Method: Data from the 2005-2006 NHANES were analyzed for this study. A total of 1883 (Mean Age 62.37 ± 14.67) participants provided valid responses to the 20-item PFQ. The four-category response options are 1 = “No Difficulty”, 2 = “Some Difficulty”, 3 = “Much Difficulty”, and 4 = “Unable to do.” Data analysis was conducted using the WINSTEPS Rasch measurement software. The effectiveness of the category structure was evaluated using 4 criteria: (1) Did the average measure advance with each category? (2) Was the mean square residual (<2) appropriate? (3) Were the boundaries between categories ordered? (4) Were the discrimination (separation) indexes appropriate?
Analysis/Results: The original four-category structure (1234) did not function well: the mean square residual exceeded the acceptable level (>2). Through Rasch analysis and collapsing categories, six different category combinations (1223, 1233, 1123, 1222, 1122, and 1112) were developed and examined to determine the optimal categorization of the PFQ. For all six newly created category combinations, the average measure advanced with each category, the mean square residuals were < 2, and the boundaries between categories were ordered. Collapsing categories 2, 3, and 4 together (1222) and collapsing categories 2 and 3 together (1223) were preferred over other combinations with better item and person separation indexes.
Conclusions: The current PFQ with four categories may not be effective. Although the two-category structure (1222) was superior, a three-category structure (1223) may be preferred to discriminate participants with functional limitations versus those with severe functional limitations who may be unable to perform the selected task. Further investigation is needed to provide confirmatory cross-validation evidence of the PFQ with a two- or three-category structure.