Adjusting to random demands of patient care a predictive model for nursing staff scheduling at Naval Medical Center San Diego

摘要

In this thesis, time series methods were used to forecast the monthly number of nursing Full Time Equivalents (FTEs) required to meet patient care needs at Naval Medical Center San Diego. In order to capture both patient census and patient acuities, the monthly total required workload hours given by the Res-Q system was used. The monthly number of nursing FTEs was calculated by dividing the total monthly workload hours required by 168 hours (per DoD 6010.13-M). The Holt-Winters' time series models were fit using both Excel and JMP software packages. Using three years of historical data to fit the models, the number of nursing FTEs that would be required every month for the fiscal year 2008 for the entire hospital was forecasted with a Mean Absolute Percentage Error (MAPE) of 17.83. Fitting the model to data starting from December 2005, to eliminate historical anomalies, further reduced the MAPE to 8.80. The overall model was, subsequently, partitioned into five sub-models, one for each of the five nursing units, reflecting the hospital's patient and nursing staff mixes. Again after adjusting for missing data points and outliers, the monthly number of nursing FTEs required for 4West, Adult ICU, Surgical, Medical, and Medical Oncology were forecasted with MAPE's of 20.77, 11.42, 13.63, 13.85, and 6.98, respectively.