Hospitals, like many service firms, operate under uncertainty about future demand for their services, but must contract in advance for labor, technology, support services, and other inputs. They face the risk that the capacities of the inputs they contract for any future day are too large, resulting in inefficiency, or too small, resulting in congestion. Congestion is a state of production in which processes are slowed by undercapacity of critical resources along the production pathway. This research develops a theoretical and empirical approach to the question of whether congestion in general hospitals affects hospital mortality. Empirical analysis which examines 3.2 million admissions to Pennsylvania hospitals in 1988 demonstrates the following results. First, congestion is not related to patient characteristics and is rather affected by the state of production in the facility. Second, there is a significant relationship between congestion and mortality rates for all hospitals. Hospitals exhibit different threshold levels of congestion at which mortality occurs, and different rates of mortality after this threshold. Third, mean length of stay is not significantly related to congestion. Through the insights gained from the empirical analysis, a theory of congestion which extends previous research and which considers the unique nature of hospital production is developed. This theory allows quality of care (excess mortality) and total treatment time (length of stay) to be determined in part by congestion. Hypothesis about the optimal level of congestion in facilities are drawn from the theory, and can guide further research in this area. The following conclusions are supported by this research. Congestion is an important aspect of hospital production because it is associated with excess mortality. Managers should therefore improve demand forecasting capabilities, manage the flow of arrivals, or reorganize critical components of service delivery in order to reduce congestion. Hospitals may reduce congestion by creating autonomous production units instead of traditional hospital wards, so that hospital-wide effects of congestion can be minimized. More research should be performed about dynamic aspects of hospital care, and data collection should be aimed at providing such data.
