Background: An important question of public and policy concern is the extent to which mental health problems in young people has changed in prevalence over time. Diagnoses and treatment of problems such as autism, ADHD, and depression have shown substantial increases over recent decades, but changes may reflect changes in help seeking, clinical recognition and diagnostic practice. To address whether the population prevalence of child and adolescent mental health problems has changed comparison of unselected representative population cohorts is required (using comparable measures of mental health at each time).

Evidence of this kind shows that adolescent emotional and conduct problems have become increasingly common since the 1970s (Collishaw et 2004, 2010; Kosidou et al 2010; Sigfusdottir et al 2008; Sourander et al 2004; Sweeting et al 2009; Ticket et al 2008). These trends are important for service planning, as well as offering novel approaches to examining risk. Time changes in prevalence must be attributable to population-level changes in environmentally mediated risks. By contrast with findings in adolescence, there is a marked knowledge gap on trends in younger children's mental health. This is important as, knowing whether observed trends in adolecent mental health have their origins in childhood can help narrow down likely causal explanations. However, to date, findings on pre-adolescent children are limited and inconsistent (Achenbach et al 2003; McArdle et al 2003; Sourander et al 2008; Tick et al 2007).

Time trends research provides a method for identifying explanatory environmental risk at a population-level. Thus far, however, very few studies have attempted to go beyond documenting trends, and studies that have tested possible explanations have focused on adolescents (Collishaw et al 2007, 2011; Schepman et al 2011; Gore et al 2011; Sweeting et al 2010). It is important to examine whether population-level change in risk prevalence has contributed to change in child mental health problems.

It is also important to understand any consequences of changes in rates of child psychopathology. There is extensive evidence that mental health symptoms are associated with concurrent impairment and adverse later mental health and psychosocial outcomes (Green et al 2005; Costello et al 2011; Maughan & Kim-Cohen 2005; Scott et al 2001). To date, there has been very little attempt to investigate these associated features in studies of time trends, but it is important to do so, not least because it can provide further evidence of the validity of observed trends (Collishaw et al., 2004).

The project will capitalise on comparable assessments of the mental health (and related risk factors) in seven-year-olds across five UK population cohorts studied from 1965 to 2008 (NCDS, ALSPAC, BCAMHS99, BCAMHS04, MCS). Cross-cohort analyses of the type proposed here offer a unique opportunity to better understand changes in child mental health, and specifcally to address the following aims.

Aims

(1) To test trends in prevalence of child mental health problems over a 40-year period in the UK using comparable assessments of emotional, conduct, and ADHD problems.

(2) To test whether the prevalence and impact of pre-/peri-natal, neurodevelopmental and early psychosocial risks has changed and how such changes have contributed to child mental health trends.

(3) To test trends in the impact of child mental health problems on current functioning, and on later mental health and psychosocial adaptation (using the three longitudinal cohorts).

Hypotheses: Findings regarding trends in child mental health problems are inconsistent. However, it is hypothesised that any increases in symptom levels may be associated with rising rates of impairment and poorer later outcomes

Exposure variable(s)

(1) The primary 'exposure variable' with respect to aim 1 will be year of study.

(2) Exposure variables of interest with respect to aim 2 include

a) Pre- and peri-natal factors: gestational age, birth weight, prematurity, birth complications, maternal smoking in preganancy

b) Neurodevelopmental factors: neurological abnormalities, epilepsy, learning disabilities, language delay, chronic ill health

c) Early psychosocial factors: family composition, social disadvantage, early parental involvement

Outcome variable(s):

1) The primary outcome variable is the parent-rated Strength and Difficulties Questionnaire (SDQ) at age 7 years. This measure is available in three of the other cohorts. The predecessor of the SDQ (the Rutter A scale) will be used in NCDS with comparability ensured using calibration methods developed in our previous studies of time trends (Collishaw et al., 2004).

2) Additional outcomes of interest available in some (but not all) of the cohorts include the teacher SDQ at age 7, and later outcome data: SDQ at ages 11, 14 16 years; peer relations at age 11 and 16 years; school exam attainment/occupation at age 16 years.

Confounding variables: Predictors of non-response assessed at birth, child ethnic origin, child gender, child age (in months).