Sandokji I, Yamamoto Y, Biswas A, et al. - Given that targeted interventions can be implemented as a result of timely prediction of AKI in children, but unique modeling challenges are posed by the wealth of data in the electronic health record, researchers analyzed retrospectively all children younger than 18 years old, using electronic medical records, who had at least two creatinine values recorded during a hospital admission from January 2014 through January 2018. This study involved derivation, and internal and external validation cohorts. From the electronic health records, experts selected 10 of 720 potentially predictive variables. AKI occurrence (per the Kidney Disease Improving Global Outcomes creatinine definition) within a moving 48-hour window was considered as the primary outcome. A machine learning-based genetic algorithm was employed in the model showing the highest-performance, yielding an overall receiver operating characteristic curve of 0.76, 0.79, and 0.81 for AKI, severe AKI, and for neonatal AKI, respectively, in the internal validation cohort. Experts defined high- and low-risk threshold points to translate this prediction model into a clinical risk-stratification tool. Overall, a time-updated prediction model incorporating ten easily available electronic health record variables was identified as well as validated employing various machine learning algorithms, and this model affords a tool to accurately predict imminent AKI among hospitalized children.