Predicting the effect of hospital closure on hospital utilization patterns.

Geographic patterns of hospital utilization were analyzed before and after the closure of Sydenham hospital in New York City. The purpose was to determine the accuracy with which hospital utilization patterns after closure could be predicted using standard spatial interaction modeling procedures. Gravity models were calibrated to represent travel to hospitals before and after closure for patients residing in Sydenham's primary service area. Using three variables, hospital size, distance and type, the models accurately described utilization patterns in each year. The distance parameter, however, changed substantially between the 2 years. In addition large errors were observed when the model calibrated before closure was used to predict utilization patterns afterward. The geographic distribution and likely causes of errors were analyzed, along with their implications for spatial modeling efforts.