Default Handling of the Non-Assessable Verbal Glasgow Coma Scale Misclassifies Illness Severity in Mechanically Ventilated Patients: A Retrospective Analysis
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
The Glasgow Coma Scale (GCS) is the intensive care unit’s universal neurologic severity score and is embedded in the Acute Physiology and Chronic Health Evaluation (APACHE) and Sequential Organ Failure Assessment (SOFA), risk-adjusted mortality models, and ICU benchmarking. In the mechanically ventilated patient the verbal component cannot be obtained, and the conventions used to fill that gap (defaulting the total to a normal score of 15, or dropping the patient from the analysis) may relabel the sickest, most-sedated patients as neurologically normal and distort illness-severity estimates.
Objective
To quantify non-assessable GCS verbal examinations after acute brain injury and how the handling convention changes severity scoring and mortality-model behavior, in two independent critical care databases.
Materials and Methods
Retrospective cohort of adults with acute brain injury, first intensive care unit stay, MIMIC-IV. A verbal examination was non-assessable when documented “No Response-ETT.” We quantified burden and determinants, compared the MIMIC-IV derived GCS convention, which defaults intubated patients to 15, with a component-aware GCS, and audited mortality-model handling strategies.
Results
Among 14,230 patients, 45.2% had a non-assessable verbal examination; 47.5% of ventilated patients had none in the first 24 hours. Non-assessability was driven by mechanical ventilation (odds ratio, 27.4; 95% CI, 22.8 to 33.0) and associated with death (odds ratio, 1.56; 95% CI, 1.32 to 1.85). The MIMIC-IV derived GCS was 15 in 42.9% of patients, placed 11.6% in the lowest severity category despite eye and motor indicating GCS 9 or less, and discriminated mortality worse than a component-aware GCS (AUROC, 0.746 vs 0.783). Complete-case handling excluded 28.5% of patients with 50.2% of deaths. Both the default-to-normal convention and the complete-case selection trap replicated in the multi-hospital eICU-CRD database (among intubated stays, 25.2% carried a total GCS of 15; non-assessable stays had higher mortality than assessable stays, 16.6% vs 9.2%).
Discussion
A derived-score convention can make the sickest intubated patients appear neurologically normal, while complete-case handling removes the highest-risk patients from analysis.
Conclusion
In mechanically ventilated patients the default handling of the non-assessable verbal GCS misclassifies illness severity, and this distortion replicated across two independent databases including 171 hospitals in eICU-CRD/APACHE, so it is a general property of ICU scoring rather than a single-database quirk. ICU severity scores, risk-adjusted mortality models, and GCS-based benchmarking or quality metrics should flag and report how the non-assessable verbal examination was handled.
