US minorities disproportionately suffer from disorders of physical and mental health. However, little awareness exists regarding culture's possible effects on disease measurement. Measurement bias refers to the possibility that individuals with identical true levels of a variable may have dissimilar responses to a measure as a function of group membership. Statistically significant measurement bias can attenuate or accentuate group differences, lead to inaccurate diagnoses, decrease reliability and validity, and render group comparisons impossible. Impact addresses the degree to which statistically significant measurement bias affects observed scores. Few studies have examined measurement bias and impact across ethnicity in assessments of leading health indicators. Without these examinations, public health research cannot accurately estimate the extent of health disparities, assess efforts to address disparities, provide proper treatment, examine health outcomes, and inform public health policy. Measurement models such as confirmatory factor analysis (CFA) and item response theory (IRT) can powerfully investigate measurement bias and its impact. Method: The purpose of this secondary data analysis is to examine measurement bias'role and impact in the assessment of several leading health indicators across ethnicity using IRT and CFA in a large (n = 43,093) nationally representative sample of the US.
Specific Aim 1 : Examine measurement bias across non-Hispanic Caucasians, non-Hispanic African-Americans, and Hispanics on standardized measures of Depression, Tobacco Use, Alcohol Abuse, Alcohol Dependence, Substance Abuse, and Substance Dependence in recent data.
Aim 1 Hypothesis: Measurement bias will present across ethnicity on each of the studied leading health indicators.
Specific Aim 2 : Assess the impact of observed measurement bias on health disparity estimates. Specific 2 Hypotheses: a) Observed measurement bias will meaningfully impact estimates. b) Measurement bias will attenuate differences across the groups;correcting measurement bias will accentuate group differences.
Specific Aim 3 : Provide psychometric information for each of the studied leading health indicators so new research can: include empirically tested, cross-culturally valid, bias free items and scales;estimate latent trait scores in new data using the application's parameter estimates;investigate and correct bias'cause;empirically create reliable standardized short scales;or add the items to computerized adaptive test (CAT) banks. Significance: Statistically significant and impactful measurement bias in the predicted direction would suggest that health disparities have been underestimated. However, non-statistically significant findings would suggest that health disparities have received accurate measurement across ethnicity. Finally, psychometric results will generally inform and improve future research efforts utilizing these items. Thus, regardless of outcome, results will provide essential information in public health efforts to estimate, understand, and close the health disparities gap.
Inaccurate measurement across ethnic groups substantially impedes public health efforts to estimate, understand, prevent, and close the health disparities gap. This application: examines whether and how culture affects the accuracy of leading public health indicator measurements, augments information regarding the role culture plays in the measurement and estimation of disease, and provides psychometric information to inform and enhance future research utilizing these items.
