Development and external validation of a machine learning model for the prediction of persistent acute kidney injury stage 3 in multi-centric, multi-national intensive care cohorts
Authors' conclusion
A machine learning approach fed with the proper data pipeline can accurately predict onset of Persistent AKI Stage 3 during ICU patient stay in retrospective, multi-centric and international datasets.
Study context
- Population
- Adults admitted to intensive care unit (ICU) with acute kidney injury (AKI) stage 2 or 3 (KDIGO criteria)enrolled for analysis
- Participants
- 7,759 adults
- Exposure
- US external validation vs Italian external validation
- Outcome
- Ability to predict AKI stage 3 lasting ≥72 h while in the ICU
Methodological detailsmachine learning prediction model development and external validation, multi-centric, multi-national, Explainable tree regressor trained and calibrated on 2 tertiary, urban, academic, single-centre databases; externally validated on 2 multicentre databases
- Machine learning prediction model development and external validation
- Multi-centric
- Multi-national
- Explainable tree regressor trained and calibrated on 2 tertiary, urban, academic, single-centre databases; externally validated on 2 multicentre databases
Main finding
As reported in the abstract
Prediction of persistent stage 3 AKI (AUC)
| Group | AUC |
|---|---|
| US external validation | 0.94 (95% CI 0.92-0.95) |
| Italian external validation | 0.85 (95% CI 0.83-0.88) |
AUC
US external validation0.94 (95% CI 0.92-0.95)
Italian external validation0.85 (95% CI 0.83-0.88)
Observational data: associations, not effects.
Further findings
Incidence of persistent stage 3 AKI
- Development cohort
- 11%
- Internal validation cohort
- 6%
- External validation cohorts
- 19%
Original publication
Development and external validation of a machine learning model for the prediction of persistent acute kidney injury stage 3 in multi-centric, multi-national intensive care cohorts. Crit Care. Published 4 June 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →