Multivariable prognostic models combine several characteristics to provide predictions for individual patients. Prognostic models can be applied in research and clinical practice, for instance to assist clinicians with decisions regarding treatment choices or informing patients and family members on prognosis (1). Before application in clinical practice, prognostic models should be validated to judge their generalizability. Although guidelines have been proposed to improve development and reporting of prognostic models, a majority of the published models is not thoroughly validated (1,2). In this viewpoint, we focus on design and analysis of validation studies for prognostic models. For illustration, we consider the validation of the International Mission for Prognosis and Analysis of Clinical Trials in Traumatic Brain Injury (IMPACT) prognostic models for patients with moderate and severe traumatic brain injury. These models combine clinical, radiological and laboratory admission characteristics to predict risk of mortality and unfavorable outcome (3). A second example is on computed tomography (CT) decision rules in patients with minor head injury (4).

Additional Metadata
Persistent URL dx.doi.org/10.21037/jeccm.2018.10.10, hdl.handle.net/1765/115275
Journal Journal of Emergency and Critical Care Medicine
Citation
Dijkland, S.A, I.R.A. Retel Helmrich (Isabel), & Steyerberg, E.W. (2018). Validation of prognostic models: challenges and opportunities. Journal of Emergency and Critical Care Medicine, 2(91), 1–4. doi:10.21037/jeccm.2018.10.10