Construction of an early alert system for intradialytic hypotension before initiating hemodialysis based on artificial intelligence (Preprint)

BACKGROUND Intradialytic hypotension (IDH) is one of the most common complications during hemodialysis,and is an important risk factor for cardiovascular events, hospitalization, and mortality. The reported incidence of IDH ranges from 7.5% to 69% according to different definitions.According to the recommended definition of Kidney Disease Outcome Quality Initiative (KDOQI), the incidence rate is 20%-30%. The mechanism of IDH is complicated and is associated with decreased blood volume and insufficient vascular response. The occurrence of IDH depends on various risk factors, including demographic factors (age, sex), comorbidities (diabetes, coronary heart disease, left ventricular hypertrophy), dialysis characteristics, complications (dialysis vintage, weight gain between dialysis sessions, anemia), dialysis treatment prescription (ultrafiltration), medications (antihypertensive medications), and laboratory findings.Therefore, pre-assessment and monitoring of all patients during hemodialysis are critical for IDH prevention and treatment.However, IDH is associated with many factors and is difficult to predict, even for experienced nephrologists. Many efforts have been made to predict and initiate interventions to prevent IDH, however, in clinical settings, the accuracy and applicability remain poor. Artificial intelligence (AI) is a branch of computer science that can analyze complex medical data and can be used for clinical diagnosis, treatment, prediction, and efficacy evaluation.[14]Machine learning can merge a large amount of different data into a unified algorithm, such that a large amount of data can be used to develop reliable predictive analysis. AI has been applied to predict IDH,however, they can only predict IDH more than 1 h after initiating hemodialysis, which results in a compromised possibility of early intervention. The purpose of this research was to establish a software-based early alert system using artificial intelligence to identify patients at a high risk of IDH just before the initiation of hemodialysis, thereby allowing the medical team to manage and prevent IDH during hemodialysis. OBJECTIVE The purpose of this study was to use artificial intelligence for establishing an early alert system before hemodialysis sessions to identify patients at a high risk of IDH. METHODS We obtained data on 314534 hemodialysis sessions conducted at Sichuan Provincial People's Hospital from the Renal Disease Treatment Information System. IDH was defined as a systolic blood pressure drop≥20 mmHg, mean arterial pressure drop≥10 mmHg during dialysis, or the occurrence of clinical hypotensive events requiring nursing intervention. After pre-processing, the data were randomly divided into training (70%) and testing (30%) sets. Four interpolation methods, three feature selection methods, and 18 machine learning algorithms were used to construct predictive models. The area under the receiver operating characteristic curve (AUROC) was the main indicator for evaluating the performance of the models, while Shapley Additive ExPlanation (SHAP) was used to explain the contribution of each variable to the best predictive model. RESULTS A total of 3906 patients and 314534 dialysis sessions were included, of which 142237 cases showed IDH (incidence rate, 45.2%). Nineteen parameters were identified through artificial intelligence feature screening, including age, weight before dialysis, dry weight, pre-dialysis blood pressure, heart rate, prescribed ultrafiltration, blood cell counts (neutrophil, lymphocyte, monocyte, eosinophil, lymphocyte, and platelet counts), hematocrit, serum calcium, creatinine, urea, glucose, and uric acid. Random forest, gradient boosting, and logistic regression were the three best-performing models, and the AUROCs were 0.812 (95% CI, 0.811-0.813), 0.748 (95% CI, 0.747-0.749), and 0.743 (95% CI, 0.742¬-0.744), respectively. CONCLUSIONS Our dialysis software-based artificial intelligence alert system can be used to predict IDH occurrence, there by contributing to initiate relevant interventions.