A mixed-mode design for the self-administrated questionnaire of the Belgian health interview survey: A feasibility study

Abstract

Introduction In addition to Face-To-Face (F2F) interviews, the Belgian Health Interview Survey (BHIS) includes paper-and-pencil (PP they were invited to complete the web SAQ and only in case of an explicit refusal the P&P option was proposed. The participants who agreed to complete the web SAQ but did not complete it were sent two postal reminders. Based on this study, the acceptance rate of the web SAQ among the participants eligible to complete the web SAQ was assessed. In addition, determinants of accepting the web SAQ were assessed using a logistical regression analysis. Furthermore, the response rate of the web SAQ and the total response rate of the SAQ were calculated. Finally, to evaluate data quality, an item non-response rate was calculated for web and P&P respondents. A multivariate ordinary least square analysis was conducted to assess the impact of the mode on item non-response while controlling for sociodemographic characteristics. Results Among the participants, 60% were eligible to respond to the web SAQ (declared having computer access, home internet connection and recent usage of the internet through a computer). The agreement to complete the web SAQ was 76% for this eligible group. This agreement varied according to the interviewer involved (P = 0.0001). Only 79% of the respondents who agreed to complete the web SAQ finally completed it. At the end, a total response rate of only 74% was achieved for the SAQ (P&P and web) while actually all interviewed participants should have completed the SAQ. The mean item non-response rate was 3.0% (SD = 4.91) in the web and 8.0% (SD = 11.22) in the P&P mode. Web respondents had significantly lower item non-response rates than P&P respondents (P = 0.0059) when controlling for sociodemographic characteristics. Conclusions The importance of using a mixed-mode design combining both web and P&P SAQs is shown by the fact that only 60% of the participants were eligible to complete the web SAQ. Among this eligible group, a substantial proportion has agreed to complete the web SAQ (76%). It is noteworthy to mention that this agreement was highly associated with the involved interviewer illustrating that their persuasiveness plays an important role. Furthermore, even after sending two reminders, we had about 20% non-response for the web SAQ. This suggests that leaving respondents free to complete the SAQ after the F2F interview can have a negative impact on the SAQ response. Finally, 26% of the interviewed participants did not complete the SAQ in any mode compared to 13% in the BHIS edition of 2013. This further confirms that our mixed-mode approach is not the solution for the declining response rate. Nevertheless, regarding data quality, it should be stated that web SAQs are superior in terms of item non-response than P&P SAQs.