Modeling the distribution and progression of motor ability among children with cerebral palsy: An analysis of three reference centile sets.

Background: Reference centiles describing gross motor function in children with cerebral palsy (CP) are used in clinical and research settings to guide treatments and evaluate interventions. However, it is unknown how existing references generalize to populations in novel settings. Aims: The aim of this study is to evaluate the cross-sectional and longitudinal performance of three reference centiles to describe the motor function of children with CP aged 2-12 years at a large urban US pediatric hospital through a retrospective observational study. Methods: and Procedures: We assessed cross-sectional performance by ranking our clinical population by quartile distributions described by the references. We assessed longitudinal performance by analyzing the distribution of prediction errors and correlations between predictions and observed scores. Outcomes and Results: For cross-sectional distribution, the reference centiles by Hanna more closely described our population than those by Duran. For longitudinal progression, prediction error was less than 6 GMFM points for most records at 24-, 12-, and 6-month time scales for all three sets of reference centiles, but higher at a 48-month time scale. Prediction errors increased at younger ages and higher motor ability. Conclusions and implications: Despite differences in cross-sectional performance, all three reference centiles achieved similar longitudinal performance and are sufficient for most clinical and research uses. Caution should be used when applying these curves to locations with different standards of care.