Diagnostic tests are the starting point of clinical decision making and are therefore central to the quality of patient care. High quality diagnostic studies are essential to identify the most appropriate test for a specific person in a given clinical context. While statistical methods for therapeutic studies are well established, methodological approaches for diagnostic studies are less advanced. A key gap concerns the handling of clustered data. Multiple observations per individual, for example from several lesions or repeated measurements over time, are often not adequately considered in conventional analyses of diagnostic accuracy measures such as sensitivity, specificity, and the area under the ROC curve.

ClusterDiag addresses this gap: The aim of this project is to develop a comprehensive methodological framework for the design and analysis of diagnostic studies with spatially or temporally clustered data and to translate it into research practice. To this end, currently available statistical approaches will be systematically reviewed and compared in simulation studies. Building on an extension of the existing estimand framework to clustered data, new methodological strategies will be developed to obtain valid estimates. In addition, methods for predictive values as well as approaches for sample size planning in clustered study settings will be advanced.

All results will be consolidated in a practical guidance document and an open access R package, providing researchers with concrete tools for application in practice.

In this way, ClusterDiag will contribute to improving the quality and reliability of diagnostic studies and strengthen their translation into clinical research.