pay based on the procedure. Asked what alternatives he was suggesting, Mr. Daley replied that the plans hadn't discussed any, but he said they needed to look at ICD-9-CM: ICD-9-CM didn't have any alphas except for V&E codes. He questioned whether anyone had pursued modifying ICD-9-CM, Vol. 3 for a new technology, and suggested a variety of things could be investigated.
Dr. Cohn remarked that ICD-10-PCS had about 180,000 codes for hospital, surgical, and procedures, but typically people were paid based on DRGs, which were heavily diagnosis with a bit of procedure attached; what already existed was collapsed into fewer categories. Dr. Cohn asked if anyone foresaw an upcoming human issue where use of anything this granular would allow CMS to do something the Subcommittee should consider. Ms. Brooks noted that last year's proposed notice of the DRG indicated changes people wanted made and Congress had asked CMS to be more responsive to new technology and do a quicker job of updating the code system and being more specific. CMS had laid out a number of new technology issues (would they keep ICD-9-CM and do the best they could, make up a new system, or move to ICD-10-PCS) and the bottom line was, without room for DRGs, they couldn't be responsive to Congress's mandate.
Ms. Brooks said if they went to ICD-10-PCS it would be in the transition year. CMS would take the ICD-10-PCS codes with the existing conversion table and put back codes that affected DRG. So, the first year, new DRGs wouldn't have a payment impact. CMS would map the new coding system over the old one. CMS wasn't planning to double the number of DRGs, but as they responded to hospitals' need to be more specific they would sometimes utilize more codes.
Responding to the issue about aneurysm repair, Ms. Brooks confirmed that ICD-9-CM only had one code for that repair, but she noted it couldn't convey a particular part of the body. ICD-10-PCS had a number of codes because they told what you were doing. Ms. Brooks said sometimes there were more codes, but still only one was selected. In CMS's experience with testing, ICD-10-PCS codes didn't lead to more codes, but to more information on the bill.
Noting the speakers mentioned impacts on fee schedules, Ms. Greenberg said she understood that ICD-9-CM, Volume 3, for which ICD-10-PCS would be a replacement, wasn't used in physician fee schedules. Ms. Brooks concurred. ICD-9-CM wasn't used by any private insurance carriers for fee schedules. Most fee schedules were for payments like lab fee, durable medical equipment, and physician fee schedules that used CPT and wouldn't be replaced. Ms. Trudel explained that, because HIPAA standards went into effect before these decisions were made, the only possible use of ICD-9-CM procedure codes was inpatient hospital. Dr. McDonald asked if insurance companies used ICD-9-CM procedure codes to decide anything but what they paid the hospitals. Ms. Greenberg said the codes affected DRG. Fee schedules applied to physician services in hospitals, but those were based on CPT; fee schedules didn't apply to inpatient hospital costs.
Discussion: Standards and Security Issues - Draft Recommendations for Code Sets
Dr. Cohn recounted that the Subcommittee held a number of hearings on this relatively complex, contentious issue and that it was hard to tease diagnosis and procedures apart. Dr. Cohn outlined options: a letter recommending putting both together, moving to an NPRM or NOI, doing nothing, or further Subcommittee deliberation (which wouldn't be for months). They'd heard conflicting testimony that the industry preferred to implement together but staggering them made conceptual sense. They'd also heard additional studies of implementation costs and benefits had to be done before an NPRM or NOI.
Mr. Daley said, in certain conditions, ICD-10's increased specificity could identify which conditions helped expedite payment, care management, fraud and abuse protection. BCBSA was concerned about the complexity and degree of impact. AAHP worried about the administrative and financial costs of dealing with additional complexity. Noting Dr. Cohn asked for a cost/benefit study, Mr. Blair said he wanted to include providers and improving patient outcomes, clinical research, and information hospital administrators needed to improve the cost structure. He noted testifiers indicated that the cost/benefit for the payer sector wasn't generous enough.
Dr. McDonald reported that the written comments (two-and-a-half for; three-and-a-half against) left him cynical. AHA's studies boiled down to 100 charts evaluated by CMS's contractors. Dr. McDonald called for an independent reviewer's detailed numerics.
Nellie Leon-Chisen said AHA believed ICD-10-PCS was sufficient for inpatient hospital reporting. Recalling Dr. McDonald's observation at the June meeting that physicians weren't clamoring for more codes but wanted guidance, Kathryn Serkes, speaking for AAPS, noted physicians wanted to address issues of audits, prosecutions, and fraud.
Mr. Beebe reported that AMA contracted with PricewaterhouseCoopers for a cost/benefit analysis of different scenarios for the inpatient setting or achieving a single procedure coding system.
Mr. Augustine pointed out that managed care and payers were good at collecting and analyzing data ahead of the curve. He remarked on their progress utilizing medical and disease management and said he'd like to see more information or put what they had into a better format.
Dr. McDonald said he wouldn't be opposed to an NOI with these caveats, but expressed concern about acting on the basis of 100 sample records. Dr. Cohn suggested a letter to the Secretary, noting the need for a cost/benefit analysis within six months. Meanwhile the Subcommittee could collect similar data.
Ms. Brooks predicted that, if the Committee did an NOI or NPRM, hospitals would do their own independent studies. An NPRM would force government to analyze the cost and calculate a projection the industry would read carefully and respond to through several iterations.
Ms. Trudel explained that a NOI provided a view of potential issues, laid out options, and, being in the Federal Register, drew wide public comment. After assessing that input, an NPRM could present options and solicit further comment. An NOI wouldn't require a full-fledged impact analysis. Ms. Greenberg cautioned that an NPRM had to follow an NOI before doing a final rule and would delay the process. Delay wasn't always a negative; an NOI for the unique identifier led to a mandate not to proceed. But they'd heard concerns about how much longer organizations could continue with codes 25 years old that couldn't be updated and reflected antiquated terminology discarded by the rest of the world. She noted Ms. Trudel had pointed out that they could accomplish everything in an NPRM that they might with an NOI and force the Department to do the impact or cost/benefit analysis.
Ms. Brooks predicted that, with ICD-10-PCS, hospitals would continue to code what they did today under DRGs, even though more codes could be used for internal purposes.
Dr. McDonald said , with the knowledge they had now, he'd go with an NPRM. Mr. Augustine reiterated a preference for an NPRM with options for an impact analysis and adopting both, one, or a future approach.
Noting an NOI could tie the General Counsel's hands, Ms. Trudel proposed recommending specific concerns and issues, including cost/benefit analysis and documented testing of actual coding using real-life charts.
Mr. Blair concurred with Dr. McDonald's concerns about broader, more comprehensive testing. Testifiers reiterated the need for the business case and Mr. Blair said he wanted to ensure that case was across the board and broader than cost/benefit for payers. He expressed concern over further delay. Consensus that implementation wouldn't be until two years after October 2003 already meant compliance and adoption couldn't happen before October 2005. With an NOI, that date might slip uncontrollably later. An NPRM would allow for the cost/benefit issues and additional testing. Dr. McDonald said making the wrong decision fast wouldn't help. Mr. Blair replied that they could draw upon negative feedback.
Dr. McDonald said he'd looked: there were 7,000 more codes available in ICD-9-CM, Volume 3, and ways to code another year. Ms. Brooks noted a write up on the issue of extending existing codes. She noted the way the tree added alphanumeric digits didn't get at the root of the structure's problem. She recalled public meetings 10 years ago with the ICD-9-CM Coordination Maintenance Committee that explored ways to expand the four digits. People had been adamantly against it.
Dr. Cohn noted Mr. Blair proposed an NPRM. With the right caveats, Dr. McDonald opted for an NOI. Dr. Cohn said key issues were differences between mechanisms (e.g.,, NPRMs and NOIs), open issues needing resolution, and uncertainty about a solution. They were talking multiple things. He asked if members wanted a letter expressing concerns and next steps or to recommend an NOI or NPRM. Mr. Augustine reiterated the possibility of an NPRM, with the option of implementing separately or both internally. Mr. Blair noted they'd discussed caveats and testing. Dr. Fitzmaurice cautioned against stretching this beyond when many members remained at the table. Though an NPRM might solve some problems, he understood Dr. McDonald's concern about ensuring the right decision and he acknowledged an NOI was appealing. But an NPRM would force the government to do an impact or benefit/cost analysis before making comments and framed issues so people could comment and trust government to follow that NPRM and turn out a good final rule. Dr. Fitzmaurice said he was sensitive to the time an NPRM took, but it worked with other HIPAA standards and everybody agreed they needed information. He concurred with Ms. Trudel who'd urged the Subcommittee not to tie the hands of the Department, but encouraged it to provide enough information to make a better recommendation at a future time.
Ms. Trudel reiterated that the Subcommittee had four alternatives: do nothing, recommend a specific code set, hold more hearings, or do additional analysis using the Administrative Procedural Act process to gain input and involvement--CMS would decide on an NOI, NPRM, open town hall meeting, or industry forum announced in the Federal Register. Dr. McDonald said he'd like to stop short of an NPRM.
Ms. Greenberg said today they'd heard, "Go with diagnosis first." The last hearing, they were told, "Go with ICD-10-PCS and ICD-10-CM together." Dr. McDonald said he didn't consider ICD-10-CM much different, but expressed concern about ICD-10-PCS. Ms. Greenberg noted some European countries had seven years experience with the diagnosis code and many had implemented 10 or clinical modifications; 10-PCS hadn't been available.
Noting his hesitancy when Dr. McDonald held a different view, Mr. Blair suggested the plus for an NOI was it didn't set the process in a bound time line; the downside was delays. With an NPRM, people weren't so vulnerable to open-ended delay and might determine if their direction was misguided, incomplete or incorrect. Dr. McDonald noted CMS's enthusiasm for moving with ICD-10-PCS as fast as possible, but emphasized that a hundred cases didn't indicate anything. Dr. McDonald said he was interested in comparisons, not continuous improvement, and wouldn't make decisions on any process change in his hospital based on that data. With 280 million Americans and a trillion dollar budget, more should be invested in the solidity of the data used in decision-making.
Dr. Zubeldia considered issues about ICD-10-CM diagnosis codes, ICD-10-PCS institutional procedures, and ICD-10-PCS for professional use (not recommended) quite different and advised separate NOIs or NPRMs. Noting issues were different for each, after everything they'd heard he said he favored the NPRM because it forced government to make an impact analysis. But he stressed there had to be separate analyses for procedure and diagnosis codes, as well as procedure codes for professional services. An additional issue was whether adoption should be synchronized. Observing that even today's testifiers weren't as concerned about implementing diagnosis codes as procedure codes, Dr. Zubeldia voted for the NPRM, with the caveat that components be addressed separately.
Reiterating that an NPRM would allow them to surface issues and correct where they were misguided, Mr. Blair endorsed it with Dr. McDonald's caveat about testing. Mr. Blair observed that Dr. Zubeldia's suggestion to separate sections in the NPRM for ICD-10-PCS, ICD-10-CM, and other pieces was similar to their first draft with the deadline of no sooner than October 2005.
Reminding everyone that the NPRM process hadn't identified problems with the NDC code that they only began to understand later, Theresa Dalmer, Blue Cross and Blue Shield Association, cautioned that identifiers and modification to the transactions and code sets were also being done and an operational date of 2005 was ambitious. She questioned that in 60 days they'd get the information needed to accurately anticipate impact. Noting comments about impact on hospitals emphasized evaluating across the industry, Ms. Dalmer noted many players didn't look at the Federal Register or NPRMs. They'd have to reach out in other ways.
Mr. Blair suggested extending the 60-day response period to 90 or 120 days. He envisioned the NPRM extending not only a time frame that could be changed by public testimony and critiques, but also establishing a mechanism that formally solicited comments and drew in the business case. Dr. Zubeldia expressed concern that an impact analysis might not reflect private industry's view. Recalling Mr. Beebe mentioned that AMA might do its own impact analysis, he asked how the process could reflect private industry impact analyses. Dr. Fitzmaurice noted perspectives submitted as comments to the NPRM had to be considered. Dr. Zubeldia pointed out that comments didn't have to be published in the final rule; he suggested before the NPRM the government could informally publish a white paper depicting its impact analysis alongside private industry's analysis, putting both in the NPRM. Ms. Greenberg noted WEDI's impact analysis was highly referenced in the NPRM on transactions and code sets. Other analyses could be referenced and points of agreement stated. Asked if the government could publish its impact analysis independent of and prior to the NPRM, Ms. Greenberg clarified that the rule-making process went through numerous parts of the government; publishing prior to the NPRM didn't make sense, because one couldn't go through that vetting process.
Mr. Augustine agreed with Dr. Zubeldia's original proposal of an NPRM, but separated out aspects. Dr. Zubeldia emphasized that the industry needed to see the government's and other impact analyses before evaluating the NPRM. He encouraged going for the NOI if they could get all the extra things done.
Dr. Zubeldia proposed that the government do a full impact analysis, compare it with private analyses, and move expediently as possible to rule making. He advocated doing an NPRM, if required to do an impact analysis, pre-publishing it before the NPRM. The staff said that wasn't possible. Mr. Blair suggested extending the public comment period. Dr. Cohn reminded everyone that they were a subcommittee of an advisory body to HHS and limited in what they could do. Members discussed whether the Committee could contract for an impact analysis. Mr. Augustine added that he'd prefer doing their own impact analysis before taking that next step, if they could do it themselves and ensure it happened in a timely manner. Dr. Fitzmaurice suggested a study with hard coders double coding and comparing 2,000-5,000 forms, acknowledging that probably was beyond NCVHS's resources.
Ms. Trudel reiterated their four options. She noted Dr. Zubeldia and Mr. Augustine voted for further study. Ms. Greenberg said the Subcommittee could commission a parallel study; it wouldn't accomplish what the NPRM process would, but they could hold hearings about other studies. But she cautioned that was no different than delaying another few years and didn't inform the field. Dr. McDonald advised that, if feasible given the time frame, a NOI was more sensible because it generated studies. Ms. Greenberg reiterated that a NOI added years to the process. Dr. Cohn recorded a vote in favor of going forward with an NPRM, with one negative.
Members discussed asking the Secretary to inform the public of their intent and urging industry to perform its own impact and cost/benefit studies in advance of release of any NPRM. Reflecting on Mr. Blair's desire for full industry appraisal, Ms. Greenberg suggested commissioning a study that "quantified benefits. Mr. Augustine cautioned that was extremely difficult.
Members approved the direction of the original document. A revision will be distributed before the next meeting. Dr. McDonald was assured that he'd have an opportunity to discuss his minority opinion to the full Committee. Dr. Cohn explained he'd abstained because this dealt with procedure codes as well as diagnosis codes, and wouldn't give an opinion to the Committee. Noting the implementation date recommended in the document seemed impossible, Dr. Cohn suggested thinking about it as they moved forward.
Discussion: DSM-IV
Dr. Cohn noted the Secretary received letters from APA requesting that DSM-IV be named a code standard for use in psychiatry and mental health and that APA be named the DSMO. Staff had prepared drafts backgrounding issues and potential recommendations. The initial view was these activities should be separate; typically, DSMOs maintained transactions rather than diagnostic criteria. The Department received no comments from the industry recommending adoption of DSM-IV during the 60-day comment period for the transactions and code sets NPRM. But they'd heard concern that, once ICD-9-CM was the standard for diagnosis reporting, mental health couldn't use the diagnostic criteria in DSM-IV. The Subcommittee's interest was that mental health professionals and psychiatrists could continue using the diagnostic criteria in DSM, as well as crosswalks between DSM-IV to ICD-9-CM. Dr. Cohn sought members' and APA's perspective. If there was agreement, the Subcommittee or staff would draft a letter to the Secretary for review at the next meeting stating continued use of DSM-IV diagnostic criteria that crosswalked to ICD-9-CM codes didn't pose a conflict with code sets adopted under HIPAA and that adoption of diagnostic criteria used to establish diagnosis was outside HIPAA's scope.
Dr. Darrel Regier, Director, Research, APA and Executive Director, the American Psychiatric Institute for Research and Education, conveyed APA's concern that the legal interpretation of a code set had to be addressed by HHS's Office of General Counsel. APA's counsel, Mr. Robert Charrow, evaluated this after APA received concerns and concluded that, according to 45 CFR 162.103's definition, without DSM-IV encoded in the HIPAA statute, the default descriptor would be the ICD-9-CM glossary and definitions. The Subcommittee will ask the Department's General Counsel if the concept of diagnostic criteria was, indeed, separate and distinct from the HIPAA code sets and whether that made a difference in what the Committee's view might be. Dr. Regier said the law explicitly acknowledging DSM-IV criteria as the descriptor for the code numbers would advance things considerably.
Turning to non-legal issues, Ms. Trudel recalled that Dr. Cohn had indicated the DSMOs were specific to the transaction standards and not clinical code set maintainers. AMA's CPT Editorial Panel and CMS weren't DSMOs and the participants didn't have clinical code set expertise and modified the transaction standards. Beyond this threshold issue, the discussion was about the clinical code sets. Dr. Regier agreed that the issue was the code set and mechanism necessary for a publicly approved code maintenance organization. Mr. Charrow clarified that there were two issues: the scope of the regulatory provision and the legal impact of a single official code set that had no diagnostic criteria. He emphasized that eliminating DSM-IV as an official code set had profound implications on 675 state laws and that, once HIPAA was fully effective, the informal crosswalk would no longer be viable.
Clarifying that the Subcommittee's main concern was that the crosswalk was appropriate and people could continue to use diagnostic criteria, Dr. Cohn said they were searching for a reasonable, speedily implementable solution and the lawyers would advise if this provided a viable option. He cautioned against hopes that this could be done with an NPRM before HIPAA was implemented, but he assured everyone that the Subcommittee wanted this to turn out right for practitioners and believed there was a solution. The next step was obtaining CMS's legal counsel's review. Hopefully, lawyer-to-lawyer discussions would resolve this by the Subcommittee's next breakout session. Dr. Regier said there's likely to be a court challenge. Dr. Cohn said Ms. Trudel will work within CMS and the Subcommittee will continue the conversation. The meeting adjourned at 4:15 p.m.
I hereby certify that, to the best of my knowledge, the foregoing summary of minutes is accurate and complete.
/s/ June 3, 2003
_________________________________________________
Chair Date