Potential age bias in the NEO-PI-R: differential item functioning in older versus younger adults.

Objectives. Geriatric researchers and clinicians are often confronted with a lack of valid personality measures for older age groups, which hampers a reliable assessment of personality in later life. However, an age-neutral measurement system is one of the basic conditions for an accurate personality assessment across the lifespan. Therefore, we empirically investigate the age-neutrality of the NEO-PI-R, by examining potential differential item functioning (DIF) for older versus younger adults.
Methods. In order to detect possible DIF as a function of age the Mantel Chi-square, the Liu-Agresti Cumulative Common Log-Odds Ratio (L-A Lor), and the Cox's Noncentrality Parameter Estimator (Cox's B) were employed for polytomous items, using DIFAS 5.0 (Penfield, 2007).
Results. In general, all 5 NEO-PI-R domain scales and 24 of the 30 facets exhibited some DIF. The initial Mantel Chi-square test of all 240 NEO-PI-R items revealed DIF at a stringent cut-off (p <.00035) for 73 items (30,4%). The L-A Lor and the Cox's B statistics only confirmed large DIF for 47 items (19.6%).
Conclusions. Overall, results indicate that the majority (80.4%) of the NEO-PI-R items were similarly endorsed by younger and older age groups having the same degree of the underlying personality trait, corroborating the NEO-PI-R's age neutrality. However, 47 items did reveal large DIF between the two age groups. Results will be discussed in terms of their implications for using the current format of the NEO-PI-R in the elderly.