Generalizing the plurality method for forming hospital service areas.

The upcoming Health Care Financing Administration's Fourth Scope of Work for peer review organizations (PROs) envisions much use of geographic analysis of utilization rates and quality of care. Proper analysis of utilization rates requires each PRO to form multiple sets of hospital service areas. The method used most often in the literature is the plurality method. Because this method can create fractured service areas and can leave hospitals without a service area, the service areas and their associated hospitals often are reworked by hand. This last step drastically raises the effort required to form service areas and makes the method nonreproducible. This report defines the generalized plurality method for forming the hospital service areas that are central to the study of use patterns via small area analysis. This new method is a true generalization of the plurality method. Like the plurality method, it forms service areas by allowing geographic areas to "vote" for their preferred hospital. The generalization is achieved by allowing near-ties in the voting to cause clustering of hospitals. Hence it is a clustering algorithm that operates on both the geographic areas (sources) and the hospitals (destinations) at the same time. It is a nonhierarchical, nonagglomerative cluster method. There are several free parameters that may be chosen to adjust the effect of the clustering by adjusting the definition of a near-tie, as well as the sensitivity of the clustering to near-ties from sources with a small number of votes. This automated method enjoys many major advantages over the methods commonly appearing in the literature: it is completely reproducible, it is quick, and it does not require the a priori convening of a panel of experts. It thus can be applied easily to a wide variety of types of care that would not necessarily have the same service areas.