The 131st Annual Meeting (November 15-19, 2003) of APHA

Abstract #58684

Martha Shumway, PhD¹, George J. Unick, MSW¹, Tetine Sentell, MA¹, Wynne Bamberg¹, and Lonnie Snowden, PhD². (1) Psychiatry Department, University of California, San Francisco, 1001 Potrero Ave, Rm 7M-W21, San Francisco, CA 94110, 415-206-5577, shumway@itsa.ucsf.edu, (2) School of Social Welfare, University of California, Berkeley, 120 Haviland Hall, Berkeley, CA 94720-7400

Question characteristics and respondent race/ethnicity systematically affect responses to opinion surveys, regardless of question content. This study tested the hypotheses that question characteristics and race/ethnicity are associated with similar response effects in a standardized mental health measure, potentially compromising measurement equivalence. Brief Symptom Inventory (BSI) data from 721 psychosocial medicine clinic outpatients (192 African Americans, 297 Latinos, 122 other non-whites, 249 whites) were examined in mixed model analyses. Independent variables were respondent race/ethnicity and question characteristics (length, order, readability, linguistic complexity). Dependent variables were objectively identifiable response effects associated with comprehension difficulties (choosing the first, last or middle response, repeating the previous response, and no response). Question characteristics predicted response effects (p<.01). Long, complex questions and those appearing towards the measure's end were more prone to hypothesized effects. Race/ethnicity was additionally predictive of some response effects. Compared to other groups, African Americans were less likely to choose the first response and more likely to choose the last (P<.05). Whites were more likely than African Americans or Latinos to choose the middle response (p<.05). These results suggest that the BSI, a standardized mental health measure, is prone to the same response effects seen in opinion surveys. Respondents faced with long, complex items appear to satisfice, or take cognitive "short cuts," choosing the first, last or middle response, repeating their last response, or giving no answer. However, respondents from different racial/ethnic groups appear to prefer different satisficing strategies. This variation could compromise the equivalence of standard mental health measures across racial/ethnic groups.

Learning Objectives:

Participants should be able to:
• Identify question characteristics that affect responses to standardized mental health measures.
• Recognize that respondent race/ethnicity can affect responses.
• Understand that response variation across racial/ethnic groups can compromise measurement equivalence.

Keywords: Outcome Measures, Ethnicity

Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.