Early prediction of mortality at sepsis diagnosis time in critically ill patients by using interpretable machine learning
Authors' conclusion
Our study demonstrated the feasibility of using the interpretable machine learning model to predict mortality at sepsis diagnosis time.
Study context
- Population
- Critically ill patients with sepsis diagnosis time < 24 h after intensive care unit (ICU) admission, MIMIC-IV database
- Participants
- 24,377 patients
- Outcome
- 30-day mortality prediction at sepsis diagnosis time (discrimination and calibration)
- Follow-up
- 30 days
Methodological detailsprediction model development and validation, temporal split (training, validation, testing), interpretable machine learning (XGBoost, SHAP)
- Prediction model development and validation
- Temporal split (training, validation, testing)
- Interpretable machine learning (XGBoost, SHAP)
Main finding
As reported in the abstract
30-day mortality prediction, testing dataset (2017-2019) · 30-day
AUROC
0.853
95% CI 0.837–0.868
Area under the precision-recall curve
0.581
95% CI 0.541–0.619
Calibration slope
1.03
95% CI 0.94–1.12
Calibration intercept
0.14
95% CI 0.04–0.25
Observational data: associations, not effects.
Further findings
30-day mortality rate, testing dataset
- 30-day mortality rate
- 17.9%
Top features (SHAP)
- Top three features: age, increased red blood cell distribution width, respiratory rate
Original publication
Early prediction of mortality at sepsis diagnosis time in critically ill patients by using interpretable machine learning. J Clin Monit Comput. Published 27 December 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →