ARS | Publication request: The Usefulness of Differential Item Functioning Methodology in Longitudinal Intervention Studies

Submitted to: International Society for Behavioral Nutrition and Physical Activity
Publication Type: Abstract Only
Publication Acceptance Date: May 24, 2008
Publication Date: May 24, 2008
Citation: Watson, K., Baranowski, T., Baranowski, J., Jago, R., Thompson, D.J. 2008. The usefulness of differential item functioning methodology in longitudinal intervention studies [abstract]. Proceedings, International Society for Behavioral Nutrition and Physical Activity, May 21-24, 2008, Banff, Alberta, Canada. p. 158-159.

Technical Abstract: Perceived self-efficacy (SE) for engaging in physical activity (PA) is a key variable mediating PA change in interventions. The purpose of this study is to demonstrate the usefulness of item response modeling-based (IRM) differential item functioning (DIF) in the investigation of group differences at the item level in longitudinal PASE. Secondary analyses was preformed on a sample of 473 boy scouts that participated in a randomized intervention consisting of either a 9-week Internet-based physical activity (intervention) or nutrition (control) program. A 19-item PASE scale was administered pre (T1) and post (T2) intervention. IRM-DIF analysis was employed to ascertain group differences in item interpretation over time. Longitudinal IRM-DIF analysis included group and time factors. IRM-DIF results did not yield a significant group or time main effect or group*time interaction. However, significant group *time* item interactions were observed for eight items, thus indicating item functioning between T1 and T2 was different among the groups. The magnitude of the DIF was small. The control group found 4 items (primarily dealing with barriers, e.g., friends, busy, and tired) easier to respond more favorably to at T1 and T2. The treatment group found 4 items (primarily referencing a time element, e.g., 15, 30, and 45 minutes) easier to respond more favorably to at T1 and T2. IRM-DIF analysis suggests some item interpretations changed over time; partial evidence suggests small DIF due to intervention participation. Research is needed to clarify whether observed DIFs were indicative of true change or real differences in interpretation.