Heterogeneous effects of obesity on all‐cause mortality: A causal forest analysis of hospitalized patients in China

To confirm the 'obesity paradox' in hospitalized populations and examine the heterogeneous effects of obesity on all-cause mortality risk. We included 5967 hospitalized patients from the Real-world Data of Cardiometabolic Protection (RED-CARPET) study (ChiCTR2000039901) in China. After 1:1 k-nearest neighbours matching, a causal forest model classified the population into four subgroups. Cox models were used to assess the association between obesity and all-cause mortality, with external validation in the Atherosclerosis Risk in Communities (ARIC) cohort. During a median follow-up of 63.8 months, 919 (15.4%) deaths occurred. A U-shaped association between body mass index (BMI) and all-cause mortality was observed, illustrating the 'obesity paradox', with the highest mortality rate (18.5%) observed in the normal weight group. Furthermore, 911 participants with obesity and 911 participants with normal weight, matched for homogeneity, were categorized into four subgroups using the causal forest model. In subgroup 3 (with good renal function, well-controlled blood glucose and favourable nutritional status), patients with obesity had a higher risk of all-cause mortality compared to those with normal weight (HR, 2.12; 95% CI, 1.06-4.22; p = 0.033). No significant association was observed in the other subgroups (p > 0.05). Similar results were verified in the ARIC study cohort. The association between obesity and all-cause mortality is heterogeneous, as individuals with good renal function, well-controlled blood glucose and favourable nutritional status may experience a higher mortality risk. These findings emphasize the need for personalized management strategies in clinical practice to address the varying effects of obesity across different health conditions.