Introduction
Exclusion from school is a disciplinary tool used around the world; in the UK, exclusions can be of either fixed period(s) of up to a total of 45 days per academic year, or a permanent expulsion that terminates the child's attendance at the excluding school. The effects and application of exclusion from school remain a contentious issue; previous research has suggested that it is associated with both internalising and externalising psychopathology, as well as poor occupational and academic outcomes (Parker et al. Reference Parker, Whear, Ukoumunne, Bethel, Thompson-Coon, Stein and Parker2014; Whear et al. Reference Whear, Marlow, Boddy, Ukoumunne, Parker, Ford, Thompson-Coon and Stein2014; Obsuth et al. Reference Obsuth, Sutherland, Cope, Pilbeam, Murray and Eisner2017). According to the latest English government figures (Department for Education, 2016a ), the number of permanent exclusions for 2014/15 was 5800 (0.07% of school population), and for fixed term, the number was 302 980 (3.88% of school population). The commonest reason for exclusion was persistent disruptive behaviour. Characteristics of pupils who appear to be over-represented in these statistics include boys, children with special educational needs, eligibility for free school meals (FSM), young people aged 14 plus and those from black and minority ethnic groups (BME; excluding Asian and Chinese).
Childhood psychiatric disorders are common (8–18% of the school age population), persistent, possibly increasingly prevalent and associated with several adverse outcomes including educational failure, adult mental illness, risk-taking behaviour and criminality (Kim-Cohen et al. Reference Kim-Cohen, Caspi, Moffitt, Harrington, Milne and Poulton2003; Collishaw et al. Reference Collishaw, Maughan, Goodman and Pickles2004; Costello et al. Reference Costello, Egger and Angold2005; Ford et al. Reference Ford, MacDiarmid, Russell, Racey and Goodman2017). Population-based studies demonstrate that when psychopathology is measured using a dimensional approach, there is a continuous spectrum of psychological functioning (Ford & Parker, Reference Ford and Parker2016), although Goodman & Goodman (Reference Goodman and Goodman2011) reported a linear association between psychopathology scores and the likelihood of psychiatric disorder. These findings suggest that for every child who meets diagnostic criteria, there will be several others who are struggling. Poor childhood mental health is associated with disruptive behaviour (Vorhaus & Vorhaus, Reference Vorhaus and Vorhaus2012) and poor academic attainment (Copeland et al. Reference Copeland, Angold, Shanahan and Costello2014) both of which may increase the likelihood that a child may be excluded. Previous research (Ford et al. Reference Ford, Goodman and Meltzer2004) reported that socio-economic deprivation, poor general health, family dysfunction, parental psychiatric illness, adverse life events and ethnicity were associated with an increased prevalence of psychiatric disorder. This suggests an overlap between the characteristics of children who are most likely to have a psychiatric disorder and those most likely to be excluded from school. In addition, most child mental health-related contacts with services occur within the education sector, and similar proportions of children with psychiatric disorder access specialist education professionals as attend child and adolescent mental health services with relatively few attending both (Ford et al. Reference Ford, Hamilton, Meltzer and Goodman2007). Two linked systematic reviews revealed a gap in the research literature, with very few studies that have explicitly explored the link between exclusion from school and psychopathology (Parker et al. Reference Parker, Whear, Ukoumunne, Bethel, Thompson-Coon, Stein and Parker2014; Whear et al. Reference Whear, Marlow, Boddy, Ukoumunne, Parker, Ford, Thompson-Coon and Stein2014). Few studies have suggested that exclusion from school may be commoner among children with psychiatric disorder than their mentally healthier peers, while psychopathology is more prevalent and/or severe among children excluded from school as compared with those who are not excluded (Parker et al. Reference Parker, Whear, Ukoumunne, Bethel, Thompson-Coon, Stein and Parker2014; Whear et al. Reference Whear, Marlow, Boddy, Ukoumunne, Parker, Ford, Thompson-Coon and Stein2014). Notably, none of the studies detected by these reviews were primarily focused on this topic.
With this in mind, we undertook a secondary analysis of the British Child and Adolescent Mental Health Survey (BCAMHS) 2004 (Green et al. Reference Green, McGinnity, Meltzer, Ford and Goodman2005) and its 3-year follow-up (Parry-Langdon, Reference Parry-Langdon2008). We predicted that children with poor mental health (regardless of whether they met diagnostic criteria for psychiatric disorder) would be more likely to be excluded from school than their peers in both 2004 and 2007. Similarly, we hypothesised that children who had been excluded from school in 2004 would have poorer mental health in 2007 compared with those children who have not been excluded from school at baseline.
Methods
The original survey had approval from Medical Research Ethics Committees (MREC); the Peninsula College of Medicine and Dentistry granted approval for this secondary analysis.
Sampling strategy and response rates
A representative sample of children and young people aged 5–16 years living in private households in Great Britain was selected from a sampling frame for England, Wales and Scotland using the Child Benefits register (Green et al. Reference Green, McGinnity, Meltzer, Ford and Goodman2005). Child Benefit was at that time a universal benefit payable to British parents for each child, with near 100% take up. Families were excluded if they did not have a valid postcode, lived in postal sectors that were deemed too small (<100 families; 0.25% of addresses) or were considered too sensitive to approach. Coverage of children aged 5–16 years was estimated to be 90%.
Four hundred and twenty-six postal sectors were sampled with a probability related to size of the sector, and stratified by regional health authority and social economic group. Figure 1 describes the recruitment process. Parents (n = 12 294) were invited to take part in the study by letter from the Office for National Statistics. In both surveys, all parents were interviewed, as were children aged 11 and over; when the family agreed, a brief questionnaire was mailed to a teacher nominated by the family. The final sample size at baseline (2004) was 7977, which represented 65% of those approached. In the follow-up study in 2007, 73% of the 7329 parents who were contacted completed interviews; the final sample size at follow-up was 5326 (72% response rate).
Fig. 1. Flow diagram of participant recruitment to the British Child Mental Health Survey in 2004, and its follow-up in 2007.
Young people were aged between 7 and 19 years at follow-up, with 1704 aged 16 and over; of these 469 (28%) were reported to be in full time education in 2007.
Measures
Psychopathology was measured using the Strengths & Difficulties Questionnaire(SDQ) (Goodman, Reference Goodman2001, http://www.sdqinfo.org) and the Development and Well-Being Assessment (DAWBA) (Goodman et al. Reference Goodman, Ford, Richards, Gatward and Meltzer2000) in both surveys.
All parents, teachers and children over 11 years were invited to complete the SDQ, which is a measure of common childhood psychopathology, validated across multiple populations (Goodman, Reference Goodman2001, http://www.sdqinfo.org). The SDQ comprises of 25 items that make up five sub-scales, which include emotional symptoms, conduct problems, hyperactivity/inattention, peer problems and pro-social behaviour. A total difficulties score is calculated by adding the sub-totals from the first four sub-scales. The SDQ impact supplement asks the informant whether they consider the child to have a significant mental health problem and if so how long any difficulties have been present, distress to the child, the impact for the child on their home life, friendships, classroom learning and leisure activities, and the burden on the informant. Questions for teachers exclude home life and leisure activities.
The DAWBA (Goodman et al. Reference Goodman, Ford, Richards, Gatward and Meltzer2000) is a standardised diagnostic interview that combines both structured and semi-structured features; information is collected from parents, young people aged 11 or over and teachers. The structured questions relate to diagnostic criteria in Diagnostic and Statistical Manual of Mental Disorders Fourth Edition (DSM-IV) (American Psychiatric Association, 1994) and International Classification of Diseases, 10th Edition (ICD-10) (World Health Organisation, 1993), which are complemented by a series of open-ended questions where problems were identified. The quantitative and qualitative information from all available informants was reviewed by a small team of experienced child psychiatrists (including TF and RG) who assigned psychiatric diagnoses according to DSM-IV classification (American Psychiatric Association, 1994). Each rater worked independently with regular group discussion of complex and borderline cases. These were reviewed by the programme developer (RG) for consistency. The κ statistic for chance-corrected agreement between two clinicians who independently rated 500 children was 0.86 for any disorder (standard error s.e. 0.04), 0.57 for internalising disorders (s.e. 0.11) and 0.98 for externalising disorders (s.e. 0.02). The test–retest reliability of the DAWBA has not been ascertained as it is doubtful that you could obtain valid responses to such a detailed assessment over a short enough period of time to ensure that the child's symptoms had not changed. Having a small clinical team made it easier to maintain diagnostic consistency.
Exclusion from school
Both surveys asked parents ‘Has [Name Child] ever been excluded from school’ (Green et al. Reference Green, McGinnity, Meltzer, Ford and Goodman2005; Parry-Langdon, Reference Parry-Langdon2008); the parent could respond ‘yes’ or ‘no’. The 2007 survey asked a series of questions about the type, reason and length of exclusion and what educational provision the child received afterwards. As these details were not available for 2004 and there were so few permanent exclusions reported in 2007 (see Table 1), all reported exclusions were analysed together. Exclusion status was classified into four groups: no exclusion in either survey (n = 3879), exclusion in both surveys (n = 54), exclusion in 2004 only (n = 19) and exclusion in 2007 only (n = 129).
Table 1. Summary of the descriptive statistics of those who had been excluded in 2007
a One child was missing for this question.
b Three children were missing data for this question; the survey question did not specify in relation to which exclusion if more than one and parents could endorse one option.
c One child was missing for this question; the survey question did not specify in relation to which exclusion if more than one and parents could endorse one option.
Background information
Demographic details, such as family type, ethnicity, parental educational qualifications and weekly household income, were obtained from the interview with parents. Housing tenure was grouped into whether families owned or rented their accommodation.
Neighbourhood environment was assessed using the ACORN (A Classification Of Residential Neighbourhoods; CACI Information Services, 1993). Parents rated their child's general health using a five-point Likert scale from very good (1) to very bad health (5), which was dichotomised by combining reports of very good and good health (n = 7401) or poor health (fair, bad and very bad; n = 464). Parent's mental health was measured using the 12-item General Health Questionnaire (GHQ, Goldberg & Williams, Reference Goldberg and Williams1988) with a cut point of 3 or more to indicate distress (Green et al. Reference Green, McGinnity, Meltzer, Ford and Goodman2005).
A child was deemed to have a learning disability if one or both of the parents or teachers had estimated that the child's mental age was 60% of the chronological age or less (e.g. a mental age of 6 or less at a chronological age of 10) (Liddle et al. Reference Liddle, Batty and Goodman2009). Teachers also provided information about the child's level of attainment in comparison to their peers. This was coded into a binary variable no learning disability or moderate/severe learning difficulty (n = 7768, 161 respectively).
Analysis
Descriptive statistics
Analysis was conducted using STATA 13.0 (StataCorp, 2013). Tests of association between categorical variables and exclusion status were conducted using χ2 tests; with one-way analysis of variance for continuous variables. Trends in parental total difficulties SDQ scores both at baseline and follow-up were explored. Logically all the children with parents who reported exclusion in 2004 should have been reported to have had exclusions in 2007 as the question asked about ‘ever’, but as some parents only reported exclusion in 2004, these four groups were analysed separately at the bivariable level.
Adjusting for survey design and probability weights
Sampling weights adjusted for the probability of postal sector selection in the sampling frame and to compensate for differential response rates by region and strata at the time of the initial survey in the reported prevalence estimates. The remaining analyses were conducted on unweighted data because analyses of the initial BCAMHS showed very small design effects on most estimates (Heyman et al. Reference Heyman, Fombonne, Simmons, Ford, Meltzer and Goodman2001). Importantly, for an outcome that might cluster in families, only one child per family was selected.
Regression models
Unadjusted models were fitted to establish the impact of individual factors on the outcome of exclusion from school (logistic regression) or psychological distress measured by parental SDQ total difficulties score (linear regression). Multi-variable regression models controlled for relevant confounding factors suggested by the background literature (Hayden, Reference Hayden1997; Hayden & Dunne, Reference Hayden and Dunne2001; Daniels et al. Reference Daniels, Cole, Sellman, Sutton, Visser and Bedward2003; Hemphill et al. Reference Hemphill, Toumbourou, Smith, Kendall, Rowland, Freiberg and Williams2010; Parsons, Reference Parsons2010; Skiba et al. Reference Skiba, Horner, Chung, Rausch, May and Tobin2011). The expected probability of exclusion from school was calculated for baseline SDQ score on exclusion at follow-up, stratified by gender because of the over-representation of boys among children who are excluded, and adjusted for other detected independent predictors to avoid overestimating the influence of psychological distress. A backwards stepwise approach was adopted where non-significant variables were individually removed until all variables retained were significant, aside from gender, age and ethnicity. The potential confounding variables considered included baseline parental mental health, exclusion status at baseline, age, household occupation, neighbourhood deprivation, household income, ethnicity, parental general mental health, mother's highest educational qualification, general health of the child and general learning disability of the child. This analysis omitted children who were excluded only in 2004 (n = 19) and those excluded at both time points (n = 54) as we wanted to test the relationships of baseline mental health on future exclusion. Interactions were studied between gender and age. The comparison group was children who had not been excluded from school. The SDQ score and psychiatric diagnosis were not included as covariates when the other was the outcome due to collinearity.
Prospective models were based on new exclusions/diagnosis in 2007. The wording of the question parents were asked about their child's exclusion did not distinguish exclusions that predated 2004 (baseline) from those that had occurred between the surveys. Thus, children who were excluded only in 2004 or at both time points were omitted from these analyses. Equally those who had a disorder only in 2004 or at both time points were absent from these models.
Results
Prevalence of exclusion from school of the overall dataset
At baseline, 3.9% (n = 313) of the sample had been excluded; 75% were boys (n = 236) and most were aged 11–16 years (87.5%, n = 274). At follow-up, 4.5% (n = 183) reported exclusion, of which 70% (n = 129) were ‘new’ exclusions; over half of the children who had been excluded by 2007 had experienced more than one exclusion (Table 1), although permanent exclusions were uncommon (10% of those excluded). In addition, 14% had moved from the school that excluded them, whereas only a third reported that they received additional support after the exclusion.
Description of the sample according to exclusion status
In both surveys, the experience of exclusion was commoner among boys, secondary school pupils and those with socio-economic deprivation, but the expected relationship with BME was not detected (Table 2). Poor child general health and learning disability and poor parental mental health were also associated with exclusion at both time points.
Table 2. Unadjusted characteristics of the sample at baseline in relation to exclusion status across both surveys
Exclusion status and SDQ parental total difficulties
Figure 2 demonstrates consistently high levels of psychological distress among those who had experienced exclusions at both time points that exceeded the commonly quoted clinical cut point of 16 (http://www.sdqinfo.org). Mean parental SDQ scores were raised at the time that data were gathered among children with exclusions reported only once and the levels of psychological distress were consistently higher among children reported to have experienced exclusion at any time point compared with their non-excluded peers.
Fig. 2. Mean parental SDQ score by exclusion status in both surveys.
Mean SDQ total difficulties score at baseline was associated with exclusion from school in cross-sectional analysis [odds ratio (OR) 1.16, 95% confidence interval (CI) 1.14–1.18] and prospectively (OR 1.11, 95% CI 1.08–1.14) as illustrated by Table 3. A significant interaction was detected between the age and psychological distress in the cross-sectional analysis (but not longitudinally) with exclusion (adjusted OR = 0.93, 95% CI 0.88–0.97, p = 0.002). For every point increase in the SDQ, the odds of exclusion increased by 15% among those aged 11–15 years compared with 23% of those aged 5–10 years.
Table 3. The impact of baseline characteristics on the child's likelihood to be excluded at follow-up
Neighbourhood, weekly income parents’ mental health, child's general health and learning disability status were also tested in this model but were not significant and therefore removed.
The probability of exclusion at follow-up (see Fig. 3) was based on an adjusted model presented in Table 3. This graph presents the predicted probability for boys (other variables coded as 0 to avoid overestimating the probability of exclusion by omitting the influence of other independent predictors) and suggests that the likelihood of exclusion from school at follow-up accelerates from about a score of 20 on the SDQ. Data for girls were too sparse to present a similar graph.
Fig. 3. Predicted probability of exclusion at follow-up from baseline parental SDQ total difficulties score among boys (based on adjusted analysis with other independent predictors coded to 0 to avoid over estimating the influence exclusion).
Prospective unadjusted models showed that children excluded from school at baseline have significantly higher SDQ scores at follow-up [β coefficient 6.76 (5.85–7.66) p < 0.001], and higher odds of a new psychiatric disorder [OR 7.09 (5.07–9.91) p < 0.001], compared with children who had not been excluded from school in 2004. This association remained after controlling for potential confounders (Table 4).
Table 4. The impact of baseline exclusion status on the parent-reported Strengths and Difficulties Questionnaire total difficulty scores in 2007
Discussion
We found associations with psychopathology in BCAMHS 2004 and 2007 among children excluded from education. High levels of psychological distress were consistently detected among excluded children, while baseline psychopathology was a significant predictor of a child's likelihood of being excluded despite adjusting for common correlates of exclusion (Hayden, Reference Hayden1997; Hayden & Dunne, Reference Hayden and Dunne2001; Daniels et al. Reference Daniels, Cole, Sellman, Sutton, Visser and Bedward2003; Hemphill et al. Reference Hemphill, Toumbourou, Smith, Kendall, Rowland, Freiberg and Williams2010; Parsons, Reference Parsons2010; Skiba et al. Reference Skiba, Horner, Chung, Rausch, May and Tobin2011). Furthermore, the impact of psychopathology on the likelihood of being excluded was greater when experienced at a younger age. Exclusion from school was likewise associated with increased psychopathology. This bi-directional association suggests that remediation and support for children whose behaviour challenges school systems is important. Timely intervention may prevent exclusion from school as well as future psychopathology. The notion of early identification of difficulty for children who are struggling is acknowledged throughout literature and policy (Kim-Cohen et al. Reference Kim-Cohen, Caspi, Moffitt, Harrington, Milne and Poulton2003; Taggart et al. Reference Taggart, Sammons, Smees, Sylva, Melhuish, Siraj-Blatchford, Elliot and Lunt2006; Department for Education and Department of Health, 2014), while studies suggest that early intervention may have a beneficial impact (Beckett et al. Reference Beckett, Beecham, Doolan, Ford, Kallitsoglou, Scott and Sylva2010; Patton et al. Reference Patton, Coffey, Romaniuk, Mackinnon, Carlin, Degenhardt, Olsson and Moran2014).
Boys, secondary school pupils, and children from a deprived socio-economic background, or with poor general health or learning disabilities were significantly more likely to be excluded at both time points. These factors correlate with government statistics for the same years, although we failed to detect significant association with BME status shown repeatedly in national statistics (Department for Children, Schools and Families, 2005, Department for Children, Schools and Families, 2009; Department for Education, 2013, 2016a ). This may be due to the small numbers of children from ethnic minorities in the sample analysed. In addition, parental mental health disorders were also related to exclusions, which are less readily gauged from governmental statistics.
At baseline, 3.9% of the BCAMHS sample reported an exclusion compared with 4.7% of the national school population for the same period. This difference may be due to selection bias; the participation and drop-out rates in cohort studies of populations with psychiatric disorders is high, although the impact of drop-outs on the validity of regression models may be less than commonly believed, or indeed negligible (Wolke et al. Reference Wolke, Waylen, Samara, Steer, Goodman, Ford and Lamberts2009). In keeping with the contemporaneous national statistics, permanent exclusions in the 2007 follow-up were a rare event and most children in the sample remained at the school that excluded them. However, parents reported little support following the exclusion. This may be due to a failure of communication of reintegration strategies to parents and/or a lack of engagement by the child and parent with support that was offered, but suggests that there is considerable room for remediation that might reduce the number of children who experience multiple exclusions from school.
Few epidemiological studies have explored the impact of mental health on school exclusions. Those published have demonstrated associations between children with impairing psychopathology and exclusion from school, particularly among children with ADHD (Barkley et al. Reference Barkley, Anastopoulos, Guevremont and Fletcher1991; Rohde et al. Reference Rohde, Biederman, Busnello, Zimmermann, Schmitz, Martins and Tramontina1999; Norwich, Reference Norwich2002; Bauermeister et al. Reference Bauermeister, Shrout, Ramírez, Bravo, Alegría, Martínez-Taboas, Chaves, Rubio-Stipec, Ribera and Canino2007; Miller et al. Reference Miller, Nevado-Montenegro and Hinshaw2012) and depression (Meyer et al. Reference Meyer, Garrison, Jackson, Addy, McKeown and Waller1993; Rushton et al. Reference Rushton, Forcier and Schectman2002). Our findings reinforce the need for larger longitudinal studies to investigate these links in greater depth.
Given the established link between children's behaviour, classroom climate and teachers’ mental health, burn out and self-efficacy, greater availability of timely support for children whose behaviour is challenging might improve teachers’ productivity and school effectiveness (Aronsson et al. Reference Aronsson, Svensson and Gustafsson2003; Maguire & O'Connell, Reference Maguire and O'Connell2007; Kidger et al. Reference Kidger, Brockman, Tilling, Campbell, Ford, Araya, King and Gunnell2016). Current guidance from the Department for Education (2016b ) focuses on authoritarian approaches to discipline and disruptive behaviour. In contrast, evidence-based programmes for conduct disorder emphasise the effectiveness of clear rules and instructions combined with promotion of positive behaviour through praise and encouragement (National Institute for Health & Care Excellence, 2013; Whear et al. Reference Whear, Thompson-Coon, Boddy, Ford, Racey and Stein2013). In contrast, current policy guidance also specifically recommend exploring whether continuing disruptive behaviour is a sign of unmet needs, and a number of vulnerable children may face exclusion from school that might be avoided with suitable interventions (Donno et al. Reference Donno, Parker, Gilmour and Skuse2010; O'Regan, Reference O'Regan2010). There is also an increasing focus on the promotion of mental health and well-being in schools (Department of Health, 2015; Department for Education, 2016b ; House of Commons, 2017) with recommendations to improve communication between schools and child mental health services; schools are encouraged to undertake needs assessment, planned support and regular review with changes where necessary for pupils with poor mental health (Department for Education, 2016b ). Early detection is a key theme, highlighting a need for teachers to have a low threshold to refer for specialist educational needs services. Specifically, the guidance refers to the use of the SDQ to aid detection and referral. This approach is potentially unethical if CAMHS or specialist educational needs services lack the capacity to respond and/or school budgets, are not allocated to support the recommendations made after specialist assessment. Previous work conducted by this team (Parker et al. Reference Parker, Marlow, Kastner, May, Mitrofan, Henley and Ford2016a ) suggests that children whose poor mental health is recognised by parents and/or teachers are MORE likely to be excluded than those whose psychiatric disorder is not recognised. Early identification without adequate support will be insufficient.
Parents report that teachers are the most commonly contacted ‘service’ in relation to children's mental health (Ford et al. Reference Ford, Hamilton, Meltzer and Goodman2007; Newlove-Delgado et al. Reference Newlove-Delgado, Ukoumunne, Stein and Ford2015). In the 1999 BCAMHS, similar numbers of families accessed mental health as did specialist education resources with little overlap between access to the two services (Ford et al. Reference Ford, Hamilton, Meltzer and Goodman2007). The additional mental health-related activities imposed substantial costs on schools (£799.2 million using 2008 prices) and specialist educational services (£508.8 million), which greatly exceeded those to other public sector services (£162.8 million for health and welfare combined; Snell et al. Reference Snell, Knapp, Healey, Guglani, Evans-Lacko, Fernandez, Meltzer and Ford2013). Marked inter-individual variation in costs suggests inefficiencies in the use of resources (Knapp et al. Reference Knapp, Snell, Healey, Guglani, Lacko-Evans, Ferbandez, Meltzer and Ford2015). Anecdotally, these costs are mostly sunk in internal and/or multi-agency meetings rather than therapeutic activity; the diversion of professional time from meetings could potentially therefore improve outcomes without additional overall costs to the education system. While some economists would argue that the time involved for school staff is not an additional cost, it is certainly an opportunity cost as it diverts them away from alternative educational activity. Furthermore, characteristics other than the severity of psychological distress predicted service costs, and included some tractable issues, such as reading attainment and parental psychopathology (Knapp et al. Reference Knapp, Snell, Healey, Guglani, Lacko-Evans, Ferbandez, Meltzer and Ford2015). Effective reading remediation or the active treatment of parental depression might also support the recovery of some children's mental health and may reduce the burden of mental health-related demands on the education system.
These analyses benefit from the large nationally representative sample, validated measures and prospective follow-up, but secondary analyses are constrained by the data available and the original questions asked. For example, the question to parents about types of exclusion and educational and other provision after exclusion did not specify which exclusion for children who experienced more than one, and while the options for educational provision are mutually exclusive, the options for mental health provision are not (see Table 1). As more than half reported multiple exclusions, the reports of no access to educational or mental health provision are even more striking. Similarly, we had no measure of eligibility or uptake of FSMs, although we did have access to multiple other socio-economic indicators. Exclusion from school results from a complex interaction of factors (Parsons, Reference Parsons2010; Parker et al. Reference Parker, Paget, Ford and Gwernan-Jones2016b ); including social, family and community issues in addition to mental health and learning. Adjustments in the models were made to account for some of these factors, but the direction of influence in relation to the impact of mental health on exclusion from school and the effect exclusion from school had on children's mental health are difficult to untangle. Data on the timing of exclusions and additional time points would have offered the potential to conduct a survival analysis, while data on the number of exclusions would have permitted a more nuanced descriptive analysis.
Not all parents consented to contact with schools, and not all teachers contacted responded; hence, our decision to use parent-reported psychopathology to allow more children to be included in the regression models, but this may not reflect the child's function in the classroom, which parents do not directly witness. Studies have shown relatively low inter-informant agreement about childhood psychopathology, which may have been present here (Achenbach et al. Reference Achenbach, McConaughy and Howell1987; Collishaw et al. Reference Collishaw, Goodman, Ford, Rabe-Hesketh and Pickles2009). Ideally parent-reported exclusions would be supplemented with teacher and child reports or links to administrative data. Parents may under-report, given the stigma surrounding exclusion from school, but this may be balanced by reporting of unofficial/illegal exclusions that would not be included in official statistics (Children's Commissioner, 2013). Indeed, as 19 children were reported to be excluded ‘only’ in 2004 when the question at both time points asked if a child had EVER been excluded demonstrates this, although these children did have lower levels of psychopathology reported by their parents at follow-up than those whose parents reported exclusions at both time points.
In summary, we detected evidence of an independent bi-directional association between child mental health and exclusion from school that suggests that prompt assessment and suitable support for children whose behaviour challenges their school placement may both avert some exclusions and improve the child's mental health.
Key points
-
• Exclusion from school is a common disciplinary procedure, and although there is a suggested link between childhood psychopathology and exclusion, there is a lack of research focussed on this topic.
-
• Our study shows a bi-directional relationship between exclusion from school and psychopathology in children seen in a large population-based survey of childhood mental health in Great Britain and its follow-up 3 years later.
-
• Prompt identification and intervention to support children suffering psychological distress and demonstrating challenging behaviour may avert exclusions and improve their future mental health.
-
• Given the lack of large-scale longitudinal studies into exclusion and childhood mental health, our research reinforces the need for more in-depth studies addressing these issues and testing the effectiveness/cost-effectiveness of intervention.
Acknowledgements
Claire Parker's Ph.D. studentship was supported by the National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health Research and Care South West Peninsula. Javid Salim worked on this paper as an Academic Clinical Fellow, also funded by NIHR. The initial surveys were funded by the English Departments of Health with contributions from their Scottish and Welsh counterparts, and data collection was led by the Office for National Statistics. The authors would like to thank the children, their parents and their teachers, as well as our colleagues at the Office for National Statistics, particularly Howard Meltzer, for their role in the original surveys.
Declaration of Interest
Robert Goodman is the owner of Youthinmind Limited, which provides no-cost and low-cost websites related to the DAWBA and SDQ.
