This paper describes the advantages and disadvantages of two distinct approaches to computing hospital daily occupancy in large datasets. The first approach employs an incremental method for computing occupancy. Initial occupancy is calculated by using the SQL procedure and then a Retain statement in a datastep creates a running sum of daily occupancy. The second approach uses a data expansion technique; A do loop in a datastep converts the patient level data to patient-day grain. After the data is expanded, dummy (zero/one) variables are created to indicate whether a patient has been admitted, discharged, or is currently in the hospital on a given day. The counter variables for admissions, discharges, and current occupancy are summed to create a daily tabulation of hospital census, discharges, and admissions. Although both approaches yield similar results in the absence of data quality issues, they are quite different in their logic and programming technique. The two methodologies were developed in the context of research into how fluctuations in demand affect hospital behavior and are relevant and applicable to any business situation involving a stock and flow relationship. The two techniques can also pinpoint deficiencies in the data. The computations yield results that are used to analyze patterns in hospital occupancy.
conference:
