Forty years of unwarranted variation--and still counting.

1 A subsequent household interview that compared six Vermont hospital service areas with widely varying spending and utilization rates showed no differences in self-reported illness, insurance coverage, ethnicIn 1967, fresh out of my medical training at Johns Hopins, I took a job as director of Vermont’s Regional Medical rogram (RMP) at the University of Vermont. The Vermont MP was part of a national program, sponsored by the ational Institutes of Health, whose goal was to ensure that ll Americans had access to the great advances in medcal care now available at academic medical centers. At ohns Hopkins, I had witnessed the advent of renal dialysis, hemotherapy, open-heart surgery and coronary intensive are units, and came to my new job with a good deal f enthusiasm for dealing with the underuse of medical are. But I also came as one trained in epidemiology and cquainted with the quantitative methods of the social scinces. It was thus quite natural for me to want to develop a atabase for identifying underuse of care among Vermont ommunities. I worked with Alan Gittelsohn, a biostatistician from ohns Hopkins who had also been my teacher, to develop a ethod for comparing the population-based rates of care mong neighboring hospital service areas. We called this he “small area analysis of health care delivery”. Our small reas were designed to maximize variation associated with ifferences in behavior of physicians according to the “maret” in which they practiced. We first conducted a patient rigin study that defined the geographic boundaries of local ealth care markets, of which there were 13 in Vermont, anging in population size from just under 5000 to about 00,000. (In each area, the large majority of services were rovided locally.) We then measured medical spending, esource inputs (such as physician labor and hospital beds), tilization of hospitals, nursing homes, physician services, iagnostic tests and surgical procedures, and, to the extent ossible, medical need and outcomes. The first view of the results brought a big surprise. While e had expected to find underservice in many parts of ermont, we found instead a typology of care characterzed by vast variations in the deployment of resources and he utilization of services among neighboring communiies, without apparent rhyme or reason. Here is a brief synopsis of what we reported in our first ublication, the 1973 article in Science [1]: