|
Abstract:
|
Medical diagnosis can be based on various clinical data (e.g., from imaging techniques, physical exams, and laboratory). Physicians combine these to arrive at a diagnostic conclusion. From that perspective multi-marker panels for diagnosis are intended to pre-aggregate some of the clinical data in an objective manner. Often the diagnostic result of some marker-panel is expressed by some score that is larger the more likely the medical condition at question is. As there is no obvious choice of scale for the score, physicians need advice on how to interpret these results. It is tempting to advocate a straight probability interpretation. Yet, what type of probability is appropriate? How should the validity of the probability claims be demonstrated? We will present some strategies to scale scores along with benefits and risks in their interpretation and performance evaluation.
|
