Biases arising from linked administrative data for epidemiological research: a conceptual framework from registration to analyses

Occupational differences in COVID-19 hospital admission and mortality risks between women and men in Scotland: a population-based study using linked administrative data

OBJECTIVES

Occupations vary with respect to workplace factors that influence exposure to COVID-19, such as ventilation, social contacts and protective equipment. Variations between women and men may arise because they have different occupational roles or behavioural responses. We estimated occupational differences in COVID-19 hospital admission and mortality risks by sex.

METHODS

We combined (1) individual-level data from 2011 Census with (2) health records and (3) household-level information from residential identifiers, using a Scottish cohort of 1.7 million adults aged 40-64 years between 1 March 2020 and 31 January 2021. We estimated age-standardised COVID-19 hospital admission and mortality rates, stratified by sex and occupation. Cox proportional hazards models were adjusted for pre-pandemic health and occupational exposure factors, including interaction effects between occupation and sex.

RESULTS

Women had lower age-standardised COVID-19 hospital admission and mortality rates than men. Among women, adjusted death risks were lowest for health professionals, and those in associate professional and technical occupations (paramedics and medical technicians), with the latter supported by results from the interaction model. Among men, elevated adjusted admission and death risks were observed for large vehicle and taxi drivers. Additionally, admission risks remained high among men in caring personal services (including home and care workers), while elevated risks were observed among women in customer service occupations (call centre operators) and process, plant and machine operative roles (assemblers and sorters).

CONCLUSIONS

Occupational differences in COVID-19 hospital admission and mortality risks between women and men highlight the need to account for sex differences when developing interventions to reduce infections among vulnerable occupational groups.

Validity of administrative health data case definitions for identifying polycystic ovary syndrome: a systematic review and meta-analysis

STUDY QUESTION

What is the validity of published administrative health data case definitions of polycystic ovary syndrome (PCOS) compared with reference standards?

SUMMARY ANSWER

Due to the limited number of eligible studies, drawing definitive conclusions is challenging; however, this review highlights significant gaps and variability in current PCOS case definitions, underscoring the need for standardized case definitions in future research.

WHAT IS KNOWN ALREADY

Administrative health data offer the opportunity to evaluate health outcomes and disease epidemiology at a population-level. Currently, the validity of existing administrative health data case definitions for PCOS is unknown.

STUDY DESIGN, SIZE, DURATION

A systematic review of the literature was conducted on full-text English-language articles up to July 2023, using the MEDLINE and EMBASE databases.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Two reviewers independently screened titles, abstracts and full texts, extracted data, assessed study quality and graded validity. A random effects meta-analysis was conducted to pool reported validity measures and heterogeneity was examined.

MAIN RESULTS AND THE ROLE OF CHANCE

The review included four eligible articles consisting of three cross-sectional studies and one retrospective cohort study. Two studies defined PCOS using the Rotterdam Criteria, one study used self-report, and one used a clinical gold standard. All case definitions included the International Classification of Diseases (ICD)-9 code 256.4 for 'polycystic ovaries' and three studies used E28.2 for 'polycystic ovarian syndrome'. Three studies reported positive predictive value (PPV), which ranged from 30 to 96%. One study reported both PPV (96%) and sensitivity (50%) for one case definition. The pooled PPV estimate for the ICD code-based case definitions was 88% (95% confidence interval 82-95%; I2 = 100%). One study reported fair agreement (percent agreement= 90.3, κ = 0.27, percent agreement bias adjusted κ = 0.81). Overall, the risk of bias of the included studies was low.

LIMITATIONS, REASONS FOR CAUTION

There were limited number of validations and precision indices of validations.

WIDER IMPLICATIONS OF THE FINDINGS

Further validation of these case definitions in other administrative health datasets, and development of novel coding algorithms is required to inform future population-based studies in PCOS.

STUDY FUNDING/COMPETING INTEREST(S)

No external funding was used and there are no disclosures.

REGISTRATION NUMBER

PROSPERO CRD42023385617.

Advancing birth cohort studies using administrative and other research-independent data repositories: Opportunities and challenges

The birth cohort study design is an essential epidemiologic tool for investigating the developmental origins of health and disease. Birth cohorts have greatly improved the etiologic understanding of asthma and allergic diseases, setting the stage for advancements in translational interventions. Increasingly, investigators leverage data repositories that have been compiled and maintained independently of research investigations (administrative data) to establish large birth cohorts or to augment data generated through active participant interaction. In many cases, administrative data can greatly enhance the capacity of birth cohorts to achieve their scientific goals. However, investigators must be wary of common pitfalls and carefully consider whether administrative data are well suited to the scientific inquiry. This article reviews the strengths and challenges of using administrative data and the pragmatic solutions that have been developed to optimize their use in birth cohorts. As birth cohorts continue to play an important role in understanding the etiology of early-life disease, unleashing the power of administrative data will greatly assist in this scientific process.

Associations between different measures of SARS-CoV-2 infection status and subsequent economic inactivity: A pooled analysis of five longitudinal surveys linked to healthcare records

INTRODUCTION

Following the acute phase of the COVID-19 pandemic, a record number of people became economically inactive in the UK. We investigated the association between coronavirus infection and subsequent economic inactivity among people employed pre-pandemic, and whether this association varied between self-report versus healthcare recorded infection status.

METHODS

We pooled data from five longitudinal studies (1970 British Cohort Study, English Longitudinal Study of Ageing, 1958 National Child Development Study, Next Steps, and Understanding Society), in two databases: the UK Longitudinal Linkage Collaboration (UKLLC), which links study data to NHS England records, and the UK Data Service (UKDS), which does not. The study population were aged 25-65 years between April 2020 to March 2021. The outcome was economic inactivity measured at the time of the last survey (November 2020 to March 2021). The exposures were COVID-19 status, indicated by a positive SARS-CoV-2 test in NHS records (UKLLC sample only), or by self-reported measures of coronavirus infection (both samples). Logistic regression models estimated odds ratios (ORs) adjusting for potential confounders including sociodemographic variables and pre-pandemic health.

RESULTS

Within the UKLLC sample (N = 8,174), both a positive SARS-CoV-2 test in NHS records (5.9% of the sample; OR 1.08, 95%CI 0.68-1.73) and self-reported positive tests (6.5% of the sample; OR 1.07, 95%CI 0.68-1.69), were marginally and non-significantly associated with economic inactivity (5.3% of the sample) in adjusted analyses. Within the larger UKDS sample (n = 13,881) reliant on self-reported ascertainment of infection (6.4% of the sample), the coefficient indicated a null relationship (OR 0.98, 95%CI 0.68-1.40) with economic inactivity (5.0% of sample).

CONCLUSIONS

Among people employed pre-pandemic, testing positive for SARS-CoV-2 was not associated with increased economic inactivity, although we could not exclude small effects. Ascertaining infection through healthcare records or self-report made little difference to results. However, processes related to record linkage may introduce small biases.

Examining the Potential Mediating Role of Maternal Mental Health in the Association Between Socioeconomic Deprivation and Child Development Outcomes

BACKGROUND

Socioeconomic deprivation has been linked to negative child developmental outcomes including brain development, psychological well-being, educational attainment, and social-emotional well-being. Maternal mental health has also been linked to mothers' parenting practices and their children's developmental outcomes. However, limited evidence exists regarding the role of maternal mental health (prenatal and postnatal) in the association between socioeconomic deprivation and children's developmental outcomes.

METHODS

We examined the potential role of maternal mental health in the association between socioeconomic deprivation (SED) and child development outcomes. We used a large linked administrative health dataset covering children born between 2011 and 2015 in Greater Glasgow and Clyde, Scotland. Of the 76,483 participants, 55,856 mothers with matched children's developmental outcome data were included. A mediation analysis model, adjusted for confounders and covariates, was used.

RESULTS

Maternal mental health assessed by a history of hospital admissions mediated, but to a small extent, the relationship between SED and children's developmental outcomes. The average direct effect (ADE), of SED in the first model with a history of hospital admissions, was ADE: ES = - 0.0875 (95% CI = - 0.097, - 0.08; p < 0.001) and ACME: ES = - 0.0002 (95% CI = - 0.001, - 0.0001; p = 0.01). The proportion mediated by the history of mental health admission was 0.3%.

CONCLUSION

The association between SED and children's developmental outcomes appears to be partially mediated by maternal mental health, although the proportional-mediated effect was very small.

Impact of Primary Health Care Data Quality on Infectious Disease Surveillance in Brazil: Case Study

Background

The increase in emerging and re-emerging infectious disease outbreaks underscores the need for robust early warning systems (EWSs) to guide mitigation and response measures. Administrative health care databases provide valuable epidemiological insights without imposing additional burdens on health services. However, these datasets are primarily collected for operational use, making data quality assessment essential to ensure an accurate interpretation of epidemiological analysis. This study focuses on the development and implementation of a data quality index (DQI) for surveillance integrated into an EWS for influenza-like illness (ILI) outbreaks using Brazil's a nationwide Primary Health Care (PHC) dataset.

Objective

We aimed to evaluate the impact of data completeness and timeliness on the performance of an EWS for ILI outbreaks and establish optimal thresholds for a suitable DQI, thereby improving the accuracy of outbreak detection and supporting public health surveillance.

Methods

A composite DQI was established to measure the completeness and timeliness of PHC data from the Brazilian National Information System on Primary Health Care. Completeness was defined as the proportion of weeks within an 8-week rolling window with any register of encounters. Timeliness was calculated as the interval between the date of encounter and its corresponding registry in the information system. The backfilled PHC dataset served as the gold standard to evaluate the impact of varying data quality levels from the weekly updated real-time PHC dataset on the EWS for ILI outbreaks across 5570 Brazilian municipalities from October 10, 2023, to March 10, 2024.

Results

During the study period, the backfilled dataset recorded 198,335,762 ILI-related encounters, averaging 8,623,294 encounters per week. The EWS detected a median of 4 (IQR 2-5) ILI outbreak warnings per municipality using the backfilled dataset. Using the real-time dataset, 12,538 (65%) warnings were concordant with the backfilled dataset. Our analysis revealed that 100% completeness yielded 76.7% concordant warnings, while 80% timeliness resulted in at least 50% concordant warnings. These thresholds were considered optimal for a suitable DQI. Restricting the analysis to municipalities with a suitable DQI increased concordant warnings to 80.4%. A median of 71% (IQR 54%-71.9%) of municipalities met the suitable DQI threshold weekly. Municipalities with ≥60% of weeks achieving a suitable DQI demonstrated the highest concordance between backfilled and real-time datasets, with those achieving ≥80% of weeks showing 82.3% concordance.

Conclusions

Our findings highlight the critical role of data quality in improving the EWS' performance based on PHC data for detecting ILI outbreaks. The proposed framework for real-time DQI monitoring is a practical approach and can be adapted to other surveillance systems, providing insights for similar implementations. We demonstrate that optimal completeness and timeliness of data significantly impact the EWS' ability to detect ILI outbreaks. Continuous monitoring and improvement of data quality should remain a priority to strengthen the reliability and effectiveness of surveillance systems.

Leveraging Administrative Health Databases to Address Health Challenges in Farming Populations: Scoping Review and Bibliometric Analysis (1975-2024)

BACKGROUND

Although agricultural health has gained importance, to date, much of the existing research relies on traditional epidemiological approaches that often face limitations related to sample size, geographic scope, temporal coverage, and the range of health events examined. To address these challenges, a complementary approach involves leveraging and reusing data beyond its original purpose. Administrative health databases (AHDs) are increasingly reused in population-based research and digital public health, especially for populations such as farmers, who face distinct environmental risks.

OBJECTIVE

We aimed to explore the reuse of AHDs in addressing health issues within farming populations by summarizing the current landscape of AHD-based research and identifying key areas of interest, research gaps, and unmet needs.

METHODS

We conducted a scoping review and bibliometric analysis using PubMed and Web of Science. Building upon previous reviews of AHD-based public health research, we conducted a comprehensive literature search using 72 terms related to the farming population and AHDs. To identify research hot spots, directions, and gaps, we used keyword frequency, co-occurrence, and thematic mapping. We also explored the bibliometric profile of the farming exposome by mapping keyword co-occurrences between environmental factors and health outcomes.

RESULTS

Between 1975 and April 2024, 296 publications across 118 journals, predominantly from high-income countries, were identified. Nearly one-third of these publications were associated with well-established cohorts, such as Agriculture and Cancer and Agricultural Health Study. The most frequently used AHDs included disease registers (158/296, 53.4%), electronic health records (124/296, 41.9%), insurance claims (106/296, 35.8%), population registers (95/296, 32.1%), and hospital discharge databases (41/296, 13.9%). Fifty (16.9%) of 296 studies involved >1 million participants. Although a broad range of exposure proxies were used, most studies (254/296, 85.8%) relied on broad proxies, which failed to capture the specifics of farming tasks. Research on the farming exposome remains underexplored, with a predominant focus on the specific external exposome, particularly pesticide exposure. A limited range of health events have been examined, primarily cancer, mortality, and injuries.

CONCLUSIONS

The increasing use of AHDs holds major potential to advance public health research within farming populations. However, substantial research gaps persist, particularly in low-income regions and among underrepresented farming subgroups, such as women, children, and contingent workers. Emerging issues, including exposure to per- and polyfluoroalkyl substances, biological agents, microbiome, microplastics, and climate change, warrant further research. Major gaps also persist in understanding various health conditions, including cardiovascular, reproductive, ocular, sleep-related, age-related, and autoimmune diseases. Addressing these overlooked areas is essential for comprehending the health risks faced by farming communities and guiding public health policies. Within this context, promoting AHD-based research, in conjunction with other digital data sources (eg, mobile health, social health data, and wearables) and artificial intelligence approaches, represents a promising avenue for future exploration.

Child welfare worker perspectives on documentation and case recording practices in Canada: A mixed-methods study protocol

In health care and child welfare, clinical records and case notes serve multiple functions. When records are aggregated and processed to create administrative data, they can be analyzed and used to inform policy development and decision-making. To be useful, such data should be complete, accurate, and recorded in a standardized way. However, sources of bias and error can impact the quality of administrative data. During the development of national child welfare data in Canada, child welfare sector partners expressed concerns about the accuracy and completeness of data about children and families. This protocol describes a study that seeks to answer two questions: 1) What individual and institutional factors influence how client data is recorded by child welfare workers in Canada? 2) What data quality issues are created through documentation and case recording practices that may impact the use of clinical case management system data for public health statistics? In this protocol, we describe an exploratory mixed methods study that involves an online survey, interviews with a purposive sample of child welfare workers, and a document review of case recording guidelines. To be eligible for the study, participants must have worked at a child welfare agency or department with clinical documentation responsibilities as a part of their job. We will use descriptive statistics to analyze the survey data and thematic analysis to analyze the qualitative data. This study will help uncover strengths, limitations, and possible sources of bias created through case recording and documentation practices in child welfare. Study results will be shared through presentations to interest holders and will inform the further development of national child welfare data in Canada.

Epidemiological methods in transition: Minimizing biases in classical and digital approaches

Epidemiology and Public Health have increasingly relied on structured and unstructured data, collected inside and outside of typical health systems, to study, identify, and mitigate diseases at the population level. Focusing on infectious diseases, we review the state of Digital Epidemiology at the beginning of 2020 and how it changed after the COVID-19 pandemic, in both nature and breadth. We argue that Epidemiology's progressive use of data generated outside of clinical and public health systems creates several technical challenges, particularly in carrying specific biases that are almost impossible to correct for a priori. Using a statistical perspective, we discuss how a definition of Digital Epidemiology that emphasizes "data-type" instead of "data-source," may be more operationally useful, by clarifying key methodological differences and gaps. Therefore, we briefly describe some of the possible biases arising from varied collection methods and sources, and offer some recommendations to better explore the potential of Digital Epidemiology, particularly on how to help reduce inequity.

Tackling algorithmic bias and promoting transparency in health datasets: the STANDING Together consensus recommendations

Without careful dissection of the ways in which biases can be encoded into artificial intelligence (AI) health technologies, there is a risk of perpetuating existing health inequalities at scale. One major source of bias is the data that underpins such technologies. The STANDING Together recommendations aim to encourage transparency regarding limitations of health datasets and proactive evaluation of their effect across population groups. Draft recommendation items were informed by a systematic review and stakeholder survey. The recommendations were developed using a Delphi approach, supplemented by a public consultation and international interview study. Overall, more than 350 representatives from 58 countries provided input into this initiative. 194 Delphi participants from 25 countries voted and provided comments on 32 candidate items across three electronic survey rounds and one in-person consensus meeting. The 29 STANDING Together consensus recommendations are presented here in two parts. Recommendations for Documentation of Health Datasets provide guidance for dataset curators to enable transparency around data composition and limitations. Recommendations for Use of Health Datasets aim to enable identification and mitigation of algorithmic biases that might exacerbate health inequalities. These recommendations are intended to prompt proactive inquiry rather than acting as a checklist. We hope to raise awareness that no dataset is free of limitations, so transparent communication of data limitations should be perceived as valuable, and absence of this information as a limitation. We hope that adoption of the STANDING Together recommendations by stakeholders across the AI health technology lifecycle will enable everyone in society to benefit from technologies which are safe and effective.

Prenatal maternal infections and early childhood developmental outcomes: analysis of linked administrative health data for Greater Glasgow & Clyde, Scotland

BACKGROUND

Previous research has linked prenatal maternal infections to later childhood developmental outcomes and socioemotional difficulties. However, existing studies have relied on retrospectively self-reported survey data, or data on hospital-recorded infections only, resulting in gaps in data collection.

METHODS

This study used a large linked administrative health dataset, bringing together data from birth records, hospital records, prescriptions and routine child health reviews for 55,856 children born in Greater Glasgow & Clyde, Scotland, 2011-2015, and their mothers. Logistic regression models examined associations between prenatal infections, measured as both hospital-diagnosed prenatal infections and receipt of infection-related prescription(s) during pregnancy, and childhood developmental concern(s) identified by health visitors during 6-8 week or 27-30 month health reviews. Secondary analyses examined whether results varied by (a) specific developmental outcome types (gross-motor-skills, hearing-communication, vision-social-awareness, personal-social, emotional-behavioural-attention and speech-language-communication) and (b) the trimester(s) in which infections occurred.

RESULTS

After confounder/covariate adjustment, hospital-diagnosed infections were associated with increased odds of having at least one developmental concern (OR: 1.30; 95% CI: 1.19-1.42). This was broadly consistent across all developmental outcome types and appeared to be specifically linked to infections occurring in pregnancy trimesters 2 (OR: 1.34; 95% CI: 1.07-1.67) and 3 (OR: 1.33; 95% CI: 1.21-1.47), that is the trimesters in which foetal brain myelination occurs. Infection-related prescriptions were not associated with any clear increase in odds of having at least one developmental concern after confounder/covariate adjustment (OR: 1.03; 95% CI: 0.98-1.08), but were associated with slightly increased odds of concerns specifically related to personal-social (OR: 1.12; 95% CI: 1.03-1.22) and emotional-behavioural-attention (OR: 1.15; 95% CI: 1.08-1.22) development.

CONCLUSIONS

Prenatal infections, particularly those which are hospital-diagnosed (and likely more severe), are associated with early childhood developmental outcomes. Prevention of prenatal infections, and monitoring of support needs of affected children, may improve childhood development, but causality remains to be established.

COVID-19 risk by work-related factors: pooled analysis of individual linked data from 14 cohorts

BACKGROUND

SARS-CoV-2 infection rates vary by occupation, but the association with work-related characteristics (such as home working, keyworker or furlough) are not fully understood and may depend on ascertainment approach. We assessed infection risks across work-related characteristics and compared findings using different ascertainment approaches.

METHODS

Participants of 14 UK-based longitudinal cohort studies completed surveys before and during the COVID-19 pandemic about their health, work and behaviour. These data were linked to the National Health Service digital health records, including COVID-19 diagnostic testing, within the UK Longitudinal Linkage Collaboration (UK LLC) research environment. Poisson regression modelled self-reported infection and diagnostic test confirmed infection within each cohort for work-related characteristics. Relative Risk (RR) were then combined using random effects meta-analysis.

RESULTS

Between March 2020 and March 2021, 74 757 individuals completed 167 302 surveys. Overall, 15 174 survey responses self-reported an infection, whereas 3053 had a linked positive test. Self-reported infection risk was greater in keyworkers versus not (RR=1.24 (95% CI 1.17, 1.31), among non-home working (1.08 (0.98, 1.19)) or some home working (1.06 (0.97, 1.17)) versus all home working. Part-time workers versus full time (0.94 (0.89, 0.99)) and furlough versus not (0.93 (0.88, 0.99)) had reduced risk. Results for the linked positive test outcome were comparable in direction but greater in magnitude, for example, a 1.85 (1.56, 2.20) in keyworkers.

CONCLUSION

The UK LLC provides new opportunities for researchers to investigate risk factors, including occupational factors, for ill-health events in multiple largescale UK cohorts. Risk of SARS-CoV-2 infection and COVID-19 illness appeared to be associated with work-related characteristics. Associations using linked diagnostic test data appeared stronger than self-reported infection status.

Air Pollution and Temperature in Seizures and Epilepsy: A Scoping Review of Epidemiological Studies

PURPOSE OF THE REVIEW

Seizures and epilepsy can be debilitating neurological conditions and have few known causes. Emerging evidence has highlighted the potential contribution of environmental exposures to the etiology of these conditions, possibly manifesting via neuroinflammation and increased oxidative stress in the brain. We conducted a scoping review of epidemiological literature linking air pollution and temperature exposures with incidence and acute aggravation of seizures and epilepsy. We systematically searched PubMed, Embase, Web of Science, and APA PsycINFO databases for peer-reviewed journal articles published in English from inception to February 7, 2024.

RECENT FINDINGS

We identified a total of 34 studies: 16 examined air pollution exposure, 12 ambient temperature, and six examined both air pollution and ambient temperature. Most studies were conducted in Asia (China, Taiwan, South Korea, and Japan). Nearly all studies retrospectively derived acute (daily average), ambient, and postnatal exposure estimates from ground monitoring systems and ascertained epilepsy cases or seizure events through record linkage with medical records, health registry systems, or insurance claims data. Commonly assessed exposures were particulate matter (PM2.5, PM10), nitrogen dioxide (NO2), sulfur dioxide (SO2), carbon monoxide (CO), ozone (O3), and daily mean ambient temperature. Overall, the main findings across studies lacked consistency, with mixed results reported for the associations of air pollutants and temperature metrics with both seizure incidence and acute aggravations of epilepsy.

A maternal and child health administrative cohort in Scotland: the utility of linked administrative data for understanding early years' outcomes and inequalities

Introduction

The early years are considered one of the most impactful points in the life course to intervene to improve population health and reduce health inequalities because, for example, both ill health and social disadvantage can track into adulthood. Scotland's outstanding systems for data linkage offer untapped potential to further our understanding of when and why inequalities in child health, development and wellbeing emerge. This understanding is vital for the consideration of policy options for their reduction.

Methods

Birth registrations, hospital episodes, dispensed community prescriptions, child health reviews and immunisation records were linked for 198,483 mother-child pairs for babies born in Scotland from October 2009 to the end of March 2013, followed up until April 2018 (average age 6 years).

Results

Outcomes include birthweight and newborn health, dispensed prescriptions for mental health medications, tobacco smoke exposure, infant feeding, immunisations, hospitalisation for unintentional injuries, socio-emotional, cognitive and motor development, and overweight and obesity. Several measures are repeated throughout childhood allowing examination of timing, change and persistence. Socio-economic circumstances (SECs) include neighbourhood deprivation, relationship status of the parents, and occupational status. Descriptive analyses highlight large inequalities across all outcomes. Inequalities are greater when measured by family-level as opposed to area-level, aspects of socio-economic circumstances and for persistent or more severe outcomes. For example, 41.4% of the most disadvantaged children (living with a lone, economically inactive mother in the most deprived fifth of areas) were exposed to tobacco smoke in utero and in infancy/toddlerhood compared to <1% in the least disadvantaged children (living with a married, managerial/professional mother in the least deprived quintile of areas).

Conclusion

This novel linkage provides a longitudinal picture of health throughout the early years and how this varies according to family- and area-level measures of SECs. Future linkages could include other family members (e.g. siblings, grandmothers) and other sectors (e.g. education, social care). The creation of additional cohorts would allow for long-term and efficient evaluation of policies as natural experiments.

Creating a learning health system to include environmental determinants of health: The GroundsWell experience

Introduction

Policies aiming to prevent ill health and reduce health inequalities need to consider the full complexity of health systems, including environmental determinants. A learning health system that incorporates environmental factors needs healthcare, social care and non-health data linkage at individual and small-area levels. Our objective was to establish privacy-preserving household record linkage for England to ensure person-level data remain secure and private when linked with data from households or the wider environment.

Methods

A stakeholder workshop with participants from our regional health board, together with the regional data processor, and the national data provider. The workshop discussed the risks and benefits of household linkages. This group then co-designed actionable dataflows between national and local data controllers and processors.

Results

A process was defined whereby the Personal Demographics Service, which includes the addresses of all patients of the National Health Service (NHS) in England, was used to match patients to a home identifier, for the time they are recorded as living at that address. Discussions with NHS England resulted in secure and quality-assured data linkages and a plan to flow these pseudonymised data onwards into regional health boards. Methods were established, including the generation of matching algorithms, transfer processes and information governance approvals. Our collaboration accelerated the development of a new data governance application, facilitating future public health intervention evaluations.

Conclusion

These activities have established a secure method for protecting the privacy of NHS patients in England, while allowing linkage of wider environmental data. This enables local health systems to learn from their data and improve health by optimizing non-health factors. Proportionate governance of health and linked non-health data is practical in England for incorporating key environmental factors into a learning health system.

Time-varying living arrangements and suicide death in the general population sample: 14-year causal survival analysis via pooled logistic regression

AIMS

While past research suggested that living arrangements are associated with suicide death, no study has examined the impact of sustained living arrangements and the change in living arrangements. Also, previous survival analysis studies only reported a single hazard ratio (HR), whereas the actual HR may change over time. We aimed to address these limitations using causal inference approaches.

METHODS

Multi-point data from a general Japanese population sample were used. Participants reported their living arrangements twice within a 5-year time interval. After that, suicide death, non-suicide death and all-cause mortality were evaluated over 14 years. We used inverse probability weighted pooled logistic regression and cumulative incidence curve, evaluating the association of time-varying living arrangements with suicide death. We also studied non-suicide death and all-cause mortality to contextualize the association. Missing data for covariates were handled using random forest imputation.

RESULTS

A total of 86,749 participants were analysed, with a mean age (standard deviation) of 51.7 (7.90) at baseline. Of these, 306 died by suicide during the 14-year follow-up. Persistently living alone was associated with an increased risk of suicide death (risk difference [RD]: 1.1%, 95% confidence interval [CI]: 0.3-2.5%; risk ratio [RR]: 4.00, 95% CI: 1.83-7.41), non-suicide death (RD: 7.8%, 95% CI: 5.2-10.5%; RR: 1.56, 95% CI: 1.38-1.74) and all-cause mortality (RD: 8.7%, 95% CI: 6.2-11.3%; RR: 1.60, 95% CI: 1.42-1.79) at the end of the follow-up. The cumulative incidence curve showed that these associations were consistent throughout the follow-up. Across all types of mortality, the increased risk was smaller for those who started to live with someone and those who transitioned to living alone. The results remained robust in sensitivity analyses.

CONCLUSIONS

Individuals who persistently live alone have an increased risk of suicide death as well as non-suicide death and all-cause mortality, whereas this impact is weaker for those who change their living arrangements.

Linking migration and hospital data in England: linkage process and evaluation of bias

Introduction

Difficulties ascertaining migrant status in national data sources such as hospital records have limited large-scale evaluation of migrant healthcare needs in many countries, including England. Linkage of immigration data for migrants and refugees, with National Health Service (NHS) hospital care data enables research into the relationship between migration and health for a large cohort of international migrants.

Objectives

We aimed to describe the linkage process and compare linkage rates between migrant sub-groups to evaluate for potential bias for data on non-EU migrants and resettled refugees linked to Hospital Episode Statistics (HES) in England.

Methods

We used stepwise deterministic linkage to match records from migrants and refugees to a unique healthcare identifier indicating interaction with the NHS (linkage stage 1 to NHS Personal Demographic Services, PDS), and then to hospital records (linkage stage 2 to HES). We calculated linkage rates and compared linked and unlinked migrant characteristics for each linkage stage.

Results

Of the 1,799,307 unique migrant records, 1,134,007 (63%) linked to PDS and 451,689 (25%) linked to at least one hospital record between 01/01/2005 and 23/03/2020. Individuals on work, student, or working holiday visas were less likely to link to a hospital record than those on settlement and dependent visas and refugees. Migrants from the Middle East and North Africa and South Asia were four times more likely to link to at least one hospital record, compared to those from East Asia and the Pacific. Differences in age, sex, visa type, and region of origin between linked and unlinked samples were small to moderate.

Conclusion

This linked dataset represents a unique opportunity to explore healthcare use in migrants. However, lower linkage rates disproportionately affected individuals on shorter-term visas so future studies of these groups may be more biased as a result. Increasing the quality and completeness of identifiers recorded in administrative data could improve data linkage quality.

Machine learning prediction of major adverse cardiac events after elective bariatric surgery

BACKGROUND

Machine learning (ML) is an emerging technology with the potential to predict and improve clinical outcomes including adverse events, based on complex pattern recognition. Major adverse cardiac events (MACE) after bariatric surgery have an incidence of 0.1% but carry significant morbidity and mortality. Prior studies have investigated these events using traditional statistical methods, however, studies reporting ML for MACE prediction in bariatric surgery remain limited. As such, the objective of this study was to evaluate and compare MACE prediction models in bariatric surgery using traditional statistical methods and ML.

METHODS

Cross-sectional study of the MBSAQIP database, from 2015 to 2019. A binary-outcome MACE prediction model was generated using three different modeling methods: (1) main-effects-only logistic regression, (2) neural network with a single hidden layer, and (3) XGBoost model with a max depth of 3. The same set of predictor variables and random split of the total data (50/50) were used to train and validate each model. Overall performance was compared based on the area under the receiver operating curve (AUC).

RESULTS

A total of 755,506 patients were included, of which 0.1% experienced MACE. Of the total sample, 79.6% were female, 47.8% had hypertension, 26.2% had diabetes, 23.7% had hyperlipidemia, 8.4% used tobacco within 1 year, 1.9% had previous percutaneous cardiac intervention, 1.2% had a history of myocardial infarction, 1.1% had previous cardiac surgery, and 0.6% had renal insufficiency. The AUC for the three different MACE prediction models was: 0.790 for logistic regression, 0.798 for neural network and 0.787 for XGBoost. While the AUC implies similar discriminant function, the risk prediction histogram for the neural network shifted in a smoother fashion.

CONCLUSION

The ML models developed achieved good discriminant function in predicting MACE. ML can help clinicians with patient selection and identify individuals who may be at elevated risk for MACE after bariatric surgery.

Ethnic inequalities in positive SARS-CoV-2 tests, infection prognosis, COVID-19 hospitalisations and deaths: analysis of 2 years of a record linked national cohort study in Scotland

BACKGROUND

This study aims to estimate ethnic inequalities in risk for positive SARS-CoV-2 tests, COVID-19 hospitalisations and deaths over time in Scotland.

METHODS

We conducted a population-based cohort study where the 2011 Scottish Census was linked to health records. We included all individuals ≥ 16 years living in Scotland on 1 March 2020. The study period was from 1 March 2020 to 17 April 2022. Self-reported ethnic group was taken from the census and Cox proportional hazard models estimated HRs for positive SARS-CoV-2 tests, hospitalisations and deaths, adjusted for age, sex and health board. We also conducted separate analyses for each of the four waves of COVID-19 to assess changes in risk over time.

FINDINGS

Of the 4 358 339 individuals analysed, 1 093 234 positive SARS-CoV-2 tests, 37 437 hospitalisations and 14 158 deaths occurred. The risk of COVID-19 hospitalisation or death among ethnic minority groups was often higher for White Gypsy/Traveller (HR 2.21, 95% CI (1.61 to 3.06)) and Pakistani 2.09 (1.90 to 2.29) groups compared with the white Scottish group. The risk of COVID-19 hospitalisation or death following confirmed positive SARS-CoV-2 test was particularly higher for White Gypsy/Traveller 2.55 (1.81-3.58), Pakistani 1.75 (1.59-1.73) and African 1.61 (1.28-2.03) individuals relative to white Scottish individuals. However, the risk of COVID-19-related death following hospitalisation did not differ. The risk of COVID-19 outcomes for ethnic minority groups was higher in the first three waves compared with the fourth wave.

INTERPRETATION

Most ethnic minority groups were at increased risk of adverse COVID-19 outcomes in Scotland, especially White Gypsy/Traveller and Pakistani groups. Ethnic inequalities persisted following community infection but not following hospitalisation, suggesting differences in hospital treatment did not substantially contribute to ethnic inequalities.

Time-sensitive testing pressures and COVID-19 outcomes: are socioeconomic inequalities over the first year of the pandemic explained by selection bias?

BACKGROUND

There are many ways in which selection bias might impact COVID-19 research. Here we focus on selection for receiving a polymerase-chain-reaction (PCR) SARS-CoV-2 test and how known changes to selection pressures over time may bias research into COVID-19 infection.

METHODS

Using UK Biobank (N = 420,231; 55% female; mean age = 66.8 [SD = 8·11]) we estimate the association between socio-economic position (SEP) and (i) being tested for SARS-CoV-2 infection versus not being tested (ii) testing positive for SARS-CoV-2 infection versus testing negative and (iii) testing negative for SARS-CoV-2 infection versus not being tested. We construct four distinct time-periods between March 2020 and March 2021, representing distinct periods of testing pressures and lockdown restrictions and specify both time-stratified and combined models for each outcome. We explore potential selection bias by examining associations with positive and negative control exposures.

RESULTS

The association between more disadvantaged SEP and receiving a SARS-CoV-2 test attenuated over time. Compared to individuals with a degree, individuals whose highest educational qualification was a GCSE or equivalent had an OR of 1·27 (95% CI: 1·18 to 1·37) in March-May 2020 and 1·13 (95% CI: 1.·10 to 1·16) in January-March 2021. The magnitude of the association between educational attainment and testing positive for SARS-CoV-2 infection increased over the same period. For the equivalent comparison, the OR for testing positive increased from 1·25 (95% CI: 1·04 to 1·47), to 1·69 (95% CI: 1·55 to 1·83). We found little evidence of an association between control exposures, and any considered outcome.

CONCLUSIONS

The association between SEP and SARS-CoV-2 testing changed over time, highlighting the potential of time-specific selection pressures to bias analyses of COVID-19. Positive and negative control analyses suggest that changes in the association between SEP and SARS-CoV-2 infection over time likely reflect true increases in socioeconomic inequalities.

Associations between SARS-CoV-2 infection and subsequent economic inactivity and employment status: pooled analyses of five linked longitudinal surveys

Introduction

Following the acute phase of the COVID-19 pandemic, record numbers of people became economically inactive (i.e., neither working nor looking for work), or non-employed (including unemployed job seekers and economically inactive people). A possible explanation is people leaving the workforce after contracting COVID-19. We investigated whether testing positive for SARS-CoV-2 is related to subsequent economic inactivity and non-employment, among people employed pre-pandemic.

Methods

The data came from five UK longitudinal population studies held by both the UK Longitudinal Linkage Collaboration (UK LLC; primary analyses) and the UK Data Service (UKDS; secondary analyses). We pooled data from five long established studies (1970 British Cohort Study, English Longitudinal Study of Ageing, 1958 National Child Development Study, Next Steps, and Understanding Society). The study population were aged 25-65 years between March 2020 to March 2021 and employed pre-pandemic. Outcomes were economic inactivity and non-employment measured at the time of the last follow-up survey (November 2020 to March 2021, depending on study). For the UK LLC sample (n=8,174), COVID-19 infection was indicated by a positive SARS-CoV-2 test in NHS England records. For the UKDS sample we used self-reported measures of COVID-19 infection (n=13,881). Logistic regression models estimated odds ratios (ORs) with 95% confidence intervals (95%CIs) adjusting for potential confounders including sociodemographic variables, pre-pandemic health and occupational class.

Results

Testing positive for SARS-CoV-2 was very weakly associated with economic inactivity (OR 1.08 95%CI 0.68-1.73) and non-employment status (OR 1.09. 95%CI 0.77-1.55) in the primary analyses. In secondary analyses, self-reported test-confirmed COVID-19 was not associated with either economic inactivity (OR 1.01 95%CI 0.70-1.44) or non-employment status (OR 1.03 95%CI 0.79-1.35).

Conclusions

Among people employed pre-pandemic, testing positive for SARS-CoV-2 was either weakly or not associated with increased economic inactivity or non-employment. Research on the recent increases in economic inactivity should focus on other potential causes.

Associations between self-reported healthcare disruption due to covid-19 and avoidable hospital admission: evidence from seven linked longitudinal studies for England

OBJECTIVES

To examine whether there is an association between people who experienced disrupted access to healthcare during the covid-19 pandemic and risk of an avoidable hospital admission.

DESIGN

Observational analysis using evidence from seven linked longitudinal cohort studies for England.

SETTING

Studies linked to electronic health records from NHS Digital from 1 March 2020 to 25 August 2022. Data were accessed using the UK Longitudinal Linkage Collaboration trusted research environment.

PARTICIPANTS

Individual level records for 29 276 people.

MAIN OUTCOME MEASURES

Avoidable hospital admissions defined as emergency hospital admissions for ambulatory care sensitive and emergency urgent care sensitive conditions.

RESULTS

9742 participants (weighted percentage 35%, adjusted for sample structure of longitudinal cohorts) self-reported some form of disrupted access to healthcare during the covid-19 pandemic. People with disrupted access were at increased risk of any (odds ratio 1.80, 95% confidence interval 1.39 to 2.34), acute (2.01, 1.39 to 2.92), and chronic (1.80, 1.31 to 2.48) ambulatory care sensitive hospital admissions. For people who experienced disrupted access to appointments (eg, visiting their doctor or an outpatient department) and procedures (eg, surgery, cancer treatment), positive associations were found with measures of avoidable hospital admissions.

CONCLUSIONS

Evidence from linked individual level data shows that people whose access to healthcare was disrupted were more likely to have a potentially preventable hospital admission. The findings highlight the need to increase healthcare investment to tackle the short and long term implications of the pandemic, and to protect treatments and procedures during future pandemics.

Thirty-three myths and misconceptions about population data: from data capture and processing to linkage

Databases covering all individuals of a population are increasingly used for research and decision-making. The massive size of such databases is often mistaken as a guarantee for valid inferences. However, population data have characteristics that make them challenging to use. Various assumptions on population coverage and data quality are commonly made, including how such data were captured and what types of processing have been applied to them. Furthermore, the full potential of population data can often only be unlocked when such data are linked to other databases. Record linkage often implies subtle technical problems, which are easily missed. We discuss a diverse range of myths and misconceptions relevant for anybody capturing, processing, linking, or analysing population data. Remarkably, many of these myths and misconceptions are due to the social nature of data collections and are therefore missed by purely technical accounts of data processing. Many are also not well documented in scientific publications. We conclude with a set of recommendations for using population data.