# Association between serum osmolality trajectories and mortality in sepsis patients: a retrospective multi-cohort study

## TL;DR

Changes in blood osmolality over time in sepsis patients predict mortality, suggesting personalized fluid therapy could improve outcomes.

## Contribution

Dynamic serum osmolality trajectories are shown to predict mortality and inform personalized fluid management in sepsis.

## Key findings

- Three serum osmolality trajectories (stable, ascending, descending) were linked to higher 28-day mortality in sepsis patients.
- Renal dysfunction and fluid imbalance partially explain the mortality risk associated with abnormal osmolality trajectories.
- Fluid management strategies should be tailored to specific trajectory patterns for optimal sepsis care.

## Abstract

Sepsis commonly leads to fluid and electrolyte imbalances, often reflected in abnormal serum osmolality. Although static measurements of osmolality have been investigated, the prognostic significance of dynamic changes in serum osmolality over time remains poorly understood in patients with sepsis.

We conducted a comprehensive analysis of 22,737 septic patients from the MIMIC-IV and eICU-CRD databases, with external validation performed using an independent cohort of 303 patients from Zhejiang Provincial People's Hospital. Latent class trajectory modeling was applied to identify distinct patterns of serum osmolality during the first 4 days in the ICU. The association between trajectory patterns and mortality was assessed using Cox proportional hazards regression. Mediation analysis was employed to explore potential biological mechanisms, while parametric g-formula simulations were used to evaluate the impact of trajectory-specific fluid management strategies.

Three distinct serum osmolality trajectories were identified: stable (ST), ascending (AS), and descending (DS). Compared to the ST group, both the AS and DS groups were associated with significantly higher 28-day mortality (HR 1.80, 95% CI 1.61–2.01 for AS; HR 1.83, 95% CI 1.62–2.06 for DS). Mediation analysis indicated that renal dysfunction (accounting for 11.16% of the total effect, P < 0.001) and cumulative positive fluid balance (11.39%, P < 0.001) partially mediated the observed associations. Parametric g-formula simulations revealed substantial heterogeneity in responses to fluid management across trajectory groups, with optimal fluid strategies varying significantly by trajectory pattern and patient characteristics.

Dynamic serum osmolality trajectories are independent predictors of mortality in sepsis, with effects partially mediated through renal dysfunction and fluid imbalance. These findings support the implementation of trajectory-guided precision fluid therapy as a novel framework for individualized sepsis care, challenging one-size-fits-all approaches and offering evidence-based guidance for personalized treatment strategies.

The online version contains supplementary material available at 10.1186/s40001-025-03699-6.