Our study synthesizes the available evidence, in particular that from high-quality evaluations. We provide an up-to-date, comprehensive estimate of the potential savings from parenting programmes, using empirically derived estimates for the UK. Our analysis is based on the results of an extensive literature search, using estimates from systematic reviews where possible. Both the base case and sensitivity analyses rely on evidence-based, conservative assumptions. The cost of parenting programmes is based on data provided by developers rather than a hypothetical intervention.
Existing studies of service use and cost commonly include only a small number of children. By combining data from several studies we have tried to obtain a more robust estimate of costs for younger children, but as far as we are aware, only one study reports the longer term costs of conduct disorder in England , leading to uncertainty in the estimate of future savings.
While the cost analysis aimed to be comprehensive, a number of potential cost savings had to be excluded. Adults with a history of childhood conduct disorder experience a range of negative outcomes . Overall, they are likely to earn lower-than-average wages  and have a higher probability of being unemployed, although those in employment may earn more than their peers . Childhood conduct disorder is also associated with other adult mental health problems and disability [31, 33]. However, there is insufficient evidence to suggest parenting programmes would have an impact on these outcomes. For the same reason, social security benefit receipt has also been excluded, although these costs may be substantial [16, 26].
We have also had to exclude from the model other potential positive effects of the intervention such as impacts on the child's social network (parental mental health and employment effects, benefits to siblings and peers, intergenerational effects), those that overlap with other cost categories (educational attainment, teenage pregnancy, smoking, drug and alcohol abuse) and excess mortality due to higher rates of accidents and suicide [6, 33]. While the aim was to look widely at all possible cost savings from providing parenting programmes for childhood conduct disorder, the lack of evidence means that the model is limited to reduced public expenditure and savings from the prevention of crime.
Problems arose from the absence of large longitudinal studies of children with conduct disorder in the UK. For example, to estimate crime rates, the cross-sectional Offending, Crime and Justice Survey has been interpreted as longitudinal data. Moreover, while the sample is weighted to be nationally representative, this survey excludes some potentially high offending groups such as those in prison, and offences such as homicide and sexual offences which will have high cost implications. We have also been unable to adjust for types of crime committed by those with conduct disorder; it may be that their pattern of offending behaviour is different from that of the rest of the population. Each of these factors will have an impact on the accuracy of our calculated savings to the criminal justice services but their inclusion is unlikely to change the overall findings.
Our model focuses on the savings to be achieved by reducing the probability of persistent conduct disorder for children with the most severe conduct problems. In analysis published elsewhere, we demonstrate that parenting interventions are cost-saving even when accounting for the fact that some children will improve but still exhibit behaviour problems . The strongest evidence base both in terms of costs and intervention effectiveness, however, exists for children with clinical levels of conduct disorder and while conduct problems exist on a continuum and non-clinical levels may be associated with increased service costs compared to children without behaviour problems , the cost implications are by far largest for those with the most severe problems . To illustrate this point, the 8% of males classed as 'prolific offenders' account for two thirds of all criminal convictions , and most of them will have exhibited behaviour problems since childhood. Furthermore, meta-analyses have found that studies of parenting programmes for children with clinical levels of behaviour problems have bigger effect sizes than those for children with sub-clinical levels [23, 35], meaning that the biggest improvements can be achieved for this group. Consequently, reducing the prevalence of clinical conduct disorder will generate the largest savings and the largest benefit to wider society.
The model results are sensitive to changes in the assumptions about the natural course of conduct disorder. If more (fewer) people recovered without intervention, this would decrease (increase) the savings from the intervention.
Finally, although the intervention cost is derived from information provided by developers of evidence-based parenting programmes, there may be additional staff costs associated with engagement, preparation and follow-up support for families, and any additional organisational costs associated with rolling-out parenting programmes are not captured in the model.