Linking longitudinal studies of ageing with administrative data: Second interim report

Gasteen, Anne, Douglas, Elaine and Bell, David (2022) Linking longitudinal studies of ageing with administrative data: Second interim report (The ESRC Centre for Population Change Report Series) ESRC Centre for Population Change 59pp.

Record type: Monograph (Project Report)

Abstract

The Linking Longitudinal Studies of Ageing with Administrative Data project is an ESRC Centre for Population Change (CPC) project. The project remit is to evaluate the administrative data linkage of the surveys belonging to the Health and Retirement Studies (HRS) family on the Gateway to Global Ageing (G2G) platform. It aims to provide a detailed understanding of linked studies’ content, use, challenges and value added.
This second interim project report details the findings from a survey of 14 HRS family research teams conducted between June 2020 and June 2021. The survey focused on the benefits and challenges of linkage, examining application, approval and access processes, linkage mechanisms, linked data usage and impact.
Our assessment of the linkage landscape found that linked datasets were invariably owned by central or local government departments. Health and mortality records were the types of data most often linked. Linkage was intermittent, reflecting fragmented administrative data infrastructures with
multiple data controllers. Most studies were reliant on respondent consent to undertake linkage.
Linkage mechanisms were determined primarily by national administrative data
infrastructures. Linkage keys were usually dataset specific as data infrastructures tended to be fragmented and lacked unique ID numbers that would allow access to multiple datasets. Linkage was most often carried out using exact matching of dataset-specific ID numbers, rather than through a single linkage key that allows access to respondent records across multiple official administrative databases.
Establishing and renewing linkages was critically dependent on relationships with data controllers, bureaucratic processes, resources, and national administrative data architecture. The length of time taken to build relationships and navigate the application/approval process varied between two and ten years. The longer the time taken, the greater the risk of disruption from changes in regulations and/or regimes.
Ease of access to linked datasets varied across studies, ranging from the use of remote secure terminals, to visits to safe haven sites. The use of safe havens presents a barrier to researchers that can be costly and time consuming. The Covid-19 Pandemic appears to have initiated a move towards 2 greater use of remote secure terminal access. Linked data access protocols could vary within studies, with different access regimes in place for different data linkages.
Where linked datasets were available to third party researchers, recording of access to the data or outputs from this was patchy; either because it was not deemed necessary to track use, or it was not possible given data access arrangements. Data share agreements tended to include clauses requesting researchers to notify studies of outputs, but these were not enforceable.
Linked data usage was greatest, in general, where linkages had been in place longest and thirdparty access was not restricted. In the case of six studies – ELSA, HRS, JSTAR, NICOLA, SHARE, TILDA – linked data outputs were seen to have influenced either government policy, public debate and/or the future direction of research. While the validation of survey data was the single most cited benefit of administrative data linkage (seven studies), eight of the ten linked studies (at time of interview) also cited the mutual complementarity of longitudinal study data and administrative data and the broadening of research possibilities that linkage enables.
HRS family linked data research outputs suggest that this is where the greatest value of linkage lies. Longitudinal study linkage with administrative data allows research questions to be explored that would not otherwise be possible with one type of data alone. It enables the before and after of life events or policy impacts and mitigations to be studied providing a better understanding of causality,
risk factors and outcomes.
The main challenges that the studies faced when undertaking data linkage centred on the building of relationships with both data controllers and respondents, data protection issues, the availability of resources and fragmented administrative data infrastructures. The development of trust was cited as key to building relationships with data controllers and, crucially, respondents. Given the reliance of most studies on respondent consent for linkage, if respondent trust was low this was likely to be reflected in low consent rates. Low consent rates had the potential to reduce linked data sample sizes to the extent that any marginal benefits were unlikely to outweigh resource costs.
Of the four studies aspiring to future micro-level data linkage at the time of interview, MARS was best placed for this with the integration of official databases and unique personal identification numbers for the population. Future ELSI-Brazil and LASI linkage should be able to utilize unique personal identification numbers in Brazil and India respectively. HART linkage would be taking place against a background of fragmented data infrastructures and no unique ID.
The MHAS experience demonstrated that, in some cases, the administrative data architecture may be simply too fragmented for linkage to be feasible.
Better informed government policy was seen as the key benefit of linkage by the currently unlinked studies while the building of relationships, political buy-in and respondent trust were cited as the main challenges.
We make the following recommendations:
Studies should seek respondent consent at the outset of any linkage process, given that linkage agreements with data controllers often depend on having respondent consent and can take several years to negotiate.
Simultaneously, studies should explore whether, given their governing data
protection legislation, respondent consent is strictly necessary, as the removal of the need for respondent consent simplifies linkage processes considerably.
The HRS family should pool and share linkage expertise in a formal forum, perhaps under the auspices of Gateway to Global Ageing (G2G), to avoid duplication of effort.
HRS Studies should monitor and document the outputs from linked data and their
impacts to build support for the case for future linkage.
Where there is no system of identifying individuals uniquely and administrative data systems are fragmented, administrative bodies should seek to establish such a system in order to enable linkage across administrative datasets (e.g. employment, tax, benefit, health, education records).
Where administrative data architecture is fragmented, infrastructure to centrally
manage administrative datasets, creating a single point of access, should be
developed by administrative bodies.
The feasibility of linkage in real time, with studies able to link to multiple
administrative databases simultaneously, on an as-needed, live basis, should be
explored by administrative bodies.