Prediction models for mortality in patients with sepsis: a systematic review and meta-analysis
By
Siyuan Lei
Huanrong Ruan
Jun Wang
Guixiang Zhao
Hulei Zhao
Jianping Liu
Jiansheng Li
June 10, 2026
Clinical Scorecard: Systematic Review and Meta-Analysis of Mortality Prediction Models in Sepsis Patients
At a Glance
Category Detail
Condition Sepsis Key Mechanisms Dysregulated host response to infection leading to organ dysfunction. Target Population Critically ill patients with sepsis. Care Setting Intensive care units (ICUs) and hospital settings.
Key Highlights
84 studies included, reporting 235 prediction models for sepsis mortality. Pooled AUC of externally validated models was 0.794, indicating moderate performance. Common predictors of mortality include age, lactate, albumin, SOFA score, and vasopressor use. High risk of bias observed in 76.19% of model development studies. Majority of studies were retrospective cohort studies. Guideline-Based Recommendations
Diagnosis
Utilize existing scoring systems like SOFA, APACHE II, and SAPS II for prognosis assessment. Management
Implement machine learning-based models for improved predictive performance. Monitoring & Follow-up
Regularly assess the predictive performance of models in clinical settings. Risks
High risk of bias in many prediction models may affect reliability. Patient & Prescribing Data
Approximately 2.7 million patient records analyzed.
Mortality rates in sepsis patients are significant, with 26.7% hospital mortality.
Clinical Best Practices
Encourage multicenter external validation of prediction models. Focus on larger cohorts and rigorous study designs for future research. Related Resources & Content
