Unser Doktorand Daniel Dümmler hält am Donnerstag, den 11. Dezember 2025 von 14:00 bis 15:00 Uhr einen hochschulöffentlichen Vortrag mit Bezug zu seinem Promotionsthema. Der Vortrag findet im großen Seminarraum des Instituts für Epidemiologie und Sozialmedizin und per Zoom statt. 

Vortragssprache / Language: English

Meeting-ID: 657 9764 6976

Kenncode: 181271

Abstract:

Diagnostic test accuracy (DTA) studies play a central role in evidence-based diagnostics and regulatory evaluation. While DTA metrics are well established, certain testing inherently generate complex clustered data structures, which require caution in estimation and interpretation. Clustered data arise when multiple observations are obtained from one individual: such as tests conducted at several lesions or repeated over time. These observations are typically positively correlated, and failing to account for intra-individual dependencies may lead to biased estimates and underestimated variance. Although various statistical methods have been proposed, no comprehensive synthesis or practice guideline currently exists to inform method selection, and ad hoc choices risk generating inconsistent evidence. This review aims to identify, categorise, and evaluate statistical methods for analysing DTA in clustered data settings, focusing on approaches that address temporal and spatial clustering across common measures such as sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC).