Machine Learning Outperforms ACC / AHA CVD Risk Calculator in MESA

Abstract

Background Studies have demonstrated that the current US guidelines based on American College of Cardiology/American Heart Association (ACC/AHA) Pooled Cohort Equations Risk Calculator may underestimate risk of atherosclerotic cardiovascular disease ( CVD ) in certain high-risk individuals, therefore missing opportunities for intensive therapy and preventing CVD events. Similarly, the guidelines may overestimate risk in low risk populations resulting in unnecessary statin therapy. We used Machine Learning ( ML ) to tackle this problem. Methods and Results We developed a ML Risk Calculator based on Support Vector Machines ( SVM s) using a 13-year follow up data set from MESA (the Multi-Ethnic Study of Atherosclerosis) of 6459 participants who were atherosclerotic CVD-free at baseline. We provided identical input to both risk calculators and compared their performance. We then used the FLEMENGHO study (the Flemish Study of Environment, Genes and Health Outcomes) to validate the model in an external cohort. ACC / AHA Risk Calculator, based on 7.5% 10-year risk threshold, recommended statin to 46.0%. Despite this high proportion, 23.8% of the 480 "Hard CVD " events occurred in those not recommended statin, resulting in sensitivity 0.76, specificity 0.56, and AUC 0.71. In contrast, ML Risk Calculator recommended only 11.4% to take statin, and only 14.4% of "Hard CVD " events occurred in those not recommended statin, resulting in sensitivity 0.86, specificity 0.95, and AUC 0.92. Similar results were found for prediction of "All CVD " events. Conclusions The ML Risk Calculator outperformed the ACC/AHA Risk Calculator by recommending less drug therapy, yet missing fewer events. Additional studies are underway to validate the ML model in other cohorts and to explore its ability in short-term CVD risk prediction.

Keywords: Artificial intelligence; Machine learning; atherosclerosis; cardiovascular disease prevention; cardiovascular disease risk factors; cardiovascular risk; clinical decision support; prediction statistics; statin.

Figure 1

Overview of ML approach. For each ML model, we divided the study population 50/50 into training and prediction subset cohorts. Next, we augmented the training subset using NEATER and trained the SVM prediction model. During prediction, each sample in the prediction cohort was analyzed and classified. Then, the cohorts switched places (ie, prediction becomes training, and vice versa) and the process was repeated. The overall iterative process was repeated 10 times for each ML model, and the results were averaged. CVD indicates cardiovascular disease; FLEMENGHO study, the Flemish Study of Environment, Genes and Health Outcomes; HDL, high‐density lipoprotein; MESA, the Multi‐Ethnic Study of Atherosclerosis; NEATER, a method for the filtering of oversampled data using non‐cooperative game theory; SVM, Support Vector Machine.

Figure 2

Receiver operating characteristic (ROC) curves for prediction of (A) “Hard CVD” events and (B) “All CVD” events comparing the ML Risk Calculator (blue) with the American College of Cardiology/American Heart Association (ACC/AHA) Risk Calculator (red). AUC indicates area under the curve; CVD, cardiovascular disease; ML, Machine Learning.

Figure 3

Risk calculator comparison: statin eligibility and missed treatment opportunities. Pie graphs illustrate the performance comparison between ML Risk Calculator and ACC/AHA Risk Calculator for predicting “Hard CVD” events (left) and “All CVD” events (right). ACC/AHA indicates American College of Cardiology/American Heart Association; CBG, coronary bypass grafts; CHD, coronary heart disease; CHF, congestive heart failures; CVD, cardiovascular disease; MI, myocardial infarction; PTCA, percutaneous transluminal coronary angioplasties; PVD, peripheral vascular diseases; TIA, transient ischemic attacks.