Multiple-regression hospitalization-cost model for pharmacy cost analysis

Abstract

A multiple-regression model that used hospital data for a single diagnosis-related group (DRG) was developed and tested as a tool for measuring the effect of pharmaceutical services on costs. Computer-generated data from one university teaching hospital for fiscal year 1983 were used to develop a model for identifying patient-specific and treatment-specific factors that correlated with total hospital costs and adjusted total costs (total costs minus room and board). Costs were analyzed for patients hospitalized for cardiothoracic procedures (DRG 108). Data for 100 patients were included. The demographic, severity, treatment-intensity, and drug-use factors included in the model accounted for 95.6% of the variation in adjusted total cost. Length of stay, number of secondary surgical procedures, quantity of supplies or drugs used in the operating room, and several drug therapies--tobramycin, i.v. electrolytes, and i.v. fluids--had significant effects on adjusted total cost. This regression model was useful for assessing the effects of the choice of drug therapy or specific services on the costs for one DRG; it should be tested for other DRGs in other institutions. The usefulness of the model for planning cost-effective services is limited by the need for better billing information, cost-estimation techniques, and records of clinical services provided.