Background: In reporting economic evaluations of clinical trials, results are often stated without a description of the methodology used to derive the cost estimates. We compared methods for measuring costs from multicenter clinical trials to determine the extent to which the methodology affects results.
Methods: Patient-level data (n = 1849) from 3 multicenter clinical trials of percutaneous coronary revascularization were used to compare 4 methods of estimating costs: 1) hospital charges; 2) hospital charges converted to costs by use of hospital-level cost-to-charge ratios; 3) hospital charges converted to costs by use of department-level cost-to-charge ratios; 4) itemized catheterization laboratory costs with nonprocedural hospital costs generated from department-level cost-to-charge ratios.
Results: The method used to approximate costs did not affect the main results of the economic comparisons for any of the trials. The magnitude of the cost estimates and the cost differences between treatment groups varied considerably by method, however. Charges were approximately twice as high as hospital cost estimates. At the patient level, costs generated by use of method 1 were within 10% of those generated by use of method 4 for only 5% of patients, compared with 34% and 22% of patients with methods 2 and 3, respectively. Only method 3 produced estimates of between-group cost differences that were consistently within $500 of the reference standard.
Conclusion: Cost estimates derived from clinical trials in the cardiovascular arena vary substantially according to accounting methodology. Thus, in reporting the results of economic analyses, a detailed description of cost derivation is necessary, particularly when the absolute magnitude of the cost estimates is important to clinical decision-making or policy-level recommendations. For the purposes of group-level comparisons, conversion of hospital charges to costs on the basis of department-level cost-to-charge ratios appears to represent a reasonable compromise between accuracy and ease of implementation.