Researchers from Yonsei University in South Korea developed and tested a machine learning model to predict the 2-year risk of progression in patients with IgA nephropathy (IgAN). This condition is one of the causes of primary kidney diseases, especially in the Asian population.
Using a retrospective cohort from two hospitals involving 1,301 patients with confirmed IgAN, the researchers applied a random forest algorithm to predict two outcomes: the primary outcome tested was a significant decline in kidney function (30% reduction in estimated glomerular filtration rate) or the need for kidney replacement therapy. The secondary outcome was improvement in proteinuria. The model performed reasonably well in predicting both outcomes.
The most critical feature for predicting both outcomes was the baseline level of proteinuria. Additionally, the model stratified patients into low, moderate, and high risk for poor 10-year kidney outcomes. Those in the high and moderate-risk categories for the primary outcome had a significantly higher risk of adverse kidney events over ten years when compared to the low-risk group.
Ref: Kim Y, Jong Hyun Jhee, Chan Min Park, et al. Machine learning-based 2-year risk prediction tool in immunoglobulin A nephropathy. Kidney research and clinical practice. Published online October 27, 2023. doi: https://doi.org/10.23876/j.krcp.23.076