ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Abstract: TH-PO022

Machine Learning Models for the Prediction of Kidney Stone Composition and Recurrence

Session Information

Category: Augmented Intelligence, Digital Health, and Data Science

  • 300 Augmented Intelligence, Digital Health, and Data Science

Authors

  • Bargagli, Matteo, Department of Nephrology and Hypertension, Inselspital, Bern University Hospital, University of Bern, Bern, Switzerland
  • Fuster, Daniel G., Department of Nephrology and Hypertension, Inselspital, Bern University Hospital, University of Bern, Bern, Switzerland
Background

Kidney stones are prevalent and cause high patient morbidity and healthcare cost. Kidney stone treatment depends on stone type and recurrence risk. We hypothesized that machine learning (ML) methods provide higher accuracy compared to current approaches for stone type and recurrence risk prediction.

Methods

Data from three comprehensively phenotyped Swiss cohorts comprising 1505 kidney stone formers with demographic, anthropometric and clinical information, stone composition analysis, and 24-h urine measurements were included. Several supervised ML models, including logistic regression, parallel-tree boosting (XGBoost), random forests, and neural networks, were trained independently to predict the stone type and the 5-year recurrence risk.

Results

XGBoost performed with generally high specificity (>90%), except for calcium oxalate stones which demonstrated lower sensitivity. The algorithm achieved an accuracy of 85% when distinguishing between uric acid-containing and calcium phosphate-containing stones. Key features informing the model included age at first stone event, body mass index, 24h urine calcium and pH. The 5-year recurrence risk predicted by the neural network ranged between 48% and 83%, closely aligning with observed recurrence risk (R2 = 0.913).

Conclusion

The developed ML models demonstrated remarkable accuracy in predicting the risk of stone recurrence. These findings have the potential to address an unmet clinical need by assisting healthcare specialists in clinical decision-making, and ultimately enhancing patient outcomes and quality of life for those affected by recurrent kidney stone disease.