Interobserver reliability in the interpretation of three-dimensional gait analysis in children with gait disorders

To assess interobserver reliability in the interpretation of three-dimensional gait analysis (3DGA) of children with gait disorders within a single institution.

Seven experienced interpreters in our institution participated in a quality-assurance program reviewing one unique patient's 3DGA data every 3 months. Between 2014 and 2017, 15 patients' data were interpreted (14 with spastic cerebral palsy, 1 with myelodysplasia). Interpreters were asked to select 'yes', 'no', or 'indeterminate' from a list of problems and treatment recommendations. Kappa and percent agreement calculations were performed to evaluate consistency.

Average percentage agreement in problem identification and treatment recommendation was greater than 84 percent and 90 percent for all interpreters respectively. Average kappa for the 10 most consistently identified problems and recommended treatments were 0.69 and 0.59 respectively. Interpreter consistency was moderate or better for the most commonly performed operations at our institution (0.44-0.59). Sagittal plane abnormalities of the hip and knee had the highest consistency.

When institutional differences in data collection and regional variations in management philosophies are removed, interobserver consistency in 3DGA interpretation is moderate to substantial for many commonly selected items. Identification of areas with poor consistency may help address underlying causes and improve data processes.

Consistency in three-dimensional gait analysis interpretation and treatment recommendation is high within a single institution. There is moderate or better consistency for most commonly identified problems and recommended treatments. Sagittal plane problem identification of the hip and knee have the highest consistency. Lower consistency is seen in areas with poor objective measures, such as dystonia and balance.