Nutritional indices predict mortality risk in elderly patients in CCU/CVICU

Among elderly patients in the cardiac care unit (CCU) or cardiovascular intensive care unit (CVICU), nutritional indices appear predictive of mortality risk, suggests a study.

A total of 811 elderly patients in the Medical Information Mart for Intensive Care IV database provided data for this retrospective observational study. All-cause mortality at 360 days was the primary outcome.

The investigators evaluated the prognostic nutritional index (PNI) and geriatric nutritional risk index (GNRI) using restricted cubic spline (RCS), Cox proportional hazards model, Kaplan‒Meier curve, and subgroup analyses.

Machine learning (ML) algorithms were also used to develop predictive models, and the Shapley Additive Explanation algorithm was applied to assess the predictive values of feature variables.

In RCS and Cox models, higher nutritional indices correlated with a reduced risk of mortality, while Kaplan‒Meier curves showed a higher risk of death among patients with lower indices.

Among the four ML algorithms constructed, the logistic regression algorithm showed superior performance relative to the others, identifying PNI and GNRI as the key predictors of mortality. In subgroup analyses, PNI and GNRI showed consistent effects across congestive heart failure, myocardial infarction, acute kidney injury, and type 2 diabetes mellitus, with no significant interactions.

“The PNI and GNRI are reliable alternative biomarkers of nutrition,” the investigators said. “However, the relationship between nutritional indicators and mortality of elderly patients in the CCU and CVICU remains unknown.”