This project deals with the development of risk prediction models basedon multicentre data. Risk prediction models estimate the risk of having(diagnosis) or developing (prognosis) a disease. One reason to carry out multicentre studies is that they enhance the generalization of the results. The recruitment of subjects from a wider population and a broader range of clinical settings results in a sample that is more typical of the target population. Multicentre data have a clustered nature:patients from the same centre are likely to be more similar than patients from different centres. Standard analysis methods do not take thisinto account, leading to biased or invalid results. Methods that adjustfor centre exist but are not commonly used. The goal of this study is to optimize strategies for the development of prediction models based on multicentre data. Four aspects of the planning and analysis of multicentre research will be investigated: sample size, variable selection, the choice of analysis method, and the application of the model in new centres. Guidelines for the clinical researcher conducting multicentre studieswill be provided that strengthen the validity, generalizability and local applicability of the developed models.