Admission GCS, Age, and Pupillary Response as a Multivariable Triad for Predicting Outcomes Following Emergent Surgery for Traumatic Brain Injury

Abstract

Introduction: Early prognostication for patients with moderate-to-severe traumatic brain injury (TBI) requiring emergent surgery and intensive care is critical but complex. While the Glasgow Coma Scale (GCS) is foundational, its standalone predictive power, especially when unadjusted for known confounders, can be misleading. This study aimed to determine the independent predictive value of admission GCS within a multivariable model including other key clinical predictors.

Methods: We conducted a retrospective, descriptive-analytic study at a tertiary referral center in Indonesia, analyzing a specific cohort of 150 patients with moderate-to-severe TBI (GCS 3–12) who all underwent the emergent ED-OR-ICU pathway between July and December 2024. Data on admission GCS, patient age, pupillary reactivity, and CT findings (Marshall score) were extracted. We built multivariable logistic regression models to predict two primary outcomes: (1) In-Hospital Mortality and (2) Unfavorable Functional Outcome (a composite of mortality or discharge to a skilled nursing/palliative care facility).

Results: A univariate analysis identifying a GCS cut-off of 9.5 produced a statistically unstable odds ratio (OR) for mortality of 104.87, consistent with quasi-complete separation. However, in the multivariable model, this effect was resolved. After adjusting for confounders, GCS remained a significant independent predictor of mortality (Adjusted OR 2.78 per point decrease) and unfavorable outcome (aOR 3.11 per point decrease). Crucially, non-reactive pupils (aOR 5.12 for mortality) and patient age (aOR 1.07 per year for unfavorable outcome) were found to be equally, if not more, powerful independent predictors.

Conclusion: Admission GCS is a robust and independent predictor of outcome in high-risk surgical TBI patients, but its true value is only revealed when used as part of a multivariable assessment. The statistical power of univariate GCS is easily inflated by confounding. We conclude that prognostication in this cohort must be a multivariable exercise, incorporating GCS, pupillary response, and age as an essential prognostic triad.