Demographic information.

We obtained participants’ sex; social class defined as the employment of the mother’s partner at birth using The Registrar-General’s Social Class Scheme: The Stevenson Version [15] (RG I, II, III non-manual, III manual, IV, V); ethnic groups (White/White other, Black/Black British, Asian/Asian British, Mixed race, other or unsure) from responses at ages 7 and 16; legal marital status was defined at age 44 (Single/Never married, Married, separated/divorced/widowed).

IQ in childhood.

Participants completed cognitive tests at ages 7 and 11. We used test results at age 11 when the widest range of the tests were completed to calculate childhood IQ. Participants completed four tests at this age: General Ability (80 Verbal and Non-Verbal multiple-choice questions), Reading Comprehension test (completing 35 sentences), Mathematics test (number skills, fractions, measures, and geometry), and Copying Designs (copying 6 drawings), all tests were performed by schoolteachers. We calculated completers’ childhood IQ by conducting a Principal Component Analysis (PCA) on these scores (N = 14,094), to obtain component scores, which was also done in a previous study [16]. We transformed the resulting scores to a scale with a mean of 100 (Standard deviation [SD] = 15) to correspond to IQ scores (min = 50.1, max = 139.6). These scores were normally distributed. We categorised participants into three groups “>85”, “borderline intellectual functioning (IQ 70–85)”, and “mild intellectual disability (50<IQ<70)” based on the standard intellectual disability definition [3].

Social participation: Social contact with relatives and friends.

‘Social participation’ has been conceptualised in a variety of ways with different terms to describe social domains, e.g., connections and engagement [17]. In this study, we define social participation as direct participation in interactions with others including confiding relationships, in line with a recent consensus definition [18]. At age 44 participants completed questionnaires about their social life, using six questions from the Berkman Syme social network index [1, 19] (Table A in S1 Appendix). These measures previously showed an association with subsequent dementia and declining cognition in a study of UK civil servants [1]. Each question is answered on a Likert-scaled for the number or frequency of people seen, with a higher score indicating more social contact. We generated a ‘relatives’ subscale from questions 1 to 3, a ‘friends’ subscale from questions 4 to 6, and a total social contact score by summing all 6 scores. The possible total social contact index is 30, with 15 each for relatives or friends.

Social participation: Confiding/emotional support from the closest person.

At age 44, participants completed the Close Person Questionnaire (CPQ) [20], which assesses emotional and practical social support from one close person nominated by the respondent, either "husband/wife/partner", "boyfriend/girlfriend", "Parent", "Siblings"(brother/sister), "Children" (son/daughter), "Other relatives”, "Neighbour", "Friend from work", "Other friends" and "Others". We used the Confiding/emotional support subscale which has eight questions (possible scores from 0 to 24) and asks the respondent to consider the emotional support from the person they felt closest to in the preceding 12 months (Table B in S1 Appendix). All questions have four response options: Not at all = 0, A little = 1, Quite a lot = 2, A great deal = 3. Higher scores correspond to a more confiding relationship or higher social support.

Social participation: Other confiding relationships.

At age 50, the participants were asked “If you needed to talk about your problems and private feelings how much would the people around you be willing to listen?”. This question has four response options: Not at all = 0, A little = 1, Somewhat = 2, A great deal = 3.

Measurement of cognitive ability in midlife.

Cognition was tested at age 50. We performed PCA on three tests: memory (word list recall), speed of processing (letter cancellation) and verbal fluency (animal naming) [21] which were completed at age 50, then standardised test scores to z-score (mean = 0, SD = 1).

Quality of life.

QoL was measured at age 50 using the 12-item Control, Autonomy, Self-realization, and Pleasure Version 2 (CASP-12 v.2) [22] which has four dimensions: control, autonomy, self-realization, and pleasure [23] (Table C in S1 Appendix). Each Likert-scaled item has four options: “Often”, “Sometimes”, “Not often”, and “Never” (score 0 to 3). Possible total score ranges from 0 to 36, with higher scores indicating better QoL.

Health status.

General health was rated at age 44 when participants were asked: “How would you describe your health generally” with potential responses of “Excellent”, “Good”, “Fair”, and “Poor”. Mental Health was measured using the validated Clinical Interview Schedule-Revised (CIS-R) [24] at age 44. The scores are summed to give an overall severity score from 0 to 4 for fatigue, concentration and forgetfulness, sleep problems, irritability, depression, depressive ideas, anxiety, phobias, and panic, with possible total scores ranging from 0 to 36. A score of 12 or more indicates a significant level of symptoms, and a score of 18 or more suggests that treatment is needed.

Statistical analysis.

We used Stata MP 17.0 for statistical analyses. We first described participants’ demographic characteristics and social participation.

1. To calculate the association between childhood IQ classifications and mid-life social participation. We used Multiple Imputation by Chained Equations (MICE) [25] to impute missing covariate data, including ethnic group, social class, legal marital status at age 44, self-reported general health at age 44, the continuous total CIS-R scores at age 44. We then used linear regression using those with IQ above 85 as the reference group, to examine the relationships between IQ classifications in childhood and: 1) Social contact with relatives and friends at age 44, 2) confiding relationship with one person at age 44 and with anyone at age 50. The model was initially unadjusted (model 1), then adjusted for sex, ethnic group, and social class (model 2), then additionally adjusted for legal marital status at age 44 and self-reported general health at age 44 (model 3), and finally additionally adjusted for continuous CIS-R scores at age 44 (fully adjusted model- model 4).
2. To examine whether social participation mediates the relationships between IQ in childhood and cognition, and QoL at age 50. We used Baron and Kenny’s mediation test [26] (Text in S2 Appendix). It consists of two main steps, 1) linear regression and 2) statistical significance tests (Sobel’s z-test, Bootstrapping [27]). The possible results are no, partial or full mediation [26, 28].

Participants.

We summarised the demographic characteristics of our sample in Table 1. There were 235 people with mild intellectual disability (1.7%), 2234 people with borderline intellectual functioning (15.9%) and 11625 people with IQ above 85 (normal group) (82.5%). Overall, 51.4% of participants were male, but 54.7% of those with borderline intellectual functioning and 59.2% of people with mild intellectual disability were male. At age 44, 45.3% of people with mild intellectual disability, 65.9% of people with borderline intellectual functioning and 73.1% of people with IQ above 85 were married. At age 44, 56.6% of people with mild intellectual disability, 70% of people with borderline intellectual functioning, and 82.7% of people with IQ above 85 reported their health as excellent or good.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Socio-demographic of study participants.

https://doi.org/10.1371/journal.pone.0302411.t001

Social network.

The mean score of Berkman-Syme social network index for adults with mild intellectual disability was 23.5 (SD = 3.1), for those with borderline intellectual functioning was 22.6 (SD = 3.4), and for those with IQ above 85 22.6(SD = 3.1). The mean score for social contact with relatives for adults with mild intellectual disability was 11.8 (SD = 1.9), for adults with borderline intellectual functioning was 11.3 (SD = 2.1), and 11.1 (SD = 1.9) for people with IQ above 85. The mean score of social contact with friends was 11.8 (SD = 2.1) for adults with mild intellectual disability), 11.2 (SD = 2.2) for people with borderline intellectual functioning and 11.5 (SD = 2.2) for adults with IQ above 85.

Confiding/emotional support.

6792 (84.5%) participants nominated a close person at age 44. For people with mild intellectual disability, their closest relationships were husband/wife/partner (47.2%), or parents (27.8%), followed by others (8.3%), siblings and other friends (both 5.6%), boyfriend/girlfriend and neighbours (both 2.8%). Unlike the other two groups, none reported their closest person as children or other relatives or friends from work. People with borderline intellectual functioning reported their closest relationship to be a spouse or partner (71.8%), other friends (5.9%), parents (5.6%), and followed by siblings (5.1%), children (4.8%), boyfriend/girlfriend (3.9%), other relatives (1.2%), friends from work (1.1%) and neighbours (0.5%) and others (0.3%). The majority of people with IQ above 85, reported their closest relationships are with their spouse or partner (76.1%), followed by other friends (7.1%), parents (3.8%), boyfriend/girlfriend (3.7%), children (3.4%), siblings (3.3%), friends from work (1.5%), other relatives and others (both 0.4%), and neighbours (0.3%).

The mean score of confiding/emotional support from the closest person for people with mild intellectual disability was 15.0 (SD = 5.1), for people with borderline intellectual functioning was 14.7 (SD = 4.4), and for adults with IQ above 85 was 15.3 (SD = 4.0). Mean scores for confiding relationships with anyone were 3.5 (SD = 0.9) for people with mild intellectual disability, 3.5 (SD = 0.9) for people with borderline intellectual functioning, and 3.7 (SD = 0.7) for people with IQ above 85

QoL.

The mean QoL at age 50 for people with mild intellectual disability was 23.0 (SD = 5.9); it was 25.0 (SD = 6.1) for people with borderline intellectual functioning and 26.3 (SD = 5.7) for people with IQ above 85.

Strengths and limitations.

This is the first study to use a nationally representative sample to investigate whether confiding relationships with others can protect cognition for people with borderline intellectual function and people with IQ above 85. This study also provides knowledge about the structures of social participation in a representative cohort of adults with borderline intellectual functioning and mild intellectual disability, in comparison to that of people with IQ above 85. To maintain representativeness, we imputed values of missing data using MICE.

This study has limitations. We were unable to determine whether the confidants of adults with intellectual disability are individuals themselves with or without intellectual disability, or to compare the difference in mediation effect of social participation between people with different severity of intellectual disability due to the lack of information and power. We only considered social contact with relatives and friends, and confiding relationships but did not include group activities or work, so we may have underestimated social participation, particularly for people with borderline intellectual functioning as their patterns of social participation may be different from people without intellectual disability. Also, people with mild intellectual disability who said they had a partner or boyfriend/girlfriend might understand the question not as an exclusive, intimate partner but as someone special to support them, as the number of partners reported was much larger than previously reported in people with mild intellectual disability.

Furthermore, for people with mild intellectual disability, the available data did not fulfil the criteria of Baron and Kenny’s Mediation test, so mediation is not calculated. This may be due to the small samples or because there is no mediation effect. We were unable to examine whether the apparent effect of social participation on cognition and QoL will continue as these people become older. And since the incidence of dementia is low at age 50, we were unable to explore whether social participation in midlife affects dementia risk. We need to explore it in the future.