Evaluation of Clinical Usefulness

Abstract

Performance measures such as discrimination and calibration consider the full range of risk predictions. We may also want to know whether a prediction model is useful to support medical decision-making: is the model beneficial to guide selection of subjects for screening, for diagnostic work-up, or decision-making on therapy? For such decisions, we need a cutoff for the predicted probability (“decision threshold”, or “classification cutoff”, see Chap. 2). Patients with predictions above the cutoff are classified as positive; those under the cutoff as negative. We will use the term “clinical usefulness” for a model’s ability to make such classifications better than a default policy without the prediction model.