AI model to predict mortality of heart attack survivors

NEW DELHI: Every year, nearly 13-14 lakh people in India suffer from a very serious type of heart attack during which one of the heart’s major arteries is blocked. Of them, 8-9% die within 30 days due to complications as well as lack of timely intervention.

This condition is known to cardiologists as ST-Elevation Myocardial Infarction ( STEMI ).

An artificial intelligence or machine-learning model developed by Indraprastha Institute of Information Technology (IIIT)-Delhi in collaboration with cardiologists from Delhi’s G B Pant Hospital and Harvard Medical School, among others, promises to change this.

The model, which is available online, has been developed by Dr Anubha Gupta of IIIT-Delhi. It can predict the risk of all-cause mortality within 30 days of a heart attack by analysing 31 key characteristics of a patient.

These include age, gender, family history of heart disease, treatment history, time taken to reach hospital and the time taken to do angioplasty.

Dr Mohit Gupta, professor of cardiology at G B Pant Hospital, who was the principal investigator of the project, told TOI that the models used currently to predict the mortality risk have been validated mostly in the western population.

“This is the first time an algorithm or model that identifies risk factors that are unique to the Indian population, for example delayed presentation and lack of angioplasty, has been developed,” he said. He added that the results of the study conducted to validate the accuracy of the model have been published in International Journal of Cardiology.

Dr Gupta said G B Pant is running a registry which collates data of all heart attack patients in north India. The model to predict mortality in STEMI cases was developed using the data of 3,191 STEMI patients from the registry.

“At 30 days, the mortality was 7.7%. On the validation dataset, the machine learning model could pick up chances of all-cause mortality in more than 85% cases correctly,” said Dr Gupta.

Dr Ashok Seth , chairman of Fortis Escorts Heart Institute, said the study was a landmark one. “In-hospital mortality rates in STEMI cases are approaching 4-5% in developed nations. The model developed by IIIT-Delhi and G B Pant Hospital may serve as an important tool to reduce mortality rates in India too. This could help in triaging STEMI cases and prioritising intervention, medical therapy or angioplasty, based on the risk score of the patients,” he said.

TIMI (Thrombolysis in Myocardial Infarction) and GRACE (Global Registry of Acute Coronary Events) scores, two of the most popular models used currently for risk stratification in STEMI cases in India, were developed and validated abroad. “In India, a major factor for all-cause mortality in STEMI cases is delayed hospitalisation. Also, in many peripheral hospitals, angioplasty facilities are not available. So, they give only medical therapy and refer the patient. All these factors aren’t considered as strongly in models developed abroad. The indigenous model takes the local factors into account,” say doctors.