Search results

IntroductionNovel data platforms are needed to expedite the linkage of geospatial and natural environment data with longitudinal population study(ies) (LPS) data. However, the intrinsic relationship between geospatial data and participant identities raises confidentiality concerns amongst participants which need to be accommodated within platform designs to ensure acceptable data use.
Objectives and ApproachWe aimed to establish generalisable mechanisms for linking spatial records into a cohort study LPS databank. Using 'Data Safe Haven' (Burton et al. 2015) approaches we developed a pipeline of technical processes (e.g. geocoding participants' residential and school address records, developing an 'engine' for linking exposure data to LPS data), which we tested for participant acceptability (through a focus group testing acceptable use of personal identifiers in this context) and tuned to meet regulatory requirements (e.g. privacy impact assessments, information security accreditation). We demonstrated our approach through an exemplar investigation assessing in utero NO2 exposure with later childhood respiratory outcomes.
ResultsParticipants' expressed clear expectations that the research use of location data should be restricted to trusted study staff (as distinct from the wider research community), although this expectation is context specific and does not represent a carte blanche for using granular information (e.g. GPS tracing data). This necessitated a 'split stage' protocol, where personal identifiers are handled separately from data/analysis. Daily NO2 exposure was modeled using road, other local and regional records; with validation exposure data collected from city-based sensors. Participants' location information was geocoded and linked to NO2 using the ALGAE (ALorithms for Generating address histories and Exposures) privacy-preserving geocoding engine. We will summarise participant views, our exemplar findings, and describe the linkage engine we have developed and its availability via an open-source repository.
Conclusion/ImplicationsParticipants' expect that studies control for confidentiality risks introduced by research using spatial identifiers. It is not realistic to expect that all LPS have the capacity to undertake specialist spatial linkages. Our generalisable approaches and open-source software could provide the basis for geospatial and natural environment epidemiology platforms in LPS.

BackgroundPolicing in the UK is increasingly taking a public health approach to tackling crime. Birth cohort studies are potentially well-placed to provide evidence on the most important risk and preventative factors. Linking official police records to cohort data offers considerable potential for future research, but data privacy concerns mean achieving such linkage is challenging. As a pilot, the Ministry of Justice provided ALSPAC with an anonymous extract (cannot be linked to study data) detailing consenting participants' criminal convictions and cautions held within the Police National Computer (PNC).
Main AimTo use anonymised PNC data to inform the potential for future research using ALSPAC and individual-level linked crime data.
MethodsWe examine criminality (type, when and where committed) in the cohort participants who were sent for linkage to the PNC in 2013 when they were aged approximately 21 years (n=7358). We consider the representativeness of these participants by comparing their self-reported crime-related behaviours during adolescence to those of participants who did not consent to linkage (n=7604). Finally, we consider these self-reported measures by present-day crime linkage consent status.
ResultsOf those sent for PNC linkage, 885 were linked to a criminal record(s) comprising 2635 convictions and 1365 cautions. A third related to serious crimes, and >90% were committed in Avon and Somerset constabulary. Those sent for linkage self-reported fewer criminal behaviours than those not sent for linkage. Regarding present-day consent status, those who have explicitly opted-in generally self-report fewer criminal behaviours and have less missing data than those who have not responded to the consent campaign.
ConclusionIf future linkage to individual police records was only achieved for explicit consenters, this would under-estimate criminality in the cohort. The geographical clustering of the crimes means local police records, which are more detailed than those of the PNC, could also be used.

Historical records and the research databases of completed studies have the potential either to establish new research studies or to inform follow-up studies assessing long-term health and social outcomes. Yet, such records are at risk of destruction resulting from misconceptions about data protection legislation and research ethics. The recent destruction of the Windrush disembarkation cards, which potentially could have formed the basis of a retrospective cohort study, illustrates this risk. As organisations across Europe transition to the EU General Data Protection Regulation (GDPR), this risk is being amplified due to uncertainty as to how to comply with complex new rules, and the requirement under GDPR that data owners catalogue their data and set data retention and destruction rules. The combination of these factors suggests there is a new meaningful risk that scientifically important historical records will be destroyed, despite the fact that GDPR provides a clear legal basis to hold historical records and to repurpose them for research for the public good. This letter describes this risk; details the legal basis enabling the retention and repurposing of these data; makes recommendations as to how to alleviate this risk; and finally encourages the research and research-active clinical community to contact their 'Data Protection Officers' to promote safe-keeping of historical records.

We aimed to develop a framework for generating synthetic identifier datasets to support development and evaluation of data linkage methods. We evaluated whether replicating associations between attributes and identifiers improved the utility of the synthetic data for assessing linkage error.
We determined the steps required to generate synthetic identifiers that replicate the properties of real-world data collection. We generated synthetic versions of a large United Kingdom cohort study (the Avon Longitudinal Study of Parents and Children), according to the quality and completeness of identifiers recorded over several waves of the cohort. We evaluated the utility of the synthetic identifier data in terms of assessing linkage quality (false matches and missed matches).
Comparing data from two collection points in ALSPAC, we found within-person disagreement in identifiers (differences in recording due to both natural change and non-valid entries) in 18% of surnames and 12% of forenames. Rates of disagreement varied by maternal age and ethnic group. Synthetic data provided accurate estimates of linkage quality metrics compared with the original data (within 0.13-0.55% for missed matches and 0.00-0.04% for false matches). Incorporating associations between identifier errors and maternal age/ethnicity improved synthetic data utility.
Replicating dependencies between attribute values (e.g. ethnicity), values of identifiers (e.g. name), identifier disagreements (e.g. missing values, errors or changes over time), and their patterns and distribution structure enables generation of realistic synthetic data that can be used for robust evaluation of linkage methods.
Our framework provides a novel and generalisable mechanism for developing and benchmarking record linkage algorithms.

IntroductionCareful development and evaluation of data linkage methods is limited by researcher access to personal identifiers. One solution is to generate synthetic identifiers, which do not pose equivalent privacy concerns, but can form a 'gold-standard' linkage algorithm training dataset. Such data could help inform choices about appropriate linkage strategies in different settings.
ObjectivesWe aimed to develop and demonstrate a framework for generating synthetic identifier datasets to support development and evaluation of data linkage methods. We evaluated whether replicating associations between attributes and identifiers improved the utility of the synthetic data for assessing linkage error.
MethodsWe determined the steps required to generate synthetic identifiers that replicate the properties of real-world data collection. We then generated synthetic versions of a large UK cohort study (the Avon Longitudinal Study of Parents and Children; ALSPAC), according to the quality and completeness of identifiers recorded over several waves of the cohort. We evaluated the utility of the synthetic identifier data in terms of assessing linkage quality (false matches and missed matches).
ResultsComparing data from two collection points in ALSPAC, we found within-person disagreement in identifiers (differences in recording due to both natural change and non-valid entries) in 18% of surnames and 12% of forenames. Rates of disagreement varied by maternal age and ethnic group. Synthetic data provided accurate estimates of linkage quality metrics compared with the original data (within 0.13-0.55% for missed matches and 0.00-0.04% for false matches). Incorporating associations between identifier errors and maternal age/ethnicity improved synthetic data utility.
ConclusionsWe show that replicating dependencies between attribute values (e.g. ethnicity), values of identifiers (e.g. name), identifier disagreements (e.g. missing values, errors or changes over time), and their patterns and distribution structure enables generation of realistic synthetic data that can be used for robust evaluation of linkage methods.

BackgroundLongitudinal Population Studies (LPS) typically operate as data resources for global bona-fide research, and therefore routinely share data to inform new research investigations. Current data sharing agreements between LPS and NHS Digital generally do not permit the onward-sharing of linked health records beyond the LPS's host institution or provide mechanisms to dynamically add new research hypotheses.
AimTo develop a framework of standardised onward-sharing models which control for risk and legal requirements and fit the context of different LPS.
ApproachThe CLOSER longitudinal consortium and NHS Digital held consultation meetings with 13 UK longitudinal studies, consortia studies and infrastructure providers to identify requirements and potential solutions.
ResultsFive onward-sharing models were identified. Two rely on the principle that anonymised data can be shared in a light-touch regulatory framework, and that this principle extends to 'effectively anonymised' data processed in line with the Information Commissioner's Office (ICO) anonymisation code of practice. Three methods utilise differing permutations of contractual agreements in order to control for risk and enforce NHS Digital's legal obligations. Establishing and maintaining these contractual arrangements will involve close working arrangements between the LPS and NHS Digital and further work is needed to clarify requirements for international data sharing. All models rely on the use of socio-technical controls used by LPS and infrastructure providers operating 'Data Safe Haven' approaches. Secure research facilities offered by national infrastructure providers – such as UK Secure eResearch Platform and UK Data Service – provide powerful risk control tools.
ConclusionThe identified models provide varied and flexible options for the onward-sharing of linked LPS and NHS Digital data. Greater clarity is needed in order to share effectively anonymised data and to share data internationally. There are opportunities for the LPS community to facilitate elements of these processes through partnership approaches.

ABSTRACT
Objectives"Clearly, details about an individual's mental health, for example, are generally much more 'sensitive' than whether they have a broken leg."
UK Information Commissioners Office
There is a perceived wisdom – based on issues such as social taboos, religious sensitivities, or financial implications linked to health status – that some health data is more sensitive than others. This distinction is present in many of the regulatory interpretations of privacy law (e.g. the UK Information Commissioners Office interpretation of the EU Data Directive, illustrated above), and is factored into the thinking of ethics and other regulatory decision-making committees. However, these particularly 'sensitive' data are defined at a regulatory level in broad terms (e.g. mental health), yet need implementing by researchers in precise terms. In 2013 our longitudinal research study was given approval by the UK Secretary of State for Health to access identifiable patient health records with the exception of those relating to mental health, sexual health or termination of pregnancy. Our objective therefore was to develop a generalisable informatics approach which enabled us to filter out sensitive records at the point of extraction.
ApproachWe developed a methodology based on the Cochrane systematic review approach: firstly using internationally recognised definitions of health concepts and reference texts (e.g. British National Formulary drug manual) we identified keywords associated with sensitive health events (including symptom and diagnostic terms, drug and appliance codes, community and secondary care references); secondly, through data-mining code terminologies – using both code terms and information embedded within the structure of the schema itself - we identified code values relating to these terms; thirdly we minimised our results through filtering out spurious results via manual review; finally, the resulting code lists were then crossed-referenced with other terminologies to ensure interoperability.
ResultsWe produced separate definitions of mental health and sexual health events initially using Read codes. Using NHS cross-reference tables we were able to translate Read observation and diagnostic codes to the SNOMED CT vocabulary, but were unable to translate Read drug codes into the SNOMED/DM+D vocabulary.
ConclusionWe have demonstrated a systematic and partially interoperable approach to defining 'sensitive' health information. However, any such exercise is likely to include decisions which will be open to interpretation and open to change over time. As such, the application of this technique should be embedded within an appropriate governance framework which can accommodate misclassification while minimising potential patient harm.

ABSTRACT
ObjectivesRates of common mental disorders may be increasing among children and adolescents, though evidence of this is mixed. Symptom questionnaires in population surveys may overestimate clinical disease. Conversely, lower participation of individuals with mental disorders may lead to underestimates in surveys. Clinical databases may have greater population coverage and contain information of more obvious clinical validity; however, several factors, including the help-seeking behaviour of individuals and the recording practices of clinicians, may influence burden-of-disease estimates based on these databases. The aim of the current investigation was to compare case definitions of common mental disorders (CMD) using linked electronic primary care data to definitions derived from self-reported data obtained in an observational study.
ApproachWe studied 1,562 adolescents who had completed the Revised Clinical Interview Schedule (CIS-R) in the Avon Longitudinal Study of Parents and Children (ALSPAC) at age 17-18 years and had linkage established to their electronic primary care records for at least 6 months after the time they completed the CIS-R. We used lists of Read codes corresponding to diagnoses, symptoms and treatment to create twelve definitions of CMD and also of depression alone. We calculated sensitivities and specificities of these, using CIS-R case definitions as the reference standard. All analyses were carried in Stata 13.0.
ResultsSensitivities ranged from 5.2% to 24.3% for depression and from 3.8% to 19.2% for CMD. The specificities of all definitions were above 98% for depression and above 96% for CMD. For both depression and CMD, the definition that included current diagnosis, treatment or symptoms identified the highest proportion of cases.
ConclusionsMost individuals meeting case definitions for CMD based on information in clinical records also met CMD case definitions based on symptoms reported in a contemporaneous survey. Conversely, many individuals identified as CMD cases based on reported symptoms had no evidence of CMD in their clinical records. A small number of individuals with CMD recorded in their clinical records had not reported symptoms of this in the survey. Overall, these data suggest that clinical databases are likely to yield underestimates of the burden of CMD in the population. Clinical records appear to yield highly valid diagnoses of common mental disorders which may be useful for studying risk factors and outcomes of these conditions. The greatest epidemiological value is likely to be obtained when the combination of information from both survey and clinical records is possible.

Objectives: Care.data was a 2013 UK government initiative to extract patient data from general practices in England to form a centralised whole-population database for service planning and health research. After a public outcry, the scheme was postponed and cancelled. Public views of care.data have previously been analysed; this study aimed to understand contemporary general practitioners' (GPs) views of the scheme, which may have been influential in its downfall. Design: Systematic search of media articles, followed by media content analysis. Setting: UK-based mainstream and GP-facing media in 2013 and 2014. Participants: Articles were eligible if they focused on care.data, and GPs were quoted, authored the article, or if articles were written for a majority GP audience. Interventions: N/A. Primary and secondary outcome measures: Themes which explained GPs' reactions to care.data and which could explain support for or opposition to the scheme. Results: 162 media articles met inclusion criteria and were drawn from newspapers, news websites and GP-facing websites. GPs recognised care.data's potential value for research and improving care, but had grave concerns about the scheme's implementation. These centred the lack of safeguards and purpose around the scheme which meant patients were not able to make informed decisions about opt-out. GPs perceived they were poorly resourced to meet competing demands to both share patients' data and protect confidentiality. They distrusted the government's likely uses of the data and perceived a risk of patient reidentification if the data were sold onto commercial entities. Conclusions: Findings show specific concerns which GPs had about care.data which led to the withdrawal of support. Future NHS patient data-sharing schemes should engage with GPs and other clinicians as key stakeholders from the earliest moments of planning, so that their views and needs are incorporated into the design of such schemes.

ABSTRACT
ObjectivesIn the UK, some children undertake National Cycle Proficiency Scheme [NCPS]/Bikeability training in primary school. It aims to promote cycling and safe cycling behaviours but there has been little scientific evaluation of its effectiveness. We aimed to examine the impact of cycle training on cycle-related behaviours and accidents in adolescence using self-report and hospital episode statistics (HES) data.
ApproachThe sample (n=5415) were participants in the Avon Longitudinal Study of Parents and Children who reported whether or not they had received NCPS training. Self-reported outcomes at 14 and 16 years included: cycling to school, ownership of cycle helmet, use of cycle helmet and high-visibility clothing on last cycle, and involvement in a cycle accident. Hospital admittance due to a cycle accident from 11-16 years was also included for a subsample (n=2222) who have been linked to HES. The association between cycle training and each of the outcomes was analysed using multilevel logistic regression (individual level 1, school level 2 [pseudo-ID of school attended obtained from linkage to National Pupil Database]), adjusted for a range of potential confounders.
ResultsApproximately 40% had received NCPS training. At 14 and 16 years, trained children were more likely to cycle to school (at 16 years: adjusted OR 1.56, 95% CI 1.20-2.02) and to own a helmet (16 years: 2.03, 1.72-2.41) than those who had not attended a course. They were also more likely to have worn a helmet on their last cycle at age 14 (1.26, 1.07-1.49), and to have worn high-visibility clothing at age 16 (1.70, 1.22-2.39). Training was not associated with self-reported involvement in a cycle accident, and only six participants with HES data had been admitted to hospital due to a cycle accident. Irrespective of training, results indicate very low use of high-visibility clothing (5%), very few girls cycling to school (<2%), and less than half of helmet owners wearing one on their last cycle. Ownership and use of helmets was particularly low in children from lower socio-economic position families.
ConclusionCycle training for children can have benefits that persist into adolescence. However, the low use of cycle helmets and high-visibility clothing indicate the further potential for interventions to encourage safe cycling behaviours in young people. Our hospital admissions outcome only captures individuals who sustained serious injuries; more minor cycle injuries would have been treated in A&E but a lack of detail in admission codes precludes analysis of such data.

Objective Trusted Research Environments (TREs) rely on Data Contributor Agreements (DCAs) and Data Access Agreements (DAAs) to bind organisations to terms and conditions in place to protect data. Our objective is to develop and drive adoption of DCA and DAA standardised templates which are optimised for data science in TREs. This is intended to reduce delays arising from the legal review process yet must not compromise on the safeguards needed for trustworthy data science.
ApproachA template DAA and DCA have been developed by the Pan-UK Data Governance Steering Group of the UK Health Data Research Alliance, underpinned by a core set of principles, and informed by a benchmarking exercise of existing DCAs/DAAs and meaningful involvement of public members and stakeholders.
ResultsPrinciples for TRE data contribution and data access in TREs have been mapped to core DCA and DAA terms for UK-based TREs and assigned to Five Safes' categories. Flexibility will be ensured by including terms specific to TREs or specific data/data owners in customisable annexes to the agreements. Public views obtained through public involvement and engagement (PIE) activities have been formative in the development of the outputs.
Conclusions By providing a familiar structure and terms, the templates should build trust among data owners and the UK public and provide clarity to all parties on their data protection obligations.
ImplicationsThis familiar structure and terms are intended to accelerate data research by enabling faster ingress of data into TREs and approval of projects, enabling more timely research.

Objectives Care.data was a 2013 UK government initiative to extract patient data from general practices in England to form a centralised whole-population database for service planning and health research. After a public outcry, the scheme was postponed and cancelled. Public views of care.data have previously been analysed; this study aimed to understand contemporary general practitioners' (GPs) views of the scheme, which may have been influential in its downfall. Design Systematic search of media articles, followed by media content analysis. Setting UK-based mainstream and GP-facing media in 2013 and 2014. Participants Articles were eligible if they focused on care.data, and GPs were quoted, authored the article, or if articles were written for a majority GP audience. Interventions N/A. Primary and secondary outcome measures Themes which explained GPs' reactions to care.data and which could explain support for or opposition to the scheme. Results 162 media articles met inclusion criteria and were drawn from newspapers, news websites and GP-facing websites. GPs recognised care.data's potential value for research and improving care, but had grave concerns about the scheme's implementation. These centred the lack of safeguards and purpose around the scheme which meant patients were not able to make informed decisions about opt-out. GPs perceived they were poorly resourced to meet competing demands to both share patients' data and protect confidentiality. They distrusted the government's likely uses of the data and perceived a risk of patient reidentification if the data were sold onto commercial entities. Conclusions Findings show specific concerns which GPs had about care.data which led to the withdrawal of support. Future NHS patient data-sharing schemes should engage with GPs and other clinicians as key stakeholders from the earliest moments of planning, so that their views and needs are incorporated into the design of such schemes.

ObjectivesDeveloping a rigorous application process for linked data within the UK Longitudinal Linkage Collaboration's Trusted Research Environment (TRE) included balancing the requirements of >20 longitudinal studies and several data providers with the need to provision data to researchers in a safe, ethical and efficient manner.
ApproachUK LLC was commissioned as a resource to enable the investigation of COVID-19 questions: a secure setting where longitudinal population studies (LPS) can be linked efficiently to regularly refreshed NHS electronic health records, and then accessed by researchers. UK LLC received funding to continue as a resource for any research in public good and to link to the administrative records (education, occupation, tax and benefits). To accommodate the constraints imposed by a new set of data controllers and to account for wide variety of proposals expected, we expanded our data access process.
ResultsUK LLC has a three-stage application and review process. The addition of administrative data meant that there is additional review for those applications requesting that data (Stage 4) and need to align processes with the UK Statistics Authority. There is also need to review complexity of the data requests and ethical implications of any application to access the data. We have developed an ethics decision tree to help guide the decision making around ethics requirements.
ConclusionOur process is evolving as we are trying to meet the challenge of balancing the needs of >20 studies, stakeholders, applicants, and participant/public contributors, to create a process that can handle complex needs and requirements.

IntroductionFetal Alcohol Spectrum Disorder (FASD) is one of the leading non-genetic causes of developmental disability worldwide and is thought to be particularly common in the UK. Despite this, there is a lack of data on FASD in the UK.
ObjectiveTo conduct public and professional involvement work to establish stakeholder views on the feasibility, acceptability, key purposes, and design of a national linked longitudinal research database for FASD in the UK.
MethodsWe consulted with stakeholders using online workshops (one for adults with FASD [and their supporters] N\,$=$\,5; one for caregivers of people with FASD (N=7), 1:1/small-team video calls/email communication with clinicians, policymakers, data-governance experts, third-sector representatives, and researchers [N=35]), and one hybrid clinical workshop (N\,$=$\,17). Discussions covered data availability, benefits, challenges, and design preferences for a national pseudonymised linked database for FASD. We derived key themes from the notes and recordings collected across all involvement activities.
ResultsOur tailored, multi-method approach generated high levels of stakeholder engagement. Stakeholders expressed support for a pseudonymised national linked database for FASD. Key anticipated benefits were the potential for: increased awareness and understanding of FASD leading to better support; new insights into clinical profiles leading to greater diagnostic efficiency; facilitating international collaboration; and increased knowledge of the long-term impacts of FASD on health, social care, education, economic and criminal justice outcomes. Given the rich data infrastructure established in the UK, stakeholders expressed that a national linked FASD database could be world-leading. Common stakeholder concerns were around privacy and data-sharing and the importance of retaining space for clinical judgement alongside insights gained from quantitative analyses.
ConclusionsMulti-method and multidisciplinary public and professional involvement activities demonstrated support for a national linked database for FASD in the UK. Flexible, diverse, embedded stakeholder collaboration will be essential as we establish this database.

ObjectivesOur Trusted Research Environment (TRE) provides a centralised infrastructure to pool Longitudinal Population Studies' (LPS) data and systematically link participants' routine health, administrative and environmental records. All data are held in a centralised research resource which is now certified by UK Statistics Authority as meeting the Digital Economy Act standard.
ApproachWe have created an unprecedented infrastructure integrating data from interdisciplinary and pan-UK LPS linked to participants' NHS England records with delegated access responsibilities. Integrated and curated data are made available for pooled analysis within a functionally anonymous DEA and ISO 27001 accredited TRE. We developed a bespoke governance and data curation framework with LPS data managers and Public/participant contributors. New data pipelines are being built with partners at ADRUK and the Office of National Statistics to link non-health records. Our design supports long-term sustainability, linkage accuracy and the ability to link data at both an individual and household level.
ResultsThis organisation is a collaboration of >24 LPS with ~280,000 participants. Participants' data are linked to NHS records and geo-coded environmental exposures. This resource is now accessible for public benefit research for bona fide UK researchers. Administrative data including tax, work and pensions, and education are being added to the resource. This data flow is enabled by: (1) a model where TTP processes participant identifiers for many different data owners; (2) creation of a novel longitudinal data pipeline, enabling linkage, data extraction and update of records over time; (3) an access framework where Linked Data Access Panel considers applications on behalf of data owners (e.g., the NHS), with review by a Public Panel and distributing applications to LPS for approval of appropriate data use.
ConclusionOur organisation provides a strategic research-ready platform for longitudinal research. We are extending linkages of LPS participants to previously inaccessible datasets. The research resource is positioned to allow researchers to investigate cross-cutting themes such as understanding health and social inequalities, health-social-environmental interactions, and managing the COVID-19 recovery.