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

ASN / Education & Meetings / Kidney Week /

Please note that you are viewing an archived section from 2023 and some content may be unavailable. To unlock all content for 2023, please visit the archives.

Abstract: SA-PO1130

Dynamic Bayesian Networks (DBN) Predicted ≥40% Decline in eGFR over Six Years

Session Information

Category: CKD (Non-Dialysis)

  • 2301 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention

Authors

  • Petousis, Panayiotis, University of California Los Angeles, Los Angeles, California, United States
  • Gordon, David, University of California Los Angeles, Los Angeles, California, United States
  • Daratha, Kenn B., Providence Medical Research Center, Spokane, Washington, United States
  • Norris, Keith C., University of California Los Angeles, Los Angeles, California, United States
  • Tuttle, Katherine R., Providence Medical Research Center, Spokane, Washington, United States
  • Nicholas, Susanne B., University of California Los Angeles, Los Angeles, California, United States
  • Bui, Alex, University of California Los Angeles, Los Angeles, California, United States

Group or Team Name

  • Center for Kidney Disease Research, Education, and Hope (CURE-CKD).
Background

A ≥40% decline in eGFR over two years is associated with an adjusted hazard ratio of ~10 for progression to kidney failure in patients with chronic kidney disease (CKD). Here, we predicted ≥40% eGFR decline at least one year in advance and simulated missing values along a patient’s CKD trajectory.

Methods

Electronic health records from patients with and at-risk (diabetes, prediabetes, hypertension) for CKD (by diagnosis code and eGFR <60 mL/min/1.73 m2) from the CURE-CKD Registry (1/1/2006-12/31/2020; N=2,250,806) at Providence (N=1,917,619) and UCLA (N=333,187) Health were used. A DBN for CURE-CKD was created using the Ranking Approaches for Unknown Structures software. The DBN was trained, tuned, and validated on a blind test set. We included demographics, comorbidities, lab values, medications, and vitals over 6 years. The primary outcome was an annual ≥40% decline in eGFR from baseline.

Results

Using observations before the prediction year, the DBN predicted the outcome of the target year despite an average rate of 29% missing values across most variables. Model performance improved over time, suggesting that longer follow-up improved the simulation of missing values and the prediction of ≥40% eGFR decline. By the 6th year, the model achieved an area under the receiver operating characteristic curve (AUCROC) of 0.83 and an average precision (AP) of 0.21 (Figure). When stratified, patients with albuminuria/proteinuria had the highest prevalence of ≥40% eGFR decline and were most accurately predicted (Table), and higher performance was observed in Hispanic and Black groups, who had a higher prevalence of ≥40% eGFR decline versus the whole population.

Conclusion

DBNs can simulate missing data to predict ≥40% eGFR decline in patients with CKD and in racial and ethnic minority groups. Model performance dramatically improved with longer follow-up.

Model performance for >=40% eGFR decline in the 6th year

Funding

  • Other NIH Support