Weight Changes and All-Cause Mortality in Critically Ill Patients: A Multi-Center Retrospective Cohort Study

Why It Matters

Dynamic weight monitoring offers clinicians a simple, real‑time prognostic tool that can guide fluid and nutrition strategies, potentially reducing mortality in critically ill patients.

Key Takeaways

• Weight‑change rate independently predicts ICU mortality (OR 1.04 per 1%).
• Hospital mortality rises similarly with each percent of weight gain.
• Risk curve shows a threshold effect near 5 % weight change.
• Weight‑change outperforms admission weight, discharge weight, and BMI in AUC.
• Study used Boruta algorithm on >30,000 eICU patients.

Pulse Analysis

In the intensive‑care setting, rapid shifts in body weight reflect a tug‑of‑war between fluid resuscitation, capillary leak, and catabolic muscle loss. While clinicians have long recognized fluid overload and severe wasting as mortality drivers, quantifying these processes at the bedside has remained elusive. Weight‑change rate captures both phenomena in a single, easily obtainable metric, offering a composite view of a patient’s evolving physiological stress.

The recent multi‑center study leveraged the eICU Collaborative Research Database, encompassing over 30,000 adult admissions from more than 200 U.S. ICUs. By applying the Boruta machine‑learning algorithm, researchers isolated weight‑change rate as a top predictor among hundreds of variables. Logistic models, adjusted for demographics, severity scores, comorbidities, and interventions, demonstrated a consistent, dose‑responsive rise in mortality risk with each percent of weight gain, and identified a non‑linear inflection point near 5 %—a threshold where prognostic impact accelerates.

These findings suggest that routine, serial weighing could be integrated into ICU electronic health‑record dashboards to flag patients crossing the 5 % threshold, prompting early reassessment of fluid balance, diuretic use, and nutritional support. Moreover, the superior discriminative ability of weight‑change rate over static indices supports its inclusion in risk‑adjusted quality metrics. Future research should explore whether targeted interventions based on real‑time weight trends can translate the observed prognostic signal into tangible survival benefits.