Deep Learning Model for Predicting Intradialytic Hypotension Without Privacy Infringement: A Retrospective Two-Center Study

Abstract

ObjectivePreviously developed Intradialytic hypotension (IDH) prediction models utilize clinical variables with potential privacy protection issues. We developed an IDH prediction model using minimal variables, without the risk of privacy infringement.MethodsUnidentifiable data from 63,640 hemodialysis sessions (26,746 of 79 patients for internal validation, 36,894 of 255 patients for external validation) from two Korean hospital hemodialysis databases were finally analyzed, using three IDH definitions: (1) systolic blood pressure (SBP) nadir <90 mmHg (Nadir90); (2) SBP decrease ≥20 mmHg from baseline (Fall20); and (3) SBP decrease ≥20 mmHg and/or mean arterial pressure decrease ≥10 mmHg (Fall20/MAP10). The developed models use 30 min information to predict an IDH event in the following 10 min window. Area under the receiver operating characteristic curves (AUROCs) and precision-recall curves were used to compare machine learning and deep learning models by logistic regression, XGBoost, and convolutional neural networks.ResultsAmong 344,714 segments, 9,154 (2.7%), 134,988 (39.2%), and 149,674 (43.4%) IDH events occurred according to three different IDH definitions (Nadir90, Fall20, and Fall20/MAP10, respectively). Compared with models including logistic regression, random forest, and XGBoost, the deep learning model achieved the best performance in predicting IDH (AUROCs: Nadir90, 0.905; Fall20, 0.864; Fall20/MAP10, 0.863) only using measurements from hemodialysis machine during dialysis session.ConclusionsThe deep learning model performed well only using monitoring measurement of hemodialysis machine in predicting IDH without any personal information that could risk privacy infringement.

Advances in Human-Centered Care to Address Contemporary Unmet Needs in Chronic Dialysis

“…Current studies have shown that AI could help to ensure optimum haemoglobin by recommending suitable erythropoietin-stimulating agents based on the patient profile 117 or predict the risk of intradialytic hypotension by considering patient clinical factors. 118 Furthermore, a recent cross-sectional study of 195 patients showed that patients rated AI responses to questions higher in terms of quality and empathy than physicians, 119 which could potentially pave the way for the use of AI in pre-dialysis education. In the future, there could be a system that integrates multiple facets of a patient’s clinical and social situation into a thorough management plan.…”

Prevention of Intradialytic Hypotension in Hemodialysis Patients: Current Challenges and Future Prospects

“…97 Additionally, a Korean study demonstrated the efficacy of computer-derived deep learning in predicting IDH using data derived solely from the HD machine. 98 Further, a very recent paper described utilization of digitized EKG recording obtained within 48 h prior to dialysis session, to predict IDH. 99 Incorporating AI learning methods into dialysis supervision platforms could offer an advantage for the ongoing tasks of adjustment to and reaccommodation of local patterns and biases that impact care delivery.…”

Wearables in Nephrology: Fanciful Gadgetry or Prêt-à-Porter?

“…Safe and effective vascular access has thus far been a large barrier to wearable HD devices, with recent conceptualisations hoping to bridge this gap [ 61 ], in addition to the dilemma of safely anticoagulating the extracorporeal circuit [ 49 , 50 ]. Furthermore, cardiovascular stability must be considered, and although deep learning algorithms have been used to predict intradialytic hypotension in ESKD patients on HD [ 62 ], this was done with variables gleaned from a conventional dialysis machine that are not yet available on wearable technology.…”