Prognostic Performance of the Glasgow Coma Scale Versus the Four Score for Predicting Mortality in Adult Intensive Care Patients: A Systematic Review and Meta-Analysis

Background: Level of consciousness is a core element of risk assessment in intensive care. The verbal component of the Glasgow Coma Scale (GCS) cannot be assessed in intubated patients, whereas the Full Outline of UnResponsiveness (FOUR) score replaces it with brainstem and respiratory items. Whether this design translates into better mortality prediction in adult intensive care unit (ICU) patients remains uncertain. Purpose: This study aimed to compare the discrimination of the FOUR score and the GCS for mortality in adult ICU patients, and to quantify the difference in the area under the receiver operating characteristic curve (AUC) between the two scores when both were measured in the same patients. Methods: This review was registered in PROSPERO (CRD420251229732) and reported according to PRISMA 2020. PubMed, Scopus and EBSCOhost (MEDLINE and CINAHL) were searched to 4 October 2026, with no supplementary searching. Observational studies of adult ICU patients in which both scores were recorded and mortality was reported were eligible; studies published more than ten years before the search and studies whose only outcome was functional status were excluded. Risk of bias was assessed with the QUIPS tool. Studies whose AUCs were reported inconsistently across the abstract, text and tables, or derived from repeated ratings of the same patients, were excluded from the primary analysis and examined separately. AUCs were pooled on the logit scale with a random-effects model (REML, Hartung–Knapp–Sidik–Jonkman confidence intervals). The within-study paired difference in AUC (FOUR minus GCS) was pooled using an assumed correlation between the two AUCs (r = 0.5, 0.7 and 0.9), because no study reported it. Heterogeneity was quantified with I-squared, tau-squared and a prediction interval. Results: Eleven studies (2016–2022; 55 to 300 patients; 1,365 patients in total) from Iran (five), India (three), Taiwan, Poland and Nepal were included. Six studies (712 patients, 192 deaths) entered the primary analysis. The pooled AUC was 0.80 (95% CI 0.77 to 0.83) for the GCS and 0.85 (95% CI 0.81 to 0.88) for the FOUR score, with no detectable heterogeneity (I-squared 0% for both). The pooled within-study difference was +0.053 in favour of the FOUR score (95% CI +0.023 to +0.082; p = 0.006; assumed r = 0.7), with a 95% prediction interval of +0.021 to +0.084 that excludes zero. Leave-one-out analysis left the difference between +0.043 and +0.057 and significant throughout. One choice changed it: restoring the study that had been set aside for inconsistent reporting returned +0.036 (95% CI −0.008 to +0.080; p = 0.094). Five studies did not enter the primary analysis, five studies were at high risk of bias and six at moderate risk, and none was at low risk. Conclusion: In adult ICU patients the FOU score discriminated mortality better than the GCS, by about 0.05 in AUC. The difference was consistent across studies and statistically significant in the primary analysis, but it disappeared when one study with inconsistent reporting was restored, and it comes from small studies at moderate to high risk of bias. The margin is too narrow, and the evidence too fragile, to justify replacing the GCS in routine practice.