BACKGROUND: Decisions about reimbursement of new medical technologies are taken under uncertainty. Making the wrong decision comes with a cost. The purpose of an expected value of information (EVPI) analysis is to quantify the costs of taking the wrong decision. An EVPI analysis may also provide information on the parameters for which additional research is most useful to reduce uncertainty (i.e. expected value of perfect partial information; EVPPI). To perform an EVPPI, two-stage sampling from the original model is required. In practice, it is often not possible to perform an EVPPI because of computer time constraints. OBJECTIVES: The aim of this study is to investigate the usefulness and feasibility of constructing a metamodel to estimate the EVPPI for individual model parameters. METHODS: The metamodel was based on a cost-effectiveness Markov model with 5000 iterations for the treatment of patients with chronic obstructive pulmonary disease. The metamodel constituted a linear regression, using the net monetary benefit of the 5000 simulations as the dependent variable and the input values of the model parameters (e.g. probabilities for exacerbations, hospital admissions, etc.) as independent variables. This regression model was used to calculate the EVPPI for each individual input parameter of the original cost-effectiveness model. RESULTS: The EVPI was €3053. The analysis showed that the regression-based metamodel provided a good fit of the original model. The adjusted R2 of the regression model was 98%. The results obtained with the metamodel closely resembled the results of a two-stage sampling EVPPI based on the original model. CONCLUSIONS: In this study, a regression-based metamodel provided reliable estimates of the EVPPI. We conclude that for cost-effectiveness models that meet the linearity criterion, regression based metamodelling promises to be a very efficient and timesaving approach for the calculation of the VOI for individual model parameters.