Search

Background & Objectives Trial Emulation methods can be applied to Population-Scale Linked Electronic Health Records (EHR) to generalise the results of Randomised Controlled Trials (RCTs) to a broader patient population. However, the Data Generating Mechanisms (DGMs) associated with EHR data mean that there is a need to adjust for informative observations, i.e. worse prognosis patients may have more observations. Approach Using the Anglo-Scandinavian Cardiac Outcomes Trial-Blood Pressure Lowering Arm (ASCOT-BPLA) as a case study: a RCT of a calcium-channel blocker-based regimen versus a β-blocker-based regimen, an emulated trial was created from population-scale, individual-level, linked anonymized EHR data curated as part of the Wales Multimorbidity e-cohort (WMC) for individuals who would have been eligible for ASCOT-BPLA but who also had Multiple Long-Term Conditions (MLTCs). Outcomes were systolic and diastolic blood pressure estimated/predicted at treatment initiation, 6 weeks, 3 months, and 6 monthly until 4 years. Methods & Results Predictions were made using a fully probabilistic Bayesian joint model which allowed for dropout due to death and which either did or did not also adjust for the frequency of blood pressure observations available. In previous simulation studies such a joint modelling approach has been shown to out-perform other adjustment methods in terms of bias and coverage. Conclusions & Implications Whilst EHR data offer an attractive platform in which to evaluate interventions in people living with MLTCs, the inherent DGMs associated with them need to be taken into account, and the use of a joint modelling approach is an attractive way to achieve this.

Background and Objectives Polypharmacy is increasingly common in people living with multiple long-term conditions (MLTC), but clinical evidence often does not consider medicine interactions beyond two or three treatments at a time. The aim of this research is to develop statistical models to better understand interactions between treatments leading to adverse events for people living with MLTC. Approach We will develop models of risk of adverse events in people with MLTC and polypharmacy using data in the Secure Anonymised Information Linkage (SAIL) Databank - a trusted research environment with linked electronic health record data at the individual-level for the population of Wales, UK. Taking a flexible Bayesian joint modelling approach, where the posterior risk is the product of several sub-models, will enable the characterisation of this complex system in a way that is interpretable for healthcare decision-makers, leading to improved understanding of risk factors for adverse events. Results The Welsh Multimorbidity e-Cohort (WMC) includes 2.9 million people alive and living in Wales on 1st Jan 2000 with follow-up for 20 years. Individuals in the cohort have 932,552,061 medication items prescribed and 277,279 coded adverse events recorded in primary care data. Modelling approaches will focus on feasibility of fitting models to population-level datasets, and the interpretability of the model results for healthcare decision-making. Conclusions and Implications Results have the potential to inform healthcare policy and practice for the management of people living with MLTC and polypharmacy, thus improving patient outcomes and making the best use of limited healthcare resources.

Objective and Approach Managers, clinicians and policymakers have limited information on near-real time patterns of attendance at emergency departments (ED). Our aim was to develop a surveillance system through extracting information from the presenting complaint to develop and feed activity dashboards, improve injury surveillance and support research. We developed a Natural Language Processing (NLP) pipeline using the General Architecture for Text Engineering (GATE) system. The pipeline produced Unified Medical Language System (UMLS) codes from ED presenting complaint records and also implemented the Joint Action on Monitoring Injuries in Europe (JAMIE) Minimum Dataset. We classified complaints into 31 common problem presentations (Coughs, Bleeds, Mental Health, Dizziness...), along with categories from JAMIE (Mechanism of Injury, Location, Activity, Intent…). Results Preliminary results are from a pilot study of 86,416 presenting complaints from Morriston Hospital in 2014. The most common terms used for grouped symptoms/categories were “Injury” (38.6%, n=28691), “Pain” (22.8%, n=19420), “Unwell” (7.2%, n=6117), “Bleeding” (6.7%, n=5760), and “Shortness of Breath” (3.8%, n=3255). 32.3% of records contained Mechanism of Injury codes, including, falls (15.2%, n=13174), struck by/against (6.9%, n=5960), cuts (5.6%, n=4833), road traffic collisions (2.5%, n=2124) and poisoning (1%, n=900). We are continuing to refine and validate the categories using precision and recall and with comparison with coded diagnoses. Conclusions and Implications NLP can be used to extract valuable additional detailed information from ED data. We will further develop the algorithms for all hospitals in Wales for data from 2023 and design dashboards to produce a near-real time surveillance system.

Objective and Approach The study aimed to utilise multiple population-wide linked administrative, health, and survey data sources to identify victims from violence across various sociodemographic groups using healthcare service interactions. We calculated violence rates for the population of Wales to identify groups of people that experience higher rates of violence compared to the general population. Our approach involved creating an observational population-wide e-cohort utilising data held in the Secure Anonymised Information Linkage (SAIL) Databank. General practice records, emergency department attendances, hospital admissions, and the Office of National Statistics (ONS) deaths register were linked to the Welsh Demographic Service dataset to include individuals living in Wales, UK between 2013-2022, allowing for a 10-year analysis. Results In total, 3.3 million individuals were included in the study, with 80,079 individuals having interacted with health services or died due to violence over the 10-year period. On average, every 4 per 1000 of the study population had experienced violence (2013-2022) with the majority of these being captured through emergency department attendances. 55% were male, of which 71% were aged under 35. The majority of individuals lived in the more deprived areas of Wales (61%). Conclusions and Implications It is important to understand and quantify violence over time to produce insight into which sociodemographic groups of individuals are being disproportionately impacted by violence for designing disparity-reducing policy interventions. ONS Census 2021 data will be used to examine the impact of violence by protected characteristics (e.g. disability, race, sexual orientation) under the UK Equality Act 2010.

Background and Objectives Healthcare decision-making has previously focused on developing recommendations for single conditions. This research aimed to develop a modeling framework to estimate the effects of multi-indication treatment in people living with multiple long-term conditions. Approach Using a case study in type 2 diabetes mellitus [T2DM], chronic kidney disease [CKD], and heart failure [HF], age-adjusted multistate models were used to estimate the effect of sodium-glucose cotransporter-2 (SGLT2) inhibitors as a multi-indication treatment. Baseline hazards were estimated using population-scale, individual-level, linked anonymized electronic health record (EHR) data for 613,195 individuals aged 55 to 85 years in Wales over a 20-year period. Hazard ratios for treatment effects compared to standard of care were obtained from randomized controlled trials for combinations of multiple-long term conditions and incorporated in a patient-level simulation. Results SGLT2 inhibitors increased the estimated mean life expectancy from 0.02 (95% Confidence Interval (CI): -0.02, 0.05) to 1.44 (95% CI: 1.12, 1.74) years in people living with coexisting T2DM, CKD, and HF. The estimated gain in life expectancy was less than 0.5 years for individuals with T2DM, CKD, and HF in different temporally ordered sequences. The estimated mean time to develop HF increased by 3.1 (95% CI: 3.00, 3.24) and 1.34 (95% CI: 1.12, 1.56) years in individuals with CKD, and in those with CKD followed by T2DM. Conclusions and Implications There is an ever increasing need to appraise interventions in people living with multiple long-term conditions to identify optimal treatment strategies and reduce polypharmacy. Multistate models applied to linked EHRs allow for a more rigorous assessment of treatment effects to inform healthcare policy.

Background and Objective Healthcare research faces challenges in developing metrics that resonate with the general public or policymakers. We created a Desirable Health Service Utilisation Indicator (DHSUI) to address this gap, centred around New Year’s wishes for survival and non-occurrence of undesired events in the following year, for the population of Wales, UK, following discussions with policy-makers. Methods We accessed and linked individual electronic healthcare records from primary and secondary care sources within the Secure Anonymised Information Linkage (SAIL) Databank. The study used data for all Welsh General Practice registered individuals on the 1st January 2015 with follow-up until residency break, death, or 31st December 2022. The DHSUI was calculated per person per year and quantified the distribution of the population who survive calendar years and do not use selected health services (not admitted to hospital; no emergency department attendance; and not prescribed antibiotics, analgesics, or mental health drugs). We linked area level demographic data to measure socio-economic differences in DHSUI distribution. Results Focussing on 2022, a population of 2,668,977 people were included, with 1,150,347 (43%) identified as meeting the indicator. The proportion of desirable health decreased significantly with age and with increasing socio-economic deprivation. A higher proportion of males (50%) had desirable health compared to females (37%). The most prevalent sub-indicator among those with undesirable health was antibiotic prescriptions accounting for 52% of cases. Conclusion This DHSUI indicator is easy to interpret and will assist policymakers and the public in understanding the magnitude and distribution of desirable healthcare utilisation.

Objective This study aims to create a national ethnicity spine based on all available ethnicity records in linkable anonymised electronic health record and administrative data sources.Design A longitudinal study using anonymised individual-level population-scale ethnicity data from 26 data sources available within the Secure Anonymised Information Linkage Databank.Setting The national ethnicity spine is created based on longitudinal national data for the population of Wales-UK over 22 years (between 2000 and 2021).Procedure and participants A total of 46 million ethnicity records for 4 297 694 individuals have been extracted, harmonised, deduplicated and made available within a longitudinal research ready data asset.Outcome measures (1) Comparing the distribution of ethnicity records over time for four different selection approaches (latest, mode, weighted mode and composite) across age bands, sex, deprivation quintiles, health board and residential location and (2) distribution and completeness of records against the ONS census 2011.Results The distribution of the dominant group (white) is minimally affected based on the four different selection approaches. Across all other ethnic group categorisations, the mixed group was most susceptible to variation in distribution depending on the selection approach used and varied from a 0.6% prevalence across the latest and mode approach to a 1.1% prevalence for the weighted mode, compared with the 3.1% prevalence for the composite approach. Substantial alignment was observed with ONS 2011 census with the Latest group method (kappa=0.68, 95% CI (0.67 to 0.71)) across all subgroups. The record completeness rate was over 95% in 2021.Conclusion In conclusion, our development of the population-scale ethnicity spine provides robust ethnicity measures for healthcare research in Wales and a template which can easily be deployed in other trusted research environments in the UK and beyond.

Objectives To compare the patterns of multimorbidity between people with and without rheumatic and musculoskeletal diseases (RMDs) and to describe how these patterns change by age and sex over time, between 2010 and 2019.Participants 103 426 people with RMDs and 2.9 million comparators registered in 395 Wales general practices (GPs). Each patient with an RMD aged 0–100 years between January 2010 and December 2019 registered in Clinical Practice Research Welsh practices was matched with up to five comparators without an RMD, based on age, gender and GP code.Primary outcome measures The prevalence of 29 Elixhauser-defined comorbidities in people with RMDs and comparators categorised by age, gender and GP practices. Conditional logistic regression models were fitted to calculate differences (OR, 95% CI) in associations with comorbidities between cohorts.Results The most prevalent comorbidities were cardiovascular risk factors, hypertension and diabetes. Having an RMD diagnosis was associated with a significantly higher odds for many conditions including deficiency anaemia (OR 1.39, 95% CI (1.32 to 1.46)), hypothyroidism (OR 1.34, 95% CI (1.19 to 1.50)), pulmonary circulation disorders (OR 1.39, 95% CI 1.12 to 1.73) diabetes (OR 1.17, 95% CI (1.11 to 1.23)) and fluid and electrolyte disorders (OR 1.27, 95% CI (1.17 to 1.38)). RMDs have a higher proportion of multimorbidity (two or more conditions in addition to the RMD) compared with non-RMD group (81% and 73%, respectively in 2019) and the mean number of comorbidities was higher in women from the age of 25 and 50 in men than in non-RMDs group.Conclusion People with RMDs are approximately 1.5 times as likely to have multimorbidity as the general population and provide a high-risk group for targeted intervention studies. The individuals with RMDs experience a greater load of coexisting health conditions, which tend to manifest at earlier ages. This phenomenon is particularly pronounced among women. Additionally, there is an under-reporting of comorbidities in individuals with RMDs.

Abstract Background The EVITE Immunity study investigated the effects of shielding Clinically Extremely Vulnerable (CEV) people during the COVID-19 pandemic on health outcomes and healthcare costs in Wales, United Kingdom, to help prepare for future pandemics. Shielding was intended to protect those at highest risk of serious harm from COVID-19. We report the cost of implementing shielding in Wales. Methods The number of people shielding was extracted from the Secure Anonymised Information Linkage Databank. Resources supporting shielding between March and June 2020 were mapped using published reports, web pages, freedom of information requests to Welsh Government and personal communications (e.g. with the office of the Chief Medical Officer for Wales). Results At the beginning of shielding, 117,415 people were on the shielding list. The total additional cost to support those advised to stay home during the initial 14 weeks of the pandemic was £13,307,654 (£113 per person shielded). This included the new resources required to compile the shielding list, inform CEV people of the shielding intervention and provide medicine and food deliveries. The list was adjusted weekly over the 3-month period (130,000 people identified by June 2020). Therefore the cost per person shielded lies between £102 and £113 per person. Conclusion This is the first evaluation of the cost of the measures put in place to support those identified to shield in Wales. However, no data on opportunity cost was available. The true costs of shielding including its budget impact and opportunity costs need to be investigated to decide whether shielding is a worthwhile policy for future health emergencies.

Abstract Background Measurement of multimorbidity in research is variable, including the choice of the data source used to ascertain conditions. We compared the estimated prevalence of multimorbidity and associations with mortality using different data sources. Methods A cross-sectional study of SAIL Databank data including 2,340,027 individuals of all ages living in Wales on 01 January 2019. Comparison of prevalence of multimorbidity and constituent 47 conditions using data from primary care (PC), hospital inpatient (HI), and linked PC-HI data sources and examination of associations between condition count and 12-month mortality. Results Using linked PC-HI compared with only HI data, multimorbidity was more prevalent (32.2% versus 16.5%), and the population of people identified as having multimorbidity was younger (mean age 62.5 versus 66.8 years) and included more women (54.2% versus 52.6%). Individuals with multimorbidity in both PC and HI data had stronger associations with mortality than those with multimorbidity only in HI data (adjusted odds ratio 8.34 [95% CI 8.02-8.68] versus 6.95 (95%CI 6.79-7.12] in people with ≥ 4 conditions). The prevalence of conditions identified using only PC versus only HI data was significantly higher for 37/47 and significantly lower for 10/47: the highest PC/HI ratio was for depression (14.2 [95% CI 14.1–14.4]) and the lowest for aneurysm (0.51 [95% CI 0.5–0.5]). Agreement in ascertainment of conditions between the two data sources varied considerably, being slight for five (kappa

Abstract Background The CVD-COVID-UK consortium was formed to understand the relationship between COVID-19 and cardiovascular diseases through analyses of harmonised electronic health records (EHRs) across the four UK nations. Beyond COVID-19, data harmonisation and common approaches enable analysis within and across independent Trusted Research Environments. Here we describe the reproducible harmonisation method developed using large-scale EHRs in Wales to accommodate the fast and efficient implementation of cross-nation analysis in England and Wales as part of the CVD-COVID-UK programme. We characterise current challenges and share lessons learnt. Methods Serving the scope and scalability of multiple study protocols, we used linked, anonymised individual-level EHR, demographic and administrative data held within the SAIL Databank for the population of Wales. The harmonisation method was implemented as a four-layer reproducible process, starting from raw data in the first layer. Then each of the layers two to four is framed by, but not limited to, the characterised challenges and lessons learnt. We achieved curated data as part of our second layer, followed by extracting phenotyped data in the third layer. We captured any project-specific requirements in the fourth layer. Results Using the implemented four-layer harmonisation method, we retrieved approximately 100 health-related variables for the 3.2 million individuals in Wales, which are harmonised with corresponding variables for > 56 million individuals in England. We processed 13 data sources into the first layer of our harmonisation method: five of these are updated daily or weekly, and the rest at various frequencies providing sufficient data flow updates for frequent capturing of up-to-date demographic, administrative and clinical information. Conclusions We implemented an efficient, transparent, scalable, and reproducible harmonisation method that enables multi-nation collaborative research. With a current focus on COVID-19 and its relationship with cardiovascular outcomes, the harmonised data has supported a wide range of research activities across the UK.

Objectives Active travel to school (ATS), such as walking and cycling, not only reduces carbon emissions and air pollution but also contributes to a myriad of health benefits. Understanding the ‘potential’ for ATS across Wales is poorly understood yet vital to inform policy and practice aimed at increasing ATS. Methods Using geospatial techniques, network distances have been calculated for all residences to all schools in Wales. Distances for each residence were de-identified and imported into the Secure Anonymised Information Linkage (SAIL) databank using SAIL’s split file approach. Within SAIL, anonymised distances were linked to encrypted school locations and a cohort of approximately 440,000 children attending a state school in Wales as recorded in the Education Wales (EDUW) dataset. Travel distance to school was subsequently filtered for each child within our cohort. Results The incorporation of geospatial network derived origin-destination pairs (n=2,205,000,000) allows us to explore how households can engage in ATS, and how this varies by primary/secondary school, welsh-speaking, urban/rural, and socioeconomic status. Initially we will present findings based on distance alone, with future refinements to encompass other factors influencing ATS including: active travel routes, road safety, and topography. We will link our results to anonymised ATS survey responses in SAIL, to calculate specific ATS distance thresholds by age, deprivation, and urban/rural status, including social indicators such as household composition. Conclusion Currently there is limited local or national comparable information on the potential for ATS in Wales. This baseline information is urgently needed to inform ATS policy and planning, and to ensure appropriate ATS interventions are prioritised for schools and communities where need is greatest.

Summary: Background: Understanding and quantifying the differences in disease development in different socioeconomic groups of people across the lifespan is important for planning healthcare and preventive services. The study aimed to measure chronic disease accrual, and examine the differences in time to individual morbidities, multimorbidity, and mortality between socioeconomic groups in Wales, UK. Methods: Population-wide electronic linked cohort study, following Welsh residents for up to 20 years (2000–2019). Chronic disease diagnoses were obtained from general practice and hospitalisation records using the CALIBER disease phenotype register. Multi-state models were used to examine trajectories of accrual of 132 diseases and mortality, adjusted for sex, age and area-level deprivation. Restricted mean survival time was calculated to measure time spent free of chronic disease(s) or mortality between socioeconomic groups. Findings: In total, 965,905 individuals aged 5–104 were included, from a possible 2.9 m individuals following a 5-year clearance period, with an average follow-up of 13.2 years (12.7 million person-years). Some 673,189 (69.7%) individuals developed at least one chronic disease or died within the study period. From ages 10 years upwards, the individuals living in the most deprived areas consistently experienced reduced time between health states, demonstrating accelerated transitions to first and subsequent morbidities and death compared to their demographic equivalent living in the least deprived areas. The largest difference were observed in 10 and 20 year old males developing multimorbidity (−0.45 years (99% CI: −0.45, −0.44)) and in 70 year old males dying after developing multimorbidity (−1.98 years (99% CI: −2.01, −1.95)). Interpretation: This study adds to the existing literature on health inequalities by demonstrating that individuals living in more deprived areas consistently experience accelerated time to diagnosis of chronic disease and death across all ages, accounting for competing risks. Funding: UK Medical Research Council, Health Data Research UK, and Administrative Data Research Wales.

Abstract There is a need for better understanding of the risk of thrombocytopenic, haemorrhagic, thromboembolic disorders following first, second and booster vaccination doses and testing positive for SARS-CoV-2. Self-controlled cases series analysis of 2.1 million linked patient records in Wales between 7th December 2020 and 31st December 2021. Outcomes were the first diagnosis of thrombocytopenic, haemorrhagic and thromboembolic events in primary or secondary care datasets, exposure was defined as 0–28 days post-vaccination or a positive reverse transcription polymerase chain reaction test for SARS-CoV-2. 36,136 individuals experienced either a thrombocytopenic, haemorrhagic or thromboembolic event during the study period. Relative to baseline, our observations show greater risk of outcomes in the periods post-first dose of BNT162b2 for haemorrhagic (IRR 1.47, 95%CI: 1.04–2.08) and idiopathic thrombocytopenic purpura (IRR 2.80, 95%CI: 1.21–6.49) events; post-second dose of ChAdOx1 for arterial thrombosis (IRR 1.14, 95%CI: 1.01–1.29); post-booster greater risk of venous thromboembolic (VTE) (IRR-Moderna 3.62, 95%CI: 0.99–13.17) (IRR-BNT162b2 1.39, 95%CI: 1.04–1.87) and arterial thrombosis (IRR-Moderna 3.14, 95%CI: 1.14–8.64) (IRR-BNT162b2 1.34, 95%CI: 1.15–1.58). Similarly, post SARS-CoV-2 infection the risk was increased for haemorrhagic (IRR 1.49, 95%CI: 1.15–1.92), VTE (IRR 5.63, 95%CI: 4.91, 6.4), arterial thrombosis (IRR 2.46, 95%CI: 2.22–2.71). We found that there was a measurable risk of thrombocytopenic, haemorrhagic, thromboembolic events after COVID-19 vaccination and infection.

BackgroundMultimorbidity prevalence rates vary considerably depending on the conditions considered in the morbidity count, but there is no standardised approach to the number or selection of conditions to include.Methods and findingsWe conducted a cross-sectional study using English primary care data for 1,168,260 participants who were all people alive and permanently registered with 149 included general practices. Outcome measures of the study were prevalence estimates of multimorbidity (defined as ≥2 conditions) when varying the number and selection of conditions considered for 80 conditions. Included conditions featured in ≥1 of the 9 published lists of conditions examined in the study and/or phenotyping algorithms in the Health Data Research UK (HDR-UK) Phenotype Library. First, multimorbidity prevalence was calculated when considering the individually most common 2 conditions, 3 conditions, etc., up to 80 conditions. Second, prevalence was calculated using 9 condition-lists from published studies. Analyses were stratified by dependent variables age, socioeconomic position, and sex. Prevalence when only the 2 commonest conditions were considered was 4.6% (95% CI [4.6, 4.6] p < 0.001), rising to 29.5% (95% CI [29.5, 29.6] p < 0.001) considering the 10 commonest, 35.2% (95% CI [35.1, 35.3] p < 0.001) considering the 20 commonest, and 40.5% (95% CI [40.4, 40.6] p < 0.001) when considering all 80 conditions. The threshold number of conditions at which multimorbidity prevalence was >99% of that measured when considering all 80 conditions was 52 for the whole population but was lower in older people (29 in >80 years) and higher in younger people (71 in 0- to 9-year-olds). Nine published condition-lists were examined; these were either recommended for measuring multimorbidity, used in previous highly cited studies of multimorbidity prevalence, or widely applied measures of "comorbidity." Multimorbidity prevalence using these lists varied from 11.1% to 36.4%. A limitation of the study is that conditions were not always replicated using the same ascertainment rules as previous studies to improve comparability across condition-lists, but this highlights further variability in prevalence estimates across studies.ConclusionsIn this study, we observed that varying the number and selection of conditions results in very large differences in multimorbidity prevalence, and different numbers of conditions are needed to reach ceiling rates of multimorbidity prevalence in certain groups of people. These findings imply that there is a need for a standardised approach to defining multimorbidity, and to facilitate this, researchers can use existing condition-lists associated with highest multimorbidity prevalence.

There is still limited understanding of how chronic conditions co-occur in patients with multimorbidity and what are the consequences for patients and the health care system. Most reported clusters of conditions have not considered the demographic characteristics of these patients during the clustering process. The study used data for all registered patients that were resident in Fife or Tayside, Scotland and aged 25 years or more on 1st January 2000 and who were followed up until 31st December 2018. We used linked demographic information, and secondary care electronic health records from 1st January 2000. Individuals with at least two of the 31 Elixhauser Comorbidity Index conditions were identified as having multimorbidity. Market basket analysis was used to cluster the conditions for the whole population and then repeatedly stratified by age, sex and deprivation. 318,235 individuals were included in the analysis, with 67,728 (21·3%) having multimorbidity. We identified five distinct clusters of conditions in the population with multimorbidity: alcohol misuse, cancer, obesity, renal failure, and heart failure. Clusters of long-term conditions differed by age, sex and socioeconomic deprivation, with some clusters not present for specific strata and others including additional conditions. These findings highlight the importance of considering demographic factors during both clustering analysis and intervention planning for individuals with multiple long-term conditions. By taking these factors into account, the healthcare system may be better equipped to develop tailored interventions that address the needs of complex patients.

Rates of Multimorbidity (also called Multiple Long Term Conditions, MLTC) are increasing in many developed nations. People with multimorbidity experience poorer outcomes and require more healthcare intervention. Grouping of conditions by health service utilisation is poorly researched. The study population consisted of a cohort of people living in Wales, UK aged 20 years or older in 2000 who were followed up until the end of 2017. Multimorbidity clusters by prevalence and healthcare resource use (HRU) were modelled using hypergraphs, mathematical objects relating diseases via links which can connect any number of diseases, thus capturing information about sets of diseases of any size. The cohort included 2,178,938 people. The most prevalent diseases were hypertension (13.3%), diabetes (6.9%), depression (6.7%) and chronic obstructive pulmonary disease (5.9%). The most important sets of diseases when considering prevalence generally contained a small number of diseases, while the most important sets of diseases when considering HRU were sets containing many diseases. The most important set of diseases taking prevalence and HRU into account was diabetes & hypertension and this combined measure of importance featured hypertension most often in the most important sets of diseases. We have used a single approach to find the most important sets of diseases based on co-occurrence and HRU measures, demonstrating the flexibility of the hypergraph approach. Hypertension, the most important single disease, is silent, underdiagnosed and increases the risk of life threatening co-morbidities. Co-occurrence of endocrine and cardiovascular diseases was common in the most important sets. Combining measures of prevalence with HRU provides insights which would be helpful for those planning and delivering services.

BackgroundMultimorbidity is one of the greatest challenges facing health and social care systems globally. It is associated with high rates of health service use, adverse healthcare events, and premature death. Despite its importance, little is known about the effects of contextual determinants such as household and area characteristics on health and care outcomes for people with multimorbidity. This study protocol presents a plan for the examination of associations between individual, household, and area characteristics with important health and social care outcomes.MethodsThe study will use a cross-section of data from the SAIL Databank on 01 January 2019 and include all people alive and registered with a Welsh GP. The cohort will be stratified according to the presence or absence of multimorbidity, defined as two or more long-term conditions. Multilevel models will be used to examine covariates measured for individuals, households, and areas to account for social processes operating at different levels. The intra-class correlation coefficient will be calculated to determine the strength of association at each level of the hierarchy. Model outcomes will be any emergency department attendance, emergency hospital or care home admission, or mortality, within the study follow-up period.DiscussionHousehold and area characteristics might act as protective or risk factors for health and care outcomes for people with multimorbidity, in which case results of the analyses can be used to guide clinical and policy responses for effective targeting of limited resources.

IntroductionAt the start of the COVID-19 pandemic there was an urgent need to identify individuals at highest risk of severe outcomes, such as hospitalisation and death following infection. The QCOVID risk prediction algorithms emerged as key tools in facilitating this which were further developed during the second wave of the COVID-19 pandemic to identify groups of people at highest risk of severe COVID-19 related outcomes following one or two doses of vaccine.ObjectivesTo externally validate the QCOVID3 algorithm based on primary and secondary care records for Wales, UK.MethodsWe conducted an observational, prospective cohort based on electronic health care records for 1.66m vaccinated adults living in Wales on 8th December 2020, with follow-up until 15th June 2021. Follow-up started from day 14 post vaccination to allow the full effect of the vaccine.ResultsThe scores produced by the QCOVID3 risk algorithm showed high levels of discrimination for both COVID-19 related deaths and hospital admissions and good calibration (Harrell C statistic: ≥ 0.828).ConclusionThis validation of the updated QCOVID3 risk algorithms in the adult vaccinated Welsh population has shown that the algorithms are valid for use in the Welsh population, and applicable on a population independent of the original study, which has not been previously reported. This study provides further evidence that the QCOVID algorithms can help inform public health risk management on the ongoing surveillance and intervention to manage COVID-19 related risks.

To inform the public and policy makers, we investigated and compared the risk of cerebral venous sinus thrombosis (CVST) after SARS-Cov-2 vaccination or infection using a national cohort of 2,643,699 individuals aged 17 y and above, alive, and resident in Wales on 1 January 2020 followed up through multiple linked data sources until 28 March 2021. Exposures were first dose of Oxford-ChAdOx1 or Pfizer-BioNTech vaccine or polymerase chain reaction (PCR)-confirmed SARS-Cov-2 infection. The outcome was an incident record of CVST. Hazard ratios (HR) were calculated using multivariable Cox regression, adjusted for confounders. HR from SARS-Cov-2 infection was compared with that for SARS-Cov-2 vaccination. We identified 910,556 (34.4%) records of first SARS-Cov-2 vaccination and 165,862 (6.3%) of SARS-Cov-2 infection. A total of 1,372 CVST events were recorded during the study period, of which 52 (3.8%) and 48 (3.5%) occurred within 28 d after vaccination and infection, respectively. We observed slight non-significant risk of CVST within 28 d of vaccination [aHR: 1.34, 95% CI: 0.95-1.90], which remained after stratifying by vaccine [BNT162b2, aHR: 1.18 (95% CI: 0.63-2.21); ChAdOx1, aHR: 1.40 (95% CI: 0.95-2.05)]. Three times the number of CVST events is observed within 28 d of a positive SARS-Cov-2 test [aHR: 3.02 (95% CI: 2.17-4.21)]. The risk of CVST following SARS-Cov-2 infection is 2.3 times that following SARS-Cov-2 vaccine. This is important information both for those designing COVID-19 vaccination programs and for individuals making their own informed decisions on the risk-benefit of vaccination. This record-linkage approach will be useful in monitoring the safety of future vaccine programs.

Introduction Shielding aimed to protect those predicted to be at highest risk from COVID-19 and was uniquely implemented in the UK during the COVID-19 pandemic. Clinically extremely vulnerable people identified through algorithms and screening of routine National Health Service (NHS) data were individually and strongly advised to stay at home and strictly self-isolate even from others in their household. This study will generate a logic model of the intervention and evaluate the effects and costs of shielding to inform policy development and delivery during future pandemics.Methods and analysis This is a quasiexperimental study undertaken in Wales where records for people who were identified for shielding were already anonymously linked into integrated data systems for public health decision-making. We will: interview policy-makers to understand rationale for shielding advice to inform analysis and interpretation of results; use anonymised individual-level data to select people identified for shielding advice in March 2020 and a matched cohort, from routine electronic health data sources, to compare outcomes; survey a stratified random sample of each group about activities and quality of life at 12 months; use routine and newly collected blood data to assess immunity; interview people who were identified for shielding and their carers and NHS staff who delivered healthcare during shielding, to explore compliance and experiences; collect healthcare resource use data to calculate implementation costs and cost–consequences. Our team includes people who were shielding, who used their experience to help design and deliver this study.Ethics and dissemination The study has received approval from the Newcastle North Tyneside 2 Research Ethics Committee (IRAS 295050). We will disseminate results directly to UK government policy-makers, publish in peer-reviewed journals, present at scientific and policy conferences and share accessible summaries of results online and through public and patient networks.

Objectives Various analgesics are frequently prescribed to cancer patients for whom pain contributes to poor physical and emotional health and well-being. We examined changes in trends of analgesic prescribing in over 35,000 cancer patients diagnosed in the Welsh population before and during the COVID-19 pandemic in order to gain insight into the COVID-19 pandemic effects on cancer patients’ ability to receive analgesia and their potential ability to control their pain via medications. Approach Within the Secure Anonymised Information Linkage (SAIL) Databank trusted research environment (TRE), patients diagnosed with incident primary breast, lung, colorectal or prostate cancers during 2017–2021 were obtained from Cancer Network Information System Cymru (CaNISC) dataset and patients’ prescription records were identified from Welsh Longitudinal General Practitioner (WLGP) dataset before being linked to their oncology e-record. We calculated opioid and non-opioid analgesic items prescribed per patient per year (PPPY) since cancer by clinical and demographic factors including cancer type, stage at diagnosis, diagnosis year, age at diagnosis, sex, comorbidities and patients’ socioeconomic status. These factors were included to model the effects of the COVID-19 pandemic on trends in analgesic prescribing for each cancer group. Results We detected significant differences in the number of analgesic items prescribed PPPY in patients diagnosed before the COVID-19 pandemic (2017–2019) and those during the pandemic (2020–2021), with 1.3 more items PPPY prescribed for the latter group (p

Objectives Having multiple long term health conditions (MLTCs), also known as multimorbidity, is becoming increasingly common as populations age. Understanding how clusters of diseases are likely to lead to other diseases and the effect of multimorbidity on healthcare resource use (HRU) will be of great importance as this trend continues. Approach Graph-based approaches, also called network analysis in the literature, have been used previously to study multimorbidity. The use of hypergraphs, which are generalisations of graphs where edges can connect to any number of nodes, and their application to the problem of understanding multimorbidity will be discussed. Analysis using hypergraphs was carried out using a population-scale cohort of people in the Secure Anonymised Information Linkage (SAIL) Databank to find the diseases and disease sets which are most important based on a measure of prevalence and measures of healthcare resource utilisation in secondary care. Results The most important sets of diseases based on the centrality of a hypergraph weighted by a measure of prevalence featured hypertension, and the most important was hypertension and diabetes. The most important sets of diseases based on the centrality of a hypergraph weighted by a measure of unplanned inpatient HRU were arrhythmia, heart failure and hypertension while for a measure of outpatient HRU the most important set of diseases was diabetes and hypertension. Conclusion Hypergraphs are very flexible and general mathematical objects and there is still a great deal of development that can be done to make them more useful in epidemiological settings and beyond.

The COVID-19 pandemic has placed a spotlight on existing and enduring health inequalities experienced by different ethnic groups. There has been a longstanding call to generate and improve the use of ethnicity data available across different data sources, in order to improve our understanding of health risks, behaviours and outcomes. We used multiple anonymised individual-level population-scale data sources available within the Secure Anonymised Information Linkage (SAIL) Databank to develop a harmonised ethnicity spine for the population of Wales. We documented ethnicity information in multiple longitudinal records from January 2000 onwards. Data sources included: health and social care, birth and mortality records, national census records, specialist clinical audits and registers, surveys and other routine electronic data. To enable multi-source harmonisation, we explored the ethnicity categorisation as well as temporal changes in recording and classifications by obtaining distribution of records for population, which informed our harmonisation algorithm for standardisation of ethnicity records. We used over 20-data sources on ~5-million individuals, spanning varying time-periods starting from January 2000 upto a maximum of 22-years. We harmonised available recorded ethnicity values into standardised ethnic classification groups within a national ethnicity-spine. Furthermore, we investigated the impact of different harmonisation methods, including composite, latest date of recording, modal and weighted modal results. With the main focus of the methodological development being in response to the COVID-19 pandemic, when linked to the ~3.1 million individuals alive and resident in Wales from January 2020, we generated harmonised ethnic groups towards ~95% completeness in data coverage for the whole population of Wales. The predominant ethnic group in Wales observed was White, accounting for 89% of the population when using the latest date of recording method. This research highlights challenges in using longitudinal ethnicity data across different sources. Further work is needed to understand the basis on which individuals / organisations record ethnicity overtime. We recommend improvements recognising differences between ethnicity and other social constructs (e.g. ancestry, nationality, country of origin) are better documented / understood.

Objectives The COVID-19 pandemic has had a detrimental impact on healthcare utilisation, resulting in increased mortality both directly and indirectly associated with COVID-19. We aimed to assess the impact of the COVID-19 pandemic on all-cause and disease-specific mortality and further explore the impact of potential inequalities, deprivation status and ethnicity. Approach Population-scale, individual-level, anonymised linked, routinely-collected electronic health records from demographic and administrative sources were used for two cohorts: i) C19-COHORT16 included individuals alive and resident in Wales on the 1st January 2016 with follow-up until death, break-in Welsh residency, or 31st December 2019; ii) C19-COHORT20 included individuals alive and resident in Wales on 1st January 2020 with follow-up until death, break-in Welsh residency, or study end. We used time-series analysis to investigate trends in mortality over time. We fitted negative binomial models to estimate expected all-cause and disease-specific mortality and compared these estimates to observed mortality in C19-COHORT20. Results Excess all-cause and COVID19-related deaths were higher during the period where the alpha variant was dominant. The trend in deaths decreased during the omicron dominant period. The Asian population had increased mortality during the period where the delta variant was dominant. Mortality was increased for most deprived groups compared to least deprived groups, however, the magnitude of this effect remained unchanged during the pandemic. COVID-19 indirectly affected cancer, circulatory, trauma, digestive and mental health related deaths, with a higher than expected mortality. The majority of trauma related deaths occurred early on in the pandemic, where a higher than expected number of deaths occurred outside of an NHS establishment. Mortality associated with respiratory disease (unrelated to COVID-19) was significantly lower than expected during the COVID-19 pandemic. Conclusion Increased all-cause and disease-specific mortality was observed during the COVID-19 pandemic. Excess deaths may be a result of reduced healthcare utilisation, delayed investigation and/or treatment of chronic diseases. As healthcare systems recover from COVID-19, investigation of mortality trends will play a central role in healthcare planning, utilisation and resource use.

Background Population-level information on dispensed medication provides insight on the distribution of treated morbidities, particularly if linked to other population-scale data at an individual-level. Objective To evaluate the impact of COVID-19 on dispensing patterns of medications. Methods Retrospective observational study using population-scale, individual-level dispensing records in Wales, UK. Total dispensed drug items for the population between 1st January 2016 and 31st December 2019 (3-years, pre-COVID-19) were compared to 2020 with follow up until 27th July 2021 (COVID-19 period). We compared trends across all years and British National Formulary (BNF) chapters and highlighted the trends in three major chapters for 2019-21: 1-Cardiovascular system (CVD); 2-Central Nervous System (CNS); 3-Immunological & Vaccine. We developed an interactive dashboard to enable monitoring of changes as the pandemic evolves. Result Amongst all BNF chapters, 73,410,543 items were dispensed in 2020 compared to 74,121,180 items in 2019 demonstrating -0.96% relative decrease in 2020. Comparison of monthly patterns showed average difference (D) of -59,220 and average Relative Change (RC) of -0.74% between the number of dispensed items in 2020 and 2019. Maximum RC was observed in March 2020 (D= +1,224,909 and RC= +20.62%), followed by second peak in June 2020 (D= +257,920, RC= +4.50%). A third peak was observed in September 2020 (D= +264,138, RC= +4.35%). Large increases in March 2020 were observed for CVD and CNS medications across all age groups. The Immunological and Vaccine products dropped to very low levels across all age groups and all months (including the March dispensing peak). Conclusions Reconfiguration of routine clinical services during COVID-19 led to substantial changes in community pharmacy drug dispensing. This change may contribute to a long-term burden of COVID-19, raising the importance of a comprehensive and timely monitoring of changes for evaluation of the potential impact on clinical care and outcomes.

Introduction COVID-19 risk prediction algorithms can be used to identify at-risk individuals from short-term serious adverse COVID-19 outcomes such as hospitalisation and death. It is important to validate these algorithms in different and diverse populations to help guide risk management decisions and target vaccination and treatment programs to the most vulnerable individuals in society. Objectives To validate externally the QCOVID risk prediction algorithm that predicts mortality outcomes from COVID-19 in the adult population of Wales, UK. Methods We conducted a retrospective cohort study using routinely collected individual-level data held in the Secure Anonymised Information Linkage (SAIL) Databank. The cohort included individuals aged between 19 and 100 years, living in Wales on 24th January 2020, registered with a SAIL-providing general practice, and followed-up to death or study end (28th July 2020). Demographic, primary and secondary healthcare, and dispensing data were used to derive all the predictor variables used to develop the published QCOVID algorithm. Mortality data were used to define time to confirmed or suspected COVID-19 death. Performance metrics, including R2 values (explained variation), Brier scores, and measures of discrimination and calibration were calculated for two periods (24th January–30th April 2020 and 1st May–28th July 2020) to assess algorithm performance. Results 1,956,760 individuals were included. 1,192 (0.06%) and 610 (0.03%) COVID-19 deaths occurred in the first and second time periods, respectively. The algorithms fitted the Welsh data and population well, explaining 68.8% (95% CI: 66.9-70.4) of the variation in time to death, Harrell’s C statistic: 0.929 (95% CI: 0.921-0.937) and D statistic: 3.036 (95% CI: 2.913-3.159) for males in the first period. Similar results were found for females and in the second time period for both sexes. Conclusions The QCOVID algorithm developed in England can be used for public health risk management for the adult Welsh population.

Vaccination programs against COVID-19 vary globally with estimates of vaccine effectiveness (VE) affected by vaccine type, schedule, strain, outcome, and recipient characteristics. This study assessed VE of BNT162b2 and ChAdOx1 vaccines against PCR positive SARS-CoV-2 infection, hospital admission, and death among adults aged 50 years and older in Wales, UK during the period 7 December 2020 to 18 July 2021, when Alpha, followed by Delta, were the predominant variants. We used individual-level linked routinely collected data within the Secure Anonymized Information Linkage (SAIL) Databank. Data were available for 1,262,689 adults aged 50 years and over; coverage of one dose of any COVID-19 vaccine in this population was 92.6%, with coverage of two doses 90.4%. VE against PCR positive infection at 28-days or more post first dose of any COVID-19 vaccine was 16.0% (95%CI 9.6–22.0), and 42.0% (95%CI 36.5–47.1) seven or more days after a second dose. VE against hospital admission was higher at 72.9% (95%CI 63.6–79.8) 28 days or more post vaccination with one dose of any vaccine type, and 84.9% (95%CI 78.2–89.5) at 7 or more days post two doses. VE for one dose against death was estimated to be 80.9% (95%CI 72.1–86.9). VE against PCR positive infection and hospital admission was higher for BNT162b2 compared to ChAdOx1. In conclusion, vaccine uptake has been high among adults in Wales and VE estimates are encouraging, with two doses providing considerable protection against severe outcomes. Continued roll-out of the vaccination programme within Wales, and globally, is crucial in our fight against COVID-19.

Background Better understanding of the role that children and school staff play in the transmission of SARS-CoV-2 is essential to guide policy development on controlling infection while minimising disruption to children’s education and well-being.Methods Our national e-cohort (n=464531) study used anonymised linked data for pupils, staff and associated households linked via educational settings in Wales. We estimated the odds of testing positive for SARS-CoV-2 infection for staff and pupils over the period August– December 2020, dependent on measures of recent exposure to known cases linked to their educational settings.Results The total number of cases in a school was not associated with a subsequent increase in the odds of testing positive (staff OR per case: 0.92, 95% CI 0.85 to 1.00; pupil OR per case: 0.98, 95% CI 0.93 to 1.02). Among pupils, the number of recent cases within the same year group was significantly associated with subsequent increased odds of testing positive (OR per case: 1.12, 95% CI 1.08 to 1.15). These effects were adjusted for a range of demographic covariates, and in particular any known cases within the same household, which had the strongest association with testing positive (staff OR: 39.86, 95% CI 35.01 to 45.38; pupil OR: 9.39, 95% CI 8.94 to 9.88).Conclusions In a national school cohort, the odds of staff testing positive for SARS-CoV-2 infection were not significantly increased in the 14-day period after case detection in the school. However, pupils were found to be at increased odds, following cases appearing within their own year group, where most of their contacts occur. Strong mitigation measures over the whole of the study period may have reduced wider spread within the school environment.

Introduction Multimorbidity is widely recognised as the presence of two or more concurrent long-term conditions, yet remains a poorly understood global issue despite increasing in prevalence.We have created the Wales Multimorbidity e-Cohort (WMC) to provide an accessible research ready data asset to further the understanding of multimorbidity. Our objectives are to create a platform to support research which would help to understand prevalence, trajectories and determinants in multimorbidity, characterise clusters that lead to highest burden on individuals and healthcare services, and evaluate and provide new multimorbidity phenotypes and algorithms to the National Health Service and research communities to support prevention, healthcare planning and the management of individuals with multimorbidity.Methods and analysis The WMC has been created and derived from multisourced demographic, administrative and electronic health record data relating to the Welsh population in the Secure Anonymised Information Linkage (SAIL) Databank. The WMC consists of 2.9 million people alive and living in Wales on the 1 January 2000 with follow-up until 31 December 2019, Welsh residency break or death. Published comorbidity indices and phenotype code lists will be used to measure and conceptualise multimorbidity.Study outcomes will include: (1) a description of multimorbidity using published data phenotype algorithms/ontologies, (2) investigation of the associations between baseline demographic factors and multimorbidity, (3) identification of temporal trajectories of clusters of conditions and multimorbidity and (4) investigation of multimorbidity clusters with poor outcomes such as mortality and high healthcare service utilisation.Ethics and dissemination The SAIL Databank independent Information Governance Review Panel has approved this study (SAIL Project: 0911). Study findings will be presented to policy groups, public meetings, national and international conferences, and published in peer-reviewed journals.

Introduction Multi-morbidity is a widely recognised but poorly understood global issue that appears to be increasing in prevalence, according to the UK’s Academy of Medical Sciences (AMS) report in 2018. Disease clustering, their determinants and consequences are poorly researched. Better understanding would help drive prevention and improved clinical care, services and patient outcomes. Objectives and Approach Development of two comprehensive population-wide e-cohorts, derived utilising data linkage techniques and including multi-sourced anonymised routine health and demographic data held within the SAIL Databank. The objective is to characterise multi-morbidity and its clustering, determinants and outcomes and compare methods using a) prospective cohort design using multiple data sources in Wales and b) retrospective cohort design to examine household level and environment clustering using GP data in demographically diverse populations (Wales and North East London). The prospective e-cohort focuses on adults living in Wales on 1 st January 2000 and followed up to 2020, including data from the NHS population register, deaths, inpatients, outpatients, Emergency Department, GP, disease registries, laboratory data, and population surveys with QoL measures. This e-cohort will be harmonised with other sites across the UK. The retrospective e-cohort is designed to harmonise with a North East London e-cohort, including all individuals living in Wales on 24 th April 2018 and registered with a GP. Results 2.8 and 2.2 million individuals have been included in the prospective and retrospective cohorts respectively, with 43.6 million person years of follow up. Established comorbidity indices and published phenotypes from libraries are being applied to the data to create initial prevalence and incidence estimates for further analysis. Important clusters will be determined by associations with mortality and excess healthcare utilisation. Conclusion/Implications Building the e-cohorts has involved multiple disciplines across organisations. Multi-morbidity prevalence estimates and study designs will be compared prior to statistical analyses and machine learning methods to evaluate clustering and determinants.

Background with rationale New insights into the wider demographic context of multimorbidity has been prioritised, notably among disadvantaged and ethnically-diverse populations with a high disease burden. We propose an innovative approach to quantify health burden and disease clustering at household level, to enable predictors of household multimorbidity to be investigated and understood. Main Aim To quantify multi-morbidity at the household level using general practitioner (GP) electronic health records (EHRs) in a geographically-defined ethnically-diverse inner city population. Methods We extracted clinical and patient address data from GP EHRs from four east London boroughs (Tower Hamlets, Newham, Waltham Forest and City & Hackney). We included currently registered patients aged ≥18 years as at July 2018, and excluded those with duplicate or complex registrations, new registrations in the previous 12 months, or registrations without historical clinical data or occurring prior to 1948, as well as inactive patients with no recorded EHR activity in varying years depending on age and gender. We defined multimorbidity using 16 long-term condition Read codesets defined in the Quality and Outcomes Framework. We grouped patients into households, defined as those sharing the same home address on their GP registration, represented by a pseudonymised Unique Property Reference Number. Results Provisional data are presented. We identified 737,920 patients (51% female) eligible for this study out of a total population of 1,171,483 currently registered patients. Of these, 23% aged =65, 38% White ethnicity, 3% Mixed, 30% Asian, 14% Black, 5% Other and 12% Not Stated/Null. We identified 312,582 shared households among 737,920 patients. Analyses to derive household-level summary characteristics and relate these to multimorbidity burden are in progress and will be presented. Conclusion Household-level multi-morbidity can be quantified using clinical and patient address data in GP electronic health records.

Background with rationale Injury is a leading cause of premature mortality, morbidity and disability worldwide. In Wales, injuries result in approximately 3 deaths, 107 hospital admissions and 868 emergency department (ED) attendances every day. Health indicators are quantifiable measures designed to summarise a population’s health, and ultimately inform policy and practice. In 2017, several experts identified a need for injury indicators in Wales, to support injury surveillance and prevention efforts. Main Aim To develop a suite of injury indicators, accessible to the public via an online tool, to inform policy and practice across Wales. Methods/Approach Injury specialists in the All Wales Injury Surveillance System (AWISS) proposed a list of measurable injury indicators, based on: data availability, major areas of interest from a Welsh Government strategy perspective, injuries contributing the greatest burden, and injuries for which effective interventions existed. A consultation process with experts and stakeholders resulted in 25 agreed indicators covering all-cause injuries, falls in older adults, hip fractures, road traffic injuries, injuries in the home and at leisure, burns and scalds, poisonings and intentional injuries. Anonymised, routinely collected ED data, inpatient and mortality data in the Secure Anonymised Information Linkage (SAIL) databank at Swansea University were used to generate indicator estimates. Estimates will be updated annually and are accessible via an online interactive tool. Results The development of a free, online injury indicator tool provides practitioners and policy makers across Wales, with the information required to make informed decisions. However, data quality issues hamper the extent to which conclusions can be made. Conclusions Injury indicators have the potential to inform policy and practice across Wales. The adoption of a simplified, standardised data collection system in EDs across Wales is recommended to improve data validity and reliability.

Introduction Multi-morbidity and polypharmacy are increasing but are under investigated. Data linkage has much to offer in understanding trends in prevalence, inter-relationships between variables and impact on healthcare activity. We created Welsh population e-cohorts in 2000 and 2014 to study these issues, using the Secure Anonymised Information Linkage (SAIL) Databank. Objectives and Approach The aim of this study was to measure changing prevalences of multimorbidity, initially through disease chapter prescribing and then to explore the relationship between the number of morbidities recorded in primary care and use of different hospital based outpatient services. Data linkage was used to create cohorts of Welsh residents registered to SAIL providing General Practices (GPs) for at least 360 days in 2000 and 2014. The 13 Read code drug chapters were used to calculate morbidity scores between 0 and 13. Proportional odds or cumulative logit models were used to relate GP recorded morbidities to outpatient attendance patterns. Results The GP cohorts included 1.6 million and 2.1 million population with 56.6% and 73.4% having at least one recorded morbidity for 2000 and 2014 respectively. In 2014, 5+ morbidities were most prevalent (61.3%) in 85+ year olds and least common (2.7%) in 5-9 year olds. Some 35% of individuals attended at least one outpatient specialty in 2014, varying from 22.4% for 5-9 year olds and 63.2% for 80-84 year olds. Preliminary modelling results show the number of GP recorded morbidity chapters was strongly related to increasing outpatient attendances at different specialties, e.g. OR 15.3 (95%CI: 15.1-15.4) of being in a higher outpatient attendance category for the 5+ morbidity group relative to the zero morbidity group. Increasing age and female gender were associated with increased numbers of specialists attended whilst deprivation had a more modest impact. Conclusion/Implications There has been a large increase in recorded multimorbidity across all age groups in Wales. In this exploratory cross-sectional design, multimorbidity was strongly related to increasing use of outpatient services. Further work is ongoing to define and utilise more refined multimorbidity metrics and incorporate longitudinal designs in analysis.

Introduction There is a lack of evidence of the adverse effects of air pollution and pollen on cognition for people with air quality related health conditions. This study explored the effects of air quality and respiratory health conditions on educational attainment for 18,241 pupils across the city of Cardiff, United Kingdom. Objectives and Approach Anonymised, routinely collected health and education data were linked at the household and school level with modelled high spatial resolution pollution data, and daily pollen measurements using the Secure Anonymised Information Linkage (SAIL) databank. This created 7 repeated cross-sectional cohorts (2009-2015). Multilevel linear regression analysis examined whether exam performance was associated with health status and/or air quality levels averaged at school and home locations during revision and examination periods. We also investigated the combined effects of air quality and associations with educational attainment for pupils who were treated for asthma and/or Severe Allergic Rhinitis (SAR), and those who were not. Results The cohort contained 9337 males and 8904 female pupils. There were 871 treated for asthma, 2091 for SAR, and 634 treated for both. Asthma was not associated with exam performance (p=0.700). However, SAR was positively associated with exam performance (p 2) was negatively associated with educational attainment (p = 0.002). Other indicators of air quality (pollutants: Ozone, Particulate Matter - PM2.5, and pollen) were not associated with educational attainment (p> 0.05). Exposure to NO2 was negatively associated with educational attainment irrespective of treatment for asthma or SAR. There was no combined effect of air quality on the variation in educational attainment between those who are treated for asthma and/or SAR and those who were not. Conclusion/Implications Irrespective of health status, exposure to NO2 was negatively associated with educational attainment. Treatment seeking behaviour may be a possible explanation for the positive association between SAR and educational attainment. For a more accurate reflection of health status, health outcomes not subject to treatment seeking behaviour should be investigated.

Introduction The Secure Anonymised Information Linkage (SAIL) databank facilitated linkage of routinely collected health and education data, high spatial resolution pollution modelling and daily pollen measurements for 18,241 pupils in 7 cross-sectional cohorts across Cardiff city, UK, to investigate effects of air quality and respiratory health conditions on education attainment. Objectives and Approach An urban atmospheric dispersion and chemistry modelling system (ADMS-Urban) simulated modelled hourly concentrations of air pollutants: PM2.5, PM10, NO2 and ozone levels. These were summarised into minimum, average and maximum daily readings for 4 time periods (e.g. school hours 9am-3pm) for all home and school locations across Cardiff between 2009 and 2015. The combination of different pollutants, measurements and time-periods created a comprehensive multi-row dataset per location. We transformed the dimensionality of this high-resolution data to create one row of summarised data per pupil per cohort, in preparation for statistical analysis. Results 157,361 school and home locations across Cardiff were anonymised and household linkage fields were appended to combine pollution estimates at the household/school to individual health data. The pollution dataset contained 369 columns, 472,083 rows per year with one column per location, pollutant type, pollutant measurement, daily time-period, and day of year. Dataset transformation reduced algorithm computation by creating a single date column, producing a five column, 3,446,205,900-row matrix per year dataset. The algorithm adjusted for weekends, school/bank holidays and allowed location to vary 3pm-5pm on school days when pupil location was uncertain. The algorithm calculated tailored pollution exposures per pupil for revision and examination periods, creating one row per pupil and reducing 7 years of data and 24 billion rows to 18,241. Conclusion/Implications We successfully linked 95% of the cohorts’ household/school pollution data to their corresponding health and education data. This demonstrates data linking retrospective exposures for total populations using multiple daily locations, and extends our analysis platform for natural experiments to include daily exposure. Future work includes adding modelled route exposures.

Introduction There is a lack of evidence on the adverse effects of air pollution on cognition for people with air quality-related health conditions. We propose that educational attainment, as a proxy for cognition, may increase with improved air quality. This study will explore whether asthma and seasonal allergic rhinitis, when exacerbated by acute exposure to air pollution, is associated with educational attainment. Objectives To describe the preparation of individual and household-level linked environmental and health data for analysis within an anonymised safe haven. Also to introduce our statistical analysis plan for our study: COgnition, Respiratory Tract illness and Effects of eXposure (CORTEX). Methods We imported daily air pollution and aeroallergen data, and individual level education data into the SAIL databank, an anonymised safe haven for person-based records. We linked individual-level education, socioeconomic and health data to air quality data for home and school locations, creating tailored exposures for individuals across a city. We developed daily exposure data for all pupils in repeated cross sectional exam cohorts (2009-2015). Conclusion We have used the SAIL databank, an innovative, data safe haven to create individual-level exposures to air pollution and pollen for multiple daily home and school locations. The analysis platform will allow us to evaluate retrospectively the impact of air quality on attainment for multiple cross-sectional cohorts of pupils. Our methods will allow us to distinguish between the pollution impacts on educational attainment for pupils with and without respiratory health conditions. The results from this study will further our understanding of the effects of air quality and respiratory-related health conditions on cognition. Highlights • This city-wide study includes longitudinal routinely-recorded educational attainment data for all pupils taking exams over seven years; • High spatial resolution air pollution data were linked within a privacy protected databank to obtain individual exposure at multiple daily locations; • This study will use health data linked at the individual level to explore associations between air pollution, related morbidity, and educational attainment.

Introduction Monitoring social wellbeing and its relationship to health service utilisation by means of appropriate measurement tools can potentially provide a complementary view for influencing service development. Aspects of wellbeing have been collected in the Welsh Health Survey (WHS) while routine health data captures health service utilisation. Objectives and Approach WHS was used to link self-reported wellbeing to health outcomes, prior to linking to routinely collected data. Initially, a measure for personal wellbeing was developed using the four personal wellbeing questions defined by The Office of National Statistics (ONS), included in national surveys from 2011 onward. We conducted regression analysis to identify potential predictors of personal wellbeing scores our model included self-reported lifestyle behaviour, self-reported health, education, work, household and general demographics. Links to primary care, hospital and emergency department datasets are being developed to provide insight into the relationship between wellbeing, multi-morbidity and health service utilisation. Results Four wellbeing questions had similar scoring patterns across age groups which is different to most health indicators that tend to show a marked health decline with increasing age. There is a difference between mean wellbeing score for males and females. Our finding showed that self-reported of ‘excellent’ or ‘very good’ general health has the largest positive effect on wellbeing while positive viewpoint on self-health has the second largest effect on our model. In addition, being retired from a paid job, eating at least one or 5+ portion of fruit and vegetables and giving up smoking have positive impact on population wellbeing. In contrast, not being able to work, intermediate household occupancy, consuming alcohol in last 12-months, having long-standing illness, showed a negative impact on wellbeing. Conclusion/Implications This project established robust methodology on utilizing survey and routine health data for monitoring and evaluation purposes. We also evaluated the linkability of survey data The latest release of National Survey for Wales (NSW) will cover a combination of self-reported health measures and aims for a higher linkage consent rate.

Background and Objectives There is a lack of evidence of the adverse effects of air pollution and pollen on cognition for people with air quality-related health conditions. The CORTEX project combined routinely collected health and education data, high spatial resolution air pollution modelling, and daily pollen measurements for 18,241 pupils living in Cardiff, UK, between 2009 and 2015, to investigate the acute effects of air quality and respiratory conditions on education attainment. Datasets Air pollutants PM2.5, PM10, NO2, and ozone levels were modelled for 157,361 home and school locations, anonymised into the Secure Anonymised Information Linkage (SAIL) Databank, and summarised into minimum, average and maximum readings for 4 daily time periods reflecting pupil home/school exposure. Adding a unique Residential Anonymised Linking Field (RALF) allowed linkage of pollution estimates to individual level data. Annual pollution datasets contained 369 columns and 472,083-rows, with one column per location, pollutant, daily time-period and day of year. Dataset transformation produced a 5 column, 3,446,205,900-row matrix per year. Methods and Conclusions An algorithm using Structured Query Language (SQL) to manage data held within a relational database management system, was designed to reduce dimensionality from 24 billion to 18,241 rows of data. The algorithm calculated average means for each pollutant (PM2.5, PM10, NO2, and ozone levels) over the revision and examination periods, and summarised data into one row per pupil. The algorithm adjusted for weekends, school, and bank holidays, it calculated daily pollutant exposure for each pupil, and successfully linked 95% of pupil pollution exposures to their health and education data.

Objectives Multi-morbidity and polypharmacy are increasing and interrelated phenomena but are poorly understood. The aim of this study is to contribute to the understanding of these issues, measure the changing prevalence’s of multimorbidity/ polypharmacy and explore the relationship between multimorbidity as recorded in primary care and the use of outpatient services. Methods The Secure Anonymised Information Linkage (SAIL) Databank facilitated linkage techniques to create population based e-cohorts of de-identified Welsh residents. Individuals were registered to a SAIL providing General Practice (GP) for at least 360 days in 2000 and 2014. Categories of morbidity were created using the 13 Read drug code chapters. In an initial cross sectional exploratory analysis proportional odds and cumulative logit models were used to relate GP recorded morbidities to outpatient attendance patterns in the same year. Findings The GP e-cohorts included 1.6 million (2000) and 2.1 million (2001) people, with 56.6% and 73.4% having ≥1 recorded morbidity for 2000 and 2014, respectively. In 2014, groups with 5+ morbidities were most prevalent (61.3%) in 85+ year olds and least (2.7%) in 5-9 year olds. Some 35% of individuals attended ≥1 outpatient specialty in 2014; 22.4% in 5-9 year olds and 63.2% for 80-84 year olds. Results from preliminary models showed the number of GP recorded morbidities was strongly related to increasing outpatient attendances at different specialties, OR=15.3 (95%CI:15.1-15.4) of being in a higher outpatient attendance category for the 5+ morbidity group relative to the zero morbidity group. Conclusion Preliminary analysis has shown large increases in GP recorded multimorbidity across Wales over fifteen years and strong relationships and NHS service utilisation in cross-sectional analyses. Further work will include creating more refined definitions for multimorbidity metrics, linkage to hospital admission data, comparisons across healthcare settings and the development of longitudinal models.

Background There is a lack of evidence of the adverse effects which air quality has on cognition for people with air quality-related health conditions, these are not widely documented in the literature. Educational attainment, as a proxy for cognition, may increase with improved air quality. Objectives Prepare individual and household level linked environmental and health data for analysis within an anonymised safe haven; analyse the linked dataset for our study investigating: Cognition, Respiratory Tract illness and Effects of eXposure (CORTEX). Methods Anonymised, routinely collected health and education data were linked with high spatial resolution pollution measurements and daily pollen measurements to provide repeated cross-sectional cohorts (2009-2015) on 18,241 pupils across the city of Cardiff, using the SAIL databank. A fully adjusted multilevel linear regression analysis examined associations between health status and/or air quality. Cohort, school and individual level confounders were controlled for. We hope that using individual-level multi-location daily exposure assessment will help to clarify the role of traffic and prevent potential community-level confounding. Combined effects of air quality on variation in educational attainment between those treated for asthma and/or Severe Allergic Rhinitis (SAR), and those not treated, was also investigated. Findings Asthma was not associated with exam performance (p=0.7). However, SAR was positively associated with exam performance (p

ABSTRACT Objectives Electronic healthcare records (EHR) are the main data sources that facilitate epidemiology research. Routinely collected data such as primary and secondary care are now easily linked to produce novel and high impact research. There are, however, rich data locked in the free text of clinical letters that are not otherwise translated into EHRs. It is highly desirable to be able to extract this information to strengthen the body of information in existing EHRs. The Swansea Collaborative in Analysis of NLP Research (SCANR) group at Swansea University has been established to evaluate the usage of Natural Language Processing platforms for obtaining new clinical data. To use Clix Enrich to extract SNOMED concepts from a variety of clinical free texts and produce EHRs from the extraction process. Approach SNOMED concepts contain common items of interest such as diagnosis, medication and symptoms, as well as contextual concepts such as historical reference and negation. Clix Enrich uses the SNOMED dictionary to encode clinical free text (pre-co-ordinated) and find contextually correct SNOMED concepts (post co-ordinated). We used Clix Enrich to extract meaningful clinical terms from MS and Epilepsy consultant letters, as well as presenting complaint fields from a Welsh Emergency Department (ED). Results We tailored Clix Enrich to extract a wide variety of clinical terms from each source (fourty texts per source) and validated the extraction accuracy with clinical experts in each domain. Clix Enrich was able to accurately extract the correct diagnosis for MS, Epilepsy and ED attendance (100%, 95% and 80%), dosage and frequency of anti-epileptic medication and MS modifying therapy (90%, 100%) and EDDS score (94%). We note a probable source of discrepancy in extraction accuracy between letter sources in the frequency of abbreviated terms, particularly within the presenting complaint field of the ED sample. Conclusion Clix Enrich can be used to accurately extract SNOMED concepts from clinical letters. The resulting datasets are readily available to link to existing EHRs, and can be linked to EHRs that adopt the SNOMED coding structure, or backward compatible hierarchies. Clix Enrich comes with out-of-the-box extraction methods but the optimum way to extract the correct information would be to build in custom queries, thus requiring clinical expertise to validate extraction.

ABSTRACT Objectives EMRTS launched in April 2015 which saw consultants join critical care practitioners on air ambulance missions and emergency response vehicles. EMRTS delivers pre hospital, Emergency Department equivalent critical care for people in Wales with life or limb threatening injuries or illness from scene to definitive care. The evaluation will compare patients treated by EMRTS to cases both pre-EMRTs and EMRTS-offline (at night or bad weather) in improving survival and functional and quality of life outcomes. Approach The evaluation will require the use of multiple datasets to follow patients from time of incident, throughout hospital care and to calculate post care health status. Electronic reporting forms will be used to measure incident details, patient vitals and all treatments/procedures pre hospital, creating the EMRTS dataset and cohort. Patients’ data will be linked to the Emergency Department Dataset (EDDS), Patient Episode Database for Wales (PEDW), Critical Care dataset, Trauma Audit and Research Network (TARN) and Intensive Care National Audit and Research Centre (ICNARC). TARN and ICNARC will provide data on patients transferred to English specialist care facilities. ONS (Office of National Statistics) and WDS (Welsh Demographic Service) data will be used for demographic and survival analysis. GIS techniques will be used on WAST (Welsh Ambulance Service Trust) data for time saved analysis in transferring patients to definitive care. Following discharge from hospital patients will be interviewed six months and one-year post incident for longer-term functional and quality of life outcomes. Results Datasets are currently being assembled for the service evaluation. Data will be encrypted and anonymized through NWIS (NHS Wales Informatics Service) to enable research through the Secure Anonymised Information Linkage (SAIL) facility. Results will include rate of survival, effect on length of hospital stay, time to definitive care, the use and appropriateness of EMRTs specific interventions and functional and quality of life outcomes. Conclusions To allow for data collection the service evaluation is still in the early stages and will be compared with local and international data.