Background The child mental health epidemiology literature focuses almost exclusively on reporting the prevalence and predictors of child mental disorders. However, there is growing recognition of positive mental health or mental health competence as an independent outcome that cannot be inferred from the absence of problems, and requires epidemiological investigation in its own right.
Methods We developed a novel measure of child mental health competence within the framework of the Australian Early Development Index, a three-yearly national census of early child development. Predictors of this outcome were investigated by linking these census data at individual level to detailed background information collected by a large longitudinal cohort study.
Results Predictors of competence were consistent with previously described theoretical and empirical models. Overall, boys were significantly less likely than girls to demonstrate a high level of competence (OR 0.60, 95% CI 0.39 to 0.91). Other strong predictors of competence were parent education and a relative absence of maternal psychological distress; these factors also appeared to attenuate the negative effect of family hardship on child competence.
Conclusions This measure of mental health competence shows promise as a population-level indicator with the potential benefit of informing and evaluating evidence-based public health intervention strategies that promote positive mental health.
- Child Health
- Mental Health
- Public Health Policy
- Measurement tool Development
- Lifecourse / Childhood Circumstances
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
- Child Health
- Mental Health
- Public Health Policy
- Measurement tool Development
- Lifecourse / Childhood Circumstances
Child mental illness is a public health issue of global importance:1 ,2 The World Health Organization (WHO) reports that up to a fifth of all children experience disabling mental disorders.3 There is an urgent need to address this burden given that developmental trajectories are set early, with opportunities for intervention most effective and cost effective while children are young.4 Recently, there has been a shift in focus towards child mental health promotion and the prevention of mental disorders.5 ,6
This approach is supported by evidence from the psychological literature where positive mental health, or mental health competence, is considered an independent predictor of improved health, educational and employment outcomes.5 ,7–9 Indeed, this same literature notes that mental health and mental disorder are separate states rather than poles on a continuum. The WHO definition of health8 is once more affirmed by emerging evidence that positive mental health cannot and should not be inferred from the absence of mental disorder.9–11
The public health community is beginning to take note. A recent WHO report stated that ‘there is now compelling evidence for the need to promote positive mental health through interventions that promote competence and psychological strengths’.5 In subscribing to this approach, it is therefore clear that there must be evidence to guide the planning and evaluation of such interventions. In particular, an understanding of the determinants of positive mental health is essential to inform decision makers who seek optimal prevention and promotion strategies.12–15
We note, however, that while numerous epidemiological studies report the population prevalence16–18 and risk factors18 ,19 for mental health problems, less is known about the epidemiology of mental health competence, particularly in children. One reason for this may be a lack of suitable measures. Currently, most measures of competence are designed for testing individuals and may demand resources that are not achievable in population-based studies.20 Thus, although pathways to competence have been explored conceptually and to a lesser extent empirically in the psychology literature,21 ,22 there is a notable lack of population-level research.
We aimed to address this evidence gap by (1) developing a measure of mental health competence in a population-based sample of children and (2) investigating the predictors of this competence measure using an empirical framework derived from the existing psychological literature. To do this, we interrogated a unique dataset in which outcome data from a national census of child development were linked with detailed early-life information from a large longitudinal cohort study; thus, our approach draws on lessons from the individual paradigm to develop a population-level approach to positive child mental health.
The principal data source for this study was the Australian Early Development Index (AEDI),23 a national census of children's development at school entry. Five developmental domains are reported by the class teacher on every child in their class using a structured survey. Brief socio-demographic characteristics including gender, age, Indigenous status and language background are recorded for all children, and area-based measures of socioeconomic status and remoteness are assigned on the basis of the child's residential suburb.
Growing Up in Australia: the Longitudinal Study of Australian Children (LSAC) is a nationally representative clustered cross-sequential sample of two cohorts of Australian children which commenced in May 2004. The sampling design has been described in detail elsewhere.24 The LSAC records detailed information across a wide range of domains including the child's development and the family environment.25
The linked AEDI/LSAC dataset comprises a subset of 720 children enrolled in LSAC who had started school or preschool at the time of the first wave of data collection in 2004. Consent was sought for teachers to complete the AEDI questionnaire on these children, allowing AEDI outcomes to be linked at the individual level with the richly detailed information collected by the LSAC.
Development and ascertainment of measures
Dependent variable: mental health competence
The construct of mental health competence was based on the well-established definition of competence provided by Masten and Curtis:
… adaptational success in the developmental tasks expected of individuals of a given age in a particular cultural and historical context. Competence by this definition is inherently multi-dimensional, because there are multiple developmental tasks salient in a given age period in a given place and time in society.11
This definition aligns well with the AEDI given its age-specificity and school-based cultural context, requiring teachers to evaluate each child relative to their expectations of new school entrants in the school setting.
An important characteristic of this outcome is that it measures strengths rather than the absence of problems, incorporating five key positive mental health constructs measured and previously validated25 in the AEDI: overall social competence; responsibility and respect; approaches to learning; readiness to explore new things; and prosocial and helping behaviour (32 data items in total; age-standardised; table 1). The measure is thus multi-dimensional and focuses on developmental tasks as outlined by Masten and Curtis.11 The outcome measure was derived by averaging each child's score for these five constructs (Cronbach's α=0.87). Interitem correlations ranged from 0.72 (responsibility and respect vs approaches to learning) to 0.46 (responsibility and respect vs readiness to explore new things), consistent with the conceptualisation of this outcome as measuring different components of mental health competency that are related, but distinct, constructs. The threshold for a high level of competence was defined as an average score in the top quintile, that is, the top 20%.
We tested the validity of this newly derived measure using the Strengths and Difficulties Questionnaire (SDQ),26 a measure of child mental health, available in LSAC for the same population. We found that teacher-reported competence was strongly associated with both teacher- and parent-reported SDQ scores. A logistic regression model with mental health competence as the dependent variable and parent-reported SDQ subscales as independent variables offered evidence of convergent and divergent validity: there was a positive association with the prosociality scale (OR 1.21; 95% CI 1.05 to 1.40) and an inverse relationship with the peer problems scale (OR 0.68; 95% CI 0.56 to 0.83), with ORs corresponding to a 1 unit increase in the relevant scores. Discriminant validity was indicated by a lack of association (either positive or negative) with the hyperactivity, emotional symptoms and conduct problems scales; these had ORs of 0.91, 0.97 and 0.91 respectively and the ORs were not significant at the 95% confidence level.
To develop a strong causal model from prior information, we reviewed the child development and mental health literature for previously-described predictors of positive mental health.
Although the ecological model of child development27 recognises influences at child, family and community levels, there is evidence in the literature for a stronger effect of family characteristics compared with area characteristics on child mental health.28 This is likely to be particularly true of the study children, given their young age. This consideration influenced the selection of covariates towards factors located in the child and family, although still mindful of broader environmental influences.
Predictors of positive outcomes are poorly understood in many respects, having been neglected in favour of studying risk or preventative factors for mental health disorders. Nevertheless, to avoid constructing post hoc hypotheses about these relationships from the data, variable selection for the analysis was limited as far as possible to factors that had been postulated as likely to promote competence rather than primarily preventing mental health problems.22 ,29–32 Factors fitting this criterion are summarised in figure 1, and details of the measures are provided in table 1.
Design weights originating from the LSAC dataset35 were used for all logistic regression analyses to adjust for population proportions and differential non-response patterns, and the complex survey design was accounted for using first-order Taylor linearisation, implemented using Stata survey analysis procedures.
In the bivariate analysis, children with competence scores in the top quintile were compared with the standard population across individual, parental and family characteristics. A further multivariable logistic regression model was then specified from prior information. Covariates for this full model were gender; whether either parent had completed high school; family hardship; mothers’ psychological distress score; family structure (parents living together and both biological parents present); and use of inductive reasoning.
Of the 720 children in the linked dataset, 688 (95.6%) had sufficient data on outcome measures to be included in this study. Across the covariates, missing data ranged from 0% to 33%. The covariate with the highest proportion of missingness in this dataset was the fathers’ psychological distress scores which had only been collected for resident fathers at this wave of data collection; this measure was therefore excluded from the logistic regression analysis. There were no significant differences in the distribution of high competence, gender, parent education and parenting styles between children with complete data on all covariates and the 146 children with missing data who were omitted from the analysis. For the children included in the full model, families were less likely to have experienced hardship and mothers were more likely to have a relative absence of psychological distress.
Analysis was performed using Stata/IC V.12.1 for Windows (copyright 1985–2011 StataCorp LP, revision 22 October 2012). ORs refer to exponentiated coefficients obtained in logistic regression analyses, and CIs are given at the 95% limit throughout.
Demographic characteristics of the study population
Child, parent and family characteristics of the study cohort are summarised in table 2. There was an even gender distribution in the study sample, and the mean age was 4.7 years. Despite the use of design weights, the study families were less diverse, family structures were more stable, parents were better-educated and parent employment levels were higher when compared with the broader LSAC child population. However, a quarter of study families stated that they had experienced recent hardship, for example, being unable to pay bills on time or needing to limit food purchases, indicating that the sample included families in diverse circumstances.
Mental health competence outcome variable
We identified 153 children with an average competence score equal to or above the 80th centile for the cohort. These children were regarded as demonstrating a high level of competence. Because of a clustering of scores at the top end of the scale, the percentage of children with scores at or above the 80th centile score constituted 22.2% of the total.
There was a substantive impact of gender on competence: the OR of high competence for boys was 0.57 (95% CI 0.39 to 0.84) (table 3). Other significant associations in the bivariate analysis were high school education, and relative absence of family hardship, recent stressful life events or psychological distress reported by the mother.
Table 3 also shows adjusted estimations of the effects of key influences identified in the literature: gender, parent education, family hardship, parent mental health and parenting style.
After adjustment, aside from gender the most important predictors of high competence were maternal education, with an OR of 1.57 (1.01 to 2.44) if the child's mother had completed high school, and a relative absence of psychological distress in mothers (OR 1.80; 1.15 to 2.81).
The negative effect of family hardship, a significant predictor in the bivariate analysis, was attenuated by adjustment for parent education and mental health. High family income was a significant predictor of high competence in isolation, but the OR was closer to the null if income was entered into the model together with hardship.
Use of inductive reasoning (reasoning with the child when he/she misbehaves) was selected for the model as being the parenting characteristic which most closely resembled previously-described competence-promoting parenting behaviours.21 However, it was not significantly associated with competence in this population. The other parenting characteristics (angry, warm, consistent, observed to praise child) had similarly minimal effects when substituted for inductive reasoning in the model.
There were indications of effect modification by gender in this dataset, for example, when high school data of mothers and fathers were entered separately into the model. There was significant interaction between mothers’ education and child gender (p<0.001) which was not seen with the combined parent education measure (p=0.091). When the model was stratified by child gender, the adjusted OR for mother completing high school was 2.71 (1.42 to 5.17) for girls and 0.86 (0.45 to 1.65) for boys. However, stratified results were difficult to explore due to lack of power in the resulting reduced sample sizes.
We report the predictors of child mental health competence using a novel population-based outcome measure. To our knowledge, this is the first description of a mental health competence measure for which data potentially exist in a full national dataset. Our findings align with previous work on this topic in the psychology literature, suggesting that this measure could be suitable for reporting and monitoring child mental health competence at a national level, particularly for young children.
Our conceptual model of predictors of mental health competence, based on existing literature, took into account hypothesised direct and mediated effects. Key findings from our analyses are outlined below and support previous work, particularly demonstrating the importance of maternal education and psychological resources on child competence.21 ,32 Specifically, we found that it was mothers’ concurrent psychological health (as opposed to their previous psychological health) that held the strongest implications for child competence. This is consistent with past research36 where teacher-reported adaptive functioning was lower for children of mothers with concurrent depression, but not perinatal depression. The findings are further strengthened by the use of teacher-report for children's behaviour; this avoids the likely influence of mothers’ own mental health on their perception of children's competence.
Family hardship was inversely associated with competence in the bivariate analysis, but the effect was attenuated by adjustment for maternal mental health and education. Hardship has previously been shown to increase parent stress,22 while educational attainment may predict greater use of adaptive coping methods.21 We note also that income did not appear to influence competence independently of hardship. A number of population studies have similarly failed to demonstrate a direct effect of income on child mental health,37 suggesting that hardship may be a more sensitive predictor.
Not surprisingly gender was important, with girls more likely than boys to demonstrate a high level of competence. The impact of gender as an individual predictor is consistent with previous studies of teacher-reported prosocial behaviour in young children,30 and is posited to be a result of both biology and socialisation.38
Finally, there was an unexpected lack of association between parenting style and mental health competence in this study. This divergence from previous studies21 may be attributable to different measures and constructs. We found that parenting styles were themselves predicted by maternal psychological distress, maternal education and/or family hardship (data not shown), suggesting that our parenting style constructs were valid descriptors, but that these specific behaviours were not strongly competence-promoting once other contextual factors were accounted for. We also noted a possible threshold effect: while parenting style did not predict a high level of competence (membership of the top 20%), it did predict membership of the top 40% in this population (data not shown).
Caution must be exercised when generalising the results of our study to the total Australian population in light of key demographic differences. Children in the AEDI subsample were less disadvantaged than in the full LSAC sample (table 2) and the LSAC children in turn were less disadvantaged than in the full Australian population. Compared with the national population, our sample had fewer Indigenous children (1.2 vs 2.6%), more mothers who were university educated (27.5% vs 21.0%) and more two-parent families (83.9% vs 72.0%).25 We note also that children excluded from the multivariable model for missing data were more likely to be from families experiencing hardship or maternal psychological distress; as a result, our analysis may underestimate the effect of these factors on competence. Sample size considerations limited detailed modelling of these complex relationships.
A further limitation was the traditional emphasis on mental disorder rather than health in the existing datasets. This posed a challenge when selecting variables for this positively-focused study from data items which were primarily designed to identify problems, not strengths. For example, we had data on maternal psychological distress but did not have an indicator of positive mental health. If designing items de novo we would have sought to measure positive and negative mental health in parents as well as children.
A degree of pragmatism is called for when investigating health at the population level as a loss of information is often inevitable. This outcome measure was constructed from existing public health data and is not presented as a psychometric instrument. Instead, our aim was to develop a population indicator as outlined by Hamilton and Redmond,39 recognising that a measure with sufficient data available for reporting population trends may not capture the detail required for clinical decisions about individuals.40
We are encouraged to find that despite the above limitations our competence measure has a high level of face validity, correlates with related constructs as expected and is sensitive to predictors described in the psychological literature, suggesting that this is a sufficiently robust measure of competence at least for use in existing Australian population-level data. We are also pleased to note convergent validity for the teacher- and parent-rated measures as school is a more cost-effective setting than home visits for large-scale surveillance of competence in the child population.
Following this study, the AEDI full datasets provide an opportunity to investigate mental health competence data in over 97% of the estimated 5-year-old Australian populations in 2009 and 2012: a total of half a million children to date. The three-yearly AEDI cycles offer an opportunity to measure competence in successive full national cohorts of school entrants, with the potential to provide both outcome data for early childhood interventions and baseline data for school-based interventions. In addition, the large sample sizes thus generated will confer sufficient study power to investigate vulnerable subpopulations, to compare effects at individual and area level, and to measure population trends over time, providing further public health- and policy-level insights.
This investigation was instigated in response to a growing conviction in the public health community of the importance of understanding positive as well as negative child mental health. The development of a positive outcome measure that can be applied to large populations of children is one step towards this goal. In addition, the association of maternal education and mental health with child competence detected in this preliminary study provides an indication of critical, yet modifiable, influences on young children which may inform public health planning. Further population-level investigation of mental health competence can provide a solid evidence base with which to realise the WHO agenda for positive mental health.
What is already known on this subject?
Child mental health is a public health issue of global importance. It is now recognised that mental health is not equivalent to the absence of disorder and that there is a need to promote positive mental health through interventions that promote competence and psychological strengths. However, the public health research community has been slow to define and investigate positive child mental health outcomes.
What this study adds
We present a teacher-reported measure of competence that can be applied to national-level data. Our analysis indicates that parent education and mental health are important predictors of competence. As for mental disorder, there is a social gradient operating for positive mental health outcomes.
Effective child mental health policy requires a detailed understanding of positive as well as negative outcomes. Reporting mental health competence at the population level can help to guide policy and interventions to promote the WHO agenda for positive mental health.
We are grateful to Dr Joanne Tarasuik for preparing the study dataset and assisting with the data description and results tables, and to Dr Sarah Howell-Meurs for assistance with the literature review. We would also like to thank the anonymous reviewers for their helpful comments. The authors thank the families participating in the study.
Contributors SG determined the aims and scope of the investigation. MO'C and EI advised on the psychological aspects of the study. AK developed the outcome measure and analysed the data, with guidance from FM. All authors contributed to interpretation of the results. AK and EI drafted the initial manuscript with critical revisions by all authors.
Funding The Australian Government and State and Territory Governments are working in partnership with The Royal Children's Hospital Centre for Community Child Health in Melbourne, the Murdoch Childrens Research Institute and the Telethon Institute for Child Health Research, Perth, to deliver the AEDI. This article uses a confidentialised unit record file from the Longitudinal Study of Australian Children (LSAC), initiated and funded by the Commonwealth Department of Families, Community Services and Indigenous Affairs, and managed by the Australian Institute of Family Studies. Murdoch Childrens Research Institute research is supported by the Victorian Government's Operational Infrastructure Program. Dr Mensah was supported by Australian National Health and Medical Research Council (NHMRC) Early Career Fellowship 1037449.
Competing interest None.
Ethics approval Royal Children's Hospital, Melbourne, Human Research Ethics Committee—HREC 24051.
Provenance and peer review Not commissioned; externally peer reviewed.