Successful validation of the CAT-MH Scales in a sample of Latin American migrants in the United States and Spain

Abstract

We examined cultural differences in the item characteristic functions of self-reported of symptoms of depression, anxiety, and mania-hypomania in a Latino population taking Computerized Adaptive Tests for Mental Health (CAT-MH) in Spanish versus a non-Latino sample taking the tests in English. We studied differential item functioning (DIF) of the most common adaptively administered symptom items out of a bank of 1,008 items between Latino (n = 1276) and non-Latino (n = 798) subjects. For depression, we identified 4 items with DIF that were good discriminators for non-Latinos but poor discriminators for Latinos. These items were related to cheerfulness, life satisfaction, concentration, and fatigue. The correlation between the original calibration and a Latino-only new calibration after eliminating these items was r = .990. For anxiety, no items with DIF were identified. The correlation between the original and new calibrations was r = .993. For mania-hypomania, we identified 4 items with differential item functioning that were good discriminators for non-Latinos but poor discriminators for Latinos. These items were related to risk-taking, self-assurance, and sexual activity. The correlation between the original and new calibration was r = .962. Once the identified items were removed, the correlation between the original calibration and a Latino-only calibration was r = .96 or greater. These findings reveal that the CAT-MH can be reliably used to measure depression, anxiety, and mania in Latinos taking these tests in Spanish. (PsycINFO Database Record (c) 2018 APA, all rights reserved).

Figure 1.

Bivariate relationship between the estimated primary severity scores based on the original non-Latino calibration and the new Latino calibration after eliminating poor discriminating items. X-axis refers to the primary severity score estimate based on the original non-Latino calibration. Y-axis refers to the primary severity score estimate based on the Latino calibration. The scale is the underlying N(0,1) distribution of the test scores produced by the bifactor model.

Figure 1.

Bivariate relationship between the estimated primary severity scores based on the original non-Latino calibration and the new Latino calibration after eliminating poor discriminating items. X-axis refers to the primary severity score estimate based on the original non-Latino calibration. Y-axis refers to the primary severity score estimate based on the Latino calibration. The scale is the underlying N(0,1) distribution of the test scores produced by the bifactor model.

Figure 1.

Bivariate relationship between the estimated primary severity scores based on the original non-Latino calibration and the new Latino calibration after eliminating poor discriminating items. X-axis refers to the primary severity score estimate based on the original non-Latino calibration. Y-axis refers to the primary severity score estimate based on the Latino calibration. The scale is the underlying N(0,1) distribution of the test scores produced by the bifactor model.

Figure 2.

Model fit. Observed and expected marginal response category proportions based on the bifactor model for the Latino sample data. Panel (a) displays the 377 categories for the 77 depression items most of which had 5 categories. Panel (b) displays the 400 categories for the 84 anxiety items most of which had 5 categories. Panel (c) displays the 110 categories for the 55 mania/hypomania items most of which had 2 categories.

Figure 2.

Model fit. Observed and expected marginal response category proportions based on the bifactor model for the Latino sample data. Panel (a) displays the 377 categories for the 77 depression items most of which had 5 categories. Panel (b) displays the 400 categories for the 84 anxiety items most of which had 5 categories. Panel (c) displays the 110 categories for the 55 mania/hypomania items most of which had 2 categories.

Figure 2.

Model fit. Observed and expected marginal response category proportions based on the bifactor model for the Latino sample data. Panel (a) displays the 377 categories for the 77 depression items most of which had 5 categories. Panel (b) displays the 400 categories for the 84 anxiety items most of which had 5 categories. Panel (c) displays the 110 categories for the 55 mania/hypomania items most of which had 2 categories.