Method variance in scales due to negatively worded items

This study evaluated the impact of negatively worded items on psychometric model fit using secondary data from the Caregiver Provider Relationship Assessment (CPRA; Moore, 2012). The CPRA includes three subscales: Caregiver Affect (α = 0.88), Caregiver Communication (α = 0.92), and Caregiver Information (α = 0.85). A three-factor model aligned with these subscales showed poor fit (RMSEA = 0.13, CI = 0.12–0.14; CFI = 0.85; TLI = 0.82; SRMR = 0.10). A four-factor model that added a method factor for negatively worded items showed improved fit (RMSEA = 0.08, CI = 0.06–0.09; CFI = 0.95; TLI = 0.93; SRMR = 0.04). These findings suggest that if method effects from negatively worded items are not modeled, scale scores may be biased despite high reliability. Reference: Moore, C. D. (2012). The caregiver–provider relationship assessment: Measuring family caregivers’ perceptions of relationship quality with health care providers. Evaluation & the health professions, 35(1), 104-110.

Carolyn E. Pickering; Qiyun Liu; Henrietta Bennett