Researchers built an AI model that predicts mortality risk in patients with complex heart and kidney conditions
A new XGBoost model uses eight clinical variables to accurately group patients with cardiovascular-kidney-metabolic syndrome by their survival odds.
Medical researchers recently published a study in npj Digital Medicine detailing how machine learning can improve outcomes for patients with cardiovascular-kidney-metabolic (CKM) syndrome. This condition affects the heart, kidneys, and metabolism simultaneously, creating complex health risks.
The team analyzed medical records from 219,561 hospitalized patients across 29 centers in China. They trained two models, XGBoost and LASSO, to calculate the likelihood of death within three years.
This parsimonious 8-variable XGBoost model provides accurate mortality risk stratification and may support personalized management of patients with CKM syndrome.
How it works
The XGBoost model proved more effective, relying on just eight specific clinical variables to generate an accurate risk score. In their tests, it performed well across different validation groups, showing a strong ability to distinguish between low, moderate, and high-risk patients.
Patients identified as high-risk by the model showed significantly higher rates of mortality compared to those in the low-risk category. This suggests the tool could help doctors prioritize care for those who need it most.
To make the tool practical, the researchers created a web-based calculator. Clinicians can input patient data to receive a risk assessment during a regular check-up. Could such tools eventually become a standard part of every hospital admission?
However, the study has limits. It was tested primarily on hospitalized patients in China, meaning its accuracy in different global populations or outpatient settings remains unproven. The model also only accounts for 3-year mortality and does not replace the comprehensive judgment of a medical professional.
Source: npj Digital Medicine