Assessing the level of consciousness in children: a plea for the Glasgow Coma Motor subscore

The Glasgow Coma Scale (GCS) is not always easy to score and its reliability has been questioned. In adults the GCS Motor score has proven a valuable alternative, as it is easier to assess yet shows similar predictive capacity for outcome. We wanted to test the non-inferiority of the Glasgow Coma Motor score GCS-M versus the Total score GCS-T for predicting outcome in children.

As part of the Flemish paediatric trauma registry (PENTA) we collected data on 96 consecutive children (0-18 years) with moderate to severe traumatic brain injury. Outcome was evaluated using a three level ordinal scale: [normal to mild disability, moderate to severe disability and death]. A number of proportional odds models were fitted for various choices of predictive variables (GCS-T, GCS-M, age, sex, and injury severity score ISS). For each model we calculated Somers'D(xy) rank correlation and NagelKerke's R(2)N index, both measures of the predictive performance of the model.

All children had an injury to the brain that resulted in a hospital stay of more than 48h. Half of them had a "best" initial GCS of 15; 60%, a Motor score of 6. The median Injury Severity Score ISS was 16. Outcome was 'normal to mild' in 79 children, 'moderate to severe' in 7, and 'death' in 10. D(xy) values were 0.983 for the model with the Motor score and 0.972 for that with the total GCS, indicating excellent predictive performance for both. R(2)N indices were 0.862 and 0.813, respectively. Overall the difference between all models was small.

The GCS Motor subscore was shown to have at least the same predictive ability for outcome as the total GCS. It is our opinion that the total GCS is unnecessarily complicated (especially in children). Using the Motor score alone will improve scoring compliance and statistical performance. We do not believe that the reduction in number of potential scores from 13 to 6 would decrease the descriptive capacity significantly, since clinical algorithms typically group values of the total GCS into five or fewer ranges.