External Validation, Recalibration, and Extension of a Prediction Model of Early Acute Kidney Injury in Critically Ill Children Using Multicenter Data

Year of Publication:

Categories of Publication:

Lay Abstract and Manuscript Citation:

Title: Helping Doctors Spot Kidney Problems Early in the ICU

Key Words: Acute Kidney Injury (AKI), PEDSnet, ICU, prediction model, children

Doctors want to find kidney problems early in children who are very sick in the ICU. This study looked at how well a computer model could predict which kids might develop acute kidney injury (AKI) within the first three days. The researchers used two large hospital networks, including PEDSnet, which combines health records from many children’s hospitals.

The team tested an older model and found it didn’t work very well across different hospitals. They improved the model by adding new information available within the first 12 hours after a child enters the ICU. With these updates, the model did a better job of spotting children at higher risk for AKI.

The study included over 186,000 ICU visits, and only a small number of children developed AKI early. Still, the improved model was much more accurate and balanced in predicting risk. This means doctors may be able to act sooner to help protect children’s kidneys.

Takeaways

Why does this matter? Finding kidney problems early can help doctors treat children faster and prevent serious health issues.

What does this mean for patients and families? Better prediction tools—built using big networks like PEDSnet—can help hospitals give safer, more personalized care to children in the ICU.

Citation

Sanchez-Pinto LN, Nishisaki A, Weiss SL, et al. External Validation, Recalibration, and Extension of a Prediction Model of Early Acute Kidney Injury in Critically Ill Children Using Multicenter Data Critical Care Explorations. 2026;8(6):e1425. doi:10.1097/CCE.0000000000001425

Find publication:

https://www.ovid.com/jnls/ccejournal/fulltext/10.1097/cce.0000000000001425~external-validation-recalibration-and-extension-of-a

The lay abstract for this publication was generated using an AI model and reviewed by the lead author and the PEDSnet Engagement Core team.

Scroll to Top