Search

The healthcare sector's adoption of data and digital technologies is hindered by stringent data privacy regulations. Synthetic data, common data models (CDMs) and federated data ecosystems present promising solutions to these challenges. This workshop explored the potential of synthetic data to revolutionise health research, education and innovation (REI). By discussing applications, benefits, and ethical implications, the workshop addressed the role of synthetic data in enhancing data interoperability, scalability, and privacy preservation within healthcare REI systems. It also examined the relationship between synthetic data and CDMs, such as the Observational Medical Outcomes Partnership (OMOP) CDM, and the potential of federated data ecosystems to enable collaborative analysis while maintaining data ownership and control. The workshop delved into specific use cases to assess the feasibility, benefits, and limitations of synthetic data in various healthcare contexts. Through case studies and discussions, participants explored and developed best practices for effective utilisation. Ultimately, the workshop aimed to explore whether the combination of synthetic data, CDMs, and federation could facilitate privacy-preserving data sharing, seamless data integration, and scalable analysis across distributed data sources. By addressing privacy, interoperability, and scalability challenges, these approaches can drive innovation and insights within the healthcare sector.

Objective and Approach Health Data Research UK, the UK national institute for health data science, is coordinating efforts alongside national academic partners to streamline data curation at disease, population, and data structure level to enhance data offerings and provide networked data infrastructure supporting whole-UK research. Due to clinical, coding, and system differences across the constituent countries of the UK, data is often not standardised for whole-UK analyses, creating burden on research teams and leading to long data preparation times in order to run even distributed analyses. The approach to solve this is multi-faceted, including deploying data curation and cohort creation algorithms into health data providers’ environments, and through the novel integration of federated analytics solutions (such as those piloted through recent national infrastructure programmes) improving data access and research deployment efficiency. Results Standardising data through clinical and structural data curation directly deployed to health data providers creates a framework for whole-UK studies to be readily achievable, and provide the data infrastructure base to integrate new technical federated analytics solutions to deploy and reproduce analytics without unnecessary large scale data migration. Conclusions and Implications Systematic curation of health data within national data providing organisations provides flexibility and choice for researchers in terms of the data they will apply for to answer vital research questions affecting the UK populace. Such advances will improve the quality and efficiency of research for all corners of the UK, and create a community of practice in terms of developing data from health systems to research environments.

Introduction Globally, 55 million individuals have dementia, with an increasing annual incident of 10 million. Enabling the development of new multi-modal models can improve the current diagnostic pathways and potentially contribute to early diagnosis and treatment of dementia. Here, we report how multi-modal resources is achieved within the successful Trusted Research Environment (TRE). Objectives We aimed to identify the challenges of the storage, distribution and analysis of neuroimaging data and how we could implement a comprehensive infrastructure to deal with these. The problems we specifically aimed to address were how to: anonymise scans, store large amounts of data, standardise datasets to a common format, extract metadata, provision the data, and allow for analysis. Methods We document various stages and capacities required for multi-modal neuroimaging analysis for dementia and conclude that achieving research ready assets to enable neuroimaging analysis for dementia from existing resources requires an engineered process to facilitate multiple aspects of curation, provisioning and large-scale analysis. Results We developed an ingest pipeline for neuroimaging data to meet the requirements set out in the objectives. This involved standardising all datasets to the Brain Imaging Data Structure, defacing scans and anonymising data, using MinIO for data storage and extracting metadata from header information for data discovery and provisioning. Conclusion The neuroimaging ingest pipeline developed has allowed for the distribution of imaging datasets within DPUK which has facilitated multi-modal research on anonymised and standardised data and enabled linkage with phenotypic and genomic data.

In recent years, great strides have been made towards the deployment of federated systems for data research, including exploring federated trusted research environments (TREs). These federated TREs allow data analysts to securely use data within multiple institutions with minimal overheads while preserving information governance. However, due to the sensitive nature of data utilised in data linkage pipelines, especially within privacy-preserving applications, the real-world deployment of federated data linkage is in its early stages. There are several challenges to overcome with privacy-preserving federated linkage, including securely and efficiently sharing information that enables the identification of record similarities. Another key consideration is the resilience of the federated linkage ecosystem to information being added or removed as data providers move in and out of the initiative, as changes in the record set can cause intermediary links to be destroyed or created. Within non-federated linkage, a spine dataset such as a national census can provide an anchor on which incoming datasets can be linked. In federated linkage this may not explicitly be the case, especially if providers are globally dispersed or such a broad spine does not exist for the use case, and so there may be critical datasets within the overall system. We present a discussion over the real-world challenges to federated linkage on population-scale applications, including approaches which allow privacy-preserving linkage across multiple TREs and handling information governance requirements. We also explore the impact of the collaborative nature of federated linkage, looking at how quality and record links are affected in practice.

Linked data is a powerful resource within data analytics and population-level research. However, methods for linkage vary and the choice of approach can impact downstream usage of data by introducing assumptions and biases in resulting links. Selecting stringent linkage methods helps strengthen identified links at risk of missing links; meanwhile, lenient rules or ill-considered comparisons may introduce false positive links. Therefore, the approach is non-trivial, requiring careful selection of preprocessing steps, model development and quality review to ensure suitable outputs, which can require significant human expertise and insight. Real-world population-scale linkage can benefit from automation and scalability offered within modern data centres, with many tasks eligible for pipelining, such as applying predefined cleaning routines, training defined models, and generating mapping tables. Despite this, there are still pinch points requiring human interaction, such as selecting appropriate linkage fields, blocking rules and comparison methods, and reviewing quality of predictions. We present an approach to provide scalable automation in linkage pipelines, whilst retaining transparency of the linkage process for downstream users, providing them with a dataset’s life history. The work output for a given dataset is a versioned catalogue documenting the dataset’s journey, with transparent reporting of data origin, linkage settings, routines, and privacy-preserving quality analysis for inspection. This gives researchers insight into how it may affect their data and provides confidence in data usage. These insights also work in both directions, allowing users to provide feedback and iteratively refine linkage approaches.

The Dementias Platform UK (DPUK) Data Portal holds over 60 cohort datasets from over 3 million participants with a range of multi-modal data including neuroimaging and genomics. This has meant we have seen an increasing interest in the development of AI models with the potential for clinical implementation. However, this presents a unique challenge to disclosure control when it comes to assessing these AI models for release due to the risk of attacks such as membership inference, model inversion or even just vulnerabilities in the models like overfitting. This is why we hosted a series of workshops bringing together members of the public, expert researchers, and data owners across the UK to build a framework for allowing the safe development and release of AI models trained on sensitive healthcare data. From this, we have put together recommendations and guidelines for the use of privacy-preserving techniques in AI models to protect patient privacy, and to allow the safe deployment of these models outside of trusted research environments. We also identified the unique challenges to privacy and implementation of privacy-preserving techniques in AI models regarding the use of complex data such as neuroimaging and genomics. This framework has created a path forward for supporting safe AI model development which takes into consideration rapidly evolving ethical, legal and privacy considerations.

The Cambrian explosion of Trusted Research Environments (TREs) will inevitably lead to increased data siloing and potentially decrease achievable data science. Federated access to data is a mitigation against this risk. The presentation will cover two major infrastructure demonstrator projects to address this need and start the UK journey towards interconnected TRE’s. TELEPORT was an infrastructure to connect TREs and allow all the data assets to be seen through a single pane of glass while leaving them in situ; the analysis can be performed on each data asset, as well as combining data from multiple TREs to derive knowledge. The analysis is performed within the construct of a pop-up TRE, which is a temporary project-specific TRE. The data governance model is operated in an as-is state, with all egress controls being applied upon extraction of research output. This method has been termed Federated Data. TREFX is a set of infrastructure based around international GA4GH standards and the RO-CRATE concept to provide a platform to send complex workflows to the TRE’s to be performed upon the data assets, with the output post egress checking to be re-packed into the inbound RO-CRATE prior to being returned to the researcher. This method addresses both federated analysis and method reproducibility. This project was extended to cover ADDI workloads for dementia. Both solutions were jointly developed by Swansea University / SeRP and share common elements. Future work will see both these products being combined into a hybrid approach as well as spin-out products and services.

This paper presents the formal semantics of concurrency in Core Erlang, an intermediate language for Erlang, along with a notion of program equivalence (based on barbed bisimulation) that is able to model equivalence between programs that have different communication structures but the same observable behaviour. The novelty in our formalisation is its extent: it includes semantics for messages and exit and link signals, in addition to most of Core Erlang’s sequential features. Furthermore, unlike previous studies, this work formalises message receipt using primitive operations, consistent with the standard as of Erlang/OTP 23. In this novel formalisation, we show some generally applicable program equivalences (such as process identifier renaming and silent evaluation) and present a practical case study featuring the equivalence of sequential and concurrent list processing.

Introduction: Positional changes in intracranial pressure (ICP) have been described in humans when measured over minutes or hours in a static posture, with ICP higher when lying supine than when sitting or standing upright. However, humans are often ambulant with frequent changes in position self-generated by active movement. Research question: We explored how ICP changes during movement between body positions. Material and methods: Sixty-two patients undergoing clinical ICP monitoring were recruited. Patients were relatively well, ambulatory and of mixed age, body habitus and pathology. We instructed patients to move back and forth between sitting and standing or lying and sitting positions at 20 s intervals after an initial 60s at rest. We simultaneously measured body position kinematics from inertial measurement units and ICP from an intraparenchymal probe at 100 Hz. Results: ICP increased transiently during movements beyond the level expected by body position alone. The amplitude of the increase varied between participants but was on average ∼5 mmHg during sit-to-stand, stand-to-sit and sit-to-lie movements and 10.8 mmHg [95%CI: 9.3,12.4] during lie-to-sit movements. The amplitude increased slightly with age, was greater in males, and increased with median 24-h ICP. For lie-to-sit and sit-to-lie movements, higher BMI was associated with greater mid-movement increase (β = 0.99 [0.78,1.20]; β = 0.49 [0.34,0.64], respectively). Discussion and conclusion: ICP increases during movement between body positions. The amplitude of the increase in ICP varies with type of movement, age, sex, and BMI. This could be a marker of disturbed ICP dynamics and may be particularly relevant for patients with CSF-diverting shunts in situ.

Existing models for the analysis of concurrent processes tend to focus on fail-stop failures, where processes are either working or permanently stopped, and their state (working/stopped) is known. In fact, systems are often affected by grey failures: failures that are latent, possibly transient, and may affect the system in subtle ways that later lead to major issues (such as crashes, limited availability, overload). We introduce a model of actor-based systems with grey failures, based on two interlinked layers: an actor model, given as an asynchronous process calculus with discrete time, and a failure model that represents failure patterns to inject in the system. Our failure model captures not only fail-stop node and link failures, but also grey failures (e.g., partial, transient). We give a behavioural equivalence relation based on weak barbed bisimulation to compare systems on the basis of their ability to recover from failures, and on this basis we define some desirable properties of reliable systems. By doing so, we reduce the problem of checking reliability properties of systems to the problem of checking bisimulation.

Background Trusted Research Environments provide a legitimate basis for data access along with a set of technologies to support implementation of the "five-safes" framework for privacy protection. Lack of standard approaches in achieving compliance with the "five-safes" framework results in a diversity of approaches across different TREs. Data access and analysis across multiple TREs has a range of benefits including improved precision of analysis due to larger sample sizes and broader availability of out-of-sample records, particularly in the study of rare conditions. Knowledge of governance approaches used across UK-TREs is limited. Objective To document key governance features in major UK-TRE contributing to UK wide analysis and to identify elements that would directly facilitate multi TRE collaborations and federated analysis in future. Method We summarised three main characteristics across 15 major UK-based TREs: 1) data access environment; 2) data access requests and disclosure control procedures; and 3) governance models. We undertook case studies of collaborative analyses conducted in more than one TRE. We identified an array of TREs operating on an equivalent level of governance. We further identify commonly governed TREs with architectural considerations for achieving an equivalent level of information security management system standards to facilitate multi TRE functionality and federated analytics. Results All 15 UK-TREs allow pooling and analysis of aggregated research outputs only when they have passed human-operated disclosure control checks. Data access requests procedures are unique to each TRE. We also observed a variability in disclosure control procedures across various TREs with no or minimal researcher guidance on best practices for file out request procedures. In 2023, six TREs (40.0%) held ISO 20071 accreditation, while 9 TREs (56.2%) participated in four-nation analyses. Conclusion Secure analysis of individual-level data from multiple TREs is possible through existing technical solutions but requires development of a well-established governance framework meeting all stakeholder requirements and addressing public and patient concerns. Formation of a standard model could act as the catalyst for evolution of current TREs governance models to a multi TRE ecosystem within the UK and beyond.

There is common consensus that data sharing accelerates science. Data sharing enhances the utility of data and promotes the creation and competition of scientific ideas. Within the Alzheimer’s disease and related dementias (ADRD) community, data types and modalities are spread across many organizations, geographies, and governance structures. The ADRD community is not alone in facing these challenges, however, the problem is even more difficult because of the need to share complex biomarker data from centers around the world. Heavy-handed data sharing mandates have, to date, been met with limited success and often outright resistance. Interest in making data Findable, Accessible, Interoperable, and Reusable (FAIR) has often resulted in centralized platforms. However, when data governance and sovereignty structures do not allow the movement of data, other methods, such as federation, must be pursued. Implementation of fully federated data approaches are not without their challenges. The user experience may become more complicated, and federated analysis of unstructured data types remains challenging. Advancement in federated data sharing should be accompanied by improvement in federated learning methodologies so that federated data sharing becomes functionally equivalent to direct access to record level data. In this article, we discuss federated data sharing approaches implemented by three data platforms in the ADRD field: Dementia’s Platform UK (DPUK) in 2014, the Global Alzheimer’s Association Interactive Network (GAAIN) in 2012, and the Alzheimer’s Disease Data Initiative (ADDI) in 2020. We conclude by addressing open questions that the research community needs to solve together.

This research aims to show the effectiveness of Operator Event Sequence Diagrams (OESDs) in the normative modelling of vehicle automation to human drivers’ handovers and validate the models with observations from a study in a driving simulator. The handover of control from automation to human operators has proved problematic, and in the most extreme circumstances catastrophic. This is currently a topic of much concern in the design of automated vehicles. OESDs were used to inform the design of the interaction, which was then tested in a driving simulator. This test provided, for the first time, the opportunity to validate OESDs with data gathered from videoing the handover processes. The findings show that the normative predictions of driver activity determined during the handover from vehicle automation in a driving simulator performed well, and similar to other Human Factors methods. It is concluded that OESDs provided a useful method for the human-centred automation design and, as the predictive validity shows, can continue to be used with some confidence. The research in this paper has shown that OESDs can be used to anticipate normative behaviour of drivers engaged in handover activities with vehicle automation in a driving simulator. Therefore, OESDs offer a useful modelling tool for the Human Factors profession and could be applied to a wide range of applications and domains.

Objectives Under section 31 (s.31) of the UK Children Act 1989, public law care proceedings can be issued if there is concern a child is subject to, or at risk of significant harm. We examined health vulnerabilities of parents involved in public law care proceedings in the two-year period prior to involvement. Approach Our study created an anonymised individual-level population-based cohort, with a matched comparison group of parents in Wales who were not subject to care proceedings, matched on age, sex and deprivation. Family court data provided by Cafcass Cymru were linked to population-level healthcare records held within the Secure Anonymised Information Linkage (SAIL) Databank. Demographic characteristics, overall health service use and health profiles of parents of children subject to s.31 care proceedings between 2011 and 2019 were examined. Results Data were available for 8,821 parents involved in care proceedings between 2011 and 2019, with a comparison group of 32,006 parents. Nearly half (47.6%) of cohort parents resided in the most deprived quintile. Higher levels of healthcare use were found for cohort mothers and fathers compared to the comparison group across multiple healthcare settings, with the most pronounced differences for emergency department attendances (59.3% vs 37.0%). Health conditions with the largest variation between groups were related to mental health (43.6% vs 16.0%), substance use (19.4% vs 1.6%) and injuries (41.5% vs 23.6%). Conclusion This study highlights the heightened socioeconomic and health vulnerabilities of parents who experience care proceedings concerning a child. Better understanding of the needs and vulnerabilities of this population may provide opportunities to improve a range of support and preventative interventions that respond to crises in the community.

Across various domains, such as health and social care, law, news, and social media, there are increasing quantities of unstructured texts being produced. These potential data sources often contain rich information that could be used for domain-specific and research purposes. However, the unstructured nature of free-text data poses a significant challenge for its utilisation due to the necessity of substantial manual intervention from domain-experts to label embedded information. Annotation tools can assist with this process by providing functionality that enables the accurate capture and transformation of unstructured texts into structured annotations, which can be used individually, or as part of larger Natural Language Processing (NLP) pipelines. We present Markup (https://www.getmarkup.com/) an open-source, web-based annotation tool that is undergoing continued development for use across all domains. Markup incorporates NLP and Active Learning (AL) technologies to enable rapid and accurate annotation using custom user configurations, predictive annotation suggestions, and automated mapping suggestions to both domain-specific ontologies, such as the Unified Medical Language System (UMLS), and custom, user-defined ontologies. We demonstrate a real-world use case of how Markup has been used in a healthcare setting to annotate structured information from unstructured clinic letters, where captured annotations were used to build and test NLP applications.

Abstract Background Africa is the origin of modern humans within the past 300 thousand years. To infer the complex demographic history of African populations and adaptation to diverse environments, we sequenced the genomes of 92 individuals from 44 indigenous African populations. Results Genetic structure analyses indicate that among Africans, genetic ancestry is largely partitioned by geography and language, though we observe mixed ancestry in many individuals, consistent with both short- and long-range migration events followed by admixture. Phylogenetic analysis indicates that the San genetic lineage is basal to all modern human lineages. The San and Niger-Congo, Afroasiatic, and Nilo-Saharan lineages were substantially diverged by 160 kya (thousand years ago). In contrast, the San and Central African rainforest hunter-gatherer (CRHG), Hadza hunter-gatherer, and Sandawe hunter-gatherer lineages were diverged by ~ 120–100 kya. Niger-Congo, Nilo-Saharan, and Afroasiatic lineages diverged more recently by ~ 54–16 kya. Eastern and western CRHG lineages diverged by ~ 50–31 kya, and the western CRHG lineages diverged by ~ 18–12 kya. The San and CRHG populations maintained the largest effective population size compared to other populations prior to 60 kya. Further, we observed signatures of positive selection at genes involved in muscle development, bone synthesis, reproduction, immune function, energy metabolism, and cell signaling, which may contribute to local adaptation of African populations. Conclusions We observe high levels of genomic variation between ethnically diverse Africans which is largely correlated with geography and language. Our study indicates ancient population substructure and local adaptation of Africans.

Autoimmune hypothyroidism may result in a wide range of neuromuscular disorders. The frequently observed neurological manifestations of acquired hypothyroidism include mild to moderate myopathy and sensorimotor neuropathy, which usually resolve by clinical and electrophysiological criteria, in adults treated with thyroid hormone replacement. We report a case of a 30-year-old male with severe hypothyroidism secondary to chronic autoimmune thyroiditis who presented with a 2-year history of progressive fatigue, upper and lower limb weakness, myalgia, and intermittent paraesthesia. His neurological exam demonstrated proximal and distal muscle weakness, lower limb areflexia, and relatively intact sensory modalities. The patient’s biochemistry revealed unusually and profoundly raised the thyroid stimulating hormone (TSH) level of 405.5 mIU/L (reference range (RR): 0.27–4.2 mIU/L) and creatine kinase (CK) level of 20,804 U/L (RR: 45–250 U/L), while his nerve conduction studies (NCS) demonstrated severe sensorimotor polyneuropathy with both axonal and demyelinating features. Thyroid hormone replacement therapy over the first 3 months resulted in biochemical normalization of his extremely deranged thyroid function tests (TFTs) and CK levels. At 12 months, despite maintaining euthyroidism and noticeable improvement in strength, his nerve conduction studies (NCS) demonstrated the continued absence of distal motor and sensory responses in his lower limbs with only partial improvement in sensory amplitudes and conduction velocities in his upper limbs. This report highlights the potential for severe neuromuscular consequences from advanced and chronic autoimmune hypothyroidism. The patient’s myopathy has resolved over a period of three months with prompt normalization of CK levels. Concerningly, the patient achieved significant but incomplete recovery from his mixed axonal and demyelinating neuropathy with residual mild distal weakness and areflexia in his lower limbs and persistent motor and sensory impairments on his NCS. The severity and incomplete resolution of our patient’s neurological manifestations emphasize the importance of early diagnosis and the need for prompt therapeutic intervention for hypothyroidism.

Introduction Unstructured free-text clinical notes often contain valuable information relating to patient symptoms, prescriptions and diagnoses. These can assist with better care for patients and novel healthcare research if transformed into accessible, structured clinical text. In particular, Natural Language Processing (NLP) algorithms can produce such structured outputs, but require gold standard data to train and validate their accuracy. While existing tools such as Brat and Webanno provide interfaces to manually annotate text, there is a lack of capability to efficiently annotate complex clinical information. Objectives and Approach We present Markup, an open-source, web-based annotation tool developed for use within clinical contexts by domain experts to produce gold standard annotations for NLP development. Markup incorporates NLP and Active Learning technologies to enable rapid and accurate annotation of unstructured documents. Markup supports custom user configurations, automated annotation suggestions, and automated mapping to existing clinical ontologies such as the Unified Medical Language System (UMLS), the Systematized Nomenclature of Medicine - Clinical Terms (SNOMED-CT), or custom, user-defined ontologies. Results Markup has been tested on Epilepsy clinic letters, where captured annotations were used to build and test NLP applications. Markup allowed for inter-annotator statistics to be calculated in the case of multiple annotators. Re-annotation, following iterations of annotation definitions, was incorporated for flexibility. UMLS codes, certainty context, and multiple components from complex phrases were all able to be captured and exported in a structured format. Conclusions / Implications Markup allows gold standard annotations to be collected efficiently across unstructured text and is optimized to capture health-specific information. These annotations are important to develop and validate NLP algorithms that automate the capture of important information from clinic letters at scale.

Introduction Research can often be slow to start and require duplication of effort which in lots of cases has previously been completed to generate research-ready-data-assets (RRDA). Within the UK, two programmes: Dementias Platform UK (DPUK) that brings together over 50 different dementia-related cohorts and Secure Anonymised Information Linkage (SAIL) Databank, which provides access to longitudinal population-scale person-level data for every person in Wales have tried to tackle this challenge of creation, use and management of RRDA’s. Objectives and Approach Combining clinical, data and management expertise from DPUK and SAIL, we hoped to construct a RRDA that was easily accessible and well described for a dementia e-cohort. Welsh residents with available primary care records were included, with clinical and demographic information including follow-up times, several dementia indicators using validated diagnostic code lists, information on every dementia-related diagnostic event and several covariates and co-morbidities. SDeC was made available to researchers and can be modified according to appropriate study designs, with learning from projects used to update the SDeC to improve future uses. Interactive visualisations effectively summarise cohort characteristics, aiding researchers to quickly determine cohort eligibility for dementia studies. Results SDeC contains data from 4.6 million participants in SAIL, with 1.5 million meeting cohort inclusion criteria, resulting in 24.3 million person-years of follow-up. Of these, 146,323 (10%) developed all-cause dementia during follow-up, with 90,150 (60%) having dementia subtype codes. We made this resource available to researchers who had never used SAIL before, with limited experience of population-scale routine-data, and projects have proceeded with one managing to proceed from point of initial access to submission of publication in less than 6-months. Conclusion / Implications SDeC provides a reproducible dynamic method for completing dementia research, and expediting learning and understanding of the use of these data, with further developments and maintenance planned to increase the complexity and detail available to researchers over time.

Introduction There is no power for change greater than a community discovering what it cares about.1 The Health Data Platform (HDP) will democratize British Columbia’s (population of approximately 4.6 million) health sector data by creating common enabling infrastructure that supports cross-organization analytics and research used by both decision makers and cademics. HDP will provide streamlined, proportionate processes that provide timelier access to data with increased transparency for the data consumer and provide shared data related services that elevate best practices by enabling consistency across data contributors, while maintaining continued stewardship of their data. HDP will be built in collaboration with Swansea University following an agile pragmatic approach starting with a minimum viable product. Objectives and Approach Build a data sharing environment that harnesses the data and the understanding and expertise about health data across academe, decision makers, and clinicians in the province by: • Enabling a common harmonized approach across the sector on: • Data stewardship • Data access • Data security and privacy • Data management • Data standards • To: • Enhance data consumer data access experience • Increase process consistency and transparency • Reduce burden of liberating data from a data source • Build trust in the data and what it is telling us and therefore the decisions made • Increase data accessibility safely and responsibly Working within the jurisdiction’s existing legislation, the Five Safes Privacy and Security Framework will be implemented, tailored to address the requirements of data contributors. Results The minimum viable product will provide the necessary enabling infrastructure including governance to enable timelier access, safely to administrative data to a limited set of data consumers. The MVP will be expanded with another release planned for early 2021. Conclusion / Implications Collaboration with Swansea University has enabled BC to accelerate its journey to increasing timelier access to data, safely and increasing the maturity of analytics by creating the enabling infrastructure that promotes collaboration and sharing of data and data approaches. 1 Margaret Wheatley

Introduction The emergence of the novel respiratory SARS-CoV-2 and subsequent COVID-19 pandemic have required rapid assimilation of population-level data to understand and control the spread of infection in the general and vulnerable populations. Rapid analyses are needed to inform policy development and target interventions to at-risk groups to prevent serious health outcomes. We aim to provide an accessible research platform to determine demographic, socioeconomic and clinical risk factors for infection, morbidity and mortality of COVID-19, to measure the impact of COVID-19 on healthcare utilisation and long-term health, and to enable the evaluation of natural experiments of policy interventions.Methods and analysis Two privacy-protecting population-level cohorts have been created and derived from multisourced demographic and healthcare data. The C20 cohort consists of 3.2 million people in Wales on the 1 January 2020 with follow-up until 31 May 2020. The complete cohort dataset will be updated monthly with some individual datasets available daily. The C16 cohort consists of 3 million people in Wales on the 1 January 2016 with follow-up to 31 December 2019. C16 is designed as a counterfactual cohort to provide contextual comparative population data on disease, health service utilisation and mortality. Study outcomes will: (a) characterise the epidemiology of COVID-19, (b) assess socioeconomic and demographic influences on infection and outcomes, (c) measure the impact of COVID-19 on short -term and longer-term population outcomes and (d) undertake studies on the transmission and spatial spread of infection.Ethics and dissemination The Secure Anonymised Information Linkage-independent Information Governance Review Panel has approved this study. The study findings will be presented to policy groups, public meetings, national and international conferences, and published in peer-reviewed journals.

Introduction Better use of administrative data is essential to enhance understanding about the family justice system, and characteristics and outcomes for children and families. The Nuffield Family Justice Observatory Data Partnership supports this aim through analyses of core family justice datasets. When a child is involved in family court proceedings in Wales, Cafcass Cymru are employed to represent a child’s best interests. This paper provides an overview of the Cafcass Cymru data, and linkage to population level health and other administrative datasets held within the Secure Anonymised Information Linkage (SAIL) Databank. Two data linkage example analyses are described. Further research opportunities are outlined. Methods Cafcass Cymru data was transferred to SAIL using a standardised approach to provide de-identified data with Anonymised Linking Fields (ALF) for successfully matched records. Three cohorts were created: all individuals involved in family court applications; all individuals with an ALF allowing subsequent health data linkage; and all individuals with a Residential Anonymised Linking Field (RALF) and Lower Super Output Area (LSOA) enabling area level deprivation analysis. Results Cafcass Cymru data are available containing 12,745 public law applications between 2011 and 2019, with 52,023 applications from 2005 to 2019 for private law. The overall match rate was 80%, with variations observed by time, law type, roles, gender and age. Forty per cent had hospital inpatient admissions 2 years prior or after application receipt at Cafcass Cymru, of which 27% were for emergency admissions; 54% had an emergency department attendance and 61% an outpatient appointment during the same period. Individuals involved in public or private law applications were more likely to reside in deprived areas. Conclusion The Nuffield Family Justice Observatory Data Partnership will enhance research opportunities to better understand the family justice system and outcomes for children and families. Population level Cafcass Cymru data can be accessed through the SAIL Databank. Forthcoming data acquisition will also facilitate further analyses and insight.

Background The Secure Anonymised Information Linkage (SAIL) Databank is a national data safe haven of de‑identified datasets principally about the population of Wales, made available in anonymised form to researchers across the world. It was established to enable the vast arrays of data collected about individuals in the course of health and other public service delivery to be made available to answer important questions that could not otherwise be addressed without prohibitive effort. The SAIL Databank is the bedrock of other funded centres relying on the data for research. Approach SAIL is a data repository surrounded by a suite of physical, technical and procedural control measures embodying a proportionate privacy-by-design governance model, informed by public engagement, to safeguard the data and facilitate data utility. SAIL operates on the UK Secure Research Platform (SeRP), which is a customisable technology and analysis platform. Researchers access anonymised data via this secure research environment, from which results can be released following scrutiny for disclosure risk. SAIL data are being used in multiple research areas to evaluate the impact of health and social exposures and policy interventions. Discussion Lessons learned and their applications include: managing evolving legislative and regulatory requirements; employing multiple, tiered security mechanisms; working hard to increase analytical capacity efficiency; and developing a multi-faceted programme of public engagement. Further work includes: incorporating new data types; enabling alternative means of data access; and developing further efficiencies across our operations. Conclusion SAIL represents an ongoing programme of work to develop and maintain an extensive, whole population data resource for research. Its privacy-by-design model and UK SeRP technology have received international acclaim, and we continually endeavour to demonstrate trustworthiness to support data provider assurance and public acceptability in data use. We strive for further improvement and continue a mutual learning process with our contemporaries in this rapidly developing field.

Background Not enough is known about how the family justice system is working, the children and families using these services, and their wider outcomes beyond their involvement with the family court. The Nuffield Family Justice Observatory (FJO) Data Partnership, comprising a bespoke analysis platform hosted within the Secure Anonymised Information Linkage (SAIL) Databank and analytical teams at Lancaster and Swansea Universities, has been established to address this knowledge deficit. Methods Family justice data is being deposited in the SAIL Databank. Data are acquired using a standardised split-file approach, stored in an anonymised format and made available to projects as linked data using a unique encrypted project anonymised linkage field. These data can be augmented with a wealth of available health, education and other governmental/social routinely collected datasets, and future data acquired from other sources, for a range of research projects. Results Children and Family Court Advisory and Support Service (Cafcass) Cymru data has been transferred to the SAIL Databank, and agreements to transfer the Cafcass England data are being finalised. Applications are now welcomed to use these data to enhance understanding of the family justice system and children and families involved with the family courts in public and private law. Access will be facilitated through the SAIL Databank, subject to relevant governance procedures. Conclusion We will discuss the rationale of the Nuffield FJO Data Partnership, and how it aims to a) increase capacity and capability of researchers and data scientists utilising family justice and other relevant administrative datasets, b) improve understanding about the family justice system using data from the Cafcass in England and Wales, c) demonstrate the complexities and value of data linkage, and d) assist future policy and practice development. We will discuss matching rates for the Cafcass Cymru data and linkages that have been made to other datasets within the SAIL Databank. We will also set out the support available from the partnership to those wishing to access and utilise family justice data.

Background The use of Geographic Information Science in administrative and health data research provides researchers and policymakers with powerful modelling capabilities when exploring spatial variations in health and social outcomes. However, there are challenges around using spatial data and associated modelling techniques in terms of computing power, skills, data quality and disclosure controls. Main Aim In this case study, we describe our approach to overcoming some of these problems in conjunction with the Secure Anonymised Information Linkage Databank (SAIL). Approach The UK Secure E-Research Platform (UKSeRP) is a suite of technologies which can be linked together to provide a secure system to support a particular use case or data resource. We have taken this foundation and built a Geographic Information UKSeRP which is solely focused on the storage, manipulation and generation of spatial data and models to support administrative and health research. The GIS system is built around the Unique Property Reference Number (UPRN) which can be anonymised into a Residential Anonymised Linking Field (RALF) and provides a precise location of households across the UK. This system allows the user to generate high-resolution spatiotemporal models of geographically varying phenomena (e.g. access to services, fast food, alcohol, greenspace and pollution) in a secure managed environment. To support the geographic data stored in the platform there is a range of GIS tools (Python, R, QGIS, PostGIS, OSRM) and standardised methods available to users which sit alongside a data catalogue. Access to the platform is managed via an information governance review process on a project by project basis to ensure best practice in disclosure control and integrity. Conclusion Spatial data and GIS techniques are an important resource for academic and policy researchers. The GRAPHITE platform demonstrates best practice for using spatial data when used in conjunction with sensitive data.

Tool-assisted refactoring transformations must be trustworthy if programmers are to be confident in applying them on arbitrarily extensive and complex code in order to improve style or efficiency. We propose a simple, high-level but rigorous, notation for defining refactoring transformations in Erlang, and show that this notation provides an extensible, verifiable and executable specification language for refactoring. To demonstrate the applicability of our approach, we show how to define and verify a number of example refactorings in the system.

Is it ever permissible for the state to formally recognize one religion? This article focuses on just one aspect of this complex question. To be specific, a critical examination is conducted of what will be called the alienation argument against establishment, according to which this sort of connection between religion and the state is wrong to the extent that it makes some people feel alienated. Whilst this argument is often dismissed in contemporary academic analyses, considerations of this kind are frequently heard in political commentary and public debate. In this article, the three most important criticisms made of the alienation argument, which are referred to as psychologism, indeterminacy and falsity, are discussed. By arguing that it may be possible to rebut these three objections, at least to some extent, the aim of the article is to mount a partial defence of the alienation argument against religious establishment. Its conclusion is that feelings of alienation should play a significant role in making all-things-considered judgements about the justifiability of particular instances of establishment.

Introduction The Dementias Platform UK (DPUK) Data Portal is a secure, accessible environment facilitating provision of rich data towards the largest Dementia, cognition and ageing community of cohort studies in the world. DPUK is also providing services for cohort studies and researchers to maximise the research potential of the programme’s community. Objectives and Approach As part of the engagement of DPUK cohorts with the Data Portal, cohorts will upload data onto the DPUK instance of UK Secure eResearch Platform infrastructure. The Data Portal allows access to a collaborative working space that allows cohorts to enrich their own data, perform their own analysis, and enhance the research potential of their data whilst making use of expertise at various DPUK sites, such as data linking, curation and multi-modal specialism. Cohort data divided into ontologies allows researchers to access data specific to their study needs and can be requested from multiple cohorts simultaneously. Results By utilising the Data Portal researchers have access to cohort data that has been prepared for dementia epidemiology using the agreed ontologies, providing more rapid access to cohort data that otherwise may be large and complex. The knowledge and experience of DPUK staff and collaborators can also help to guide nascent cohorts and feasibility studies into producing research-ready datasets, enabling them to achieve greater impact with their data. A range of analytical tools are provided on the Data Portal making analysis of a cohort’s own data or multiple independent datasets more accessible. Alongside data curation, DPUK also facilitates data linkage to routine sources, beginning with a Wales-wide use case that will expand to the UK over the course of the project. Conclusion/Implications Data from international sources accessible using a central platform permits international collaboration, with ontologies allowing previously disparate data to be combined and analysed to build knowledge and research impact. DPUK projects create policy leading results and operational research standards, enhancing cohort impact and discovery of benefits for Dementia patients.

Introduction Electronic health records (EHR) are a powerful resource in enabling large-scale healthcare research. EHRs often lack detailed disease-specific information that is collected in free text within clinical settings. This challenge can be addressed by using Natural Language Processing (NLP) to derive and extract detailed clinical information from free text. Objectives and Approach Using a training sample of 40 letters, we used the General Architecture for Text Engineering (GATE) framework to build custom rule sets for nine categories of epilepsy information as well as clinic date and date of birth. We used a validation set of 200 clinic letters to compare the results of our algorithm to a separate manual review by a clinician, where we evaluated a “per item” and a “per letter” approach for each category. Results The “per letter” approach identified 1,939 items of information with overall precision, recall and F1-score of 92.7%, 77.7% and 85.6%. Precision and recall for epilepsy specific categories were: diagnosis (85.3%,92.4%), type (93.7%,83.2%), focal seizure (99.0%,68.3%), generalised seizure (92.5%,57.0%), seizure frequency (92.0%,52.3%), medication (96.1%,94.0%), CT (66.7%,47.1%), MRI (96.6%,51.4%) and EEG (95.8%,40.6%). By combining all items per category, per letter we were able to achieve higher precision, recall and F1-scores of 94.6%, 84.2% and 89.0% across all categories. Conclusion/Implications Our results demonstrate that NLP techniques can be used to accurately extract rich phenotypic details from clinic letters that is often missing from routinely-collected data. Capturing these new data types provides a platform for conducting novel precision neurology research, in addition to potential applicability to other disease areas.

Introduction New non-vitamin K Target Specific Oral Anticoagulants (TSOACs) have a favourable risk-benefit profile and debatable cost effectiveness. Large numbers and data from multiple countries in a European study are required to investigate safety issue of TSOACs in subgroups, e.g. people with an intracranial haemorrhage. Objectives and Approach We developed an approach to rapidly replicate data and analyses to support cross-country distributed research within the UK/EU using Electronic Health Records (EHRs). This project was conceptualised and initialled by linking relevant datasets held in multiple data warehouses, in Scotland with the Scottish National Data Safe Haven, and in Wales through the Secure Anonymised Information Linkage (SAIL) databank. Analysts in Edinburgh and Swansea had remote access to each other’s datasets and worked collaboratively to harmonise variables and analysis scripts. . A common R code script has been produced to harmonise individual data as well as the outputs from the study. Results The study screened data on 8M people to develop a cohort that included pseudonymised information of 4,153 individuals in Scotland and 2,676 individuals in Wales, 6,829 individuals in total. Standardised risk analyses were completed in both settings with ongoing work in combining the analyses. In Wales, 39.5% of the patients in the cohort had been admitted to hospitals due to serious vascular events or died caused by these events, after intracranial haemorrhage. Incident rates for male and female are 0.63 and 0.7 respectively. Within the cohort, 0.5% were prescribed with TSOACs and 3% with Warfarin (included as reference). The project is also in the process of including other European jurisdictions. Conclusion/Implications The adopted approach was the simplest, yet most efficient and cost-effective method to ensure consistency in analysis and coherence with currently available governance systems of both safe havens. It can also be considered as an initialisation of developing infrastructure to support research using EHRs across the UK and EU.

Introduction Although healthcare systems generate significant amounts of structured data, there remains a untapped wealth of unstructured narrative data. In the UK, 70% of all NHS digital information is in unstructured form. The NHS has no plans to computerise this data, as it is simply would not be cost effective. Objectives and Approach Our aim was to make all digitised free text within partner organisations accessible for NLP processing for research, while overcoming information governance challenges. We developed a distributed GATE-based NLP platform enabling NLP models to be automatically distributed and materialised against the free text data in each organisation to create new conventional datasets, which can then be transmitted back using an established governance model. This work adds NLP capability to the UK’s National Research Data Appliances, deployed throughout Wales and beyond and uses many open source components enabling a deployment without additional software licence costs, leading to increased potential use cases. Results We have been able to demonstrate a fully federated network of analytical nodes into NHS Wales, which takes the analytical NLP model to the free text data, as opposed to the data having to travel. Under a common, acceptable, governance model, an approval system enables organisations such as health boards to give permission for projects and NLP models to be used against their data. In a proof of concept project, we have run a number of NLP models over large numbers of documents, which the platform has ingested, converted and analysed. We have developed a proposal for a common NLP model definition format to enable models to be interchangeable between different research groups and systems. Sharing/discovery of established NLP models is key deliverable. Conclusion/Implications The implications of being able to send the query to the data, enables access to this untapped data source, finally enabling the realisation of new datasets, while abiding by any IG framework. The low cost and simplicity will enable a many research opportunities, some of which are already being realised.

Introduction Modern team science requires effective sharing of data and skills. The DPUK Data Portal is a collection of tools, datasets and networks that allows for epidemiologists and specialist researchers alike to access, analyse and investigate cohort and different modalities of routine data across UK and international sources. Objectives and Approach The Portal is housed on an instance of UKSeRP (UK Secure eResearch Platform), that allows customisable infrastructure to be used for multi-modal research (thus far live in genetics, imaging and clinical data) for researchers across the world using remote access technology whilst allowing governance to remain with the data provider. A central team at Swansea University is responsible for data curation and processing, and runs an access procedure for researchers to apply to use data from multiple sources to be analysed in a central analysis environment. Other modalities are similarly hosted, with input from partner sites in Cardiff and Oxford. Results DPUK facilitates data access and research on 49 cohorts, 40 UK-based and 9 international. The centralised repository model including remote access and ability to store and make available different modalities of data, from phenotypic data, to genetic and imaging data, has allowed DPUK to begin to support research of varying topics, from those studying cognitive decline and Dementia as a disease, to those maturing analytical models. By providing access to data platforms specialising in genetics, imaging and routine clinical data, as well as to specialists in disease and biology to aid with its understanding, DPUK has realised a large-scale research exercise combining major data modalities on a central platform, and allow access to such rich data across the world under an umbrella of robust governance. Conclusion/Implications Globally, cohorts are pooling data, expertise and desire to enrich their own aims in partnership with a federated research community to enable in-depth scrutiny of the biological origins of dementia and the development and evaluation of novel approach to disease prevention and cure.

DPUK relaunched the Data Portal in November 2017 to present openly available information on the data availability and technical capability of the Data Portal, which supports multi-modal research studies with various objectives from disease model validation to observation investigation. DPUK not only brings clinical data together from cohorts, but is now supporting multi-modal studies in genetics and imaging, as well as linkage opportunities to routine data using world-leading technical solutions to data sharing. The capacity, adaptability and sophistication of the UK Secure eResearch Platform which the Portal is housed on, allows for unprecedented levels of centralised access to rich cohort and routine data, which is consequentially leading to international collaboration and development ambition within epidemiology, bioinformatics, research methodology and technical research solutions. As of March 2018, DPUK is supporting 50 cohorts, 41 from the UK and 9 from across the rest of the world, alongside furthering links and access to routine data held in the UK and across the world. 20 research studies are underway, and the DPUK mission to enhance data science within dementia research is leading the conversation for developing a community of excellence in this field and across other research genres.

ABSTRACT Objectives Linkage of “big data” can provide the answers to a variety of health questions that benefit the delivery of patient care, impact of policies, system planning and evaluation. In some jurisdictions, legal and operational barriers may prevent data linkage for research and system evaluation. Collaboration between international research institutions in Canada, Australia and Wales was formed at the Farr Institute International Conference in 2015. This partnership will test privacy-preserving record linkage (PPRL) techniques for linkage accuracy on real datasets held in a Canadian data repository. Approach Bloom filter PPRL techniques have been incorporated into a prototype linkage system. Evaluations on probabilistic linkage using Bloom filters method have shown potential for large-scale record linkage, performing both accurately and efficiently under experimental conditions. The prototype will be used to evaluate the Bloom filter PPRL techniques in 3 phases. Phase 1: 3 tests using simulated data relating to 20 million individuals will be matched to a sub-cohort of 1 million individuals. Phase 2: 100,000 people from hospital inpatient records will be matched to 18 million people in a health system registration file. These tests will inform whether the method can achieve high levels of privacy protection without negatively impacting performance and linkage quality. Performance indicators include match rate and processing efficiency based on record volumes. Results Linkage quality will be assessed by the number of true matches and non matches identified as links and non-links. This method will be evaluated using synthetic and real-world datasets, where the true match status is known. Initial performance testing linked a file of 3,000 records to 30,000 with a 100% match result. Subsequent test phases as above will continue to be evaluated and these results will be presented. Conclusion Completion of the phased tests will confirm the ability to link datasets while preserving privacy. This international collaboration will expand the utility of this prototype linkage system and expand the global knowledge bank focusing on PPRL methods in general. It will also inform how to adapt to local requirements by providing a solution to many common legal and administrative challenges.

ABSTRACT Background As part of an initiative funded by FARR institute, Swansea University Health Informatics Unit embarked on the creation of the UK Secure e-Research Platform (UKSeRP) to both build on the experience developed by the SAIL databank over many years and to create an infrastructure that can support, at scale, multiple national research programmes. Method The system was created utilising the in-house technical staff and combines multiple vender solutions all orchestrated through bespoke developments. The same team also developed the National Research Data Appliance (NRDA) and elements of this initiate were integrated into the core of this development. Results The system went live at the end of February 2016 and is already supporting five national level research programmes. The presentation will show now the system is architected and how it evolved over time as additional opportunities arose. It is imperative to both provide common architecture and systems that can scale to meet the research programmes needs and yet still allow for additional unique aspects to be provisioned that deliver against the programmes research goals and differing governance models. The platform has gained an ISO 27001 accreditation which provided clear control and management over information security but also allows programmes utilising the platform to inherit these key controls and if desirable to gain their own accreditation easier, as the SAIL Databank has demonstrated. The platform has an ambitus road map and is currently adding imaging capabilities supported by an imaging processing pipeline. Genomics handling, processing and analysis capabilities are being added in the next couple of months. Conclusion The demand for UKSeRP has been higher than initially envisaged and the possible use cases are constantly being imagined. We are also introducing an entry level UKSeRP to support individual projects and their requirements.

ABSTRACT Background Established by the Medical Research Council (MRC) in June 2014, it is a £53 million collaboration between universities and drug companies to transform the best dementia research into the best treatments as quickly as possible. DPUK will bring together 33 different dementia cohorts as well as imaging, brain bank, genomics and NLP. Method To achieve this DPUK has a large number of work packages developing knowledge, capability and science. Work package 2 - DPUK Data portal has many cross cutting themes but its main aim is to provide a central access mechanism for the researchers to access the diverse datasets in a secure research platform which is delivered using the UK Secure e-Research Platform (UKSeRP). The Portal supports the work of WP1 Cohort Profiling by providing information capture systems as well as automated publicly accessible summaries and more detailed reporting for DPUK members. There are obvious sensitives around the sharing of the cohort datasets, especially given the many years of investment and commitment by the cohorts PI. However, many of these cohorts are no longer funded and risk of becoming obsolete and possibly being lost unless they become adopted by a programme such as DPUK. UKSeRP will provide a technical platform far more advanced than is often available to the cohorts and therefore the cohorts will benefit from these additional capabilities. The system leaves data ownership with the cohort PI and the governance structure puts them in the centre of the process and ensures that the cohort data is only used in projects that the cohorts are comfortable with. The Portal supports the project application and approval process, and supports the governance structure and cohort engagement process. The application process is common regardless of the type of dataset required (phenotypic, genomic, imaging, physical samples) Conclusion DPUK has extended UKSeRP to add the technical capabilities developed jointly with other work packages, most notably the addition of a fully federated image retrieval and analysis system and the genomics analysis system. The platform will be fully operational by the end of the year.

Following publication of the original article [1], a typographical error in the formula for calculating d i in the “Scans for local adaptation” subsection in the Method section, was identified. The correct formula should be