Background: This study used item response theory (IRT) to create a brief PF scale (BPFS) and subsequently examined its relationship with several health characteristics. Methods: Data were used from N=1,716 adults 50+ years of age participating in a large health survey. A pool of 19 PF items were dichotomized to either 1 (any amount of difficulty) or 0 (no difficulty). A 2-parameter logistic (2PL) IRT model was used to evaluate item fit to the unidimensional PF construct. Criteria used to eliminate an item was 1) a small discrimination (slope) parameter, 2) a significant chi-square statistic for cell residuals, and 3) a large root mean square error of approximation (RMSEA). The IRT model was continually re-fitted until all remaining items met criteria. SAS PROC IRT and R ltm were used for scale construction. Results: The IRT analysis resulted in 8 well-fitting items with large item discrimination (as > 2.03), moderate item difficulty range (bs: -0.07 - 1.35), and adequate item fit (RMSEAs < .036). After full adjustment, each additional BPFS point significantly (ps < .05) increased stepwise the odds of reporting poor HRQOL (OR = 1.59), being depressed (OR = 1.46), having thoughts of suicide (OR = 1.35), not meeting PA guidelines (OR = 1.29), being BMI-obese (OR = 1.23), being WC-obese (OR = 1.13), experiencing poor sleep (OR = 1.29), and reporting sleepiness (OR = 1.16). Conclusion: Results from this study show that the IRT-constructed BPFS is an efficient and valid tool that can predict health status in older adults.

physical functioning, Item response theory (IRT), psychometrics, aging and health