Clinical library · Outcome measures
GMFM and GMFCS
Two instruments constantly confused with each other. One measures change and one classifies ability — and the systematic review says in terms that the classification should not be used to detect change.
In one line. The Gross Motor Function Measure is a criterion-referenced observational measure of how much gross motor function a child has, across five dimensions from lying and rolling to walking, running and jumping. The Gross Motor Function Classification System is a five-level classification of how a child with cerebral palsy usually moves, based on self-initiated movement with emphasis on sitting and walking.
The GMFM is an outcome measure. The GMFCS is a description of who the child is. Using one for the other's job is the commonest error made with both.
The GMFCS should not be used to detect change. That is the explicit conclusion of the COSMIN systematic review and meta-analysis of 44 articles: strong positive evidence for reliability and construct validity, moderate positive evidence for measurement error, an unknown level of evidence for responsiveness, and the plain statement that these classification systems should not be used to detect change. [1] If your GMFCS level moves, the usual explanation is reclassification, not treatment effect.
The numbers you actually need
| Property | Value | Source and caveat |
|---|---|---|
| GMFM-88 relative reliability | ICC 0.952 to 1.000 | Prospective cohort in children with cerebral palsy [2] |
| GMFM-88 measurement error | SEM 1.60; smallest real difference 3.14 | Total score. Below about 3 points, an individual change is inside the error [2] |
| GMFM-88 responsiveness | Effect size rose 0.5 to 0.8 (GMFCS I–II) and 0.4 to 0.9 (GMFCS III–V) across three follow-ups | Standardised response means 1.3 to 2.0 and 1.5 to 2.0 respectively [2] |
| GMFM as a change measure | Useful, on moderate-to-high quality studies | 21 studies from 62 identified; ages 10 months to 16 years; evidence levels II to IV, and populations of convenience [3] |
| GMFCS reliability and construct validity | Strong positive evidence | 44 articles, 37 in meta-analysis. Content validity: unknown level of evidence. No study investigated cross-cultural validity [1] |
| GMFCS stability over time | 85.9% stable; weighted kappa 0.91 (95% CI 0.89 to 0.92) | 1,670 children, mean age 11 years 4 months. Likelihood of change was associated with initial level only [4] |
| Stability across repeated assessment | 56% received the same rating at every assessment | 736 children, 7,922 assessments, median 11 ratings each by a median of three assessors. Changes were often transient [5] |
What they measure
The GMFM-88 scores 88 items across five dimensions: lying and rolling; sitting; crawling and kneeling; standing; and walking, running and jumping. The GMFM-66 is a Rasch-derived subset of 66 items giving interval-level scoring, at the cost of being unusable in children who cannot be scored on enough items.
The GMFCS assigns one of five levels describing usual performance, not best capacity, with separate descriptors for each age band. Level I walks without limitation; level V is transported in a manual wheelchair. It is a stable characteristic, which is exactly what makes it useful for prognosis and useless as an outcome.
Where it misleads
1. Treating a GMFCS change as a treatment effect
The classification is designed to be stable and is: 85.9% of 1,670 children maintained the same level, with a weighted kappa of 0.91. [4] In a registry of 736 children with 7,922 assessments, 56% had the identical rating at every single assessment, and where changes occurred they were often transient. [5] Combined with the review's finding that responsiveness has an unknown evidence level and that these systems should not be used to detect change, [1] a shift in level is more likely to be measurement variation than improvement.
2. Reading GMFM change without its measurement error
The smallest real difference for the GMFM-88 total is 3.14 points with a standard error of measurement of 1.60. [2] Small percentage gains reported in intervention studies routinely sit below that threshold for an individual child, even when the group difference is statistically significant.
3. The GMFM evidence base is thinner than its ubiquity implies
The review of its ability to detect change found 21 eligible studies from 62 identified, with designs at evidence levels II to IV, and describes them as using participant populations of convenience. [3] The instrument is effective as far as this goes; "as far as this goes" is a small and heterogeneous literature spanning ages 10 months to 16 years.
4. Cross-cultural validity has not been studied at all
The COSMIN review states that no studies investigated cross-cultural validity for the GMFCS and related classifications, and that content validity carried an unknown level of evidence. [1] For an instrument used worldwide, including across Indian services, that is a real gap rather than a technicality.
5. Which level a child sits at changes what responsiveness to expect
Effect sizes across three follow-up assessments rose from 0.5 to 0.8 in children at GMFCS levels I and II, and from 0.4 to 0.9 at levels III to V. [2] The instrument behaves differently across the classification, so a single expectation of change across a mixed caseload will misjudge both ends.
What the evidence supports — and what it does not
Supported
- GMFM-88 as a reliable measure of gross motor function, ICC 0.952 to 1.000. [2]
- GMFM-88 and GMFM-66 to detect change in children with cerebral palsy undergoing intervention. [3]
- GMFCS for classification and prognosis, with strong evidence for reliability and construct validity. [1]
- GMFCS as predictive of long-term motor function for most children. [4][5]
Not supported
- Using the GMFCS to detect change. Stated explicitly. [1]
- Interpreting GMFM-88 change below about 3 points in one child. Smallest real difference 3.14. [2]
- Assuming cross-cultural validity. Never studied. [1]
- A single expectation of GMFM change across all GMFCS levels. [2]
- Reading a one-level GMFCS shift as clinical improvement. Changes are often transient. [5]
How certain is this?
Evidence grade: Moderate.
The classification evidence is strong where it has been studied: a COSMIN systematic review of 44 articles with 37 meta-analysed, grading reliability and construct validity as strong positive. [1] Its honesty about what is unknown — content validity, responsiveness, cross-cultural validity — is what makes it usable.
The stability findings are large and consistent: 1,670 children in one registry [4] and 736 children with 7,922 assessments in another. [5] These are among the better-evidenced claims in this whole section.
The GMFM reliability and responsiveness figures come from a single prospective cohort, [2] and the change-detection review from 21 heterogeneous studies of moderate to high methodological quality using convenience populations. [3]
What would change the grade: cross-cultural validation, and GMFM minimal clinically important difference values derived by level rather than pooled across the classification.
Common questions
Can I use the GMFCS as an outcome measure?
No. The systematic review states that the GMFCS and related classification systems should not be used to detect change, and rates the level of evidence for responsiveness as unknown. [1] It is a classification of usual performance, deliberately stable — 85.9% of children keep the same level over years. [4] Use the GMFM for change and the GMFCS for describing and predicting.
How much GMFM change is real in one child?
More than about 3 points on the GMFM-88 total. The standard error of measurement was 1.60 and the smallest real difference 3.14. [2] Note also that responsiveness differs by classification level: effect sizes rose from 0.5 to 0.8 at GMFCS levels I to II and from 0.4 to 0.9 at levels III to V across three assessments.
GMFM-88 or GMFM-66?
Both have been shown useful for detecting change in children with cerebral palsy undergoing intervention. [3] The GMFM-66 gives interval-level scoring from a Rasch-derived item subset; the GMFM-88 retains all items and is usable in children who would score too sparsely for the 66-item version. The measurement-error figures quoted on this page are for the GMFM-88.
My patient's GMFCS level changed. What does that mean?
Most often, that the classification was applied slightly differently. In a registry of 736 children with 7,922 assessments by a median of three different assessors, only 56% had the same rating every time and changes were often transient. [5] In a separate cohort of 1,670 children, the likelihood of change was associated with initial GMFCS level only. [4] Look for a consistent, sustained change across assessors before treating it as real.
Do these tools apply to children in India?
Their reliability and construct validity are strongly evidenced, but the review found that no studies investigated cross-cultural validity, and content validity for the GMFCS carried an unknown level of evidence. [1] Use them, and be aware that the cross-cultural work has not been done. See paediatric physiotherapy for the intervention evidence.
References
- Piscitelli D, Ferrarello F, Ugolini A, et al. Measurement properties of the Gross Motor Function Classification System, Gross Motor Function Classification System-Expanded & Revised, Manual Ability Classification System, and Communication Function Classification System in cerebral palsy: a systematic review with meta-analysis. Developmental Medicine & Child Neurology. 2021 Nov;63(11):1251–1261. doi:10.1111/dmcn.14910 PMID 34028793 Systematic review and meta-analysis (COSMIN)
- Ko J, Kim M. Reliability and responsiveness of the gross motor function measure-88 in children with cerebral palsy. Physical Therapy. 2013 Mar;93(3):393–400. doi:10.2522/ptj.20110374 PMID 23139425 Prospective cohort study
- Alotaibi M, Long T, Kennedy E, et al. The efficacy of GMFM-88 and GMFM-66 to detect changes in gross motor function in children with cerebral palsy (CP): a literature review. Disability and Rehabilitation. 2014;36(8):617–27. doi:10.3109/09638288.2013.805820 PMID 23802141 Literature review
- Huroy M, Behlim T, Andersen J, et al. Stability of the Gross Motor Function Classification System over time in children with cerebral palsy. Developmental Medicine & Child Neurology. 2022 Dec;64(12):1487–1493. doi:10.1111/dmcn.15375 PMID 35941090 Registry cohort study
- Alriksson-Schmidt A, Nordmark E, Czuba T, et al. Stability of the Gross Motor Function Classification System in children and adolescents with cerebral palsy: a retrospective cohort registry study. Developmental Medicine & Child Neurology. 2017 Jun;59(6):641–646. doi:10.1111/dmcn.13385 PMID 28083887 Retrospective cohort registry study
About this resource
Using this in clinic
Every figure here is traceable to its source.
This page keeps the measure and the classification apart throughout, because the single most common error with these two instruments is using one to do the other's job. Where a value could not be verified against the paper it came from, it is not on this page, and the omission is stated rather than filled with a number from a secondary source.