A new study has been published in Intensive Care Medicine Experimental, reporting the prospective and external validation of AKIRA, U-Care Medical’s artificial intelligence model for the continuous and early prediction of moderate and severe acute kidney injury (AKI) in critically ill patients.
The model was retrospectively evaluated using 70,107 ICU admissions across 176 intensive care units in the United States, the Netherlands and Italy. It was subsequently prospectively validated in 329 patients enrolled at three European hospitals in Italy and Spain.
In the retrospective cohorts, the model achieved auROC values above 0.89. In the prospective validation, predictive performance ranged from 0.82 to 0.96 across the different centers, with an average lead time of approximately 14 hours before the onset of stage 2 or 3 AKI.
These results support the model’s ability to maintain strong predictive performance across different patient populations and clinical settings, representing a further step in the validation of artificial intelligence applied to critical care.