8/18/2019 Good Man 1997

1/7

J. Child Psychol. Psychiat. Vol, 38. No, 5, pp. 581-586,Cambridge University P

© 1997 Association for Child Psychology and PsycPrinted in Great Britain, All rights rese

0021-9630/97 $15.00 +

The Strengths and Difficulties Questionnaire: A Research Note

Robert GoodmanInstitute of Psychiatry, London, U.K.

A novel behavioural screening questionnaire, the Strengths and Difficulties Questionnaire(SDQ), was administered along with Rutter questionnaires to parents and teachers of 403children drawn from dental and psychiatric clinics. Scores derived from the SDQ and Rutterquestionnaires were highly correlated; parent-teacher correlations for the two sets ofmeasures were comparable or favoured the SDQ. The two sets of measures did not differ intheir ability to discriminate between psychiatric and dental clinic attenders. Thesepreliminary findings suggest that the SDQ functions as well as the Rutter questionnaireswhile offering the following ad dition al a dva nta ges: a focus on streng ths as well as difficulties;better coverage of inattention , peer relationships, and prosocial behaviour; a shorter format;and a single form suitable for both parents and teachers, perhaps thereby increasing

parent-teacher correlations.

Keywords: Questionnaire, child behaviour, psychopathology, strengths.

Abbreviations: CBCL: Child Behavior Checklist; ROC: Receiver Operating Characteristic;S D Q : Strengths and Difficulties Questionnaire.

In t roduct ion

This paper describes a brief behavioural screeningquestionnaire that provides balanced coverage of childrenand young people's behaviours, emotions, and relation-ships. The value of this novel Strengths and DifficultiesQuestionnaire (SD Q) is evaluated against the benchm arkset by the Rutter parent and teacher questionnaires. TheSDQ has been designed to meet the needs of researchers,clinicians, and educationalists.

The Rutter questionnaires are long-established andhighly respected behavioural screening questionnairesthat have proved valid and reliable in many contexts(Elander & Rutter, 1996). Though substantially shorterand therefore quicker to complete than the Child

Behavior Checklist (CBCL; Achenbach, 1991a), theRutter parent questionnaire seem no less useful for manypurposes (Berg, Lucas, & McGuire, 1992; Elander &Rutter, 1995; Fom bonn e, 1989). Developed three decadesago, the Rutter questionnaires have generally worn well,though they do show their age in some ways. Thus allitems are about undesirable traits whereas the recenttrend, particularly in education, has been to emphasisechildren's strengths and no t just their deficits. In addition,the range of behavioural items covered by the Rutterquestionnaires is now somewhat dated. Thus nail-bitingand thumb-sucking are included whereas many areasof contemporary interest—including concentration,

Requests for reprints or sample questi^pnnaires (available

impulsivity-reflectiveness, having friends, being timised, and acting prosocially—are poorly cov

Finally, whereas one version of the Achenbach qtionnaire is designed for completion by young pethemselves (Achenbach, 1991b), there is no equivRutter questionnaire for self-completion.

A previous research note (Goodman, 1994) descan expanded Rutter parent questionnaire that inporated all of the original Rutter items as well as additional items, mostly on children's strengths. inclusion of these additional items did not appeaattenuate the valuable properties of the original Rquestionnaire as a behaviour screening instrumthough the extra items presumably did make questionnaire somewhat more time-consuming to plete. Factor analyses suggested that among chiof normal intelligence the expanded questionnairetapping five distinct dimensions; conduct problemotional symptoms, hyperactivity, peer problemsprosocial behaviour.

Using these findings as a guide, the SDQ was desto meet the following specifications; it should fit easone side of pap er; it should be applicable to childrenyoung people ranging from 4 to 16 years; the version should be completed by parents and teachesimilar version should be available for self-report;strengths and difficulties should be well represented

there should be equal num bers of items on each orelevant dimensions, namely conduct problems, emoal symptoms, hyperactivity, peer relationships, and

i l b h i Thi i f

8/18/2019 Good Man 1997

2/7

S2 R. G OO D M AN

Materials and MethodsamplingQuestionnaires were obtained on 403 children aged 4-16

ears attending one of two London child psychiatric clinics orhe children's department of a London dental hospital. Thearents of children attending these clinics were recruited into

he study until a planned total of roughly 150-250 children hadeen attained for both dental and psychiatric samples. In theental clinic and one of the psychiatric clinics (Clinic A),aren ts who had given informed consent were asked to completewo behavioural screening questionnaires while awaiting theirlinic appointment. Participating parents were subsequentlysked for permission for their child's teacher to be approachedn a similar basis. The other psychiatric clinic (Clinic B)outinely used questionnaires prior to the first assessment,ending them to all paren ts and, when permission was obtaine d,o teachers as well. In this clinic, parents were routinely sentoth behavioural screening questionnaires and asked if they

would be willing for their answers (and the teacher's answers) toe used not only for clinical purposes but also for research.

Some of the parents from Clinic B did not completeuestionnaires themselves but did give permission for teacheruestionnaires to be used for research. The proportion ofefusals was not systematically recorded since, as explainedater, the statistical analyses did not require the samples to beepresentative.

MethodsRespondents were administered a Rutter questionnaire and a

Strengths and Difficulties Questionnaire (SDQ) in randomisedrder. Parents were given the Rutter A(2) Questionnaire andeachers the Rutte r B(2) Que stionna ire; both were scored in thetandard way to generate scores for total deviance, conduct

problems, emotional symptoms, and hyperactivity (Rutter,1967; Rutter, Tizard, & Whitmore, 1970; Schachar, Rutter, &Smith, 1981).

The informant-rated version ofthe SDQ was administered toboth parents and teachers. This version of the SDQ iseproduced in full in Appendix A for information only. The

SDQ asks about 25 attributes, 10 of which would generally behoug ht of as strengths, 14 of which would generally be thou ght

of as difficulties, and on e of wh ich— gets on better with ad ultshan with other children —is neutral. Though no SDQ item isdentically worded to any Rutter item, five items are similarly

worded. The initial choice of items was guided by the factor

oadings and frequency distributions that had previously beenobtained on an expanded Rutter parent questionnaire (Good-man, 1994); items were subsequently modified and amalga-mated on the basis of a succession of informal trials as well asadvice from colleagues. The 25 SDQ items are divided between5 scales of 5 items each, as shown below.

Hyperactivity Scale. Restless, overa ctive, can not stay stillor long ; Con stantly fidgeting or squirm ing ; Easily

distracted, concentration wanders ; '•'Thinks things out beforeacting '; an d Sees tasks through to the end, good attentionpan''.

Emotional Symptoms Scale. Ofte n com plains of head-aches, stomach -ache or sick nes s ; M an y worries, often seemswo rried ; Often unhapp y, down-hearted or tearful ;

Nervous or clingy in new situations, easily loses confidence ;and M an y fears, easily sca red .

Conduct Problems Scale. Often has temper tantrums or

Peer Problems Scale. Ra the r solitary, tends to pa lone ; 'Has at least one good friend ; ''Generally liked bother children ; Pick ed on or bullied by other ch ild ren ; Gets on better with adults than with other children .

Prosocial Scale. Considerate of other people's feelin Sh ares readily with other children (treats, toys, pencils, e Helpful if someone is hurt, upset or feeling ill ; Kiyounger childre n ; and Often volunteers to help

(parents, teachers, other children) .Each item can be marked no t tr ue , som ewha t tru certainly true . For all of the items except the five pabov e in italics, the item is scored 0 for n ot tr ue , somewhat tru e , and 2 for certain ly tru e . Fo r the fiveprinted abo ve in italics, the item is scored 2 for not true 1 f somewh at tru e , and 0 for certainly true . The scoeach of the five scales is generate d by sum ming t he scores five items that m ake u p that scale, thereby generating ascore rangin g from 0 to 10. The scores for hypera cemotional symptoms, conduct problems, and peer procan be summed to generate a total difficulties score rfrom 0 to 40; the prosocial score is not incorporated reverse direction in to the to tal difficulties score since the aof prosocial behaviours is conceptually different fropresence of psychological difficulties.

The Rutter A(2) and the SDQ were both completed parents of 346 children: 158 dental clinic attenders anpsychiatric clinic attenders. The R utter B(2) and the SDQboth completed by the teachers of 185 children: 39 dentaattenders and 146 psychiatric clinic attenders. Most teacher reports were on psychiatric clinic attende rs becauparents of children attending Child Psychiatric Cligenerally agreed to the clinic sending questionnaires to tfor clinical as well as research purposes; parents of chattending the dental clinic or Child Psychiatric Clinic Aless likely to give permission for teacher questionnaires

obtained solely for research purposes.

Statistical AnalysisAs in previous studies comparing the validity of di

screening questionnaires (e.g. Berg et al, 1992), analyReceiver Operating Characteristic (ROC) curves were uestablish how well each questionnaire was able to distbetween high- and low-risk samples, determining the areathe curve for each questionnaire (Hanley & M cNeil, 198this purpose, the only underlying assumption is that crecruited from the two psychiatric clinics were substamore likely to have psychiatric disorders than children refrom the dental clinic. There is no assumption that all srecruited from the psychiatric clinics had psychiatric dinor that all subjects recruited from the dental clinic wefrom psychiatric disorder. Equally, there is no assum ptithe psychiatric samp le was representative of all children b4 and 16 who attend psychiatric clinics, nor that thesample was representative of all children attending clinics, let alone of all children aged between 4 an d 16. SROC curves for the SDQ and Rutter questionnairederived from the same set of patients, the statistical comof the areas under these ROC curves allowed for thenatu re ofthe d ata (Hanley & McN eil, 1982). Com parisoparent-teacher correlations of the SDQ and questionnaires also allowed for the paired nature of thusing structural equation modelling (EQS, BMDP StaSoftware) and examining whether constraining the twolations to be the same resulted in a significantly poo

8/18/2019 Good Man 1997

3/7

S T R E N G T H S A N D D I F F I C U LT I E S Q U E S T I O N N A I R E 58

could securely be generalised to representative epidemiologicalor clinical samples; such estimates will subsequently be derivedfrom other studies in progress.

Reported correlations are Pearson product-moment corre-lations, but the pattern of findings was not changed whenSpearman correlations were used instead. Intraclasscorrelations—which are often appro priate for reliabilityestimates—were not used to measure parent-teacher agreementeven though this agreement could be construed as an index ofinter-rater reliability. Parents and teachers make ratings basedon different sources of information, whereas measures of inter-rater reliability are more appropriately derived from inde-pendent ratings based on the same source of information.Furthermore, employing intraclass correlations would haveinvolved mixing parent- and teacher-derived scores, and thiswould have been inappropriate since mean scores differedsystematically between parent and teacher ratings—a differenceallowed for when interpreting these scores (Rutter, 1967; Rutteret al., 1970; and see Appendix B).

ResultsAge and Gender

The mean age {SD) ofthe dental sample was 10.8 years(3.1) while that ofthe psychiatric sample was 9.8 years(3.3), a significant difference [t (401) = 3 .00, /? < .01]. Asexpected, the propor t ion of males was higher in thepsychiatric sample (63%, 153/244) than in the dentalsample (5 3% , 85/159) [cont inu i ty-adjus ted/ (1) = 3.03,p < .05, 1-tailed]. The results reported here are for th esample as a whole, tho ug h closely similar results wereobta ined when ROC and corre lational analyses w ererepeated separately for boys and girls, and separately forchildren aged 4-10 and 11-16.

Sensitivity

0 8

0 6

0.4

0.2

0 0 2 0 4 0 6 0 8

1 1

0.8

0.6

0.4

0 2

Sensitivity

°

I D

1

_:_^ : - . ; - —

ŜDQ

° Rutter

0 0 2 0 . 8.4 0.61-Specif ic i ty

Pigure 2. RO C curves for teacher-rated questionnair

Discriminating between Psychiatric andNonpsychiatric Samples

The ability of the two questionnaires to distinguibetween dental and psychiatric cases is reflected in thReceiver Operating Characteristic (ROC) curves sin Figs. 1 and 2 for parent and teacher reports respectively. The R OC curves for the Rutter questionare based on total de viance scores, whereas the curves are based on tota l difficulties score s. The cparable ability of the two measures to discriminbetween the two samples is evident from the extent towhich the two curves almost superimpose on one anQua ntitatively, this comparability can be judged froarea un der each o fth e curves, which is a measure owell that m easure discriminates between the two samthe area under the curve would be 1.0 for a measur

discriminated perfectly, and .5 for a measure that dcriminated with no better tha n chance accuracy. Forpare nt re ports, the area un der the curve (95 % confiinterval) was .87 (.83-.91) for the SDQ as compared.87 (.83-.91) for the R utter A(2) parent questionnainonsignificant difference (z = .13, = .9). Fo r teachreports, the corresponding values were .85 (.78-.93the SDQ as compared with .84 (.76-.93) for the RB(2) teacher question naire— a nonsignificant diffe(2 = .41 , ; ; = .7).

SDQ-Rutter CorrelationsTable 1 shows the correlations between SD Q and

R Th l i l

8/18/2019 Good Man 1997

4/7

4 R. GOODMAN

Table 1Inter-measure Correlation for Each Type of Rater

Total Deviance/Difficulties scoreConduct Problems scoreEmotional Symptoms scoreHyperactivity score

SDQ-Rut te r

Parent report

.88

.88

.78

.82

correlation

Teacher reportA^= 185

.92

.91

.87

.90

ble 2er-rater Correlations for Each Type of Measure

tal Deviance/Difficulties scoreonduct Problems scoremotional Symptoms scoreyperactivity scoreer Problems scoreosocial Behaviour score

Parent-Teachercorrelation(N =

SDQ.62*.65.41.54.59.37

= 128)

Rutter.52.57.47.55——

* Correlation significantly higher than the com parabletter correlation (p < .02); all othe r comp arisons non-nificant.

ions can be presented on two SDQ scores—the peeroblems score and the prosocial behaviou r score—sinceey have no Rutter counterpart.

rent-Teacher Correlations

Table 2 presents the correlation coefficients betweenacher- and parent-derived scores when both are usinge SDQ or when both are using Rutter questionnaires.r comparable scores, the cross-situation correlations

the SDQ and Rutter measures were generally similar,art from the higher SDQ correlation for total score ̂ (1) = 5.90, p < .02]. Thou gh the co rrelations were

wer when the analyses were repeated for the dental andychiatric samples separately, these correlations werenerally comparable for the SDQ and Rutter measures,art from a higher SDQ co rrelation for total score in theychiatric sample [x ̂ (1) = 4.05, p < .05).

Discussion

Given the well-established validity and reliability ofth etter questionna ires (Elander & Rutte r, 1996), the highrrelation between the total scores generated by the

used identical items for parents and teachers whereRutter questionnaires were somewhat differenparents and teachers. The ROC analyses showed thtwo measures had equivalent predictive validitjudged by their ability to distinguish between psycand nonpsychiatric samples. Of course, discrimibetween psychiatric and dental clinic attendersrelatively easy task, but the high correlation beSDQ and Rutter scores within each clinic group suthat the two measures are also likely to be compdiscriminating in more demanding screening tasksas detecting nonreferred cases of child mental problems in the community; further empirical swould be needed to confirm this. Since previous shave shown that CBCL and Rutter parent questioscores are highly correlated (Berg et al., 1992; Fom bon1989), and that these two sets of questionnaires acomparable predictive vahdity (Berg et al., 1992)likely that the SDQ and CBCL will also be hcorrelated and have comparable validity; directparisons are currently under way.

The SDQ and Rutter questionnaires can each beto generate separate scores for conduct probemotional symptoms, and hyperactivity. For eathese three scores, there was a high correlation bethe SDQ score and the Rutter score; and parent-tecorrelations were comparable for the two semeasures. Despite its brevity, the SDQ also gentwo scores that have no Rutter counterparts; aproblems score and a prosocial behaviour score.

The performance of the SDQ could potentiallybeen undermined by three of its design features; incof strengths as well as difficulties; use of an idequestionnaire for both parents and teachers; acompact presentation on just one side of paperequivalence of the SDQ and Rutter scores suggesthese three features have not had an adverse effectshould encourage researchers and clinicians whcontemplating incorporating similar features intoquestionnaires.

Rutter questionnaires are routinely used to catechildren as likely psychiatric ca se s or no n- caccording to whether their to tal deviance score is eqor greater than a standard cut-off; 13 on the Ruparent questionnaire and 9 on the Rutter teachertionnaire (Rutter, 1967; Rutter et al., 1970). Us

8/18/2019 Good Man 1997

5/7

STRENGTHS AND DIFEICULTIES QUESTIONNAIRE 8

ifferent studies simply because those studies havemployed the same cut-off. Comparability is particularlyikely to be lost when high- and low-risk samples areontrasted. A worked example may be helpful. Study Xnvolves 100 children from a high-risk population with arue rate of psychiatric disorder of 50 %; if the screening

questionnaire has a sensitivity of 8 and a specificity of .8when using the standard cut-off, the questionnaire willdentify 40 true positives and 10 false positives. Study Ynvolves 100 children from a low-risk population with arue rate of disorder of 10%; even with the sameensitivity and specificity, the questionnaire will identify 8rue positives and 18 false positives. Despite using the

same questionnaire and the same cut-off, a comparison ofcases from studies X and Y will primarily be a

comparison of true positives from study X with falsepositives from study Y.

Given these problems, the best strategy for researchersmay be to choose cut-offs according to the likely disorderrate in the sample being studied, and according to the

relative importance for that study of false positives andfalse negatives. It may also be appropriate to adjust cut-offs for age and gender. Ongoing clinical and epidemio-logical studies using the SDQ should provide the basis forcut-offs adjusted for these sample characteristics. Inaddition, planned trials should establish if the predictivevalidity of the SDQ can further be improved by analgorithm that combines SDQ scores with scores from anadditional and even briefer screening instrument thatelicits the respondent's view on whether the child hassignificant emotional or behavioural difficulties, and onthe extent to which these difficulties result in socialimpairment or distress for the child, or burden for others.Until these various studies are completed, SDQ users canuse the provisional cut-off scores shown in Appendix B,which are derived partly from the samples used for thisstudy and partly from other ongoing epidemiologicalsurveys using the SDQ . The bo rde rlin e cut-offs can beused for studies of high-risk samples where false positivesare not a major conc ern; the ab no rm al cut-offs can beused for studies of low-risk samples where it is moreimportant to reduce the rate of false positives.

ConclusionThese initial findings suggest that the SDQ may

function as well as the Rutter questionnaires (and, byinference, the Achenbach questionnaires) while offering

the following additional advantages: a compact foa focus on strengths as well as difficulties; better covof inattention, peer relationships, and prosocial biour ; and a single form suitable for both parents teachers, perhaps thereby increasing parent-tecorrelations.

Acknowledgements —I am very grateful for the willin

operation of parents and teachers, and for the invaassistance of E)r. Hilary Richards and the staff of theLondon clinics that participated in the study: the Depaof Paediatric Dentistry of King's Dental Institute, CambChild Guidance Centre, and the Department of ChilAdolescent Psychiatry, Hounslow.

References

Achen bach, T. M. (1991a). M anual for the Child BehavChecklist 14-18 and 99 Profile. Burlington, VT : U niveof Vermont Department of Psychiatry.

Achenbach, T. M. (1991b). Manua l for the Youth Self-Reand 1991 Profile. Burlington, vf : University of Ver

Department of Psychiatry.Berg, I., Lucas, C , & McGu ire, R. (1992). Measurembehaviour difficulties in children using standard administered to mothers by computer: Reliability aidity. European Child and Adolescent Psychiatry, 1, 14-2

Elande r, J., & Rutter, M. (1996). Use and developmenRutter Parents' and Teachers' Scales. International JournalMethods in Psychiatric Research, 6, 63-78.

Fombonne, E. (1989). The Child Behavior Checklist Rutter Parental Questionnaire: A comparison betwescreening instruments. Psychological Medicine, 19, 111-

Goodman, R. (1994). A modified version ofthe Rutterquestionnaire including items on children's strengresearch note. Journal of Child Psychology and Psychi35, 1483-1494.

Hanley, J. A., & M cNeil, B. J. (1982). The meaning anthe area under a receiver operating characteristic curve. Radiology, 143, 29-36.

Hanley, J. A., & McN eil, B. J. (1983). A method of comthe areas under receiver operating characteristic derived from the same cases. Radiology, 148, 839-843.

Rutter, M. (1967). A children's behaviour questionncompletion by teachers: Preliminary findings. JournalChild Psychology and Psvehiatry, 8, 1-11.

Rutter, M., Tizard. J., & Whitmore, K. (1970). Educathealth and behaviour. London: Longman.

Schachar, R., Rutter, M., & Smith, A. (1981). The

teristics of situationally and pervasively hyperactive cImplications for syndrome definition. Journal of Child chology and Psychiatry, 22, 375-392.

Accepted manuscript received 26 Septembe

8/18/2019 Good Man 1997

6/7

86 R. GOODMAN

Appendix A: Strengths and Difficulties Questionnaire

r each item, please mark the box for Not True, Somewhat True or Certainly True. It would help us if you answered all items asst you ean even if you are not absolutely certain or the item seems daftl Please give your answers on the basis of the child'shaviour over the last six months or this school year.

hilds Name

te of Birth.

gnature

Considerate of other people's feelingsRestless, overactive. cannot stay still for longOften complains of headaches, stomach-aches or sicknessShares readily with other children (treats, toys, pencils etc)Often has temper tantrums or hot tempersRather solitary, tends to play aloneGenerally obedient, usually does what adults requestMany worries, often seems worriedHelpful if someone is hurt, upset or feeling illConstantly fidgeting or squirmingHas at least one good friendOften fights with other children or bullies themOften unhappy, down-hearted or tearfulGenerally liked by other childrenEasily distracted, concentration wandersNervous or clingy in new situations, easily loses confidenceKind to younger childrenOften lies or cheatsPicked on or bullied by other childrenOften volunteers to help others (parents, teachers, other children)Thinks things out before actingSteals from home, school or elsewhereGets on better with adults than with other childrenMany fears, easily scaredSees tasks through to the end. good attention span

Date

arent/Teacher/Other (please specify:)Thank you very much or your help

N otTrue

n

n

n

nn

n

SomewhatTrue

n

n

Male/Fem

CertainlyTrue

n

n

nn

Robert Goodman, 1977

Appendix B: Provisional Banding of SDQ Scores

hese bands, which are not adjusted for age or gender, have been chosen so that roughly 80 of children in the community arermal, 10 are borderline, and 10 are abnormal.

Parent completedTotal Difficulties ScoreEmotional Symptoms ScoreConduct Problems ScoreHyperactivity ScorePeer Problems ScoreProsocial Behaviour Score

Teacher completedTotal Difficulties ScoreEmotional Symptoms ScoreC d P bl S

Normal Borderline Abnormal

0-130-30-20-50-2

6-10

0-110-4

14^1643635

12-155

17-405-104^107-104-100 - 4

l^-*06-10

8/18/2019 Good Man 1997

7/7