The Strength and Difficulties Questionnaire: Predictive validity of parent and teacher ratings for help-seeking behaviour over one year

Latent growth modelling was applied to investigate the predictive validity of the subscale and total difficulties scores from the Strengths and Difficulties Questionnaire (SDQ). Prospective data were collected on a community sample of n 659 children (aged 7 to 11 years) over a one-year period. Outcomes at one year after baseline were defined in terms of parental help-seeking behaviour and worry. Teacher-reported SDQ summary scores were more predictive of help-seeking behaviour from both the general practitioner and the school than parent-reported SDQ scores. Changes (increases) in SDQ scores proved more useful in predicting help-seeking than initial scores. Gender, age and socioeconomic status were not related to help-seeking behaviour, but the parents of children with higher IQ scores were more likely to seek help. The usefulness of the SDQ to predict help-seeking behaviour for emotional-behaviour difficulties in community settings and its implications for service-use issues in the UK context are discussed. classifying children and adolescents as potential ‘cases’ using thresholds on screening questionnaires, or scoring them on multidimensional symptom scales (Costello et al., 1993; Goodman et al., 2003; Lahey et al., 1996; Shaffer et al., 1996). The Strengths and Difficulties Questionnaire (Goodman, 1997, 2001) is frequently used in UK samples to screen children aged 3–16 years for emotional, behavioural and social or relationship difficulties. The SDQ is ideal for use in community or primary care settings because it is shorter than established instruments, can be administered to parents, teachers, or for 11–16-year-olds through selfreport. Apart from its brevity (it takes approximately five minutes to complete), it provides ratings from multiple informants that can be used for multi-informant comparisons. In most areas it has been shown to function as well as the established Rutter questionnaires and it has some advantages over the Child Behaviour Checklist (Achenbach, 1991) which is more time consuming to administer (Goodman, 1997). In addition, the SDQ is free to use, with versions for download available on the Internet (http:// www. sdqinfo.com). International versions are available, with the SDQ translated into over 40 languages. Empirical studies have begun to reveal the SDQ’s usefulness in detecting psychiatric disorders in the community. Multi-informant SDQs have identified individuals with a psychiatric diagnosis with a specificity of 94.6 per cent (95 per cent confidence interval (CI), 94.1–95.1) and a sensitivity of 63.3 per cent (59.7–66.9 per cent) (Goodman et al., 2003). Similar results were reported by criterion validity studies for the SDQ in Australia (Mathai et al., 2003), the Nordic countries (Obel et al., 2004), Dhaka (Goodman et al., 2000), southern European countries (Marzocchi et al., 2004), the Netherlands (van Widenfelt et al., 2003) and Germany (Woerner et al., 2002). Several studies in the UK and abroad have demonstrated the construct validity of the SDQ (see Table 1 for a summary), but only four studies (Ford et al., 2003; Goodman et al., 2002; Hawes & Dadds, 2004; Mathai et al., 2003) have reported longitudinal data, of which only two are UK-based (Ford et al., 2003; Goodman et al., 2002). An aspect that has not yet been addressed is the ability of SDQ scores to predict later, not concurrent, help-seeking behaviour. SDQ scores have been shown to correlate with concurrent treatment status (Hawes & Dadds, 2004) and help-seeking (Koskelainen et al., 2000), but neither of these studies was from UK samples and neither approached this issue using prospective longitudinal data. If the SDQ is to play a role in the early detection and identification of children in the community who may cause enough concern to parents or teachers to trigger help-seeking, then its predictive validity in this context needs to be demonstrated. The first aim of the current study was therefore to test the ability of SDQ scores to predict help-seeking behaviour at one year. For this purpose, we divided sources of help and advice into three levels, following the approach of Ford et al. (2003): informal contacts included discussing child mental health concerns with family members or friends; front-line services comprised the general practitioner (GP) and within-school resources, such as teachers and head-teachers; specialist services referred to CAMHS. We also included a fourth outcome measure, parental worry, to investigate whether SDQ baseline scores would predict how much a parent would feel worried or concerned about their child one year later. Many other factors have been shown to influence referral and help-seeking, including the nature, severity and comorbidity of mental health problems, psychosocial family stress, socio economic status (SES), parental psychiatric illness, poor peer and school performance, age and gender (Garralda & Bailey, 1988; Verhulst & van der Ende, 1997; Zahner & Daskalakis, 1997; Zwaanswijk, Van der Ende et al., 2003; Zwaanswijk, Verhaak, et al., 2003). We therefore also considered child age, gender, IQ (as an index of Educational & Child Psychology Vol 22 No 3 29 The strength and difficulties questionnaire