ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

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

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: TH-PO917

Prediction of Rapid Kidney Function Decline in Type 2 Diabetes Using Machine Learning Combining Blood Biomarkers and Electronic Health Record Data

Session Information

Category: Diabetic Kidney Disease

  • 602 Diabetic Kidney Disease: Clinical

Authors

  • Nadkarni, Girish N., Icahn School of Medicine at Mount Sinai, New York, New York, United States
  • Fleming, Fergus, RenalytixAI, New York, New York, United States
  • Mccullough, James, RenalytixAI, New York, New York, United States
  • Chauhan, Kinsuk, Icahn School of Medicine at Mount Sinai, New York, New York, United States
  • Verghese, Divya Anna, Icahn School of Medicine at Mount Sinai, New York, New York, United States
  • He, John Cijiang, Icahn School of Medicine at Mount Sinai, New York, New York, United States
  • Quackenbush, John, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, United States
  • Bonventre, Joseph V., Brigham and Women's Hospital, Boston, Massachusetts, United States
  • Murphy, Barbara T., Icahn School of Medicine at Mount Sinai, New York, New York, United States
  • Parikh, Chirag R., Johns Hopkins University, Baltimore, Maryland, United States
  • Donovan, Michael J., Icahn School of Medicine at Mount Sinai, New York, New York, United States
  • Coca, Steven G., Icahn School of Medicine at Mount Sinai, New York, New York, United States
Background

Individuals with type 2 diabetes (T2DM) are at increased risk of rapid kidney function decline (RKFD). The application of machine learning to integrate biomarkers with EHR data may lead to improved prediction of RKFD.

Methods

We selected individuals with T2DM with a baseline eGFR ≥45 and <90 ml/min/1.73 m2 from the Mount Sinai BioMe Biobank (n=871). We measured plasma levels of tumor necrosis factor (TNFR)1 & 2, and kidney injury molecule(KIM)-1 and employed random forest (RF) models to combine the biomarkers with longitudinal clinical features extracted from the electronic health record (EHR) to predict RKFD (eGFR decline of ≥5 ml/min/1.73 m2/year).

Results

In 871 participants, baseline eGFR was 74 ml/min/1.73 m2, and median UACR was 13 mg/g. Overall, 164 (19%) of individuals experienced RKFD over a median follow-up of 4.7 years from the baseline specimen collection. In the training and test sets respectively, the combined RF model (clinical features plus biomarkers) had an AUC of 0.82 (95% CI, 0.81-0.83) and 0.80 (95% CI, 0.78-0.82), which outperformed a standard clinical model via logistic regression (AUC 0.64, 95% CI 0.63-0.65), a biomarker model alone (AUC 0.76, 95% CI 0.72-0.79), and RF model using clinical features alone (AUC 0.74, 95% CI 0.73-0.76). The RKFD score stratified 18%, 49%, and 33% of patients in the entire cohort to high, intermediate, and low-probability strata, respectively, with a PPV of 53% in the high-probability group and an NPV of 97% in the low-probability group (Figure).

Conclusion

In patients with type 2 DM, a RF model combining plasma biomarkers and longitudinal EHR data significantly improved prediction of RFKD over standard clinical or biomarker-only models. Further validation of such approaches is needed.

Distribution of RKFD Scores and Probability of Rapid Kidney Function Decline by Continuous and Categorical Strata

Funding

  • NIDDK Support