A Machine Learning Model for Pre-Dialysis Prediction of Intradialytic Hypotension Risk

10.6221/AN.202212_36(4).0005

I-Ning Yang；Chung-Feng Liu；Wei-Jian Lim；Chih-Chiang Chien；Hsien-Yi Wang；Jhi-Joung Wang；Yu-Ting Shen；Chia-Jung Chen

artificial intelligence ； hemodialysis ； intradialytic hypotension ； machine learning ； pre-dialysis

Acta Nephrologica

36卷4期（2022 / 12 / 01）

212 - 221

英文

BACKGROUND: Intradialytic hypotension (IDH) is a distressful complication for chronic hemodialysis patients. Despite its confirmed clinical significance with high mortality and morbidity, IDH remains challenging to predict and manage. Hence, our study aims to apply a machine learning model to predict the risk of IDH before a dialysis session and engage in early preventive interventions. METHODS: We used artificial intelligence (AI) and machine learning (ML) technologies to build the predictive model for IDH risk. In addition to the current status of the patient, the other features that we incorporated were sourced from the records of previous dialysis sessions to maximize the model performance. Seven ML algorithms were used for modeling, and the best model was chosen for AI predictive system implementation. The electronic medical records of patients who had undergone hemodialysis at Chi Mei Medical Center between September 1, 2020 and December 31, 2020 were included in this research. Preliminary impact analysis of AI assistance was conducted through the IDH rate. RESULTS: The results showed that the performance of our prediction model was excellent. A preliminary impact analysis indicated that the IDH rate significantly decreased after implementing AI assistance. The dialysis team was highly satisfied with the AI prediction for providing real-time scientific probabilities of IDH risks for each dialysis patient in a graphic style. CONCLUSION: We demonstrated a ML model that identifies patients likely to suffer from IDH before hemodialysis starts. Knowing the individual IDH risks in advance enables the dialysis staff to have enough time to handle this complex intradialytic complication more efficiently.