Associations between mental health competence and indicators of physical health and cognitive development in eleven year olds: findings from the UK Millennium Cohort Study

We used the Millennium Cohort Study (MCS), a nationally-representative contemporary cohort of 19,517 children born in the UK between September 2000 and January 2002 [14]. We analysed data on children who were present at the age 11 sweep (13,112 singletons interviewed in 2012/13, 68% response rate). Only children living in England and Wales (n = 10,378) were eligible for inclusion in the teacher survey, and 7430 (72% response rate) participated [15]. After exclusion of those with missing data on mental health competence measures (described below), and where the main respondent was someone other than the mother, the analytic samples comprised 12,082 children for maternal report (92% of children present at that sweep) and 6739 children for teacher report (91% of children with data from the teacher survey). Multiple imputation was carried out for 20 imputed datasets under a missing at random assumption, to address missing outcome and covariate data, using Stata ‘mi impute chained’. Complete case analyses are included in the supplementary materials (Additional file 1: Tables S1-S4). Results are similar to those using the imputed data reported in the paper.

Ethical approval was granted by the Northern and Yorkshire multicentre research ethics committee in July 2011 (Ethics Committee reference: 11/YH/0203). Further information about the MCS can be found elsewhere (http://​www.​cls.​ioe.​ac.​uk/​MCS). Data were downloaded from the UK Data Service, University of Essex and University of Manchester, in March 2014.

We investigated MHC in relation to a series of covariates in order to examine the socio-demographic patterning of MHC and to account for potential confounding between MHC and the outcomes: child sex and ethnicity; maternal age at cohort child’s birth and highest academic qualification; family structure (natural couple, reconstituted family, lone parent); and maternal mental health problems (Kessler-6 scale [21], summed and dichotomised using an existing cut-off for medium (including high) psychological distress [22]). We also investigated MHD as a potential confounder of the association between MHC and physical health and cognitive development in sensitivity analyses. This adjustment was not included in the main analyses due to uncertainties over the direction of the relationship between MHC and MHD (that is, if MHD is a mediator of the association between MHC and health and development, it would be an over-adjustment to account for it).

Four class latent measures were selected for both maternal and teacher reported MHC. The characteristics of MHC classes are detailed in Table 1 and described below.

Indicators of physical health and cognitive development varied by classes of MHC reported by mothers and teachers, with children rated as having higher MHC generally having better outcomes than those with lower MHC (Supplementary Materials: Additional file 1: Table S8). Multinomial regression was used to estimate the associations between MHC classes and indicators of health and cognitive development.

In a nationally-representative contemporary sample of 11 year old UK children, we drew on existing survey items to develop a measure of mental health competence (MHC) for public health research comprising two age-relevant competencies: prosocial behaviours and learning skills, reported by mothers and teachers. Using a data-driven technique to summarise the combination and level of both competencies, four classes were identified ranging from high MHC (high prosocial behaviour and high learning skills) to low MHC (moderate prosocial behaviour and low learning skills). Although most children displayed high or high-moderate MHC, levels were lower in children from less advantaged socio-economic circumstances, boys, and children of mothers with psychological distress. MHC and mental health difficulties (MHD) were not the inverse of one another, with notable proportions of children rated with high MHC and high MHD, or low MHC and low MHD.

Using indicators of physical health and cognitive development that could be plausibly influenced by MHC, we showed variability in outcomes and degree of severity according to MHC class. MHC was not, or only weakly, associated with experiencing a single unintentional injury between ages 7–11 years, being overweight or having only a single symptom of asthma. MHC was more strongly related to injuries on multiple occasions, obesity and two or more asthma symptoms, although associations generally did not show a dose-response relationship according to level of MHC. For example, the likelihood of poor physical health on the indicators assessed was generally greater for those with low MHC (compared to those in the high MHC class); however, the elevated risks for those with moderate MHC were often smaller than those with high-moderate MHC. MHC was strongly associated with measures of cognitive development available in the MCS at 11 years (verbal, spatial working memory and risk-taking), with low MHC (and to a lesser extent high-moderate) associated with the lowest scores. In the MCS, as in many other studies, children from more disadvantaged circumstances are at greater risk of poorer physical health and cognitive development [19, 25]. While adjustment for confounding attenuated the associations between MHC and outcomes in some cases, elevated risks remained. Results were also similar after accounting for MHD as a covariate. However, given potential bidirectional pathways between mental health competence and difficulties, these sensitivity analyses represent a potential over-adjustment.

Positive mental health and mental illness do not necessarily represent opposite ends of a continuum of mental health, and, while related constructs, they may develop along separate pathways [26]. As with earlier research on MHC from Australia [13], we found that competence and mental health difficulties were distinct constructs, and that high MHC was not necessarily accompanied by low MHD (or low MHC with high MHD). Our findings are also consistent with earlier studies that have revealed social patterning in diverse measures of positive mental health [27‐32], including mental health competence [11], whereby advantaged children are most likely to exhibit positive mental health. However, we have gone further by differentiating high, moderate and low functioning on prosocial behaviour and learning skills domains. We have shown that, although less prevalent in children from disadvantaged families, high MHC was common in MCS children at age 11 across all social groups.

There is limited research investigating associations between positive mental health and indicators of child health. For childhood overweight, evidence is inconsistent, with some research showing an association between higher levels of learning skills and prosocial behaviours and smaller risks of overweight and obesity [33‐36], while others have found an association between higher levels of social skills and greater risks of overweight and obesity [37, 38]. Although there is evidence linking childhood problem behaviours with risk of injury [39], to our knowledge this is the first study to have demonstrated an association between positive mental health (specifically MHC) and injury. Asthma management has been associated in a number of studies with psychological factors (including coping styles and mental health problems) [40], with some evidence that specific positive mental health factors (such as self-efficacy) predict better asthma control [41]. In terms of cognitive development, Australian research has shown similar results to those observed in the MCS, with higher MHC at age 5 years associated with classroom oral communication skills [4], and language and cognitive skills [13]. Attitudes to risk may have an impact on important life decisions [42] and risky behaviours [43], and skills related to mental health competence are associated with a lower likelihood of initiating risk-taking behaviours in adolescence [44]. In this context, relationships between MHC and risk-taking scores at age 11 years may presage future patterns of behaviour.

Inconsistencies between studies may in part reflect non-linear relationships between positive mental health and outcomes. For example, our study has shown that children classed as moderate MHC sometimes had physical health and cognitive development outcomes as good as children with high MHC and better than children with high-moderate MHC. Patterns of association between MHC and outcomes are likely to, in part, reflect the greater differentiation manifest in learning skills (which could be classed into three levels) compared to prosocial behaviour (for which only high and moderate levels were identified).

This study was carried out using a large, nationally-representative cohort of children. We were able to investigate competencies, in contrast to measures of mental illness which seek to identify children with mental health difficulties versus a heterogeneous comparison group of children without difficulties. Using latent class analysis, we developed a measure of MHC which allowed questions on prosocial behaviours and learning skills to be summarised while differentiating levels of high and moderate functioning rather than producing a single score. Unlike other measures of positive mental health based on measuring subjective mood or affect, we focused on competencies than could be assessed by mothers and teachers. A limitation of the study is the relatively narrow range of questions pertinent to the measurement of MHC, which captured only learning skills and prosocial behaviours. As operationalised elsewhere [8], MHC (depending on the age of the child) might also include responsibility and respect, and readiness to explore new things. In the MCS, as in many population surveys that include assessments of health and well-being, negative markers of mental health predominate rather than strengths or competencies. Our review of relevant survey questions resulted in the selection of the majority of MHC items from the widely-used Strengths and Difficulties Questionnaire (SDQ), including all items from the prosocial scale and three positively-framed items from the hyperactivity and conduct disorders scales of the total difficulties scale. The fact that the items used to capture MHC in the present study are commonly collected in surveys and at different ages is, however, also a potential strength. We do not claim to have developed a complete measure of MHC, rather to have made efficient use of existing data to explore positive mental health at a population level, with a focus on the use of such measures in a public health and epidemiology setting. This provides opportunities for similar measures of MHC to be developed and used in other studies and at different stages of childhood and adolescence. There are challenges in deciding the most appropriate reporter for mental health-related measures as there tends to be relatively low agreement [45]. However, there is a general preference for taking more than one perspective, including in population studies [46]. In our study, the same classes of MHC were identified from reports provided by mothers and teachers, indicating a consistent structure for this measure of MHC, although prevalence of classes differed slightly. The patterning of results were largely consistent for mother and teacher reported MHC, although the strength of associations tended be greater for teacher-report MHC. Differences (or similarities) between mother and teacher-report MHC do not necessarily indicate bias, and may instead show the importance of context, with children’s behaviours varying between home and school settings [47].

We were able to assess physical health outcomes that are common in children and plausibly predicted by MHC, together with measured aspects of cognitive development. The MCS has data on many explanatory covariates which could be examined in relation to MHC, and accounted for when investigating associations between MHC and the physical health and cognitive development of the child. As with many cohort studies, missing data was an issue. Weights were used in all analyses to account for sample design and attrition up to age 11. In addition, missing MHC items were automatically imputed during the LCA procedure. Although there were missing data due to non-response on other items, the maternal-report sample comprised 92% of the MCS children who took part in the 11 year sweep, and we used multiple imputation to address missing covariate and outcome data. We have shown that a relationship between MHC and child health and cognitive development largely persisted after adjustment for potential confounding by socio-demographic characteristics and maternal psychological distress. Nevertheless it remains possible that these cross-sectional associations are susceptible to residual confounding or reverse causality.