economic impact of childhood psychiatric disorder on public sector services in britain: estimates...

Economic impact of childhood psychiatric disorder onpublic sector services in Britain: estimates from

national survey data

Tom Snell,1 Martin Knapp,2,3 Andrew Healey,4 Sacha Guglani,5 Sara Evans-Lacko,6

Jose-Luis Fernandez,7 Howard Meltzer,8 and Tamsin Ford9

1Personal Social Services Research Unit, London School of Economics and Political Science, London, UK; 2PersonalSocial Services Research Unit, London School of Economics and Political Science, London, UK; 3Institute ofPsychiatry, King’s College London, London, UK; 4Personal Social Services Research Unit, London School of

Economics and Political Science, London, UK; 5Croydon Adolescent Mental Health Team, Croydon, UK; 6Institute ofPsychiatry, King’s College London, London, UK; 7Personal Social Services Research Unit, London School of

Economics and Political Science, London, UK; 8Department of Health Sciences, University of Leicester, Leicester, UK;9Institute of Health Services Research, Peninsula Medical School, University of Exeter, Exeter, UK

Background: Approximately one in ten children aged 5–15 in Britain has a conduct, hyperactivity or emotionaldisorder.Methods: The British Child and Adolescent Mental Health Surveys (BCAMHS) identified children aged 5–15with a psychiatric disorder, and their use of health, education and social care services. Service costs were estimatedfor each child and weighted to estimate the overall economic impact at national level. Results: Additional health,social care and education costs associated with child psychiatric disorders totalled £1.47bn in 2008. The lion’s shareof the costs falls to frontline education and special education services. Conclusions: There are huge costs to thepublic sector associated with child psychiatric disorder, particularly the education system. There is a pressing needto explore ways to reduce these costs while improving health and well-being. Keywords: Cost, economic impact,child, adolescent, mental health, psychiatric, hyperactivity, conduct, emotional, disorder.

IntroductionWith many countries now facing economic chal-lenges on a scale not seen for generations, govern-ments face difficult public expenditure decisions.The UK government has embarked on a programmeof cuts that will inevitably reduce access to and levelof support from health, education, social care andother services for some people. It is therefore imper-ative to understand the economic consequences ofaddressing the needs of different population groupsto provide a platform for discussing short-termpriorities and longer term aspirations.

We looked at one particular and complex area ofpublic responsibility: the care, support and treat-ment of young people with psychiatric disorders.Using data from the British Child and AdolescentMental Health Surveys (BCAMHS) (Meltzer, Gat-wood, Goodman & Ford, 2000), we estimated costs– at individual level and nationally – of health, socialcare and education service use by children andyoung people associated with psychiatric disorder.

MethodsData sources

The BCAMHS used centralised records held by theChild Benefit Centre as a sampling frame to obtain a

nationally representative sample of 5–15 year olds inBritain (Meltzer et al., 2000). It is the only suchnational survey conducted in Britain, and one of onlya few internationally. All children were drawn frompostal code sectors, covering 90% of all Britishchildren.

Figure 1 summarises the survey process. Themain ‘baseline’ survey (time 1) in 1999 was followedup with surveys conducted 20 months later (time 2)and 3 years later (time 3).

At time 2, parents of all 929 children identifiedas having a psychiatric disorder at baseline and a 1-in-3 sample of those without any disorder (3,063children) were sent a postal questionnaire. Parentswho indicated contact with frontline professionals orspecialised services, those whose children had sig-nificant psychological difficulties but reported noservice use and a stratified sample of those reportingmeetings with teachers but no other services wereinvited to participate in semi-structured telephoneinterviews covering frequency of service use since thebaseline survey.

The time 3 follow-up aimed to recruit all childrensampled at time 2 regardless of whether postalquestionnaires were returned. All parents whoreported any service contact were approached fortelephone interview at time 3, together with 63parents of children with a psychiatric disorder butno reported service contacts.

Costs described here build on analysis of parent-reported service use over the time 2 and 3 periodsConflict of interest statement: No conflicts declared.

© 2013 The Authors. Journal of Child Psychology and Psychiatry © 2013 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

Journal of Child Psychology and Psychiatry 54:9 (2013), pp 977–985 doi:10.1111/jcpp.12055

(amounting to 3 years) for 2,461 children and youngpeople covered by the initial baseline survey suc-cessfully followed up across all time points (Ford,Hamilton & Goodman, 2005).

Assessment of psychiatric disorder

The Development and Well-being Assessment (DAW-BA) assessed presence of psychiatric disorder atbaseline and time 3 (Goodman, Ford, Richards,Gatward & Meltzer, 2000). The DAWBA involveshighly structured interviews with parents and chil-dren aged 11 or over, with a shortened questionnaireto teachers where families consented; questionsrelated directly to information required to meetICD-10 (World Health Organization, 1990) andDSM-IV diagnostic criteria (American PsychiatricAssociation, 1994). All informants were encouragedto provide additional qualitative information aboutareas of reported difficulty. A small team of clinicalraters used data provided from all available infor-mants to assign diagnoses according to ICD-10criteria, combining information as they would inthe clinic. In the DAWBA validation study, there werehigh levels of agreement between DAWBA and casenotes among the clinical sample (Kendall’s tau

b = 0.47–0.70) (Goodman et al., 2000). The kappastatistic in the baseline survey for chance-correctedagreement on 500 children between two raters was0.86 for any disorder, 0.57 for emotional disordersand 0.98 for behavioural disorders.

Service use

Parents were asked about service contacts sincebaseline at time 2 and over a 1-year retrospectiveperiod at time 3. Telephone interviews also asked fordetails regarding contact frequency and duration.The semi-structured telephone interview developedfor these follow-up studies is at least moderatelyreliable with good validity when assessed againstrecords from a tertiary referral clinic (Ford, Hamil-ton, Dosani, Burke & Goodman, 2007). It was basedon questions about service contacts drawn fromexisting research tools (Ascher, Farmer, Burns &Angold, 2007; Beecham & Knapp, 2001; Stiffmanet al., 2000).

Service use data were constructed by combininginformation in the time 2 postal questionnaire andtime 3 survey interview (Ford et al., 2005) withinformation on volume of resource use reported inthe telephone interviews, where available. As data

Time 3

Time 2

Sample 3992 (929 with disorder + 3063 without) Invited2954 Surveys completed2393 Willing to be contacted by telephoneSurvey modalityPostal questionnaireMeasuresParental SDQ + service use

Sample499 Parents invited439 Interviews completedSurvey modalityTelephone interviewsMeasuresContacts with any source of help since Time 1

Sample3245 (887 with disorder + 2358 without) Invited2899 Surveys completed2667 Willing to be contacted by telephoneSurvey modalityIn person interview (children and parents) and postal questionnaire (teachers)Measures Development and Well-Being AssessmentSDQSocio-demographic variables

Sample474 Parents invited403 Interviews completedSurvey modality: Telephone interviewsMeasures Contacts with any source of help over the preceding year

Time 1

Sample10,438 RespondentsSurvey modalityIn-person interview (children and parents) and postal questionnaire (teachers)MeasuresDevelopment and Well-Being AssessmentStrengths and Difficulties Questionnaire (SDQ)British Picture Vocabulary Test and British Ability Reading ScalesSocio-demographic variables

Initial interview/questionnaire (Jan-May 1999)

Postal questionnaire (Oct 2000) Phone interview (Nov 2000-Mar 2001)

Interview / questionnaire (Jan-May 2002) Phone interview (Mar-June 2002)

Figure 1 Study design and sampling strategy

© 2013 The Authors. Journal of Child Psychology and Psychiatry © 2013 Association for Child and Adolescent Mental Health.

978 Tom Snell et al. J Child Psychol Psychiatry 2013; 54(9): 977–85

were not available for volume or frequency of contactwith social workers and other social services profes-sionals, it was assumed that one assessment wascarried out per follow-up period. Respite care receiptwas estimated at 2 weeks were reported. All costestimations were carried out in relation to totalservice contacts for the entire 3-yearperiod, basedon data collected at each follow-up point.

Parents were asked to indicate service contactsmade specifically in response to concerns about achild’s ‘emotions, behaviour and concentration’.However, during the interviews it became clear thatparents were also indicating professional contactsthat were not strictly related to these kinds ofdifficulties. Therefore, interview responses weregraded on a 0–4 scale based on how closely relatedto emotional and behavioural problems each contactwith services had been (0 = no relevance, 4 = com-pletely related): only contacts graded 3 or 4 wereincluded as an assumed response to emotional-behavioural difficulties. Service use and cost esti-mations therefore only reflect costs directly related topsychiatric disorder. For example, additional educa-tional support related to dyslexia was not counted asrelated to psychiatric disorder (coded 0), while seeinga school doctor for assessment of special educationalneeds in relation to autism would be classified asmostly related (code 3).

Unit costs

Unit costs at 2007/2008 values were attached toservice use frequencies to reflect long-run socialmarginal opportunity costs (Beecham & Knapp,2001).

Costs were grouped into six broad categories:

1. Primary care costs – contact with GPs and healthvisitors.

2. Paediatrics and child health service costs – con-tact with paediatricians, paediatric inpatientstays, community nurses, school nurses, dieti-cians, physiotherapists, occupational therapists,speech therapists and visits to accident andemergency departments.

3. Mental health service costs – contact with childpsychiatrists, child psychiatric inpatient staysand child psychiatric day hospital visits, coun-selling services provided in school and elsewhere,psychologists, family therapists, and communitypsychiatric nursing staff.

4. Frontline education resources – parental meetingswith teachers, extra help provided in the schoolby teaching staff and learning support assistants,contact with special educational needs officersand involvement with special educational needstribunals.

5. Special education resource costs – attendance atspecial schools and contact with educationalsocial workers and educational psychologists.

6. Social care services – social services assessments,contact with a social worker and use of respitecare.

The main source of health and social service unitcost data was the Personal Social Services ResearchUnit annual handbook (Curtis, 2008), or inflatedvalues from previous years (Curtis, 2007; Netten &Curtis, 2003) if unavailable. The cost of socialservices assessment was updated from a review ofthe Assessment Framework (Department of Health,2003).

Unit costs associated with special schools provi-sion for children and young people were inflated to2008 levels from Education Cost Statistics publishedonline by the Chartered Institute for Public Financeand Accountancy. These estimates were adjustedupward for children with funded residential place-ments and downward for day placements based onproportional differences in local authority residentialand day care costs for elderly people (Netten &Curtis, 2003).

Teacher costs were derived from salary scale mid-points published by the National Union of Teacherswith an add-on for related costs and institutionaloverheads. Costs of special educational needs offi-cers were derived from senior teachers’ salary scale.Contacts with teaching support staff (e.g. learningassistants) were costed using the mid-point salaryon the unqualified teacher pay scale. Estimates ofannual costs of conducting Special EducationalNeeds Tribunals (covering salaries, administration,accommodation and staff training) were derived fromthe Report of the Review of Tribunals (Lord Chan-cellor’s Department’s, 2001).

Unit costs for a small number of health andeducation professionals for whom we could not findpublished estimates were approximated using esti-mates for similar services. London adjustments weremade (Netten & Curtis, 2003).

Cost estimation

The estimation of mean costs for each service cate-gory introduces three statistical issues: estimation ofstatistical uncertainty surrounding the average esti-mates; adjustment of mean values to reflect surveydesign and selective follow-up; and missing data onreported service use and cost.

Statistical uncertainty surrounding estimatedmeans was measured using standard errors, and95% confidence intervals were generated via non-parametric bootstrapping of cost data. The right-skewed nature of costs within the sample invalidatesstandard methods of inference and uncertaintymeasurement which assume normality (Thompson& Barber, 2000). The bootstrap involves repeatedsampling with replacement from the cost data (1,000replications in this instance) to generate a samplingdistribution of mean costs upon which the


doi:10.1111/jcpp.12055 Economic impact of childhood psychiatric disorder on public sector services 979

bootstrapped standard errors are based. This non-parametric approach makes no prior assumptionsregarding the distribution of cost data across thepopulation of interest.

Adjustments to the cost estimates for samplingdesign were made by applying existing samplingweights developed by the survey investigators andthe Office of National Statistics (ONS) to accuratelyweight according to prevalence of types of disorderat general population level (Ford et al., 2005).These weights were multiplied by sampling weightsdeveloped by the ONS to adjust for regional over-sampling and to weight main baseline survey databack to the general population with respect to ageand gender. Population weights and adjustmentsfor sample clustering were also applied to the costdata to account for clustering effects withinprimary sampling units. Taylor series linearizationmethods (Heeringa and Liu, 1998; Stata Corpora-tion, 2003) were used to adjust estimated meansand standard errors using derived weights andadjustment factors.

Estimates of total costs at individual level werebased upon the sum of interactions with nearly 30types of services or professionals at each time period.Not all records included complete data on service usefrequency or intensity: complete data on frontlineeducation resources were available for less than halfthe sample; between 19% and 24% of the samplewere missing some data for other services. Multipleimputation was used to replace missing individualcost values (Graham, 2009; Schafer, 1997).Reflecting the stochastic nature of the imputedvalues, multiple imputation was used to derive fivecomplete datasets on the basis of observed values.A separate variable for intensity was imputed as aconditional variable on the basis of an imputeddichotomous indicator for receipt to maximise homo-geneity between observed and imputed values. Tominimise the amount of observed data discarded,missing values were imputed for individual servicesseparately by follow-up period. This also allowed usto utilise the strong correlation between time 2 andtime 3 service use indicators in the imputationprocess. Values generated were checked manuallyto ensure that no nonsensical values were generated.

ResultsTable 1 summarises the cost findings over the entire3-year follow-up period, based solely on observed(non-imputed) cost data.

Table 2 presents estimates of mean cost per childfor the 3-year period for all children with psychiatricdisorders, alongside annual equivalents. Estimatesrelate to all children and young people identified ashaving a disorder at baseline irrespective of whetherparents reported service use. Aggregated servicecategory costs have been calculated using multipleimputation techniques described earlier.

The findings illustrate stark differences in thedistribution of resources provided to children andyoung people with psychiatric disorder, and empha-sise the large impact on the education sector. Over3 years, the mean cost of support per child deliv-ered by frontline education services across allchildren with a disorder is over 11 times the meanfor mental health service contacts. Bootstrappedconfidence intervals are fairly wide, and reflectvariance in costs as well as the relatively smallsample size and level of missing data. Costs aregenerally higher for children with behavioural dis-orders (hyperkinetic and conduct) compared withchildren with emotional disorders (Table 3 andFigure 2). The difference between disorders for eachservice setting are not significant, although thesample may be inadequately powered for detectingdifferences in mean costs by disorder type. Thegeneral pattern again points to a higher resourceimpact on education services. Primary care andpaediatric/children’s health services are the lowestcost service category.

National annual cost estimates for the populationof children with any behavioural or emotionaldisorder aged 5–15 living in Britain were calculatedby combining mean costs in Table 2 with currentprevalence estimates derived from the BCAMHS(Meltzer et al., 2000) and population estimates forthe 5–15 age group using ONS mid-2008population estimates. Results are summarised inTable 4. Overall, the health, education and socialservice costs of mental health problems forchildren aged 5–15 years amounted to at least£1.465 billion.

DiscussionApproximately one in ten children aged 5–15 inBritain has conduct disorder, hyperactivity or emo-tional disorders as defined by ICD-10 criteria (Melt-zer et al., 2000). Responding to the needs of thesechildren has inevitable economic implications. Thehigher costs associated with education service con-tacts indicated by our findings arose because of highlevels of utilisation relative to other services both interms of prevalence and frequency of contacts (Fordet al., 2005). This may relate to easier access or morechildren with low levels of need within schools whowould not reach current thresholds for interventionby specialist mental health services. In many cases,parents would not be aware that their child had amental disorder. Many children with significantpsychosocial problems were being taught withinspecial schools, the most costly of the educationresource items. By contrast, only three children inthe sample were admitted to an inpatient psychiatricunit.

The survey could not collect data about the extentto which psychiatric disorder was recognised and thetypes of intervention delivered within these service



contacts. Some children with repeated behaviour-related education and specialist educational inputmight have had undetected psychiatric disorders.Therefore there might be scope to reduce the costs tothe education sector by improving detection andappropriate intervention within schools.

Differences in costs across disorder types wereapparent, mirroring previous findings in both the UK(e.g. Petrou et al., 2010) and the U.S. (e.g. Committeeon Prevention of Mental Disorders & SubstanceAbuse among Children, Youth & Young Adults,2009). The higher mean costs for behavioural disor-

Table 2 Mean costs per child over 3-year follow-up for children with a psychiatric disorder: mean costs for all disorders (N = 445)

Mean cost per user:all children with a psychiatric

disorder over 3 years (£) 95% CI (lower) 95% CI (upper)Mean annual

cost (£)

Primary care 58.19 32.42 83.97 19.40Paediatrics/children’shealth services

96.86 58.07 135.65 32.29

Mental health services 236.75 162.29 311.22 78.92Frontline education 2,949.22 1,627.31 4,271.12 983.07Special education 1,858.89 967.89 2,749.90 619.63Social care 208.85 96.65 321.06 69.62Total cost 5,408.77 3,687.27 7,130.27 1,802.92

Costs estimated for all 445 cases with a psychiatric disorder included in 3-year follow-up. Costs adjusted for selection and non-response in follow-up sample and for sample clustering in main survey design at baseline. Imputed values used for missing serviceuse data. Total costs are the sum of costs across each service category for each child/young person.

Table 1 Mental health-related service use and costs by service setting over 3-year follow-up for children and young people withpsychiatric disorder (N = 445)

Proportion inreceipt (%)

Mean cost peruser (£) SD (£) N

Primary careGeneral practitioner 24.9% 190.11 435.09 361Health visitor 2.5% 317.00 576.31 366

Paediatrics/children’shealth servicePaediatric inpatient 1.4% 335.74 122.40 367Paediatrician 9.6% 572.83 516.17 363A & E 2.2% 86.25 31.94 361Community nurse 0.5% 401.37 538.94 368School nurse 4.6% 47.66 51.48 351Dietician 1.1% 32.34 12.94 366Occupational therapy 0.8% 265.93 209.08 368Speech therapy 1.9% 2,007.46 3,174.12 365Physiotherapy 1.6% 319.48 192.19 367

Mental health servicesPsychiatric inpatient 0.3% 1,988.00 367Child psychiatrist 7.1% 1,457.12 1,470.88 366Psychologist 10.7% 334.62 582.26 366Family therapy 0.5% 398.97 43.40 368Counselling 3.8% 192.97 209.17 368School counselling 4.7% 746.76 1,132.81 360Community psychiatric nurse 0.8% 738.28 436.24 367

Frontline educationParental meetings with teachers 45.1% 256.87 511.69 355Extra time with teachers and teaching assistants 13.1% 13,907.29 20,606.40 321Special educational needs coordinators 12.8% 139.99 146.28 360Special educational needs tribunals 3.5% 2,550.35 852.74 351

Special educationEducational social worker 6.6% 120.50 141.96 365Educational psychologist 7.4% 196.51 194.91 351Special school status 4.3% 38,104.65 27,768.39 368

Social careSocial services assessments 5.2% 1,924.70 408.48 442Social worker 6.6% 1,643.74 3,420.93 440Respite care 0.7% 1,512.00 654.72 439



ders partly reflect the greater likelihood of contactwith services among children and young people witheither a behavioural and particularly a hyperkineticdisorder (Ford, Goodman & Meltzer, 2003). The

difference may also reflect genuine differences inresource requirements for dealing with differentdisorders once service contact has been established(Evans-Lacko, dosReis & Kastelic, 2011). Behavio-ural problems – because of their more externalisingand potentially more disruptive nature – may bemore likely to provoke responses from parents interms of seeking professional help, as well as fromteachers and other education professionals (cf. Robbet al., 2011). Previous studies have highlighted thehigh and wide-ranging costs of hyperkinetic andbehavioural problems in childhood and adolescence(e.g. Fletcher & Wolfe, 2009; Foster & Jones, 2005;Jones & Foster, 2009; Knapp, Scott & Davies, 1999;Romeo, Knapp & Scott, 2006).

By contrast, internalising difficulties arising fromanxietyanddepressionmaybemoredifficult todetect,or may be interpreted by parents or teachers as insome sense ‘normal’ or not in need of treatment. Otherstudies have demonstrated that depressive symptomsamong children are associated with lower levels ofhealth service use (Riley et al., 1993), although costscan be high (Bodden, Dirksen & Bogels, 2008). More-over, early-onset anxiety anddepressionmaygeneratesignificant costs in adulthood in terms of psychiatricdisorder, substance misuse and ensuing impacts onproductivity (Kim-Cohen et al., 2003; Knapp, King,Healey & Thomas, 2011).

Table 3 Mean costs per child over 3-year follow-up by psychiatric disorder (N = 445)

Mean cost per user:all children with a psychiatric

disorder over 3 years (£) 95% CI (lower) 95% CI (upper)Mean annual

cost (£)

Hyperkinetic disordersPrimary care 46.02 18.61 73.43 15.34Paediatrics/children’shealth services

209.10 57.15 361.05 69.70

Mental health services 551.27 287.79 814.75 183.76Frontline education 5,286.88 1,697.03 8,876.73 1,762.29Special education 2,888.47 360.42 5,416.53 962.82Social care 342.33 0 866.29 114.11Total cost 9,324.08 4,609.53 1,4038.60 3,108.03

Conduct disordersPrimary care 58.46 26.80 90.12 19.49Paediatrics/children’shealth services

78.55 37.18 119.92 26.18

Mental health services 247.24 132.95 361.53 82.41Frontline education 3,041.13 1,125.30 4,956.96 1,013.71Special education 1,854.68 367.35 3,342.02 618.23Social care 289.26 122.77 455.76 96.42Total cost 5,569.32 3,093.25 8,045.39 1,856.44

Emotional disordersPrimary care 61.19 19.24 103.15 20.40Paediatrics/children’shealth services

79.33 19.91 138.76 26.44

Mental health services 126.19 28.87 223.52 42.06Frontline education 2,087.20 276.24 3,898.17 695.73Special education 1,056.97 0 2,252.62 352.32Social care 84.80 10.76 158.85 28.27Total cost 3,495.70 1,041.62 5,949.78 1,165.23

Costs estimated for all 445 cases with a psychiatric disorder included in 3-year follow-up. All costs adjusted for selection and non-response in follow-up sample and for sample clustering in main survey design at baseline. Imputed values used for missing serviceuse data. Total costs are the sum of costs across each service category for each child/young person.

£0

£1000

£2000

£3000

£4000

£5000

£6000

£7000

£8000

£9000

£10,000

Hyperkine cdisorders

Conductdisorders

Emo onaldisorders

Mea

n co

st (£

)

Primary care Paediatrics/children’s health services

Mental health services Frontline educa on resources

Special educa on resources Social care services

Figure 2 Health, social service and education resource use: meancost over 3-year follow-up for all children/young people with adisorder



Strengths and limitations

Like any cost-of-illness calculation, the present studycan contribute to policy and practice discussions bysetting out the wide-ranging economic impacts ofdisorders. In this case, we focused on the impacts onthree major public sector systems: health, educationand social care. Although informative, cost-of-illnessstudies have their limitations (Byford, Torgerson &Raftery, 2000). In particular, they cannot provide thebasis for resource allocation decisions because theydo not offer insights into the efficiency with whichresources are used; for that purpose, decision-mak-ers need evidence from cost-effectiveness analyses.

Our estimates were constructed from the firstnational epidemiological survey of child and adoles-cent mental health problems in Britain, using datacollected from parents through face-to-face andtelephone interviews. Earlier reviews of this fieldnoted a dearth of economic evaluations (Knapp,1997; Romeo, Byford & Knapp, 2005). Since then,studies have begun to offer decision-makers insightsinto economic implications of different courses ofaction. However, almost all previous economic stud-ies are based on local data collections, postal ques-tionnaires or (incomplete) agency administrativerecords. Consequently, although we set out limita-tions to our study below, the data analysed hereprovide considerable incremental methodologicalgains, not least because the data were generatedfrom an unusually large, national, population-basedsurvey, with carefully designed follow-ups.

Children and parents from more disadvantagedsocial backgrounds were under-represented in thefollow-up sample (Ford et al., 2003): this type ofdrop-out may bias downwards health service util-isation estimates (Reijneveld & Stronks, 1999). Thisgroup would have been expected to have higherdisorder prevalence, particularly the disruptive dis-orders that lead to extensive contacts within theeducation system. Whether this is as serious aproblem as regards the reporting of education ser-vice contacts is open to question, although, as notedearlier, not all parents who reported contact withschool staff were approached for interviews at time 2.

Ford et al. (2003, 2005) previously highlightedissues associated with the service utilisation esti-mates. As we were unable to interview all parents atfollow-up, this is likely to have led to underestima-tion of service contacts. in addition, a few childrenwere assigned diagnoses on the basis of teacher andself-report. Exclusion of children who are lookedafter by local authorities who have high rates ofpsychiatric disorder and complex needs will alsomean that we are underestimating actual costs of allchildren with psychiatric disorder (Meltzer, Gatward,Corbin, Goodman & Ford, 2003).

Cost estimates can only be as reliable as theresource utilisation data upon which they are based.All service contact data were based on parent reportusing semi-structured interviews. Self-report datamay be open to recall bias in. Despite these caveatsthere is evidence that parent-reported or patient self-reported service contacts are at least as reliable asother sources, including administrative records(Patel, Rendu, Moran, Mann & Knapp, 2005; Stiff-man et al., 2000). Moreover, the instrument used toassess service use in this study has evidence ofmoderate to strong reliability and validity (Fordet al., 2007).

The statistical uncertainty surrounding estimatedmean costs is considerable, as evidenced by wideconfidence intervals, partly due to the limited num-ber of children with identified disorders in the follow-up dataset, combined with both large variation inobserved costs and use of imputations for missingdata. The degree of uncertainty is likely to have beenunderestimated given that unit costs were implicitlytreated as nonstochastic and without samplingerror.

We were unable to study contact with police andother youth justice services due to the low frequen-cies and high rate of missing data. Crime-relatedcosts (criminal justice system costs only) are likely tobe a significant component of overall costs amongyounger people aged between 10 and 28 (Scott,Knapp, Henderson & Maughan, 2001).

Mental health service costs reported here rely onpublished unit cost data that include cost of medi-cation administered to children with psychiatricdisorders, but this element cannot be separated.While levels of receipt are likely to be high (Fordet al., 2003), the associated costs are generally smallin comparison to many other services described.

Policy considerations

The education system bears the largest public sectorcost burden in relation to psychiatric disordersamong children and young people. These costs arenot restricted to delivery of specialist services tar-geting children and young people with mental healthneeds; there are also sizeable costs incurred withinmainstream schools, including provision of addi-tional teaching inputs and time spent meeting and

Table 4 Annual national costs of mental health service use forpopulation aged 5–15 with emotional/behavioural disorder

Population aged5–15 with disorder Service type/setting

National costestimate(£ million)

813,000 Primary care 15.8Paediatric/children’shealth services

26.2

Mental health services 64.2Frontline education 799.2Special education 503.8Social care 56.6Total cost 1,465.8

Figures may not add up to total shown due to rounding.



discussing problems with parents. Importantly, con-tact with services was not synonymous with appro-priate identification and management of mentalhealth needs. From a public policy perspective, theseresource commitments may imply diversion of edu-cation resources away from other socially beneficialactivities within or outside the education sector,although considerable work has been done recentlyto increase awareness and effective management ofemotional and behavioural issues within schools inrecognition of their impact on the ability of pupils toaccess the school curriculum. Although not evalu-ated here, early identification and effective interven-tion before problems develop to the levels requiringspecialist mental health services and/or specialisteducational input might reduce costs, particularly inthe education sector (Allen, 2011). Our findingsunderline the importance of evaluating such initia-tives, particularly in an era of fiscal austerity, whenthey might be especially vulnerable.

Longer term and wide-ranging societal costs inadulthood associated with childhood hyperkinetic,behavioural and emotional problems observed inother studies (Aos, Lieb, Mayfield, Miller & Pennucci,2004; Fletcher & Wolfe, 2009; Healey, Knapp &Farrington, 2004; Knapp, McCrone, Fombonne, Bee-cham & Wostear, 2002; Knapp et al., 2011; Scottet al., 2001) suggest that the individual, family andsocietal benefits of limiting or preventing these typesof difficulties could be considerable. When theyreach early adulthood, children with emotionalproblems or ADHD are likely to have more employ-

ment difficulties and lower earnings (Knapp et al.,2011). Many adult psychiatric disorders have theirroots in childhood: half the adults with psychiatricdisorder at age 26 in the Dunedin cohort had apsychiatric disorder before age 15, increasing tothree-quarters by age 18 (Kim-Cohen et al., 2003). Itis possible that effective intervention in childhoodmight alter developmental trajectories in a way thatcould provide significant savings in adulthood aswell as childhood (Aos et al., 2004; Bonin, Stevens,Beecham, Byford & Parsonage, 2011).

In summary, this study provides national-levelevidence on the enormous impact of childhoodpsychiatric disorder on public sector service bud-gets. The highest service-related cost impacts fell onthe education system, particularly on schools. Thesenew figures help describe the context within which itis possible to explore the cost-effectiveness of inter-ventions to support children with emotional andbehavioural disorders, both in schools and else-where.

AcknowledgementsThis article reports independent research funded by theUK Department of Health (grant 035/0045). Viewsexpressed are those of the authors and not necessarilythose of the Department of Health

CorrespondenceMartin Knapp, LSE, Houghton Street, LondonWC2A2AE, UK; Email: [email protected]

Key points

• One in ten children aged 5–15 in Britain has a conduct, hyperactivity or emotional disorder.

• Estimated additional health, social care and education costs associated with childhood psychiatric disorderstotalled £1.47bn in 2008.

• Higher mean costs were estimated for children with hyperkinetic disorders than for those with conduct oremotional disorders.

• The education system bears by far the largest cost burden of all public sector services.

References

Allen, G. (2011). Early intervention; the next steps. London:Cabinet Office.

American Psychiatric Association. (1994). Diagnostic andstatistical manual of mental health disorders (4th edn).Washington, DC: American Psychiatric Association.

Aos, S., Lieb, R., Mayfield, J., Miller, M., & Pennucci, A. (2004).Benefits and costs of prevention and early interventionprograms for youth. Olympia: Washington State Institutefor Public Policy.

Ascher, B.H., Farmer, E.M.Z., Burns, B.J., & Angold, A. (2007).The Child and Adolescent Services Assessment (CASA):Description and psychometrics. Journal of Emotional andBehavioral Disorders, 4, 12–20.

Beecham, J., & Knapp, M. (2001). Costing psychiatricinterventions in measuring mental health needs. In G.

Thornicroft (Ed.), Measuring mental health needs (2nd edn).London: Gaskell.

Bodden, D.H., Dirksen, C.D., & Bogels, S.M. (2008). Societalburden of clinically anxious youth referred for treatment: Acost-of-illness study. Journal of Abnormal Child Psychology,36, 487–497.

Bonin, E.-M., Stevens, M., Beecham, J., Byford, S., &Parsonage, M. (2011). Costs and longer-term savings ofparenting programmes for the prevention of persistentconduct disorder: A modelling study. BMC Public Health,11, 803.

Byford, S., Torgerson, D., & Raftery, J. (2000). Cost of illnessstudies. British Medical Journal, 320, 1335.

Committee on Prevention of Mental Disorders and SubstanceAbuse among Children, Youth and Young Adults. (2009).Preventing mental, emotional, and behavioral disorders



among young people: Progress and possibilities. Washington,DC: National Academies Press.

Curtis, L. (2007). Unit costs of health and social care 2007.Canterbury: Personal Social Services Research Unit,University of Kent.

Curtis, L. (2008). Unit costs of health and social care 2008.Canterbury: Personal Social Services Research Unit,University of Kent.

Department of Health. (2003). Assessing children’s needs andcircumstances: The impact of the assessment framework.London: Department of Health.

Evans-Lacko, S., dosReis, S., & Kastelic, E. (2011). Mentalhealth service use before and after diagnosis of early-onsetbipolar disorder. Journal of Behavioral Health Services andResearch, 38, 398–413.

Fletcher, J., & Wolfe, B. (2009). Long-term consequences ofchildhood ADHD on criminal activities. Journal of MentalHealth Policy and Economics, 12, 119–138.

Ford, T., Goodman, R., & Meltzer, H. (2003). Service SSE over18 months among a nationally representative sample ofBritish children with psychiatric disorder. Clinical ChildPsychology and Psychiatry, 8, 37–52.

Ford, T., Hamilton, H., Dosani, S., Burke, L., & Goodman, R.(2007). The children’s service interview: validity and reliability.Social Psychiatry and Psychiatric Epidemiology, 42, 36–49.

Ford, T., Hamilton, H., & Goodman, R. (2005). Service contactsamong the children participating in the British Child andAdolescent Mental Health Surveys. Children and AdolescentMental Health, 10, 2–48.

Foster, M., & Jones, D. (2005). The high costs of aggression:Public expenditures resulting from conduct disorder.American Journal of Public Health, 95, 1767–1772.

Goodman, R., Ford, T., Richards, H., Gatward, R., & Meltzer,H. (2000). The development and well-being assessment:Description and initial validation of an integratedassessment of child and adolescent psychopathology.Journal of Child Psychology and Psychiatry, 41, 645–655.

Graham, J.W. (2009). Missing data analysis: Making it work inthe real world. Annual Review of Psychology, 60, 549–576.

Healey, A., Knapp, M., & Farrington, D. (2004). Adult labourmarket implications of antisocial behaviour in childhoodand adolescence: Findings from a UK longitudinal study.Applied Economics, 36, 93–105.

Heeringa, S.G., & Liu, J. (1998). Complex sample design effectsand inference for mental health survey data. InternationalJournal of Methods in Psychiatric Research, 7, 56–65.

Jones, D.E., & Foster, E.M. (2009). Service use patterns foradolescents with ADHD and comorbid conduct disorder.Journal of Behavioral Health Services Research, 36, 436–449.

Kim-Cohen, J., Caspi, A.,Moffitt, T.E.,Harrington,H.,Milne,B.J.,& Poulton, P.J. (2003). Prior juvenile diagnoses in adults:Developmental follow back of a prospective longitudinal cohort.Archives of General Psychiatry, 60, 709–717.

Knapp, M. (1997). Economic evaluation and interventions forchildren and adolescents with mental health problems.Journal of Child Psychology and Psychiatry, 38, 3–26.

Knapp, M., King, D., Healey, A., & Thomas, C. (2011).Economic outcomes in adulthood and their associationswith antisocial conduct, attention deficit and anxietyproblems in childhood. Journal of Mental Health Policy andEconomics, 14, 122–132.

Knapp, M., McCrone, P., Fombonne, E., Beecham, J., &Wostear, G. (2002). The Maudsley long-term follow-up ofchild and adolescent depression: Impact of comorbid

conduct disorder on service use and costs in adulthood.British Journal of Psychiatry, 180, 19–23.

Knapp, M., Scott, S., & Davies, J. (1999). The cost of antisocialbehaviour in younger children. Clinical Child Psychology andPsychiatry, 4, 457–473.

Lord Chancellor’s Department. (2001). Tribunals for users: Onesystem one service (Leggatt review). London: The StationeryOffice.

Meltzer, M., Gatward, R., Corbin, T., Goodman, R., & Ford, T.(2003). Persistence, onset, risk factors and outcomes ofchildhood mental disorders. London: The Stationery Office.

Meltzer, H., Gatwood, R., Goodman, R., & Ford, T. (2000).Mental health of children and adolescents in Great Britain.London: The Stationery Office.

Netten, A., & Curtis, L. (2003). Unit costs of health and socialcare 2003. Canterbury: Personal Social Services ResearchUnit, University of Kent.

Patel, A., Rendu, A., Moran, P., Mann, A., & Knapp, M. (2005).A comparison of two methods of collecting economic data inprimary care. Family Practice, 22, 323–327.

Petrou, S., Johnson, S., Wolke, D., Hollis, C., Kochhar, P., &Marlow, N. (2010). Economic costs and preference-basedhealth-related quality of life outcomes associated withchildhood psychiatric disorders. British Journal ofPsychiatry, 197, 395–404.

Reijneveld, S.A., & Stronks, K. (1999). The impact of responsebias on estimates of health care utilization in a metropolitanarea: The use of administrative data. International Journal ofEpidemiology, 28, 1134–1140.

Riley, A.W., Finney, J.W., Mellits, D., Starfield, B., Kidwell, S.,Quaskey, S., … & Shematek, J.P. (1993). Determinants ofchildren’s health care use: An investigation of psychosocialfactors. Medical Care, 31, 767–783.

Robb, J.A., Sibley, M.H., Pelham, W., Foster, E.M., Molina, B.,& Gnagy, E. (2011). The estimated annual cost of ADHD tothe US education system. School Mental Health, 3, 169–177.

Romeo, R., Byford, S., & Knapp, M. (2005). Economicevaluations of child and adolescent mental healthinterventions: A systematic review. Journal of ChildPsychology and Psychiatry, 46, 919–930.

Romeo, R., Knapp, M., & Scott, S. (2006). Economic cost ofsevere antisocial behaviour in children – And who for pays it.British Journal of Psychiatry, 188, 547–553.

Schafer, J.L. (1997). Analysis of incomplete multivariate data.New York: Chapman & Hall.

Scott, S., Knapp, M., Henderson, J., & Maughan, B. (2001).Financial cost of social exclusion: follow up study ofantisocial children into adulthood. British Medical Journal,323, 191–193.

Stata Corporation. (2003). Statistics/data analysis. Texas:College Stationery, Stata Press.

Stiffman, A.R., Horwitz, S.M., Compton, W., Cottler, L., Bean,D.L., Narrow, W.E., & Weisz, J.R. (2000). The ServiceAssessment for Children and Adolescents (SACA): Adultand child reports. Journal of the American Academy of Childand Adolescent Psychiatry, 39, 1032–1039.

Thompson, S.G., & Barber, J.A. (2000). How should cost datain pragmatic randomised trials be analysed? British MedicalJournal, 320, 1197–1200.

World Health Organization. (1990). International classificationof diseases (ICD-10). Geneva: World Health Organization.

Accepted for publication: 17 December 2012Published online: 27 February 2013



economic impact of childhood psychiatric disorder on public sector services in britain: estimates...

Documents