Artificial intelligence showed promise in predicting sudden cardiac deaths by analyzing past medical records, according to researchers in France.
The AI was fed electronic medical records and databases from Paris, France and Seattle,
Washington including almost 13,000 people from 2011-2015 and about 11,000 people from 2016-2020 who died from sudden cardiac arrest. These cases were matched by about 70,000 people from the general population, equalizing the data across age, sex and residential area to serve as a control.
The data was recorded as far back as ten years prior to the sudden cardiac deaths, providing a solid basis by which the AI could form its predictions.
The researchers created almost 24,000 equations for analyzing risk factors to identify individuals with a “very high risk of sudden cardiac arrest.” The equations used predictive factors like hypertension medications, history of heart disease and mental and behavioral disorders, like alcohol abuse.
From this, the AI was able to develop an accurate model to identify people relatively accurately with a 90% risk of sudden cardiac death. Overall, the models built by the AI “achieved an area under the curve (AUC) of 0.80, a positive predictive value of 77% and a sensitivity of 68%,” a uniquely promising result.
“We have been working for almost 30 years in the field of sudden cardiac death prediction; however, we did not expect to reach such a high level of accuracy,” Xavier Jouven, M.D., Ph.D., the lead author of the study and Professor of Cardiology and Epidemiology at the Paris Cardiovascular Research Center said in a statement.
Jouven also highlighted one of the key areas where AI models can offer predictive value beyond what cardiologists may have the range to consider.
“We also discovered that the personalized risk factors are very different between the participants and are often issued from different medical fields (a mix of neurological, psychiatric, metabolic and cardiovascular data) – a picture difficult to catch for the medical eyes and brain of a specialist in one given field,” Jouven said.
In other words, cardiologists trained for a specific purpose may miss key details because they are focused solely on the cardiac results, without considering the other risk factors that arise from other aspects of the patient’s life.
Jouven emphasized that he believed that the integration of AI in medical decisions is necessary to detect and analyze past medical information in a short period of time in a way that can streamline and inform decisions without succumbing to biases that the cardiologist may have.
The study’s limitations included its reliance on past medical data, a concern with any predictive model.
While medicine is more resistant to breakneck changes compared to fields such as technology, the landscape of medicine is ever-changing and predictive models may lag new trends.
One pertinent aspect that may not be properly considered yet is the impact of COVID-19 on heart health. The research focused on years 2011-2020, revealing a gap that must be at the very least acknowledged.
The research was presented at the Resuscitation Science Symposium, an international meeting “showcasing the most recent advances in resuscitation science and practices” held by the American Heart Association on Nov. 11-12.
