Background: Attention Deficit Hyperactivity Disorder (ADHD) is a childhood and adult disorder characterised by nonnormative inattentive, impulsive, and hyperactive behaviour. Over time the condition has become increasingly medicalised, and whilst it is estimated to affect 5-7% of schoolchildren internationally (Sayal et al., 2018), only 1.6% are diagnosed with ADHD in the UK (NHS Digital, 2018). Reviews report that childhood ADHD leads to poor adult outcomes in all areas of life (e.g. Costello & Maughan, 2015; Erskine et al., 2016). Although about 50% of ADHD children function well as adults, knowledge is limited about psychosocial factors in outcomes, (Costello & Maughan, 2015) such as those related to stress.

State regulation theory, (Sanders, 1983; Sergeant, 2000) was the basis for an investigation using data from the age 0, 5, 10, 34, and 42 sweeps of the 1970 British Cohort Study (BCS70; Centre for Longitudinal Studies: UCL/IoE, 2019). Stress and protective factors were operationalised as stressful life events, chronic stressors, self-esteem, and locus of control. The following questions were examined :

1. What robust measures of DSM-5 ADHD can be retrospectively measured and validated?
2. What is the relationship between childhood ADHD and stress?
3. What is the effect of childhood ADHD on adult a) subjective wellbeing, and b) educational attainment, the latter as a proxy for SES and objective wellbeing?

Method: Innovative data science methods were applied, including:

1. A data mining framework (Kurgan & Musilek, 2006) to derive new constructs in old data;
2. Robust linear and logistic regression models (e.g. MLR, FIML; Muthen & Muthen, 2017);
3. Zero-inflated mixture modelling (Wall et al., 2015) to estimate an ADHD severity score;
4. Machine learning (vselect; Lindsey & Sheather, 2010) to aid selection of an optimal set of covariates for quasi-experimental matching; and
5. Coarsened Exact Matching (CEM; Iacus et al., 2014) to derive a weighted matched sample of ADHD children and similar controls.

Key findings: A DSM-5 ADHD subgroup and subtypes were retrospectively derived and validated using age 10 BCS70 data (N=11,426; nADHD=594, 5.2% prevalence, 30% girls, 46% inattentive subtype). Overall prevalence aligned with epidemiology estimates, but the relatively high percentages of ADHD girls and inattentive cases enabled rare new insights for these groups. The distribution of the ADHD severity score (N=11,426, M=0.06, SD=0.91) supported dimensionality of the construct.

Stressful life events, chronic stressors, self-esteem and locus of control significantly predicted DSM-5 ADHD symptomatology and explained 19.5% of the ADHD severity score at age 10 (N=11,426), supporting State Regulation Theory at the psychosocial construct level.

Quasi-experimental methods were employed to create a pruned longitudinal sample of ADHD and control cohort members matched on evidence-based confounds (N=6,207). Regression models on this sample did not support a significant effect of childhood ADHD on adult outcomes, contrary to prevailing evidence from mostly clinical samples matched on fewer confounds. Matching confounds used were sex, father’s education, depressed mother, mother smoked during pregnancy, childhood wheezing, and low standard home. Replication and refinement are needed, but the finding suggests future experimental studies should consider stratifying samples on these factors, and that ADHD per se may not drive poor outcomes.

In the matched sample (N=6,207), age 10 maths scores (boys and girls), externalising problems, and engagement in leisure activity (girls only), were significant factors predicting a continuous composite measure of adult subjective wellbeing. Parent education, age 10 maths, reading (boys and girls), locus of control, and authoritarian child-rearing views (girls only), were significant childhood factors predicting a dichotomous academic qualification measure of adult educational attainment, as a proxy for SES/objective wellbeing. All effect sizes were small .

In a longitudinal ADHD subsample (n=369), age 10 chronic stressors, externalising problems, and reading significantly predicted adult subjective wellbeing, explaining 7.1% of variance (boys and girls). Father’s education and age 10 reading significantly predicted adult educational attainment. The effects of chronic stressors and reading, and the higher proportion of girls and inattentive ADHD cases in the sample provide novel insights which should be translatable into teacher training and practice.

Findings are applicable internationally, subject to demographic generalisability parameters.