In an effort to establish a reliable and valid measure of mental illness stigma among adolescents, the investigators examined the underlying factor structure of the r-AQ and reliability of this factor structure. The EFA yielded a single factor structure that consisted of five items, which the investigators call the modified r-AQ. The modified r-AQ was internally consistent. The CFA confirmed the single factor structure identified in the EFA and supported construct validity of the modified r-AQ; the single factor structure had excellent model fit to these data. The five item modified r-AQ measures a dimension of mental illness stigma, which the investigators called emotional reaction to people with mental illness, and is based on the work of Link, Yang, Phelan, and Collins (Link et al., 2004
) and Angermeyer, Matschinger, and Riedel-Heller (1999)
. Emotional reaction is the affective response of the person who is stigmatizing (Link et al., 2004
) and consists of aggressive emotions (anger), pro-social reaction (help, sympathy), and feelings of anxiety (scared) (Angermeyer et al., 1999
). Item 8 describes feelings of anxiety, items 3, 4, and 7 describe pro-social reaction, and item 6 describes feelings of aggressive emotions. Emotional reactions to people with mental illness are important because they are one of the strongest predictors of discrimination and social distance from persons with mental illness (Angermeyer, Holzinger, & Matschinger, 2010
). Emotional reactions also communicate to the stigmatized person how he or she is perceived by others (Angermeyer et al., 2010
). Emotional reactions in interpersonal interactions may verify the misconceptions of mental illness held by the public and shape future behaviors of avoidance and social withdrawal from persons with mental illness (Link et al., 2004
According to Link, et al., (Link et al., 2004
) stigma is a multidimensional construct that includes the domains of behavior, labeling, stereotyping, cognitive separating, emotional reactions, status loss, structural discrimination, and behavioral responses. Although the EFA supports construct validity of the r-AQ, only a third of the variance was captured, which suggests there may be multiple dimensions of the construct. It is possible that items 2 and 5 may demonstrate validity with other samples or when included with more items that enhance the conceptual fit of the items to emotional reaction. However, the EFA did not support the retention of these items in this sample of adolescents. Because this instrument has not been evaluated in previous studies with adolescents, we cannot compare our findings to others. This is first investigation of the r-AQ among adolescents; it is possible that more items are needed to adequately measure feelings of aggressive emotion and pro-social reaction, and the reduction of items to five items is insufficient to fully capture the construct. Since the r-AQ is based on attribution theory, and has evidence to support validity in adults, items 2 and 5 should be considered for inclusion in future studies. The inclusion of items 2 and 5, along with new items that are theoretically consistent with emotional reactions, may better capture the multidimensional construct of stigma and consequently enhance the r-AQ. Further instrument development is needed to encompass these domains in adolescent populations.
Several limitations should be considered in interpretation of these results. First, this study employed a convenience sample from three high schools in the southern area of the United States. Adolescents in this study may not be representative of all adolescents, and future studies should examine the r-AQ in other populations to enhance generalizability of study findings. Second, although the samples size was 20 participants per item and exceeded conventional standards for adequate sample size, (Costello & Osborne, 2005
) the majority of participants in this study were Caucasian adolescent girls of moderate socioeconomic level. Future studies should include more adolescent boys and subset populations of adolescents that are of lower socioeconomic level and ethnically diverse and aim to explain more variance with the addition of items. Finally, instrument stability is not known. Longitudinal data is needed to determine if the factor model is invariant over time.