RESUMO
OBJECTIVES: To assess "alternate forms" reliability and inter-rater reliability of Gross Motor Function Classification System (GMFCS) scores. METHODS: Fifty randomly selected children with cerebral palsy were divided into two groups: (1) GMFCS score assigned during gait assessment ("GMFCS previously assigned") and (2) no GMFCS score assigned. Using database information, two physiotherapists independently determined GMFCS scores for 25 children from the "previously assigned" group, and 25 from the "no score assigned" group. Therapists compared their recently assigned scores for the "previously assigned" group, discussing discrepancies until attaining agreement. This group's consensus scores were compared to GMFCS scores assigned at time of actual assessment to calculate "alternate forms" reliability. RESULTS: Between-therapist agreements were kappa = 0.84 for "GMFCS previously assigned" group and 0.95 for "no GMFCS assigned" group. Kappa agreement between direct assessment and retrospectively assigned scores for the "GMFCS previously assigned" group was 0.79. CONCLUSIONS: Retrospective GMFCS scores can be reliably assigned.