J Clin Epidemiol Vol.44, No. 7, pp. 701-713, 1991 0895-4356/91 $3.00 + 0.00 Printedin Great Britain. All rightsreserved Copyright 0 1991 Pergamon Press plc DATA ANALYSIS AND SAMPLE SIZE ISSUES IN EVALUATIONS OF COMMUNITY-BASED HEALTH PROMOTION AND DISEASE PREVENTION PROGRAMS: A MIXED-MODEL ANALYSIS OF VARIANCE APPROACH THOMAS D. KoEPsELL,‘.~+~* DONALD C. MARTIN,~ PAULA H. DIEHR,“~ BRUCE M. PsATY,I’~~~ EDWARD H. WAGNER,‘*’ EDWARD B. PERRIN”~ and ALLEN CHEADLE’ Departments of ‘Health Services, *Epidemiology, ‘Medicine, and “Biostatistics, University of Washington, Seattle, WA 98195 and ‘Center for Health Studies, Group Health Cooperative of Puget Sound, Seattle, Washington, U.S.A. (Received in revised form 26 September 1990) Abstract-The growing interest in community-based approaches to health promotion and disease prevention (HP/DP) has been accompanied by a growing need to evaluate the effectiveness of such programs. Special issues that arise in these evaluation studies include (1) entire communities are assigned to intervention and control groups, (2) only a small number of communities can usually be studied, (3) the time course of changes in behavior and other outcomes is often of interest, and (4) surveys to measure such changes over time can be conducted with either repeated cross-sectional samples or with longitudinal samples. This paper shows how these issues can be addressed under a mixed-model analysis of variance approach. This approach serves to unify several ideas in the literature on evaluation of community studies, including use of time-series regression and the question of whether the individual or the community should be the unit of analysis. We also describe how the method can be used to estimate sample size requirements, statistical power, or minimum detectable program effect. Sample size Statistical power Evaluation INTRODUCTION Over the past decade there has been increasing interest and investment in community-based approaches to health promotion and disease prevention (HP/DP), stimulated in part by promising results from such early programs as the North Karelia Project [l] and the Stanford Three Community Study [2]. These programs *All correspondence should be addressed to: Thomas D. Koepsell, M.D., M.P.H., Department of Health Ser- vices, SC-37, University of Washington, Seattle, WA 98195, U.S.A. [Tel: (206) 543-79521. Supported in part by Grant No. CA 34847 from the National Cancer Institute, and by a grant from the Henry J. Kaiser Family Foundation. target large segments of the population in a designated geographic area and use unselective intervention modalities such as the mass media and activation of community organizations in order to bring about changes in community norms, the social and physical environment, individual health behavior, and ultimately health outcomes [3]. Growth in the number and scope of community-based HP/DP programs has been accompanied by growing recognition of several difficulties inherent in evaluating their effectiveness. First, because entire communities are the targets of interventions, controlled evaluation designs must involve the comparison of a group 701