Admission body mass index categories and short-term mortality in critically ill adults: a retrospective cohort study using Asian-oriented cut points

Admission body mass index (BMI) is routinely available in intensive care, but BMI-mortality associations are difficult to interpret because BMI may reflect adiposity, nutritional reserve, sarcopenia, fluid status, illness severity, and selection processes. We evaluated admission BMI category in relation to in-hospital mortality and intensive care unit (ICU) mortality in critically ill adults using Asian-oriented BMI cut points. We conducted a single-center retrospective cohort study of adult ICU admissions from 2018 to 2020. BMI was categorized using Asian-oriented cut points as < 18.5, 18.5–23.9, 24.0-26.9, and > = 27.0 kg/m². The primary outcome was in-hospital mortality; ICU mortality was secondary. Multivariable logistic regression with patient-level cluster-robust standard errors adjusted for age, sex, APACHE II score, ICU type, admission source, and selected comorbidities. A conceptual directed acyclic graph was added to clarify the assumed covariate structure and to guide interpretation of the adjusted associations. Among 16,692 ICU admissions, BMI was available for 11,923/16,692 (71.4%). Adjusted analyses included 11,450/11,923 admissions (96.0%) for in-hospital mortality and 11,379/11,923 admissions (95.4%) for ICU mortality. In-hospital mortality was 22.8% (314/1,376), 15.8% (747/4,729), 12.5% (352/2,806), and 9.4% (283/3,012) across increasing BMI categories. Compared with normal BMI, adjusted odds ratios for in-hospital mortality were 1.31 (95% confidence interval [CI] 1.07–1.62; p = 0.011) for underweight, 0.79 (95% CI 0.66–0.95; p = 0.012) for overweight, and 0.61 (95% CI 0.50–0.74; p < 0.001) for obesity. Obesity was also associated with lower ICU mortality (adjusted odds ratio 0.72, 95% CI 0.58–0.90; p = 0.004). Admission BMI category was associated with short-term mortality in this cohort. The findings should be interpreted as adjusted associations rather than evidence that higher BMI is causally protective. The substantial proportion of missing BMI data, uncertainty around anthropometric measurement, and residual confounding limit causal interpretation.