A method for identifying comparison counties: application to a rural telemedicine evaluation project

The evaluation of telemedicine activity in rural communities is complicated by the fact that most telemedicine sites are chosen because of their existing telecommunications infrastructure and institutional relationships, not by a random selection process. In addition, it is difficult to draw conclusions about the effects of telemedicine without a careful analysis of parallel changes in communities which do not have access to telemedicine services. We have developed a method of identifying comparable counties based on an aggregate measure of health status. A set of 66 variables was collected in a previous project to develop a model to evaluate the relative health status of the population in Missouri. A stepwise regression was used to identify a subset of 15 variables that had the highest predictive value for the health status of a county. Distance measures were then used to identify six counties which were most similar to three telemedicine counties. The method can be used with any study set chosen non-randomly, to identify similar objects that can be used for comparative purposes.