Development and validation of a machine learning model to predict the use of renal replacement therapy in 14,374 patients with COVID-19
Authors' conclusion
An early ML model using easily available clinical and laboratory data accurately predicted the use of RRT in critically ill patients with COVID-19.
Study context
- Population
- Critically ill adults with COVID-19 requiring ventilatory support, 126 Brazilian intensive care unit (ICUs), February 2020 to December 2021 (development) and January to May 2022 (validation)1,924 (13%) required renal replacement therapy (RRT)
- Participants
- 14,374 adults
- Outcome
- Prediction of use of RRT
Methodological detailsprediction model development and validation, multicentre, machine learning, temporal validation, cross-validation
- Prediction model development and validation
- Multicentre
- Machine learning
- Temporal validation
- Cross-validation
Main finding
As reported in the abstract
Discrimination and calibration for predicting RRT
| Group | AUC (95% CI) | Brier's Score (95% CI) |
|---|---|---|
| Random Forest | 0.78 [0.75-0.81] | 0.09 [95%CI: 0.08-0.10] |
Random Forest
AUC (95% CI)0.78 [0.75-0.81]
Brier's Score (95% CI)0.09 [95%CI: 0.08-0.10]
Observational data: associations, not effects.
Further findings
| Outcome | Values |
|---|---|
| Temporal validation (2022 data) | Random Forest AUC (95% CI) 0.79 [0.75-0.84] · Brier's Score (95% CI) 0.08 [95%CI: 0.08-0.1] |
Temporal validation (2022 data)
- Comparable performance in temporal validation
Characteristics of RRT vs non-RRT patients
- Age, RRT vs non-RRT
- 65 [53-75] vs. 55 [42-68]
- Charlson's Comorbidity Index, RRT vs non-RRT
- 1.0 [0.00-2.00] vs 0.0 [0.00-1.00]
- SAPS-3 median, RRT vs non-RRT
- 61 [51-74] vs 48 [41-58]
- In-hospital mortality, RRT vs non-RRT
- 71% vs 22%
Risk factors for RRT (13 predictors selected by Recursive Feature Elimination (RFE))
- Creatinine, GCS, Urea, Invasive Mechanical Ventilation, Age, CKD, Platelets count, Vasopressors, NIV, Hypertension, DM, Modified Frailty Index (mFI) and Gender
Original publication
Development and validation of a machine learning model to predict the use of renal replacement therapy in 14,374 patients with COVID-19. J Crit Care. Published 27 November 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →