Cohort description: The Northern Ireland Longitudinal Study (NILS)

Northern Ireland has traditionally been well served with cross-sectional studies, including episodic surveys of poverty and social exclusion,1 and of health and social well-being.2 However, such studies are of limited use in either analysis of life-course transitions or in the separation of cause and effect, both important goals of current research strategies where renewed interest in equity and social exclusion is stimulating research into the effects of disadvantage on individuals over time. This is all the more urgent given the increased social and geographical mobility and greater fluidity in people’s lives. Such information can only be derived longitudinally, and there was a dearth of such large-scale general purpose studies in Northern Ireland. Information had not been included in the three British Birth Cohort studies of 1946, 1958 and 19703–5 and although other longitudinal studies, such as PRIME6 and Young Hearts,7 are available they were designed to answer research questions related to specific diseases. Although the British Household Panel Study8 has recently been extended to include Northern Ireland and there is now a Northern Ireland component to the Millennium Birth Cohort Study, neither is large enough to function as a general purpose longitudinal study to meet the general research or policy needs of Northern Ireland.9 In the early 2000s, a group of senior statisticians from the Northern Ireland Statistics and Research Agency (NISRA) and academics was convened to estimate the cost of a Northern Ireland Longitudinal Study (NILS), equivalent to others either already available in England and Wales10 or, at the time, in development in Scotland.11 The aim was to have a multi-cohort study that would fulfil a range of academic and policy-related purposes, with a sample size large enough to enable robust analysis of population sub-groups and of areas of policy relevance. Because the experience of the Scottish Longitudinal Study (SLS) team had shown that linkage to the 1991 census would incur considerable costs, especially if social class for earlier years had to be recoded to the NS-SEC classification12 used in the 2001 census, it was decided to start with the 2001 census. Funding for the establishment and maintenance of NILS (and its sister study NIMS, see below) was jointly secured in 2003 from the Department of Health, Social Services and Public Safety and the Research and Development Office of the then Health and Personal Social Services. All funding for the development and maintenance of NILS and NIMS now comes from the Health and Social Care Research and Development Division of the Public Health Agency (HSC R&D Division). NISRA helps to fund the NILS/NIMS project both through the provision of accommodation to house all aspects of the NILS/NIMS operation and staff to maintain and develop the databases and provide strategic management of the project. Both NILS and NIMS were launched in December 2006. The goal for NILS is to provide a mechanism for understanding the population health dynamics of Northern Ireland equivalent to that provided by the ONS-LS which has, for the past 30 years, fulfilled this function for England and Wales.10 The ONS-LS used 4 days of birth in a year to define an initial cohort of ∼1% of the population (500 000 people) from which the 1971 census information was extracted and to which routinely collected vital statistics data such as births, deaths, cancer registrations and migration data were linked on an annual basis. Using the same days of birth equivalent, cohorts were drawn from censuses of 1981, 1991 and 2001 and linked to the initial cohort as they became available. A similar study has recently been established in Scotland (SLS). This is based on 20 birthdays and currently includes 265 321 people (5.3% of the population), with initial linkages to the 1991 and 2001 censuses.11 Although the ONS-LS has been the basis of many academic and government papers,13 it was thought that, because of the distinct historic, demographic and social conditions of Northern Ireland, a locally implemented longitudinal study would be more directly applicable to these conditions. For example, Northern Ireland has traditionally had the highest fertility and the youngest age structure of any UK region. It is also one of the most deprived regions of the UK and the highest rates of both unemployment and long-term unemployment and lowest proportions of the working age population defined as economically active.14 Average gross weekly earnings in Northern Ireland are ∼15% less than in the rest of the UK and there is a heavier reliance on state benefits, especially means-tested benefits.15 Northern Ireland also experiences some of the highest levels of poor self-reported health in the UK.16 Finally, it has a need to monitor its unique social equality legislation,17 with particular emphasis on the religious divisions and 30 years of civil unrest known locally as ‘the Troubles’.18 NILS is a representative sample of the Northern Ireland population (estimated at 1.78 million in June 2008). If someone within the centralized Northern Ireland Health Card registration system has one of 104 pre-designated days of birth they are included in the sample and linked to the census data. In Northern Ireland, people are entitled to ‘free at the point of use’ health and social care, but they must be registered. Almost 100% of entitled people are contained in this Health Card registration system. This marks an important difference with the ONS-LS and SLS, where inclusion is defined using the census data and then linked to the respective health card registration systems to facilitate linkage to other data sources. Therefore, although the NILS uses date of birth, sex and location (home postcode) from the health card registration system, the majority of cohort attributes are derived from the census record (Figure 1). The health card registrations are arguably a more robust source for the basic demographic data as the census data are captured only once and in 2001 depended on accurate interpretation of electronically scanned census forms. NILS composition: relationship between health card registration and census records The NILS cohort of c508 000 represents ∼28% of health card registrations (Figure 2). An aim of the 2001 census was to produce detailed estimates of under-enumeration and to adjust the census counts for this undercount using a process termed the One Number Census19 approach. Therefore, an additional 4.8% of people were statistically imputed to produce an adjusted total on which the final census results were based. For NILS, this equates to c23 500 people estimated to have been alive and resident in the country but not enumerated in the census. Finally, although the health card registration system contains almost 100% of the population it also includes some records that cannot be matched to the census. These relate primarily to people who have emigrated but not informed the health service, cross-border workers, as well as a small number of duplicate entries. This ‘list inflation’ is estimated at 4.7%, or approximately 24 000 records of people who ought not to be on the register and therefore cannot have a 2001 census record match. The initial linkage between census and health card registration records was based on an exact electronic match of forename, surname, gender, date of birth and location. For those remaining unlinked records, the matching process was relaxed to include fuzzy matching techniques with clerical assessment and confirmation, and finally, where this proved unsuccessful, wholly manual detailed searching and manual processes were undertaken. Structure of the NILS and linkage of contextual and event data sets The second study, the Northern Ireland Mortality Study (NIMS), is census based and comprises the whole of the 2001 census-enumerated population (1 603 641 people), to which subsequently registered deaths among residents are linked using a sequence of processes similar to that outlined above for use with the NILS.20 NIMS is recommended to researchers whose primary interest is mortality, where the larger cohort size enables more robust examination of sub-populations and of specific causes of death. Table 1 compares the NILS and NIMS databases with the 2001 census returns. This shows the capacity of the data sets to represent even relatively small sub-populations and demonstrates that both the NILS and NIMS cohorts are representative of official census outputs. The census returns include imputed persons and therefore have a slightly greater proportion of younger adults, males and people from larger conurbations than in NILS and NIMS. Numbers and percentage of NILS and NIMS members and the 2001 census counts aOther category relates to unassigned responses including those residing in communal establishments. Numbers and percentage of NILS and NIMS members and the 2001 census counts aOther category relates to unassigned responses including those residing in communal establishments. Because with NILS the major linkages are across censuses, the first ‘follow-up’ point will be when the 2011 NILS census cohort is linked. However, a partial linkage with the 1991 census data has been completed; this is classed as partial because it is based on exact electronic matching only. After matching, no personal identifiers that would allow identification of cohort members are included on the research data sets. Vital events such as births and deaths are updated regularly from routine registration data. Linkage to health and social care data is also possible (see below). Table 2 describes the cohort attributes and the outcomes measured. The census is a rich source of information and includes data on demographic and socio-economic characteristics, employment status and housing tenure. It also includes two indicators of self-reported health status—limiting long-term illness and general health in the year preceding the census. As the census form is based on households NILS can also provide contextual information at this level including: household composition and family structure; the relationship of the cohort member to other household members; access to household amenities; presence of a carer; and number of dependents in the household. Only data for the cohort member are linked over time. One unique household characteristic linked to the NILS database is the rateable value of the residence of the cohort member. This had been derived previously, in 2005, as part of an exercise to determine the level of local tax payable by each household. In addition, NILS receives information on change of address and date of move from the health card registration system, providing data on migration between census dates. Currently, for NILS sample members, these data provide migration information (from 2001 to 2010) on immigration into the sample, emigration out of and re-entries into Northern Ireland and internal migration within Northern Ireland. Data sets currently held in the NILS Age, sex and marital status Religion and community background Family, household or communal establishment type Housing, including tenure, rooms and amenities Country of birth, ethnicity Educational qualifications Economic activity, occupation and social class Migration (between 2000 and 2001) Limiting, long-term illness, self-reported general health, caregiving Travel to work New births into the sample Births to sample mothers and fathers Stillbirths to sample mothers (on demand) Infant mortality of children of sample mothers and fathers (on demand) Deaths of sample members Marriages 2004–06 Widow(er)hoods 2004–06 Age, sex and marital status Religion and community background Family, household or communal establishment type Housing, including tenure, rooms and amenities Country of birth, ethnicity Educational qualifications Economic activity, occupation and social class Migration (between 2000 and 2001) Limiting, long-term illness, self-reported general health, caregiving Travel to work New births into the sample Births to sample mothers and fathers Stillbirths to sample mothers (on demand) Infant mortality of children of sample mothers and fathers (on demand) Deaths of sample members Marriages 2004–06 Widow(er)hoods 2004–06 Capital and rating value (based on 2005 valuation exercise) Household characteristics (no. of rooms, property type, floor space, central heating) Immigrants into the sample Emigration out of Northern Ireland of sample members Re-entries into Northern Ireland after previous emigrations of sample members Migration within Northern Ireland of sample members Capital and rating value (based on 2005 valuation exercise) Household characteristics (no. of rooms, property type, floor space, central heating) Immigrants into the sample Emigration out of Northern Ireland of sample members Re-entries into Northern Ireland after previous emigrations of sample members Migration within Northern Ireland of sample members aThese data are available for NILS members and for all members of households with at least one NILS member. Data sets currently held in the NILS Age, sex and marital status Religion and community background Family, household or communal establishment type Housing, including tenure, rooms and amenities Country of birth, ethnicity Educational qualifications Economic activity, occupation and social class Migration (between 2000 and 2001) Limiting, long-term illness, self-reported general health, caregiving Travel to work New births into the sample Births to sample mothers and fathers Stillbirths to sample mothers (on demand) Infant mortality of children of sample mothers and fathers (on demand) Deaths of sample members Marriages 2004–06 Widow(er)hoods 2004–06 Age, sex and marital status Religion and community background Family, household or communal establishment type Housing, including tenure, rooms and amenities Country of birth, ethnicity Educational qualifications Economic activity, occupation and social class Migration (between 2000 and 2001) Limiting, long-term illness, self-reported general health, caregiving Travel to work New births into the sample Births to sample mothers and fathers Stillbirths to sample mothers (on demand) Infant mortality of children of sample mothers and fathers (on demand) Deaths of sample members Marriages 2004–06 Widow(er)hoods 2004–06 Capital and rating value (based on 2005 valuation exercise) Household characteristics (no. of rooms, property type, floor space, central heating) Immigrants into the sample Emigration out of Northern Ireland of sample members Re-entries into Northern Ireland after previous emigrations of sample members Migration within Northern Ireland of sample members Capital and rating value (based on 2005 valuation exercise) Household characteristics (no. of rooms, property type, floor space, central heating) Immigrants into the sample Emigration out of Northern Ireland of sample members Re-entries into Northern Ireland after previous emigrations of sample members Migration within Northern Ireland of sample members aThese data are available for NILS members and for all members of households with at least one NILS member. The lowest level of geography normally available for analysis of NILS/NIMS data is Super Output Area (average size of approximately 2000 persons), allowing a wide array of area-level characteristics to be added as contextual data, for example an indicator of urban/rural residence;21 relative deprivation;22 modelled indicator of relative poverty and income inequality;23 or others supplied by the researcher. Current routine NILS linkages to vital events data include births to NILS members (mothers and fathers), births of NILS members (babies) and deaths of NILS members. Births data are currently available from 1997 to 2007, whereas mortality data are available from census day 2001 to 2007. Other events, such as deaths among NILS members in infancy, infant deaths to NILS members and stillbirths to NILS members are also available (though this latter is not routinely linked at present). Marriage data sets have been linked for the period 2004–06 and there is potential to link to more updated marriage data based on research demand. Widow(er)hood data sets have also been linked for the period 2001–06; these linkages can also be routinely updated depending on researchers’ needs. Ethical approval has been pre-obtained for all research which utilizes these regular linkages providing the research proposal itself has obtained approval from the NILS Research Approvals Group. These data sets will not include information about the health condition, practitioners or treatments of members. With NILS the inclusion of the Health and Care number facilitates linkage to health and social care and data the electronic of and of Northern Ireland, the other of the has health and social the same unique Health and Care it is possible to link data to the care of people in the community or on to care data from to the are held by the Northern Ireland and can also be linked to to this type of data not routinely as are Linkage and each must have specific approval from the Office of Research for Northern Ireland With these a process of is on the of each data to and of the data. This is on a project by project basis. Northern Ireland has a relatively but in the from 2000 to this has to an of approximately 000 persons a range of to Because of the of the NILS databases this is in the NILS sample at However, these additional cohort members will not have any census contextual data the linkage of the 2011 census. As with its sister UK longitudinal studies, linkage to routine databases no is not a The majority of are of studies using NIMS, this is to change as NILS One of the studies used a unique by the NIMS linkage of all deaths to Northern Ireland to study the characteristics of people whose registration events not be linked to the enumerated In other longitudinal studies this is estimated by mortality rates for the country as a In the NIMS data although all deaths can be the linkage between mortality data and the census was with other mortality linkage studies as New and The characteristics of the unlinked deaths records related primarily to at the census and to in the information on the census form and which These are to be applicable to all other longitudinal The of NIMS more robust of population than is possible using other UK longitudinal For example, the one of research has on health outcomes with the religious in Northern one study mortality and that those who be as from a more mortality when with other religious especially from related to and NIMS also facilitates study of less causes of such as and mortality related to and on either cause of death. One of the unique of NILS is the to study migration In although researchers using longitudinal studies have to for the census to determine change of address and but because NILS is the health card registration system, to registration are regularly with NILS providing information on population the Although such data are known to provide limited information on migration they are robust for internal with some address are registered NILS has used these data to study the relationship between migration and by to of registered the of to these The inclusion of the unique Health and Care number in NILS enables the potential linkage of the census to health care data for distinct two Linkage have been one at the and area-level related to of and a second in registrations for care and of among younger A of and and in of and is on the NILS (see below). The primary of NILS is that it is large and It is also as in the presence of the Health and Care number a for potential linkage of health Finally, other studies NILS includes residents in care and it has been used to mortality among residents in care and mortality between those in and NILS, has the potential to study a range of outcomes of more to policy than the vital events such as mortality with such NILS is similar in structure to the ONS-LS and given the in census questions across the this that three equivalent longitudinal studies now the whole of the allowing the potential for and A study data sets from NILS with equivalent from its sister studies to in the relationship between self-reported health and mortality across the UK that are of difference in the and of The census, which is the basis for of the and of NILS, is also the source of some of its of the Northern Ireland population was not enumerated in the 2001 census and this is thought to younger males from deprived For those who are not enumerated vital events and other outcomes cannot be linked to a census record and this be by data matching which is when there is no unique as in with a population The cohort attributes are on the and of the census The two self-reported health of mortality and health are especially with to of be to some by an array of questions to be included in the Northern Ireland 2011 census. UK censuses not information on such as as is the in New of income is with NILS this is for to some by the inclusion of the value of the household residence which can be used as a of socio-economic A with NILS and similar studies is that they with a census, a cross-sectional study from in its initial it is to cause from for health effects was a by the researchers using the ONS-LS to the relationship between unemployment and but it has also with NIMS when the mortality with Both NILS and NIMS data and care is to and are by NISRA census and access is and by and that data All NILS research project data sets are held on a access to which is on which are not to any and with all to this is possible only from within the in NISRA in central to which must be are at all Only the small group of individuals for the databases have access to the days of birth used for to Data sets are at the point of use and all electronic and are to they not example, with counts of less than are not to All aspects of NILS are by a and by a Research Approvals who access to the data. All researchers with a NILS project must a of to with and to data and using the data. a NILS Research is available to through these processes and to provide and to of the data. to use NILS or NIMS data the who will in a and an form for to the Research Approvals Group. a project has been will a project data and with analysis and interpretation of Although researchers are to out it is possible for researchers to out these by to be on data, with the results and to in NILS or NIMS where they can and more detailed information about the data sets and a through the potential researchers can the at The is by the of and the and Social of the Office of the and for Northern Ireland The NILS and the NIMS are by both the Health and Social Care Research and Development Division of the Public Health Agency and The provided by the Northern Ireland Statistics and Research Agency (NISRA) staff within the NILS is as is the from the and the of Services at the Services The are for the included of