Daily Sepsis Research Analysis

Using 6,929 patients from MIMIC-IV, the investigators compared 12 machine learning algorithms and developed an interpretable CatBoost model using 32 clinically accessible predictors. The model achieved an area under the receiver operating characteristic curve of 0.828 in the development dataset and 0.793 in external eICU testing, supporting transportable early risk stratification.

Impact: This study moves beyond single-center prediction by combining a large derivation cohort, interpretable feature selection, web deployment, and external validation. Its focus on the clinically important subgroup of diabetic patients with sepsis-associated acute kidney injury increases practical relevance.

Clinical Implications: The model could support early identification of diabetic patients with sepsis-associated acute kidney injury who are at high risk of in-hospital death and may benefit from intensified monitoring and timely escalation of care. It should be used as decision support rather than as a replacement for clinical judgment until prospective impact studies confirm clinical benefit.

Future Directions: Prospective multicenter impact studies should evaluate calibration across healthcare systems, workflow integration, treatment effects after model implementation, and whether the web tool improves recognition, resource allocation, or survival.

This prospective age-stratified study measured heparanase-1 activity and heparan sulfate disaccharides using liquid chromatography-tandem mass spectrometry in children and adults with sepsis. Heparanase activity and circulating heparan sulfate were elevated in both age groups, but adult activity was approximately tenfold higher and correlated with markers of vascular endotheliopathy and organ dysfunction.

Impact: The study identifies a substantial age-dependent difference in a mechanistically relevant endothelial glycocalyx pathway, challenging the assumption that sepsis endothelial biology is uniform across ages. These findings provide a rationale for age-stratified biomarker interpretation and therapeutic development.

Clinical Implications: Circulating heparanase-1 may become a candidate biomarker for endothelial injury and age-specific risk assessment in sepsis. However, assay standardization, larger age-stratified validation cohorts, and studies linking activity-guided treatment to outcomes are required before clinical use.

Future Directions: Larger longitudinal studies should define age-specific reference ranges, identify the cellular sources of circulating heparanase-1, determine temporal relationships with organ failure, and test age-adapted glycocalyx-protective interventions.

In 17,403 adults meeting Sepsis-3 criteria in MIMIC-IV, late-ICU estimated plasma-volume status and its change were independently associated with in-hospital mortality, whereas the early value was not significant after adjustment. Restricted cubic spline and threshold analyses suggested increased risk at late ePVS of at least 8 dL/g or an ePVS change of at least 2.

Impact: This large cohort evaluates a routinely calculable, dynamic physiologic measure rather than relying only on admission severity. The nonlinear findings may refine volume-status interpretation and prognostic stratification, while also illustrating why static single-time-point measurements can be misleading.

Clinical Implications: Late-ICU ePVS and its trajectory may complement clinical examination, fluid balance, and hemodynamic assessment when estimating mortality risk in sepsis. The thresholds should not be used to prescribe fluid removal or administration because the study does not establish a treatment target or causal benefit.

Future Directions: Prospective multicenter studies should validate the thresholds in diverse sepsis populations, test time-updated models against standard hemodynamic measures, and determine whether ePVS-guided management improves outcomes without increasing kidney injury or other complications.