Suchergebnisse

ObjectivesTo make research ready data from transcribed historic civil registration records: birth, marriage and death (from mid-19th century onwards). To use historic records effectively for large-scale research they must not only be made machine-readable, but also coded in a suitable format – along with classification of transcribed information. Included on the digitised historic records are textual descriptions of occupations and causes of death. Thus, to code transcribed occupations to HISCO and causes of death to ICD.
ApproachIt is impractical to hand-code the records manually (34 million occupations and 8 million causes of death), especially for deaths where more than one cause can be given. As such, coding is viewed as a text classification task and the process is automated. To facilitate auto-coding, a proportion of records were hand-coded as training data (90,000 occupations and 102,000 deaths). Ahead of the auto-coding, initial pre-processing, cleaning and standardising is done on both the occupations and deaths.
ResultsPreliminary experiments undertaken obtained reasonable results from a combination of exact matching and statistical classification. Experiments using a larger pilot uncovered that since some occupations are very common, the training data set covers a very large proportion of the records (ie exact match). This proportion is not as high for deaths given the different ways causes are written.
ConclusionThis is work in progress to create the research ready data, and the poster will include the results from experiments using the full training data (90,000 occupations and 102,000 deaths).

ObjectivesAims of this research, involving data linkage and health outcomes, is to gain a full understanding of the impact of both fertility histories and childlessness on health outcomes mid-life accounting for socio-economic background and area of residence. The research draws on and extends work on reproductive histories and life-course outcomes.
ApproachWe aim to extend this area of research, specifically for Scotland, using Census data (1991-2011) from the Scottish Longitudinal Study (SLS) linked to health data. The Census health measures – including the 2011 Census health condition question on mental health - are the research outcomes and the explanatory information is from Census socio-economic data (captured around peak fertility for the research cohort in 1991), along with the SMR02 Maternity and SMR04 Mental Health datasets. The time-frame for available data allows 20 year follow-up from the 1991 Census to mid-life for specific female SLS birth cohorts (born 1959-1966, aged ~45-52 in 2011).
ResultsFrom preliminary modelling we initially find, for this specific female research cohort, high birth parity to be an important factor in relation to self-reported mental health conditions at follow-up in 2011, but not once socio-economic and area-level variables are controlled for.
ConclusionPreliminary modelling also highlights that relationship status – single, married or cohabiting – to be important over that of legal marital status as recorded at Census. For limiting long-term illness as a health outcome the findings are similar.

ObjectivesIn Spring 2020, Public Health Scotland (PHS) were tasked with seeding the NHS Scotland Community Health Index (CHI) with a Unique Property Reference Number (UPRN). This was to allow understanding of the impact on people residing with someone infected with COVID-19, but utility goes much further than this.
ApproachAddress fields were extracted from CHI records (August 2020) of persons still alive in Scotland or who died since January 2020.
Some pre-formatting was carried out on unique address strings that were then processed at DataHub, a service, provided by the Improvement Service, helping public sector organisations secure access to a data matching service.
Files were returned with a best matching UPRN attached, and a category indicating the quality of the match.
To enhance the seeding of care homes addresses, unmatched CHI records were manually searched, and UPRN added where expert knowledge determined it was a care home address.
ResultsThere were 3,208,951 unique address strings processed, relating to 5,828,951 CHI registrations, including people deceased since January 2020. In total, 5,207,389 (89.3%) people had a UPRN returned with match categories EXCELLENT (78.8%), GOOD (8.3%), or FAIR (2.2%).
Only 60% of the 40,196 current CHI records that were pre-indicated with an institutional care home flag could be automatically matched to a UPRN. Following the manual expert review, 42,334 people in adult care homes could be identified with a specific UPRN.
A random sample of 1,724 address pairs, stratified by match category and urban-rural category, was selected for quality assurance review by an independent checker. Results have still to be verified, but precision is estimated to be more than 95% of matches to the correct UPRN.
ConclusionThe creation of the CHI-UPRN Residential Linkage (CURL) is a major development to CHI as a data linkage tool in Scotland. CHI has long been used as a population spine for patient-level data linkage in research projects but introducing a property-level dimension through CURL provides new opportunities for researchers.

Background Life expectancy in Glasgow is lower than other Scottish cities and comparable English cities. One explanation for this excess mortality is selective migration, specifically policy from the late 1940s onwards to move younger, more skilled workers and their families from Glasgow to five Scottish New Towns, established between 1948 and 1966, leaving behind those who are more deprived and unhealthy.
Aim To examine the impact of migration to New Towns on premature mortality in Glasgow using linked Scottish Mental Survey 1947 and NHS Central Register data.
Methods This is a new data linkage using historical data for health board moves dating back to 1948.
The Scottish Mental Survey 1947 encompasses a cohort born in 1936 that had cognitive ability assessed at age 11. This cohort were of young working age at the time of New Towns policy. The cohort has been linked to NHS Central Register data, which contains information on current and previous health board location dating back to the late 1940s, as well as date of death. A time-appropriate move to the relevant health board area is used as a proxy for New Town migration. We can put people "back" to their starting geography to see what impact this has on premature mortality in Glasgow.
We will examine factors associated with geographical mobility. Logistic regression and cox regression models will be used to analyse the association between migration status and early mortality.
Results Preliminary results suggest that, as hypothesised, those who have migrated out of Glasgow to new Towns have lower likelihood of premature mortality than those who remain. Analysis is ongoing and we will present our final findings.
Conclusion With New Towns being re-considered as an answer to the current acute UK housing shortage, it is useful to understand the, perhaps unexpected, consequences of such policies.

IntroductionIn the natural sciences biomonitors, (organisms, such as pine needles, shells, lichen, that can provide quantitative information on the quality of their past and present environments) have been developed for environmental measurement. In recent year's physiologists have started to explore if human tissue could also be used.
Objectives and ApproachThe costs of collecting and processing human tissues for biomonitoring, may be too prohibitive for its use in wide scale monitoring. In contrast if biomonitors could be identified within a routine data collection process that were part of standard medical recording a low cost, widely available large sample would be available to scientists. Pregnancy is known to be effected by air pollution, therefore we explore whether birthweight, recorded in maternity records, could be a biomonitor for air pollution.
ResultsWe use maternity records (~1 million births) in Scotland between 2000 and 2015. We modelled, at the individual mother level birthweight, controlling mother's age, estimated household income, local area crime rates and area level of multiple deprivation. We then aggregate and calculate the averaged of the residuals for this model for all mothers within an intermediate datazones (small areas of around 4000 residents). These mean deviations were then compared with pollution modelled figures produced by AEA for the Scottish government averaged over the same spatial units and time period as the maternity data. "We find a relatively strong correlations (between -0.37 and -0.39) between our 'biomonitor estimates' of air pollution derived from the maternity records and the entirely separated modelled air pollution data."
Conclusion/ImplicationsAs far as we know this is the first study to demonstrate that it may be possible to use routine health data to derive 'biomonitors' information. Importantly if this method proves to be reliable it will be a relatively cheap method for collecting information that is actually personally monitoring.

IntroductionWe know that rates of adverse pregnancy outcome such as low birthweight and preterm births vary both by socio-economic status and across geographical areas. Furthermore, more deprived neighbourhoods show higher rates of adverse outcomes. However, much of the evidence comes from cross-sectional observational studies where causality is difficult to assess.
Objectives and ApproachNeighbourhood crime and a resultant increase in maternal stress is a plausible explanation for the deprivation-birth outcome relationship. Without an experiment, isolating this pathway from other confounding effects is challenging. Because many mothers (within the Scottish maternity dataset) experience multiple pregnancies and varying exposure for the same person, we can assume a quasi-experimental allocation to exposure. We model associations as both within mother (comparing exposures between pregnancies to the same mother) and between mother (comparing different people between more and less deprived areas) partitioning the variation and effects accordingly.
ResultsWe find that both composite neighbourhood deprivation and neighbourhood crime specifically is associated with reduced birthweight. However, the composite neighbourhood deprivation measure is only significant when modelled as a between mother effect. Neighbourhood crime on the other hand is significantly associated with both between mother and within mother effects (though the latter is reduced compared to the former). These results suggest that much of the observed relationship with deprivation in general is therefore confounded by factors associated with the composition of individuals in these areas rather than an area deprivation effect per se. In contrast, the association with neighbourhood crime can be considered in isolation from the composition of individuals and is therefore likely to be an independent "area" effect.
Conclusion/ImplicationsThese results are important because outcomes of pregnancy including birthweight and premature births influence health and social outcomes not just in infancy but across the life course. This study suggests policies targeting neighbourhood crime may help to reduce neighbourhood disparities in pregnancy outcomes and therefore health inequalities in general.

BackgroundIn recent years, promoting cycling as a means of transport has been put on multiple policy agendas, including health, transport and climate change. Provision of cycling facilities is important in promotion of bicycle use. However, there are no conclusive findings in terms of what cycling facilities are associated with higher levels of bicycle use.
ObjectivesIn this paper we aim to assess the influence of off-road cycle paths and on-road lanes, by examining whether proximity to two types of cycle route is associated to the level of cycling to work. The research findings has potential to provide further evidence on policy interventions in promoting cycling to work through development of new cycling infrastructures.
MethodsThe cycling to work data are from the 2001 and 2011 censuses at the output area level. Proximity to cycling facilities is defined for output areas if they are within 400 metres from their centroid to a cycle route. We fit regression models to examine the association of cycle paths and cycle lanes with levels of cycling to work adjusting for a number of socioeconomic factors.
FindingsThe modelling results show that proximity to off-road paths is associated with increased levels of cycling to work but proximity to on-road lanes shows no effect.
ConclusionsAlthough this is by no means indicative of a causal relationship, this research provides further evidence on the effects of cycle facilities on cycling behaviour, lending support on policy interventions that cycle facilities can be built to promote cycling to work. We will extend this research using the individual level census data which will give us opportunity to overcome the ecological fallacy. In addition we will use coarsened exact matching to identify a control group in order to make causal inference on the effect of cycle paths.

ObjectivesWe aim to estimate occupational differences in COVID-19 hospital admission and mortality by sex. Occupations vary with respect to environmental factors that influence exposure to COVID-19 such as ventilation, social contact and protective equipment. Variations between women and men may arise because they occupy different roles within an occupation or from behavioural differences.
ApproachWe established an innovative linked data collection combining individual-level data from 2011 Census with health administrative records and household identifiers from Ordnance Survey's Unique Property Reference Number (UPRN). Using data for a cohort of approximately 1.7 million Scottish adults aged 40 to 64 years, we analysed occupational differences in COVID-19 hospital admission and mortality, during the period between 1 March 2020 and 31 January 2021. We estimated age-standardised incidence rates (ASIRs) for COVID-19 hospital admission and death per 100,000 adults, stratified by sex and occupation. Using Cox proportional hazards models, we estimated COVID-19 hospital admission and death risks, adjusting for socio-economic and pre-pandemic health factors.
ResultsGenerally, women had lower ASIRs for COVID-19 hospital admission and mortality compared to men. For COVID-19 mortality, ASIRs were highest among women employed in elementary trades and related occupations (e.g. packers/canners). For hospital admission, high rates were observed among women working in caring personal services (e.g. nursing assistants/ambulance staff). Among men, ASIRs were highest among those in elementary services occupations (kitchen assistants/waiters) and taxi drivers, who also had the highest admission rate. After adjusting for pre-pandemic health factors, we observed lower death risks for women employed as health professionals, and those in associate professional and technical occupations (medical technicians/paramedics), compared to the baseline model. Lower adjusted admission risks were also observed for both women and men in professional occupations (e.g. in science/engineering and teaching/education). Among men, adjusted death risks remained elevated for large vehicle and taxi drivers, who also displayed high admission risks.
ConclusionOccupational differences in COVID-19 hospital admission and mortality between women and men may be explained by a mixture of social, workplace and behavioural factors. These need to be considered when designing intervention policies which aim to reduce infections in the workplace.

AbstractAdministrative data are widely used to construct indicators of social disadvantage, such as Free School Meals eligibility and Indices of Multiple Deprivation, for policy purposes. For research these indicators are often a compromise between accuracy and simplicity, because they rely on cross-sectional data. The growing availability of longitudinal administrative data may aid construction of more accurate indicators for research. To illustrate this potential, we use administrative data on welfare benefits from DWP's National Benefits Database and annual earnings from employment from HMRC's P14/P60 data to reconstruct individual labour market histories over a 5-year period. These administrative datasets were linked to survey data from the Poverty and Social Exclusion UK 2012. Results from descriptive and logistic regression analyses show that longitudinal measures correlate highly with survey responses on the same topic and are stronger predictors of poverty risks than measures based on cross-sectional data. These results suggest that longitudinal administrative measures would have potentially wide-ranging applications in policy as well as poverty research.

BackgroundThe status of the Scottish Offenders Index (SOI) as a census (n=all) of conviction proceedings in Scotland allows the adoption of research designs that would be prohibitively costly and time consuming to implement with more traditional survey approaches. This paper argues that this flexibility, combined with the large time span covered by the SOI (1989-current), provides a way to bridge the typical individual (micro-) focus of developmental criminology and the (macro-) analysis of aggregate crime rates.
ObjectivesBy examining conviction patterns for multiple cohorts across multiple years, we can infer whether recent falls in aggregate conviction rates are most plausibly explained by period or cohort effects. In turn, this distinction between period and cohort effects can both suggest which potential explanations for these recent falls in conviction rates in Scotland are most feasible - thus helping to refine explanations for the recent 'crime drop' - and help us understand the factors leading to change in criminal careers over time.
Methods and FindingsTo support these claims, this paper presents the results of an exploratory analysis of change in convictions patterns in the SOI between 1989 and 2011, drawing on demographic information to account for change in population structure. Results show a complex pattern of change over time, with differences in convictions trends for young and old, for men and women and in different periods. Explanations of the crime drop must be able to account for these complex trends.
ConclusionsAdministrative data provide a unique window into the recent crime drop, and we encourage similar studies on comparable datasets where possible.

We describe results on the creation and use of synthetic data that were derived in the context of a project to make synthetic extracts available for users of the UK Longitudinal Studies. A critical review of existing methods of inference from large synthetic data sets is presented. We introduce new variance estimates for use with large samples of completely synthesised data that do not require them to be generated from the posterior predictive distribution derived from the observed data and can be used with a single synthetic data set. We make recommendations on how to synthesise data based on these results. The practical consequences of these results are illustrated with an example from the Scottish Longitudinal Study.

As the population ages, older adults are expected to work for longer into the life course. However, older adults experience particular problems staying economically active, even prior to reaching statutory retirement. Recent work has suggested that economic activity in midlife can be predicted by the far-reaching effects of early life, such as childhood socio-economic circumstances, cognitive ability and education. The present study investigates whether these same early-life factors predict the odds of being economically active much later in life, from age 55 to age 75. We capitalise on data linkage conducted between a subsample of the Scottish Mental Survey 1947 cohort and the Scottish Longitudinal Study, which includes three waves of national census data (1991, 2001 and 2011). The structural association between early-life factors and later-life economic activity was assessed using latent growth curve analyses conducted for males and females separately. In both males and females, the odds of being economically active decreased non-linearly across the 20-year follow-up period. For males, greater odds of being economically active at age 55 were predicted by higher childhood cognitive ability and higher educational attainment. For females, greater odds of being economically active at age 55 were predicted by higher childhood socio-economic status and higher childhood cognitive ability. In contrast, early-life factors did not predict the odds of becoming inactive over the 20-year follow-up period. We suggest that early-life advantage may contribute to the capacity for work in later life, but that it does not necessarily protect from subsequent decline in this capacity.