Associations of weight-derived markers with mortality in patients with Corona virus disease 2019: evidence from hospitals and patients.

BMI consistently demonstrated a protective effect against in-hospital mortality across all three analytical models.

• Crude model: OR (95% CI) = 0.84 (0.77-0.92), P < 0.05
• Adjusted model 1: OR (95% CI) = 0.84 (0.77-0.92), P < 0.05
• Propensity Score Matching (PSM) model: OR (95% CI) = 0.85 (0.74-0.97), P < 0.05
• BMI was the only weight-derived marker to consistently demonstrate a protective effect among all markers examined

Restricted Cubic Spline regression revealed significant nonlinear associations between BMI, Weight, LBM, and WBFM with in-hospital mortality.

• P for overall < 0.05 for BMI, Weight, Lean Body Mass (LBM), and Whole-Body Fat Mass (WBFM)
• No significant nonlinear associations were observed between Body Fat Percentage (BFP) and in-hospital mortality
• No significant nonlinear associations were observed between Basal Metabolic Rate (BMR) and in-hospital mortality

The BMI-based Random Forest model effectively forecasted in-hospital mortality in COVID-19 patients.

• ROC (95% CI) = 0.84 (0.76-0.92) for the BMI-based Random Forest model
• Four machine-learning models were developed: Decision Tree Classifier, Random Forest, Gaussian Naive Bayes, and Gradient Boosting Classifier
• The Random Forest model demonstrated the strongest predictive capability among the models assessed

A total of 509 COVID-19 patients were included, with body composition markers calculated from height, weight, gender, and age.

• Sample size: 509 patients with COVID-19
• Weight-derived markers examined included Weight, BMI, Body Fat Percentage (BFP), Whole-Body Fat Mass (WBFM), Lean Body Mass (LBM), and Basal Metabolic Rate (BMR)
• All body composition variables were derived from height, weight, gender, and age
• In-hospital mortality served as the primary clinical outcome
• Associations were evaluated using a crude model, a logistic model adjusted for confounders, and a PSM model