Maximizing rates of consent to record linkage and additional data collection

This experimental study of alternative approaches to obtaining informed consent in the GDPR context of the European Union documents the advantages, for survey response rates, of using condensed information and in-text legal references (as opposed to a hyperlink).

We report on an experiment in a supplemental web survey as part of a longitudinal study in the United Kingdom where we ask survey respondents to consent to two forms of data linkage to health records and to consent to be mailed a serology kit. We varied the placement (early, early in context, or late in the survey) and order (linkage first or serology first) of the consent requests. We also examine reasons for consent or non-consent. We find that order of the requests does not make much difference, but making the requests early in the survey significantly increases consent rates over asking them after a series of content-related questions (by 3.4 percentage points) or later in the survey (by 7.2 percentage points). This is consistent with previous research showing that early requests for consent in a survey have a positive effect. The main reason chosen for not consenting related to the personal nature of the information requested.

This presentation provides an overview of an NIMLAS pilot study that aims to examine consent patterns from multi-national, multi-lingual, multi-cultural contexts of the Survey of Health, Ageing and Retirement in Europe (SHARE).

We report on an experiment in a supplemental web survey as part of a longitudinal study in the United Kingdom where we ask survey respondents to consent to two forms of data linkage to health records and to consent to be mailed a serology kit. We varied the placement (early, early in context, or late in the survey) and order (linkage first or serology first) of the consent requests. We also examine reasons for consent or non-consent. We find that order of the requests does not make much difference, but making the requests early in the survey significantly increases consent rates over asking them after a series of content-related questions (by 3.4 percentage points) or later in the survey (by 7.2 percentage points). This is consistent with previous research showing that early requests for consent in a survey have a positive effect. The main reason chosen for not consenting related to the personal nature of the information requested.

This study aimed to understand how different aspects of consent request design in surveys can impact informed consent rates for linking survey data to administrative records. In order to do this, a series of experiments including factors such as readability, placement, default options, additional information, and trust priming were varied. The findings suggest that asking for consent early in the survey and priming respondents to consider their trust in the data holder can increase consent rates without compromising understanding, while factors like default wording and offering additional information did not significantly impact consent rates.

In a context of multiple requests for linking different administrative records, the formatting of the multiple requests (all at once, one per page, one statement for all requests) does not matter, but the order can matter! Less sensitive requests first, more sensitive requests later. This work has primarily focused on admin data linkages; more work is needed on physical measures, biological measures, etc.

This study explored factors influencing individuals’ willingness to participate in smartphone-based data collection methods. Findings indicate that participants are more likely to engage when they have control over data collection, can review data before submission, receive invitations via postal letters instead of phone calls, and are offered unconditional incentives rather than conditional ones.

Collecting additional data from a variety of sources can help to enrich survey data. Regardless of the source, consent must be provided in order to collect this additional information. When comparing administrative data, smartphone usage data, bank data, biomarkers, Facebook data, health insurance data, and sensor data, survey respondents were most likely to consent to sharing their Facebook, smartphone usage, and biomarkers, whereas they are more reluctant to share bank data.

We use experimental mixed-mode data from a probability survey in Great Britain to examine why respondents are less likely to consent to data linkage in online than face-to-face interviews. We find that the 30 percentage point difference in consent rates is a causal effect of the mode on willingness to consent; it is not due to selection of different types of respondents into web and face-to-face interviews. We find that respondents are less likely to understand the data linkage request, less likely to process the consent request thoroughly, and more likely to be concerned about privacy and data security when answering online rather than in a face-to-face interview. Using digital audio-recordings of the face-to-face interviews, we find that verbal behaviours of interviewers do not explain the mode effects: respondents only rarely ask questions or express concern, and interviewers only rarely offer additional information about the data linkage. We also examine which devices respondents used to complete the web survey and find that these do not explain the mode effects either. Finally, we test the effects of simplifying the consent request, by reducing the reading difficulty: while the easier wording increases understanding of the request, it does not increase consent in either mode.

Data from social media platforms such as Twitter can be very useful when it is linked to survey data, as it provides a vast amount of information on the thoughts and opinions on various topics of those who utilize these platforms, and do so in almost real-time. Due to privacy requirements, additional consent from respondents is needed in order to link their Twitter data with survey data. This study found that consent rates are higher if the request is placed at the beginning rather than the end of the interview or survey. Additionally, there are many additional individual-level factors that impact consent to data linkage.

The COVID-19 pandemic has highlighted the importance of high-quality health research and data for use in political decision-making surrounding population health issues. Therefore, it is important to understand factors such as social trust, privacy concerns, technical affinity, altruism, age, and device ownership, as they either encourage or deter people from sharing their health data.

Linking data from longitudinal surveys to administrative records (whether held by government or private entities) is an increasingly attractive option for several reasons. As longitudinal surveys continue to respond to pressures to increase efficiency, and as new survey modes are developed, the use of mixed-mode data collection is increasing. Much of the work on informed consent for administrative data linkages has focused on exploring correlates of consent at both the respondent and interviewer level and examining non-consent bias. In this chapter, the authors focus on understanding the process by which respondents decide whether to give consent. They use both quantitative analyses of existing data from Understanding Society and qualitative work to explore the consent process. The quantitative analyses document the extent to which respondents make consistent decisions, and the extent to which the mode of data collection affects this decision, by addressing some research questions.

Surveying transient and other hard-to-reach populations can be especially difficult. In an effort to increase data collection in a group of refugees, a mobile app was offered after an interview in an attempt to collect additional passive data. Unfortunately, many refugees did not install the app, and low literacy rates are highly correlated with app nonparticipation. On the contrary, introducing an incentive was shown to significantly increase participation in the passive data collection efforts.

Combining survey and digital trace data from sources such as Facebook, Twitter, Spotify, and smartphone health applications can enhance the richness and quality of the overall data. Using incentives is shown to increase the likelihood of a participant agreeing to share digital trace data.

Interviewers can have very strong effects on record linkage consent rates (and potentially consent bias), and additional studies of explanations for these effects are needed.

The increase in smartphone usage and improvement in related technologies provides researchers with a unique opportunity to collect data from users through passive data collection with smartphone applications. Participant’s willingness to participate in passive mobile data collection is often contingent upon the following: the incentive promised for study participation, sponsor of the research, length of data collection period, option to opt out at any time, and ensuring participants privacy will be maintained.

In a telephone setting, better to place linkage consent requests at the beginning of the survey; cognitive load gets higher as the survey goes on, so it is important to ask earlier. Interviewer effects can also be quite strong on consent requests, and can depend on interviewer attitudes to linkage consent. Interviewers who themselves would consent tend to have higher consent rates.