Forecasting patient tray census for hospital food service.

Five computerized forecasting models were tested with data on daily patient tray demand in a large medical center food service, and results were compared with intuitive forecasts made by the food service supervisor. All five models gave more accurate results than the intuitive procedure; an adaptive exponential smoothing model was most accurate. The effects of model complexity and data storage requirements are discussed, and simple exponential smoothing is suggested for forecasting patient tray demand in this setting.