The Estimand Framework in Diagnostic Accuracy Studies.

Abstract

Diagnostic accuracy studies evaluate how well a diagnostic test can detect or rule out a medical condition. Different events can interfere with the conduct of the test, affecting the test result. Before starting a diagnostic test accuracy study, the clinical question of interest should be precisely defined. Based on that, strategies can be chosen for dealing with the interfering event. We introduce six different strategies for how such events could be handled. We introduce the estimand concept for diagnostic accuracy studies, which consists of the attributes population, target condition, index test, accuracy measure, and the strategies for handling interfering events. The estimand determines which effect is estimated based on the study objective. To bridge the gap between the clinical study objective and the method for the estimation, we illustrate the necessary steps using a fictitious computed tomography scan study. The defined estimand improves the structure of the planning phase, enhances the interdisciplinary exchange, and supports the interpretation based on the study objective.