Personalized predictions about treatment outcomes can be made by prospectively collecting the required data or by using routinely collected big data. As the first option is very resource intensive, it should be assessed whether the use of the available data is appropriate for specific outcomes and conditions. In this project, this will be done for the prediction of return to work in patients with traumatic brain injury (TBI).
The current project aims to make predictions about the probability of different employment situations, one year after TBI. Big data on hospital stay, healthcare utilization and employment will be linked at an individual level. The additional collection of an extended dataset, retrieved from electronic health records and questionnaires, will allow for an evaluation of the appropriateness of big data for the prediction of post-injury employment. First, this will be done by validation of the predictions resulting from these data. Second, it will be assessed whether the additional variables improve the accuracy of the predictions. In case of a superior extended model, it will be verified which of the additional variables contribute most to the increase in predictive performance. For these variables, the cost of gathering
and pooling data for Flanders will be estimated. The best resulting model can be used to inform patients, but also clinicians and policymakers on how to optimize current practices regarding the trajectory towards reinstatement.