Confirmatory Factor Analysis of a Multidimensional Attitude Toward Condom Use Scale: Determinants of Condom Use Among Young Ghanaians

Background and Purpose: Much of the existing research on young people’s attitudes toward condom use relies on measurement scales developed primarily in North America and/or developed for the general adult population. Although latent constructs are applicable across contexts, the content of the constructs may be different from one setting to another and across various groups. Many of the existing scales have not been validated cross-culturally and tested with young people. Further, some scales are not informed by existing theories. Failing to exploit theory in developing a scale may provide a limited description of the phenomenon we are interested in. A measurement scale based on a well-elaborated theory may be better – particularly when a scale aims to assess risks and inform interventions - than an atheoretical scale. Given these limitations, we developed a condom use scale using the health belief model (HBM) (Rosenstock, Strecher, & Becker, 1988). This study aims to confirm the factor structure and quality of this new scale.

Methods:This study used the YouthSave Ghana experiment baseline data. 6,252 youth, ages 10 to 27, attending 100 schools from eight of Ghana’s ten regions constituted the sample. The analysis focused on a 21-item condom use attitude scale developed for the YouthSave project. We used confirmatory factor analysis (CFA) to test the relationships among hypothesized latent variables as supported by theory and prior research. Mplus was used to conduct CFA because of Mplus’ ability to appropriately handle characteristics of our data including clustering of students in 100 schools, missing data, and ordinal-level variables.   

Results: Multidimensional scales fit our data better compared to a one-dimensional condom scale. A one-dimensional scale did not meet any of the five predetermined fit criteria. When we tested two competing multidimensional scales, a first-order factor model with four HBM dimensions met three of five fit criteria (RMSEA = 0.041, 90% CI: 0.037 – 0.045; CFI = 0.96; TLI = 0.95). The other first-order factor model with two additional latent factors (self-efficacy and social support) met only one fit criteria (RMSEA = 0.035, 90% CI: 0.033 – 0.038). Results of the analyses using a validation sample supported the findings based on the model with a calibration sample. All factor loadings and variances were statistically significant (p< .001).

Conclusions and Implications: CFA findings suggest the condom attitude scale being used in the Ghana YouthSave project adequately measures the hypothesized dimensions, based on HBM, of latent variables. Our data support a hypothesized multidimensional scale that includes perceived benefits of and barriers to condom use, as well as perceived susceptibility and severity of HIV/AIDS. Consistent with prior research in North America (Brown, 1984; Helweg-Larsen & Collins, 1994), the results suggest that condom attitudes are multidimensional. Our data did not support a one-dimensional scale which sums all HBM dimensions into a single score. In order to predict or change condom attitudes or behavior among young people, social workers should use multidimensional scales to specify which specific components of condom attitude will be considered when assessing risks and informing programs.