Analysis of the pediatric outcomes data collection instrument in ambulatory children with cerebral palsy using confirmatory factor analysis and item response theory methods

BACKGROUND: Traditional use of the Pediatric Outcomes Data Collection Instrument (PODCI) assumes that all items have the same structure, are measuring the intended constructs, and assess the right levels of function to show change after orthopaedic or neurological intervention. Item response theory (IRT) methods can statistically account for inherent differences in PODCI item characteristics and thus reveal attributes of the measure important to effectiveness research. Our study uses IRT methods to determine whether PODCI items fit the projected dimensional structure of the PODCI, assess function on each dimension at the right level for a population of ambulatory children with cerebral palsy (CP), and reveal changes after intervention in this population. METHODS: Proxy-reported PODCI questionnaires for 570 ambulatory children with CP were randomly divided into 2 groups for model creation and model testing using exploratory and then confirmatory factor analysis. The resulting model was compared with the projected dimensional structure, tested for fit of individual items, and examined for gaps and ceiling effects. Response changes at 1 year were compared between those with (n = 91) and without (n = 284) surgical intervention using paired t tests. RESULTS: Factor analysis reduced the projected dimensions from 5 to 4 for this population, resulting in dimensions for mobility, upper extremity function (UEF), comfort and general health, and self-worth. All but 3 items fit their respective dimensions; ceiling effects were noted in 3 dimensions. Responses showed changes in the comfort and general health, mobility, and UEF dimensions in those who had surgery; in those children who did not have surgery, only the UEF responses changed. CONCLUSIONS: The PODCI can show change after intervention when data are analyzed using IRT methods. Ceiling effects in 3 dimensions may limit the amount of change the PODCI can show in a population of ambulatory children with CP. LEVEL OF EVIDENCE: Level II. This was a retrospective investigation of a diagnostic tool, the PODCI, using a randomized cross-sectional design for model development, and a case-control design to assess sensitivity to change.