Personalization of Cover Letter and Response Rate - a Randomized Controlled Trial in the Danish Health and Morbidity Survey

Background and aim: In the Danish Health and Morbidity Surveys the participation rate has declined from 61% in 2010 to 57% in 2013. A declining participation rate may affect the representativeness and the results of the survey. Several studies indicate that a survey that adapts personalized design features achieves higher participation rates. A feature can be a personalized cover letter which appears more relevant and interesting for the sample member. The research question, then, is whether cover letters with targeted content can perform better than a standard letter. The proposition is that such letters should increase the willingness of some sample members to participate, and that this will be reflected in higher response rates. However, little is known about which targeted content is the most effective.

Thus, in the spirit of adaptive design, this study seeks to establish whether it can be advantageous for participation overall to use cover letters with targeted content and whether the effect on participation of different versions of the cover letter varies in sample subgroups in order to optimize design features in future surveys. Further, given that the majority of sample members participate, these sample members who are swayed by the targeted letter must have relatively low response propensities (with the standard letter). Thus, we hypothesize that targeted letters should particularly improve response rates in low response propensity subgroups (i.e. young men, unmarried, and elderly women). A secondary research question is, therefore, whether any effect of targeted letters on response rate is greater among sample subgroups with low response propensities.

Design: A randomized trial will be incorporated the Danish Health and Morbidity Survey 2017 where a paragraph in the cover letter intended to heighten perceptions of relevance of the survey is varied among sample members. The Danish Health and Morbidity Survey 2017 will be based on a nationally representative random sample of 25,000 individuals (including institutionalized individuals) aged 16 years or older and resident in Denmark per 1 January 2017. The sample will be drawn from the adult population in Denmark using the Danish Civil Registration System and the survey will be conducted between February 3 and Maj 5, 2017. All sample members will be randomly allocated, with equal probabilities, to one of 11 treatment groups. One group will receive a standard letter that is designed to have general appeal and the other 10 groups will receive one of 10 versions of the targeted cover letter. In the previous Health and Morbidity Surveys, all sample members received the same standard letter. Much of the content of the standard letter and the targeted letter will be the same. The intention is to hold constant features designed to demonstrate the credibility of the survey, to allay fears about confidentiality, to appeal to self-interest, and to provide basic information about the task of participation. All sample members will be invited either to complete a web questionnaire or to fill out a paper questionnaire concerning health and morbidity.

Statistics: The analysis will be based on logistic regression modelling of the 25,000 persons invited to the Health and Morbidity Survey 2017. The independent variable is a dichotomous indicator of treatment group: targeted or standard letter. The dependent variable indicates whether the sample member has fully or partially completed the questionnaire. The mediator variable indicates membership of each of three operationally important low response propensity groups (i.e. young men, unmarried, and elderly women). The approach is to first test for a main effect of the independent variable in each of the treatment groups. A second step is to conduct the same analysis stratified by treatment group. A third step will then seek to identify whether any significant effects occurs between the mediator variable and dependent variable.