The model accurately estimates the risk of major adverse cardiovascular events over a decade by analyzing a single chest radiograph, a new study finds.

Key Points:

  • A deep-learning model, validated on 8,869 outpatients with unknown ASCVD risk, shows 1.5-fold higher accuracy in identifying patients eligible for statin therapy compared to traditional methods.
  • CXR CVD-Risk, utilizing routine chest radiographs, offers potential for population-based screening to pinpoint high-risk individuals for primary prevention of cardiovascular disease.
  • The AI tool’s performance is comparable to traditional ASCVD risk scores in patients with complete data, adding value to current cardiovascular risk assessment protocols.

A risk prediction study of cardiovascular disease prevention efforts found that a deep-learning model better predicts 10-year risk for major adverse cardiovascular events (MACE) beyond the current clinical risk score, even for patients whose score cannot be calculated due to missing data. 

The study is published in Annals of Internal Medicine.

Guidelines from the American College of Cardiology and American Heart Association on the primary prevention of atherosclerotic cardiovascular disease (ASCVD) recommend a risk calculator for nondiabetic adults aged 40 to 75 years with low-density lipoprotein cholesterol levels between 1.81 and 4.91 mmol/L (between 70 and 190 mg/dL) to estimate the 10-year risk for ASCVD as a guide to pharmacologic and other primary prevention. 

However, because the necessary input variables to calculate the ASCVD risk score are often not available in the electronic medical record, other approaches for population-based screening are desirable to identify individuals at high risk who are likely to benefit from a statin.

Researchers from Massachusetts General Hospital and Harvard Medical School conducted a risk prediction study of a deep-learning model (CXR CVD-Risk) that estimates a 10-year risk for MACE from a routine chest radiograph (CXR). The model was validated using data from 8,869 outpatients with unknown ASCVD risk because of missing inputs to calculate the ASCVD risk score and in 2,132 outpatients with known risk whose ASCVD risk score could be calculated. 

The authors found that, for the 81% of patients who had unknown ASCVD risk due to missing inputs to calculate the traditional ASCVD risk score, 10-year risk for incident MACE was 1.5-fold higher for persons identified as statin-eligible by CXR CVD-Risk than for those classified as ineligible, independent of available baseline cardiovascular risk factors. 

They also found that for the 19% of patients who had all necessary inputs available to calculate the traditional, guideline-recommended ASCVD risk score, CXR CVD-Risk had similar performance and additive value to the traditional risk score. 

According to the authors, their findings suggest that CXR CVD-Risk could enable population-based opportunistic screening using routine CXRs to identify persons at high risk who would benefit from primary ASCVD prevention with statins but are currently unrecognized.

Photo 170690137 © Алеся Фесенко | Dreamstime.com