Objective: The objective of this study was to test the hypothesis that dental practice characteristics are associated with tooth loss incidence with both tooth-specific and patient-specific characteristics already taken into account. Research Design: A population-based prospective cohort study was conducted. In-person interviews and clinical examinations were done at baseline, 24, and 48 months, with telephone interviews every 6 months. Practices that coincidentally served participants in the study completed practice characteristics questionnaires. To increase inferential power when testing practice-level effects, detailed tooth-specific and patient-specific data were simultaneously taken into account in tests for association between practice-level effects and tooth loss. Setting: Data were from the Florida Dental Care Study. The key health outcome was tooth loss, a leading measure of a population's oral health. Participants: Eight hundred seventy-three African-Americans and non-Hispanic whites who had at least 1 tooth. Results: Certain practice characteristics were associated with tooth loss, including the racial mix of the practice's patient population; persons who attended practices with higher percentages of African-Americans were more likely to receive a dental extraction regardless of the individual patient's race. Conclusions: This is the first longitudinal report of increased risk for tooth loss resulting from practice-level effects. Although a patient-level racial disparity remained evident, and even with detailed tooth-specific and patient-level characteristics taken into account, racial differences in characteristics of practices attended independently contributed to the patient-level racial disparity in health. Copyright © 2006 by Lippincott Williams & Wilkins.