# Agreement Between the Gross Motor Ability Estimator-3 and the Reduced Gross Motor Function Measure-66 Based on Artificial Intelligence

## TL;DR

This study shows that a shortened version of a motor function test for children with cerebral palsy is highly accurate and saves time compared to the full version.

## Contribution

The study validates the rGMFM-66 using the updated GMAE-3 and compares agreement between GMFM-66 versions.

## Key findings

- The rGMFM-66 showed excellent agreement with GMFM-66v3 (ICC = 0.994).
- Administration time was reduced by 42% using the rGMFM-66.
- Bland–Altman plots confirmed close agreement across all comparisons.

## Abstract

Background: The reduced Gross Motor Function Measure-66 (rGMFM-66) has already demonstrated its validity compared to the standard GMFM-66 using the Gross Motor Ability Estimator-2 (GMAE-2). This study aimed to evaluate its validity using the updated Gross Motor Ability Estimator-3 (GMAE-3) and to compare agreement between GMFM-66v2 and GMFM-66v3. Methods: A retrospective analysis was conducted on 250 children with cerebral palsy (CP) enrolled in a rehabilitation program between 2015 and 2024. All GMFCS levels (I–V) were represented. The sample included 107 females and 143 males, with a mean age of 6.9 years (SD 3.4). Agreement between scoring methods was assessed using intraclass correlation coefficients (ICCs) and Bland–Altman analyses. Results: The rGMFM-66 showed excellent agreement with GMFM-66v3 (ICC = 0.994; 95% CI 0.992–0.996). Similar agreement was found between GMFM-66v2 andGMFM-66v3 (ICC = 0.994; 95% CI 0.991–0.996). Bland–Altman plots confirmed close agreement across all comparisons. The rGMFM-66 reduces administration time from 45 to 26 min, offering a 42% time saving in clinical use. Conclusions: The rGMFM-66 demonstrates very high agreement with GMFM-66v3 and appears to be a valid alternative. Its strong concordance supports its applicability in both clinical and research settings. Although agreement was high, minor differences between scoring methods indicate that results should be interpreted in light of the scoring algorithm applied.