(PS7-C69) Measurement Invariance of the OCI-12 Across Racial and Ethnic Groups

The OCI-12, an adaptation of the OCI-R (Foa et al., 2002), is a brief, theoretically consistent measure of obsessive-compulsive symptoms (Abramovitch et al., 2021). That is, the neutralizing and hoarding items were removed from the OCI-R, consistent with literature indicating that they do not represent distinct domains of obsessive-compulsive symptoms. As such, the OCI-12 proposes a 4-factor structure, with 3 items loading onto each of 4 subscales (obsessing, washing, checking, and ordering), and those subscales are proposed to load onto a single latent factor. Abramovitch et al. (2021) supported the 4-factor and higher-order structure of the OCI-12 items in both clinical and nonclinical samples. However, because their sample was 75.5% White, the authors called for examination of the structural and psychometric properties of the OCI-12 across diverse racial and ethnic groups. The major aim of the current study was to heed this call and offer psychometric evaluation of the OCI-12 in a large, racially diverse respondent sample. Specifically, the focus was on measurement invariance.

The sample was 5,815 students from a midwestern university. Participants self-reported their racial/ethnic identities as 8.3% Asian (n = 480), 28.8% Black (n = 1,676), 18.0% Hispanic (n = 1,045), and 45.0% White (n = 2,614). Each participant completed a battery of questionnaires, including the 18-item OCI-R. We used multiple-group confirmatory factor analysis (MGCFA; Meredith, 1993) to examine potential measurement invariance of the OCI-12 across the four groups. Following Hu and Bentler (1999), acceptable model fit was operationalized as CFI/TLI values > .90, and RMSEA/SRMR values < .08; good model fit was operationalized as CFI/TLI values > .95 and RMSEA/SRMR values < .05. Due to its sensitivity to large sample sizes (Brown, 2015), chi square was not used to assess model fit.

In CFAs run on each group, the 4-factor higher-order solution demonstrated good or acceptable fit on all indicators (CFI = .957-.976, TLI = .943-.968, RMSEA = .055-.072, SRMR = .027-.038). Full configural, metric, and scalar measurement invariance were supported by the MGCFA, as no significant decrease in model fit—defined per Chen (2007) as ΔCFI ≤ –.01 or ΔRMSEA ≥ .015—was observed as each additional component was constrained. This suggests that the OCI-12 total and subscale scores may be interpreted equivalently across these participant groups; namely, the latent structure, factor loadings, and item loadings are equivalent across groups. Additionally, observed mean differences among groups may be interpreted as being due to valid differences rather than measurement variance.

Our results support that the OCI-12 is psychometrically invariant across Asian, Black, Hispanic, and White Americans. This permits a conclusion that it is appropriate for use in research with racially/ethnically diverse nonclinical samples. Future examinations of the measure should be carried out in clinical samples, to assess its use in the diagnosis and treatment of OCD, and the appropriateness of the cut-off scores proposed by Abramovitch et al. (2021) across racial/ethnic groups.