AI may predict patients' risk of premature death, study finds

  • Small
  • Medium
  • Large

U.K.-based researchers developed an artificial intelligence-powered system to calculate the risk of early death among middle-aged chronic disease patients, according to a study published in PLOS One.

Five notes:

1. Researchers analyzed data from 502,628 participants who were recruited to the U.K. Biobank, the country's biobank study on genetic predisposition and environmental exposure's effect on disease development, from 2006-10. Participants were aged 40-69, and researchers followed up with them until 2016.

2. Participants were assessed on various factors, including demographic, biometric, clinical and lifestyle. Participants also underwent physical assessments and had biometric, blood and saliva samples analyzed.

3. Researchers made predictions to premature mortality data from the cohort using Office for National Statistics death records, the U.K. cancer registry and hospital statistics.

4. Once the data was collected, researchers developed deep learning and random forest, a supervised learning algorithm, machine learning models to predict participants' premature mortality rates.

The team also used the Cox regression model, which predicts outcomes based on age and gender, as a measure of comparison.

5. Results of the study showed the AI models random forest and deep learning, respectively predicted participants' premature mortality rates 9.4 percent and 10.1 percent more effectively than the standard Cox regression model.

Study authors concluded that AI and machine learning algorithms can improve prediction accuracy of premature mortality amongst middle-aged patient population, and the research needs to be further explored in other populations.

To access the full report, click here.

Copyright © 2021 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.


Featured Whitepapers

Featured Webinars