Statistical forecasting in a hospital clinical laboratory

Three forecasting methodologies were applied to monthly laboratory test count data in order to arrive at a best procedure for forecasting ahead to cover the next fiscal year. The purpose of the forecasting was, first, to aid in reimbursement and income decisions and, second, to assist in operations management decisions within the laboratory itself. The Box-Jenkins ARIMA models were found to be superior in all cases, and forecasts for individual test counts (as opposed to packages of tests billed as a unit) were improved if forecasts for inpatients and outpatients were done separately and then aggregated. With 2 years of experience to go on, the annual forecast error stands at around 4.5%.