Abstract 16867: Predictors of Long-Term Mortality in Chronic Limb Threatening Ischemia Derived From a Random Forest Machine-Learning Model

Abstract

Introduction: Optimizing limb survival in patients with chronic limb-threatening ischemia (CLTI) undergoing peripheral vascular intervention (PVI) is a major focus. Care plans that consider patients’ high-mortality risk are less common. Understanding which patients have the highest mortality risk would enable more personalized care plan designs to enable person-centered care. We aimed to leverage a random forest machine learning (RFML) model to predict long-term mortality in patients with CLTI undergoing PVI. Methods: A cohort of patients with CLTI undergoing PVI from 2017 to 2019 from the Medicare linked Vascular Quality Initiative registry was derived. Using 89 pre-procedural variables, we constructed a RFML model predicting 3-year mortality. Missing data were addressed by random forest imputation. The data were divided into a training sample (80%) and a test sample (20%) using random sampling without replacement comprising. Variables that most frequently caused branch splitting closest to the tree′s origin of were defined as the most predictive factors, with a total of 1,000 trees being built. Accuracy was assessed using the Continuous Ranked Probability Score (CRPS) and Harrell’s C-index. Results: A total of 10,114 patients were included (mean age 72 ± 11 years, 59% male, 74% white). The 3-year mortality rate was 39.1% (Median survival: IQR: ), Among the 15 most predictive variables, 14 variables are potentially modifiable, Remarkably, 5 functional or psychiatric variables were among the most predictive factors. CRPS and Harrerl′s C-index were 0.172 and 0.70, respectively. Conclusion: Our RFML model successfully predicted 3-year all-cause mortality in patients with CLTI undergoing PVI. Predictors of mortality in CLTI extend beyond medical history and encompass functional and behavioral variables. A shared decision-making individualized plan focusing on patient-centered care is needed to improve long-term mortality in patients with CLTI.