Parent ratings of school behaviour inchildren at risk of attention deficit/hyperactivity disorder

Introduction

Attention deficit/hyperactivity disorder (ADHD) isa childhood neuro-developmental disorder withhigh impact for affected children and their families.It is defined by persistent traits of pervasiveinattention or overactivity/impulsiveness resultingin impairment and affects up to 5% of school-agechildren (1). The accurate identification of affectedchildren is important as effective evidence-basedtreatments are available (2). ADHD has importantimplications for later life and increases the risk forother psychiatric disorders including conduct andpersonality disorders; substance misuse; educa-tional and employment difficulties; and offending(3, 4).Current diagnostic criteria for hyperactivity

disorders specify that symptoms or associatedimpairment need to be pervasive (occurring in atleast two settings such as home and school). Inboth clinical and research practice, parents areoften asked about their child’s behaviour andfunctioning at school. However, there has beenlittle systematic work comparing these ratings with

actual teacher ratings. Amongst children withpervasive hyperactivity, this study investigates therelationships between ratings completed by theparent, teacher, and parent accounts of schoolbehaviour with regard to child hyperactivity symp-toms and associated impairment. Such parentalaccounts reflect parental perceptions about schooland are an accessible source of information whenreferral is being considered or during routinemonitoring in clinics. As well as concordancebetween respondents, a better understandingabout differences can be clinically informative indeveloping management plans, for example whe-ther the initial setting for intervention should behome or school.Although many studies (5–7) have compared

parent and teacher ratings of hyperactivity, to dateonly one published study has also obtained parentreports about in-school behaviour (8). This foundthat parents under-estimate levels of symptoms atschool. However, it involved a clinically referredsample leading to the possibility that factorsassociated with parents� ability to report theirchild’s symptoms were confounded by factors that

Sayal K, Taylor E. Parent ratings of school behaviour in children at riskof attention deficit/hyperactivity disorder.Acta Psychiatr Scand 2005: 111: 460–465.�BlackwellMunksgaard 2005.

Objective: To investigate whether parents are accurate informants ofchild hyperactivity symptoms and impairment at school.Method: Parents of a community sample of 93 children with pervasivehyperactivity completed rating scales about their child’s behaviour athome and school. These were compared with teacher ratings.Results: Parent ratings about school correlate more closely with parent(home) than teacher ratings. Such ratings systematically under-estimate teacher ratings and are influenced by the child’s behaviour atboth home and school as well as parental mental health. However, aparental report of impairment for the child at school is likely to beaccurate.Conclusion: There are limitations in relying on parental accounts ofschool behaviour if teacher ratings are unavailable. As such ratingsmay under-identify children with ADHD and discrepancies betweenparent and teacher ratings may reflect actual differences in behaviour,this suggests that ratings are required from both sets of informants.

K. Sayal, E. TaylorDepartment of Child Psychiatry, Institute of Psychiatry,London, UK

Key words: parent; children; school; attention-deficithyperactivity disorder

Dr Kapil Sayal, Department of Child Psychiatry, BoxP085, Institute of Psychiatry, Denmark Hill, London SE58AF, UK.E-mail: k.sayal@iop.kcl.ac.uk

Accepted for publication November 8, 2004

Acta Psychiatr Scand 2005: 111: 460–465All rights reservedDOI: 10.1111/j.1600-0447.2004.00487.x

Copyright � Blackwell Munksgaard 2005

ACTA PSYCHIATRICASCANDINAVICA

460

led to the referral. For example, compared withhyperactive children in the community, referredchildren are more likely to have comorbid symp-toms (especially conduct problems) and familyadversity but not greater severity of hyperactivity(9–12). The present investigation is unique as itinvolves a community sample that was identifiedafter obtaining teacher ratings that established thepresence of pervasive hyperactivity. Even if suchsymptoms do not reach criteria for ADHD,pervasive hyperactivity is a persistent mentalhealth problem and a risk factor for later develop-ment (13). The majority of clinically referredchildren have attention or hyperactivity difficulties;such problems being strongly associated withservice use (14–16). However, the converse situ-ation does not apply; few children with pervasivehyperactivity in the community receive specialistservices (9, 10). Therefore, the present sample bothreflects referred children and is also representativeof children in the community with pervasivehyperactivity.Previous research in referred samples has also

indicated that an affirmative parent report aboutparticular symptoms that leads to a diagnosis ofADHD (based on parental account) also predictsa positive teacher report for the same symptoms(17–19). Although diagnostic criteria stress thesalience of impairment across settings, it is notknown whether parents can accurately reportimpairment at school. This study investigateswhether these previous findings involving referredsamples generalize to community samples. Afurther advantage of investigating the relation-ship between parent and teacher ratings in apervasively hyperactive sample is that there isalready some agreement about the child’s beha-viour across settings. This means that any furtherdifferences in ratings might be more likely torelate to the source of informant than to reflectsituational factors.

Aims of the study

In a community sample of children with pervasivehyperactivity, this study systematically evaluateswhether parents are accurate informants abouttheir child’s hyperactivity symptoms and associ-ated impairment at school. It examines whetherpossible differences in ratings reflect the type ofinformant (parent or teacher) or the setting (homeor school) of the behaviour. It is hypothesized thatparent–teacher correlations would be weaker thanthose involving the other sets of combinations(ratings about the child in the same setting orwhere made by the same rater).

Material and methods

Sample

The parents of 2992, 5–11 year old children regis-tered with five randomly selected primary carepractices in South London were invited to partici-pate in the study. They were asked to complete theStrengths and Difficulties Questionnaire (SDQ)which is a well-validated measure of childhoodmental health (20, 21). Pervasive hyperactivity wasdefined as a score of 6 or above on the hyperac-tivity scale (range 0–10) of both the parent andteacher SDQs.Parental SDQs were received for 1194 (40%)

children; the response rate across the five practicesranged from 34 to 57%. The possibility of aresponse bias was investigated at the individuallevel. The proportion of children (21%; 248/1194)scoring above the hyperactivity cut-off was asexpected (population proportion of 20%; 20). Thedistributions of the other SDQ symptom scoresand child gender were also as expected. Althoughresponse was associated with both child age andthe household Jarman under-privileged area scores(UPA; 22), these had small effect sizes (0.10 and0.27, respectively). The Jarman UPA score is aneighbourhood-level deprivation index derivedfrom national census data. As it is a populationmeasure, scores cannot be ascribed to individualhouseholds. Further details about sample recruit-ment are described elsewhere (10).Teacher SDQs were received for 96% (237/248)

of children with parent-identified hyperactivity. Ofthese, 110 (9% of the whole sample) children hadpervasive hyperactivity. Parents of 93 of thesechildren were interviewed and they constitute thesample for this investigation. Interviews wereusually (95%; 88/93) carried out with the mother.The mean (sd) age of the children was 8.18 (1.69)years and 76 (82%) were male.

Measures

Hyperactivity, difficulties, impact and burden rat-ings from the SDQ were utilized. During theinterview, parents were asked to complete part ofthe teacher SDQ (�parent-completed teacher SDQ�)in relation to their understanding of their child’sbehaviour at school. This information included thehyperactivity scale, whether the parent thought theschool regard the behaviour as a problem and, ifapplicable, impact for the child (social impairmentand distress) and burden for the teacher and theclass. These ratings were compared with the parent(home) and teacher (school) SDQs. Thus data fromthe parent-completed teacher SDQ inform whether

Parent ratings of ADHD

461

parent ratings can substitute for actual teacherratings. For the parent SDQ impact measure, onlythe three components that are also in the teacherSDQ were included (distress, peer relationships,classroom learning). The item on classroom ratingsis also reported separately as this reflects an aspectthat is specific to school. The parent also comple-ted the Malaise Inventory (23) which is a com-monly used questionnaire to assess parentalpsychiatric morbidity. It is a 24-item scale withitems covering emotional disturbance and associ-ated somatic symptoms.

Analysis

The comparisons reflect situations where theinformant source or the setting is systematicallyaltered. The baseline situation involves parent andteacher ratings. First, for the school setting, theinformant source is altered from the teacher to theparent (using the parent-completed teacher rating).Second, for the parent as the informant source, thesetting is altered from home to school (using theparent-completed teacher rating). Therefore, com-pared with the baseline situation, the combinationof teacher and parent-completed teacher ratingsreflects a change in rater and the combination ofparent and parent-completed teacher ratingsreflects a change in setting.Correlations and differences (using paired

t-tests) between the three sets of measures werecalculated. Linear regression analyses examinedthe proportion of variance in hyperactivity ratingsthat could be explained by the malaise score. Withregard to impairment, the probability [positivepredictive value (PPV)] is examined that theteacher report for an item will be positive giventhat the parent report for the same item is positive.In clinical practice, these values may be moreuseful than correlations and sensitivity or specific-ity measures. The PPV is appropriate for sampleswhere all children score above the cut-off for aparticular informant. The analysis is extendedfurther by also examining the probability that theteacher report for an item will be positive giventhat the parent-completed teacher report for thesame item is positive. Conditional probabilityanalyses examined positive (PPV) and negativepredictive values (NPV) based on dichotomousyes/no measures for two sets of ratings. A score of0 or 1 on the impact measure was coded as �no� andhigher scores as �yes� (21). Based on previousstudies (17–19), Table 1 shows the example of thecomparison between parent and teacher ratings.The PPV is the conditional probability that theteacher reports an item (Teacher yes), given that

the parent’s report was positive (Parent yes). TheNPV is the conditional probability that the teacherdoes not report an item (Teacher no), given thatthe parent’s report was negative (Parent no).

Results

All pairs of ratings were positively correlated(Table 2). The magnitude of these correlationswas low or moderate. As expected, correlationswere stronger if both ratings were done by the sameperson or were about the child in the same setting.Furthermore, although the parent-completed tea-cher ratings were moderately correlated with tea-cher ratings, they were more highly correlated withparent ratings. However, all correlations betweenthe teacher and parent-completed teacher ratingsremained significant (at P < 0.01 level), althoughslightly diminished in magnitude, when controllingfor the parent ratings.In terms of differences between the measures,

Table 3 shows that the parent-completed teacherscores were lower than the other two sets of scores.They were significantly lower than parent ratingsfor two of the measures and lower than teacherratings for all of the measures. The only significantdifferences between parent and teacher ratings werein the direction of lower parent ratings. Theindividual contributions to the variance in theparent-completed teacher hyperactivity scores wereteacher hyperactivity rating (7%), parent hyperac-tivity rating (20%), and parent malaise score

Table 1. Statistical measures describing conditional probability analyses involvingparent and teacher ratings

Teacher yes Teacher no

Parent yes a bParent no c dBase rate Parent a + b/nBase rate Teacher a + c/nPPV a/a + bNPV d/c + d

n, number of children; Base rate, proportion of children with a positive report of anitem.PPV, positive predictive value; NPV, negative predictive value.

Table 2. Correlations between the three sets of measures

SDQ measures(range)

Parent andteacher

Teacher andparent–teacher

Parent andparent–teacher

Hyperactivity (0–10) 0.13 0.27* 0.47**Difficulties (0–3) 0.27** 0.38** 0.56**Classroom learning (0–2) 0.17 0.30** 0.36**Impact (0–6) 0.28** 0.40** 0.52**Burden (0–3) 0.19 0.42** 0.44**

Pearsons Correlation Coefficients; *P < 0.05; **P < 0.01.

Sayal and Taylor

462

(25%). After controlling for child age and genderand household Jarman UPA score, these variablescollectively accounted for 42% of the variance withthe malaise score making an additional contribu-tion once the parent and teacher ratings wereaccounted for. In contrast, the malaise score madeno contribution to the teacher hyperactivity ratingand only accounted for 4% of the variance for theparent hyperactivity rating.Table 4 shows findings from the conditional

probability analyses comparing teacher ratingswith both parent and parent-completed teacherratings. The higher PPV involving the parent-completed teacher impact measure implies that aparent report of impact at school was usuallyconfirmed by the teacher. However, the absence ofreported impact was less likely to be confirmed bythe teacher.

Discussion

This study investigated whether parent ratingsabout hyperactivity-related symptoms andimpairment at school can substitute for teacherratings. Parent-completed teacher ratingsappeared to reflect the child’s behaviour at bothhome and school as well as maternal mood.These ratings systematically under-estimated tea-cher ratings suggesting that their routine use mayunder-identify children with ADHD and thatfurther information is required in these situa-tions. However, the conditional probability

analyses indicated that a parent report of impair-ment at school is likely to be an accurateaccount. As hypothesized, the correlationsbetween parent and teacher ratings were weakerthan the other two combinations of ratings. Thissuggests that children exhibit different behavioursacross settings and/or that there are variationsbetween raters� perceptions.The finding, in this community sample, that the

correlations between teacher ratings and parentratings about the child’s behaviour at school wereintermediate compared with the two other sets ofcorrelations is in contrast to previous findings in areferred sample (8). Parental perception of prob-lems and impact on the family are importantdeterminants of referral (11). It is possible thatparental accounts of the school behaviour ofreferred children are coloured by the impact ofthese symptoms on home life. In contrast, whenconsidered free of possible referral bias, the find-ings suggest that such parental accounts mightcontribute additional information about teacherratings. However, despite a moderate correlationbetween parent-completed and actual teacher rat-ings, this parental proxy rating was influenced bythe child’s behaviour at home as well as parentalmental health. The finding that parent ratingsabout the child’s behaviour at school were morehighly correlated with their own ratings abouthome behaviour than with teacher ratings aboutschool behaviour replicates and extends similarfindings in referred samples (8). These findingsconfirm a �rater effect� involving parents wherebytheir ratings are determined by both the child’sbehaviour at home as well as the setting they areasked to consider. Furthermore, parental mentalhealth did not influence home ratings but played arole when the parent was asked to consider thechild’s behaviour at school.Although parent-completed teacher ratings

remain correlated with teacher ratings even whencontrolling for home ratings, they only provide apartial picture. The findings that scores on bothparent-completed teacher ratings were lower thanparent and teacher ratings were also consistentwith previous work (8). Clinically, these data areimportant in establishing that parent-completedteacher ratings systematically under-estimate theteacher’s perspective of the child’s behaviour atschool. Apriori, it might have been expected thatthese severity ratings would be intermediatebetween parent and teacher ratings. It is possiblethat parental uncertainty is reflected in caution inratings.With regard to the conditional probability ana-

lyses, the findings relating to impact measures were

Table 3. Comparisons between the three sets of measures

SDQ measures(range)

Parent(a)

Teacher(b)

Parent–teacher(c)

Pairedt-tests

Hyperactivity (0–10) 8.11 (1.46) 8.11 (1.47) 7.03 (2.13) a, b > c***Difficulties (0–3) 1.45 (0.79) 1.79 (0.75) 1.37 (0.83) b > a**; b > c***Classroomlearning (0–2)

0.94 (0.80) 1.15 (0.73) 0.77 (0.74) b > c***

Impact (0–6) 1.74 (1.56) 2.18 (1.65) 1.64 (1.59) b > a*; b > c**Burden (0–3) 1.38 (1.01) 1.62 (0.89) 1.14 (0.89) a > c*; b > c***

Mean (SD)*P < 0.05; **P < 0.01; ***P < 0.001.

Table 4. Positive Predictive Values of comparisons between teacher ratings andparent and parent-completed teacher ratings

Base rate Base rate T PPV NPV

Impact (P) 58 57 67 56Impact (PT) 45 57 78 60

Values represent percentages.Base rate, proportion of children with a positive report of an item; P, parent; T,teacher; PT, parent-completed teacher; PPV, positive predictive value; NPV, negativepredictive value.

Parent ratings of ADHD

463

consistent with previous research that examinedthe role of symptoms (17–19). Although there wasonly modest correlation between raters, concor-dance rates were higher when considered in termsof PPVs. Both PPVs and NPVs were higher whenparent-completed teacher ratings were used inpreference to parent ratings. The PPV of 78% forthe composite impact measure (reflecting distressand impairment) suggests that if the parent reportsimpact for the child at school, then the teacher islikely to agree. However the lower NPV (60%)implies that a negative parent report of impact forthe child at school does not accurately indicate theteacher’s opinion.

Research and clinical implications

These findings relating to parent reports aboutschool only apply to children with pervasivehyperactivity and should be considered prelimin-ary. The nature of this sample reduces the rangeof scores and this may under-estimate correla-tions in the population. However, the reportedfindings are free of the confounds of referral. Thelow parental response rate is a limitation of thisstudy and wider generalization of the findingsshould only be made cautiously. Despite this,responders appeared to be representative andthere was a very high rate of teacher participa-tion. Furthermore, these analyses are primarilywithin-subject (involving reports on the samechild). Although it is appropriate to examine theutility of parent-completed teacher ratings in asample where parents have noticed symptoms, itis uncertain whether these findings will generalizeacross the full range of teacher hyperactivityscores. Further research is required to explore thepossible clinical applications of asking parents tocomplete teacher ratings. This needs to beexamined amongst all children with parent-iden-tified hyperactivity. In this group, the full rangeof teacher and parent-completed teacher ratingscould then be compared. These studies couldinform approaches to screening and case identi-fication in research and clinical practice.This study also raises several other methodo-

logical issues. The findings are restricted toprimary school children (ages 5–11 years) whohave a single main teacher for each academicyear. In contrast, similar research in adolescentsposes considerable methodological challengessince they usually have several teachers whomay only see them for relatively short periods oftime. From a methodological perspective, it hasalso been assumed that the rating by the currentteacher is the �gold standard�. However, the

parental report of school behaviour may reflectthe child’s history of schooling and the views ofa range of teachers. It may also be influenced bythe level of contact the parent has with theschool as well as by factors unrelated to school(such as parental mental health). Similarly, thereare also likely to be individual differences at theteacher level that influence their ratings and thatare affected by factors unrelated to the child.Finally, although the measures employed corre-late well with the diagnosis of ADHD andinclude a rating of impairment, they are notbased on operationalized diagnostic criteria (1).It is possible that findings may vary with the useof specific ADHD criteria-based rating scales.Further research is also required to identify the

reasons for disagreement between parent andteacher ratings. These discrepancies may reflect acombination of child, parent, teacher, school, andsocio-economic factors. Gaining a better under-standing of factors that determine differences andsimilarities between informant reports requires acomparison of parent and teachers ratings ofsymptoms and impairment, across the full rangeof scores. Such information will be useful clinicallyin interpreting informant reports. Currently, thereare practical challenges in determining how best tocombine data from different informants. Varia-tions in approaches lead to wide differences inprevalence rates of ADHD (24, 25). However,children’s pathways to effective treatments invol-ving decisions about referral to services, diagnosticformulations and choice of treatment may welldepend on the subjective clinical interpretations ofthese reports.This study’s findings support the necessity of

obtaining teacher ratings in practice. First, theseare crucial in improving diagnostic accuracy. Thepossible risks associated with the misdiagnosis ofand long-term treatment for ADHD mean thatteachers should be consulted at some pointduring the assessment process. Teacher ratingsalso have the advantage of identifying particularsymptoms that can be targeted in the manage-ment plan. Second, they play an important rolein the identification of children who have undi-agnosed ADHD. The omission of informationfrom teachers also considerably under-estimatesprevalence rates of ADHD (24). This providesfurther evidence that the limited concordancebetween parent and teacher reports is in thedirection of parents systematically under-report-ing the levels of problems at school. The omis-sion of teacher ratings could act as a furtherbarrier to the identification of children withpervasive hyperactivity or ADHD.

Sayal and Taylor

464

References

1. American Psychiatric Association. Diagnostic and Statis-tical Manual of Mental Disorders, 4th edn (DSM-IV).Washington, DC: APA, 1994.

2. The MTA Cooperative Group. A 14-month randomizedclinical trial of treatment strategies for attention deficit/hyperactivity disorder. Arch Gen Psychiatry 1999;56:1073–1086.

3. Klein RG, Mannuzza S. Long-term outcome of hyperactivechildren: a review. J Am Acad Child Adolesc Psychiatry1991;30:383–387.

4. Steinhausen H-C. Attention-deficit hyperactivity disorder ina life perspective. Acta Psychiatr Scand 2003;107:321–322.

5. Achenbach TM, Mcconaughy SH, Howell CT. Child/adol-escent behavioral and emotional problems: implications ofcross-informant correlations for situational specificity.Psychol Bull 1987;101:213–232.

6. Kumpulainen K, Rasanen E, Henttonen I et al. Children’sbehavioural/emotional problems: a comparison of parents�and teachers� reports for elementary school-aged children.Eur Child Adolesc Psychiatry 1999;8(suppl. 4):41–47.

7. Verhulst FC, Akkerhuis GW. Agreement between parents�and teachers� ratings of behavioral/emotional problems ofchildren aged 4–12. J Child Psychol Psychiatry 1989;30:123–136.

8. Mitsis EM, Mckay KE, Schulz KP, Newcorn JH, HalperinJM. Parent–teacher concordance for DSM-IV attention-deficit/hyperactivity disorder in a clinic-referred sample.J Am Acad Child Adolesc Psychiatry 2000;39:308–313.

9. Bussing R, Zima BT, Gary FA, Garven CW. Barriers todetection, help-seeking, and service use for children withADHD symptoms. J Behav Health Services Res2003;30:176–189.

10. Sayal K, Taylor E, Beecham J, Byrne P. Pathways to care inchildren at-risk of attention deficit/hyperactivity disorder.Br J Psychiatry 2002;181:43–48.

11. Sayal K, Taylor E, Beecham J. Parental Perception ofproblems and mental health service use for hyperactivity.J Am Acad Child Adolesc Psychiatry 2003;42:1410–1414.

12. Woodward L, Dowdney L, Taylor E. Child and familyfactors influencing the clinical referral of children withhyperactivity: a research note. J Child Psychol Psychiatry1997;38:479–485.

13. Taylor E, Chadwick O, Heptinstall E, Danckaerts M.

Hyperactivity and conduct problems as risk factors foradolescent development. J Am Acad Child Adolesc Psy-chiatry 1996;35:1213–1226.

14. Cantwell DP. Attention deficit disorder: a review of thepast 10 years. J Am Acad Child Adolesc Psychiatry 1996;35:978–987.

15. Sprafkin J, Gadow K, Nolan E. The utility of a DSM-IV-referenced screening instrument for attention-deficit/hyperactivity disorder. J Emotion Behav Disord 2001;9:182–191.

16. Verhulst FC, van der Ende J. Factors associated with childmental health service use in the community. J Am AcadChild Adolesc Psychiatry 1997;36:901–909.

17. Biederman J, Keenan K, Faraone SV. Parent-based diagnosisof attention deficit disorder predicts a diagnosis based onteacher report. J Am Acad Child Adolesc Psychiatry 1990;29:698–701.

18. Biederman J, Faraone SV, Milberger S, Doyle A. Diagnosesof attention-deficit hyperactivity disorder from parent re-ports predict diagnoses based on teacher reports. J AmAcad Child Adolesc Psychiatry 1993;32:315–317.

19. Zeiner P. Parent-reported symptoms of hyperactivity andattention deficits predict teacher-reported symptoms. ActaPaediatr 1997;86:178–182.

20. Goodman R. The strengths and difficulties questionnaire: aresearch note. J Child Psychol Psychiatry 1997;38:581–586.

21. Goodman R. The extended version of the Strengths andDifficulties Questionnaire: as a guide to child psychiatriccaseness and consequent burden. J Child Psychol Psychi-atry 1999;40:791–799.

22. Jarman B. Underprivileged areas: validation and distribu-tion of scores. BMJ 1984;289:1587–1592.

23. Rutter M, Tizard J, Whitmore K. Education, Health, andBehaviour. London: Longman, 1970.

24. Ford T, Goodman R, Meltzer H. The British Child andAdolescent Mental Health Survey 1999: the prevalence ofDSM-IV disorders. J Am Acad Child Adolesc Psychiatry2003;42:1203–1211.

25. Wolraich ML, Lambert EW, Bickman L, Simmons T,Doffing MA, Worley KA. Assessing the impact of parentand teacher agreement on diagnosing attention-deficithyperactivity disorder. J Dev Behav Pediatr 2004;25:41–47.

Parent ratings of ADHD

465