Multicenter Development and Prospective Validation of eCARTv5: A Gradient-Boosted Machine-Learning Early Warning Score
Authors:
- Matthew M. MD ChurpekMCMatthew M. MD Churpek
- Kyle A. CareyKCKyle A. Carey
- Ashley SnyderASAshley Snyder
- Christopher J WinslowCWChristopher J Winslow
- Emily GilbertEGEmily Gilbert
- Nirav S ShahNSNirav S Shah
- Brian W. PattersonBPBrian W. Patterson
- Majid Md AfsharMAMajid Md Afshar
- Alan Md WeissAWAlan Md Weiss
- Devendra N. AminDADevendra N. Amin
- Deborah J. RhodesDRDeborah J. Rhodes
- Dana P. EdelsonDEDana P. Edelson
Open Access
- 26 March 2025
- journal article
- research article
- Published by Wolters Kluwer Health in Critical Care Explorations
- Vol. 7 (4) , e1232
Abstract
BACKGROUND: Early detection of clinical deterioration using machine-learning early warning scores may improve outcomes. However, most implemented scores were developed using logistic regression, only underwent retrospective validation, and were not tested in important subgroups. OBJECTIVE: The objective of our multicenter retrospective and prospective observational study was to develop and prospectively validate a gradient-boosted machine model (eCARTv5) for identifying clinical deterioration on the wards. DERIVATION COHORT: All adult patients admitted to the inpatient medical-surgical wards at seven hospitals in three health systems for model development (2006–2022). VALIDATION COHORT: All adult patients admitted to the inpatient medical-surgical wards and at 21 hospitals from three health systems for retrospective (2009–2023) and prospective (2023–2024) external validation. PREDICTION MODEL: Predictor variables (demographics, vital signs, documentation, and laboratory values) were used in a gradient-boosted trees algorithm to predict ICU transfer or death in the next 24 hours. The developed model (eCARTv5) was compared with the Modified Early Warning Score (MEWS), the National Early Warning Score (NEWS), and eCARTv2 using the area under the receiver operating characteristic curve (AUROC). RESULTS: The development cohort included 901,491 admissions, the retrospective validation cohort included 1,769,461 admissions, and the prospective validation cohort included 205,946 admissions. In retrospective validation, eCARTv5 had the highest AUROC (0.834; 95% CI, 0.834–0.835), followed by eCARTv2 (0.775 [95% CI, 0.775–0.776]), NEWS (0.766 [95% CI, 0.766–0.767]), and MEWS (0.704 [95% CI, 0.703–0.704]). eCARTv5’s performance remained high (AUROC ≥0.80) across a range of patient demographics, clinical conditions, and during prospective validation. CONCLUSION: We developed eCARTv5, which performed better than eCARTv2, NEWS, and MEWS retrospectively, prospectively, and across a range of subgroups. These results served as the foundation for Food and Drug Administration clearance for its use in identifying deterioration in hospitalized ward patients.Keywords
All Related Versions
This publication has 26 references indexed in Scilit:
- Delayed Rapid Response Team Activation Is Associated With Increased Hospital Mortality, Morbidity, and Length of Stay in a Tertiary Care Institution*Critical Care Medicine, 2016
- Subsampling Versus Bootstrapping in Resampling-Based Model Selection for Multivariable RegressionBiometrics, 2015
- Multicenter Development and Validation of a Risk Stratification Tool for Ward PatientsAmerican Journal of Respiratory and Critical Care Medicine, 2014
- Risk Stratification of Hospitalized Patients on the WardsChest, 2013
- The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and deathResuscitation, 2013
- A controlled trial of electronic automated advisory vital signs monitoring in general hospital wards*Critical Care Medicine, 2012
- Incidence, location and reasons for avoidable in-hospital cardiac arrest in a district general hospitalResuscitation, 2002
- Validation of a modified Early Warning Score in medical admissionsQJM: An International Journal of Medicine, 2001
- Investigating onset, cessation, relapse, and recovery: Why you should, and how you can, use discrete-time survival analysis to examine event occurrence.Journal of Consulting and Clinical Psychology, 1993
- Comparing the Areas under Two or More Correlated Receiver Operating Characteristic Curves: A Nonparametric ApproachBiometrics, 1988