Social Psychiatry and Psychiatric Epidemiology

, Volume 45, Issue 12, pp 1179–1186 | Cite as

Validation of the Chinese version of the strengths and difficulties questionnaire in Hong Kong

  • Kelly Y. C. Lai
  • Ernest S. L. Luk
  • Patrick W. L. Leung
  • Ann S. Y. Wong
  • Lawrence Law
  • Karen Ho
Original Paper

Abstract

Background

The strengths and difficulties questionnaire (SDQ) is now one of the most commonly used instruments for screening child psychiatric morbidities. Psychometric studies in the West affirm its reliability and validity, but similar studies are scarce among non-Western populations. This is an important gap because cultural differences can influence how children’s behaviours are perceived and rated. This study explores the psychometric properties of the Chinese version of the SDQ among children in Hong Kong.

Method

The SDQ was translated into Chinese. A community sample of 3,722 students between 6 and 12 years were recruited by stratified random sampling from across the whole of Hong Kong. Comparison group consisted of 494 consecutive children attending a general child psychiatric clinic. SDQ and basic socio-demographic data were collected from parents and teachers. Reliability was determined by internal consistency and test–retest stability. Validity was assessed by the questionnaire’s ability to discriminate between community and clinic samples, and ROC curves. Cutoff scores and their sensitivity, specificity, positive predictive value and negative predictive value were calculated.

Results

Our results confirm the questionnaire’s reliability and validity. The total difficulties scale and hyperactivity subscale are potentially the most useful in discriminating between community and clinic children. The emotional subscale was relatively weaker, especially with respect to teachers’ ratings. Of note also is that our normative scores are significantly higher than those reported in the West, highlighting once again the importance of examining a questionnaire’s cultural applicability.

Conclusion

Our data support the use of the Chinese version of the SDQ, especially the total difficulties scale, as a screening instrument for psychiatric morbidities among children in Hong Kong.

Keywords

Questionnaire Reliability Validity Screening Chinese 

Introduction

The strengths and difficulties questionnaire (SDQ) [1] is one of the most widely used screening instruments for the detection of mental health problems among children and adolescents. It has been translated into many languages, and validation studies (mostly from Western countries) have affirmed its reliability and validity [2]. Consisting of 25 items, it is user-friendly and can readily be incorporated into clinical practice and epidemiological research. The availability of parallel parent-, teacher- and self-report versions allow information to be collected from different informants. An extended version with impact questions gives useful information on impairment and carers’ burden, and has been found to be highly discriminatory [3].

The psychometric properties of the questionnaire among non-Western populations have not been well studied [4]. This is an important gap, because expectations and evaluations of children’s behaviours are culturally sensitive. An instrument’s applicability cannot be assumed when it is used in a culture that is very different from where it originated. To the best of the authors’ knowledge, where Chinese culture is concerned, only one psychometric study on the SDQ has been reported. The study was carried out in Shanghai, China [5]. The results were mixed; neither the reliability nor validity was as sound as expected. The Shanghai study was carried out among Chinese parents and teachers in mainland China, whose attitudes towards children and child rearing practices may still be quite different from those in Hong Kong. The aim of this study is therefore to explore the applicability of the parent and teacher versions of the Chinese SDQ among children in Hong Kong and to further explore its psychometric properties.

Method

Instrument

The extended version of the parent- and teacher-SDQ was translated into Chinese by a clinician and back translated into English by an independent bi-lingual university graduate. A panel of experienced child and adolescent psychiatrists compared the back-translation with the original English version. Discrepancies were discussed. When necessary, the Chinese translation was revised and back translated again. The final translated version was approved by consensus. The questionnaire was scored according to standard instructions [1].

Community sample

The community sample consisted of primary school pupils from government-funded primary schools across the whole of Hong Kong. The study was approved by the Hong Kong Government’s Education Bureau, which was instrumental in recruiting the community sample. Data was collected over a 3-month period in the second term of the school year (April–July 2006) to ensure that teachers had had sufficient time to observe the children’s behaviours.

Random sampling method stratified to balance the proportion of schools with different achievement bandings and different socio-economic regions was used. Private schools (which are not banded), international schools (whose students are mostly from non-Chinese families), and schools for handicapped children were not sampled. Out of a total of 675 primary schools for non-handicapped children in Hong Kong, 44 (6.5%) were private schools and 48 (7.1%) were international schools. From the remaining 583 schools, 130 were invited. Seventy-six (58.4%) agreed to participate. A second round of invitation was sent to 52 schools, of which 20 agreed. In total, 96 schools participated.

All primary classes (primary 1–6, equivalent to 6–12 years olds) were involved. In order not to overload the teachers, only two students per class were randomly selected to take part. A senior school personnel was briefed about the study and was responsible for coordinating the logistics. Parents’ written consent was sought and those who agreed to take part were asked to complete the Chinese version of the parent-SDQ and return it to the school in a sealed envelope to ensure confidentiality. If the parents of the selected child refused to participate, the teachers were instructed to select the student with the next class number. The class teachers of participating students were required to complete the Chinese version of the teacher-SDQ. In addition to SDQ data, basic demographic information was obtained from the parents, which included the age and sex of the child, marital status of the parents, and parents’ level of education. These were kept to a minimum to facilitate cooperation from parents.

A total of 3,722 questionnaires were returned. Feedback from the schools’ coordinators indicated only very few parents refused to participate. Of the 96 schools, 19 volunteered to participate in the test–retest exercise 2–4 weeks later.

Clinic sample

The clinic sample was derived from consecutive children between the ages of 6 and 12 years attending the child and adolescent psychiatry clinic of a university-affiliated district general hospital for the first time between April 2006 and March 2007. As a routine part of the assessment, parents were asked to complete the Chinese version of the parent-SDQ prior to being seen by a clinician. At the end of the assessment session, with parents’ consent, a copy of the Chinese version of the teacher-SDQ was given to the child’s class teacher, who would complete the questionnaire and return it to the clinic in a sealed addressed envelope.

Clinicians were also required to complete a clinical data form to indicate the presenting problems. Pertinent to the current study are the internalising problems (which include anxiety and mood problems), externalising problems (which includes behavioural and conduct problems), and hyperactivity and attention problems.

Data analysis

SPSS version 14.0 was used to perform the statistical analyses. Reliability was determined by internal consistency and test–retest stability. Validity was assessed by its ability to discriminate between community and clinic samples, and ROC curves. Cutoff scores and their sensitivity, specificity, positive predictive value and negative predictive value were calculated.

Results

Of the 3,722 parent questionnaire returned by the community sample, 88 (2.4%) were discarded because of incomplete data (missing one or more of the 25 items), leaving 3,634 for analysis. Of the 3,722 teacher questionnaire returned, 75 (2.0%) were discarded for similar reasons, leaving 3,647 for analysis. From the clinic sample, 494 parent questionnaires were collected, of which 44 were incomplete, leaving 450 for analysis. These same children yielded 262 teacher questionnaires. The mean age of the community sample was 9.1 years (SD 1.8 years), while that of the clinic sample was 8.4 years (SD 1.7 years). Boys comprised 51% of the community sample and 81% of the clinic sample.

Parents and teachers of 867 students from 19 schools completed the questionnaires for a second time. However, 369 of the parent questionnaire had to be excluded because either (a) the parents completed the questionnaire less than 1 week or more than 1 month after the initial completion, or (b) different parents completed the questionnaire on the two occasions. That left 471 cases for test–retest analysis. Similarly 135 of the teacher’s questionnaire were excluded, leaving 732 teacher-SDQ for analysis. The mean duration between the two completion dates was 19 days (SD 7.3).

Scale means, gender and age effects

The mean score of the subscales and total difficulties scale of the community sample were examined according to gender and age. A significant gender effect was found. On both parent- and teacher-SDQs, boys scored significantly higher than girls on all the problem subscales except emotional problems (where girls scored higher). On the prosocial subscale, boys scored lower than girls. However, effect sizes were small (Cohen’s d between 0.06 and 0.18). No significant age effect was present. The interaction effect between gender and age was insignificant (Tables 1, 2).
Table 1

Comparison of parent-SDQ mean scores from different countries

 

This study

China

Japan

Australia

US

UK

Age (years)

6–12

3–10

4–12

7–9

8–10

5–15

Total sample size

3,722

1,217

2,899

557

2,064

10,298

Male (%)

51

49

51

54

50

50

Total difficulties

 Boys

12.3 (5.3)

11.3 (4. 9) [0.19]***

8.7 (5.0) [0.7]***

9.6 (6.3) [0.51]***

7.9 (6.4) [0.77]***

9.1 (6.0) [0.55]***

 Girls

11.2 (5.4)

10.5 (4.8) [0.14]***

7.9 (4.9) [0.65]***

8.2 (5.6) [0.57]***

6.4 (5.1) [0.91]***

7.8 (5.5) [0.62]***

Emotional

 Boys

2.3 (1.9)

1.9 (1.8) [0.21]***

1.6 (1.7) [0.39]***

2.2 (2.2) [0.05]**

1.5 (1.9) [0.42]***

1.8 (2.0) [0.25]***

 Girls

2.6 (2.1)

2.3 (1.9) [0.15]***

1.9 (1.9) [0.35]***

2.3 (2.2) [0.14]***

1.5 (1.8) [0.55]***

2.0 (2.0) [0.30]***

Conduct

 Boys

2.3 (1.7)

1.8 (1.5) [0.30]***

2.1 (1.6) [0.12]***

2.0 (2.0) [0.17]***

1.5 (1.9) [0.45]***

1.7 (1.8) [0.34]***

 Girls

2.0 (1.6)

1.4 (1.3) [0.39]***

1.9 (1.5) NS

1.5 (1.6) [0.31]***

1.1 (1.4) [0.59]***

1.5 (1.6) [0.31]***

Hyperactivity

 Boys

4.9 (2.3)

4.9 (2.5) NS

3.6 (2.4) [0.55]***

3.9 (2.7) [0.42]***

3.3 (2.9) [0.63]***

4.0 (2.7) [0.35]***

 Girls

4.2 (2.2)

4.1 (2.4) NS

2.8 (2.1) [0.65]***

3.1 (2.4) [0.49]***

2.4 (2.3) [0.81]***

2.9 (2.4) [0.55]***

Peer problem

 Boys

2.7 (1.7)

2.7 (1.7) NS

1.5 (1.6) [0.72]

1.5 (1.7) [0.71]***

1.5 (1.6) [0.72]***

1.5 (1.7) [0.71]***

 Girls

2.5 (1.7)

2.7 (1.7) [0.12]***

1.4 (1.4) [0.70]***

1.4 (1.6) [0.65]***

1.4 (1.5) [0.67]***

1.4 (1.6) [0.68]***

Prosocial

 Boys

6.4 (1.9)

6.8 (1.9) [0.23]***

6.3 (2.1) NS

7.8 (1.9) [0.72]***

8.6 (1.8) [1.18]***

8.4 (1.7) [1.14]***

 Girls

7.0 (2.0)

7.5 (1.8) [0.24]***

7.0 (2.0) NS

8.4 (1.6) [0.70]***

9.0 (1.5) [1.09]***

8.9 (1.4) [1.20]***

Values are expressed as mean scores (SD) [effect sizes]

*** Significantly different from current study at p = 0.000; ** p < 0.005

Table 2

Comparison of teacher-SDQ mean scores from different countries

 

This study

China

Australia

UK

Age (years)

6–12

3–10

7–9

5–15

Total sample size

3,722

1,217

557

10,298

Male (%)

51

49

54

50

Total difficulties

 Boys

10.2 (5.7)

10.6 (6.0) [0.07]**

7.8 (6.9) [0.41]***

7.8 (6.3) [0.39]***

 Girls

7.7 (5.1)

8.0 (4.9) [0.05]*

5.7 (6.0) [0.39]***

5.3 (5.3) [0.46]***

Emotional

 Boys

1.8 (1.7)

1.8 (1.8) NS

1.5 (1.9) [0.17]***

1.4 (1.9) [0.22]***

 Girls

2.0 (1.9)

1.8 (1.8) [0.11]***

1.5 (1.8) [0.26]***

1.4 (1.9) [0.32]***

Conduct

 Boys

1.7 (1.8)

1.7 (1.8) NS

1.3 (1.7) [0.22]***

1.2 (1.8) [0.28]***

 Girls

1.0 (1.4)

1.0 (1.3) NS

0.9 (1.5) [0.07]***

0.6 (1.3) [0.30]***

Hyperactivity

     

 Boys

4.2 (2.6)

4.7 (2.8) [0.19]***

3.5 (3.0) [0.26]***

3.7 (3.0) [0.17]***

 Girls

2.6 (2.2)

3.2 (2.4) [0.27]***

1.8 (2.3) [0.36]***

2.1 (2.3) [0.22]***

Peer problem

 Boys

2.5 (1.8)

2.5 (1.8) NS

1.5 (1.9) [0.55]***

1.5 (1.9) [0.53]***

 Girls

2.1 (1.7)

2.0 (1.6) [0.02]*

1.5 (2.0) [0.35]***

1.2 (1.6) [0.55]***

Prosocial

 Boys

5.8 (2.3)

6.3 (2.5) [0.23]***

7.3 (2.3) [0.65]***

6.6 (2.5) [0.33]***

 Girls

7.1 (2.2)

7.3 (2.3) [0.07]**

8.4 (1.8) [0.60]***

7.9 (2.1) [0.38]***

Values are expressed as mean scores (SD) [effect sizes]

Significantly different from current study at * p < 0.05; ** p < 0.005; *** p < 0.001

We also compared the mean scores from our study with that from China [5], Japan [6], Australia [7], US [8] and the UK [9] and found significantly higher problem scores, and lower prosocial scores than those from Western countries, with moderate to large effect sizes. Comparing within Asian cultures, our problem scores were still significantly higher than those from Japan, but similar to those from China, while the prosocial scores across the three places were similar.

Cross-scale correlations

All subscales correlated significantly with each other at the p < 0.01 level (Pearson’s correlation). The directions of the correlations were as expected, with the prosocial subscale negatively correlated with all the problem subscales. On both the parent- and teacher-SDQs, highest correlations were found between hyperactivity and conduct problems (r = 0.55 on parent-SDQ, r = 0.67 on teacher-SDQ). Lowest correlations were found between emotional and hyperactivity problems on the parent-SDQ (r = 0.29) and emotional and conduct problems on the teacher-SDQ (r = 0.12).

Inter-informant correlations

Pearson’s correlations between parent- and teacher-SDQ scores were, as expected, moderate and ranged from r = 0.17 (emotional problem) to r = 0.47 (hyperactivity problem), while that of the total difficulties scale was 0.41.

Reliability

Internal consistency

Table 3a, b lists the internal consistency (Cronbach alpha) of the different subscales and the total difficulties scale. A reliability coefficient of 0.7 or higher is considered good in most social science research. Our results found the total difficulties scale of both parent- and teacher-SDQs to be the most homogenous, with alpha values greater than 0.8. Of the subscales, hyperactivity problems and prosocial behaviours on both parent- and teacher-SDQs were also highly consistent with alphas of greater than 0.7. The least internally consistent was peer problems which yielded low alphas on both parent- (alpha = 0.45) and teacher-SDQ (alpha = 0.55).
Table 3

(a) Internal consistency and test–retest reliability of parent-SDQ; (b) internal consistency and test–retest reliability of teacher-SDQ

 

Cronbach’s α

ICC

Pairwise t test

This study

UK [10]

Time 1 [mean (SD)]

Time 2 [mean (SD)]

Sig [effect size]a

(a)

 Emotional

0.66

0.67

0.82

2.7 (2.2)

2.4 (1.9)

p < 0.01 [0.14]

 Conduct

0.62

0.63

0.75

2.2 (1.6)

1.9 (1.5)

p < 0.01 [0.18]

 Hyperactivity

0.76

0.77

0.86

4.6 (2.2)

4.4 (2.2)

NS

 Peer problem

0.45

0.57

0.75

2.6 (1.8)

2.5 (1.7)

NS

 Prosocial

0.70

0.65

0.78

6.7 (1.9)

6.7 (2.0)

NS

 Total difficulties

0.81

0.82

0.86

12.1 (5.6)

11.3 (5.3)

p < 0.01 [0.14]

(b)

 Emotional

0.68

0.78

0.80

1.9 (1.8)

1.9 (1.8)

NS

 Conduct

0.71

0.74

0.86

1.3 (1.7)

1.3 (1.7)

NS

 Hyperactivity

0.84

0.88

0.89

3.5 (2.5)

3.4 (2.5)

NS

 Peer problem

0.55

0.70

0.81

2.3 (1.8)

2.4 (1.8)

p < 0.05 [−0.08]

 Prosocial

0.82

0.84

0.84

6.4 (2.4)

6.4 (2.4)

NS

 Total score

0.84

0.87

0.90

9.0 (5.6)

8.9 (5.8)

NS

aCohen’s d

Test–retest reliability

Intra-class correlation (ICC) of all the subscales and the total difficulties scale of both the parent-SDQ and teacher-SDQ were excellent, with a mean value of 0.8 for parents and 0.85 for teachers (Table 3a, b). To examine whether attenuation and augmentation occurred across time, repeated measures t tests were performed, which found a tendency for some of the scores on the parent-SDQ to attenuate in the second administration, although the effect sizes of these changes were very small (Cohen’s d < 0.2).

Validity

Comparison of mean scores

Table 4a, b summarises the differences in the mean scores between the community and clinic samples. For the total difficulties score, peer problem subscale and prosocial subscale, comparison was made between the community and the whole-clinic sample. For the emotional, conduct and hyperactivity subscales, comparisons were made between the community sample and the corresponding clinic subgroups. All the score differences were statistically significant and in the expected direction (i.e. worse among clinic children), and the effect sizes were generally large (Cohen’s d 0.72–2.06).
Table 4

(a) Comparison of parent-SDQ scores between community and clinic samples; (b) comparison of teacher-SDQ scores between community and clinic samples

 

Mean scores (SD)

AUC

Community

Clinic sample

Sig.

Effect size

(a)

 Emotionala

2.5 (2.0)

5.1 (2.6)

***

1.29

0.79***

 Conducta

2.1 (1.6)

5.4 (2.1)

***

2.06

0.89***

 Hyperactivitya

4.5 (2.3)

7.8 (1.9)

***

1.44

0.86***

 Peer problemb

2.6 (1.7)

4.2 (2.0)

***

0.91

0.71***

 Prosocial behaviourb

6.7 (2.0)

6.0 (2.0)

***

0.35

0.60***

 Total difficultiesb

11.7 (5.4)

19.2 (5.3)

***

1.38

0.84***

(b)

 Emotionala

1.9 (1.8)

3.2 (1.9)

***

0.72

0.70***

 Conducta

1.4 (1.7)

4.7 (2.6)

***

1.93

0.86***

 Hyperactivitya

3.5 (2.5)

7.2 (2.4)

***

1.48

0.85***

 Peer problemb

2.3 (1.7)

3.6 (2.0)

***

0.76

0.69***

 Prosocial behaviourb

6.4 (2.3)

4.7 (2.6)

***

0.75

0.69***

 Total difficultiesb

9.0 (5.5)

15.6 (6.2)

***

1.19

0.78***

*** p < 0.001

aCommunity sample versus clinical subgroups

bCommunity sample vs clinic sample

Discriminative validity

ROC (Receiver Operating Characteristics) analysis was used to examine the discriminative power of the subscales and Total Difficulties scale. An AUC of 0.8 and above indicated good discriminatory potential, while that below 0.7 would be considered unsatisfactory. When comparing the clinic sample as a whole against the community sample, the Total Difficulties scale yielded an AUC of 0.84 on the parent questionnaire and 0.78 on the teacher questionnaire. To examine the AUCs of the subscales, clinic children were selected according to their presenting problems and the corresponding SDQ subscale scores were compared against the community sample scores. On both the parent and teacher questionnaires, the hyperactivity subscale and the conduct subscales yielded AUCs of at least 0.85, but that of the emotional subscale was lower at 0.79 and 0.70 respectively (Table 4a, b).

Cutoff scores

The Total Difficulties scores from the community sample were used to calculate the cutoff scores. Following Goodman’s suggestion, where scores in the top 10% were placed in the “high difficulties” band [1], our 90th percentile scores were 19 on the parent questionnaire and 17 on the teacher questionnaire. However, because Goodman’s selection of 90th percentile was guided by an estimated disorders rate in the community to be 10% [1], while that in Hong Kong has been suggested to be in the range of 15% [10], a second set of cutoff scores were calculated by labelling the top 15% scores as high risk cases. This second set of cutoff scores were 17 on the parent questionnaire and 15 on the teacher questionnaire. There was no additional benefit by adding the impact scores to the Total Difficulties score. The sensitivities and specificities of these two cutoff scores are included in Table 5.
Table 5

Cutoff scores for total difficulties scale

 

Present study

UK sample [10]

Chinese CBCL/TRF

Total problems [13]

90th percentile

85th percentile

Boys and girls

Boys and girls

Boys and girls

Boys and girls

Parent-SDQ

>19

>17

>17

Clinical cutoff (T 64)

Specificity (%)

89

83

94

91

Sensitivity (%)

53

64

47

52

PPV (%)

47

40

46

NPV (%)

92

93

96

Teacher-SDQ

>17

>15

>16

Clinical cutoff (T 64)

Specificity (%)

89

83

95

92

Sensitivity (%)

47

58

43

64

PPV (%)

43

38

44

NPV (%)

91

92

94

In order not to bias the calculation of positive predictive values (PPVs) and negative predictive values (NPVs) of these cutoff scores, a re-constituted sample was created to resemble the best estimate of psychiatric morbidity of children in Hong Kong. Based on the prevalence figures of 15% from a study by Leung et al. [10], this re-constituted sample consisted of 238 clinic children and 1,338 community children so that the two groups were in the ratio of 15%:85%, making a total of 1,576 children. As can be seen from Table 5, the PPVs for the two sets of cutoffs were around 40% and NPVs over 90%.

Discussion

The parent and teacher versions of the SDQ have been translated into Chinese and psychometric properties examined in a large representative community sample of 6–12 years old Chinese children in Hong Kong. Our results, by and large, confirmed its reliability and validity. Reliability data found the internal consistencies to be very similar to those reported from the UK [11], US [12] and Australia [7]. All found good to excellent homogeneity on the total difficulties scale and most of the subscales, with that of the peer problem subscale being the weakest. Stability of the questionnaire was confirmed by the excellent test–retest reliability over a 4-week period, which compares well with the Chinese child behaviour checklist (CBCL) and teacher report form (TRF) [13]. It is also superior to the figures reported from the UK (mean ICC of 0.62 on the parent questionnaire and 0.74 on the teacher questionnaire) [11] and Australia (mean ICC of 0.69 on the parent questionnaire) [7], although this is only to be expected, as the duration between tests in these two studies was much longer (4–6 months in the UK study and 12 months in the Australian study). The parent–teacher correlations of three of the subscales, namely, conduct symptoms, hyperactivity symptoms and total difficulties scale, were above the meta-analytic mean of 0.28 as reported by Achenbach [14]. However, these were relatively lower on the emotional and prosocial subscales, suggesting that teachers’ and parents’ observations may be quite different in these two areas.

Our data also supported the validity of the questionnaire. The total difficulties scale, the hyperactivity subscale and conduct problem subscale were all highly discriminatory from both parents’ and teachers’ ratings. The emotional subscale, however, was less satisfactory, and particularly so from the teacher’s questionnaires. This profile is similar to that of the Chinese CBCL [13], where the discriminatory ability of the emotional/internalising subscales is weaker than the behavioural/externalising subscales. This is particularly evident from the teachers’ ratings. All these suggest that parents can be reliable reporters of children’s psychopathology, but teachers are more likely to identify behavioural problems than emotional problem. Given that the average class size in Hong Kong is between 30 and 40 students, and different teachers are responsible for different subject areas, it is perhaps not surprising that teachers are less attuned to the more subtle emotional issues of their students.

While the clinical utility of the problem scales is well supported by many studies, use of the prosocial subscale has received relatively less attention. In our study, the AUCs of the prosocial subscale on both the parent and teacher questionnaires were poor, especially the parent questionnaire. This is in keeping with data from the UK [11] and Australia [7], where the odds ratio of having a psychiatric disorder when the prosocial score was low (i.e. less prosocial) is the lowest compared to other problem scales. In a preliminary study from Bangladesh [15], the prosocial scale was also not able to distinguish between community and clinic samples. This weak association between prosocial behaviours and psychopathology has also been shown with the Chinese CBCL and TRF, where the competence scales were the least predictive of clinical diagnoses [13]. It is attractive to incorporate questions that tap into the positive aspects of children’s behaviours, but how the data may be useful for screening or epidemiological purposes will require further examination.

While the psychometric properties are comparable, of note is that our normative scores were higher than many Western reports. This was true with both parents and teachers as raters. Possibilities for such differences include age factors: many studies have documented differences in normative scores for different age groups and genders. Other possibilities include cultural nuances which affect how raters perceive the behaviours described, and how they give their ratings. Of course, real differences in children’s behaviours and/or a much more stringent expectation of children’s behaviours by parents and teachers also need to be considered [16, 17]. Hong Kong is a very competitive society where children are expected to achieve academically, socially and in extra-curricular activities. There is a tremendous pressure to perform and conform, and deviations are frowned upon. Parents and teachers are likely to be much less tolerant of disturbing behaviours. It is interesting that in the study by Du et al. [5], where SDQ data from parents and teachers were collected for 1,965 subjects between 3 and 17 years old in Shanghai, another competitive city rooted in Chinese culture, similar elevations in normative scores were documented. Two other behavioural questionnaire studies carried out on Hong Kong children, one using the Rutter’s questionnaire and the other Conners’ teachers rating scale, also found much higher normative scores compared to that of the West [18, 19]. All these highlight how culture may affect the way a questionnaire is perceived and rated. Nevertheless, without the benefit of a second stage interview with parents of the community sample, it is not possible to make further inferences at this stage.

Concerning cutoff scores, we found that by lowering to 85th percentile (which corresponds to the disorders rate of 15%), there was an expected lowering of specificity and a rise in sensitivity. If the summing of specificity and sensitivity are to be used as indication of a more optimum cutoff [13], then the 85th percentile is to be preferred, especially if the questionnaire was to be used for screening purposes. In comparing the 85th percentile cutoff with the 90th percentile, there was little change in the NPVs, which remain at 91–93%. Overall, the comparability of our data is encouraging, and as SDQ is a much shorter questionnaire than the CBCL and the TRF, it is likely to be more versatile and welcomed.

There are several obvious limitations in this study. First of all, because we have confined ourselves to children between 6 and 12 years, the psychometric properties of the SDQ pertaining to younger children and teenagers are not known. Our results cannot be generalized to these other age groups because there may be age effects on SDQ scores, as suggested by Western studies. We also did not study the youth self-report, so its applicability is unknown. Secondly, the samples used in our study did not allow us to study the validity of the impact questions, because all the children attending our clinic, almost by default, have significant impact on their daily functioning, thus making the impact questions redundant. In order to further explore their usefulness, a two-stage design employing community samples would be more informative. For the time being, we can only use these questions to gain a general impression of severity.

In conclusion, with parents and teachers as raters, and primary school children as targets, the psychometric properties of the Chinese version of the SDQ are favourable. It will be a useful adjunct in the screening for possible mental health problems in children for epidemiological purposes as well as clinical studies. Like any screening questionnaires, it is essential that cutoff scores are chosen carefully depending on the purpose of the exercise. Moreover, if the questionnaires were to be used to monitor treatment progress, clinicians have to be aware of the attenuation effect over a short re-test period, which can confound study results. This study also highlights the importance of establishing local norms, as socio-cultural factors can substantially influence the applicability of screening questionnaires.

Notes

Acknowledgments

The authors are grateful for the help of the Education Bureau of Hong Kong in the recruitment of subjects.

References

  1. 1.
    Goodman R (1997) The strengths and difficulties questionnaire: a research note. J Child Psychol Psychiatry 38:581–586CrossRefPubMedGoogle Scholar
  2. 2.
    Vostanis P (2006) Strengths and difficulties questionnaire: research and clinical applications. Curr Opin Psychiatry 19(4):367–372CrossRefPubMedGoogle Scholar
  3. 3.
    Goodman R (1999) The extended version of the strengths and difficulties questionnaire as a guide to child psychiatric caseness and consequent burden. J Child Psychol Psychiatry 40:791–799CrossRefPubMedGoogle Scholar
  4. 4.
    Woerner W, Fleitlich-Bilyk B, Martinussen R, Fletcher J, Cucchiaro G, Dalgalarrondo P, Lui M, Tannock R (2004) The strengths and difficulties questionnaire overseas: evaluations and applications of the SDQ beyond Europe. Eur Child Adolesc Psychiatry 13(Suppl 2):II/47–II/54Google Scholar
  5. 5.
    Du Y, Kou J, Coghill D (2008) The validity, reliability and normative scores of the parent, teacher and self report versions of the strengths and difficulties questionnaire in China. Child Adolesc Psychiatry Mental Health 2:8. doi:10.1186/1753-2000-2-8 CrossRefGoogle Scholar
  6. 6.
    Matsuishi T, Nagano M, Araki Y, Tanaka Y, Iwasaki M, Yamashita Y, Nagamitsu S, Iizuka C, Ohya T, Shibuya K, Hara M, Matsuda K, Tsuda A, Kakuma T (2008) Scale properties of the Japanese version of the strengths and difficulties questionnaire (SDQ): a study of infant and school children in community samples. Brain Dev 30:410–415CrossRefPubMedGoogle Scholar
  7. 7.
    Hawes DJ, Dadds MR (2004) Australian data and psychometric properties of the strengths and difficulties questionnaire. Aust N Z J Psychiatry 38(8):644–651PubMedGoogle Scholar
  8. 8.
    SDQ (2008) Information for researchers and professionals about the strengths and difficulties questionnaires. Avaliable from http://www.sdqinfo.com/bb2.html
  9. 9.
    SDQ (2008) Information for researchers and professionals about the strengths and difficulties questionnaires. Available from http://www.sdqinfo.com/bb1.html
  10. 10.
    Leung PWL, Hung SF, Ho TP, Lee CC, Liu WS, Tang CP, Kwong SL (2008) Prevalence of DSM-IV disorders in Chinese adolescents and the effects of an impairment criterion: a pilot community study in Hong Kong. Eur Child Adolesc Psychiatry 17:452–461CrossRefPubMedGoogle Scholar
  11. 11.
    Goodman R (2001) Psychometric properties of the strengths and difficulties questionnaire. J Am Acad Child Adolesc Psychiatry 40(11):1337–1345CrossRefPubMedGoogle Scholar
  12. 12.
    Bourdon K, Goodman R, Rae D, Simpson G, Koretz D (2005) Strengths and difficulties questionnaire: US normative data and psychometric properties. J Am Acad Child Adolesc Psychiatry 44(6):557–564CrossRefPubMedGoogle Scholar
  13. 13.
    Leung PWL, Kwong SL, Tang CP, Ho TP, Hung SF, Lee CC, Hong SL, Chiu CM, Liu WS (2006) Test–retest reliability and criterion validity of the Chinese version of CBCL, TRF, and YSR. J Child Psychol Psychiatry 47(9):970–973CrossRefPubMedGoogle Scholar
  14. 14.
    Achenbach TM, McConaughy SH, Howell C (1987) Child/adolescent behavioural and emotional problems: implications of cross-informant correlations for situational specificity. Psychol Bull 101(2):213–232CrossRefPubMedGoogle Scholar
  15. 15.
    Mullick MSI, Goodman R (2001) Questionnaire screening for mental health problems in Bangladeshi children: a preliminary study. Soc Psychiatry Psychiatr Epidemiol 36:94–99CrossRefPubMedGoogle Scholar
  16. 16.
    Shek DTL (1997) Mental health of chinese adolescents. In: Lau S (ed) Growing up the Chinese way. Chinese University Press, Hong KongGoogle Scholar
  17. 17.
    Man P (1991) The influence of peers and parents on youth life satisfaction in Hong Kong. Soc Indic Res 24:347–365CrossRefGoogle Scholar
  18. 18.
    Ho TP, Leung PWL, Luk ESL, Taylor E, Bacon-Shone J, Lieh-Mak F (1996) Establishing the constructs of childhood behavioural disturbances in a Chinese population: a questionnaire study. J Abnorm Child Psychol 24(4):417–431CrossRefPubMedGoogle Scholar
  19. 19.
    Luk SL, Leung PWL, Lee PLM (1988) Conners’ teacher rating scale in Chinese children in Hong Kong. J Child Psychol Psychiatry 29(2):165–174CrossRefPubMedGoogle Scholar

Copyright information

© Springer-Verlag 2009

Authors and Affiliations

  • Kelly Y. C. Lai
    • 1
  • Ernest S. L. Luk
    • 1
  • Patrick W. L. Leung
    • 2
  • Ann S. Y. Wong
    • 2
  • Lawrence Law
    • 2
  • Karen Ho
    • 2
  1. 1.Department of PsychiatryChinese University of Hong KongShatinHong Kong
  2. 2.Department of PsychologyChinese University of Hong KongShatinHong Kong

Personalised recommendations