Machine Learning Multicenter Risk Model to Predict Right Ventricular Failure After Mechanical Circulatory Support

• DOI: 10.1001/jamacardio.2023.5372
• Publication Date: 2024-01-31T16:33:18Z
• Authors: Iosif Taleb; Christos P. Kyria...; Robyn Fong; Naila Ijaz; Zachary Demertzis; Konstantinos Sideris; Omar Wever-Pinzon; Antigone G. Kolio...; Michael J. Bonios; Rohan Shad; Adithya Peruri; Thomas C. Hanff; Elizabeth Dranow; Theodoros V. Gian...; Ethan Krauspe; Cyril Zakka; Daniel G. Tang; Hassan W. Nemeh; Josef Stehlik; James C. Fang; Craig H. Selzman; Rami Alharethi; William T. Caine; Jennifer A. Cowger; William Hiesinger; Palak Shah; Stavros G. Drakos
• Keywords: 03 medical and health sciences; 0302 clinical medicine

Importance The existing models predicting right ventricular failure (RVF) after durable left assist device (LVAD) support might be limited, partly due to lack of external validation, marginal predictive power, and absence intraoperative characteristics. Objective To derive validate a risk model predict RVF LVAD implantation. Design, Setting, Participants This was hybrid prospective-retrospective multicenter cohort study conducted from April 2008 July 2019 patients with advanced heart (HF) requiring continuous-flow LVAD. derivation included enrolled at 5 institutions. validation sixth institution within the same period. Study data were analyzed October 2022 August 2023. Exposures participants underwent chronic support. Main Outcome Measures primary outcome incidence, defined as need for RV or intravenous inotropes greater than 14 days. Bootstrap imputation adaptive least absolute shrinkage selection operator variable techniques used model. An calculator (STOP-RVF) then developed subsequently externally validated, which can provide personalized quantification candidates. Its accuracy compared previously published scores. Results 798 (mean [SE] age, 56.1 [13.2] years; 668 male [83.7%]). 327 patients. in 193 (24.2%) 107 (32.7%) cohort. Preimplant variables associated postoperative nonischemic cardiomyopathy, intra-aortic balloon pump, microaxial percutaneous device/venoarterial extracorporeal membrane oxygenation, configuration, Interagency Registry Mechanically Assisted Circulatory Support profiles 1 2, atrial/pulmonary capillary wedge pressure ratio, use angiotensin-converting enzyme inhibitors, platelet count, serum sodium, albumin, creatinine levels. Inclusion characteristics did not improve performance. achieved C statistic 0.75 (95% CI, 0.71-0.79) 0.73 0.67-0.80) Cumulative survival higher composing low-risk group (estimated <20% risk) those higher-risk groups. STOP-RVF exhibited significantly better performance commonly scores proposed by Kormos et al (C statistic, 0.58; 95% 0.53-0.63) Drakos 0.62; 0.57-0.67). Conclusions Relevance Implementing routine clinical data, this derived validated assessment tool prediction RVF-associated all-cause mortality.