Key points are not available for this paper at this time.
The Minnesota Heart Health Program is a 13-year research and demonstration project to reduce morbidity and mortality from coronary heart disease in whole communities in the upper Midwest. Six communities were selected for the study: three intervention and three comparison sites, matched to increase baseline comparability. After 2-4 years of baseline observations, a 5- to 6-year program of intensive intervention was introduced in the three intervention communities. Periodic cross-sectional and cohort surveys provided data on risk factors and related behaviors. Regression adjustments within and between communities reduced the confounding influences of important covariates and the variance inflation associated with the nesting of individuals within communities and surveys. Post hoc stratification allowed exploration of the main and strata-specific effects of the intervention program. Finally, the intervention effect was modeled as a departure from the trend line fit to the nonintervention city-year means. Together, these procedures explicitly acknowledged the component of variance associated with communities, and so avoided a major source of bias created in the usual analysis when that variation is ignored. They also increased the interpretability of the analyses and reduced the mean square errors used to assess the treatment effects.
Murray et al. (Sat,) studied this question.