i i.. . i i, . I @. 1 i @@fi !u5i!! fililk@@ I# *@@)o I @ COMPUTER ASSISTED EKG ANALYSIS AND THE REGIONAL MEDICAL PROGRAMS A Report of a Subcommittee of the 'NATION7AL ADVISORY COUNCIL ON REGIONAL MEDICAL PROGRAMS August 3, 1971 ANALYSIS COIMPUTER ASSISTED EKG I. Definition II. Description III. Potential Advantages and Critique IV. Current RMPS Participation V. Current Usuage and Potential VI. Conclusions Prepared by: Kenneth S. Gimbel, M.D. Herbert P. Dunning John R. Farrell, M.D. COMPUTER ASSISTED EKG ANALYSIS 1. Definition: Electrocardiographic interpretation utilizing a digital computer programmed to recognize specific characteristics of EKG wave contour and/or rhythmn. II. Description: The digital computer is capable of analyzing a large volume of electrocardiographic variables related to wave contour magnitude, duration, direction, and frequency. The recognition of deviations from programmed normal values in these parameters permits discrimination of abnormal electrocardiograms from normal ones. Identification of patterns of deviation can then be translated into specific electrocardiographic diagnosis. The computer, therefore, may serve to select out abnormal tracings form a large population of normals, with definitive interpretation being reserved for conventional methods. Or, the computer may offer diagnostic interpretations with alternatives and exclusions. Though individual systems differ with respect to specific hardware involved, they all share some common featur es: -2- 1. An electrocardiographic preamplifier system designed to amplify the millivolt-range physiologic signals to a level which can be worked with conveniently. 2. A means of converting this signal into a mode suitable for computer input (an analog-digital conver ter). 3. A means of transmitting the converted signal to the computer, either-directly via telephone input, or by magnetic ta e recording: p on-line or delayed input.' Major differences exist with respect to the number of variables analyzed. Most systems utilize standard twelve lead electrocardiograms, while others use a special three lead (Frank orthogonal) approach. Still others record from a single lead and are principally concerned with continuous monitoring of rhythmn. Most programs approach EKG analysis utilizing concepts of binary logic. Multivariate statistical analysis has been used less frequently. Systems also differ in their modes of data transmission, computer printout form, and storage and retrieval capabilities. Potential Advantap-es and Critique: 1. Phy,,,sician Assistance - Automated electrocardiographic interpre- tation relieves the physician of a time-consuming task allowing him additional time for more critical functions. If computer-EKG programs -3- were fully automated at present, this assertion might be valid. In practice, however, most.programs are of a developmental nature and since the com uter interpretation is evaluated by a trained physician, .p no time bbnefit is accrued. Automatic measurement of intervals and durations by the computer does save time for the physician. Many of these determinations, however, are carried out in some institutions .by technicians, thus no advantage to the physician is realized. The complexity of the computer program creates new personnel problems. One obvious hindrance would be the increased demand for skilled computer programmers. In addition, the time required for acquisition of the EKG on the computer unit exceeds that of the con- ventional EKG. 2. rac - By adhering to programmed diagnostic criteria, observer variation is eliminated, uniformity of interpretation is enhanced, and accuracy is improved. In addition,.expert EKG interpre- tation would be available to areas lacking skilled interpreters. While observer variation is reduced in practice, the computer interpreted EKG has been found less accurate than EKGs read by conventional methods. Specifically, the incidence of false positive (abnormal) EI@Gs has varied from 2%-20%, and the incidence of false negative interpretations has been reported from 3%-5%. 6, 7, 9)' The inability of the computer to accurately measure certain variables, e.g., the termination of the QRS complex; failure to detect waves of diminutive amplitude, e.g., small P waves; and artifacts created by poor recording and transmitting techniques, all contribute to error. Though accuracy has improved during the past three years, considerable disagreement still is found between computer read and physician read EKGs (approximately 20%). Disagreement with respect to the nature.of the abnormality in an ab- normal EKG accounts for most of the difference, along with false posi- tive interpretations. Separation of completely normal EKGs from those with borderline or gross abnormalities is achieved with a high degree of accuracy. (9) 3. Volume Capacity If current problems in diagnostic accuracy can be overcome, then automatic (interpretation of) El@Gs will permit the increased utilization of this parameter as a diagnostic tool. Since the actual computerized interpretive operation occupies only a few minutes time, the major determinants of volume capacity are data acqui- sition time, transmission time, and printout time. Recording the standard twelve lead EKGs on a data acquisition unit takes'somewhat longer than with a conventional electrocardiograph. This is offset, however, by computer data printout and EKG plotting, freeing the tech- nician from the task of cutting and mounting the EKG. Even if the EKGs are reviewed subsequently by a physician-interpreter, he benefits from premeasured intervals provided in the computer data printout. Aside from the physician-sparing function, therefore, a major potential benefit of the computerized EKG is its volume capacity. In -5- order for this advantage to be realized, however, the computer system must be situated in the proper milieu, i.e., one providing an optimal volume of input. This optimum will vary among computer systems depend- ing upon their degree of sophistication. It has been estimated that the computer must perform 100,000 EKG readings yearly in order for maximal cost-effectiveness to be achieved. (11) 4. Cost Effectiveness - The cost of conventional EKGs is approxi- mately fifteen dollars per cardiogram. It has been estimated that a computerized EKG should cost $2.50, thus representing a considerable savings. (9) This figure, however, will be arrived at only when@two conditions are fulfilled: 1. optimal volume input is established; 2. reliability and accuracy improve so that physician supervision is no longer required (in terms of comparative interpretation). Current systems have not yet approached a level that will provide any economic advantage over conventional electrocardiographic interpre- tation. In addition to the present unfavorable operating costs, the considerable expenditure for the computer and the recording and trans- mitting equipment must be recognized. Assuming that the necessary degree of accuracy-is achievable.. identification of a suitable (in terms of minimum input volume) environ- ment is necessary. Since most hospitals require far less than 100,000 EKGs yearly, in most instances independent hospital based computer EKG systems are impractical. A single computer base subserving multiple. hospitals utilizing telephone linkages represents one approach. Similar relationships between a computer base and multiple HMOs is also fore- seeable. The potential of bringing the-expe@tise of the university hospital to outlying community hospitals and out-patient facilities is attractive from an RMPS viewp oint, both because it might improve the quality of medical care delivered in the smaller hospitals, and because it might promote interho@pital cooperation. One must question, however, whether or not a need exists for other than conventional EKG interpre- tation in the community hospital; whether the computer would fulfill that need; or whether some other alternative would be more practical. -7 - IV. Current Pu@ Participation At present, five regions are involved with computer assisted EKG analysis: Missotiri,.The Intermountain Region, Florida, Rochester and Memphis. This section will describe the variations in computer EKG application as effected in the regions mentioned, and will indicate their success or failure in fulfilling PITS goals. A. Missouri Computer-bioengineering projects accounted for a considerable part of the Regional Medical Program in Missouri during its initial three years of activity. Project #19 entitled "Automated Electrocardio- gram for Rural Areas" attempted to develop a computer system for EKG interpretation serving remote areas. Fifteen remote data acquisition units linked by ordinary telephone to a dedicated computer in Columbia, Missouri, were established to evaluate the feasibility of such a system. The remote locations included several community hospitals, a private clinic, a prison, an individual GP practice, and a medical center. This project was reviewed during a site visit in October of 1969, attended by several leading experts in the area of the automated EKG. It was noted that, (1) no need for the computer EKG service had been established and, (2) no cost savings had been demonstrated. There were also several specific defects with reference to project design and management.' The National Advisory Council acknowledged that the endeavor had potential value. Because, however, no clear advantage in terms of regionalization -8- or health care delivery was foreseeable for the immediate future, Council advised reduction in funding wit h complete RMPS withdrawal by June 1971. Continuation of the project utilizing s@pport from some other agency was recommended. Many of the technical criticisms concerning this'@project are no longer applicable as improvements in program design and management have been instituted. Cost-efficiency has improved, cardiologist back-up interpretation has been provided, and reliability is improving. In addition, a demand for this service of sufficient magnitude to make it economically feasible, has been identified. The past four years experiences have (it appears) resulted in a program of improved technical quality. Economic justification demands that 50,000 yearly EKGs be analyzed in order to make the program finan- cially self-supporting. At present, only 12,000 El@Gs are being recorded, far fewer than required. However, a survey concluded in February 1971 by Donald L. Wilson (Mo. RMP) seems to have established a need for this service. Based upon an estimate of the average number of EKGs performed in the region, the expressed interest of various physicians and hos- pitals, and the apparent lack of sufficient numbers of skilled inter- preters presently available, such a project may be justifiable. On the basis of this survey and because of their past experience with the computer EKG, project #69 has been submitted "Computer Processed Diagnostic Aids in a Rural Area." -9- The proposal aims at extending both the scope and magnitude of- the previous computer-EKG program by: 1. Placing remote stations in suitable areas to generate 50,000 EKGs early, enough to be cost effective. y 2. Increasing the range of services provided: pacemaker follow-u,p, arrhythmia analysis, exercise-testing, phonocardiography and spirography. 3. Providing for planned decremental @ support, including eventual financial independence by establishing a fee for service system on a suitable unit volume. Two major problems remain: 1. Computer accuracy, though improving, does not yet appear to be at an acceptable levei. 2. Even if the necessary volume-input is achieved, the high cost of the analog data acquisition units, in terms of purchase and maintenance, has to be reckonned with. Admittedly, these are essentially technical problems that ar e solvable; the question is when? -10- B. Florida Project #3: Regional Computerized EKG Processing Center This project was designed to provide inexpens ve, rapid and reliable EKG analysis to small hospitals in the North Central Florida area presently lacking this service. Objective data identifying a need for this service in the area mentioned is not presented. An initiation period of three years with yearly increments in the number of EKGs per- formed and the number of hospitals participating'was proposed. Financial indepe ndence was to be achieved by charging a fee for service, $3.50 per EKG (though $7.50 was indicated to be a more realistic figure). A technical site visit conducted in January 1970 was generally favorable. Continued funding was recommended for six months at which time the region would be required to provide: 1. marketing strategy 2. quality control plans 3. cost analysis, utilization and evaluation study 4. clinical EKG validation study 5. choice of arrhythmia interpretation program 6. a minimum of 100 EKGs daily. During the 02 year, 31,000 EKGs were generated from ten participating hospitals. Only one of thes6.institutions has agreed. to pay for the service, the agreed fee being $3.50 per EKG. This figure had previously been admitted to be unrealistically low by the project's director. A time schedule for initiation of fee for service EKG analysis in institutions presently receiving this service on a non- payment basis provides for $100,000 by the end of the grant period (03 year). Current interest levels of those nineteen additional hos- pitals representing potential purchasers in an area within a hundred mile radius of Gainesville need to be identified. Marketing strategy appears to be incomplete. Procedures for clinical validation of the EKGS, and quality control have been started. Project #12: Computerized EKG Screening Program A computerized EKG is one of multiple variables measured in this cardiovascular disease screening program. Patients with abnormal EKCs (or abnormalities in any of the parameters measured) are referred to a designated private physician. The computer identifies abnormal EKGs and provides a specific diagnosis. The computer program being utilized is that described in Project #3. Major expenditures are al ot- ted for purchase of data acquisition units ($42,750) and telephone service (49,000). Estimated initial costs are $102,650, and yearly cost subse- quent to start-up is estimated at $76,000. A technical site visit in January of 1970 observed that the volume of EKGs screened was small, statistical'data was lacking, and plans for patient follow-up were not clear. The projected unit cost per EKG of $3.00 appeared to be unrealistic. Project i@12 has been com- bined with Project YPIO, -"Cardiovascular Screening Program in Four Rural Counties" to form Project l@10, "A Procram for Cardiovascular Screening and EKG Analysis." -12- C. Memphis Project #8: Regional Electrocardiographic Diagnostic Center This project, now completing its third year,@was intended to: 1. Improve the efficiency of service electrocardiography in the hospitals of the University complex by utilization of a central small computer linked to data acquisition units in member hospitals by telephone transmission. 2. Provide this service on a regional basis. 3. Improve the quality of EKG analysis in the region. Because of delays in equipment delivery and installation, the program has been operational for only one year. Notably lacking in the original ,grant application were: 1. Documentation of a need for the intended services within the region. 2. Detailed plans for marketing and expansion of the service. 3. Demonstration of any potential cost-saving value. Apparently, the proposed expansion of routine computerized EKG analysis into the region met with resistance from the established medical com- munity. Plans now center around development of an arrhythmia screening center serving local and regional coronary care units on a 24-hour -13- basis. Two units are functioning, with plans to service an additional "ten or fifteen" units-in the region. Notable by their absence are: I.' Documentati on of a re ional need for this service and 9 evidence of willingness to participate by the medical community at large. 2. Marketing strategy. 3. Details concerning the specific arrhythniia program to be utilized, and provisions for clinical validation and quality control. 4. Potential cost saviIngs: The original grant application hoped to achieve a fee for service compatible with the customary electrocardiographic charges in this community." A progress report containing the specifics of the past year of opera- tion with reference to the number of EKGs performed, clinical correla- tion, accuracy and reliability of the system, problems encountered and the proposed solutions, and objective evidence demonstrating improve- ment of EKG interpretation in the region would be most'helpful. D. Intermountai ion Project l@10: Physiologic Data Monitoring System Funded since 1967,'this project provides computer monitoring and analysis of multiple physiologic variables related to the cardio- vascular and respiratory systems: 1. Au tomated data analysis for four cardiac catheterization laboratories. 2. Automated spirometry (two hospitals). 3. Automated EKG Analysis by Computer: This service is utilized at two of,the five hospitals presently involved in the computer program. To date, 14,170 EKGs have been analyzed by computer. 4. On-line patient monitoring in a variety of intensive care unit situations. A technical site visit in October 1969 commented upon the overall qualitative excellence of this project demonstrated great sophistication and technical expertise in the development of computer techniques for automating physiologic measurements of cardiovascular function." However, the member hospitals were noted to be highly sophisticated centers intimately concerned with cardiac catheterization and cardiac surgery. With respect to electrocardiography, none of -these institutions lacked skilled cardiologists in EI