Abstract B16: Ethnic differences in the psychometric properties of social networks and cancer knowledge

Abstract

Abstract Introduction: Disparities in Latino colorectal cancer (CRC) screening are critical in explaining disparities in CRC mortality. However, the mechanisms of these disparities are unknown. Two hypothesized explanations involve social networks and cancer knowledge. Measuring these variables requires tools that have been developed from research conducted with Caucasians. It is unknown if social network and cancer knowledge measures represent the same meanings among Latinos as among Caucasians, or how these measures vary among Latinos who are foreign born or US born. This study sought to examine the psychometric properties of social network variables and cancer knowledge variables among Latinos (foreign born and US born) as compared to Caucasians. This investigation could improve our ability to prevent and reduce disparities Latino cancer screening by explaining factors that contribute to CRC screening behaviors. Methods: This study used data from the 2005 Health Information National Trends Survey (HINTS). HINTS is a population-based probability survey of health communication and information among US adults aged 18 and over. HINTS provides a diverse sample of ethnic minority participants (n=491 Latino, n=4,103 Caucasian) and allows for the exploration of differences in social networks and cancer knowledge variables. Factor analysis with tetrachoric correlations for binary variables was conducted on each of the social networks items and on each of the cancer knowledge items. Caucasian, and Latino (foreign born, and US born) results were compared to identify differences in social networks and cancer knowledge. Principle component analysis with varimax orthogonal rotation was used to extract factors. Factors were retained when eigenvalues met or exceeded 1.0. Statistical analyses were conducted using Stata 11SE. Results: Social Networks: For Caucasian respondents, factor analysis produced a single factor from the social network variables (1.27),χ2 = 262.31, p < .001, where all three variables loaded to the factor at or above .6. For Latinos, factor analysis produced a single factor from the social network variables (1.23), excluding marital status for both foreign born, and US born, Latinos. Cancer Knowledge: For Caucasian respondents factor analysis produced a single factor from the cancer knowledge variables (1.54), (χ2 =1.049.45, p < .001), where all three variables loaded to the factor at or above .60. Among Latino respondents a single factor also emerged (1.33), χ2 = 54.53, p < .001, however for both foreign, and US, born Latino participants, the exercise question did not load to the factor. The other cancer knowledge items loaded at or above .79. Conclusions: Our findings indicate that the social network variables did not represent the same concepts for Latinos as compared to Caucasians. Our findings suggest that the marital status item is unsuitable for inclusion in this social network scale when examining Latino social networks. This may mean that among Latinos, the marital status item is related to a different underlying construct than the other two social network variables. Our findings also indicate that the exercise item is unsuitable for inclusion in this cancer knowledge scale when examining Latino cancer knowledge. This may mean that Latinos have a different understanding of exercise as a cancer prevention behavior. It also means that the exercise item may be related to a different underlying construct than the other two variables. These findings are relevant to cancer prevention because they help to clarify how social networks and cancer knowledge vary by ethnicity. This improves our ability to use multiple social network and cancer knowledge items as a scale to explain and prevent disparities in cancer prevention among Latinos. Citation Information: Cancer Prev Res 2011;4(10 Suppl):B16.