B0352
Title: Regularization approaches in clinical biostatistics: A review of methods and their applications
Authors: Sarah Friedrich - University of Augsburg (Germany) [presenting]
Andreas Groll - Technical University Dortmund (Germany)
Katja Ickstadt - TU Dortmund University (Germany)
Thomas Kneib - University of Goettingen (Germany)
Markus Pauly - Technical University of Dortmund (Germany)
Joerg Rahnenfuehrer - TU Dortmund University (Germany)
Tim Friede - University Medical Center Goettingen (Germany)
Abstract: A range of regularization approaches has been proposed in the data sciences to overcome overfitting, exploit sparsity or improve prediction. Using a broad definition of regularization, namely controlling model complexity by adding information in order to solve ill-posed problems or to prevent overfitting, a range of approaches within this framework is reviewed, including penalization, early stopping, ensembling and model averaging. To assess the extent to which these approaches are used in medicine, recent volumes of three journals publishing are systematically reviewed in general medicine, namely the Journal of the American Medical Association (JAMA), the New England Journal of Medicine (NEJM) and the British Medical Journal (BMJ). The literature review revealed that regularization approaches are rarely applied in practical clinical applications, with the exception of random effects models. However, statistical software is available and implementation is straightforward, as it is demonstrated in an applied data example on prostate cancer. In situations where other approaches also work well, the only downside of the regularization approaches is increased complexity in the conduct of the analyses. Hence, a more frequent use of regularization approaches in medical research is suggested.
