Estimating study costs for use in VOI, a study of dutch publicly funded drug related research

Objectives: To perform value of information (VOI) analyses, an estimate of research costs is needed. However, reference values for such costs are not available. This study aimed to analyze empirical data on research budgets and, by means of a cost tool, provide an overview of costs of several types of drug-related research, as a reference for use in VOI analyses. Methods: A wide variety of study characteristics considered to be of impact on total study costs, were identified and discussed with experts. For a set of granted study proposals from the Netherlands Organization for Health Research and Development (ZonMw), information on study type, the selected characteristics (cost drivers), and budget were extracted. After full data-extraction, exploratory analyses were performed, followed by additional selection of cost drivers, based on inter-observer variation and overlap with other variables. A cost tool was developed including the remaining cost drivers as selection criteria. Results: We analysed 74 grant proposals concerning drug-related research from ZonMw. Exploratory analyses showed little coherence between cost drivers and budgets; univariate and multivariate analyses showed that only study type and duration were significantly associated to budget. Therefore, the cost tool provides reference values, and not a 'predicted budget'. Given user-defined choices on selection criteria, e.g. sample size and study duration, the tool returns the range of previously requested budgets for comparable studies. The range can be used in expected net benefit of sampling and real options analysis, or as a reference value for interpretation of other VOI outcomes. Conclusions: The absence of an association between study characteristics and budgets may signal inconsistencies in the budgeting or granting process. The tool nonetheless provides the user with useful information on historical budgets for a certain type of study. The underlying database can be complemented with new studies being granted, keeping estimates up-to-date