Kidney Week

Abstract: SA-PO390

Prediction of Gastrointestinal Bleeding Hospitalization in Hemodialysis

Session Information

• Hemodialysis and Frequent Dialysis: CV and Risk Prediction
  November 05, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
  Abstract Time: 10:00 AM - 12:00 PM

Category: Dialysis

• 701 Dialysis: Hemodialysis and Frequent Dialysis

Authors

• Lama, Suman Kumar, Fresenius Medical Care, Global Medical Office, Waltham, Massachusetts, United States

Suman Kumar Lama,

• Larkin, John W., Fresenius Medical Care, Global Medical Office, Waltham, Massachusetts, United States

John W. Larkin,

• Chaudhuri, Sheetal, Fresenius Medical Care, Global Medical Office, Waltham, Massachusetts, United States

Sheetal Chaudhuri,

• Willetts, Joanna, Fresenius Medical Care, Global Medical Office, Waltham, Massachusetts, United States

Joanna Willetts,

• Winter, Anke, Fresenius Medical Care, Global Medical Office, Bad Homburg, Germany

Anke Winter,

• Jiao, Yue, Fresenius Medical Care, Global Medical Office, Waltham, Massachusetts, United States

Yue Jiao,

• Stauss-Grabo, Manuela, Fresenius Medical Care, Global Medical Office, Bad Homburg, Germany

Manuela Stauss-Grabo,

• Usvyat, Len A., Fresenius Medical Care, Global Medical Office, Waltham, Massachusetts, United States

Len A. Usvyat,

• Hymes, Jeffrey L., Fresenius Medical Care, Global Medical Office, Waltham, Massachusetts, United States

Jeffrey L. Hymes,

• Maddux, Franklin W., Fresenius Medical Care AG & Co KGaA, Bad Homburg, Hessen, Germany

Franklin W. Maddux,

• Stenvinkel, Peter, Karolinska Institutet, Stockholm, Sweden

Peter Stenvinkel,

• Floege, Jürgen, University Hospital Aachen, Division of Nephrology and Clinical Immunology, Aachen, Germany

Jürgen Floege,

Group or Team Name

• on behalf of the INSPIRE Core Group

Background

INitiativeS on advancing Patients’ outcomes In REnal disease (INSPIRE) is a collaboration set forth to conduct investigations fundamental to advancing nephrology. Early detection of gastrointestinal (GI) bleeding events was chosen as a priority and the group built a machine learning model to identify a hemodialysis (HD) patient’s 180-day GI bleed hospitalization risk.

Methods

Data from adult HD patients in the United States (Jan 2016-Dec 2020) were randomly split for training (50%), validation (30%), and testing (20%). Model (XGBoost) was built with >400 exposures, and refined to top 50 exposures, for classification of 180-day GI bleed hospitalization risk. Unseen testing data determined model performance and effect sizes.

Results

Among 27,796 HD patients in the test data, 322 had a GI bleed hospitalization. Model showed an area under the curve=0.70, sensitivity=56.2%, specificity=70.7%, and accuracy=70.6%. Exposures with largest effect size per Shapley values were older age (68±13 years GI bleed event vs 63±14 years no event), higher serum 25 hydroxyvitamin D level (25OH Vit D) (33.4±17.0 ng/mL GI bleed event vs 30.5±16.1 ng/mL no event), shorter days since all-cause hospitalization (203±246 days GI bleed event vs 253±269 days no event). Other important exposures included days since GI bleed hospitalization, hemoglobin and iron, time since known GI bleed, and heparin dose (Figure 1).

Conclusion

Model appears to be suitable for early detection of HD patients at risk of a GI bleed requiring hospitalization within 180 days. Prospective testing is needed with hopes of timely actions to avoid events. The association between higher 25OH Vit D and GI bleeding was unexpected and needs to be explored further in HD patients. Similar signals have been seen in warfarin users without kidney disease at 25OH Vit D levels >30 ng/mL (Keskin U, 2019).

Funding

• Commercial Support –