OBJECTIVE: Studies using claims databases usually establish initial diagnosis for a disease by requiring subjects to be disease free for some period of time before inclusion in the study. This time period should be stringent enough to minimize misclassification of cases, but liberal enough to ensure an adequate sample size. This study was to evaluate the change in study sample size when the time period subjects were required to be disease free(disease-free-time-period) was varied. METHODS: Patients diagnosed with hypertension(ICD-9-CM 410) in the South Carolina Medicaid population in 1995(n=20,829) were followed back in time(1992-1994) to determine the change in sample size when the disease-free-time-period was varied from 6 months-3 years. All patients were required to be eligible for at least 11 months in each year of interest. RESULTS: The table below shows %decreases in sample size when disease-free-time-periods were varied. Years of Interest Pts eligible for 11 mths in each year of interest % decrease in sample size. 6 12 18 24 30 36 mths. mths. mths. mths. mths. mths. 1994 13202 51.53 61.69 NA NA NA NA 1993 & 1994 10512 52.63 63.57 70.41 73.77 NA NA 1992, 1993, & 1994 8634 52.95 63.93 70.81 74.18 76.11 77.43 The disease free sample size decreased by approximately 50% in the twelve month period following patient identification. It continued to decrease at a lower rate as the disease free time period was increased to 18 months and tapered off after that. CONCLUSIONS: Studies using disease free time periods less than 12-18 months may include 10-20% of subjects identified that are not really patients with an initial diagnosis. This may introduce potential bias if researchers think they are analyzing initially diagnosed subjects.