Using artificial intelligence-based technologies to detect clinically relevant changes of gross motor function in children with cerebral palsy

Abstract

Aim: To compare the 66-item Gross Motor Function Measure (GMFM-66) with the reduced version of the GMFM-66 (rGMFM-66) with respect to the detection of clinically relevant changes in gross motor function in children with cerebral palsy (CP). Method: The study was a retrospective single centre analysis of children with CP who participated in a rehabilitation programme. Overall, 1352 pairs of GMFM-66 and rGMFM66 measurements with a time interval of 5 to 7 months were available. To measure clinically relevant changes in gross motor function, the individual effect size (iES) was calculated. Results: The study population consisted of 1352 children (539 females), mean age 6 years 4 months (SD 2 years 4 months). The iES based on the GMFM-66 and the rGMFM-66 showed a significant correlation (r = 0.84, p < 0.001). The analysis of the area under the receiver operating characteristic curve showed an excellent agreement for clinically relevant gross motor improvement (Cohen’s d ≥ 0.5; area under the curve = 0.90 [95% confidence interval 0.88–0.92]) or deterioration (Cohen’s d ≤ −0.5; area under the curve = 0.95 [95% confidence interval 0.92–0.97]). Interpretation: Performing the rGMFM-66 saves time compared to the full GMFM-66. The rGMFM-66 showed good agreement with the GMFM-66 with respect to the detection of clinically relevant changes in gross motor function in children with CP, so its use in everyday clinical practice seems justifiable. What this paper adds: The reduced version of the 66-item Gross Motor Function Measure (rGMFM-66) detects clinically relevant changes in gross motor function in children with cerebral palsy. The rGMFM-66 correlates highly with the full GMFM-66. The rGMFM-66 can be used in clinical practice when the time schedule is limited.