Future promise, current clinical ambiguity: a systematic review of machine learning algorithm outputs predicting risk of cardiovascular disease.

In 30 seconds

This systematic review evaluated machine learning algorithms predicting cardiovascular disease (CVD) risk compared to the Framingham Risk Score (FRS) in healthy adults. Out of 29 studies analyzed, 23 showed improved predictive ability over the FRS. The review highlights the potential of integrating sociodemographic factors and biological variables, but emphasizes the need for clinical insight and cost-benefit assessments.

Key findings

• 23 out of 29 studies reported improved predictive ability compared to the Framingham Risk Score.
• Many algorithms included sociodemographic predictors not present in the FRS.
• Some algorithms added costly diagnostic tests like CT angiography to risk assessment.
• Statistical significance was often used as a proxy for clinical significance.

Why it matters

Understanding the effectiveness of machine learning algorithms in predicting CVD risk is crucial for clinicians, as it may enhance risk assessment strategies and patient management. However, careful consideration of clinical relevance and cost implications is necessary.