From static fracture classification to dynamic clinical phenotypes: Feature-based longitudinal annotation and...

Traditional surgical classifications provide standardized languages for describing clinical conditions at diagnosis, but they necessarily simplify complex and varying biological and mechanical phenomena into predefined categories. In fracture surgery, the AO/OTA classification has substantially improved communication and comparability by providing a standardized morphological framework. However, patients with apparently similar morphology at diagnosis may follow substantially different clinical trajectories after treatment. We propose a complementary approach based on feature-based annotations and longitudinal documentation of the complete treatment course. Rather than representing a patient solely by a static classification at time zero, this approach describes a dynamic clinical phenotype, comprising patient characteristics, fracture features, intraoperative events, postoperative course and outcome. Such structured longitudinal data enable the identification of comparable real-world cases beyond conventional classification subgroups. Artificial intelligence may facilitate the integration and analysis of these multidimensional data and, importantly, help identify when apparently similar clinical trajectories begin to diverge. This framework can support surgical education and shared decision-making by providing easy access to comprehensively documented comparable cases.