This is a case-control study of rural hospitals which closed between 1970 and 1980. Nonmetropolitan hospitals which closed were matched with a comparable set of hospitals which remained open, yielding 148 closures (cases) and 310 controls. Univariate and multivariate analyses indicate that closed hospitals differ from open hospitals in their internal characteristics, and that their geographical service areas also are significantly different. At the multivariate level, the most significant variables, as determined by stepwise discriminant analysis, are the ownership (profit) status of the hospital, occupancy rate, competitive beds within the county, scope of service, and county population change during the preceding decade. The nature of the discriminant functions differed between counties which were adjacent to metropolitan areas, and those which were not adjacent to metropolitan areas. Closed hospitals differ from open hospitals, as ascertained by standard data sources. Hospital closure reflects the interaction of internal hospital characteristics, and the characteristics of the hospital's service area.