Clinical prediction models are an important asset in contemporary decision making. These models estimate the absolute risk or probability that a certain condition or disease is present (diagnostic models) or will occur in the future (prognostic models) for a certain subject. Hereto, they combine information from so-called predictors, such as patient history and results from physical examination or medical testing. Risk estimates from prediction models are often used to inform the referral of patients for further testing, to plan lifestyle or therapeutic decisions or to risk-stratify participants in therapeutic clinical trials. It is crucial that prediction models are (sufficiently) accurate, as otherwise their implementation may lead to harm and/or unnecessary costs.
In this talk, I will highlight common challenges in prediction research, and discuss key opportunities for improving the accuracy of clinical prediction models. Hereby, I will focus on the merits of evidence synthesis, a topic I extensively studied since the start of my scientific career.