Cheng-Jui Lin1, Chih-Yang Chen2, Pei-Chen Wu2, Chi-Feng Pan2, Hong-Mou Shih3, Ming-Yuan Huang4, Li-Hua Chou5, Jin-Sheng Tang6, Chih-Jen Wu7. 1. Division of Nephrology, Department of Internal Medicine, MacKay Memorial Hospital, Taipei, Taiwan; Department of Medicine, Mackay Medical College, New Taipei, Taiwan; Mackay Junior College of Medicine, Nursing and Management, Taipei, Taiwan. 2. Division of Nephrology, Department of Internal Medicine, MacKay Memorial Hospital, Taipei, Taiwan. 3. Division of Nephrology, Department of Internal Medicine, MacKay Memorial Hospital, Taipei, Taiwan; Graduate Institute of Physiology, College of Medicine, National Taiwan University, Taipei, Taiwan. 4. Department of Medicine, Mackay Medical College, New Taipei, Taiwan. 5. Division of Hemodialysis and Peritoneal Dialysis, Department of Nursing, MacKay Memorial Hospital, Taipei, Taiwan. 6. Department of Information Technology, MacKay Memorial Hospital, Taipei, Taiwan. 7. Division of Nephrology, Department of Internal Medicine, MacKay Memorial Hospital, Taipei, Taiwan; Department of Medicine, Mackay Medical College, New Taipei, Taiwan; Graduate Institute of Medical Sciences and Department of Pharmacology, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan; Department of Medical Research, China Medical University Hospital, China Medical University, Taichung, Taiwan. Electronic address: yaliwcj@gmail.com.
Abstract
BACKGROUND: Intradialytic hypotension (IDH) is a serious complication and a major risk factor of increased mortality during hemodialysis (HD). However, predicting the occurrence of intradialytic blood pressure (BP) fluctuations clinically is difficult. This study aimed to develop an intelligent system with capability of predicting IDH. METHODS: In developing and training the prediction models in the intelligent system, we used a database of 653 HD outpatients who underwent 55,516 HD treatment sessions, resulting in 285,705 valid BP records. We built models to predict IDH at the next BP check by applying time-dependent logistic regression analyses. RESULTS: Our results showed the sensitivity of 86% and specificity of 81% for both nadir systolic BP (SBP) of <90 mmHg and <100 mmHg, suggesting good performance of our prediction models. We obtained similar results in validating via test data and data of newly enrolled patients (new-patient data), which is important for simulating prospective situations wherein dialysis staff are unfamiliar with new patients. This compensates for the retrospective nature of the BP records used in our study. CONCLUSION: The use of this validated intelligent system can identify patients who are at risk of IDH in advance, which may facilitate well-timed personalized management and intervention.
BACKGROUND: Intradialytic hypotension (IDH) is a serious complication and a major risk factor of increased mortality during hemodialysis (HD). However, predicting the occurrence of intradialytic blood pressure (BP) fluctuations clinically is difficult. This study aimed to develop an intelligent system with capability of predicting IDH. METHODS: In developing and training the prediction models in the intelligent system, we used a database of 653 HD outpatients who underwent 55,516 HD treatment sessions, resulting in 285,705 valid BP records. We built models to predict IDH at the next BP check by applying time-dependent logistic regression analyses. RESULTS: Our results showed the sensitivity of 86% and specificity of 81% for both nadir systolic BP (SBP) of <90 mmHg and <100 mmHg, suggesting good performance of our prediction models. We obtained similar results in validating via test data and data of newly enrolled patients (new-patient data), which is important for simulating prospective situations wherein dialysis staff are unfamiliar with new patients. This compensates for the retrospective nature of the BP records used in our study. CONCLUSION: The use of this validated intelligent system can identify patients who are at risk of IDH in advance, which may facilitate well-timed personalized management and intervention.
Authors: Tae Wuk Bae; Min Seong Kim; Jong Won Park; Kee Koo Kwon; Kyu Hyung Kim Journal: Int J Environ Res Public Health Date: 2022-08-20 Impact factor: 4.614