Comparison of hospital costing methods in an economic evaluation of a multinational clinical trial

OBJECTIVES: To develop and evaluate strategies for estimating hospitalization costs in multinational clinical trials. METHODS: Hospital cost estimates for eleven diagnoses were collected from twelve countries participating in a trial of therapies for congestive heart failure. Estimates were combined with U.S.-based diagnosis-related group weights to compute country-specific unit cost estimates for all reasons for hospitalization. Variations of hospital costing methods were developed. The unit cost method assigns a country-specific unit cost estimate to each hospitalization. The other methods adjust for length of stay using a daily cost (DC) estimate for each diagnosis, based on either the mean length of stay (DC-mean method) or the median length of stay (DC-median method) for each diagnosis in each country. Additional modifications were explored through adjustment of the distribution of daily costs incurred during a hospital stay. RESULTS: The mean cost for all hospitalizations was dollars 10,242 (SD, 10,042) using the unit cost method, dollars 10,242 (SD, 12,760) using the standard DC-mean method, and dollars 13,967 (SD, 18,762) using the standard DC-median method. In comparisons of costs for all 5,486 hospitalizations incurred by a subset of 2,352 patients in the trial, the unit cost method provided 92% power to detect a dollars 1,000 cost difference. The standard DC-mean method provided 76% power, and the standard DC-median method provided 44% power. CONCLUSIONS: Hospital costing methods that adjust for differences in length of stay require a significantly larger sample to attain comparable statistical power as methods that assign unadjusted unit cost estimates to hospitalization events.