Artificial intelligence may assist India in detecting individuals at risk of cardiovascular disease earlier and support more tailored prevention strategies, yet the technology remains unprepared to direct standard clinical choices, according to a systematic review by researchers from the Indian Institute of Science, M.S. Ramaiah University of Applied Sciences, and the London School of Hygiene and Tropical Medicine.
The analysis, published in BMC Medical Informatics and Decision Making, examined 30 studies from 2017 onward that created AI models to forecast cardiovascular disease in adults without prior heart conditions.
The studies were located through a systematic search of over 6,700 records in major scientific databases. Most models used data from the United States, United Kingdom, and South Korea, drawing on routine clinical records and machine learning methods including Random Forests, Support Vector Machines, and neural networks.
Relevance for India
The results hold special importance for India, where cardiovascular disease causes nearly one-third of deaths and often strikes at younger ages than in many other nations.
Denny John, from M.S. Ramaiah University of Applied Sciences, noted that AI could refine cardiovascular risk assessment but that current evidence falls short for broad clinical application.
“AI offers an opportunity to make cardiovascular risk prediction more precise and more personalised. But our review shows that the evidence is still incomplete,” he said.
Several Indian institutions have created AI models that include local factors such as smokeless tobacco use, psychosocial stress, and physical inactivity. However, John stated these tools require thorough independent validation before use in primary care or public health programs.
“Many models demonstrate good discrimination, but very few studies examine whether the predicted risks correspond to what actually happens in different populations,” he said.
He added that strong external validation, calibration, and evaluation of clinical usefulness are essential before AI tools can inform long-term treatment choices such as starting blood pressure or cholesterol-lowering therapies.
Comparable with conventional risk scores
Twelve studies directly compared AI models with established cardiovascular risk calculators, including the Framingham Risk Score, which estimates 10-year risk of events such as heart attack or stroke.
The review showed AI models performed as well as, and sometimes slightly better than, conventional tools in identifying higher-risk individuals over five to 10 years.
Yet the researchers warned that improved statistical performance alone does not confirm better patient care.
Major validation gaps
A main concern was the absence of evidence that predicted risks matched real-world outcomes.
Nearly all studies measured discrimination but none assessed calibration. Only seven studies validated models on independent groups. Sensitivity was reported in just four studies, and none performed decision-curve analyses.
For India, this gap matters because models from the United States, United Kingdom, or South Korea may not apply equally to Indian populations with different risk factors, John said.
Need for independent testing
The researchers urged prospective validation of AI cardiovascular risk models across varied populations, including in India, before routine healthcare integration.
They also recommended that future studies follow frameworks such as TRIPOD+AI and PROBAST+AI to enhance transparency and reduce bias.
