Data-driven phenotyping of infective endocarditis using unsupervised machine learning: toward a personalized approach
BMC Infectious Diseases, cilt.26, sa.1, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 26 Sayı: 1
- Basım Tarihi: 2026
- Doi Numarası: 10.1186/s12879-026-14119-3
- Dergi Adı: BMC Infectious Diseases
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, CINAHL, EMBASE, MEDLINE, Directory of Open Access Journals, Academic Search Ultimate (EBSCO), Biomedical Reference Collection: Corporate Edition (EBSCO), Health Research Premium Collection (ProQuest)
- Anahtar Kelimeler: Infective endocarditis, Machine learning, Unsupervised clustering, Phenotyping, Clinical profiles
- Sağlık Bilimleri Üniversitesi Adresli: Evet
Özet
Purpose: Infective endocarditis (IE) is a heterogeneous disease with diverse clinical presentations, microbiological profiles, and outcomes. Conventional classification systems may not fully capture this complexity. We aimed to identify data-driven clinical phenotypes in IE using an unsupervised machine learning (ML) approach. Methods: In this multicenter retrospective study, 335 patients with definite IE from three hospitals were included. A total of 55 clinical, microbiological, and echocardiographic variables were analyzed. Missing data were imputed using the MissForest algorithm. Clustering was performed using spectral clustering based on a Gower dissimilarity matrix and supported by a Bayesian non-parametric approach. Associations between cluster assignment and clinical outcomes were assessed using regression analyses. Results: Two distinct clusters were identified, reflecting different clinical profiles. Cluster 1 was characterized by native valve endocarditis, a higher burden of comorbidities, inflammatory activation, and larger vegetations. Cluster 2 was characterized by prosthetic valve and device-related IE, higher rates of atrial fibrillation, and larger cardiac dimensions. Cluster assignment was not significantly associated with in-hospital mortality in either unadjusted or adjusted analyses. Increasing age emerged as an independent predictor of in-hospital mortality. In the adjusted model, Cluster 2 was independently associated with longer hospital stay, whereas no significant difference was observed in ICU stay. Conclusion: Unsupervised ML identified two clinically distinct phenotypes in IE, reflecting disease heterogeneity. Although these phenotypes were not associated with major clinical outcomes, they provide a data-driven framework for understanding disease heterogeneity and may contribute to future personalized approaches in the management of IE. Larger prospective studies are needed to determine their clinical applicability.