Reproducible intubation contrast without evidence of a unique threshold at Glasgow Coma Scale 8: a two-era nationwide retrospective cohort study

A Glasgow Coma Scale (GCS) score of 8 or lower is widely used in trauma classification and airway assessment, but a difference across a familiar cutoff does not necessarily mean that risk changes abruptly at that boundary. We tested whether the boundary between GCS 8 and 9 behaved as a unique prognostic or recorded-intubation threshold. We analyzed adults in two Japan Trauma Data Bank eras (2008–2014 and 2019–2023) who were transported directly from scene, had valid hospital-arrival GCS, no recorded prehospital advanced airway or cardiopulmonary support, and predefined arrival physiology. Outcomes were severe head injury, in-hospital death, urgent cranial surgery, and recorded post-arrival intubation. We asked whether GCS 8 and 9 differed, whether that difference was unusual compared with neighboring scores, and whether adding a discrete jump between GCS 8 and 9 improved model fit beyond a graded GCS-outcome relationship. Bayesian information criterion (BIC) was the prespecified criterion for added-step model comparisons; Akaike information criterion (AIC), likelihood-ratio tests, and coefficient-level inference were secondary. Cohorts included 71,988 patients from 190 hospitals and 66,417 from 209 hospitals. Prognostic and cranial-surgery differences between GCS 8 and 9 were imprecisely estimated. Recorded intubation was 9.1 and 10.6 percentage points more frequent at GCS 8 (pooled adjusted odds ratio, 1.61; 95% confidence interval, 1.36–1.91). Other adjacent-score intubation differences were also present. At the GCS 8/9 boundary, BIC favored the graded model in every outcome in both eras, including a GCS 5–12 spline sensitivity; secondary criteria were less consistent. Categorical-model estimates of local exceptionality were imprecise and inconsistent across eras. A recorded-intubation contrast was reproduced temporally within one national registry. Its local exceptionality can neither be established nor excluded. Prespecified BIC comparisons did not favor an added 8/9 step, but this does not establish the absence of a threshold. These observational data do not determine whether maintaining, changing, or abandoning an intubation rule improves outcomes.