Article Text

This article has a correction. Please see:


Development of a modified paediatric coma scale in intensive care clinical practice
  1. A Tatmana,
  2. A Warrena,
  3. A Williamsb,
  4. J E Powellc,
  5. W Whitehouseb
  1. aThe Birmingham Children’s Hospital NHS Trust: Department of Intensive Care, bDepartment of Paediatric Neurology, cDepartment of Public Health and Epidemiology, University of Birmingham
  1. Dr W Whitehouse, Department of Paediatric Neurology, Birmingham Children’s Hospital, Ladywood Middleway, Birmingham B16 8ET.


James’ adaptation of the Glasgow coma scale (JGCS) was designed for young children. Intubated patients are not allocated a verbal score, however, so important changes in a patient’s conscious level may be missed. A grimace score was therefore developed and assessed for use in intubated children.

 Two observers made a JGCS observation within 15 minutes of each other. One observer was the patient’s nurse and the other a trained investigator. Interobserver reliability was determined between the first and second observation for each component of the scale. Reliability was measured using κ and weighted κ statistics.

 Seventy three children had 104 sets of observations. Interobserver reliability was moderate to good for all components, with the grimace score better than the verbal score.

 It is concluded that the grimace score is more reliable than the verbal score and may be useful in intubated patients in whom the verbal score cannot be used.

  • coma scale
  • intensive care
  • conscious level

Statistics from

The Glasgow coma scale has been widely adopted in the management of adult and paediatric coma.1 2 It should not be used in small children as the verbal component is not appropriate.3 Several coma scores have been developed specifically for children in an attempt to compensate for their differences in verbal and motor capabilities.4-11 Three years ago, we introduced into our intensive care unit (ICU) a modified Glasgow coma scale, which is Sharples’ adaptation (personal communication) of the James’ adaptation of the Glasgow coma scale (JGCS) (table1).12

During this time, our nursing staff reported that many children who were intubated showed varying degrees of orofacial grimacing when stimulated. Therefore we developed a grimace score to replace the verbal component in intubated children. We report the results of a study to assess the reliability of our modified coma scale in this clinical setting.

Subjects and methods


After receiving local ethical committee approval, children on the ICU with coma from any cause were selected in a quasirandom manner: whenever one of the three trained investigators was available, the patient accessible, and the patient had not been studied within 24 hours nor with the same JGCS (on the routine nursing JGCS chart).

Verbal consent was obtained from parents when available. A set of observations consisted of two JGCS (table 1) scores, the second score being completed within 15 minutes of the first. These were performed sequentially by two observers, one being the child’s bedside nurse and the other being one of three trained observers. The observers were blinded to the preceding score. Children who were not intubated were given a verbal score. Children who were intubated were given a grimace score. We excluded children with cervical spinal cord injury, peripheral nerve disease, or neuromuscular disorders, including residual paralysis from neuromuscular blockade. The painful stimulus was nail bed pressure on both upper limbs, using a pencil. The best response was taken for the observation.

Table 1

Modified Glasgow coma scale. Pain as nail bed pressure with pencil; score best response


Interobserver reliability (E1–E2, V1–V2, G1–G2, M1–M2, and summated scores EVM1–EVM2 and EGM1–EGM2), that is, the level of agreement between the two observations, was measured by the κ and weighted κ statistics.13 While the κ statistic measures the level of agreement above that expected by chance, it does not take into account the degree of disagreement between observations. The weighted κ statistic measures agreement and takes into account the magnitude of the disagreement.

For both κ and weighted κ, strength of agreement is interpreted as < 0.2 = poor; 0.21–0.40 = fair; 0.41–0.6 = moderate; 0.61–0.80 = good; > 0.8 = very good or near perfect.


One hundred and four sets of observations were completed in 73 children of whom 42 were boys. Four children had severe orbital swelling and were not given an eye score. Forty one observers were involved (38 nurses and three trained observers). The children ranged in age from 1 day to 16 years (median age 73 days). Table 2 shows the diagnostic categories. Tables 3 and 4 show the raw data for each component and for summated scores using the grimace and verbal scores separately. Table 5 shows the interobserver reliability.

Table 2

Diagnostic category on admission to ICU

Table 3

Each pair of observations for each component of the adapted JGCS

Table 4

Each pair of observations for the summated adapted JGCS, with grimace in place of verbal

Table 5

Interobserver agreement


We adopted the JGCS because it takes account of developmental immaturity in small children, uses the same number of points irrespective of the child’s age, and is simple for the patient’s nurse to use without additional staff or equipment.

Several studies have examined the reliability of paediatric coma scales using two or three trained observers.5 10 14 This is useful for determining a scale’s experimental reliability, but may not necessarily translate into clinical practice.15 For example, in our ICU there are over 100 nurses with varying levels of experience. Therefore, any scale must be robust enough to produce reliable results given the observers who will be using it. Complicated scales, which are used relatively infrequently, are unlikely to be reliable.

Our results suggest that despite a large number of observers, there is moderate to good interobserver agreement for the components of this scale.

The grimace component appears to be more reliable than the verbal component. They may measure different aspects of brain function and cannot necessarily be equated clinically. Facial expression, however, is an important part of non-verbal communication, so facial grimace and verbal language are not totally independent skills. Furthermore, we believe that in intubated patients the restoration of a third variable (eye opening, motor, and grimace) in assessing coma increases the likelihood of detecting an improvement or a deterioration in the patient’s condition, particularly when the variables are measured independently.

We have included the summated values for interest. We do not summate the values clinically, as the variables have different weights and are not clinically comparable.16 17

Although the grimace score has not been validated for outcome, it is more reliable than the verbal score in this study and may be useful in intubated patients when the verbal score cannot be used.

This study has also shown the reliability of the other components of our adaptation of the JGCS when used by nurses and doctors in an ICU.


We are very grateful for the help of all the nurses at Birmingham Children’s Hospital ICU for their contribution to the study and their continuing support.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.