Prevalence: are two-fifths of young people really ‘abnormal’?

We read with interest Deighton and colleagues’ paper about mental health problems among 11- to 14-year-olds.¹ They stress the importance of understanding prevalence and report ‘findings that indicate the scale of mental health problems in England is much higher than previous estimates’. The previous estimates referred to are from England's Mental Health of Children and Young People (MHCYP) survey, which recently identified 13.6% of 11- to 15-year-olds as meeting the diagnostic criteria for a mental disorder.² They do not explain why their estimate of 42.5% is more reliable.

Their survey was conducted in six of the most deprived local authorities in the country: Blackpool, Cornwall, Hull, Kent, Newham and Wolverhampton. The MHCYP survey was nationally representative. As expected, given the deprived areas sampled, children eligible for free school meals were overrepresented, as well as White pupils. These characteristics are associated with higher rates of disorder,³ but are not addressed with the use of survey weights. The MHCYP survey used a complex weighting strategy to correct for selection and non-response biases to ensure that the sample was representative.

Only the child self-report Strengths and Difficulties Questionnaire (SDQ) was used. The single-informant SDQ is a less reliable predictor of child mental disorder than the multi-informant SDQ, and the child self-report measure is less reliable than the parent or teacher measures.⁴ In contrast, the MHCYP used a multi-informant standardised diagnostic assessment; the Development and Wellbeing Assessment. This combines highly structured and semi-structured questions, as well as clinical rating to triangulate child, parent and teacher reports and assign ICD-10 diagnoses.⁵

Prevalence estimates rely on the thresholds applied. To identify pupils with problems in each of the four domains examined the authors have used a no longer recommended ‘three band’ approach and,⁶ crucially, do not appear to have taken account of impact.

Similarly, the overall threshold was derived by a score above the subscale cut-point on four of six possible subscales. This unusual approach was not explained, although we are sympathetic to the challenges of describing complex methodology within a short report. The standard approach would be to apply a threshold to the SDQ total difficulties score.⁴

We disagree, therefore that the authors’ findings indicate that the MHCYP's rates are underestimates. Poor mental health can be conceptualised in a number of ways, and clarity about definitions, especially when making comparisons, is essential.