Impact of a Sequential Mixed Mode Strategy on Response Rates and Reporting Errors in a Sample of US Veterans who applied for Department of Veterans Affairs Disability Benefits for Posttraumatic Stress Disorder
Background: Sequential mixed-mode methods, where one data collection strategy in a sample is followed by a second strategy, are increasingly popular for improving survey response rates and reducing non-response bias. There is risk, however, that switching data collection modes will exacerbate non-response bias or introduce interactions between reporting errors and data collection mode. Objectives: To assess the impact of adding a telephone interview to a mailed questionnaire on: 1) overall response rate and estimates of the population’s attributes, 2) differences in participants’ attributes across modes, and 3) the concordance of responses across modes among those who responded to both modes. Methods: observational, cross-sectional survey of 4,918 nationally representative US Veterans who applied for Department of Veterans Affairs disability benefits for posttraumatic stress disorder (PTSD) between 1994 and 1998. Analyses were stratified by gender. Results: In men, the response rate increased from 65.6% to 80.2% after the telephone interview and in women, from 66.7% to 80.2% (ps < 0.001). There were no differences across modes for the percentage of men and women who said they worked any hours for pay, went out socially, participated in sports, or experienced combat (ps > 0.05). The percentage of men and women saying they were sick or unable to work was 40 percentage points higher in the telephone interview than on the mailed questionnaire (ps < 0.001). Comparing the telephone interview to mailed questionnaire, the percentage reporting unwanted sexual attention while in the military was 18.5 percentage points lower in the men and 9.7 percentage points lower in the women (ps < 0.001). Total population estimates for being sick or unable to work increased by 7.3-8.3 percentage points after the telephone interviews, unwanted sexual attention in the military lowered by 1.6-3.2 percentage points. Conclusions: Adding a second data collection mode substantially improved response rates but may have introduced reporting errors that affected our estimates of why Veterans did not work for pay and whether they experienced unwanted sexual attention in the military.