Prediction of Intradialytic Blood Pressure Variation Based on Big Data

Abstract

Introduction: Cardiovascular (CV) events are the major cause of morbidity and mortality associated with blood pressure (BP) in hemodialysis (HD) patients. BP varies significantly during HD treatment, and the dramatic variation in BP is a well-recognized risk factor for increased mortality. The development of an intelligent system capable of predicting BP profiles for real-time monitoring is important. Our aim was to build a web-based system to predict changes in systolic BP (SBP) during HD. Methods: In this study, dialysis equipment connected to the Vital Info Portal gateway collected HD parameters that were linked to demographic data stored in the hospital information system. There were 3 types of patients: training, test, and new. A multiple linear regression model was built using the training group with SBP change as the dependent variable and dialysis parameters as the independent variables. We tested the model’s performance on test and new patient groups using coverage rates with different thresholds. The model’s performance was visualized using a web-based interactive system. Results: A total of 542,424 BP records were used for model building. The accuracy was greater than 80% in the prediction error range of 15%, and 20 mm Hg of true SBP in the test and new patient groups for the model of SBP changes suggested the good performance of our prediction model. In the analysis of absolute SBP values (5, 10, 15, 20, and 25 mm Hg), the accuracy of the SBP prediction increased as the threshold value increased. Discussion: This databae supported our prediction model in reducing the frequency of intradialytic SBP variability, which may help in clinical decision-making when a new patient receives HD treatment. Further investigations are needed to determine whether the introduction of the intelligent SBP prediction system decreases the incidence of CV events in HD patients.