Neighboorhood Effects in Community Trials for Cardiovascular Disease
BACKGROUND:Cluster-randomized community trials are critically important to cardiovascular (CVD) epidemiology. While studies evaluating the effect of community education programs report favorable risk factor changes, including the Pawtucket Heart Health Program (PHHP), changes in hard CVD endpoints have been less common. Few have examined the effect of neighborhood on endpoints, especially controlling for individual characteristics. In short, overall effects may be confounded with neighborhood. METHODS:In the first study of this type, we geocoded PHHP neighborhoods to 1990 Census tracts, and randomly sampled 50% (n=521) of the observations from the 6th PHHP cross-sectional survey. Endpoints included CVD knowledge (number of correct risk factors mentioned), and total serum cholesterol. Income and percent working poor (households in a tract with income < 150% of the poverty-line) served as individual- and neighborhood-level SES, respectively. SAS Proc Mixed was used to fit multilevel models. Geographic Information Systems (GIS) methods map empirical Bayes estimates. RESULTS:Choropleth maps show variation in CVD knowledge and cholesterol across the 22 tracts of Pawtucket. Multilevel models reveal that income accounts for most of this neighborhood-level variation. However, a significant negative interaction between individual- and neighborhood-level SES (p < 0.10) was observed for CVD knowledge. Predicted values for CVD knowledge show that persons with a moderate income (level 5) living in a neighborhood with a large proportion (65%) of working poor would be expected to have a knowledge score of 2.34. This same person living in a neighborhood with a small proportion of working poor (15%) would be expected to have a knowledge score of 3.19. This represents a 14% difference due to the neighborhood effect. COMMENT: These results suggest that having more individual-level income is positively related to CVD knowledge; such effects are muted if one resides in a neighborhood with a larger proportion of working poor. It may be that future community trials for CVD may benefit by targeting, or concentrating more heavily on, not only poor persons but persons living in poor neighborhoods.