Teachers' concerns about pupils' mental health in a cross‐sectional survey of a population sample of British schoolchildren

Abstract

Background
Schools are becoming central to the identification and referral of children and young people with poor mental health. Understanding how well a teacher concern predicts mental disorder in a child or young person is important for mental health teams who need to respond to referrals.

Method
This secondary analysis of the 2004 British Child and Adolescent Mental Health Survey used the first item of the Strengths and Difficulties Questionnaire (SDQ) Impact subscale to indicate concern about a child or young person’s mental health. Mental disorder according to DSM IVR criteria was assessed using the multi‐informant Development and Well‐Being Assessment. We compared the proportion with and without mental disorder according to the presence or absence of teacher concern.

Results
Teacher concern was moderately predictive (49% with teacher concern had a disorder) and sensitive (teacher concern present among 56% with disorder), while lack of teacher concern was highly predictive (only 5% had disorder) and specific (94% no disorder). Teacher concern was associated with significantly poorer mental health (mean teacher SDQ total difficulty score 19.6, SD 5.6 with disorder, mean 15.0; SD 5.1 if no disorder) compared to children without teacher concern (mean 9.6, SD 5.5 with disorder, and 4.9; SD 4.3 if no disorder; F (3, 5,931) = 1527.228, p = .001). If both teacher and parents were concerned, the child or young person was much more likely to have a disorder.

Conclusion
A lack of teacher concern can reassure mental health practitioners in the vast majority of cases. While teacher concern does identify those with poorer mental health, it is only moderately predictive of a disorder. When concerned about a child or young person, discussions with parents or others who know them may help teachers identify those who most need support.

Key Practitioner Message
The emphasis on schools as a major setting to provide support and identify the need for referral to specialist mental health services means service commissioners, providers and practitioners could benefit from insight into how predictive a teacher’s concern is of childhood mental health conditions and how this may vary with the type of disorder
If teachers are not concerned about a child, practitioners can be reassured that there is unlikely to be a significant problem with their mental health, although this will be less certain in schools whose pupils are likely to have a higher than average levels of difficulty
Teacher concerns do not necessarily differentiate between clinically impairing and mild/ moderate mental health difficulties, but do identify children in poorer mental health
Asking for corroboration of concern from other sources increases the strength of the association to severe mental health disorders

Full text

Teachers’ concerns about pupil’s mental health in a
cross-sectional survey of a population sample of
British schoolchildren
Frances Mathews
1
, Tamsin Newlove-Delgado
1
, Katie Finning
1
,
Christopher Boyle
2
, Rachel Hayes
1
, Patrick Johnston
3
& Tamsin Ford
4
1
College of Medicine and Health, University of Exeter, Exeter,
2
Graduate School of Education, University of Exeter, Exeter,
3
Place2Be, London,
4
Department of Psychiatry, University of Cambridge, Cambridge, UK
Background: Schools are becoming central to the identification and referral of children and young people
with poor mental health. Understanding how well a teacher concern predicts mental disorder in a child or
young person is important for mental health teams who need to respond to referrals. Method: This secondary
analysis of the 2004 British Child and Adolescent Mental Health Survey used the first item of the Strengths and
Difficulties Questionnaire (SDQ) Impact subscale to indicate concern about a child or young person’s mental
health. Mental disorder according to DSM IVR criteria was assessed using the multi-informant Development
and Well-Being Assessment. We compared the proportion with and without mental disorder according to the
presence or absence of teacher concern. Results: Teacher concern was moderately predictive (49% with tea-
cher concern had a disorder) and sensitive (teacher concern present among 56% with disorder), while lack of
teacher concern was highly predictive (only 5% had disorder) and specific (94% no disorder). Teacher concern
was associated with significantly poorer mental health (mean teacher SDQ total difficulty score 19.6, SD 5.6
with disorder, mean 15.0; SD 5.1 if no disorder) compared to children without teacher concern (mean 9.6, SD
5.5 with disorder, and 4.9; SD 4.3 if no disorder; F(3, 5,931) = 1527.228, p= .001). If both teacher and parents
were concerned, the child or young person was much more likely to have a disorder. Conclusion: A lack of tea-
cher concern can reassure mental health practitioners in the vast majority of cases. While teacher concern does
identify those with poorer mental health, it is only moderately predictive of a disorder. When concerned about
a child or young person, discussions with parents or others who know them may help teachers identify those
who most need support.
Key Practitioner Message
•The emphasis on schools as a major setting to provide support and identify the need for referral to specialist
mental health services means service commissioners, providers and practitioners could benefit from insight
into how predictive a teacher’s concern is of childhood mental health conditions and how this may vary
with the type of disorder
•If teachers are not concerned about a child, practitioners can be reassured that there is unlikely to be a sig-
nificant problem with their mental health, although this will be less certain in schools whose pupils are
likely to have a higher than average levels of difficulty
•Teacher concerns do not necessarily differentiate between clinically impairing and mild/ moderate mental
health difficulties, but do identify children in poorer mental health
•Asking for corroboration of concern from other sources increases the strength of the association to severe
mental health disorders
Keywords: adolescent; child; mental disorder; mental health; schools; Teacher
Introduction
Recent worldwide research shows 13% of children and
young people of compulsory school age have a mental
health disorder (Polanczyk, Salum, Sugaya, Caye, &
Rohde, 2015). While UK findings report one in eight with
disorder, only about a quarter of these children and
young people access mental health services (Green,
McGinnity, Meltzer, Ford, & Goodman, 2005; Mandalia
et al., 2018). On average, approximately a quarter of
referrals to specialist services are rejected or redirected
in the UK, with the figure in some areas being as high as
64% (Frith, 2017). In many countries, teachers are the
most commonly consulted service in relation to child
mental health, so schools have a key role in the identifi-
cation of mental health conditions in their pupils (Ford,
©2020 Association for Child and Adolescent Mental Health
Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA 02148, USA
Child and Adolescent Mental Health Volume **, No. *, 2020, pp. **–** doi:10.1111/camh.12390

Hamilton, Meltzer, & Goodman, 2007; Mental Health
Taskforce, 2016; Newlove-Delgado, Moore, Ukoumunne,
Stein, & Ford, 2015; Sadler et al., 2018). In the UK, a
recent Green paper proposed the development of school-
based mental health teams to link the Child and Adoles-
cent Mental Health Services (CAMHS) to schools and rec-
ommended that all schools have a designated mental
health lead to identify children and young people who
are struggling (Department of Health and Social Care &
Department for Education, 2017). The ultimate aim is to
increase access to early intervention for those with mild-
to-moderate mental health needs, as well as timely refer-
ral to those with more severe problems (Department of
Health and Social Care & Department for Education,
2017).
Understanding how teacher concern reflects clinically
impairing mental health conditions in children and
young people and how this may vary with different types
of disorder is important for mental health practitioners
who receive and triage referrals from schools, and for
those training and managing the education staff who will
be taking up these new roles. We explored these ques-
tions using data from the 2004 British Child and Adoles-
cent Mental Health Survey (BCAMHS; Green et al.,
2005). Specifically, we used the first question of the
Strength and Difficulties Questionnaire (SDQ) impact
supplement (Goodman, Ford, Richards, Meltzer, & Gat-
ward, 2000; www.sdqinfo.com) to indicate concern
about mental health and analysed the presence or
absence of concern in relation to the presence of absence
of a clinically impairing mental disorder assessed by
multi-informant diagnostic assessment.
Method
The University of Exeter College of Medicine and Health ethics
committee provided approval for the secondary analysis of these
data, while the original survey gained approval from the Medical
Research Ethics Committee.
The 2004 British Child and Adolescent Mental Health survey
recruited 7,977 children and young people aged 5–16 years
from England, Scotland and Wales in a stratified probability
sample using the universal child benefit register as a sample
frame (Green et al., 2005). Mental health was assessed using
the Development and Well-being Assessment (DAWBA), which
combines highly structured questions about a range of common
childhood mental disorders with semi-structured probes about
any areas of reported difficulty (Goodman et al., 2000). The
DAWBA was completed in full by the parents of all participating
children included in the study. Young people aged 11–16 were
invited to complete the DAWBA, and if the family agreed, a tea-
cher was mailed a briefer questionnaire. A small team of expert
clinical raters who were blind to the SDQ data reviewed all
responses from all informants to assign DSM IV psychiatric
diagnoses (American Psychiatric Association, 1994). Teachers,
parents and young people aged 11 years or older also completed
the Strength and Difficulties Questionnaire (SDQ), which is a
brief, reliable, validated and widely used dimensional measure
of childhood psychopathology (Goodman, 1999). The first ques-
tion of the Impact supplement was used to assess the level of
teacher or parental concern about the child (Goodman, 1999).
This asks respondents if they consider the child to have difficul-
ties in the following areas: emotions, concentration, behaviour
or being able to get on with others, to which respondents could
rate difficulties as ‘no’,‘minor’,‘definite’or ‘severe’(Goodman,
1999). The latter two responses were categorised as indicating
concern, and the former were assumed to indicate no significant
concern. Parents reported demographic characteristics, their
child’s health and access to services over the previous
12 months. Parental mental health was assessed using the
General Health Questionnaire and family function using the
McMaster’s family Assessment Device (Goldberg et al., 1997;
Miller, Epstein, Bishop, & Keitner, 1985).
SPSS version 25.0 was used to conduct our analysis. Those
children and young people without a teacher report were
removed, leaving 5,965 cases to analyse. We excluded the self-
report SDQ from this analysis because it was only available on
those over the age of 11, so would provide imprecise estimates,
particularly for the type of disorders. In addition, previous work
suggests that parent and teacher reports are more predictive of
disorder (Goodman, Ford, Simmons, Gatward, & Meltzer,
2003). Those with and without teacher data were compared for
all available background characteristics using chi-squared and
t-tests (significance level p=.05) to consider the impact of miss-
ing teacher data on our analysis. Based on the similarity
between both studies, values of sensitivity and specificity (as
defined in Table 1) from this study and prevalence data from
the recent UK national survey will calculate positive predictive
value (PPV) and negative predictive value (NPV; as defined in
Table 1) parameters to understand how generalisable our find-
ings for those with teacher reports are to populations with a dif-
ferent prevalence (Mandalia et al., 2018).
Data on the children and young people were analysed sepa-
rately by primary (5–10 years) and secondary (11–16), and by
all ages. We estimated the prevalence, PPV, NPV, sensitivity and
specificity to identify differences in teacher concern by school
age and disorder type.
A one-way ANOVA was used to compare teacher and parent
SDQ total difficulty mean scores for four possible categories of
diagnosis of a mental disorder (no disorder and no concern, dis-
order but no concern, no disorder but concern, and disorder
and concern). We analysed teacher and parent concern in rela-
tion to disorder separately.
Finally, we calculated the PPV and NPV, sensitivity and speci-
ficity when both teacher and parent had a concern, only a tea-
cher, or only a parent had a concern, in relation to the presence
or absence of a mental disorder (Prince, 2003).
Table 1. Definitions of positive predictive value (PPV), negative
predictive value (NPV), sensitivity and specificity
Epidemiological
tests Definition Calculation
Positive predictive
value
The probability that a
person with a positive
test result is a true
positive
A/
(A +B) 9100
Negative
predictive value
The probability that a
person with a negative
test result is a true
negative
D/
(D +C) 9100
Sensitivity The probability that a
diseased person (case) in
the population tested
will be identified as
diseased by the test
A/
(A +C) 9100
Specificity Is the probability that a
person without the
disease (noncase) will be
correctly identified as
nondiseased by the test
D/
(D +B) 9100
Disease No Disease
Positive (number) A(true
positive)
B(false
positive)
Test positive
Negative
(number)
C(false
negative)
D(true
negative)
Test negative
True disease True no
disease
Total
Samet et al. (2009).
©2020 Association for Child and Adolescent Mental Health
2Frances Mathews et al. Child Adolesc Ment Health 2020; *(*): **–**

Each analysis considered the presence of any mental disor-
der, more than one disorder (comorbidity), emotional disorder
(anxiety or depression), attention-deficit hyperactivity disorder
(ADHD) or conduct disorder. Less common disorders, such as
tic and eating disorders, were only analysed when included in
the ‘any mental disorder’category owing to small numbers.
Results
The sample of children that we could not include in this
analysis due to missing teacher data had poorer mental
health according to the parental SDQ, but importantly
was not more likely to have a mental disorder (see Tables
S1 and S2). They were more likely to be living in chal-
lenging circumstances (poorer parental mental health,
large, single or reconstituted families, economic depriva-
tion, more stressful life events and of black of ethnic
minority status). They were in poorer physical health
and were more likely to have a learning disability or con-
tact with mental health services or educational special-
ists. The NPV of those with missing teacher information
would not be generalisable to those with a teacher report
were it possible to calculate it, as NPV is dependent on
the prevalence of the condition under study. Current
prevalence rates of mental health disorders in children
and young people increase teacher concern PPV to 56.7
and decrease NPV to 93.6 parameters (see Table S3).
The NPV and specificity of teacher concern were high
for the whole sample as well as both primary and sec-
ondary school-aged children, which suggests that men-
tal health practitioners can be reassured by a lack of
teacher concern (see Table 2). In the sample as a whole,
teachers’concerns were more sensitive to conduct disor-
der and ADHD, than emotional disorder, which may
have fewer symptoms that are visible in the school set-
ting. They were also more sensitive to children with more
than one mental health condition (comorbidity), which
probably relates to the increased impairment experi-
enced by these children. A similar but less striking trend
was evident for the PPV in primary age children, while
teachers’concerns about ADHD among secondary aged
pupils were less predictive than their concerns about
emotional disorders.
As Figure 1 illustrates, distress according to the SDQ
was highest for those with both a disorder and adult con-
cern and lowest for those with no disorder and no con-
cern regardless of whether the teacher or parent was
reporting. Both the intermediate groups had intermedi-
ate levels of distress, but parental SDQ scores were more
likely to reflect disorder rather than the level of parental
concern, while teacher SDQ scores were lower for those
with a disorder who the teacher thought were coping
compared to significantly higher scores for those without
disorder who the teacher was concerned about. This
may be a true reflection of their function in the school
setting, as not all disorders may cause problems coping
with school. Overall, a statistically significant difference
was found between the different groups related to the
presence or absence of teacher concern and disorder for
both the teacher SDQ total difficulty score mean (F(3,
5,931) =1527.228, p=.001) and parental SDQ total dif-
ficulty score mean value (F(3, 5,931) =719,894,
p=.001). Importantly, children for whom there was tea-
cher concern but no mental disorder had considerably
worse mental health than their peers for whom teachers
reported no worries (mean teacher SDQ total difficulty
score 15.0, standard deviation 5.1 vs. mean 4.9, SD 4.3).
When reports of concern from both teachers and par-
ents were combined, they produced higher levels of sen-
sitivity across all disorders compared to teacher- or
parent-only reports, with the exception of parent-only
concern for emotional disorder. Specificity for any psy-
chiatric diagnosis was almost twice as high when both
teacher and parent were concerned (see Table 3).
Discussion
To our knowledge, this is the first paper to examine how
accurately teacher concern predicts mental disorder in
school-aged children and young people. Our findings
replicate a similar analysis of parental concern that used
data from the earlier 1999 British Child and Adolescent
Mental Health Survey (Ford, Sayal, Meltzer, & Goodman,
2005). The latter study found that the accuracy of paren-
tal concern prediction was increased by checking
whether the teacher was also worried, which mirrored
the findings from the current analysis, higher sensitivity
of teacher concern when corroborated by parental con-
cern. This previous survey also demonstrated that the
children and young people for whom parents were con-
cerned but did not meet diagnostic criteria for a mental
disorder had significantly poorer mental health, as
demonstrated by elevated SDQ total difficulty scores,
compared to the children and young people whom par-
ents were not worried about them (Ford et al., 2005).
Thus, if the parents or carers decline requests for con-
tact with teachers, practitioners can be reassured if par-
ents lack concern that the vast majority of children will
be in good mental health, but will need more information
to assess whether concerns relate to clinically impairing
mental health conditions or subclinical difficulties (Ford
et al., 2005). The effect of a lack of a teacher report war-
rants further investigation as the practical benefits of
multi-informant assessments may improve access to
services (Collishaw, Goodman, Ford, Rabe-Hesketh, &
Pickles, 2009; Mandalia et al., 2018; McNeilis,
Maughan, Goodman, & Rowe, 2017).
Our findings support previous literature that teacher
and parents do not always have the same perspectives in
their assessment of mental health, reflected in the tea-
cher SDQ total difficulty scores which relate to the level
of teacher concern more closely than parental concern
(Collishaw et al., 2009; De Los Reyes et al., 2015). Poor
levels of agreement between informants about the same
child are common and likely to be related to true differ-
ences in how the child functions in relation to different
environments, the informants’different frames of refer-
ence and measurement error (Collishaw et al., 2009; De
Los Reyes et al., 2015). Parents are likely to be predomi-
nantly influenced by family distress and child physical
health, while teachers’concerns will reflect their broader
experience of the normal range of child behaviour in the
school context, including peer relationships, child
attainment and economic disadvantages of school and
surrounding neighbourhood (Collishaw et al., 2009).
Attempts to develop a disorder-specific hierarchy of
informant perspectives, usually in a research context,
have proved unhelpful, but careful clinical review of all
the information from all informants improves diagnostic
assessment (McNeilis et al., 2017).
©2020 Association for Child and Adolescent Mental Health
doi:10.1111/camh.12390 Accuracy of teachers concerns about pupil’s mental health 3

Universal screening programmes have not been exten-
sively tested, but may arguably be the most accurate
method of identifying children and young people with
mental health disorders (Anderson et al., 2018; Good-
man et al., 2003). Our findings reinforce the concern
that universal screening programmes would yield a high
number of false-positive results, particularly if solely
reliant on teacher accounts. If the objective is to identify
disorder, teacher concerns do not necessarily differenti-
ate between severe, moderate and milder mental health
difficulties (Anderson et al., 2018). Better differentiation
about the severity of difficulties and the threshold for
referral to specialist services is important to avoid
swamping specialist services, as is the adequate provi-
sion of support for those with mild or moderate mental
health problems to prevent deterioration. The latter is
particularly important in the school context, where poor
mental health can impede children’s ability to function
in school and because schools are, by default, the front-
line service provider for child mental health (Ford et al.,
2007; Newlove-Delgado et al., 2015).
While informative, our study raised additional
research questions. The current analysis was cross sec-
tional, and it would be useful to understand more about
the future mental health trajectory with children with
mild, moderate and severe mental health difficulties in
order to understand better how to target resources. In
addition, it would be helpful to understand how predic-
tive teacher and parental concerns are of persistent diffi-
culties. The impact of training programmes on mental
health in initial teacher training and for the new desig-
nated mental health leads in schools could potentially
increase the accuracy of the recognition of more severe
problems but would need testing empirically
Table 2. Predictability of teacher concern in recognising the type of disorder for a child or young person by school age
Prevalence % Positive predictive power Negative predictive power Sensitivity Specificity
All ages
a
n=5,965
Any psychiatric diagnosis 9.6 48.6 95.3 56.3 93.7
Comorbidity 2.5 16.9 99.3 74.2 90.5
Emotional disorder 3.5 13.1 97.7 41.4 90.0
Conduct disorder 5.4 34.1 98.2 70.4 92.3
ADHD 2.2 14.8 99.4 74.8 90.3
ASD 0.8 4.7 99.7 66.0 89.3
Primary
b
n=3,134
Any psychiatric diagnosis 7.9 43.2 96.4 59.3 93.3
Comorbidity 2.1 14.1 99.4 73.8 90.5
Emotional disorder 2.3 7.9 98.4 38.0 89.8
Conduct disorder 4.7 30.0 98.4 69.9 92.0
ADHD 2.3 16.2 99.4 76.4 90.7
ASD 0.9 5.6 99.6 65.5 89.7
Secondary
c
n=2,831
Any psychiatric diagnosis 11.4 54.3 94.1 54.0 94.1
Comorbidity 3.0 19.9 99.1 74.4 90.6
Emotional disorder 4.9 18.6 96.9 43.1 90.3
Conduct disorder 6.2 38.5 98.0 70.9 92.5
ADHD 2.1 13.4 99.4 72.9 89.9
ASD 0.6 3.7 99.7 66.7 89.0
ADHD, attention-deficit hyperactivity disorder; ASD, autism spectrum disorder.
a
All children and young people in the survey between 5 and 16 years.
b
Children of primary school age between 5 and 10 years.
c
Young people of secondary school age between 11 and 16 years.
Figure 1. Teacher and parent mean total SDQ scores of children and young people in the presence or absence of a mental disorder and
are of concern
©2020 Association for Child and Adolescent Mental Health
4Frances Mathews et al. Child Adolesc Ment Health 2020; *(*): **–**

(Department of Health and Social Care & Department for
Education, 2017; Mental Health Taskforce, 2016). Moni-
toring the effect that teams in schools have in supporting
low-level mental health problems is important and may
provide data that support more accurate identification of
need (Department of Health and Social Care & Depart-
ment for Education, 2017; Mental Health Taskforce,
2016; Newlove-Delgado et al., 2015). Analysis so far has
not explored how predictive young people’s concerns are
in detail, though Goodman’s initial work suggests that
young people were less predictive than either parent or
teacher (Goodman et al., 2003).
Our analysis benefits from a large representative
population-based sample of children and the use of
validated measures to assess mental health (Good-
man, 1999; Goodman et al., 2000; Green et al.,
2005). Overall, characteristics of those with a mental
disorder identified here, irrespective of having a tea-
cher report, are the same as those described in
recent literature (Department of Health and Social
Care & Department for Education, 2017; Green
et al., 2005; Mandalia et al., 2018; Mental Health
Taskforce, 2016) However, not all families consented
to contact with a teacher, and not all teachers
responded to the survey, which reduced the sample
for analysis to almost 75% of the entire baseline
sample (Green et al., 2005). Our analysis shows that
children and young people without teacher data were
in poorer mental health and facing higher levels of
adversity than the children we were able to include.
This might therefore have reduced the estimated sen-
sitivity and PPV of teacher concern. However, emerg-
ing evidence suggests that sociodemographic,
economic and family factors, such as living in a one-
parent family, may increase the level of disagreement
between parents and teachers, a reporting bias that
would operate in the opposite direction (Cheng et al.,
2018). In addition, previous research shows that
while those with poorer mental health tend to be less
likely to participate and more likely to dropout of
studies, this does not necessarily change the nature
of associations detected in the obtained sample
(Wolke et al., 2009).
Secondary analysis is inevitably restricted to the avail-
able data, which means that we lacked data in the under
5’s and the over 16’s, so our results may not generalise
to nurseries and further education (Sadler et al., 2018).
It is worth also noting that NPV and PPV are both depen-
dent on the prevalence of the condition under study, so
that in schools with particularly high levels of need (e.g.
alternative provision or schools in highly deprived
areas), a lack of teacher concern may be less reassuring
and the presence of teacher concern more predictive
(Samet, Wipfli, Platz, & Bhavsar, 2009). Nonetheless, in
our sensitivity analysis, PPV and NPV figures altered
only marginally (see Table S3; Green et al., 2005; Man-
dalia et al., 2018). Further adjustments in PPV and NPV
of teacher concern may also be impacted by the quality
or availability of mental health training received as
teachers report the need for additional training and sup-
port (Evans et al., 2019).
We have no information about the detail of teacher’s
concerns, which we are inferring from their response to
the SDQ (Green et al., 2005). The number of children
with specific diagnoses, such as separation anxiety, was
too few to permit meaningful analysis, so we analysed
across broad groups of disorder (emotional disorder).
This may have missed some differences that may relate
both the pattern of prevalence (e.g. separation anxiety is
common in very young children, panic attacks are rare
before adolescence) and their salience to the school con-
text (social anxiety can be reflected in an unwillingness
to speak in class, while phobias that do not involve
school-based stimuli may never cause problems in
school; Collishaw et al., 2009; De Los Reyes et al., 2015).
The latter, in combination with the issue that some of
the less common disorders, such as tic and eating disor-
ders, if not comorbid with other disorders that impact
function at school, might contribute to the surprisingly
weak relationships between teacher and parent concern
and mental disorder.
Conclusion
Teacher concern is only moderately predictive of clini-
cally diagnosable mental health disorders but does iden-
tify children and young people who are more distressed
than those without disorder or concern. Lack of teacher
concern should reassure mental health practitioners in
the vast majority of cases, while asking parents or other
informants who know the child for their opinion may
help to identify those who most need support.
Table 3. Testing how well the teacher or parent or both recognise a child or young person with disorder/s by type (for teacher only values
see Table 2)
Positive predictive power Negative predictive power Sensitivity Specificity
Both (parent and teacher) Any psychiatric diagnosis 74.4 97.7 63.1 98.6
Comorbidity 48.7 93.0 92.2 51.5
Emotional disorder 32.8 43.9 50.0 27.6
Conduct disorder 67.2 72.8 80.9 56.5
ADHD 37.9 93.5 91.7 44.2
ASD 14.4 98.2 93.3 40.1
Parent only Any psychiatric diagnosis 33.3 68.6 51.9 50.3
Comorbidity 23.0 86.4 64.6 50.9
Emotional disorder 43.7 82.4 72.8 57.5
Conduct disorder 45.2 22.9 36.1 30.3
ADHD 16.3 81.6 48.9 47.4
ASD 10.4 97.6 82.4 50.2
ADHD, attention-deficit hyperactivity disorder; ASD, autism spectrum disorder.
©2020 Association for Child and Adolescent Mental Health
doi:10.1111/camh.12390 Accuracy of teachers concerns about pupil’s mental health 5

Acknowledgements
F.M.’s time was funded by Place2Be via Exeter University Col-
lege of Medicine. T.F. has a voluntary position as the Research
Chair for Place2Be. The initial survey was funded by the govern-
ments of England, Wales and Scotland. The funders had no
influence on the design, analysis or reporting of this study. The
authors would like to thank the children, their parents and their
teachers who participated in this survey, as well as their col-
leagues on the steering groups, as well as the England’s Depart-
ment of Health, the Welsh Assembly and Scottish Office who
commissioned the survey, and the Office for National Statistics
who led the data collection. The input of Robert Goodman in this
work was essential. He led the team of clinical raters as well as
designing both the mental health measures. The authors are
also grateful to Darren Moore, lecturer in Education at the
University of Exeter for his helpful comments on an early draft.
Finally, the authors would like to thank Sarah Golden at Pla-
ce2Be; the authors’conversation with her about children who
appear to have no difficulties at referral sparked this analysis.
T.F. conceptualised the aims of the study and supervised the
analysis and writing. F.M. led the analysis and writing. All
others contributed to the interpretation of the results and the
writing of the manuscript. T.F. has a voluntary position as the
Research Chair for Place2Be, which involves sitting on the
Research Advisory Group and contributing to research projects.
She receives no personal payment other than reimbursement of
travel expenses. Place2Be provide funding to her research team
to support Place2Be-related research projects. The remaining
authors have declared that they have no competing or potential
conflicts of interest.
Ethical information
UEMS ethics committee provided ethical approval for
the secondary analysis of 2004 British Child and Adoles-
cent Mental Health Survey. Medical Research Ethics
Committee provided ethical approval for the original sur-
vey in 2004. All participants provided informed consent.
Correspondence
Tamsin Ford, Department of Psychiatry, University of
Cambridge, Douglas House, 18b Trumpington Road,
Cambridge CB2 8AH, UK; Email: tjf52@medschl.
cam.ac.uk
Supporting information
Additional Supporting Information may be found in the online
version of this article:
Table S1. Teacher concern missing and non-missing cases
by characteristics.
Table S2. Teacher concern missing and non-missing mean
values by characteristics.
Table S3. Source of PPV and NPV calculations using current
prevalence rates for children and young people with a mental
health disorder of 12.8% and Sensitivity and Specificity data
from 2003 survey data with teacher concern.
References
American Psychiatric Association (1994). Diagnostic and statis-
tical manual of mental disorders: DSM-IVpK748KW3Eq.
Washington, DC: AUTHOR.
Anderson, J. K., Ford, T., Soneson, E., Thompson Coon, J.,
Humphrey, A., Rogers, M., ...& Howarth, E. (2018). A system-
atic review of effectiveness and cost-effectiveness of school-
based identification of children and young people at risk of, or
currently experiencing mental health difficulties. Psychologi-
cal Medicine,49,9–19.
Cheng, S., Keyes, K. M., Bitfoi, A., Carta, M. G., Koc
ß, C., Goelitz,
D., ... & Kovess-Masfety, V. (2018). Understanding parent–
teacher agreement of the Strengths and Difficulties Question-
naire (SDQ): Comparison across seven European countries.
International Journal of Methods in Psychiatric Research,27,
e1589.
Collishaw, S., Goodman, R., Ford, T., Rabe-Hesketh, S., & Pick-
les, A. (2009). How far are associations between child, family
and community factors and child psychopathology infor-
mant-specific and informant-general?Journal of Child Psy-
chology and Psychiatry,50, 571–580.
De Los Reyes, A., Augenstein, T. M., Wang, M., Thomas, S. A.,
Drabick, D. A. G., Burgers, D. E., & Rabinowitz, J. (2015).
The validity of the multi-informant approach to assessing
child and adolescent mental health. Psychological Bulletin,
141, 858–900.
Department of Health and Social Care & Department for Educa-
tion (2017). Transforming children and young people’s men-
tal health provision: A green paper. Retrieved from: https://
www.gov.uk/government/consultations/transforming-child
ren-and-young-peoples-mental-health-provision-a-green-
paper.
Evans, R., Parker, R., Russell, A. E., Mathews, F., Ford, T.,
Hewitt, G., ... & Janssens, A. (2019). Adolescent self-harm
prevention and intervention in secondary schools: A survey of
staff in England and Wales. Child and Adolescent Mental
Health,24, 230–238.
Ford, T., Hamilton, H., Meltzer, H., & Goodman, R. (2007). Child
mental health is everybody’s business: The prevalence of con-
tact with public sector services by type of disorder among Bri-
tish School Children in a Three-Year Period. Child and
Adolescent Mental Health,12,13–20.
Ford, T., Sayal, K., Meltzer, H., & Goodman, R. (2005). Parental
concerns about their child’s emotions and behaviour and
referral to specialist services: general population survey. Bri-
tish Medical Journal,331, 1435–1436.
Frith, E. (2017). Access and waiting times in children and young
people’s mental health services. Education policy institute,
Retrieved from https://epi.org.uk/publications-and-researc
h/access-waiting-times-children-young-peoples-mental-hea
lth-services/.
Goldberg, D. P., Gater, R., Sartorius, N., Ustun, T. B., Piccinelli,
M., Gureje, O., & Rutter, C. (1997). The validity of two ver-
sions of the GHQ in the WHO study of mental illness in gen-
eral health care. Psychological Medicine,27, 191–197.
Goodman, R. (1999). The extended version of the Strengths and
Difficulties Questionnaire as a guide to child psychiatric case-
ness and consequent burden. Journal of Child Psychology
and Psychiatry,40, 791–799.
Goodman, R., Ford, T., Richards, H., Meltzer, H., & Gatward, R.
(2000). The development and well-being assessment: descrip-
tion and initial validation of an integrated assessment of child
and adolescent psychopathology. Journal of Child Psychologi-
cal Psychiatry,41, 645–657.
Goodman, R., Ford, T., Simmons, H., Gatward, R., & Melt-
zer, H. (2003). Using the Strengths and Difficulties Ques-
tionnaire (SDQ) to screen for child psychiatric disorders in
a community sample. International Review of Psychiatry,
15, 166–172.
Green, H., McGinnity, A., Meltzer, H., Ford, T., & Goodman, R.
(2005). Mental health of children and young people in Great
Britain, 2004. London: Office for National Statistics.
Mandalia, D., Ford, T., Hill, S., Sadler, K., Vizard, T., Goodman,
A., ... & McManus, S. (2018). Mental Health of Children and
Young People in England, 2017. London: Services, Informal
Support and Education; Health and Social Care Information
Centre.
McNeilis, J., Maughan, B., Goodman, R., & Rowe, R. (2017).
Comparing the characteristics and outcomes of Parent- and
Teacher-Reported Oppositional Defiant Disorder: Findings
from a national sample. Journal of Child Psychology and Psy-
chiatry,59, 659–666.
©2020 Association for Child and Adolescent Mental Health
6Frances Mathews et al. Child Adolesc Ment Health 2020; *(*): **–**

Mental Health Taskforce (2016). The five year forward view for
mental health. NHS. England. Retrieved from: https://
www.england.nhs.uk/mental-health/taskforce/.
Miller, I. W., Epstein, N. B., Bishop, D. S., & Keitner, G. I.
(1985). The McMaster family assessment device: Reliability
and validity. Journal of Marital and Family Therapy,11, 345–
356.
Newlove-Delgado, T., Moore, D., Ukoumunne, O. C., Stein, K., &
Ford, T. (2015). Mental health related contact with education
professionals in the British Child and Adolescent Mental
Health Survey 2004. The Journal of Mental Health Training,
Education and Practice,10, 159–169.
Polanczyk, G. V., Salum, G. A., Sugaya, L. S., Caye, A., & Rohde,
L. A. (2015). Annual Research Review: A meta-analysis of the
worldwide prevalence of mental disorders in children and
adolescents. Journal of Child Psychology and Psychiatry,56,
345–365.
Prince, M. (2003). Measurement in psychiatry. In M. Prince, R.
Stewart, T. Ford, & M. Hotopf (Eds.), Practical psychiatric epi-
demiology (pp. 13–41). Oxford: Oxford University Press.
Sadler, K., Vizard, T., Ford, T., Goodman, A., Goodman, R., &
McManus, S. (2018). The Mental Health of Children and
Young People in England 2017: Trends and characteristics.
London: Health and Social Care Information Centre.
Samet, J. M., Wipfli, H., Platz, E. A., & Bhavsar, N. (2009). A dic-
tionary of epidemiology, Fifth Edition: Edited by Miquel Porta.
American Journal of Epidemiology,170, 1449–1451.
Wolke, D., Waylen, A., Samara, M., Steer, C., Goodman, R.,
Ford, T., & Lamberts, K. (2009). Selective drop-out in longitu-
dinal studies and non-biased prediction of behaviour disor-
ders. The British Journal of Psychiatry,195, 249–256.
Accepted for publication: 17 March 2020
©2020 Association for Child and Adolescent Mental Health
doi:10.1111/camh.12390 Accuracy of teachers concerns about pupil’s mental health 7