Forecasting hospital laboratory procedures

Publisher Website
Google Scholar
Abstract
Improved forecasts of hospital laboratory procedures can provide the basis for better resource planning and enhanced operating efficiency. The research reported here-in describes how multiple regression models can be both a source of insight into causal relationships and a tool for achieving accurate monthly forecasts. Past research in this area may have overstated the statistical significance of findings because of a failure to address the potential effect of serial correlation. The present study uses the Cochrane-Orcutt regression procedure, rather than OLS, to overcome this problem. A model using inpatient admissions, acuity days, length of stay, discharge days and seasonal dummy variables is shown to account for 87% of the variation in the number of billable laboratory procedures. A simpler multiple regression model and a Winters' exponential smoothing model were found to provide excellent forecasts for laboratory procedures. In a one year out of sample evaluation, the annual percent forecast error was 0.7% for the regression model. This compares favorably to a percentage forecast error of 11.6% using subjective forecasting methods.

This publication has 4 references indexed in Scilit: