Prediction of intradialytic hypotension based on heart rate variability and skin sympathetic nerve activity using LASSO-enabled feature selection: a two-center study.

Background: Intradialytic hypotension (IDH) is a prevalent complication during hemodialysis (HD). However, conventional predictive models are imperfect due to multifaceted etiologies underlying IDH.

Methods: This study enrolled 201 patients undergoing maintenance HD across two centers. Seventy percent of the patient cohort was randomly allocated to the training cohort (n = 136), while the remaining 30% formed the validation cohort (n = 65). IDH was defined as a reduction in systolic blood pressure (SBP) ≥20 mmHg or mean arterial pressure (MAP) ≥10 mmHg. Clinical data and autonomic nervous parameters, including skin sympathetic nerve activity (SKNA) and heart rate variability (HRV) during the initial 30 min of HD, were employed to construct the model. The least absolute shrinkage and selection operator (LASSO) regression facilitated variable selection associated with IDH. Subsequently, a multivariable logistic regression model was formulated to predict the risk of IDH and establish the nomogram.

Results: Sixty-six baseline features were included in the LASSO-regression model. In the final multivariable logistic regression model, 5 variables (SBP₀, aSKNA₀, △aSKNA_0-30, SDNN₀, △SDNN_0-30) were incorporated into the nomogram. The AUC was 0.920 (95% CI, 0.878-0.962) in the training cohort and 0.855 (95% CI, 0.763-0.947) in the validation cohort, indicating concordance between the nomogram prediction and actual observation of IDH.

Conclusion: The LASSO-enabled model, based on clinical characteristics and autonomic nervous system parameters from the first 30 min of HD, shows promise in accurately predicting IDH.