Administrative data is the term used to describe everyday data about individuals collected by government departments and agencies. Examples include exam results, benefit receipt and National Insurance payments.
Age effects relates to changes in an outcome as a result of getting older.
Anonymisation refers to the removal of study participants’ identifying information (e.g., name, address) in order to preserve their privacy.
Attrition is the discontinued participation of study participants in a longitudinal study. Attrition can reflect a range of factors, from the study participant not being traceable to them choosing not to take part when contacted. Attrition is problematic both because it can lead to bias in the study findings (if the attrition is higher among some groups than others) and because it reduces the size of the sample.
Baseline refers to the start of a study when initial information is collected on participation (however, in longitudinal studies, researchers may adopt an alternative ‘baseline’ for the purposes of analysis).
Biological samples is the term used for specimens collected from human subjects from which biological information, such as genetic markers, can be extracted for analysis. Common examples include blood, saliva or hair.
Body mass index is a measure used to assess if an individual is a healthy weight for their height. It is calculated by dividing the individual’s weight by the square of their height, and it is typically represented in units of kg/m2.
Boosted samples are used to overcome sample bias due to attrition or to supplement the representation of smaller sub-groups within the sample. Inclusion of boosted samples must be accompanied by appropriate survey weights.
Computer-assisted personal interviewing (CAPI) is a technique for collecting data from participants using computers to eliminate common errors such as questionnaire routing and data entry mistakes. The use of computers take place within the context of a face-to-face interview.
Computer-assisted self-interviewing (CASI) is a technique for collecting data from participants using computers to eliminate common errors such as questionnaire routing and data entry mistakes. The use of computers take place within the context of a self-completion questionnaire.
A categorical variable is a variable that can take one of a limited number of discrete values. They can be either nominal – they contain no inherent order of categories (e.g. sex; marital status) – or ordinal – they can be ranked in some meaningful order (e.g. level of satisfaction with a service).
Computer-assisted telephone interviewing (CATI) is a technique for collecting data from participants using computers to eliminate common errors such as questionnaire routing and data entry mistakes. The use of computers take place within the context of a telephone interview.
For some study participants the exact time of an event will not be known because either: the study ends (or the analysis is carried out) before they have had the event, or the participant drops out of the study before experiencing the event. It is therefore, only known that the event has not occurred up to the time that they were last observed in the study.
Census refers to a universal and systematic collection of data from all individuals within a population. In the UK, the government conducts a census every ten years with the next one due in 2021.
A codebook is a document (online or hard-copy) that contains all the information about how a dataset has been coded, such that it can be deciphered by a researcher not familiar with the original coding frame.
Coding is the process of converting survey responses into numerical codes to facilitate data analysis. All potential responses (as well as possible reasons for non-response) for each variable are assigned numerical values according to a coding frame.
Cognitive assessments are exercises used to measure thinking abilities, such as memory, reasoning and language. Longitudinal studies collecting data in this way can track the extent to which someone’s cognitive abilities change (develop or decline) over time.
Cohort studies are concerned with charting the lives of groups of individuals who experience the same life events within a given time period. The best known examples are birth cohort studies, which follow a group of people born in a particular period.
Complete case analysis is the term used to describe a statistical analysis that only includes participants for which we have no missing data on the variables of interest. Participants with any missing data are excluded.
Conditioning refers to the process whereby participants’ answers to some questions may be influenced by their participation in the study – in other words, their responses are ‘conditioned’ by their being members of a longitudinal study. Examples would include study respondents answering questions differently or even behaving differently as a result of their participation in the study.
Confounding occurs where the relationship between independent and dependent variables is distorted by one or more additional, and sometimes unmeasured, variables. A confounding variable must be associated with both the independent and dependent variables but must not be an intermediate step in the relationship between the two (i.e. not on the causal pathway).
For example, we know that physical exercise (an independent variable) can reduce a person’s risk of cardiovascular disease (a dependent variable). We can say that age is a confounder of that relationship as it is associated with, but not caused by, physical activity and is also associated with coronary health. See also ‘unobserved heterogeneity’, below.
A continuous variable is a variable that has an infinite number of uncountable values e.g. time, temperature. They are also known as quantitative variables or scale variables.
Cohort effects relates to changes in an outcome associated with being a member of a specific cohort of people (e.g. born in the same year; or starting school at the same time).
In metadata management, coverage refers to the temporal, spatial and topical aspects of the data collection to describe the comprehensiveness of a dataset. For longitudinal studies, this can relate to the topics that are covered across waves, the population to which one can generalise or the geographic extent of the dataset.
Cross-sectional surveys involve interviewing a fresh sample of people each time they are carried out. Some cross-sectional studies are repeated regularly and can include a large number of repeat questions (questions asked on each survey round).
Within the context of data protection, a data access agreement specifies the terms under which users are provided access to specified datasets. This usually forms part of the application process to the data controller to ensure that researchers adhere to a set of terms regarding data confidentiality, sensitivity and dissemination before accessing the data. See also: research ethics
Data cleaning is an important preliminary step in the data analysis process and involves preparing a dataset so that it can be correctly analysed. ‘Cleaning’ the data usually involves identifying data input errors, assessing the completeness of the dataset and verifying any anomalies (e.g. outliers).
Within the context of data protection, data confidentiality is the process of protecting participants’ data from being accessed or disclosed by those unauthorised to do so. Key methods employed in data confidentiality include anonymisation of responses (removal of personal identifying information) and data encryption (protecting the data using codes and/or passwords).
Data harmonisation involves retrospectively adjusting data collected by different surveys to make it possible to compare the data that was collected. This enables researchers to make comparisons both within and across studies. Repeating the same longitudinal analysis across a number of studies allows researchers to test whether results are consistent across studies, or differ in response to changing social conditions.
Data imputation is a technique for replacing missing data with an alternative estimate. There are a number of different approaches, including mean substitution and model-based multivariate approaches.
Data linkage simply means connecting two or more sources of administrative, educational, geographic, health or survey data relating to the same individual for research and statistical purposes. For example, linking housing or income data to exam results data could be used to investigate the impact of socioeconomic factors on educational outcomes.
Data protection refers to the broad suite of rules governing the handling and access of information about people. Data protection principles include confidentiality of responses, informed consent of participants and security of data access. These principles are legally protected by the Data Protection Act (DPA) and the General Data Protection Regulation (GDPR).
Data structure refers to the way in which data are organised and formatting in advance of data analysis.
In analysis, the dependent variable is the variable you expect to change in response to different values of your independent (or predictor) variables. For example, a students’ test results may be (partially) explained by the number of hours spent on revision. In this case, the dependent variable is students’ test score, which you expect to be different according to the amount of time spent revising.
A derived variable is a variable that is calculated from the values of other variables and not asked directly of the participants. It can involve a mathematical calculation (e.g. deriving monthly income from annual income by dividing by 12) or a recategorisation of one or more existing variables (e.g. categorising monthly income into £500 bands – £0 to £500, £501 to £1,000, etc.)
Diaries are a data collection instrument that is particularly useful in recording information about time use or other regular activity, such as food intake. They have the benefit of collecting data from participants as and when an activity occurs. As such, they can minimise recall bias and provide a more accurate record of activities than a retrospective interview.
Dissemination is the process of sharing information – particularly research findings – to other researchers, stakeholders, policy makers, and practitioners through various avenues and channels, including online, written publications and events. Dissemination is a planned process that involves consideration of target audiences in ways that will facilitate research uptake in decision-making processes and practice.
Dummy variables, also called indicator variables, are sets of dichotomous (two-category) variables we create to enable subgroup comparisons when we are analysing a categorical variable with three or more categories.
Empirical data refers to data collected through observation or experimentation. Analysis of empirical data can provide evidence for how a theory or assumption works in practice.
In metadata management, fields are the elements of a database which describes the attributes of items of data.
General ability is a term used to describe cognitive ability, and is sometimes used as a proxy for intelligent quotient (IQ) scores.
Growth curve modelling is used to analyse trajectories of longitudinal change over time allowing us to model the way participants change over time, and then to explore what characteristics or circumstances influence these patterns of longitudinal change.
Hazard rate refers to the probability that an event of interest occurs at a given time point, given that it has not occurred before.
Health assessments refers to the assessments carried out on research participants in relation to their physical characteristics or function. These can include measurements of height and weight, blood pressure or lung function.
Heterogeneity is a term that refers to differences, most commonly differences in characteristics between study participants or samples. It is the opposite of homogeneity, which is the term used when participants share the same characteristics. Where there are differences between study designs, this is sometimes referred to as methodological heterogeneity. Both participant or methodological differences can cause divergences between the findings of individual studies and if these are greater than chance alone, we call this statistical heterogeneity. See also: unobserved heterogeneity.
Household panel surveys collect information about the whole household at each wave of data collection, to allow individuals to be viewed in the context of their overall household. To remain representative of the population of households as a whole, studies will typically have rules governing how new entrants to the household are added to the study.
As a way of encouraging participants to take part in research, they may be offered an incentive or a reward. These may be monetary or, more commonly, non-monetary vouchers or tokens. Incentives are advertised beforehand and can act as an aid to recruitment; rewards are a token of gratitude to the participants for giving their time.
In analysis, an independent variable is any factor that may be associated with an outcome or dependent variable. For example, the number of hours a student spends on revision may influence their test result. In this case, the independent variable, revision time (at least partially) ‘explains’ the outcome of the test.
A key principle of research ethics, informed consent refers to the process of providing full details of the research to participants so that they are sufficiently able to choose whether or not to consent to taking part.
Kurtosis is sometimes described as a measure of ‘tailedness’. It is a characteristic of the distribution of observations on a variable and denotes the heaviness of the distribution’s tails. To put it another way, it is a measure of how thin or fat the lower and upper ends of a distribution are.
A person’s life course refers to the experiences and stages an individual passes through during their life. It centres on the individual and emphasises the changing social and contextual processes that influence their life over time.
Longitudinal studies gather data about the same individuals (‘study participants’) repeatedly over a period of time, in some cases from birth until old age. Many longitudinal studies focus upon individuals, but some look at whole households or organisations.
Metadata refers to data about data, which provides the contextual information that allows you to interpret what data mean.
Missing data refers to values that are missing and do not appear in a dataset. This may be due to item non-response, participant drop-out (or attrition) or, in longitudinal studies, some data (e.g. date of birth) may be collected only in some waves. Large amounts of missing data can be a problem and lead researchers to make erroneous inferences from their analysis. There are several ways to deal with the issue of missing data, from casewise deletion to complex multiple imputation models.
Multi-level modelling refers to statistical techniques used to analyse data that is structured in a hierarchical or nested way. For example. study participants who belong to the same household, or students who attend the same school may be expected to be more similar to each other than to participants in other households or schools (such as sharing similar contextual influences). This similarity means that the data from participants within these households/schools are not independent. Multi-level models can account for variability at both the individual level and the group (e.g. household or school) level.
Non-response bias is a type of bias introduced when those who participate in a study differ to those who do not in a way that is not random (for example, if attrition rates are particularly high among certain sub-groups). Non-random attrition over time can mean that the sample no longer remains representative of the original population being studied. Two approaches are typically adopted to deal with this type of missing data: weighting survey responses to re-balance the sample, and imputing values for the missing information.
Observational studies focus on observing the characteristics of a particular sample without attempting to influence any aspects of the participants’ lives. They can be contrasted with experimental studies, which apply a specific ‘treatment’ to some participants in order to understand its effect.
Panel studies follow the same individuals over time. They vary considerably in scope and scale. Examples include online opinion panels and short-term studies whereby people are followed up once or twice after an initial interview.
Peer review is a method of quality control in the process of academic publishing, whereby research is appraised (usually anonymously) by one or more independent academic with expertise in the subject.
Period effects relate to changes in an outcome associated with living during a particular time, regardless of age or cohort membership (e.g. living through times of war, economic recession or global pandemic).
Piloting is the process of testing your research instruments and procedures to identify potential problems or issues before implementing them in the full study. A pilot study is usually conducted on a small subset of eligible participants who are encouraged to provide feedback on the length, comprehensibility and format of the process and to highlight any other potential issues.
Population refers to all the people of interest to the study and to whom the findings will be able to be generalized (e.g. a study looking into rates of recidivism may have a [target] population encompassing everyone with a criminal conviction). Owing to the size of the population, a study will usually select a sample from which to make inferences. See also: sample, representiveness.
A percentile is a measure that allows us to explore the distribution of data on a variable. It denotes the percentage of individuals or observations that fall below a specified value on a variable. The value that splits the number of observations evenly, i.e. 50% of the observations on a variable fall below this value and 50% above, is called the 50th percentile or more commonly, the median.
Primary research refers to original research undertaken by researchers collecting new data. It has the benefit that researchers can design the study to answer specific questions and hypotheses rather than relying on data collected for similar but not necessarily identical purposes. See also: secondary research
In prospective studies, individuals are followed over time and data about them is collected as their characteristics or circumstances change.
Qualitative data are non-numeric – typically textual, audio or visual. Qualitative data are collected through interviews, focus groups or participant observation. Qualitative data are often analysed thematically to identify patterns of behaviour and attitudes that may be highly context-specific.
Quantitative data can be counted, measured and expressed numerically. They are collected through measurement or by administering structured questionnaires. Quantitative data can be analysed using statistical techniques to test hypotheses and make inferences to a population.
Questionnaires are research instruments used to elicit information from participants in a structured way. They might be administered by an interviewer (either face-to-face or over the phone), or completed by the participants on their own (either online or using a paper questionnaire. Questions can cover a wide range of topics and often include previously-validated instruments and scales (e.g. the Rosenberg Self-Esteem Scale).
Recall error or bias describes the errors that can occur when study participants are asked to recall events or experiences from the past. It can take a number of forms – participants might completely forget something happened, or misremember aspects of it, such as when it happened, how long it lasted, or other details. Certain questions are more susceptible to recall bias than others. For example, it is usually easy for a person to accurately recall the date they got married, but it is much harder to accurately recall how much they earned in a particular job, or how their mood at a particular time.
Record linkage studies involve linking together administrative records (for example, benefit receipts or census records) for the same individuals over time.
A reference group is a category on a categorical variable to which we compare other values. It is a term that is commonly used in the context of regression analyses in which categorical variables are being modelled.
Regression analysis refers to analytical techniques that use a mathematical ‘model’ to predict values of a dependent variable from values of one or many independent variable(s).
Repeated measures are measurements of the same variable at multiple time points on the same participants, allowing researchers to study change over time.
Representativeness is the extent to which a sample is representative of the population from which it is selected. Representative samples can be achieved through, for example, random sampling, systematic sampling, stratified sampling or cluster sampling.
Research ethics relates to the fundamental codes of practice associated with conducting research. Ethical issues that need to be considered include providing informed consent to participants, non-disclosure of sensitive information, confidentiality and anonymity safeguarding of vulnerable groups, and respect for participants’ well-being. Academic research proposals need be approved by an ethics committee before any actual research (either primary or secondary) can begin.
Research impact is the demonstrable contribution that research makes to society and the economy that can be realised through engagement with other researchers and academics, policy makers, stakeholders and members of the general public. It includes influencing policy development, improving practice or service provision, or advancing skills and techniques.
Residuals are the difference between your observed values (the constant and predictors in the model) and expected values (the error), i.e. the distance of the actual value from the estimated value on the regression line.
Respondent burden is a catch all phrase that describes the perceived burden faced by participants as a result of their being involved in a study. It could include time spent taking part in the interview and inconvenience this may cause, as well as any difficulties faced as a result of the content of the interview.
Response rate refers to the proportion of participants in the target sample who completed the survey. Longitudinal surveys are designed with the expectation that response rates will decline over time so will typically seek to recruit a large initial sample in order to compensate for likely attrition of participants.
In retrospective studies, individuals are sampled and information is collected about their past. This might be through interviews in which participants are asked to recall important events, or by identifying relevant administrative data to fill in information on past events and circumstances.
Sample is a subset of a population that is used to represent the population as a whole. This reflects the fact that it is often not practical or necessary to survey every member of a particular population. In the case of birth cohort studies, the larger ‘population’ from which the sample is drawn comprises those born in a particular period. In the case of a household panel study like Understanding Society, the larger population from which the sample was drawn comprised all residential addresses in the UK.
Sample size refers to the number of data units contained within a dataset. It most frequently refers to the number of respondents who took part in your study and for whom there is usable data. However, it could also relate to households, countries or other institutions. The size of a sample, relative to the size of the population, will have consequences for analysis: the larger a sample is, the smaller the margin of error of its estimates, the more reliable the results of the analysis and the greater statistical power of the study.
A sampling frame is a list of the target population from which potential study participants can be selected.
Scales are frequently used as part of a research instrument seeking to measure specific concepts in a uniform and replicable way. Typically, they are composed of multiple items that are aggregated into one or more composite scores. Examples of standardised scales include the British Ability Scale (BAS); the Malaise Inventory; and the Rosenberg Self-Esteem Scale.
A scatterplot is a way of visualising the relationship between two continuous variables by plotting the value of each associated with a single case on a set of X-Y coordinates. For example, students’ test scores in English and maths can be represented as point on a graph, with each point representing a single student’s English (x-axis) and maths (y-axis) score. Looking at data for many students allows us to build up a visualisation of the relationship between students’ scores in maths and English.
Secondary research refers to new research undertaken using data previously collected by others. It has the benefit of being more cost-effective than primary research whilst still providing important insights into research questions under investigation.
Skewness is the measure of how assymetrical the distribution of observations are on a variable. If the distribution has a more pronounced/longer tail at the upper end of the distribution (right-hand side), we say that the distribution is negatively skewed. If it is more pronounced/longer at the lower end (left-hand side), we say that it is positively skewed.
A statistical model is a mathematical representation of the relationship between variables.
Statistical software packages are specifically designed to carry out statistical analysis; these can either be open-source (e.g. R) or available through institutional or individual subscription (e.g. SPSS; Stata).
Structured metadata define the relationship between data items/objects to enable computer systems to understand the contextual meaning of the data. It uses standardised content to facilitate the use of metadata for data discovery and sharing, and the relationship between metadata elements.
Study participants are the individuals who are interviewed as part of a longitudinal study.
Also called conditional routing (sometimes called ‘filters’), survey logic refers to the flow that takes respondents through a survey. Respondents may be required to answer some questions only if they had provided a relevant response to a previous question. E.g. Only respondents who are currently at university may be asked to answer a question relating to their degree subject. This is important when considering missing data.
Survey weights can be used to adjust a survey sample so it is representative of the survey population as a whole. They may be used to reduce the impact of attrition on the sample, or to correct for certain groups being over-sampled.
Survival analysis is an analytical technique that uses time-to-event data to statistically model the probability of experiencing an event by a given time point. For example, time to retirement, disease onset or length of periods of unemployment.
The term used to refer to a round of data collection in a particular longitudinal study (for example, the age 7 sweep of the National Child Development Study refers to the data collection that took place in 1965 when the participants were aged 7). Note that the term wave often has the same meaning.
The population of people that the study team wants to research, and from which a sample will be drawn.
Time to event refers to the duration of time (e.g. in hours, days, months, etc.) from a defined baseline to the time of occurrence of an event of interest (e.g. diagnosis of an illness, first re-offence following release from prison). Survival analysis can be used to analyse such data.
Tracing (or tracking) describes the process by which study teams attempt to locate participants who have moved from the address at which they were last interviewed.
Unobserved heterogeneity is a term that describes the existence of unmeasured (unobserved) differences between study participants or samples that are associated with the (observed) variables of interest. The existence of unobserved variables means that statistical findings based on the observed data may be incorrect.
Part of the documentation that is usually provided with statistical datasets, user guides are an invaluable resource for researchers. The guides contain information about the study, including the sample, data collection procedures, and data processing. Use guides may also provide information about how to analyse the data, whether there are missing data due to survey logic, and advice on how to analyse the data such the application of survey weights.
Variables is the term that tends to be used to describe data items within a dataset. So, for example, a questionnaire might collect information about a participant’s job (its title, whether it involves any supervision, the type of organisation they work for and so on). This information would then be coded using a code-frame and the results made available in the dataset in the form of a variable about occupation. In data analysis variables can be described as ‘dependent’ and ‘independent’, with the dependent variable being a particular outcome of interest (for example, high attainment at school) and the independent variables being the variables that might have a bearing on this outcome (for example, parental education, gender and so on).
Vulnerable groups refers to research participants who may be particularly susceptible to risk or harm as a result of the research process. Different groups might be considered vulnerable in different settings. The term can encompass children and minors, adults with learning difficulties, refugees, the elderly and infirm, economically disadvantaged people, or those in institutional care. Additional consideration and mitigation of potential risk is usually required before research is carried out with vulnerable groups.
The term used to refer to a round of data collection in a particular longitudinal study (for example, the age 7 wave of the National Child Development Study refers to the data collection that took place in 1965 when the participants were aged 7). Note that the term sweep often has the same meaning.
The quality of child-parent relationships and parental mental health in childhood are strongly associated with subsequent mental well-being in adulthood, even more so than childhood socioeconomic circumstances.
Researchers from University College London, University of Oxford, University of Edinburgh, and the Health Foundation analysed data from three British birth cohort studies to investigate how multiple aspects of the childhood environment were associated with subsequent mental well-being in adulthood across different generations. Using equivalent measures across multiple studies, the researchers were able to determine how childhood experiences compare across cohorts, and to explore how these experiences influence adult mental well-being.
Mental well-being is a key aspect of healthy ageing, and thus it is important to understand what specific factors promote positive adult well-being across the life-course.
The Medical Research Council (MRC) National Survey of Health and Development (NSHD) is the oldest and longest running of the British birth cohort studies; it is a nationally representative sample (N=5,362) of men and women born in England, Scotland or Wales in March 1946.
The 1958 National Child Development Study (NCDS) follows the lives of 17,415 people born in England, Scotland and Wales in a single week of 1958.
The 1970 British Cohort Study (BCS70) follows the lives of 17,198 people born in England, Scotland and Wales in a single week of 1970.
Well-being was measured at ages 60-64 years in NSHD, age 50 in NCDS, and age 40 in BCS70, using the self-completed Warwick-Edinburgh Mental Well-Being Scale (WEMWBS). This instrument comprises 14 items, measured on a five-point Likert scale, combining hedonic (happiness, pleasure, and pain avoidance) and eudemonic (self-realisation and positive functioning) aspects of well-being into a single factor. A total well-being score on the instrument ranges from 14-70, with a higher score denoting greater well-being.
Building upon previous analysis which explored individual childhood domains in isolation, the researchers drew from a number of bespoke questionnaire instruments in each study to measure multiple indicators of the childhood environment. Where there was variability within and between the studies in the methods of assessment, the researchers utilised retrospective data harmonisation techniques to foster better comparability. The different aspects of childhood environment examined were:
Family socioeconomic circumstances: across all studies, father’s occupational social class was measured at ages 10-11 and subsequently coded according to the Registrar General’s 1990 classification, ranging from I-professional to V-unskilled.
Parental education: the age parents left full-time continuous education, reported when cohort members were age 6 in NSHD, age 16 in NCDS and ages 5 and 16 in BCS70.
Teen parenthood: defined as parents being under 20 years old at cohort member’s birth.
Lack of amenities (including sole use of the bathroom or kitchen, and access to hot running water): measured at age 2 and 11 in NSHD, age 7 and 11 in NCDS, and age 5 and 10 years in BCS70.
Overcrowding: the average number of people per room exceeding 1, was measured at ages 2, 4, 6, 8 and 11 in NSHD, ages 7, 11, and 16 in NCDS, and age 5 years in BCS70.
Housing tenure: whether accommodation was rented or owned, was measured at ages 2 and 11 in NSHD, 7 and 11 in NCDS, and 5 and 10 years in NSC70.
Whether the cohort member was breastfed was collected at age 2 in NSHD, age 7 in NCDS, and age 5 in BCS70.
Parental care and overprotection subscales were derived from the Parental Bonding Instrument using factor analysis. In NSHD, this measure was administered retrospectively at age 43, included 24 items, and asked about each parent individually with average scores derived across parents. In BCS70, a shortened 11-item version of the instrument was administered prospectively at age 16 and its items asked about both parents together. For NCDS, a measurement of how well cohort members got on with either parent was collected at age 16.
Parental interest in child’s education was reported by the mother when cohort members were aged 11 years in NSHD, and by both parents in NCDS and BSC70 when children were aged 7 and 10 years in NCDS and BSC70, with data from the most interested parent used in analysis.
Experience of parental divorce before or up to the age of 16 years was measured in each study.
The number of residential moves was collected at age 15 years in NSHD, age 16 in NCDS, and age 10 in BCS70.
Experience of separation from mother over one night or longer was measured at age 4 in NSHD and age 7 in NCDS. Conversely, in BCS70, experience of long-term separation was measured at age 10, with responses categorised into ‘separated for less than one week or never’ and ‘separated for more than one week’.
A measure for parental chronic health conditions was collected at age 15 years in NSHD, age 11 in NCDS, and age 10 in BCS70.
Family mental health was measured at age 15 years in NSHD using the self-reported Maudsley Personality Inventory to determine maternal neuroticism, age 7 in NCDS using a measure of family mental health collected by a health visitor, and age 10 in BCS70 using the self-reported Malaise Inventory, to measure maternal depression.
Despite changes in social context experienced by the cohorts in childhood and cross-study differences in the measurement of most childhood variables, indicators of child rearing, parenting quality and parental health were associated with later adult mental well-being across all cohorts. Poorer quality parent-child relationships including low parental care and high parental overprotection (NSHD and BCS70) and not getting on with parents (NCDS) were associated with lower adult well-being. Poor maternal/ family mental health (NSHD and BCS70) and health problems for the father (NCDS) were also associated with lower adult well-being.
In contrast to the NSHD, where no such association was found, in NCDS and BSC70, a lack of parental interest in education was independently associated with lower levels of adult mental well-being. Moreover, in BCS70, where parental divorce rates were highest, this was additionally found to be independently associated with poorer mental well-being.
Having a father in a more disadvantaged social class was associated with lower well-being in all three cohorts. However, compared to psychosocial childhood indicators, when comparing standardised coefficients and effect sizes, the association between socioeconomic circumstances and adulthood mental well-being was smaller in magnitude.
Many of the determinants of well-being begin early in the life course, and using longitudinal data allows researchers to access a wide range of prospectively collected measures to explore how childhood circumstances may be associated with mental well-being later in adulthood. This is in contrast to studies that utilise retrospective measurements, which may be subject to issues of recall bias.
A challenge of working with longitudinal data is attrition, as cohort members who experienced adverse childhood socioeconomic conditions may be more likely to drop out of the studies. This could lead to underestimation of the association between childhood socioeconomic circumstances and adulthood mental well-being.
An advantage of using multiple longitudinal datasets is the opportunity to make cross-cohort comparisons, although it should be acknowledged that each cohort experienced a different childhood social context which may have influenced the associations being explored. For example, the NSHD cohort grew up in the post-war period which meant they experienced different housing conditions to NCDS and BCS70 cohort members. Additionally, divorce rates in the BCS70 cohort were higher, suggesting changes occurred in family structure between cohorts. Even so, researchers found that across all cohorts the indicators of the childhood environment included were associated with adult mental well-being.
In the present paper, the researchers found differences between the studies in how they captured the relevant indicators of the childhood environment. As a consequence, they needed to employ different techniques to foster a basis for cross-study comparison and co-analysis. Addressing such measurement heterogeneity can require a range of solutions, some more complex than others.
In the case of parental social class and educational attainment, a basis for harmonisation had been already established elsewhere. Where existing harmonisation protocols were unavailable, bespoke methods had to be employed. In some cases, measures with variable scale lengths or different numbers of categories were standardised to make responses more comparable.
For characteristics where the researchers found that overtly comparable indicators were not available, they relied on variables that tapped into similar underlying constructs. As an illustration, three different instruments of family mental health were used in the present paper: a measure for maternal neuroticism (NSHD), family mental health (NCDS), and maternal depression (BCS70).
Testing the assumptions inherent in such data preparation can be an important step in data harmonisation. Moreover, techniques for evaluating measurement equivalence in latent constructs (e.g., mental well-being) can help validate harmonisation decisions and the basis for the inferences drawn. That this was not done in the current paper is a limitation of this research.
A further challenge in such cross-study longitudinal research is that the age of assessment and frequency of follow-up can vary between studies. To address this in the present study, the researchers derived means across the repeated measures or defined variables in terms of whole childhood experience (e.g., parental divorce at any point up to age 16).
Cross-study variability in age of assessment however also occurred for the outcome measure of adult mental well-being. This meant that direct comparisons could not be made to explore age or cohort effects, making it difficult to determine if the childhood social context or age at measurement influenced associations with well-being. The barrier to addressing such heterogeneity retrospectively underlines the advantage of increased alignment across cohort data collection efforts.
Unlike previous research which has typically assessed individual childhood indicators in isolation, this paper has analysed multiple domains of the childhood environment in evaluating their impact on adult mental well-being. This approach acknowledges that mental well-being is influenced by several factors. As in the case of the present paper, the use of prospectively collected longitudinal data allows researchers to effectively ascertain how these might trace back to the childhood environment. Data harmonisation allows exploration of these childhood domains across several studies to determine how these experiences might vary between cohorts, but it is not without challenges.
The findings of this research suggest that interventions should be twofold, focusing on reducing childhood poverty, and supporting positive parenting and parental mental health during childhood in order to promote lifelong mental well-being. Policies that focus on improving childhood socioeconomic circumstances and parental practices would therefore reduce the need for later life interventions.
Wood, N., Hardy, R., Bann, D., Gale, C., Stafford, M. (2021) Childhood Correlates of adult positive mental well-being in three British longitudinal studies. Journal of Epidemiology & Community Health. (75)2, pp. 177-184. https://doi.org/10.1136/jech-2019-213709
Walsh, S., Kaye, N. & O’Neill, D. (2021). Research Case Studies: Childhood environment and adult mental well-being. CLOSER Learning Hub, London, UK: CLOSER.