Machine Learning May Improve Myocardial Recovery Prediction on LVAD Support


January 09, 2022

2 minutes to read

Disclosures: Topkara reports receiving research support from NIH. Please see the study for relevant financial information from all other authors.

We have not been able to process your request. Please try again later. If you continue to experience this problem, please contact [email protected].

According to new data in Circulation: heart failure.

“Machine learning has become a powerful tool for learning complex data models and identifying nonlinear interactions within large clinical data sets to formulate disease classifications as well as to predict patient outcomes.” Veli K. Topkara, MD, M.Sc., an expert in advanced heart failure and transplant cardiology at Columbia University Irving Medical Center, and colleagues wrote. “Using the INTERMACS multi-center registry, we hypothesized that machine learning-based prediction models can improve the selection of candidates likely to recover on LVAD media over traditional regression-based models and discover new results. new clinical risk factors that are positively or negatively associated with the myocardium. recovery on LVAD media.

Heart Matrix_Adobe Stock
Source: Adobe Stock

Train machine learning models

To this end, the researchers trained the machine learning models to predict myocardial recovery after LVAD implantation using 28 unique clinical features. Half of these features were positively associated with myocardial recovery, including bridge-to-recovery implantation strategy, current tobacco abuse, postpartum cardiomyopathy, recent cardiac diagnosis, history of drug use. alcohol, limited social support, nonadherence, elevated hemoglobin, sinus rhythm, elevated heart elevated serum sodium, INTERMACS profile 3, and psychiatric illness. The other half was negatively associated with myocardial recovery and included the presence of an implantable cardioverting defibrillator, reduced post-implant ejection fraction, implantation of a right ventricular assist device with LVAD, old age, increased blood urea nitrogen, ischemic cardiomyopathy, use of a centrifugal pump, post-implant increase LV end-diastolic diameter, bypass with intra-aortic balloon pump, preimplant warfarin use, preimplant use of amiodarone, indication for bridge grafting and concomitant tricuspid valve surgery.

Researchers included 20,270 adult patients from the INTERMACS registry with sustained continuous flow LVAD.

Machine learning to predict myocardial recovery

According to the study, machine learning models reached an area under the curve of 0.813 to 0.824 and outperformed the new INTERMACS recovery risk score based on logistic regression (AUC = 0.796), as well as risk scores. previously established LVAD recovery, INTERMACS Cardiac Recover Score (AUC = 0.744) and INTERMACS recovery score of Topkara and colleagues (AUC = 0.748; P for all <.05>

“The low positive predictive value of machine learning models is not unexpected given the low incidence of myocardial recovery in patients with LVAD,” the researchers wrote. “The inclusion of additional features that are not available in INTERMACS, such as pharmacological therapy, drug doses used, optimization of pump speed, and declining echocardiographic or hemodynamic indices, could improve the ability discriminant of models. “

Researchers observed a higher incidence of myocardial recovery resulting in LVAD explantation in patients identified by machine learning models as having myocardial recovery compared to patients whose recovery was not predicted by learning models automatic (18.8% versus 2.6%).

“Using and interpreting complex machine learning models could be difficult for the clinician compared to simple logistic regression models. However, machine learning algorithms have been shown to be superior to traditional risk models in predicting the risk of readmission and heart failure mortality, ”the researchers wrote. Likewise, the current analysis suggested the superior discriminating ability of machine learning models with respect to the prediction of recovery in patients managed by LVAD.


Comments are closed.