Using Cognitive Interviews to Develop Surveys in Diverse Populations

• Published in: Medical care 2006-11, Vol.44 (11), p.S21-S30
• Main Authors: Nápoles-Springer, Anna M., Santoyo-Olsson, Jasmine, O'Brien, Helen, Stewart, Anita L.
• Format: Article
• Language: English
• Subjects: Adult; Aged; Aged, 80 and over; Analysis; Communication; Computer software; Cultural Diversity; Data coding; Ethnic groups; Ethnic Groups - psychology; Female; Health; Health behavior; Health care; Health care inequality; Health Care Surveys; Health surveys; Hispanics; Humans; Interviews; Interviews as Topic - methods; Male; Medical research; Middle Aged; Minority & ethnic groups; Multiculturalism & pluralism; Polls & surveys; Psychometrics - methods; Qualitative data; QUALITATIVE METHODS; Questionnaires; Research methodology; Survey research; Surveys and Questionnaires; United States
• ISSN: 0025-7079; 1537-1948
• DOI: 10.1097/01.mlr.0000245425.65905.1d

Background: Conceptual equivalence of measures is essential in research that compares health across diverse racial/ethnic groups. Cognitive interviews are pretest methods to explore the conceptual equivalence of survey items. Systematic approaches for using these methods are emerging. Objective: We describe an interaction analysis (IA) approach using qualitative data analysis software to analyze transcripts of cognitive interviews in a study to develop a survey instrument of the quality of interpersonal processes of care of diverse patients. Cognitive interviews included standard administration of the survey followed by retrospective probes for selected items. Subjects: Interviews were completed with 48 Latino, black, and non-Latino white respondents 18 years of age or older with at least one doctor's visit in the past 12 months. Participants averaged 45.8 years in age (standard deviation [SD] = 18.4), 58% were women, and mean education was 14.7 years (SD = 4.0). Results: Problems were identified in 126 of 159 items (79%). Behavior coding identified 32 problematic items (20%). IA of the transcript of the survey and retrospective probes identified 94 additional problematic items (59%). IA often revealed the nature of the problems, enabling decisions to modify or drop items based on respondents' comments. Behavior coding and IA identified ethnic and language similarities and differences in the use of response sets and the interpretation of items. Conclusions: IA and behavior coding of cognitive interview transcripts can identify efficiently problems with items and their source to increase the likelihood of the revised items being conceptually equivalent across ethnic groups.