PARAMETERS TO AID IN THE DESIGN AND ANALYSIS OF COMMUNITY TRIALS - INTRACLASS CORRELATIONS FROM THE MINNESOTA-HEART-HEALTH-PROGRAM

Community trials involve the assignment of intact social groups to stu dy conditions and are becoming increasingly common in epidemiologic re search. In both the design and analysis of these studies, whether cros s-sectional or cohort, allowance must be made for the dependence of el ements within intact groups if variances are to be properly estimated. In the design phase, the statistician needs estimates of the level of dependence likely to be encountered. In the analysis phase, external estimates of the level of dependence may be useful in preventing the e rosion of power associated with small numbers of intact groups assigne d to each condition. We report the intraclass correlation coefficients of the city-year component of variance as estimated in the Minnesota Heart: Health Program for a variety of community survey variables and illustrate their use in both design and analysis. Of 23 variables asse ssed, all but two showed positive estimates of city-year intraclass co rrelations. In these data, estimates of intraclass correlation coeffic ients generally were in the range 0.002-0.012.