Five retrospective studies (102,132 patients) were included in the review. Where reported, follow-up was at hospital discharge or in-hospital death. All studies scored 11.5 on the QUADAS tool; none of the studies ensured that the reference test was independent of the index test and none reported uninterpretable/intermediate test results.
Statistically significant differences in accuracy between the models for predicting mortality (Glasgow Coma Scale AUC 0.90, 95% CI 0.88 to 0.91, Simplified Motor Score AUC 0.87, 95% CI 0.86 to 0.88; p=0.01) suggested that the Glasgow Coma Scale had higher predictive value.
There were no statistically significant differences in accuracy between the Glasgow Coma Score and Simplified motor Score for predicting clinically significant brain injury (AUC 0.79 versus 0.75), neurosurgical intervention (AUC 0.83 versus 0.81) or emergency tracheal intubation (AUC 0.85 versus 0.82).
Significant statistical heterogeneity was evident for all outcomes.