Estimating hospital costs by diagnosis for population-based analysis.

Knowledge of cost or charge for a discharge with a specified diagnosis to a given hospital is a prerequisite for population-based analysis of acute hospital utilization. As this information is not usually available directly for all discharges of residents of a geographical area, estimates are required. We compared alternative estimates derived from the data being assembled in many states--from hospital discharge data systems and routine cost and statistical reports. The independent variables were the name of the hospital, its total inpatient costs, and the diagnosis and length of stay for each discharge. We verified our estimates with data from Maine, for which charges were also available for many discharges. The estimate that used all three of the variables from the discharge data bank was most accurate. It explained 77.3 percent of the variability in hospital average charges per case for eight representative diagnoses. A simpler estimate, not requiring knowledge of diagnosis, proved almost as accurate. This estimate, defined as the product of cost per day for a hospital times length of stay for a case, explained 76.1 percent of the variability in individual charges per case, and 91.9 percent of the variability in average cost per case. Thus, per capita expenditures on hospital care for any diagnosis, and for inpatient care in total, can be estimated reliably by combining information from discharge data banks with routinely reported hospital per diem costs.