Predicting Sepsis Mortality: An Elastic-Net Regression Model versus the Sequential Organ Failure Assessment Score

Abstract Introduction: The Sequential Organ Failure Assessment version 1 (SOFA-1) score defines organ dysfunction in Sepsis-3. However, this score describes illness severity rather than predicts mortality. This study evaluated whether routine continuous admission variables improve mortality prediction. Methods: This retrospective cohort included 465 adults with sepsis admitted to intensive care units between March 2022 and March 2024. Overall, 114 patients died in hospital. Candidate predictor sets included six continuous SOFA-1 proxies and 14 additional routine admission variables. Models were evaluated using 20 repeats of stratified 10-fold cross-validation. The primary model was an elastic-net penalised logistic regression. Discrimination was compared using the paired DeLong test. Results: On held-out folds, the elastic-net model achieved an area under the curve (AUC) of 0.846. In comparison, SOFA-1 achieved an AUC of 0.711 (difference 0.134, 95% confidence interval (CI) 0.074 to 0.195; p < 0.001). The six-proxy model improved AUC by 0.089 (95% CI 0.033 to 0.144). The Glasgow Coma Scale alone matched SOFA-1 discrimination (difference in AUC 0.035, 95% CI − 0.025 to 0.096; p = 0.240). Adding age and sex to SOFA-1 yielded no significant change (difference in AUC 0.000, 95% CI − 0.017 to 0.018). Conclusions: Routinely recorded continuous admission variables predicted in-hospital mortality better than SOFA-1 in this cohort. Prospective external validation is required before clinical use.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-12
DOI
https://doi.org/10.21203/rs.3.rs-11141215/v1
Primary Topic
Sepsis Diagnosis and Treatment
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Predicting Sepsis Mortality: An Elastic-Net Regression Model versus the Sequential Organ Failure Assessment Score

Ramita Shahabifar, Reza Raeesi Estabragh, Mahdieh Azizian, Morteza Hashemian et al.
Research Square
Sepsis Diagnosis and Treatment
preprint

Predicting Sepsis Mortality: An Elastic-Net Regression Model versus the Sequential Organ Failure Assessment Score

Ramita Shahabifar, Reza Raeesi Estabragh, Mahdieh Azizian, Morteza Hashemian, Mohammad Rahmanian, Mahdieh Khazaneha, Neda Hamzehnejadi
preprint en

Abstract

Abstract Introduction: The Sequential Organ Failure Assessment version 1 (SOFA-1) score defines organ dysfunction in Sepsis-3. However, this score describes illness severity rather than predicts mortality. This study evaluated whether routine continuous admission variables improve mortality prediction. Methods: This retrospective cohort included 465 adults with sepsis admitted to intensive care units between March 2022 and March 2024. Overall, 114 patients died in hospital. Candidate predictor sets included six continuous SOFA-1 proxies and 14 additional routine admission variables. Models were evaluated using 20 repeats of stratified 10-fold cross-validation. The primary model was an elastic-net penalised logistic regression. Discrimination was compared using the paired DeLong test. Results: On held-out folds, the elastic-net model achieved an area under the curve (AUC) of 0.846. In comparison, SOFA-1 achieved an AUC of 0.711 (difference 0.134, 95% confidence interval (CI) 0.074 to 0.195; p < 0.001). The six-proxy model improved AUC by 0.089 (95% CI 0.033 to 0.144). The Glasgow Coma Scale alone matched SOFA-1 discrimination (difference in AUC 0.035, 95% CI − 0.025 to 0.096; p = 0.240). Adding age and sex to SOFA-1 yielded no significant change (difference in AUC 0.000, 95% CI − 0.017 to 0.018). Conclusions: Routinely recorded continuous admission variables predicted in-hospital mortality better than SOFA-1 in this cohort. Prospective external validation is required before clinical use.

Research Square
Shiraz University of Medical Sciences (IR), Kerman University of Medical Sciences (IR)
Good health and well-being
Sepsis Diagnosis and Treatment
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon