Thirty two studies were included in the review: 29 assessed the Child Behaviour Checklist (n=25,006); and three assessed the Strengths and Difficulties Questionnaire (n=10,424). Participation rates ranged from 30% to 95%. Most studies collected scale and reference standard results for all children. In most studies Child Behaviour Checklist scores and comparison diagnoses were generated blindly. In nine studies of the Child Behaviour Checklist, data were collected via a multi-stage screening process and it was unclear whether Child Behaviour Checklist results were known when making the diagnosis.
The Child Behaviour Checklist had a sensitivity of 66% (95% CI 60% to 73%) and specificity of 83% (95% CI 81% to 85%) for total problems. Pooled sensitivity was similar for mood/anxiety (59%) and disruptive behaviour (61%) and higher for attention (71%). Estimates of specificity were similar across the diagnostic domains (72% to 79%).
The Strengths and Difficulties Questionnaire was less sensitive (pooled sensitivity 49%, 95% CI 46% to 51%), but more specific (pooled specificity 93%, 95% CI 92% to 94%) than the Child Behaviour Checklist. Sensitivity was lower for mood/anxiety (54%) than for disruptive behaviour (75%) or attention (73%). Estimates of specificity were similar across the diagnostic domains (85% to 91%).
Results of the sensitivity analysis were reported for setting, but insufficient data were reported to allow a meaningful interpretation of results. There was insufficient data to conduct any other of the proposed subgroup analyses. Data on heterogeneity were only presented for likelihood ratios, which were generally found to be homogeneous, but the analysis focused on positive likelihood ratios with no data on negative likelihood ratios and so were not included here.