Interrater reliability of Glasgow Coma Scale scores in the emergency department

• Published in: Annals of emergency medicine 2004-02, Vol.43 (2), p.215-223
• Main Authors: Gill, Michelle R, Reiley, David G, Green, Steven M
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Aged; Anesthesia. Intensive care medicine. Transfusions. Cell therapy and gene therapy; Biological and medical sciences; Emergency Medicine; Emergency Service, Hospital; Female; Glasgow Coma Scale; Humans; Intensive care medicine; Male; Medical sciences; Medical Staff, Hospital; Middle Aged; Observer Variation; Prospective Studies; Reproducibility of Results; Single-Blind Method; Statistics, Nonparametric
• ISSN: 0196-0644; 1097-6760
• DOI: 10.1016/S0196-0644(03)00814-X

Emergency physicians often use the Glasgow Coma Scale (GCS) to help guide decisions in patient care, yet the reliability of the GCS has never been tested in a typical broad sample of emergency department (ED) patients. We determined the interrater reliability of the GCS between emergency physicians when adult patients with altered levels of consciousness are assessed. In this prospective observational study at a university Level I trauma center, we enrolled a convenience sample of ED patients older than 17 years who presented with an altered level of consciousness. Two residency-trained attending emergency physicians independently assessed and recorded the GCS score and its components (eye, verbal, and motor) in blinded fashion within a 5-minute period. Data were analyzed for interrater reliability by using standard ordinal calculations. We also created scatter plots and Bland-Altman plots for each GCS component and for the GCS score. One hundred thirty-one patients were screened and enrolled in the study, with 15 excluded because of protocol violations. Of the 116 remaining patients, the agreement percentage for exact total GCS was 32% (τ-b=0.739; Spearman ρ=0.864; Spearman ρ 2=75%). Agreement percentage for GCS components were eye 74% (τ-b=0.715; Spearman ρ=0.757; Spearman ρ 2=57%), verbal 55% (τ-b=0.587; Spearman ρ=0.665; Spearman ρ 2=44%), and motor 72% (τ-b=0.742; Spearman ρ=0.808; Spearman ρ 2=65%). Our Spearman's analyses found that only approximately half (44% to 65%) of the observed variance could be explained by the relationship between the paired component measures. For GCS components, only 55% to 74% of paired measures were identical, and 6% to 17% of them were 2 or more points apart. We found only moderate degrees of interrater agreement for the GCS and its components.