HMS–Sepsis Mortality Model Published in CHEST and Released as Open-Source Tool on GitHub

The peer-reviewed study, “Development and Validation of the Hospital Medicine Safety Sepsis Initiative Mortality Model,” was published in CHEST in July 2024 by Prescott and colleagues. Using data from the HMS–Sepsis registry, the authors developed and validated a risk-adjustment model to predict 30-day mortality among patients hospitalized with community-onset sepsis.

About the CHEST 2024 Study

The study analyzed 17,514 sepsis hospitalizations across 66 hospitals participating in the HMS–Sepsis Initiative between April 2022 and September 2023. A split-sample design was used, with 70% of cases for model derivation and 30% for validation. The final model incorporated 13 physiologic variables, physiologic interactions, and 16 demographic and chronic health variables, including age, metastatic malignancy, temperature, altered mental status, and platelet count.

The HMS–Sepsis Mortality Model demonstrated strong discrimination and calibration, with C-statistics of 0.82 in the derivation cohort and 0.81 in the validation cohort. Importantly, the study showed that risk-adjusted mortality substantially changed hospital performance assessment, reclassifying more than one-third of hospitals compared with unadjusted mortality rates. These findings underscore the importance of accounting for patient case mix when benchmarking sepsis outcomes across institutions.

Model Now Live on GitHub

To extend the impact of this work, the HMS team has released the model code and accompanying data dictionary on GitHub:

🔗 HMS–Sepsis Mortality Model (GitHub): https://github.com/mi-hms/HMS-Mortality-Model

The repository enables researchers, quality improvement teams, and health systems to review, implement, and adapt the model for benchmarking, observational analyses, and evaluation of sepsis initiatives. By making the methodology fully transparent, HMS aims to support reproducible research and facilitate collaboration across institutions.

Implications for Sepsis Quality Measurement

Sepsis remains a leading cause of hospital mortality and health care utilization in the United States. Accurate risk adjustment is essential for fair hospital comparisons, monitoring improvement over time, and conducting causal inference analyses. The HMS–Sepsis Mortality Model provides a rigorously validated tool specifically designed for these purposes within real-world hospital data environments.

Citation and Use

Users of the HMS–Sepsis Mortality Model are asked to cite the CHEST 2024 publication:

Prescott HC, Heath M, Munroe ES, et al. Development and Validation of the Hospital Medicine Safety Sepsis Initiative Mortality Model. Chest. 2024;166(5):1035–1045.