Problem / Objective:

Nearly 3 million heart attacks happen in India every year, and there are 30 million people who suffer from coronary diseases. Despite the sheer number of cases, doctors in India are unable to identify the probability of cardiac illnesses when patients come for their regular health check-ups. 

There are some scores or algorithms available worldwide that predict the probability of a patient having a heart attack in the next 10 to 20 years. But doctors cannot generalize the same risk-factors and apply those to patients in India because most of those scores are derived from western studies and do not have a high degree of accuracy when it comes to the Indian population.

Solution / Approach:

Apollo Hospitals partnered with Microsoft's AI Network for Healthcare to develop an India-specific heart risk score and predict cardiac diseases for the general population with the help of Apollo's database and expertise in the field, and Microsoft's cloud and AI tools.

The team at Apollo Hospitals collected more than seven years of data from master health check-ups conducted at its hospitals across India from 2010-2017, which consisted of clinical and lab data of 400,000 patients. The hospital used Microsoft Azure to upload the data to the cloud and some of the SQL tools to put it into the data warehouse from where data scientists and clinicians could statistically correlate and train the machine learning models. They stripped off all identifiable patient information from the data sets. They also ensured the processes were in line with Indian IT laws as well as HIPAA and ISO 20071 guidelines for data privacy and security before the data was shared.

The team fed the data of these patients into the machine learning model to find out the risk factors that caused these patients to have a heart attack despite having a routine health check-up. The data scientists could identify that there were different risk-factors, which current models were either not considering or not giving enough weight in building the risk-factor score. They worked with Azure Machine Learning services, which is an end-to-end service that helps accelerate the rate of Machine Learning experimentation by rapid prototyping, scaling up on virtual machines and proactively managing model performance. The Azure services include Libraries for Apache Spark and Visual Studio Code Tools for AI.

Impact / Implementation:

The team started with 100 health check-up risk-factors and 200 lab data points and correlated each factor concerning its significance to the occurrence of the disease. Ultimately, they narrowed down to 21 risk factors to build a model to predict heart risk for the Indian population. A team of data scientists and clinicians working on the data found that nearly 60,000 patients had been admitted to a hospital for a cardiac event after undergoing health check-ups. Of these, approximately 34,000 patients had two normal check-ups in a span of five years and yet had a cardiac event. Over 7,000 others had a cardiac event preceding one, two or three years of a normal health check-up.

The accuracy of the model for the Indian population in predicting the probability of coronary disease in the future came out to be twice the existing ones. With the new heart risk score for India, Apollo Hospitals is looking at redefining how preventive health check-ups are done across its hospitals. The models help gauge a patient's risk for heart disease and provide rich insights to doctors on treatment plans, assist early diagnosis and empower doctors with predictive solutions.

The team is already working on an AI-powered Cardio API platform, which would let anyone come to the hospital and able to find their heart risk score without a detailed health check-up.