A method for estimating the population at risk in primary care practices by applying correction factors to the active patient census.

Epidemiological research based in primary care practices has been hindered by the inability to estimate the number of persons served by individual practices. This study examines one method proposed as a means of estimating practice populations. The effectiveness of this "correction factor" method depends on uniformity in the proportion of persons visiting a physician in a two-year period across various sociodemographic subpopulations in the United States. The National Health Interview Survey data were examined for evidence of such uniformity. Within broad age-sex categories there was little variation by race, educational level, income, Spanish ancestry, or location of residence in the proportion reporting a physician visit within the previous two years. Although these results are encouraging, a number of problems remain before the "correction factor" method can be credibly used to estimate practice denominators.