Opportunities for longitudinal data linkage in Scotland

Abstract

Scotland has existing data resources which are competitive internationally and available to researchers from elsewhere. The Scottish Informatics and Linkage Collaboration (SILC) was recently launched, allowing data sets to be linked within and between sectors (e.g. health to non-health). The purpose of this review article is to introduce and define key terms in data linkage, to describe the emerging data linkage resources available in Scotland and to describe the opportunities available in Scotland to researchers internationally. The review is aimed at researchers internationally who are interested in data linkage using Scottish data resources. The review makes particular reference to longitudinal health data but emphasises that linkage to non-health data allows research questions to be considered that were previously not answerable. The review is focused on longitudinal data resources (e.g. cohort studies and repeated measures designs), since they are usually the focus of data linkage research. The review concludes that any intended data linkage for research should be driven by a clear research question. The infrastructure already available and the launch of SILC will accelerate research in Scotland and generate new research questions that previously could not be considered.

Keywords

Data linkage longitudinal studies Scotland administrative data epidemiology

Introduction

Scotland already has data resources which are competitive internationally and available to researchers from elsewhere. The purpose of this review article is to describe the emerging data linkage resources available, so that researchers interested in data linkage opportunities become aware of what Scotland can offer. The review makes particular reference to health data (e.g. primary care, hospital records, cancer registries, mortality data) but emphasises that linkage to non-health data (e.g. education, social care and crime) may offer unique opportunities for answering certain research questions. The review is focused on longitudinal data resources (e.g. cohort studies and repeated measures designs), since these are often the focus of data linkage efforts, and are usually considered better quality evidence in the ‘hierarchy of evidence’ than cross-sectional data or case reports, for example.¹ Longitudinal data can, however, be analysed as cross-sectional data, by restricting to the analysis to one measurement occasion if required. The review is also focused on data that have been digitised, but it is worth noting that a great deal of historic administrative data exists only in paper records.

Data linkage

Data linkage involves bringing two records together that belong to the same individual. This might involve linking records that belong to the same individual over time in the same data set (e.g. repeated hospital admissions for a patient), or linking records that belong to the same individual from two different data sets (e.g. a patient's health records to their social care records). Data linkage can result in two kinds of data linkage errors: false matches (two people are assigned the same ID) or missed matches (the same person is assigned more than one ID).^2,3

In Scotland following development of the Scotland-wide Data Linkage Framework (SDLF), the Scottish Informatics and Linkage Collaboration (hereafter, SILC) has now been launched. Subject to approval, researchers will be able to obtain linked data from different sectors. Each database retains its own identification, but these are linked my matching different identifiers together (e.g. sex, date of birth, National Health Service (NHS) number) through an index called the population spine. This population spine has a unique ID for each individual. Existing data from different sectors can then be linked, using the population spine, although data linkage success will depend on the accuracy and quality of the identifiers available.³ Linked data will be available for researchers to analyse in secure settings called safe havens, several of which exist already.

The SILC project builds on the success of the previous ScottisH Informatics Programme (SHIP) infrastructure (now part of the Farr Institute in Scotland), which already has an indexing service and various secure databases. The project is expected to enhance research opportunities, improve how the public's administrative data are processed and attract investment into Scotland.⁴ Investment can be generated from being able to answer novel research questions by data linkage for the first time in Scotland (due to infrastructure already available and under development), innovation in data linkage methodology, attracting researchers to move to Scotland in order to take advantage of the data linkage resources available and enhanced reputation for high quality research. Administrative data refer to data that are collected routinely by government departments, not originally for research purposes. The health informatics infrastructure already developed in Scotland then, has provided the foundation for data linkage involving data sets beyond health (e.g. education, social care and crime).

Historical context

It could be argued that attempts to pool data from different sectors in Scotland are nothing new – having been attempted three times historically. An early attempt to pool data in order to map the economy, demography and geography of Scotland (1680/1790) was followed by Sinclair's enormous ‘Statistical account of Scotland’ (1790/1799) and two sequels, the ‘Statistical account of Scotland II’ (1834/1845) and a final survey (1951/1992) which had staggered completion dates. There have also been regular Censuses which collect data on individuals, from 1801 and then every 10 years except during World War II. All these exercises however, produced cross-sectional data (with the exception of the Scottish Longitudinal Study, linking together around 5% of each census from 1991 onwards). In epidemiology and public health, cross-sectional data can be helpful for determining prevalence of a disease, but not for determining incidence – the rate of new cases over a specified time period.¹ Longitudinal data, where data on the same individuals are collected more than once, or where individuals are monitored to see what happens to them, allows calculation of incidence rates and has several additional advantages over cross-sectional data. For example in epidemiology research, it is easier to make claims about cause (exposure) and effect (outcome) when the exposure was measured before the outcome. Cross-sectional surveys make it difficult to disentangle causes, effects and confounding factors; nor is it possible (without data linkage) to evaluate key end points such as cancer, cardiovascular events and death.

International context

Scotland is already internationally competitive in terms of data linkage, particularly in health informatics. Data linkage systems have been created in Canada, Australia and Scandinavia and more recently in other UK countries (e.g. the SAIL databank in Wales). The Oxford Record Linkage System is an early example of data linkage in the health sector, but this has been discontinued. Health informatics in Scotland is competitive in part to the foresight of the creators of the Community Health Index (CHI) number which was initially used in Tayside only, but then was used routinely for all people who come into contact with healthcare providers in Scotland. This meant that many different kinds of health data sets had the same ID number (e.g. primary care, hospital records, diabetes registers). The CHI contains date of birth and is therefore more disclosive than the NHS number. CHI numbers have been extremely useful in facilitating data linkage in Scotland. Today, CHI is used by the NHS National Services of Scotland indexing team for health research, whereas NHS number (from NHS Central Registry) is mainly used to socio-economic research projects by National Records of Scotland (NRS). There is regular linkage between NHS and CHI number however, with a nightly update ensuring high correspondence between the two systems. At the time of writing, the Economic and Social Research Council (ESRC) has funded Administrative Data Research Centres (ADRC) in England, Wales, Scotland and Northern Ireland. It is not yet clear whether population spines will be created in all four countries, but it is likely that the infrastructure and innovation developed in Scotland will inform what happens elsewhere.

United Kingdom context

Data research and informatics agencies within Scotland and the UK provide data sets and linkage resources for researchers. The Electronic Data Research and Innovation Service (eDRIS) provides support for health data linkage in Scotland, and is funded by NHS Scotland. It draws on NHS Scotland Information Services Division (ISD) Record Linkage Service and contributes to the Farr Institute at Scotland, incorporating elements of the earlier SHIP. Agencies across the UK do work together in order to address so-called ‘complex cases’ such as Scottish patients presenting in England for the first time, or vice versa. Administrative data are collected routinely for the UK as a whole, nationally in Scotland and by local authorities. Prior to the launch of the ADRC centres, there was often no central body for researchers to contact in order to obtain all the administrative data needed. For some linkage projects, it may be necessary to contact local authorities separately and then pool the data, as done recently for a project linking health with social care data.⁵ Many authorities do however submit their data for national processing or to the Department for Communities and Local Government (DCLG), in which case it may be sensible to contact government departments in order to obtain it.

Policy and legal context

Recent legal and policy changes in Scotland mean that now is a particularly good time to consider data linkage opportunities. Aimed at improving record keeping in Scotland, the Public Records (Scotland) Act 2011 is a key piece of legislation,⁶ the first specifically concerning data linkage since 1937. The act was introduced to improve record keeping in order to prevent abuse and protect vulnerable people,⁶ allowing data linkage between sectors to achieve this aim. Subsequent policy documents and data linkage exercises have highlighted the potential for protecting vulnerable people and other benefits of data linkage such as reducing costs.^4,5,7 The economic climate at the time of writing is important to consider. Scotland's capital budget is being cut by 37% in real terms. The budget for 2012/2013 was intended to ensure that investment was made in infrastructure, attracting investment and job creation. There is a move towards (and an increased investment in) so-called ‘preventative spending’ which includes adult social care, the early years and reducing re-offending. Sectors are encouraged to work together in order to reduce expenditure in the future. More frequent and better quality (fewer data linkage errors) data sharing across sectors is thought to contribute to this objective. Data linkage from existing research studies to administrative data is also featured in the recent review of how research is used to support the Scottish Government in its objectives.^4,7

Benefits of data linkage in Scotland

The benefits of data linkage from longitudinal studies to administrative data are numerous. A recent example is the linkage of data from Scottish Health Surveys to determine the association between shorter height and dementia death.⁸ A key advantage of linking research to administrative data is the ability to continue monitoring participants who might otherwise have been lost due to drop-out, migration and so on. While people often drop-out from longitudinal research and from clinical trials, they rarely drop-out of administrative data.⁹ If participants in a prospective cohort provide informed consent to be followed for health outcomes, their administrative data can be checked even if they no longer take part in future follow-ups by the research study itself. People recruited and retained into longitudinal research tend to have better physical health, higher levels of educational attainment and have higher socio-economic status.¹⁰ This can produce biased results when data are analysed, because the sample can suffer from health selection^11,12 and healthy survivor effects. Health selection refers to restriction of range on study variables at recruitment, for example, if people tend to have better physical health than the wider population.¹² Healthy survivor effects refer to the tendency for unhealthier participants to die or drop-out of follow-ups. This can produce biased estimates of associations between smoking and lung cancer for example, if smokers who remain active in the study are healthier than those who have left.¹³ Quantifying the extent of health selection at recruitment into a study is often difficult, because the characteristics of the non-responders are not known. Using routinely collected data means that researchers can check for differences between their participants and those not included in the study, using chi-square tests for proportions expected and observed, for example. Health selection is a particular problem for cohorts established in adulthood.¹⁴

Longitudinal data resources in Scotland

There are two kinds of longitudinal data in Scotland that might be considered useful for research purposes. First, there are longitudinal studies conducted in Scotland on residents in Scotland. Second, there are longitudinal studies conducted across the UK which contain residents in Scotland. Table 1 includes both scenarios, and shows the names of the studies, their acronym, an approximate sample size and dates. The online Supplementary Table shows the topics and sectors covered by these studies. Each had different aims and objectives, but the area of their main focus is shown. Links to study protocols or ‘cohort profiles’ are made available, which are useful for obtaining more detail about a particular study.

Table 1.

Longitudinal data resources in Scotland, sample sizes and study protocols/profiles.^a

Name of study	Acronym	N in Scotland	Start date	End date	Study protocol, where available
1946 birth cohort (MRC National Survey of Health and Development)	NCDS	343	2013	Ongoing	16
36-Day sample	36-day	2039	1947	2014	17
6-Day sample	6-day	527	1947	1963, then 2012 onwards	17
6-Day sample siblings	6-day siblings	863	1948	Not yet recruited	17
SMS1947 twins	SMS1947 twins	465	1947	Not yet recruited	17
SMS1947	SMS1947	67,969	1947	Not yet recruited	17
Aberdeen Children of the 1950s	ACONF	12,150	2010	2011	17,18
Adolescent Lifestyles In Contemporary Europe	ALICE	2937	2010	2011	19
Annual Survey of Hours and Earnings Panel	ASHE	Varies	1997	2013
Aspirin for Asymptomatic Atherosclerosis	AAA	28,980	2006	2010	20,21
Biobank	BB	36,000	1937	2003	22
Boyd Orr Cohort	BOC	2358	1970	2012	23
British Cohort Study	BCS1970	1617	2013		24
British Election Study Six-Wave Panel Survey, 2005–2009	BEPS	183	2005	2009	25
British Election Panel Study, 1997–2001; Waves 1 to 8	BEPS	465	1997	2001	25
British Election Panel Study, 2001; Waves 1 and 2	BEPS	948	2001	2001	25
British Election Panel Study, 1992-1997	BEPS	957	1992	1997	25
British General Election Panel Study, 1987–1992	BEPS	552	1987	1992	25
British Election Panel Study, 1983, 1986 and 1987	BEPS	504	1983	1987	25
British Election Study, February 1974, October 1974, June 1975, May 1979; Panel Survey	BEPS	107	1974	1979	25
British Election Study, 1969, June 1970, February 1974; Panel Survey	BEPS	234	1969	1974	25
British Household Panel Study (now Understanding Society)	BHPS	1500	1991	Ongoing	26,27
British Regional Heart Study	BRHS	7735	1975	2004	28
Edinburgh Artery Study	EAS	1592	1987	1988	21
Edinburgh Study of Youth Transitions and Crime	ESYTC	4389	1998	2009	29
Families and Children Study	FACS	7000	2002	2008	30
General Household Survey, 2005	GLS/GLS	1171	2005	2005	31
General Household Survey, 2006	GLS/GLS	893	2006	2006	31
General Household Survey: Time Series Dataset, 1972–2004	GLS/GLS	81,503	1972	2004	32
Generation Scotland: Scottish Family Health Study	GS	30,000	2006	2011	33
GoWell	GW	6000	2005	2015	34
Growing Up in Scotland	GUS	5217	2005	Ongoing	35
Health and Lifestyle Survey	HALS	925	1984	1992	36,37
Healthy Old People in Edinburgh	HOPE	603	1990	2006	38,39
Labour Force Survey	LFS	1000 to 6000	1993	2006
Life Opportunities Survey: Wave One, 2009–2011	LOS	3709	2009	2011	40
Lothian Birth Cohort 1921	LBC1921	550	1999	Ongoing	17
Lothian Birth Cohort 1936	LBC1936	1091	2006	Ongoing	17
M74 Study	M74	1322	2005	2006	41,42
MIDSPAN Family study	MIDSPAN:FS	2338	1996	2001	43
MIDSPAN: Collaborative study	MIDSPAN:CS	7028	1970	1977	43
MIDSPAN: Main and Tiree study	MIDSPAN:MT	3931	1964	1970	43
MIDSPAN: Renfrew/Paisley study (3 cohorts)	MIDSPAN:RP	15,402	1972	1979	43
Millennium Cohort Study	MCS	2370	2000		44
National Child Development Study	BCS1958	1985	1958	2009	45
National Diabetes Register Data	NDRD	52,280	2005	2006	46
National Study of Health and Growth, Phases I to III	NSHAG	2000	1972	1984
Orkney Complex Disease Study	ORCADES	2100	2005	2011	47
Peers and Level of Stress	PaLS	3950	2006	2006
Prospective Study of Pravastatin in the Elderly at Risk	PROSPER	2520	1997	2002	48
Renton Study: Housing-led Regeneration and Health	RENTON	100	2000	2005	49
Scotland's Housing And Regeneration Project	SHARP	723	2002	2008	50
Scottish Health and Ethnicity Linkage Study	SHELS	4,600,000	2001	2001	51
Scottish Health Survey 1995	SHS1995	7932	1995	1995	52
Scottish Health Survey 1998	SHS1998	9047	1998	1998	53
Scottish Health Survey 2003	SHS2003	8148	2003	2003	54
Scottish Health Survey 2008	SHS2008	6465	2008	2008	55
Scottish Health Survey 2009	SHS2009	7531	2009	2009	55
Scottish Health Survey 2010	SHS2010	9038	2010	2010	56
Scottish Health Survey 2011	SHS2011	9531	2011	2011	57
Scottish Heart and Arterial Risk Prevention Survey	SHARP	19,400	1991	1996	58
Scottish Longitudinal Study	SLS	274,000	1991	2011
Sexual Health & Relationships Study	SHARE	7616	1996	2000
Studying Physical Activity in Children's Environments across Scotland	SPACES	300	2013	2013
Survey of New Refugees	SNR	300	2005	2009
Tayside Child Health Information Service	TCHIS	20,000	1995	2002
Tayside Growth Study	TGS	40,000	1989	1996
The Feeding Your Baby Study	FYBS	355	2009	2010
UK Women's Cohort Study	UKWCS	2000	1995	2006	59
Understanding Society (wave 19 + of UK Household Longitudinal Study)	US	7500	2009	Ongoing
Viking Health Study: Shetland	VHS	∼2000	2013	Ongoing
Walker project	WALKER	48,404	1952	1966
Wealth and Assets Survey	WAS	5500
West of Scotland 11 to 16 & 16 + Study	WS11-16	2586	1994	2006
West of Scotland Coronary Prevention Study	WOSCOPS	6595	1989	1995	60
West of Scotland Twenty 07	WS20-07	8266	1987	2008
Work & Pensions Longitudinal Study	WPLS	∼1500	2002	Ongoing

Although not included in this report, it is legitimate to follow up or link cross-sectional data to administrative data and long term outcomes. It is also legitimate to convert randomised controlled trials into longitudinal studies, as done for several studies here.

It must be emphasised that any data linkage project should be driven by a clear research question, and one which cannot be answered without the linkage. Efficient data linkage in the longitudinal sector will usually involve linking a set of predictor variables from a longitudinal study, with a set of outcomes from routinely collected administrative data – but not always. Techniques such as meta-analysis of individual participant data involve pooling data from individuals in order to increase statistical power overall.⁸ Statistical power, the ability to detect an association between an exposure and an outcome, or to correctly detect that there is no association, increases with large sample sizes.¹ Many researchers are familiar with the importance of doing sample size calculations in order to recruit a sample that has sufficient statistical power to detect a particular size of effect (or difference). (1) In data linkage however, the sample sizes available are typically much larger and statistical power is enhanced for looking at smaller groups of individuals and rarer outcomes.

Researchers who want to use data linkage might want to propose a linkage that would allow them to group several research questions together, making the exercise more cost-effective. For example, they could study the association between an exposure and three different health outcomes from the same data linkage application using NHS number: cancer registration, all-cause mortality and cause-specific mortality. Those evaluating proposals however, will need to be careful that proposals are too broad and research questions are sufficiently precise, to avoid requesting data that is not actually needed. If researchers intend to request a large dataset in order to answer several different questions, they will need to be specific from the outset about which data are needed. For example, an epidemiologist could request one data extract, in order to study 10 exposures with two or three outcomes from administrative data. This is more efficient than requesting 30 different data extracts.

Conclusions

The SILC service offers opportunities for transforming cross-sectional data into longitudinal data. Data linkage from longitudinal studies to administrative data beyond health data is a promising new development. As mentioned previously, health data have been linked to social care outcomes.⁵ As mentioned previously however, a great deal of administrative data exists however but only in paper records. For example, the National Records of Scotland hold transcripts from Sheriff courts in paper format. Linkage to these records would not be practical in the short to medium term, because considerable investment would be needed in order to digitise the data.

Take home messages

Any intended data linkage for research should be driven by a clear research question. Efficient data linkage in the longitudinal sector will usually involve linking a set of predictor variables from a longitudinal study, with a set of outcomes from routinely collected administrative data. Researchers may want to identity studies in Scotland that have particularly detailed information on a set of predictors, repeated measurements of these predictors and data on the ‘outcomes’ they are interested in, whether health or non-health (see online Table S1). For example, a researcher interested in the role of early educational exposures measuring repeatedly in adolescence to early adulthood, on health and social care in old age, might find the 36-day sample particularly relevant.¹⁵ Researchers could group research questions together and propose a linkage that would allow them to address several questions with the same data, making the linkage more cost-effective. The infrastructure already available in Scotland and the launch of the SILC will accelerate research in Scotland and generate research questions that could not even be considered until now.

Footnotes

Glossary

Administrative data	Data that are collected routinely by government departments, not originally for research purposes.
Administrative Data Research Centres (ADRC)	Centres in each of four countries (England, Scotland, Wales, Northern Ireland) that facilitate data linkage, develop data linkage methodology and provide expertise on public engagement, ethics, information governance and law.
Community Health Index (CHI) number	A 10-digit number used to identify patients treated by NHS Scotland, which appears in many medical data sets in Scotland.
Cross-sectional data	Data recorded at one point in time, with no follow-up to repeat measurements on the same participants.
Data linkage	Joining two records together that belong to the same individual. Internal data linkage links records from the same person over time (e.g. repeated hospital admissions). External data linkage links records to another data set (e.g. participant in a cohort study to cancer registration data).
Data linkage error	Following an attempt to link data, the production of false matches (two people are assigned the same ID) or missed matches (the same person is assigned more than one ID).
Electronic Data Research and Innovation Service (eDRIS)	A service to support researchers who want to conduct health research involving Scottish patients or Scottish health data.
Farr Institute at Scotland	A collaborative network of Universities and NHS services in Scotland to support health informatics research in Scotland.
Health informatics	The application of information technology to healthcare, epidemiology and public health.
Hierarchy of evidence	A hierarchy considered useful for comparing the relative quality of evidence produced by different research designs, ranging from systematic reviews of randomised controlled trials (best) to case reports and anecdote (worst).
Identifiers	Characteristics that identify individuals in a data set (e.g. date of birth, sex, NHS number, postcode) and which can be used to link data sets together.
Information Services Division (ISD), NHS Scotland
Longitudinal data	Data in which the characteristics of individuals are measures more than once, or where individuals are followed over time to see what happens to them. Prospective cohort studies, epidemiological surveillance studies and repeated measures designs are all considered longitudinal.
National Health Service (NHS) number	A 10-digit number used to identify patients treated by the NHS across the UK. In theory, it is a unique identifier so that one NHS number applies to one patient.
National Records of Scotland (NRS)	Responsible for registration (e.g. births, marriages and civil partnerships, deaths) and statistical functions of the Registrar General for Scotland, including demography and the census.
Population spine	A new data set created by cross-referencing other data sets, theoretically allowing an accurate count of the entire population.
Safe haven	A secure environment in which data are linked and accessed.
Scottish Informatics and Linkage Collaboration (SILC)	A collaboration between many academic and public bodies to ensure that Scotland realises the benefits that can be derived through the legal, ethical and carefully controlled use of administrative, survey and other types of data.
ScottisH Informatics Programme (SHIP)	A precursor to the Farr Institute at Scotland.

Acknowledgements

GH-J would like to thank Gerald Donnelly and Ian Deary for their support and guidance, and staff at the Centre for Cognitive Ageing and Cognitive Epidemiology (CCACE), University of Edinburgh for hosting the author during his visit (particularly Caroline Brett). He would also like to thank David Campbell and Susan Carsley at National Records of Scotland for helpful comments on earlier versions of this review.

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: This review was originally undertaken as paid consultancy by GH-J for the Scottish Government and University of Edinburgh. It has not been published elsewhere in any form. Any views expressed are the authors' own.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Supplementary material

A Supplementary Table is available online at .

References

Hagger-Johnson

. Introduction to research methods and data analysis in the health sciences, Abingdon, Oxon: Routledge, 2014.

Moore C, Amin J, Gidding H, et al. A new method for assessing how sensitivity and specificity of linkage studies affects estimation. PloS One 2014; 9(7).

Hagger-Johnson

Harron

Gonzalez-Izquierdo

. Identifying possible false matches in anonymized hospital administrative data without patient identifiers. Health Serv Res 2014, pp. 1–17.

Scottish Government. A data linkage framework for Scotland 2014. http://www.scotland.gov.uk/Topics/Statistics/datalinkageframework (accessed 23 April 2014).

Taylor

Lynch

. Linking social care, housing & health data: data Linkage literature review 2010, Edinburgh: Scottish Government, 2010.

Public Records (Scotland) Act 2011: 2011 asp 12, 2011.

Scottish Government. Data sharing and linkage service 2014. http://www.scotland.gov.uk/Topics/Statistics/datalinkageframework/datasharingandlinkageservice (accessed 23 April 2014).

Russ

Kivimäki

Starr

. Height in relation to dementia death: individual participant meta-analysis of 18 UK prospective cohort studies. Br J Psychiatry 2014; 205(5): 348–54.

Harron

Wade

Gilbert

. Evaluating bias due to data linkage error in electronic healthcare records. BMC Med Res Methodol 2014; 14(1): 36.

10.

Nishiwaki Y, Clark H, Morton S, et al. Early life factors, childhood cognition and postal questionnaire response rate in middle age: the Aberdeen children of the 1950s study. BMC Med Res Methodol 2005; 5(1).

11.

Heckman

. Sample selection bias as a specification error. Econometrica 1979; 47(1): 153–161.

12.

Chandola

Bartley

Sacker

. Health selection in the Whitehall II study, UK. Soc Sci Med 2003; 56(10): 2059–2072.

13.

Park

. The healthy worker survivor effect and mortality at two automotive engine manufacturing plants. Am J Ind Med 1996; 30(6): 655–663.

14.

Wills

Lawlor

Matthews

. Life course trajectories of systolic blood pressure using longitudinal data from eight UK cohorts. PLoS Med 2011; 8(6): e1000440.

15.

Brett

Deary

. Realising health data linkage from a researcher's perspective: following up the 6-day sample of the Scottish Mental Survey 1947. Longitudinal Life Course Stud 2014; 5(3): 283–298.

16.

Kuh

Pierce

Adams

. Cohort profile: updating the cohort profile for the MRC National Survey of Health and Development: a new clinic-based data collection for ageing research. Int J Epidemiol 2011; 40(1): e1–e9.

17.

Deary

Whiteman

Starr

. The impact of childhood intelligence on later life: following up the Scottish Mental Surveys of 1932 and 1947. J Pers Soc Psychol 2004; 86(1): 130–147.

18.

Leon

Lawlor

Clark

. Cohort profile: the Aberdeen children of the 1950s study. Int J Epidemiol 2006; 35(5): 549–552.

19.

Sweeting H, Bhaskar A and Hunt K. Positive associations between consumerism and tobacco and alcohol use in early adolescence: cross-sectional study. BMJ Open 2012; 2(5).

20.

Fowkes

Price

Stewart

. Aspirin for prevention of cardiovascular events in a general population screened for a low ankle brachial index: a randomized controlled trial. JAMA 2010; 303(9): 841–848.

21.

Taylor

Whiteman

Fowkes

. Five factor model personality traits and all-cause mortality in the Edinburgh Artery Study cohort. Psychosom Med 2009; 71(6): 631–641.

22.

UK Biobank. UK Biobank: protocol for a large-scale prospective epidemiological resource, Stockport: UK Biobank Coordinating Centre, 2011.

23.

Martin

Gunnell

Pemberton

. Cohort profile: The Boyd Orr cohort—an historical cohort study based on the 65 year follow-up of the Carnegie Survey of Diet and Health (1937–39). Int J Epidemiol 2005; 34(4): 742–749.

24.

Elliott

Shepherd

. Cohort profile: 1970 British Birth Cohort (BCS70). Int J Epidemiol 2006; 35(4): 836–843.

25.

Scarbrough

. The British election study and electoral research. Political Stud 2000; 48(3): 391–414.

26.

Rose

Buck

Johnston

. The British household panel study - a valuable new resource for geographical research. Area 1994; 26(4): 368–376.

27.

Lynn P. Sample design for understanding society. Understanding Society Working Paper Series, 2009; 1.

28.

Walker

Whincup

Shaper

. The British Regional Heart Study 1975–2004. Int J Epidemiol 2004; 33(6): 1185–1192.

29.

Smith

McVie

. Theory and method in the Edinburgh study of youth transitions and crime. Br J Criminol 2003; 43(1): 169–195.

30.

Maplethorpe

Chanfreau

Philo

. Families with children in Britain: findings from the 2008 Families and Children Study (FACS), London: DWP, 2010.

31.

Office for National Statistics S, Vital Statistics D. General Household Survey, 2005. UK Data Archive, 2010.

32.

Office for National Statistics S, Vital Statistics D. General Household Survey: Time Series Dataset, 1972–2004. UK Data Archive, 2009.

33.

Smith

Campbell

Linksted

. Cohort profile: Generation Scotland: Scottish Family Health Study (GS:SFHS). The study, its participants and their potential for genetic research on health and illness. Int J Epidemiol 2013; 42(3): 689–700.

34.

Egan

Kearns

Mason

. Protocol for a mixed methods study investigating the impact of investment in housing, regeneration and neighbourhood renewal on the health and wellbeing of residents: the GoWell programme. BMC Med Res Methodol 2010; 10(1): 41.

35.

Scottish Centre for Social R. Growing Up in Scotland: Sweeps 1 to 3, 2005–2008. UK Data Archive, 2010.

36.

Kvaavik

Batty

Ursin

. Influence of individual and combined health behaviors on total and cause-specific mortality in men and women the United Kingdom Health and Lifestyle Survey. Arch Intern Med 2010; 170(8): 711.

37.

Shipley

Der

Taylor

. Association between mortality and cognitive change over 7 years in a large representative sample of UK residents. Psychosom Med 2007; 69(7): 640–650.

38.

Starr

Deary

Macintyre

. Associations with successful ageing in the “Healthy Old People in Edinburgh” cohort: being well, fit and healthy. Aging Clin Exp Res 2003; 15(4): 336–342.

39.

Deary

MacLennan

Starr

. Is age kinder to the initially more able? Differential ageing of a verbal ability in the healthy old people in Edinburgh study. Intelligence 1998; 26(4): 357–375.

40.

ONS. Life Opportunities Survey Technical Report 2009/10, London: Office of National Statistics, 2010.

41.

Ogilvie

Mitchell

Mutrie

. Evaluating health effects of transport interventions. Am J Prev Med 2006; 31(2): 118–126.

42.

Ogilvie

Mitchell

Mutrie

. Perceived characteristics of the environment associated with active travel: development and testing of a new scale. Int J Behav Nutr Phys Act 2008; 5(1): 32.

43.

Hart

MacKinnon

Watt

. The midspan studies. Int J Epidemiol 2005; 34(1): 28–34.

44.

Centre for Longitudinal Studies. Millenium Cohort Study: technical report, London: CLS, 2003.

45.

Power

Elliott

. Cohort profile: 1958 British birth cohort (National Child Development Study). Int J Epidemiol 2006; 35(1): 34–41.

46.

Wild

Macleod

McKnight

. Impact of deprivation on cardiovascular risk factors in people with diabetes: an observational study. Diab Med 2008; 25(2): 194–199.

47.

Wilson J. Orcades, http://http://www.orcades.ed.ac.uk/orcades/index.html (2013, accessed 25 March 2015).

48.

Sattar

McConnachie

Shaper

. Can metabolic syndrome usefully predict cardiovascular disease and diabetes? Outcome data from two prospective studies. Lancet 2008; 371(9628): 1927–1935.

49.

Thomson

Morrison

Petticrew

. The health impacts of housing-led regeneration: a prospective controlled study. J Epidemiol Community Health 2007; 61(3): 211–214.

50.

Petticrew M, Kearns A, Mason P, et al. The SHARP study: a quantitative and qualitative evaluation of the short-term outcomes of housing and neighbourhood renewal. BMC Public Health 2009; 9: 415.

51.

Bhopal

Fischbacher

Povey

. Cohort profile: Scottish Health and Ethnicity Linkage Study of 4.65 million people exploring ethnic variations in disease in Scotland. Int J Epidemiol 2011; 40(5): 1168–1175.

52.

Dong

Erens

. Scottish Health Survey 1995, London: Department of Epidemiology and Public Health, UCL, 1997.

53.

Shaw

McMunn

Field

. Scottish Health Survey 1998, London: Department of Epidemiology and Public Health, UCL, 2000.

54.

Bromley

Sproston

Shelton

. Scottish Health Survey 2003, Edinburgh: Scottish Government, 2005.

55.

Bromley

Bradshow

Given

. Scottish Centre for Social Research. Scottish Health Survey 2008, Edinburgh: Scottish Government, 2009.

56.

Bromley

Given

Ormston

. Scottish Centre for Social Research. Scottish Health Survey 2009, Edinburgh: Scottish Government, 2010.

57.

Rutherford

Sharp

Bromley

. Scottish Centre for Social Research. Scottish Health Survey 2011, Edinburgh: Scottish Government, 2012.

58.

McEwan S, Daly F, Forbes C, et al. Vascular risk factor profiles in the first phase of the Scottish Heart and Arterial Risk Prevention (SHARP) Survey, 1991–1996. Int Angiol 2000; 19(3): 197–205.

59.

Cade

Burley

Greenwood

. The UK Women's Cohort Study: comparison of vegetarians, fish-eaters and meat-eaters. Public Health Nutr 2004; 7(7): 871–878.

60.

Benzeval M, Der G, Ellaway A, et al. Cohort Profile: West of Scotland Twenty-07 Study: Health in the Community. Int J Epidemiol 2009; 38(5): 1215–1223.