Clustering of disease trajectories with explainable machine learning: A case study on postoperative delirium phenotypes

The identification of phenotypes within complex diseases is a fundamental component of personalized medicine, which aims to adapt healthcare to individual patient characteristics. Postoperative delirium (POD) is a complex neuropsychiatric condition with significant heterogeneity in its clinical manifestations and underlying pathophysiology. We hypothesize that POD comprises several distinct phenotypes, which cannot be directly observed in clinical practice. Identifying these phenotypes could enhance our understanding of POD pathogenesis and facilitate the development of targeted prevention and treatment strategies. In this paper, we propose an approach that combines supervised machine learning for personalized POD risk prediction with unsupervised clustering technique to uncover potential POD phenotypes. We first demonstrate our approach using synthetic data, where we simulate patient cohorts with predefined phenotypes based on distinct sets of informative features. We aim to mimic any clinical disease with our synthetic data generation method. By training a predictive model and computing SHapley Additive exPlanations (SHAP), we show that clustering patients in the SHAP feature scoring space successfully recovers the true underlying phenotypes, outperforming clustering in the raw feature space. We then present a case study using real-world data from a cohort of elderly POD patients. We train machine learning models on heterogeneous electronic health record data covering the preoperative, intraoperative and postoperative stages to predict personalized POD risk. Subsequent clustering of patients based on their SHAP feature scores reveals distinct subgroups with differing clinical characteristics and risk profiles, potentially representing POD phenotypes. These results showcase the utility of our approach in uncovering clinically relevant subtypes of complex disorders like POD, paving the way for more precise and personalized treatment strategies.Author summary: Our research addresses how we can identify different subtypes within complex medical conditions using advanced data analysis techniques. While traditional medical approaches often treat conditions like postoperative delirium as uniform entities, we demonstrate that they actually comprise distinct subtypes with unique underlying characteristics. We’ve developed a novel two-step approach that first predicts a patient’s risk using machine learning algorithms, then identifies the specific factors driving that risk for each individual. By grouping patients based on these personalized risk factors rather than their raw medical data, we can uncover meaningful subtypes that weren’t previously apparent. Testing this approach with both synthetic and real-world medical data proved remarkably effective. The method successfully identified distinct patient subgroups with different clinical profiles, potentially representing different forms of postoperative delirium with unique underlying mechanisms. These findings have important implications for patient care. Understanding these subtypes could help clinicians develop more targeted prevention strategies and treatments tailored to a patient’s specific condition variant rather than using a one-size-fits-all approach. This work represents a significant step toward more personalized medicine that recognizes the inherent diversity within complex medical conditions, ultimately improving outcomes through more precise interventions.