Impact of Correlation of Predictors on Discrimination of Risk Models inDevelopment and External Populations

[摘要] Prediction models to estimate disease risk and identify individualsat high risk are widely advocated for optimizing prevention andmanagement of multifactorial diseases. For several common complexdiseases including dLوٴereQt forms of cancer, diabetes, andcardiovascular disease, a number of prediction models have beendeveloped in various populations. Нe predictive performance of theserisk models is typically assessed by evaluating discrimination which isthe ability of the model to separate those with events from thosewithout events. $іer developing a risk model, it is essential toinvestigate the model’s discriminative performance within an externalpopulation to judge about the generalizability and reliability of the riskmodel. A risk model that does not have an appreciable predictiveability in an external population may have limited opportunity toinvestigate further, aiming implementation in practice.