United States Spanish Short-Form-36 Health Survey - Scaling assumptions and reliability in elderly community-dwelling Mexican Americans

Background: Measuring health status is challenging in Mexican Americans and other diverse groups, because most health measurement instruments were developed and tested in English. Thus, it is difficult to determine whether measured health disparities are the result of actual differences or due to measurement error resulting from translation or conceptual differences. Objectives: The purpose of this study was to test the metric equivalence of the United States (US) Spanish Short-Form-36 Health Survey (SF-36) in a group of elderly Mexican Americans. In addition, the SF-36 scores of elderly Mexican American women in our sample were compared with normed scores for the SF-36 scales in the general population of elderly US women. Method: Health status was measured by the US Spanish SF-36 in telephone surveys conducted entirely in Spanish. The sample (N = 65) was elderly (mean age 75.3) and primarily female (78%). Most had less than 7 years of education and annual income below $10,000. Results: Missing data were neglibible, and did not indicate difficulty with particular items. The item response values were well distributed and item response means were generally similar within a scale. Most item correlations were higher with the item's hypothesized scale than with other scales, though some items in the Mental Health, Vitality, and Social Functioning scales were highly correlated with other scales. Internal consistency reliability (Cronbach alpha) was .80 or above on all scales except Social Functioning (.69). SF-36 scale scores were lower in elderly Mexican American women than in elderly women in the general US population. Discussion: The US Spanish SF-36 was a generally satisfactory measure of health status in a sample of elderly Mexican Americans with little formal education. The performance of some items in the Mental Health, Vitality, and Social Functioning scales warrants further research.