Prediction Models and Clinical Outcomes—A Call for Papers
Stephan D. Fihn, Jesse A. Berlin, Sebastien Haneuse, Frederick P. Rivara · JAMA Network Open · 2024
that internal validation techniques, such as split samples, hold-out, k-fold, and others, are not designed to overcome the intrinsic differences between data sources and, therefore, are not suited to quantifying performance externally.While the population to which the models apply should be described explicitly, ideally any such models should be applicable to patients from the wide range of races, ethnicities, and backgrounds commonly encountered in clinic practice.Most importantly, we are interested in examples of models that have been evaluated in clinical settings, assessing their feasibility and potential clinical benefit.This includes studies with negative as well as positive outcomes.