Kidney Week

Abstract: SA-PO0076

Wearable Single-Lead Electrocardiography (ECG) Monitoring During Hemodialysis: Predicting Hyperkalemia and Intradialytic Hypotension Using ECG and Heart Rate Variability Parameters

Session Information

• Artificial Intelligence: Kidney Imaging, Pathology, and Precision Phenotyping
  October 24, 2026 | Location: Exhibit Hall A, Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Artificial Intelligence, Digital Health, and Data Science

• 300 Artificial Intelligence, Digital Health, and Data Science

Authors

• Ham, Youngrok, Chungnam National University Hospital, Daejeon, Korea (the Republic of)

Youngrok Ham, MD, PhD

• Park, Kyungho, Chungnam National University Hospital, Daejeon, Korea (the Republic of)

Kyungho Park, MD

• Yoon, Suyeon, Chungnam National University Hospital, Daejeon, Korea (the Republic of)

Suyeon Yoon, MD

• Jang, Soyoung, Chungnam National University Hospital, Daejeon, Korea (the Republic of)

Soyoung Jang, MD

• Jeong, Seongyeop, Chungnam National University Hospital, Daejeon, Korea (the Republic of)

Seongyeop Jeong

• Park, Hyerim, Chungnam National University School of Medicine, Daejeon, Korea (the Republic of)

Hyerim Park

• Lee, Eu Jin, Chungnam National University Hospital, Daejeon, Korea (the Republic of)

Eu Jin Lee, MD

• Na, Ki Ryang, Chungnam National University Hospital, Daejeon, Korea (the Republic of)

Ki Ryang Na, DrPH

• Choi, Dae Eun, Chungnam National University Hospital, Daejeon, Korea (the Republic of)

Dae Eun Choi, MD, PhD

Background

Hyperkalemia and intradialytic hypotension (IDH) are frequent, potentially life-threatening complications in maintenance hemodialysis (HD). Real-time non-invasive tools for predicting these events remain limited. Wearable ECG devices capture continuous cardiac signals throughout HD, potentially reflecting potassium-driven T-wave changes and hemodynamic instability.

Methods

Thirty-two maintenance HD patients completed 126 sessions. A wearable single-lead ECG device (HiCardi, MEZOO, Korea) was applied throughout each session. Parameters were extracted from two standardized 20-minute windows: early (+10 to +30 min) and late (−30 to −10 min), with delta (Δ) variables computed as late minus early. Among 45 candidate ECG/HRV variables, stable predictors were selected via Elastic Net with GroupKFold cross-validation (5-fold, 100 iterations; threshold >50%) and evaluated using generalized linear mixed models (GLMM) with patient-level random intercepts. Outcomes were hyperkalemia (serum potassium ≥5.0 mEq/L, 41.3% of sessions) and IDH (SBP drop ≥20 mmHg, 32.5% of sessions).

Results

For hyperkalemia, Δ T-wave amplitude (d_t_amp_mean) was the sole ECG predictor selected. Its addition to the covariate model significantly improved fit (OR 0.33, 95% CI 0.14–0.76, p=0.009; LRT p=0.001) and cross-validated AUC (0.634→0.666, p<0.001). For IDH, three ECG/HRV variables were selected: Δ mean heart rate (d_mean_hr), detrended Δ RMSSD (d_rmssd_detrend), and early ST-segment mean (early_st_segment_mean). While d_mean_hr and d_rmssd_detrend did not reach individual significance (p=0.116 and p=0.143, respectively), their joint addition with early_st_segment_mean substantially improved model fit (LRT p=0.002) and AUC (0.696→0.799, Δ+0.103 [95% CI 0.034–0.175], p<0.001). Early ST-segment mean was independently significant (OR 0.45, 95% CI 0.23–0.89, p=0.021).

Conclusion

Wearable ECG-derived parameters independently predicted hyperkalemia and IDH in maintenance HD patients. Δ T-wave amplitude emerged as a non-invasive marker of potassium-driven ECG change, while early ST-segment level was associated with IDH risk. These findings support the feasibility of continuous wearable ECG monitoring for real-time complication risk stratification during HD.

Funding

• Clinical Revenue Support