Study design for prediction models

Abstract

In this chapter, we consider several issues in the design of studies for prediction research. These include the selection of subjects or patients for a cohort study, strengths and limitations of case series from a single center, from registries, or prospective trials. We further discuss issues in choosing predictors and outcome variables for prediction models. An important question is often how large a study needs to be for sufficient statistical power. Power considerations are given for studying effects of specific predictors, and for developing a prediction model that can provide reliable predictions. We use several case studies for illustration.