This study was the first to develop an algorithm for the detection of antisocial behaviour in 8-12 year-old primary school children, using information that can be obtained during well-child visits. Our findings show that this information may indeed help CHPs to identify children who are at increased risk of antisocial activities, in general, and violence against people, in particular. However, the predictive power of the detection algorithms as measured by the AUC was relatively poor.
Our findings show that a detection algorithm based on routinely available data may be a useful first step in a multi-step detection procedure in CHP practice. In a second step, confirmative testing on antisocial acts would be needed in a selected part of the population. The resulting final group could be offered further early intervention which has been shown to decrease antisocial behaviour in about 66% [15–17]. This could yield a 20-29% reduction in antisocial behaviour in the community, albeit at the expense of a group which had been false-positive at earlier stages of the procedure.
The discriminatory power of the detection algorithm is moderate, which indicates that it needs to be improved for application in routine preventive child healthcare. Several approaches may yield such an improvement. First, other characteristics might be included in the detection algorithm. As reported by others, antisocial behaviour was associated with male gender, large family, young mother, poor child-parent relationship, and substance use [8, 26–29]. One might consider to extend the preventive child healthcare assessment procedure by other potential predictors. Candidates might be child characteristics such as (low) intelligence or academic performance, externalizing behaviour, hyperactivity and behavioural problems; parent characteristics such as (poor) parental supervision, hostile parenting, physical punishment, parent-child separation, deviant mother-child interactions, parental criminality, maternal smoking during pregnancy, and family psychiatric history [30–32]; and social factors, such as antisocial peers and high delinquency neighbourhood . If these factors would be included in the assessment procedure, it definitely requires additional study whether this could be managed in the available time per visit. In addition, it requires additional study whether data in the 'possibly available' category, i.e. substance use, can be obtained in a valid way indeed.
As a second means to improve detection, one might consider having the child fill out the same questionnaire as used in this study or a similar one. However, if the child would have to give the completed questionnaire to the CHP, this is very likely to lead to biased information compared to the setting of this study in which confidentiality was guaranteed to the child with only the researchers reading the answers after removal of all identifying data.
Third, information from teachers might be added. Petras et al. found good prediction of violence using the Teacher Observation of Classroom Adaptation (TOCA) [13, 14]. However, consent needs to be obtained from parents if applied for well-child purposes, which may limit its applicability.
Fourth, parent-reported questionnaires on antisocial behaviour might be added to the routine behavioural assessment at the time of well-child visits. However, it may in fact be quite questionable whether the parent would actually be well-informed about any such behaviour, even in cases of great concern.
We defined antisocial behaviour as an act of violence against either property or people [4, 24], leading to direct and indirect effects on health [5, 6]. This broad definition may explain the higher prevalence of antisocial behaviour in our study compared with previous studies that used a more restrictive definition based upon judicial prosecution [13, 14], or psychopathology [8, 26]. We think that our definition better reflects antisocial behaviour as perceived in the community, given the process of development of the ISRD questionnaire [4, 24]. Our definition may include transient antisocial behaviour, but early onset has been shown to be predictive for a life-long career of such behaviour. Early detection and intervention may turn the trait into a socially acceptable lifestyle [1–4]. Future studies are needed to evaluate the effectiveness of intervention in antisocial behaviour after detection based on information obtained from routine well-child assessments or school health records. Finally, one might challenge our definition of violence against people, in particular the inclusion of threatening someone. Repeating the analyses with exclusion of this item did not affect the results, however.
Strengths and limitations
The strengths of this study lie in its community-based setting using information that is commonly available from school health records to detect pre-adolescents at risk of antisocial behaviour. The limitations of our study involved missing data, the data collection procedure, and the definition of antisocial behaviour. First, we measured anti-social behaviour using self-report. This may have resulted in underreporting. However, the alternatives - observation and proxy-reporting - would likely yield much more underreporting, and previous studies have shown the ISRD questionnaire to be highly valid [4, 24]. In addition, the data might have been collected in a more rigorous way than would actually occur in routine well-child care. Therefore, our results need confirmation in routine practice.
Our findings imply that well-child health care can support the early detection of antisocial behaviour. Additional measurements on other predictors of antisocial behaviour are needed, however, to further strengthen the subsequent stages of this early detection. This could in the end contribute towards resolving what is a major threat to both the health of the individuals involved and to society as a whole.