What patient population does visit-based sampling in primary care settings represent?
BACKGROUND: Evaluations of outpatient interventions often rely on consecutive sampling of clinic visitors, and assume that study results generalize to the population of patients cared for. OBJECTIVE: The representativeness of such visit-based sampling compared with the population of patients seen during the same year, in terms of sociodemographic and clinical characteristics of the user groups that visit-based sampling yielded were assessed. METHODS: One thousand five hundred forty-six continuing patients visiting the primary care firms in an urban VA medical center were consecutively sampled, and visit frequencies were compared for these patients with subsets of the patient population. Administrative and survey data was then used to describe the types of patients visit-based sampling most represented compared with the types of patients sampled less frequently. RESULTS: The average sampled patient visited the firms significantly more often than patients in the reference population (18.7 vs. 9.5). Sampled patients were significantly older (>55 years), in poorer health (higher prevalence of cancer, stroke, hypertension), less likely to smoke, and more likely to be single than the average patient visiting the firms (P<0.05). Adjusting for age and sickness, frequent visitors were more apt to have experienced continuity of care during the prior year, to prefer VA care, and to be unemployed. CONCLUSIONS: Consecutive visit-based sampling actually selected patients with a visit pattern more typical of the patient population visiting four or more times a year. Studies using sampling of consecutive visitors will typically under-represent low users of care and should account for the degree to which results may not generalize to the broader practice population.