Objective: The objectives of this study were (i) to develop a conceptual framework for selecting hospitals for unit cost estimates in national and international multicentre trials and (ii) to test the impact of alternative hospital selection on the cost results.
Design and setting: Within the conceptual framework, the following considerations which can be used when selecting a sample of hospitals for unit cost estimates in multicentre trials were identified: the number of hospitals; the sampling method; and the desired level of geographical subanalysis. Results from a recently completed international multicentre trial were used to explore changes in cost results obtained by using alternative methods of selecting and stratifying hospitals for unit cost estimates.
Patients and participants: The study included 5041 women from 72 hospitals in 6 countries with prelabour rupture of the membranes at term.
Interventions: The women were randomly assigned to induction of labour with intravenous oxytocin, induction of labour with prostaglandin E2 gel, or expectant management for up to 4 days with labour induced if complications developed.
Main outcome measures and results: Across each of the 4 management strategies of the study, the method of selecting and stratifying hospitals resulted in a 30 to 55% difference between the lowest and highest median unit cost estimates. In some cases, the relative ranking of the least to most expensive strategy varied across methods of hospital selection. The statistical comparisons across strategies found that the method used had a substantial impact on the conclusions of the economic evaluation.
Conclusions: Unit cost information should be collected from as many hospitals as possible. Multivariate hospital cost studies are needed to identify important cost drivers that will assist with hospital selection in the future.