The present study involved secondary analysis of data collected in the first four Waves of Understanding Society, a new longitudinal study focusing on the life experiences of UK citizens. Data were downloaded from the UK Data Archive (http://www.data-archive.ac.uk/). Full details of the surveys’ development and methodology are available in a series of reports [23–30], key aspects of which are summarized below.
In the first wave of data collection (undertaken between January 2009 and December 2011), random sampling from the Postcode Address File in Great Britain and the Land and Property Services Agency list of domestic properties in Northern Ireland identified 55,684 eligible households. Interviews were completed with 50,994 individuals aged 16 or older from 30,117 households, giving a household response rate of 54 % and an individual response rate within co-operating households of 86 % [23, 29]. Participants are followed up annually. Sample sizes for subsequent Waves are: Wave 2 (January 2010 and March 2012) 54,584 individuals from 30,428 households; Wave 3 (January 2011 and July 2013) 49,708 individuals from 27,715 households; and Wave 4 (January 2012 and June 2014) 47,132 individuals from 25,814 households . Longitudinal individual re-interview rates have risen consistently from 75 % (between Waves 1 and 2) to 85 % (between Waves 3 and 4) .
Data collection for variables used in the present paper was undertaken using either Computer Assisted Personal Interviewing or by examination conducted by a visiting nurse (see below).
Understanding Society does not include information on the formal diagnosis of intellectual disability. As a result, we identified adults with intellectual disability on the basis of the results of cognitive testing undertaken at Wave 3 and self-reported educational attainment. The vast majority of children with intellectual disability have very low educational attainment . As a result, low self-reported educational attainment (no educational qualifications) was used as a selection criterion as evidence that low cognitive ability may have originated in childhood (one of the defining characteristics of intellectual disability). Due to historical changes in educational qualifications and attainment in the UK, we restricted our analysis to the age range 16–49.
In Wave 3 a battery of five cognitive tests was used to assess memory (two tests) and cognitive functioning (three tests; Number Series, Verbal Fluency, Numerical Ability) . The Number Series test was developed for use in the US Health and Retirement Study (HRS) . The Verbal Fluency test has been used in the English Longitudinal Study of Ageing (ELSA) , the German Socio-economic Panel Study  and the National Survey of Health and Development . The Numerical Ability test was taken from ELSA and some portions of it have been used in the HRS and Survey of Health, Ageing and Retirement in Europe .
First, we standardized test scores on the latter three tests to have a mean of zero and standard deviation of one. Second, we used linear regression to impute missing standardized test scores from obtained scores on completed tests. No other variables were used in the imputation process. This led to the imputation of Numeric Ability scores for 153 participants (0.7 % of the used sample), Verbal Fluency scores for 141 participants (0.6 %) and Number Series scores for 1,214 participants (4.9 %). Third, we used principal components analysis to extract the first component (which accounted for 63 % of the variance) from the three scales as an estimate of general intelligence . Fourth, we identified participants as having intellectual disability if they scored lower than two standard deviations below the mean on the extracted component (the conventional cut-off point for defining intellectual disability used in ICD-10) and had no educational qualifications. This identified 294 participants (1.2 % of the unweighted age-restricted sample) as having intellectual disability. An additional 532 participants scored less than two standard deviations below the mean on the extracted component but did have educational qualifications.
Fifth, we included in the intellectual disability group five participants who gave consent for testing but for whom all three tests were terminated due to their inability to understand the test instructions, and also had no educational qualifications. The complete procedure identified 299 participants (1.2 % of the unweighted age-restricted sample) as having intellectual disability.
Four forms of self-reported health data were collected. First, the SF-12 was used to assess physical and mental health . We derived a binary measure of SF-12 Physical Health on the basis of Wave 3 responses to the SF12 Physical Component scores falling within the bottom decile of the weighted Wave 3 sample. Second, self-rated health was evaluated by a single question incorporating five possible response options: ‘In general, would you say your health is … (1) excellent, (2) very good, (3) good, (4) fair, (5) poor’. We converted these data into a binary measure of ‘poor’ vs. better than ‘poor’ health.
Third, in Waves 1–4 participants were asked ‘Has a doctor or other health professional ever told you that you have any of the conditions listed on this card?’ Response options included: asthma, arthritis, congestive heart failure, coronary heart disease, angina, heart attack or myocardial infarction, stroke, emphysema, hyperthyroidism or an over-active thyroid, hypothyroidism or an under-active thyroid, chronic bronchitis, any kind of liver condition, cancer or malignancy, diabetes, epilepsy, high blood pressure. We combined data across Waves 1–4 to derive lifetime prevalence rates of each health condition. Due to very low prevalence rates of specific conditions we derived a measure of respiratory disorder (one or more of emphysema or chronic bronchitis), other cardio-vascular disease (one or more of congestive heart failure, coronary heart disease, angina, heart attack or myocardial infarction, stroke) and thyroid condition (one or more of hyperthyroidism or an over-active thyroid, hypothyroidism or an under-active thyroid).
Finally, participants were asked if since the previous Wave they had had a hospital admission for any newly diagnosed health conditions (using the list of conditions presented above). We combined data across Waves 2 to 4 to derive a variable of hospitalization for a newly diagnosed condition.
Health: biosocial data collected by trained nurse
In Waves 2 and 3 of Understanding Society biosocial data were collected by a trained nurse from a subsample of 11,142 respondents , including 60 respondents (0.5 %) who we had identified as having intellectual disability. In Wave 2, the nurse health assessments were conducted with a subset of the General Population Sample component of Understanding Society, with data collection extending over 24 months. In Wave 3, the measures were undertaken with the British Household Panel Survey sample component of Understanding Society. The eligibility criteria in both Waves were completion of a full face-to-face interview in the corresponding wave, being aged 16 or older, living in England, Scotland or Wales, and completion of the interview in English. In the second year of Wave 2, eligibility was further restricted to .81 of the primary sampling units in England. Nurse visits were not conducted in Northern Ireland or with members of the ethnic minority boost sample. Women who were pregnant at the time of the nurse assessment were classified as not eligible.
The measures taken included: (1) measured weight and height from which BMI was calculated and obesity categorized as a BMI of 30 or more; (2) systolic and diastolic blood pressure; (3) dominant hand grip strength; (4) lung function; (5) medications taken. A measure of stage 2 hypertension or on medication was derived from systolic and diastolic blood pressure readings of greater than 160/100 or that the participant was taking medication for hypertension. Four measures of lung function were derived from the available data: (1) forced expiratory volume in one second (FEV1); (2) forced vital capacity (FVC), the total amount of air (in liters) that can forcibly be blown out after a full inspiration; (3) the ratio of FEV1 to FVC (FEV1/FVC ratio); and (4) peak expiratory flow (PEF), the speed of air moving out of the lungs at the beginning of expiration measured in litres per second. Polypharmacy was defined from the medications data as taking five or more separate medications .
We used two indicators of socio-economic disadvantage. Self-assessed financial status was assessed at Wave 3 by a single item: ‘How well would you say you yourself are managing financially these days? Would you say you are… 1 Living comfortably, 2 Doing alright, 3 Just about getting by, 4 Finding it quite difficult or 5 finding it very difficult?’ Material hardship was assessed at Wave 1 by summing ‘cannot afford’ (questions one to seven) or ‘no’ responses (question eight) to eight questions preceded by the introduction ‘Next we have some questions about the sorts of things that some families/people have but which many people have difficulty finding the money for. For each of the following things please tell me the number from the showcard which best explains whether you (and your family/partner) have it or not. Do you (and your family partner) have: (1) Friends or family around for a drink or meal at least once a month? (2) Two pairs of all weather shoes for all adult members of the family? (3) Enough money to keep your house in a decent state of repair? (4) Enough money to make regular savings of £10 a month or more for rainy days or retirement? (5) Household contents insurance? (6) Enough money to replace any worn out furniture? (7) Enough money to replace or repair major electrical goods such as a refrigerator or a washing machine, when broken? (8) For the next question please just answer yes or no. In winter, are you able to keep this accommodation warm enough?’
Neighborhood social capital
A scale of neighborhood social capital was derived from 13 items relating to perceptions of neighborhood quality and civic and social participation (in the following list #3 contains four separate items):
‘Overall, do you like living in this neighborhood (Yes/No)?’
‘Are you able to access all services such as healthcare, food shops or learning facilities when you need to (Yes/No)?’
‘I am going to read out a set of statements that could be true about your neighborhood. Please tell me how much you agree or disagree that each statement describes your neighborhood (1 Strongly agree, 2 Agree, 3 Neither agree nor disagree, 4 Disagree, 5 Strongly disagree): (a) First, this is a close-knit neighborhood; (b) People around here are willing to help their neighbors; (c) People in this neighborhood can be trusted; (d) People in this neighborhood generally don't get along with each other.’ Data were recoded into binary variables; 1–2 v 3–5 for positively worded questions (a-c), 1–3 v 4–5 for question (d).
‘Now I have some questions about crime. Do you ever worry about the possibility that you, or anyone else who lives with you, might be the victim of crime? Is this a big worry, a bit of a worry, or an occasional doubt?’ Data were recoded into a binary variable; crime is a big worry v not.
‘How safe do you feel walking alone in this area after dark? (1 Very safe, 2 Fairly safe, 3 A bit unsafe, 4 Very unsafe, 5 SPONTANEOUS: Never goes out after dark)’. Data were recoded into a binary variable fairly safe/very safe v not.
‘How many close friends would you say you have?’ Data were recoded into a binary variable; two or more close friends v not.
‘Do you go out socially or visit friends when you feel like it (Yes/No)?’
‘Please tell me how easy or difficult you would find it to visit family or relatives when you need to (1 Very difficult, 2 Difficult, 3 Neither difficult nor easy, 4 Easy, 5 Very easy, 6 Has no family).’ Data were recoded into a binary variable; Easy/very easy v not.
‘Are you currently a member of any of the kinds of organizations on this card (1 Political party, 2 Trade Unions, 3 Environmental group, 4 Parents'/School Association, 5 Tenants'/Residents' Group or Neighborhood Watch, 6 Religious group or church organization, 7 Voluntary services group, 8 Pensioners group/organization, 9 Scouts/Guides organization, 10 Professional organization, 11 Other community or civic group, 12 Social Club/Working men's club, 13 Sports Club, 14 Women's Institute/Townswomen's Guild, 15 Women's Group/Feminist Organization, 16 Other group or organization)’. Data were recoded into a binary variable; member of one or more organization vs. not.
Whether the informant was employed for 16 hours a week or longer
Exploratory analysis of the resulting data indicated that the recoded binary variables showed acceptable internal consistency (alpha = 0.61). As a result, they were combined into a scale of ‘neighborhood social capital’ (range 0–13).
Approach to analysis
Our approach to analysis was undertaken in four stages. First, we recoded all ordinal or scale outcome variables to binary variables. The reasons underlying this decision were: (1) to allow for a common metric of strength of association (odds ratios) to be used across all outcome variables; (2) to avoid analytical problems associated with the non-normal distribution of many of the outcome variables . To create the binary outcome variables we defined low functioning on a particular variable as falling within the bottom 10 % of the population weighted sample. This decision was based on the absence of well-established clinical cut-offs for individual scales and tests. However, functioning in the lowest population decile is likely to be of clinical concern. For example, current UK guidelines suggests that a diagnosis of COPD is indicated when the FEV1/FVC ratio is less than 0.7 and the FEV1 is less than 80 % of predicted . In the present study, the cut-off point for the lowest sample decile for the FEV1/FVC ratio was 0.7.
Second, we made simple bivariate comparisons between participants with and without intellectual disability with regard to available socio-demographic characteristics that may have a potential association with health (e.g., financial strain, gender). Again, to allow for a common metric of strength of association to be used across demographic variables we converted all ordinal or scale variables to binary variables using the sample median as the cut-off (age 16–33, 34–49; self-assessed financial status ‘living comfortably’ or ‘doing alright’/not; material hardship 0,1+; neighborhood social capital (low 0–9, 10–13).
Third we made unadjusted bivariate comparisons (using binary logistic regression) between participants with and without intellectual disability with regard to health status.
Fourth, we made adjusted bivariate comparisons (using multivariate binary logistic regression) between participants with and without intellectual disability with regard to health status adjusted to take account in between-group differences in potential confounding variables. In Model 1 we controlled for between sample differences in age, gender and (for self-report measures) the number of waves in which the respondent participated. In Model 2 we controlled for between sample differences in exposure to two well-established social determinants of poorer health; socio-economic disadvantage [44, 45] and neighborhood social capital [46–48].
We report effect size categories for Odds Ratios following the recommendations of Olivier and Bell; small (OR < =0.82 or > =1.22), medium (OR < =0.54 or > =1.86), large (OR < =0.33 or > =3.00) . All analyses were undertaken using SPSS 20.
Understanding Society is designed and conducted in accordance with the ESRC Research Ethics Framework and the ISER Code of Ethics. The University of Essex Ethics Committee approved Waves 1–5 of Understanding Society. Approval from the National Research Ethics Service was obtained for the collection of biosocial data by trained nurses in Waves 2 and 3 of the main survey (Understanding Society – UK Household Longitudinal Study: A Biosocial Component, Oxfordshire A REC, Reference: 10/H0604/2).