Kidney Week

Abstract: SA-PO716

Machine-Learning Based Prediction Model for Prognosis of IgA Nephropathy Patients

Session Information

• Glomerular Diseases: Clinical, Outcomes, Trials - III
  November 05, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
  Abstract Time: 10:00 AM - 12:00 PM

Category: Glomerular Diseases

• 1303 Glomerular Diseases: Clinical‚ Outcomes‚ and Trials

Authors

• Park, Sehoon, Seoul National University Hospital, Jongno-gu, Seoul, Korea (the Republic of)

Sehoon Park,

• Koh, Eun Sil, Catholic University of Korea School of Medicine, Seoul, Korea (the Republic of)

Eun Sil Koh,

• Baek, Chung Hee, Asan Medical Center, Songpa-gu, Seoul, Korea (the Republic of)

Chung Hee Baek,

• Kim, Yong Chul, Seoul National University Hospital, Jongno-gu, Seoul, Korea (the Republic of)

Yong Chul Kim,

• Lee, Jung Pyo, Seoul National University Seoul Metropolitan Government Boramae Medical Center, Dongjak-gu, Seoul, Korea (the Republic of)

Jung Pyo Lee,

• Kim, Dong Ki, Seoul National University Hospital, Jongno-gu, Seoul, Korea (the Republic of)

Dong Ki Kim,

• Han, Seung Hyeok, Severance Hospital, Seodaemun-gu, Seoul, Korea (the Republic of)

Seung Hyeok Han,

• Chin, Ho Jun, Seoul National University Bundang Hospital, Seongnam, Korea (the Republic of)

Ho Jun Chin,

• Joo, Kwon Wook, Seoul National University Hospital, Jongno-gu, Seoul, Korea (the Republic of)

Kwon Wook Joo,

• Kim, Yon Su, Seoul National University Hospital, Jongno-gu, Seoul, Korea (the Republic of)

Yon Su Kim,

• Lee, Hajeong, Seoul National University Hospital, Jongno-gu, Seoul, Korea (the Republic of)

Hajeong Lee,

Background

IgA nephropathy is one of the most common primary glomerulonephritis and the disease shows heterogeneous prognosis. International Risk Score system has been developed to predict the prognosis of IgA nephropathy, however, further investigation for additional prognostic modeling by machine-learning based method may further improve the prediction power.

Methods

We screened total of 5387 IgA nephropathy patients from 3 tertiary university hospitals in Korea. The study population was divided into development and validation cohort. Based on the collected electronic health records, CatBoost model was used for machine-learning based modeling for the adverse composite outcome, which included halving of eGFR and end-stage kidney disease. Area under curve (AUC) values for the outcomes within 1, 3, and 5 years were calculated to assess the discriminative power.

Results

The constructed model showed good discriminative power in the developmental cohort as the AUC values ranged from 0.94 to 0.99 for the study outcomes. In the developmental cohort, there was some attenuation particularly in the hospitals with small number of samples (AUC 0.65-0.80), however, most AUC values for the study outcomes remained in acceptable range (AUC 0.81-0.97).

Conclusion

Machine-learning based prediction model for IgA nephropathy prognosis may provide a valid tool to estimate the kidney failure risk in the patients.