Heterogeneous Physiological Trends and their Implications for Diagnosis of Paroxysmal Sympathetic Hyperactivity (PSH) among Traumatic Brain Injury Patients
Introduction: Dysautonomia is common after traumatic brain injury (TBI), and paroxysmal sympathetic hyperactivity (PSH) has become increasingly recognized as a potential source of secondary injury in critically ill TBI patients. Exploration of PSH symptom evolution following TBI may improve understanding and early recognition of this syndrome. Here, we aim to identify PSH clinical feature score (CFS) trajectories in the acute post-injury period and their relationships to both admission features and outcomes of interest. Dataset and
Method: This study used a cohort of critically ill adult TBI patients (N=221) with Abbreviated Injury Scale (AIS) > 0, ICU Length of Stay (LOS) >= 3 days, and hospital (LOS) >=14 days. Patients with mild uncomplicated TBI and spinal cord injuries were excluded. We computed the daily CFS score, a sub-score of PSH-Assessment Measure (PSH-AM), based on discrete categories of vital signs including Heart Rate, Respiratory Rate, Systolic Blood Pressure, and Temperature, as well as sweating and posturing ratings. We consider consecutive 13-day CFS scores after the first 24 hours of resuscitation. To discover heterogeneous trajectory groups, we use K-Means clustering with Dynamic Time Warping (DTW) distance metrics and the elbow method to determine the optimal number of groups. We then evaluate the relationship between group membership with clinical diagnosis of PSH as a binary outcome and PSH-AM Diagnostic Likelihood Tool (DLT) score on the 14th day as an ordinal outcome, as well as other outcomes ICU LOS, hospital LOS, and Glasgow Coma Scale (GCS) at time of discharge. We also explore the relationship between group membership and ICU admission variables, such as GCS and mGCS scores.
Results: We identify five trajectory groups, each with distinct CFS trend patterns. Analyzing the mean CFS trend by aggregating all patients belonging to each group, we note that Group 1, 3, and 5 share similar trends of increasing during the first 4-5 days and then stabilizing, though with different initial CSF scores (low, medium, and high, respectively). In contrast, Groups 2 and 4 maintain high CFS scores for about 5-6 days before exhibiting a downward trend. To assess the relationship between group membership and PSH diagnosis, we computed the odds of PSH clinical diagnosis for each group. We noted that Group 5 and 3 have the highest odds of PSH positive, 1.8 and 1.4, respectively, while Group 1, 2, and 4 have odds lower than 1. The overall Chi-squared test p-value is 0.008. Furthermore, ANOVA test (p-value=0.00015) shows that group members are associated with the DLT score on day 14, with Group 5 having the highest DLT score of 7.0 [95% CI 6.01- 7.90] and Group 1 the lowest at 2.9 (95% CI 1.6 - 4.3). In addition, we note that the mGCS score at the time of admission is associated with group membership (ANOVA test p-value = 0.002), specifically, Group 1 with the highest mean mGCS of 4.3 [95% CI 3.3 - 5.3] while Group 5 with the lowest mean mGCS at 2.4 [95% CI 1.9 -3.0]. However, we did not find a significant correlation between group membership and discharge GCS or mGCS, nor with other outcome variables such as LOS, ICU LOS, and the number of mechanical ventilation days.
Conclusions: The trajectory group analysis based on daily CFS score reveals heterogeneity in physiological trends for TBI patients, which are associated with PSH diagnosis and DLT score. In addition, the GCS score at the time of admission is associated with group membership. This study suggests the value of exploring physiological trend patterns as a tool for diagnosis and managing TBI patients. Further work is planned to study how early we could predict the CFS trend and PSH diagnosis and explore the relationship between trajectory patterns and PSH intervention strategies.
