As fewer people are responding to surveys than ever before, the linking of administrative records to survey responses is a potentially important tool to obtain new, high-quality data while making surveys more efficient. While initial research on adults suggest factors possibly influencing the consent decision and potential bias in records obtained, little research has been done on consent for children.
Given the possible benefits, a growing number of survey practitioners see linking administrative data to survey responses as an opportunity to improve data quality (and quantity) while reducing costs and easing interviewer and respondent burden. However, informed consent is frequently required to link survey and administrative data. Previous studies have largely focused on the consent process in surveys for adults; however, parents may also be asked to link their children’s administrative records, which may be of particular interest in longitudinal studies. Inclusion of children’s records allow for studying changes and outcomes over time for familial units, including intergenerational change1. These linked data may lack generalizability and may be biased, however, if consent is low and/or those who consented are different to those that did not.
To understand what factors possibly influence parents’ decisions to consent for themselves and their children, a framework was developed to direct the research based on theories of survey participation generally2. These ideas have been identified as possibly important in understanding the consent decision, and include factors coming from the respondent, the respondent’s environment, the interviewer, and the design of the survey. However, additional considerations need to be made when children are the subject of the consent request. For example, parenting style is an additional respondent factor that may not otherwise be considered. Further, it may not just be the parent’s (i.e. the respondent’s) characteristics influencing decisions, but also those of the children.
To examine and test this framework, our NCRM-funded research utilized data from Understanding Society: The United Kingdom Longitudinal Household Study. The survey is the largest longitudinal household survey in the UK, and at the first wave (in 2009-2010), respondents were asked if they consented to have their health and education records linked to their survey responses. For administrative reasons, only those born after 1981 and schooled in the UK were asked for linkage to their education records, while everyone was asked to link their NHS records. Parents were also asked for consent to link these same records for any children under 16, with education records only being requested if the child was school-aged. The rules regarding which parent was asked for children led to the mother nearly always being asked (95% of the time) and as such, only mothers were examined in this analysis.
Nearly all mothers gave the same consent response for all of their children, suggesting that in this instance what are important are mother-level factors, rather than child-level factors. This does not mean that future research should necessarily discount child characteristics. Other populations or survey contexts may lead to child-level factors having more of an impact. Mothers also consented for health records at a lower rate than for education records for both themselves and their children, while mothers consented for their children at lower rates than for themselves. However, a non-trivial number of mothers consented only for their children and not themselves (2.8%).
Analyses of consent outcomes show that mothers from minority ethnic groups are generally less likely to consent for either their children or themselves. Mothers who were harder-to-contact (and possibly more resistant to taking the survey) are more likely to refuse all consent requests. These respondents may feel that accepting the survey request is the extent of their willingness to participate. Improving strategies for reducing survey resistance and increasing willingness to share further may be possible through interviewer strategies developed while interacting with the respondent.2 Building a rapport with the respondent may also lead to higher consent rates, suggested by the finding in this data that that longer interviews led to higher consent rates.
There was little impact of interviewer demographics, overall experience, and interviewer’s achieved response rate and experience within Understanding Society, with no significant effects identified for mother or child health records. Why one consent request is apparently not affected by these interviewer success measures is unclear. Further exploration of which, when, and why interviewer traits are important is needed, which can then be used in interviewer recruitment and training.
In addition to identifying the factors related to consent, this study examined characteristics of children based on whether they were consented for or not to identify potential biases. There are a number of important demographic differences across children, with records less likely to be obtained among ethnic minorities and those in Southeast England and London. Responses to the youth survey of 10-15 year olds in Understanding Society, however, suggest little differences across several behaviours and attitudes, except for internet usage. While the lack of differences may be somewhat encouraging to users of linked data, the best way to minimise bias is to increase consent rates. For example, the differences in consent across ethnicities raise other concerns when the linked data are correlated with these demographics as many health outcomes may be.
1 Lightfoot, D. & Dibben, C. (2013) Approaches to linking administrative records to studies and surveys - a review. Administrative Data Liaison Service, University of St. Andrews. Retrieved from: http://www.adls.ac.uk/wp-content/uploads/Approaches-to-linking-administrative-records-to-studies-and-surveys-a-review.pdf
2. Groves, R. M. & Couper, M.P. (1998). Nonresponse in Household Interview Surveys. New York: Wiley.