Higher quality and lower cost from improving hospital discharge decision making

Excess, wasted costs in healthcare are estimated to exceed the size of the entire defense budget. Effectively addressing this problem requires decreasing the cost of healthcare while increasing its quality by improving healthcare decision making. This paper reports research on improving decisions about hospital discharges ‒ decisions that are now made by physicians and based on mainly subjective evaluations of patients’ health status. The research has three components: (a) econometric identification of patient medical and demographic variables that discriminate between likelihood of successful discharge and likelihood of subsequent readmission within 30 days (the Medicare threshold); (b) development of decision support software that can be used in more effective, evidence-based discharge decision making; and (c) empirical evaluation of effectiveness of the decision support software. 

We begin with a large sample of data from the “data warehouse” of patient electronic medical records of a large southeastern teaching hospital. A probit model is estimated with these data. The estimated probit model provides the foundation for a decision support model that uses data from the electronic medical record for an individual patient over his course of stay in the hospital to generate patient-specific and day-specific probabilities of readmission and selects those clinical variables that are key to predicting outcomes for the individual patient. The decision support model presents the physician with the estimated daily readmission probabilities (with error bounds) and dynamically-selected clinical variables for the individual patient in a user friendly format.

Empirical evaluation of the model includes both a laboratory experiment and a field experiment, in the form of a hospital patient ward intervention.  The laboratory experiment is reported in this paper. We use a two-by-three experimental design defined over three information treatments and two time constraints on patient care. The experiments are conducted using resident physicians and fourth-year medical students at a university medical school as subjects.  

Experimental data reported herein indicate that utilizing the decision support software with change in the default option reduces (i) patient length of stay in the hospital and (ii) the probability of readmission for higher risk patients.