Predicting productivity based on EQ-5D: an explorative study.

Background: Productivity costs are often ignored in economic evaluations. In order to facilitate productivity cost inclusion, it has been suggested to estimate productivity costs indirectly using quality of life data.

Objective: This study aimed to derive and validate an algorithm for predicting productivity losses on the basis of quality-of-life data using the EQ-5D-3L.

Methods: A large representative sample of the Dutch general public (n = 1,100) was asked in a web-based questionnaire to state their expected level of productivity in terms of absenteeism and presenteeism for multiple EQ-5D health states. Based on these data, two generalized estimating equations (GEE) models were constructed: (1) a model predicting levels of absenteeism and (2) a model predicting presenteeism. The models were validated by comparing model predictions with conventionally measured productivity within a group of low back pain patients.

Results: Predicted absenteeism levels based on EQ-5D health state closely resembled conventionally measured absenteeism levels. Productivity losses related to presenteeism seemed somewhat overestimated by our prediction model. Measured and predicted productivity were moderately but highly significantly correlated.

Conclusions: Overall, it appears possible to make reasonable productivity predictions based on EQ-5D data. Further exploration and validation of prediction algorithms remains necessary, however, especially for presenteeism.