The following is a summary of ‘’Systematic review of externally validated machine learning models for predicting acute kidney injury in general hospital patients,” published in the July 2023 issue of Nephrology by Wanstein et al.

Acute kidney injury (AKI) is a common and severe complication in hospitalized patients. Early prediction can help prevent or minimize harm.

Researchers performed a retrospective study to evaluate and identify externally validated machine learning (ML) models for predicting AKI in general hospital patients. They also assessed their limitations and potential for improved applicability in diverse settings. They conducted a literature review of externally validated ML models developed from general hospital populations, using the current AKI definition to explore this. Out of 889 studies screened, only three met the criteria. Although most models exhibited good performance and sound methodology, concerns arose regarding their development and validation in limited diversity populations, comparable digital ecosystems, extensive use of predictor variables, and excessive reliance on easily accessible kidney injury biomarkers. These limitations may hinder the applicability of the models in diverse socioeconomic and cultural settings. 

The study concluded that there is a need for simpler, portable prediction models that can outperform existing tools in predicting and diagnosing AKI.

Source: frontiersin.org/articles/10.3389/fneph.2023.1220214/abstract