Internal medicine manages patients with multiple comorbidities or rare diseases, for whom the scientific literature often provides limited guidance. Patient similarity search aims to identify, within a medical database, individuals whose clinical, biological, morphological and longitudinal profile is close enough to that of a given patient for their experience to inform clinical care. This approach extends the analogical reasoning familiar clinicians. The growth of health data warehouses and recent advances in artificial intelligence, now make it possible to automate this search and integrate heterogeneous data sources. Applications include diagnostic support, prognostic evaluation, treatment personalization and clinical trial recruitment. Although intuitive and explainable, this approach faces technical limitations and a risk of automation bias, justifying regulatory oversight and appropriate physician training.
