# A novel scoring system for evaluating mortality risk of patients with sepsis during early hospitalization

## TL;DR

This paper introduces a new scoring system to predict sepsis mortality risk in patients during early hospitalization or in the emergency department.

## Contribution

The novel e-SEPSS scoring system uses early clinical and lab data to predict mortality risk in septic patients.

## Key findings

- e-SEPSS showed a significant linear increase in mortality risk with each additional point, from 4.1% to 100%.
- The scoring system demonstrated strong discrimination power with ROC AUC values of 0.718 in the development group and 0.798 in the validation group.
- Shorter survival times and time-to-death with higher scores confirmed the validity of e-SEPSS.

## Abstract

Sepsis is a complex life-threatening condition. Early initiation of treatment is crucial in reducing mortality. Current scoring systems have a lack of reliability in the emergency department (ED) or during early hospitalization (EH). Thus, a quick, reliable, and objective scoring system for assessing mortality risk during EH or ED could significantly reduce sepsis mortality.

Using the MIMIC-IV database, we identified 7546 patients hospitalized due to septicemia. We included 13 comorbidity groups and the first chronologically available values of 75 laboratory parameters from the ED or EH. To create and validate our scoring system for early prediction of in-hospital mortality (e-SEPSS), patients were assigned to model development (MD) (N = 1004) or model validation (MV) group (N = 6542), with the latter serving as internal validation of e-SEPSS. Each risk factor that contributed significantly to mortality was assigned one point. Groups with different numbers of points were compared according to mortality and hospitalization duration.

Decreased chlorides, increased mean corpuscular hemoglobin, increased red blood cell distribution width, increased phosphates, decreased pH, increased partial thromboplastin time, and increased lactate dehydrogenase were included in e-SEPSS due to the highest reliability in predicting mortality. Patients received 1 point for each parameter, creating 8 mortality risk groups. A significant linear increase in mortality with each additional point was shown, ranging from 4.1% (0 points) to 100% (7 points) in the MD group. Similar trends were observed in the MV group. High power in discriminating patients with different mortality risks was shown (MD (ROC AUC = 0.718, CI 0.682–0.754), MV (ROC AUC = 0.798, CI 0.783–0.813)). Decreased survival time and shorter time-to-death with each additional point strengthened the validity of e-SEPSS (Mantel-Cox χ2(7) = 994.2, p-value < 0.001).

e-SEPSS provides a simple, objective, reliable, and accessible way of predicting mortality in septic patients in the ED or during EH.. After external and clinical validation of e-SEPSS, it could become a useful additional tool in reducing sepsis mortality.

The online version contains supplementary material available at 10.1186/s12879-025-10920-8.