Predicting Demand

Whereas staff scheduling systems include information on future staff activity, and can accept changes to future work schedules as it happens, a prediction or forecast of the clinical demand is much more complex to obtain in a systematic way.

Legacy staffing and scheduling systems use two marginally adequate solutions to address this need. The first is to utilize a purely statistical forecast based on an analysis of historical data. These analyses produce average volumes that can pinpoint seasonal or special day variation from the norm. Statistical forecasts can be used successfully for budgeting and long-range schedule development, but due to the significant potential variation fall short for short-term day to day or hour to hour predictions.

The legacy solution to this is the push forward method employed by several vendor systems. This involves using whatever the current needs are as the prediction for the next 8, 12, or 24 hours. In a busy clinical environment where things can change rapidly this solution is far from adequate, but probably the best give the typical timeliness and quality of demand data.

A few vendors and providers have proposed systems that are labeled Predictive Modeling. Though several vendors label their pure statistical forecasting solutions, “Predictive Modeling” or “Predictive Analytics” they are not. A true predictive modeling system would need to leverage real time patient scheduling and arrival data combined with a detailed predictive pattern of the workforce needs of the patients expected stay.

CCWPP is not aware of a functioning true predictive modeling system available to the health care industry, but this is an area of immense potential for the management of the cost and quality of the healthcare system.