Home›Rheumatology› Sepsis Mortality Prediction Models Show Moderate Performance
Sepsis Mortality Prediction Models Show Moderate PerformancePrediction models show moderate success in forecasting sepsis deaths
Frontiers in MedicinePublished June 10, 2026DOI ↗Editorial oversight: Dr. Amelia Tan, PhD · Internal Medicine & Chronic Disease
AI-generated summary of the cited source, checked by automated accuracy review.
How we work
Share
Key Takeaway
Sepsis mortality prediction models have moderate accuracy (AUC 0.79), but high bias in most studies limits their clinical reliability.
This systematic review and meta-analysis evaluated prediction models for mortality in patients with sepsis, analyzing data from 84 studies encompassing approximately 2.7 million patient records. The primary outcome was the discriminative performance of externally validated models, measured by the area under the receiver operating characteristic curve (AUC). The pooled AUC was 0.794 (95% CI: 0.755–0.834), indicating moderate discriminative ability.
Common predictors of mortality identified across the models included age, lactate, albumin, SOFA score, and vasopressor use. These predictors appeared frequently, with age reported in 19 studies, lactate in 13, albumin and SOFA score in 8 each, and vasopressor use in 7. However, only 11 out of 84 studies performed full external validation, limiting the generalizability of the findings.
Importantly, the review highlighted a high risk of bias in the majority of included studies. Specifically, 64 studies (76.19%) had high risk of bias in model development, and 50 studies (65.79%) in model evaluation. This raises concerns about the reliability of the reported performance metrics. The authors note that while the models show promise, their clinical utility is tempered by methodological limitations.
In summary, externally validated prediction models for sepsis mortality demonstrate moderate discriminative performance, but the high prevalence of bias in the literature underscores the need for more rigorous model development and validation practices before these tools can be confidently integrated into clinical decision-making.
How this fits prior evidence
This meta-analysis confirms prior coverage that prediction models show moderate discrimination for sepsis outcomes, extending the finding specifically to mortality with a pooled AUC of 0.794. It contrasts with the narrative review on neutrophil modulation, which did not report specific outcomes. The high risk of bias noted here aligns with earlier cautions about model limitations. The finding does not directly address antibiotic stewardship or gut microbiota, but reinforces the need for robust validation before clinical use.
Sepsis is a life-threatening medical emergency. When it hits, doctors need to know quickly who is at the highest risk of dying so they can act fast. Researchers looked at 2.7 million records across 84 different studies to see how well computer models could predict these outcomes.
The analysis found that these prediction models have a moderate ability to tell the difference between patients who survive and those who do not. The most common factors used by these models to predict death included age, lactate levels, albumin levels, SOFA scores (a measure of organ failure), and the use of vasopressors (medicines used to raise blood pressure).
While the tools show promise, there are important hurdles to clear. Many of the studies used to build these models had a high risk of bias, and only 11 out of the 84 studies were fully validated. Because of these limitations, while the models show some success, they are still being refined to ensure they work reliably in every hospital setting.
What this means for you:
Prediction models show moderate accuracy in identifying sepsis mortality based on factors like age and lactate levels.
Common questions
What factors are used to predict if a patient will survive sepsis?
The most common predictors found in the models were age, lactate levels, albumin levels, SOFA scores, and the use of vasopressors. These specific markers appeared frequently across the 84 studies analyzed to help determine the risk of mortality for patients with sepsis.
How accurate are these prediction models?
The study found that these models have a moderate discriminative performance. This means they have a fair ability to distinguish between outcomes, but they are not perfect. The results showed a pooled score of 0.794 for how well the models predicted mortality.
Are there any concerns with the data used for these models?
Yes, there are some limitations to keep in mind. A high number of studies had a high risk of bias during both development and evaluation. Additionally, only 11 out of the 84 studies included in the review were fully validated, meaning more research is needed to ensure reliability.
BackgroundSepsis remains a leading cause of mortality among critically ill patients worldwide. Although an increasing number of prediction models have been published in recent years, their predictive performance, methodological quality, and major predictors have not been comprehensively evaluated in a systematic and quantitative manner. This study aims to evaluate the performance of these models and to identify common predictors associated with sepsis mortality.MethodsWe systematically searched PubMed, Embase, Cochrane Library, and Web of Science for studies on sepsis mortality prediction models published up to July 1, 2025. Data were extracted and appraised using the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies (CHARMS), and risk of bias was assessed with the Prediction Model Risk of Bias Assessment Tool for Artificial Intelligence (PROBAST+AI). Meta-analyses were performed to pool area under the curve of the receiver operating characteristic (AUC) metric of externally validated models and the odds ratio (OR) of common predictors. The study was registered in PROSPERO (CRD42024604119).ResultsA total of 84 eligible studies were included, reporting 235 prediction models for sepsis mortality and involving approximately 2.7 million patient records reported across studies, with 461,387 deaths. Only 11(13.10%) studies encompassed model development, internal validation, and external validation. The included studies comprised 78(92.86%) retrospective cohort studies, 57(67.86%) studies developed in intensive care unit (ICU) settings, with MIMIC databases being among the most commonly used data sources. The most prevalent mortality endpoints were in-hospital (n = 38, 45.24%), 28-day (n = 23, 27.38%), and 30-day mortality (n = 16, 19.05%). The included studies employed various modeling approaches, such as logistic regression (LR), random forest (RF), eXtreme Gradient Boosting (XGBoost), and K-nearest neighbors (KNN). Of the included studies, 46(54.76%) evaluated the calibration, 26(30.95%) conducted decision curve analysis, and 25(29.76%) applied SHAP for interpretability, while most did not provide visual model presentation (e.g., nomograms). The reported AUCs ranged from 0.590–0.976 for model development and 0.530–0.992 for internal validation. The pooled AUC of externally validated models, based on one representative model per study, was 0.794 (95% CI: 0.755–0.834), indicating moderate discriminative performance. Overall, the most commonly used predictors were age (n = 19, 22.62%), lactate (n = 13, 15.48%), albumin (n = 8, 9.52%), SOFA score (n = 8, 9.52%), and vasopressor (n = 7, 8.33%). Notably, 64 (76.19%) studies and 50 (65.79%) studies were judged to have a high risk of bias in model development and model evaluation phases, respectively.ConclusionExternally validated prediction models generally demonstrate moderate discriminative performance for predicting sepsis mortality, but a substantial proportion of these studies were evaluated as having a high risk of bias. Age, lactate, albumin, SOFA score, and vasopressor use were identified as predictors of mortality. Future studies with larger cohorts, rigorous designs, and multicenter external validation are warranted to improve their generalizability and facilitate clinical implementation.Systematic review registrationThe unique registration identifier is CRD42024604119, and the publicly accessible website is https://www.crd.york.ac.uk/prospero/.
