Kind S, Virnig B; Academy for Health Services Research and Health Policy. Meeting.
Abstr Acad Health Serv Res Health Policy Meet. 2001; 18: 23.
Health Services Research and Policy, University of Minnesota, MMC 729 Mayo, 420 Delaware Street SE, Minneapolis, MN 55455, Phone: (612) 625-5683, Fax: (612) 378-4866, E-mail: sarakind@umn.edu
RESEARCH OBJECTIVE: Explore validity of diagnosis dates derived from Medicare claims data by investigating the concordance between cancer diagnosis dates found in the Surveillance, Epidemiology, and End Results (SEER) cancer registry and date of first claim with specific cancer diagnosis in Medicare databases. A high level of agreement between these dates would point to the validity of these administrative data for estimating time of diagnosis.STUDY DESIGN: Using the SEER-Medicare linked files: Patient Entitlement and Diagnosis Summary File (PEDSF) and the associated Physician/Supplier Part B Standard Analytic Files (SAFs), Outpatient SAFs and Medicare Provider Analysis and Review (MedPAR) files for 1991-1994, initial analysis was limited to beneficiaries with a cancer diagnosis in 1992 or 1993. Claims were searched and the date of the first diagnosis with individual's specific cancer was identified. Lag time was calculated as the difference (in months) between SEER diagnosis date and date of first claim with specific cancer diagnosis. Analysis of factors influencing lag time included cancer type, stage at diagnosis, age, sex and race. General linear models (GLM) were used to simultaneously assess the influence of these factors on lag time and to assess statistical significance.POPULATION STUDIED: 82,948 Medicare beneficiaries in the SEER-Medicare linked files who were diagnosed with one of 12 common cancers (bladder, breast, colo-rectal, esophageal, kidney, liver, lung, ovarian, pancreatic, prostate, stomach, or uterine) in 1992 or 1993, had only one cancer diagnosis between 1973-1994, were 65 or older in the year of diagnosis, were not enrolled in an HMO between 1991 - 1994, were not diagnosed at autopsy or via death certificate and had matching diagnosis found in claims.PRINCIPAL FINDINGS: Overall, 70% had their first claim indicating their cancer in the same month and year as SEER diagnosis date. 10% had the first claim in the month prior to SEER diagnosis and 8% had the first claim in the month after SEER diagnosis. Overall, this resulted in 88% concordance within 1 month of SEER diagnosis date. Concordance was lowest for prostate cancer (61% matched on same month) and highest for colo-rectal cancer (80%). Lag time was significantly (p < .0001) related to cancer type, stage, age, and race. When separate analyses were run for each cancer type, age was the only consistently significant variable, and it was significant in bladder, breast, colo-rectal, lung, ovarian, prostate, stomach, and uterine cancers only. Those closer to age 65 (new enrollees) had claim dates that were an average of 2 months after SEER's. However, when the population was restricted to those age 67 or older (more likely to be Medicare enrollee at time of diagnosis), the significance disappears.CONCLUSIONS: There is a high percentage of agreement between the SEER registry data and Medicare claims with respect to diagnosis dates. This points to high validity of the administrative data for estimating diagnosis date.IMPLICATIONS FOR POLICY, DELIVERY, OR PRACTICE: Although claims are a rich source of health care information, researchers often are wary about their validity. This establishes the usefulness of claims data for research applications relying on dates such as survival analysis and epidemiologic surveillance.PRIMARY FUNDING SOURCE: National Cancer Institute
Publication Types:
Keywords:
- Animals
- Case-Control Studies
- Chronology
- Colorectal Neoplasms
- Health Maintenance Organizations
- Humans
- Male
- Medicare
- Neoplasm Staging
- Neoplasms
- SEER Program
- Time
- diagnosis
- economics
- utilization
- hsrmtgs
Other ID:
UI: 102273088
From Meeting Abstracts