First steps towards a mental health and neurodevelopmental screening of secondary school children following two fixed-term school exclusions in the UK

ABSTRACT Children excluded from school often have a range of unidentified needs that may directly contribute to their behaviour. Research and policy highlight the absence of screening for difficulties in children at risk of exclusion. This study aimed to assess and explore the feasibility and acceptability of neurodevelopmental and mental health screening of secondary school children with two or more fixed-term exclusions, as well as compare differences in rates to matched controls. Mental health and neurodevelopmental measures were collected from 40 children aged 11–16, 13 parents and 9 teachers. The screening process was feasible and acceptable for the children, however there was limited informant response. Case participants showed elevated levels across all screening measures compared to matched control participants. This study highlights the range of difficulties experienced by children at risk of exclusion from school and suggests a screening process may help to identify underlying neurodevelopmental and mental health difficulties.


Introduction
School exclusions are used as a disciplinary sanction issued in response to students' misbehaviour and involve the removal of a student from either regular teaching or the school premises (Valdebenito et al. 2018). In the United Kingdom (UK), a student can receive a fixed-term exclusion whereby they are temporarily removed from the school or classroom for a set period of time, up to 45 school days of the academic year, or a permanent exclusion in which they are expelled and cannot return to the school where they enrolled (Department for Education 2017a). The most common reason for exclusion from school in England is persistent disruptive behaviour, followed by physical assault to a student/teacher and verbal abuse or threatening behaviour (DfE 2018). National educational data suggests that school exclusions in the UK have been on the rise since 2012. The data also suggests that certain groups of school aged children are consistently overrepresented in school exclusion figures, including: boys, those eligible for free school meals, looked-after children, black and minority ethnic groups, and children with special educational needs (SEN) (DfE 2020).
Of increasing concern is the high rate of exclusion among children who have neurodevelopmental or mental health difficulties. A recent UK government review into school exclusion suggested that children aged 5 to 16 with SEN were seven times more likely to be excluded than those without (Timpson 2019). Within this group, pupils with social, emotional and mental health needs experience the highest rate of exclusions as well as those with autism spectrum disorders, attention-deficit hyperactivity disorder (ADHD), conduct disorder, speech, language and communication needs, and those with a moderate or specific learning difficulty (Department for Education 2018; Parker et al. 2019), suggesting that schools may not be adequately meeting the needs of these children. Children with SEN who are not in receipt of additional school support are more likely to be excluded than those with additional support in place (Timpson 2019).
There have been few empirical studies looking at the association between neurodevelopmental and mental health difficulties and exclusion from school. Parker et al. (2016) explored neurodevelopmental and mental health needs among primary school children aged 4 to 12 who had received exclusions from school, reporting higher rates of emotional disorders, conduct disorder, ADHD, and autism. Interestingly, findings showed that nearly all these children had needs that were recognised by teachers and parents; however, very few of them were accessing appropriate services, thus lacking support from mental health services. This builds on a previous study that examined traits of autism in primary school children aged 6 to 13 who had a history of permanent or fixed-term exclusion (Donno et al. 2010). The results indicated that many of the excluded children had unidentified social communication impairment, with 35% meeting criteria for autism. A further study looking at language impairment in boys aged 8 to 16 who had been excluded from school found that excluded boys showed greater language impairment and higher rates of previously unidentified language problems than controls (Ripley and Yuill 2005).
Although there is a growing body of evidence suggesting an association between neurodevelopmental and mental health difficulties and exclusion from school, there are likely to be many more children at risk of being excluded who may have similar -but unidentified -social, emotional or behavioural needs. Furthermore, a lack of early intervention and support is considered likely to contribute to children with unmet SEN being excluded from school. The necessity to identify and support the needs of children at risk of exclusion is imperative given the associated negative outcomes, including poor academic outcomes, antisocial behaviour, substance misuse, homelessness, unemployment, mental health difficulties and crime in later life (Hemphill et al. 2006;Fothergill et al. 2008;Noltemeyer, Ward, and Mcloughlin 2015;Ford et al. 2018;Heerde et al. 2020;Rosenbaum 2020;Madia et al. 2022).
In line with rates found among excluded children, studies have also shown a consistently higher incidence of both mental health and neurodevelopmental needs amongst young offenders (Kroll et al. 2002;Teplin et al. 2002;Hughes et al. 2012;Chitsabesan and Hughes 2016). Hughes et al. (2012) suggest that there are likely to be high numbers of children currently within the Criminal Justice System (CJS) in England who have undiagnosed neurodevelopmental conditions that have led them to becoming involved in the CJS. They suggest that a lack of identification and holistic screening has led to a significant level of unmet needs and a 'pathway' into offending.
Given the over-representation of children with neurodevelopmental and mental health needs in the criminal justice system, in 2015 the British Psychological Society (BPS) issued a position paper suggesting that screening for neuro-disabilities should take place within education and community settings, such as at the time of a second fixed-term exclusion from school (BPS 2015). Following on from the BPS' call for earlier screening, the Departments of Health and Educations Green Paper (2017b) encouraged schools to focus on early identification of mental health needs.
Both educational and health policy indicates that early identification of neurodevelopmental and mental health difficulties is essential in understanding reasons for behavioural difficulties, offering timely support, and preventing adverse outcomes such as school exclusion and later offending (Department for Health and Department of Education, 2017b). Nevertheless, there has been little empirical investigation on this topic to date. Three case-control studies (Ripley and Yuill 2005;Donno et al. 2010;Parker et al. 2016) have looked at psychopathology, development, social communication, language and learning difficulties among children excluded from primary school. However, the feasibility of carrying out a screening process for neurodevelopmental and mental health difficulties has not been explored with children of secondary school age. This is important to investigate as exclusion rates in secondary schools are much higher (Department for Education 2020).
Given that previous research has not been conducted directly within school settings, a feasibility and acceptability study to trial the research process and methods is imperative prior to conducting a larger scale study or implementing the screening process more widely within schools. Routinely, neurodevelopment and mental health screening/assessments are carried out within CAMHS settings; however, given the Departments of Health and Education's Green Paper (2017b) drive to increase access to mental health support within schools as well as to aid early identification, it would be of value to see whether schools and clinicians can work collaboratively to implement this process. Accordingly, the current study aims to explore the feasibility and acceptability of neurodevelopmental and mental health screening of secondary school children who receive two or more fixedterm exclusions from secondary school in the UK. In addition, the study aims to explore the differences in neurodevelopmental and mental health difficulties of secondary school children with two or more fixed-term exclusions compared to matched children without a fixed-term exclusion, as would be expected from the literature.

Hypotheses
(1) The screening process will be feasible in terms of recruitment, retention and completion of informant measures.
(2) Secondary school children, parents and teachers will find the screening process acceptable.
(3) Secondary school children with two or more fixed-term exclusions will show higher scores on screening measures of neurodevelopmental and mental health difficulties relative to matched children from the same school without fixed-term exclusions.

Study design
Ethical approval was granted by King's College London Psychiatry, Nursing and Midwifery Research Ethics Subcommittee (Reference number: HR-18/19-8488). A prospective case-control design was used, comparing secondary school children who had received two or more fixed-term exclusions to a matched control group who had not received exclusions. A mixed quantitative and qualitative design was utilised to collect feedback about the acceptability and feasibility of the screening process, using a questionnaire. The measures of feasibility and acceptability were developed by the main researcher (trainee clinical psychologist) and supervisor (consultant clinical psychologist to gather information that would help to inform implementation on a wider scale.

Participants
Forty secondary school children participated: 20 case participants who had received two or more fixed-term exclusions and 20 matched control participants without fixed-term exclusions. Inclusion criteria included being between the ages of 11-16 years and sufficient proficiency in English to understand and complete the assessment screening. Participants were eligible for the case group if they had received two or more fixed-term exclusions from school, operationalised as one or more days where a student had been removed from the mainstream classroom. This included internal and external exclusions as well as a move to an alternative provision for a fixed period. The reason for exclusion had to be behavioural (verbal/physical aggression, persistent disruptive behaviour, damage to property). Participants were eligible for the control group if they had not received any school exclusions. Control participants were recruited from the same school and matched on age and gender. Thirteen parents and 9 teachers completed informant measures.

Demographics
Participant demographic information was collected, including age, ethnicity, gender, first language, contact with services and mental health diagnoses. Case participants and parents were asked about number of and reasons for school exclusion.

Feasibility
To assess feasibility the following data was collected: rates of fixed-term exclusions within the school; number of secondary school children within each school who were eligible; number of children contacted to take part; rates of consent from children and parents; reasons why children/parents declined to take part; rates of completion of the screening assessment; reasons why any children/ parents dropped out of the study; rates of completion of the informant questionnaires; time taken for the screening process; number of schools approached, how many took part and reasons for declining.

Acceptability
Acceptability was assessed with questionnaires consisting of quantitative questions on Likert scales and qualitative free text responses. Questionnaires assessed satisfaction and understanding of the screening assessment, the acceptability of the timing as well as recommendations for improvements.
Informants were also asked about ease of completion of the questionnaires.

The Strengths and Difficulties self-report, parent and teacher questionnaires (SDQ; Goodman 1999)
The SDQ is a brief, 25 item screening questionnaire that can be used for children aged four to sixteen. The items are divided into five subscales; conduct, hyperactivity, emotional symptoms, peer problems and prosocial behaviour. The total difficulties score ranges from 0-40 with a score of 20 or more suggesting distress. An impact supplement assesses the presence of emotional, behavioural, concentration and peer difficulties as well as the impact and burden of these on the child's life. Scores can be categorised into four bands (close to average, slightly raised, high and very high). Higher scores on the conduct, hyperactivity, emotional symptoms and peer problems subscales reflect a higher presence of difficulties, whilst lower scores on the pro-social behaviour subscale reflect a higher presence of difficulties.

Revised Child Anxiety and Depression Scale -Self Report, Short Version (RCADS-25; Ebesutani et al. 2012)
The RCADS-25 is a 25-item scale that assesses the frequency of symptoms of anxiety and low mood and produces a total score for each. It is scored on a four-point Likert scale and can be completed by children aged eight to eighteen. Scores are converted into age and gender normed T-scores with a score of ≥ 65 falling within the 'borderline' range and a score of ≥ 70 'above clinical threshold'.

Conners-3 Self Report, Parent and Teacher Short Forms (Conners et al. 1997)
The Conners-3 is used to assess ADHD symptoms and common co-morbid problems in children and adolescents aged six to eighteen. The questionnaire consists of around 40 questions mapping onto 6 constructs; inattention, hyperactivity/impulsivity, learning problems, aggression, executive functioning (parent version only) and peer/family relations. After scoring, T-scores are provided, with a score of ≥ 65 suggesting there are 'more concerns' and a score of ≥ 70 suggesting that there are 'many more concerns' than expected for their age.

The Wechsler Abbreviated Scale of Intelligence -second edition (WASI-II; Wechsler and Hsiaopin 2011)
The WASI-II is used to measure cognitive ability for individuals aged six to ninety, consisting of four subtests: vocabulary; similarities; block design and matrix reasoning. Four composite scores are provided, which estimate verbal comprehension, perceptual reasoning abilities, a full-scale IQ and an abbreviated IQ. A full-scale IQ score of ≤ 70 is indicative of an intellectual disability.

Clinical Evaluation of Language Fundamentals -Version 4 (CELF-4; Semel, Wiig, and Secord 2003)
The CELF-4 can be used for children between five and sixteen and assesses both expressive and receptive language skills, in order to identify language disorder or delay. Four subtests make up the core language score which quantifies overall language performance and indicates whether a language disorder may be present.

Children's communication checklist 2 (CCC-2; Bishop 2003)
The CCC-2 is a parent-report measure for language impairment, communication problems and autism. It can be used for children from four to sixteen years. The CCC-2 consists of 10 scales resulting in two composite scores; the general communication composite (GCC) which identifies children with a communication problem and the social interaction deviance composite (SIDC) which highlights a communication profile characteristic of autism. The secondary school children completed the WASI-II, CELF-4, RCADS-25, SDQ and Conners-3. Parents completed the SDQ, CCC-2 and Conners-3 and teachers completed the SDQ and Conners-3.

Procedure
For pragmatic reasons, forty-three mainstream secondary schools in inner London were initially contacted by email, with eleven expressing interest in finding out more about the study. Following further contact and meetings, two mainstream, state secondary schools agreed to participate. Students who had received two or more fixed-term exclusions were identified by their teachers, given an information sheet and invited to take part in the study. If the students showed an interest, their parents/carers were also provided with the information sheet and consent form. Once parental consent was gained, a time was arranged at the student's school to conduct the screening assessment. Participants were briefed by the researcher before providing written informed assent or consent. Control participants followed the same procedure following case participant recruitment.
All participants completed a screening assessment consisting of standardised cognitive assessments and psychometric questionnaires. Self-report questionnaires were completed by the participants themselves, whilst the cognitive and language assessments were administered by the lead researcher to all participants. Participants were given a £10 voucher for their time. Parents and teachers were also contacted to complete informant screening questionnaires. Questionnaires were posted to parents to complete by paper. Mental health questionnaires completed by the secondary school children were screened by the main researcher and their supervisor immediately after completion to ensure timely, appropriate support was offered if needed. Specifically, if the children scored above the clinical cut-off or within the 'very-high' range on one or more of the measures, a letter was sent to parents within two weeks suggesting that further support may be required and they were encouraged to discuss this with the school or their GP. Parents could also contact a member of the research team for further advice. In addition, if a young person answered 'Sometimes', 'Often' or 'Always' to the statement 'I think about death', the researcher asked the young person to describe the type of thoughts to ensure there was no suicidal ideation. If there had been, a full risk assessment would have been carried out and this information would have been shared with parents/carers and the school. The schools' policies and safeguarding protocols were followed at all times.

Data analysis
Descriptive statistics were used to assess feasibility and the quantitative acceptability data. The qualitative data were themed where possible and combined into a synthesis of free text responses. Acceptability was measured by assessing whether 80% of participants and informants indicated that they were satisfied with the process and whether they would recommend others to take part. No standardised cut off point for acceptability was identified in previous literature, therefore 80% was agreed by the main research and supervisor based on clinical judgement and practicality.
The main purpose of the study was to explore feasibility and therefore it was not powered to detect significant differences between the case and control group. However, preliminary differences between groups were explored using T-tests and Mann Whitney U tests. Effect sizes were calculated using Cohen's d (0.2 = small effect, 0.5 = medium effect, 0.8 = large effect) and eta square (n 2 ) (0.01 = small effect, 0.06 = medium effect, 0.12 = large effect).
All analysis was conducted using IBM SPSS statistical software (version 25).

Participant demographics
Case and control demographics are presented in Table 1. Between group analysis (T-tests and Chi Square tests) revealed no significant differences based on gender, age, or ethnicity. However, of note is the higher percentage of Black participants (45%) in the case group, which was three times that of the control group (15%). Three case and three control participants had previous diagnoses: dyslexia, anxiety and ADHD within the case group, and dyspraxia, dyslexia and autism within the control group. Reasons for exclusion are presented in Figure 1. Persistent disruptive behaviour as well as low level deviance, such as failure to comply with school rules and disrespecting teachers, were the most common reasons for exclusion, followed by physical fighting with other pupils.

Feasibility
A flow diagram of participant recruitment is presented in Figure 2. Of the 43 schools contacted, 26% responded to the initial email. Seventy-three percent of these schools subsequently did not respond following emails with further details. Meetings were held with three schools to discuss the project, with two agreeing to participate and one declining due to lack of resources to support the project. For the control participants, 100% of the children recruited consented to and completed the assessment, whilst 77% of the recruited case participants consented to and completed the assessment. Reasons for case participants declining to participate included the assessment being too long (n = 1) and too many stressors in their life (n = 1). Data for four participants was excluded following assessment due to inadequate proficiency in understanding English. Parent response rate for completed informant questionnaires was 32.5%, whilst the teacher response rate was 45%. On average, the screening assessment took 90 minutes, ranging from 75 to 110 minutes. The modal time for teacher and parental questionnaire completion was 10 and 20 minutes respectively, ranging from five to twenty minutes and five minutes to one hour.

Acceptability
All students completed feedback and rated the screening as acceptable. When asked about anything they did not like about the assessment or anything that could have been done differently, two participants reported that it was boring, one said it was hard, and two other participants commented on individual tests which could be changed. Two case participants suggested it would be useful to ask participants more about their experience and why they behave in a certain way. Ninety percent of students reported that the length of the assessment was acceptable, 75% reported that they understood the purpose of the assessment, and 90% said they would recommend a friend to take part. Those who responded 'no' all suggested that their friends would not like it.
Eight teachers completed open ended feedback, providing predominantly positive feedback that the screening process was fair, valid, quick and easy and also enabled them to gain an insight into the child and identify potential areas of need. Eighty-one percent of teachers that completed the feedback were happy to spend time completing the questionnaires, 94% reported that they were easy to complete, 73% reported that they would be happy to complete the questionnaire for another  student, and 83% reported that the screening would be easy to implement within their school. Lastly, 94% of teachers understood the purpose of the questionnaires, and 87.5% said they would recommend other students with fixed-term exclusions to take part.
Seven parents completed open ended feedback, reporting that they liked the idea and were happy to take part, thought it would help younger children and that it was important for an independent service to assess their child. One parent did not think that some of the questionnaires were appropriate for a 14-year-old. Another parent suggested that the school should be advised on any recommendations regarding the child's needs. Eighty-five percent of parents understood the questionnaires' purpose, and 83% reported being happy to spend time completing them, with 69% reporting them easy to complete. Parents who found it difficult reported that the questionnaires were too long, too personal or they found the change in rating scales difficult. Overall respondent satisfaction ratings are presented in Figure 3.

Clinical screening outcomes
The mean scores for participant assessment measures are presented in Table 2. Due to the limited number of parent and teacher responses, statistical analysis was not conducted but descriptive results are presented in Table 3.

Psychopathology Secondary school children's outcome measures.
Self-report SDQ scores revealed that the mean total difficulties (p = .046), conduct problems (p = .005) and hyperactivity scores (p = .004) for children were significantly higher in the case than control group. Cases scored consistently higher than controls across all subscales apart from peer related problems and pro-social behaviour in which the control group scored slightly higher. Cases reported higher rates of anxiety and depression symptomology across all subscales on the RCADS-25, however, this was only significantly different from the controls for depression symptomology (p = .042).

Parent-rated outcome measures. Parent data showed large disparities between the case and control
group, with the case mean total difficulties score within the 'high' range, and the control mean within the 'close to average' range. Reflecting self-report data, cases scored consistently higher across all parent subscales, most notably for hyperactivity/ inattention, emotions and conduct problems.

Teacher outcome measures.
Teacher data did not show large differences in the total difficulties score: cases similarly scored higher for conduct problems and hyperactivity/inattention, but controls scored higher on the emotions and peer related problems scale. Across the data, the impact of psychopathology was similar for both case and control groups. Of note, parent impact scores were higher than those reported by the children and teachers. In addition, although control parents reported fewer problems than case parents, the mean impact score was higher and fell within the 'very high' range.
Neurodevelopmental difficulties ADHD measures. Cases scored significantly higher across all self-report subscales of the Conners-3 compared to controls, apart from learning problems. Scores were particularly elevated on the inattention (p = .001) and aggression (p = .001) subscales with the mean falling within the 'many more concerns' than expected range. The same pattern was observed in the parent and teacher report measures. Reflecting SDQ scores, the only subscale in which the controls scored higher was for teacher reported peer relation difficulties.

ASD measures.
Eighty percent of case parents (n = 4) reported that their children had communication difficulties of clinical significance. In line with this, the average total General Communication Composite for the cases was below the 5 th percentile when compared with normative data, while the control group scored within the average range. Furthermore, 40% of case parents (n = 2) suggested that their children had social communication profiles which were indicative of autism. Thirty-seven percent of control parents (n = 3) suggested that their child had a profile indicative of autism. However, one of these children already had a diagnosis of autism and therefore 25% (n = 2) had unidentified profiles indicative of autism. This must be interpreted with caution due to the reduced sample size.

Cognitive measures.
The cognitive ability of the cases was significantly different from the controls (p = .001) with a mean FSIQ in the lower end of the 'low average' range, whilst the controls mean fell in the 'average' range. Similar differences were obtained on Verbal Comprehension (p = .004) and Perceptual Reasoning (p = .007). Three case participants scored ≤ 70 for FSIQ, falling in the 'extremely low' range; no controls scored within this range.
Language measures. Lastly, the mean core language score for cases on the CELF-4 was significantly lower (p = .001) than the control group. The case mean fell within the 'borderline' range whereas the control mean fell within the 'average' range. Forty percent of case participants and 5% of control participants scored in the 'clinically significant' range, indicative of a moderate or severe language disorder.

Clinically significant results
The proportion of participants scoring within the 'clinically significant' range on self-report measures is displayed in Figure 4.

Discussion
The results of this study show that screening secondary school pupils at the point of a second fixedterm exclusion is acceptable and feasible for students. Screening data showed significant differences in rates of both mental health and neurodevelopmental difficulties between case and control groups suggesting high rates of unmet needs in those at high risk of fixed and/or permanent exclusion. Age, gender and ethnicity demographics were reflective of the excluded from school population within the UK (Department for Education 2020).

Feasibility and acceptability
All parents consented for their child to participate. Retention rates were high with only one young person not completing the full assessment process. From the number of eligible participants, it is suggested that an adequate number could be recruited to statistically power a larger scale study. Completion of informant measures was low, despite repeated waves of data collection in which questionnaires were sent out via mail to both parents and teachers, as well as reminder calls to parents. The low response rate would suggest that this is not a feasible way of collecting informant data and future research should explore additional methods, such as electronic versions of the questionnaires. The response rates are very different in comparison to those found in Parker et al.'s (2016) study, which had an 100% response rate from parents and 75% response rate from teachers. The different parental response rate could be partly due to the different methodology employed in the current study. Researchers were working within the school and questionnaires sent to parents, as opposed to meeting with and having direct contact. The high level of social deprivation and the limited parental engagement within the two schools, which was highlighted by school staff at the outset of the study, may have also contributed to the parental informant response rate.
With regards to teachers, it was initially thought that being positioned within the school would increase their response rate, however, previous research that collected informant data from teachers had higher response rates than the current study (Parker et al. 2016;Ripley and Yuill 2005). The main difference is that this study was conducted within a secondary school as opposed to primary schools. It could be hypothesised that primary school teachers know their pupils very well and take more responsibility for their support and wellbeing, whereas, secondary school teachers work with many students, and responsibility for them is generally more widely shared. Secondary schools also have a higher number of teachers, which makes it difficult for the researcher to liaise with them directly. The majority of teachers and parents that responded reported finding the questionnaires easy to complete and were predominantly positive about the process, suggesting that it could help to identify potential areas of need for children. Positively, 83.3% of teachers that responded thought it would be easy to implement the screening within their school.
Over 90% of the secondary school children and teachers were either satisfied or very satisfied with the process. However, the small number of parents who responded were less satisfied, with only 63% reporting being satisfied or very satisfied. Although the feedback suggests overall acceptability, to make the experience more positive, particularly for parents, it may be of benefit to explore the use of different questionnaires.

Clinical screening outcomes
The screening process revealed that secondary school children who received two or more fixed term exclusions experienced very elevated symptoms of ADHD as well as elevated depression, conduct problems and language and communication difficulties compared to a control group and normative data. Furthermore, the case group showed significantly lower cognitive ability. Although not always significant, the case group scored higher across subscales on all measures, apart from self and teacher reported peer-related problems and prosocial behaviour and teacher reported emotional difficulties. Worryingly, the vast majority of the difficulties were not reported by the student or their parent at the beginning of the assessment, indicated by a low level of existing diagnoses, suggesting a high level of unidentified need.
The most consistent finding was the high rates of ADHD symptomology reported by children, parents and teachers. This finding was expected and may be explained by the key behavioural symptoms of ADHD, such as aggression, impulsivity and hyperactivity, which closely reflect the reasons children are excluded from school (Department for Education 2020). Indeed, research has shown around 11% of children with a diagnosis of ADHD were excluded from school (O'Regan 2009) compared to exclusion rates of 0.1% in the general population (DfE 2019).
There were significant differences in cognitive functioning between the case and control group. The lower cognitive ability found among the case group would suggest that secondary school children at risk of exclusion may experience more educational difficulties, requiring greater academic support. Furthermore, 40% of case participants fell within the category indicative of a moderate or severe language disorder. Due to limited parent informant data, it is hard to draw clear conclusions about autism and social communication difficulties. Although it must be interpreted with caution due to the small number of responses, the identified rate mirrors Donno et al.'s (2010) study of case participants showing symptoms consistent with autism.
Although case participants showed elevated symptoms of depression compared to controls, the mean T-score fell in the average range. Rates of anxiety symptomology were not raised compared to controls. In addition, there were no significant difficulties reported with emotions on the SDQ by students and teachers. By contrast, teachers reported more emotional difficulties for the control group. This may be explained by attribution bias, as teachers may have been more likely to attribute emotional difficulties in the case group as displays of challenging behaviour. This finding is in contrast to Parker et al.'s (2016) study where they found children excluded from primary school to have higher rates of anxiety and depression compared to controls. However, conduct problems followed a similar pattern to Parker et al.'s (2016) study. The proportion of cases in the 'clinically significant' range for conduct problems was twice that of controls. Teachers and parents also reported elevated conduct problem scores compared to controls.
In comparison to previous data, it is interesting that there were not higher rates of anxiety and depression reported. One explanation for this may be that children who experience anxiety and depression are less likely to attend school and therefore may not have been available or self-selected to take part in the study. Previous research has demonstrated that mental health difficulties such as anxiety are associated with school absence and refusal behaviours (Richards and Hadwin 2011).
A further explanation may be due to the current study relying solely on self-report information for mental health outcomes as opposed to parent and teacher data used in previous studies (Donno et al. 2010;Parker et al. 2016). There are often large discrepancies between informant reports on measures (De Los Reyes et al. 2015).

Limitations
As the main focus of the study was to assess acceptability and feasibility, the sample size was small and the study was not powered to detect significant differences between groups. Our findings must be taken with some caution. The low response rate of informants makes it difficult to draw clear conclusions or to make comparisons between groups. One idea for future research could be to conduct an initial focus group with parents and teachers, providing more insight into what support is needed to help them to complete the questionnaires and increase involvement.
Four case participants were excluded from the study due to not meeting the criteria for fluency in English and therefore impacting the validity of their results due to the lack of normative data for this group on the standardised questionnaires and cognitive assessments used. A large proportion of students within one of the schools were classified as having limited proficiency in English and a high rate of these students had received fixed-term exclusions. Excluding these children from the study suggests that it was not an entirely representative sample of the population but, even more importantly, prevents them from having the same access to assessment and further support. Future research should explore ways in which to include secondary school children with limited English in the research, exploring the use of culturally appropriate measures.
Lastly, some outcomes proposed by the BPS (2015) position paper were not assessed, such as alcohol and drug use. Previous research has shown that children excluded from school are more likely to use substances (McCrsytal et al. 2005) and therefore screening for this at an early stage would be of benefit. Furthermore, Parker et al.'s (2016) study found a surprisingly high rate of selfharm among their case group. This was not assessed in the current study, but future studies should include measures to assess self-harm and suicidal ideation to ensure these needs are identified and supported.

Clinical and policy implications
The current findings are consistent with previous research exploring rates of neurodevelopmental and mental health difficulties among secondary school children excluded or at risk of exclusion from school (Ripley and Yuill 2005;Donno et al. 2010;Parker et al. 2016). This study extends the findings from previous case-control studies with primary school children and suggests that there are a range of complex needs that are still unrecognised and unsupported among children at risk of permanent exclusion within secondary schools. The needs experienced by these children are likely to have contributed to their repeated exclusions, with the majority of case students reporting over five fixed-term exclusions. The relationship between fixed-term exclusions and mental health difficulties could also be bi-directional as demonstrated by Ford et al. (2018), whereby mental health could contribute to exclusion from school and repeated exclusions could impact on a young person's wellbeing.
Currently, statutory guidance for exclusions suggests that a multi-agency assessment which may address underlying SEN or mental health difficulties should take place at the time of considering a permanent exclusion (DfE 2017a). This research would suggest that at the time of a permanent exclusion is too late and this should be factored in to fixed-term exclusion guidelines. The aim for future research would be to conduct a statistically powered larger scale study, taking into account the limitations discussed. Future research should also explore the use of alternative measures, particularly briefer screening measures that are being developed and used within young offender populations, such as the Child and Adolescent Intellectual Disability Screening Questionnaire (McKenzie et al. 2012). The use of briefer measures may enable more areas to be assessed within the same time-frame and children can be referred on for further assessment where necessary.
Finally, future research should consider involving children, parents, teachers and the wider school in the co-design of further projects to ensure they are as effective as possible.

Conclusion
Consistent with previous research, the current study demonstrated that secondary school children who receive school exclusions experience a higher rate of neurodevelopmental and mental health difficulties. This study extends previous findings within primary schools and suggests a range of complex needs that are still unrecognised and unsupported among children at risk of permanent exclusion within secondary schools. Although there were limited informant responses, a screening process to identify these needs was shown to be feasible and acceptable. Early identification and intervention of underlying difficulties is important to ensure these secondary school children are able to effectively engage in their education as well as to prevent further negative outcomes associated with school exclusion.