Identification of clinically relevant patient endotypes in traumatic brain injury using latent class analysis

Abstract

Abstract Traumatic brain injury (TBI) is a complex condition where heterogeneity impedes the advancement of care. Understanding the diverse presentations of TBI is crucial for personalized medicine. Our study aimed to identify clinically relevant patient endotypes in TBI using latent class analysis based on comorbidity data. We used the Medical Information Mart for Intensive Care III database, which includes 2,629 adult TBI patients. We identified five stable endotypes characterized by specific comorbidity profiles: Heart Failure and Arrhythmia, Healthy, Renal Failure with Hypertension, Alcohol Abuse, and Hypertension. Each endotype had distinct clinical characteristics and outcomes: The Heart Failure and Arrhythmia endotype had lower survival rates than the Renal Failure with Hypertension despite featuring fewer comorbidities overall. Patients in the Hypertension endotype had higher rates of neurosurgical intervention but shorter stays in contrast to the Alcohol Abuse endotype which had lower rates of neurosurgical intervention but significantly longer hospital stays. Both endotypes had high overall survival rates comparable to the Healthy endotype. Logistic regression models showed that endotypes improved the predictability of survival compared to individual comorbidities alone. This study validates clinical endotypes as an approach to addressing heterogeneity in TBI and demonstrates the potential of this methodology in other complex conditions.