Explainable Machine Learning Analysis of Right Heart Failure After Left Ventricular Assist Device Implantation

Abstract

Right heart failure (RHF) remains a common and serious complication after durable left ventricular assist device (LVAD) implantation. We used explainable machine learning (ML) methods to derive novel insights into preimplant patient factors associated with RHF. Continuous-flow LVAD implantations from 2008 to 2017 in the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) were included. A total of 186 preimplant patient factors were analyzed and the primary outcome was 30 days of severe RHF. A boosted decision tree ML algorithm and an explainable ML method were applied to identify the most important factors associated with RHF, nonlinear relationships and interactions, and risk inflection points. Out of 19,595 patients, 19.1% developed severe RHF at 30 days. Thirty top predictors of RHF were identified with the top five being INTERMACS profile, Model for End-stage Liver Disease score, the number of inotropic infusions, hemoglobin, and race. Many top factors exhibited nonlinear relationships with key risk inflection points such as INTERMACS profile between 2 and 3, right atrial pressure of 15 mmHg, pulmonary artery pressure index of 3, and prealbumin of 23 mg/dl. Finally, the most important variable interactions involved INTERMACS profile and the number of inotropes. These insights could help formulate patient optimization strategies prior to LVAD implantation.

Improving the Prediction of 1-Year Right Ventricular Failure After Left Ventricular Assist Device Implantation

“…2 This study considered four strategies: 1) splitting the data into training and test groups, 2) using repeated cross-validation for training the model, 3) employing big data that includes approximately 12,000 patients, 4) applying XGBoost which has been effectively used in the healthcare field for disease diagnosis and risk prediction. 15,16 The reported, well-calibrated XGBoost model achieved a high AUC-ROC (C statistics) of 0.80 compared to AUC of 0.53–0.65 reported in a recent (2020) external validation study of 20 RHF models by Frankfurter et al 2 and a very recently published model of early RHF by Bahl et al 17 (AUC = 0.67). The test dataset in the present model comprised 2,393 patients that were randomly chosen from a large national registry from more than 180 hospitals.…”

The Role of Artificial Intelligence and Machine Learning in the Prediction of Right Heart Failure after Left Ventricular Assist Device Implantation: A Comprehensive Review

“…For example, Loghmanpour et al found that the systolic PAP, pre-albumin, LDH, and RV EF were most important predictive parameters, and they achieved an impressive AUC rate in their study [57]. Similarly, Kilic et al and Bahl et al identified more predictive parameters and non-linear relationships between preoperative clinical parameters, which is one of the main strengths of AI-based investigations as it is very difficult to infer these relationships through statistical analysis [61,62,64].…”

“…The study that was designed by Bahl et al was one of the newest studies that focused on ML and RHF [64]. They preferred an "explainable" ML method called boosted decision trees in order to analyze the preimplant patient factors in nonlinear interactions with RHF after LVAD implantation.…”

Association between pulmonary artery pulsatility and mortality after implantation of left ventricular assist device

“…The majority of our study population had, therefore, an INTERMACS profile ranging from I to III, a condition at high risk of post-operative morbidity and mortality. 38,39 We found a threshold value of PAPi at 2.84 to predict the risk of early death within 3 months after surgery. This was the median value of PAPi in our study population.…”