Saturday, November 12
Data Quality and Measurement Error
Sat, Nov 12, 9:00 AM - 10:25 AM
Hibiscus A
Assessing Data Quality for Sensitive Topics

Change in Questionnaire Design: Respondent Interaction with Survey Mode in an Assessment of a Mental Health Population (303560)

*Rose L Krebill-Prather, Social and Economic Sciences Research Center - Washington State University 
Danna Lee Moore , Washington State University 

Keywords: mixed mode questionnaire effects, social desirability, sensitive topics

Individuals that receive mental health services are often hard to identify and survey. Reporting mental health is sensitive, with concerns of stigma associated with receiving services. The elusive nature of identifying mental health status makes it difficult to know what contact methods and questionnaire designs work best for gaining survey cooperation and obtaining measures with this population. In 2015, mixed mode questionnaire design changes were implemented with individuals pre-identified as receiving outpatient mental health services and this allowed for evaluation of questionnaire changes and effectiveness of survey design.

The Mental Health Statistics Improvement Program Survey measures mental health consumers’ ratings and provides quality measures of outpatient services. Previously, the survey was conducted exclusively by telephone. In 2015, the survey was adapted to include a web mode as an additional way of completing the survey with the hopes of increasing the response rate and to encourage more reticent and younger respondents into the survey.

A 35% response rate was achieved with 1549 adult mental health consumer responses split with 91% by telephone and 9% by internet. This study presents findings on survey mode differences that are related to consumer ratings of outpatient services. Cross tabulation and statistical analyses were used to profile response differences across subgroups of respondents. Overall, our findings show a pattern where internet respondents were more likely to give negative assessments of outpatient services or be undecided, whereas telephone respondents were more likely to give positive responses. These results suggest there is an interaction effect with telephone interviewers and the potential for satisficing. Furthermore, with the current rating scales allowing for a neutral response, our findings also suggest that without an interviewer present, respondents are more likely to choose this neutral noncommittal response.