Glasgow Coma Score

I believe that the modelling of the GCS needs updating.  Currently there are the following terms:

386554004 | Glasgow coma scale (assessment scale) |

68664003 | Glasgow coma testing and evaluation (procedure) |

3386557006 Glasgow coma scale finding (finding)

281395000 | Glasgow coma score eye opening subscore (observable entity) |

281397008 | Glasgow coma score verbal response subscore (observable entity) |

281396004 | Glasgow coma score motor response subscore (observable entity) |

There are even a whole series of coding for each score

248241002 | Glasgow coma score (observable entity) | score out of 14

444323003 | Modified Glasgow coma score (observable entity) | score out of 15  etc

However I cannot find a codes to code all the individual findings ?observations, qualifiers to the components below:

Eyes Opening

• Spontaneous  5054005 | Spontaneous (qualifier value) |
• To sound    247286002 | Finding of response to sound (finding) |  ?247297002 | Localization of sound source (observable entity) 
• To pressure --can find motor response only 299822001 | Motor response to deep glabella pressure present (finding) |
• None  ??260413007 | None (qualifier value) |

Verbal Response

• Orientated  426224004 | Oriented to person, time and place (finding) |
• Confused 40917007 | Clouded consciousness (finding) |
• Words 288612001 | Does use words (finding) |
• Sounds ?? 286256004 | Does make speech sounds (finding) | Is speech sounds intelligible? no grunt or groan sound codes except for breathing
• None  ? 260413007 | None (qualifier value) |

Motor Response

• Obey commands  248239003 | Response to commands (observable entity) |
• Localising in response to pain or sounds
• Normal flexion  in response to pain or sounds 123622004 | Normal flexion (finding) |
• Abnormal flexion   in response to pain or sounds 123623009 | Abnormal flexion (finding) |
• Extension  ?19212008 | Abnormal extension (finding) |
• None 260413007 | None (qualifier value) |

Steven, This is an area that I always find difficult as to the level of detail that could / should be modelled for assessment scales. The SNOMED editorial guide Staging and Scales doesn't address this in any great detail.  Agree that there are not all the clinical finding descriptions that relate to particular component scores of GCS.   Some help needed from the SNOMED authors and modellers here.

Andrew

I think the expectation has previously been that if you wanted to document the overall GCS score, then you should use:

EITHER

248241002 Glasgow coma score (observable entity)
plus an associated numeric value in the information model

OR

444323003 Modified Glasgow coma score (observable entity)
plus an associated numeric value in the information model

OR

26394007 Glasgow coma scale, 3 (finding)
112110007 Glasgow coma scale, 4 (finding)
74957005 Glasgow coma scale, 5 (finding)
80072008 Glasgow coma scale, 6 (finding)
18136007 Glasgow coma scale, 7 (finding)
32856008 Glasgow coma scale, 8 (finding)
5999000 Glasgow coma scale, 9 (finding)
1184008 Glasgow coma scale, 10 (finding)
61102007 Glasgow coma scale, 11 (finding)
91234001 Glasgow coma scale, 12 (finding)
54185009 Glasgow coma scale, 13 (finding)
26734006 Glasgow coma scale, 14 (finding)
70040003 Glasgow coma scale, 15 (finding)

...which explicitly (in its modelling) states that it relates only to a vanilla 248241002 GCS and not to a modified 444323003 GCS (and so the option does not exist for recording a modified GCS as a finding precoordinated with the numeric value).

The option of recording a vanilla GCS as a finding precoordinated with the numeric value should, however, be noted to be suboptimal in many respects: firstly, the numeric value is not easy to extract and so charting the timetrend for the score will be challenging. Secondly, and for the same reason, there is no simple machine transform that could compute that the two FHIR Observation expressions:

1. Observation.code=248241002, Observation.value.valueQuantity=9
2. Observation.code=5999000

...are semantically equivalent. For that reason I think I would personally recommend either deprecation of that precoordinated Finding pattern with SI moving to inactivate that content, or SI adding the currently missing concrete domain modelling that would make exactly that kind of bidirectional transform and equivalence detection possible.

Correspondingly, if you also want to document the individual subscores that contributed to that total, you should use only:

281395000 Glasgow coma score eye opening subscore (observable entity)
281396004 Glasgow coma score motor response subscore (observable entity)
281397008 Glasgow coma score verbal response subscore (observable entity)
plus associated numeric values in the information model

Of course, this representation of the subscore requires that all readers will know that e.g. "GCS eye opening subscore =1" means "subject does not open eyes either spontaneously or in response to ANY stimulus".

An alternative encoding that I can imagine folk attempting, and which I think Stephen's query echoes, is therefore to replace the numeric values with human readable strings, along the lines of:

Observation.code=281395000 Glasgow coma score eye opening subscore (observable entity)
Observation.value.valueCodeableConcept=255411008 Spontaneous event (qualifier value)

This approach is risky for a variety of reasons:

As in the case of 'spontaneous' above, for which two clinically interchangeable codes already exist, I think you'll rapidly find that more than one SNOMED code plausibly corresponds to the particular subscore 'values' you really want to capture and interoperate. And so, again, it will become challenging to reliably determine when two clinicians have said the same thing but in different ways, and so reliable interoperation fails.

You'll also have the problem of curating a 'map' between all possible adjectival values that clinicians may choose to select as a descriptive human readable representation of their underlying clinical assessment result, and the subscores in their abstracted numeric form as would be required if you're going to automatically compute the total GCS score from what the clinician's were persuaded or empowered to enter.

Finally, the 'none' option is especially problematic, as more generally is all negation. The following three are just some of the ways that a clinician might elect to record some aspect of a GCS assessment:

Observation.code=281396004 Glasgow coma score motor response subscore (observable entity)
Observation.value.valueCodeableConcept=260413007 None (qualifier value)

Observation.code=281396004 Glasgow coma score motor response subscore (observable entity)
Observation.value.valueCodeableConcept=299823006 Motor response to deep glabella pressure absent (finding)

Observation.code=298336006 No motor response to command (finding)

...but working out which are clinically equivalent, and what the actual subscore numeric value for each should be, is "challenging".

My recommendation would be to restrict clinicians (and clinical systems) to encoding GCS subscores only directly to the numeric. Onscreen data captured templates for use by clinicians can display more human readable prompts e.g. 'click here if moves limbs purposefully to remove painful stimulus of deep pressure to sternum' but that should be coded, and messaged, behind the scenes to:

Observation.code=Glasgow coma score motor response subscore (observable entity)
Observation.value.valueQuantity=5
Observation.value.valueString="moves limbs purposefully to remove painful stimulus of deep pressure to sternum"

..where the valueString can be different between clinical systems.

Following Jeremy's detailed comments and proposals and the discussions at the last Anesthesia CRG , some proposals to try and move forward the issue of modelling clinical descriptions in the Glasgow Coma Scale that enable recording of the subscore components.

This proposal requires consideration by the Modelling Advisory Group and updating and enhancement of the SNOMED editorial guide as regards assessment scales.

The test case being used by the Anesthesia CRG to establish modelling guidance is the Cormack Lehane score that assesses the view of the larynx and associated structures at direct laryngoscopy and has a correlation with the difficulty of endotracheal intubation.

The agreement reached was that for each scale point, the clinical finding would be included as one of the synonyms and should be the preferred term. These terms have been modelled and are now included in the SNOMED daily build.

Example:

416876007      F: Cormack and Lehane grade 3 (finding)

                        S: Cormack and Lehane grade 3

                        D: Cormack and Lehane grade 3 - epiglottis visible, vocal cords not  visible at direct laryngoscopy

The following seeks to extend the same methodology to the Glasgow Coma Scale for enabling clinicians to record GCS subscore findings. It is well known that clinicians may frequently know the precise top and bottom point of a scale or scale component, but be uncertain of intermediate points, hence need for clinical terms and/or on screen prompts to enable correct recording. Jeremy argues that for subscore components, this should be done using the subscore observable entity terms + a numeric value (plus the use of onscreen prompts/definitions if needed which would not be encoded as clinical concepts)

The argument of whether to use a purely numerical approach with the observable entity terms requires confirmation by the Modelling Advisory Group.

The approach suggested by the Anesthesia CRG (using the test scale methodology) and to help meet Steven's suggestions for more explicit description of the clinical components is additional finding terms as illustrated below and would rely on an implementation model to calculate scores, with inclusion of the numeric score in the concept being deprecated.

The reference source for these terms illustrating possible modelling of eye opening component of the Glasgow coma scale is https://www.glasgowcomascale.org/downloads/GCS-Assessment-Aid-English.pdf?v=3

This is the latest version of the clinical aid published by the Institute of Neurological Sciences, Glasgow  ( the originating institution of the Glasgow Coma Scale)

386557006      F: Glasgow coma scale finding (finding)

                       F: Glasgow coma scale eye opening component (finding)

                        F: Glasgow coma scale spontaneous eye opening (finding)

                        D: Glasgow coma scale eye opening before stimulus

                        S: Glasgow coma scale eye opening 4 (deprecated)

                        F: Glasgow coma scale eye opening to sound (finding)

                        S: Glasgow coma scale eye opening to command

                        D:Glasgow coma scale eye opening after spoken or shouted request                      

                        S:Glasgow coma scale eye opening 3  (deprecated)

                        F: Glasgow coma scale eye opening to pressure (finding)

                        D: Glasgow coma scale eye opening after finger tip stimulus *

                        S: Glasgow coma scale eye opening to pain (deprecated)

                        S: Glasgow coma scale eye opening 2  (deprecated)

* Accepted sites are finger tip, trapezius and supraorbital notch - ?write into D term

299823006 | Motor response to deep glabella pressure absent | is now probably obsolete as a clinical term for Glasgow coma scores as the defined methods now differ

                        F: Glasgow coma scale eye opening none (finding)

                        D: Glasgow coma scale no eye opening at any time, no interfering factor

                        S: Glasgow coma scale eye opening 1  (deprecated)

                        F: Glasgow coma scale eye opening not testable (finding)

                        D: Glasgow coma scale eye opening - eye closed by local factor

If this modelling approach is endorsed we can proceed and complete the other components, as there remains a clinical requirement in many circumstances to be aware of the component scores as well as the total Glasgow coma score.

Additional points:  248241002 | Glasgow coma score (observable entity) - the definition specifies that this is the original 1974 publication which had a maximum score of 14 (abnormal flexion was subsequently added to the motor response component).  I think that virtually everywhere uses the 3 to 15 scale now - It was in use at the Institute of Neurological Sciences in Glasgow in the early 1980's when I worked there ).  Although there is 444323003 Modified Glasgow coma score - I am unaware of any common usage of this although of course it remains technically correct

There is a lot to be considered here as it appears that a wholesale change to the way SNOMED represents the Glasgow coma score elements is desired.  It is not an issue for the Modeling Advisory Group, but rather the Editorial Advisory Group, which is responsible for setting editorial policy regarding addition and changes of content.  My suggestion is that the Anesthesia group prepare a detailed proposal, agreed by all members of the group, and submit it for consideration by the EAG.  As Jeremy mentioned, there are many different ways that these data may be recorded, which leads to challenges in determining equivalence.  The impact of the Glasgow coma score discussion extends to assessment instruments in general and thus has a very broad impact on the way SNOMED addresses this area.  I would encourage the Anesthesia group to address the broader implications of their proposed approach to recording values of assessments.

GCS is one example of assessment scales in SNOMED CT that over the years has managed to ensure that only a human can really understand what is meant to be a machine processable terminology.  'unfortunately' I am on leave this week but can I suggest that any discussions involve nurses and AHPs as the principles laid down for GCS which is multidisciplinary including medicine, may well end up be applied to many other scales that are predominantly non-medical.  Many do not represent the clinical findings in an authority dependent way and we need to have an approach that is both scalable and interoperable between different methods of recording.

Jeremy Rogers,  After some time I am finally getting back to this as it has much wider implications as to what SI should include in the terminology relative to assessment scales.

WRT your comment: "

The option of recording a vanilla GCS as a finding precoordinated with the numeric value should, however, be noted to be suboptimal in many respects: firstly, the numeric value is not easy to extract and so charting the timetrend for the score will be challenging. Secondly, and for the same reason, there is no simple machine transform that could compute that the two FHIR Observation expressions:

1. Observation.code=248241002, Observation.value.valueQuantity=9
2. Observation.code=5999000

...are semantically equivalent. For that reason I think I would personally recommend either deprecation of that precoordinated Finding pattern with SI moving to inactivate that content, or SI adding the currently missing concrete domain modelling that would make exactly that kind of bidirectional transform and equivalence detection possible."

It is not essential and possible not desirable for numeric scores on assessments to be treated as concrete numbers as they are not actual numbers but ordinals, as you cannot meaningfully perform calculations using them.  Thus, we can create equivalence between these two terms by using the concept representation of ordinal numbers in the value of an observable and as a relationship value in the finding. 

So the immediate questions we have to address here is whether SNOMED should begin to instantiate the component values for assessment scales such as the GCS and Modified GCS.  Our position in the past has been to represent the scale, the procedure to administer the instrument and an Observable or a finding to represent the summary score.  I think we need the opinion of the EAG to determine whether we are going to assume this HUGE responsibility to represent all of the components of scales, not withstanding the issues related to IP of "reproducing" the assessment within SNOMED CT. 

Jim,

Many thanks for your latest helpful comments. Further to your previous post, I have given some thought to a discussion paper on modeling and representation of assessment scales in SNOMED CT to come to the EAG if agreed by the Anesthesia CRG.  The next meeting of the Anesthesia CRG is scheduled for Tuesday 24th November.  I have nearly finished a draft and will be sending it out to the CRG during the next couple of days.  I have tried to consider assessment scales in general, although examples I have used in the paper are most often from anaesthesia and critical care as those are areas with which I am most familiar. 

My thoughts seem to be be pretty well aligned to what you have written, I also agree with Zac's comments about the need for this to be addressed in a way that is applicable across specialties and for different health care professionals.  Thanks for moving forward the discussion in this area

Hi All,

I was just reviewing ISO/DIS 13972:2020 Health informatics — Clinical information models — Characteristics, structures and requirements for a ISO TC215 ballot.  Annex D of this draft standard uses the GCS as a modelling example.  I am not sure if I can share the draft standard in this forum.  It does however refer to the website:   GlasgowComaScale-v3.2(2020EN) - Zorginformatiebouwstenen (zibs.nl) which may be helpful in our work.