Paroxysmal sympathetic hyperactivity risk modeling based on transients in time series describing the autonomic nervous system and cerebral hemodynamics.

Abstract

Overstimulation of the autonomic nervous system (ANS) in the acute phase after traumatic brain injury (TBI) may lead to paroxysmal sympathetic hyperactivity (PSH) syndrome. This study aimed to investigate the impact of the relationship between ANS activity and cerebral hemodynamics on the development of PSH syndrome. This retrospective study included 41 TBI patients admitted to Wroclaw University Hospital (Poland). Among them, 14 were classified as at risk for PSH based on the probabilistic Paroxysmal Sympathetic Hyperactivity Assessment Measure (PSH-AM), with 10 rated as 'possible' and 4 as 'probable'. High-resolution neuromonitoring data from the first 72 h post-injury included intracranial pressure (ICP), pressure reactivity index (PRx), baroreflex sensitivity (BRS), arterial blood pressure (ABP), and heart rate (HR). The correlation between ANS activity and cerebral hemodynamics was quantified using the mean, standard deviation, and zero-crossing rate (ZCR) across sliding windows of 3, 6, 12, and 24 h. Logistic regression was used to model PSH risk. The PSH risk model, including ZCR-based variability of ANS-cerebral hemodynamic correlations within a 3-h sliding window and adjusted by clinical metadata, achieved the highest performance (AUC 0.72 ± 0.27), outperforming the clinical metadata-only model (AUC 0.64 ± 0.18). Aggregated feature importance values indicated that the most predictive relationships were observed between HR-ICP and HR-PRx. Including the early post-injury interactions between ANS and cerebral hemodynamics in the clinical characteristics-based PSH risk model may improve its performance. Further studies in larger cohorts are necessary to validate these findings.