Use of a Geographic Information System to create treatment groups for group-randomized community trials: The Minnesota Heart Health Program

Abstract

Background: Group-randomized trials of communities often rely on the convenience of pre-existing administrative divisions, such as school district boundaries or census entities, to divide the study area into intervention and control sites. However, these boundaries may include substantial heterogeneity between regions, introducing unmeasured confounding variables. This challenge can be addressed by the creation of exchangeable intervention and control territories that are equally weighted by pertinent socio-demographic characteristics. The present study used territory design software as a novel approach to partitioning study areas for The Minnesota Heart Health Program's "Ask about Aspirin" Initiative. Methods: Twenty-four territories were created to be similar in terms of age, sex, and educational attainment, as factors known to modify aspirin use. To promote ease of intervention administration, the shape and spread of the territories were controlled. Means of the variables used in balancing the territories were assessed as well as other factors that were not used in the balancing process. Results: The analysis demonstrated that demographic characteristics did not differ significantly between the intervention and control territories created by the territory design software. Conclusions: The creation of exchangeable territories diminishes geographically based impact on outcomes following community interventions in group-randomized trials. The method used to identify comparable geographical units may be applied to a wide range of population-based health intervention trials.