Sensitivity of a computer adaptive assessment for measuring functional mobility changes in children enrolled in a community fitness programme

Abstract

Objective: To examine the relative sensitivity to detect functional mobility changes with a full-length parent questionnaire compared with a computerized adaptive testing version of the questionnaire after a 16-week group fitness programme. Design: Prospective, pre- and posttest study with a 16-week group fitness intervention. Setting: Three community-based fitness centres. Subjects: Convenience sample of children (n = 28) with physical or developmental disabilities. Interventions: A 16-week group exercise programme held twice a week in a community setting. Main measures: A full-length (161 items) paper version of a mobility parent questionnaire based on the Pediatric Evaluation of Disability Inventory, but expanded to include expected skills of children up to 15 years old was compared with a 15-item computer adaptive testing version. Both measures were administered at pre- and posttest intervals. Results: Both the full-length Pediatric Evaluation of Disability Inventory and the 15- item computer adaptive testing version detected significant changes between pre and posttest scores, had large effect sizes, and standardized response means, with a modest decrease in the computer adaptive test as compared with the 161-item paper version. Correlations between the computer adaptive and paper formats across pre and posttest scores ranged from r = 0.76 to 0.86. Conclusions: Both functional mobility test versions were able to detect positive functional changes at the end of the intervention period. Greater variability in score estimates was generated by the computerized adaptive testing version, which led to a relative reduction in sensitivity as defined by the standardized response mean. Extreme scores were generally more difficult for the computer adaptive format to estimate with as much accuracy as scores in the mid-range of the scale. However, the reduction in accuracy and sensitivity, which did not influence the group effect results in this study, is counterbalanced by the large reduction in testing burden.