OBJECTIVES:

According to European Society of Cardiology, cardiovascular disease (CVD) accounts for 3.9 million deaths/year in Europe. In recent years, artificial intelligence (AI)-driven machine learning (ML) has shown promise in CVD risk prediction. This study aimed to summarise the composite predictive ability of AI/ML algorithms to improve CVD risk prediction in focused populations.

METHODS:

Computerised databases (PubMed, EMBASE, and Cochrane) were searched from 2018 to 2022 to identify most recent literature reporting the use of AI/ML in predictive CVD risk analysis. A total of 50 articles in English were selected, focusing on geography and algorithms employed. The area under the receiver operating characteristic curve (AUCROC) was used to quantify the improvement over random chance (AUCROC: 0.5).

RESULTS:

In these studies, a total of 2,620,577 individuals were included and 87.5%, 7.5% and 5.0% of AI/ML models fell under supervised, unsupervised, and semi-supervised learning, respectively. Diabetes and obesity were identified as key CVD risk factors. For prediction of diabetes, AutoPrognosis, logistic regression (LR), cox regression (CR) and gradient boosting (GB) models had a pooled AUCROC of 0.71, 0.82, 0.73 and 0.68, respectively. For prediction of obesity, LR and CR models had a pooled AUCROC of 0.75. For prediction of heart failure (HF) and hypertension as CVD-related indications, LR, GB, and custom-built models had a pooled AUCROC of 0.73, 0.80, and 0.89, respectively. Notably, CVD-related hospitalisation and mortality risk was also accurately predicted by RF and AdaBoost models (AUCROC: 0.83, 0.78), respectively.

CONCLUSIONS:

Our targeted review summarises that AI/ML models may accurately predict CVD risk outcomes in European populations. This can help clinicians make informed decisions regarding early therapeutic intervention, thereby resulting in reduced disease burden. However, more appropriate studies are needed to evaluate other CVD-related risk factors and to also include ML as a part of population-based CVD risk assessment tools/databases.