Clinical Prediction Models: A Practical Approach to Development, Validation and Updating by STEYERBERG, E. W.

Clinical prediction models play a very important role in health care. These models are used to predict the risk of health outcomes that assist in the clinical management of patients, enable patients to make informed decisions about their treatment, help making fairer comparison of institutional performances, and are used to determine eligibility of patients for clinical trials. The methodology used in the development and application of clinical prediction models in medical publications is currently not adequate. The aim of this book is to provide a better understanding of the model development process and present some recommendations regarding modeling strategies to improve current practice. The book is written by Ewout Steyerberg, who is a professor of medical decision modeling at the Erasmus MC-University Medical Center, Rotterdam and has published extensively on this topic. In this book, the author discusses study designs for prediction models, the problems of model over fitting and optimism and modern estimation methods that could be used to overcome this problem, choosing between alternative statistical models, handling missing values, and continuous and categorical predictors, variable selection strategies, assumptions of additivity and linearity required by models, use of external information in model development, measures and methods of model validation and clinical usefulness, presentation and updating of models. Some case studies based on the development of risk models on real data are presented. The book is suitable for epidemiologists, health services and public health researchers, clinicians with quantitative knowledge, and applied biostatisticians. Some R codes are presented. The data sets presented in the book as examples are available from the website to enable other people to use these to practice model development. The application areas considered are public health, clinical practice, and medical research. Continuous, binary, categorical, and survival outcomes are considered.