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

Please note that you are viewing an archived section from 2022 and some content may be unavailable. To unlock all content for 2022, please visit the archives.

Abstract: TH-PO884

Performance of Machine Learning Compared With Regression Analysis in Predicting Albuminuria

Session Information

Category: CKD (Non-Dialysis)

  • 2201 CKD (Non-Dialysis): Epidemiology‚ Risk Factors‚ and Prevention

Authors

  • Arsiwala, Ali Haider, Dr. Vasantrao Pawar Medical College, Hospital & Research Centre, Nashik, India
  • Mathew, Marcella, Charite Universitatsmedizin Berlin, Berlin, Berlin, Germany
  • Scheppach, Johannes B., Charite Universitatsmedizin Berlin, Berlin, Berlin, Germany
Background

Albuminuria is a marker of kidney disease and an independent risk factor for cardiovascular events, cognitive decline, and all-cause mortality. Machine learning (ML) algorithms are increasingly used for risk prediction in medical science, given their proficiency in identifying patterns in large datasets and modeling non-linear interactions, surpassing the capacity of traditional statistical approaches. We compared the predictive performance of three ML models with a regression model in correctly classifying participants as with or without albuminuria.

Methods

We studied 18,117 participants aged 20 to 80 years from the National Health and Nutrition Examination Survey (NHANES) between 2005 and 2018. Synthetic minority oversampling technique was used to correct the class imbalances in the data resulting from the oversampling recruitment strategy applied by NHANES. Self-reported variables like age, sex, ethnicity, education status, blood pressure, diabetes status, and modified frailty phenotype were used to predict albuminuria, defined as a Urine Albumin-Creatinine Ratio > 30 mg/g. Participants were randomly grouped into a training dataset (n=13,587) and a testing dataset (n=4,530). A logistic regression model, a gradient boosting classifier model, a k-nearest neighbor model, and an artificial neural network (ANN) were used for classification. To compare classification results, accuracy, sensitivity, and specificity were computed.

Results

Out of all prediction models, the ANN had the highest accuracy (0.89; 95% Confidence Interval: [0.88, 0.90]) and highest sensitivity (0.99 [0.97, 1.00]). For comparison, the logistic regression model showed an accuracy of (0.86 [0.85, 0.87]) and a sensitivity of (0.90 [0.89, 0.92]). However, the specificity of the ANN (0.77 [0.75, 0.80]) was lower than that of the logistic regression model (0.83 [0.81, 0.85]).

Conclusion

Using ML algorithms for the prediction of albuminuria showed an advantage in performance with higher accuracy and sensitivity over traditional regression analysis. Artificial intelligence-based screening tools for albuminuria developed from self-reported data could offer diagnostic and economic advantages in the early detection and prevention of chronic kidney disease.