An Algorithm to Predict the Best Treatment for Disruptive Behaviour Problems Each Child

GGz (Dutch mental health care) professionals find themselves trapped in a catch-22: If they adhere to standardized, evidence-based treatment protocols, they fail to adapt treatment to the needs of individual clients; yet, if they do adapt such treatment protocols, they fail to comply with evidence-based treatment standards. This captures in a nutshell the pressing need for evidence-based personalization of GGz treatment. As it stands, the science and clinical tools for this are missing. With this fellowship, I will develop and implement an algorithm to predict the best treatment content for early disruptive behavior problems in individual children. Strongly embedded in a collaborative network of clinical and educational partners, the proposed project will lay the groundwork for evidence-based personalization of GGz treatment. Disruptive behavior is one of the most prevalent reasons for children’s referral to the GGz [1], and if left untreated, predicts serious deterioration of children’s healthy development (e.g., psychological, social, academic, and physical) [2,3]. Disruptive behavior problems can be identified early (i.e., around age 4), but as few as 25% of children benefit from even our most effective, evidence-based treatments [4, and Publication 3]. Current treatment approaches seek to address a multitude of risk factors for disruptive child behavior in one single protocol. From medicine, however, we know that precisely targeted interventions outperform these “scattergun approaches” [5,6]. How can we precisely target treatment for disruptive child behavior? Work by myself and others shows that individual client characteristics rarely predict treatment success [Publications 4,10,13,21,27], and that there are no “effective elements” that guarantee treatment success in all families [Publications 1,2,5]. What does predict treatment success, is how families’ personal risk factors interact with treatment content [Publications 1,3,8]. Disruptive behavior problems tend to be maintained by three core parenting risk factors: a distorted parent-child relationship, a cycle of coercive parent-child interaction, and compromised parental self-efficacy. With regard to treatment content, I argue that three requirements need to be met: • RELEVANCE: The risk factors targeted in treatment should not only be present, but actually contribute to a child’s disruptive behavior. • MALLEABILITY: It should actually be possible to change the risk factors targeted in treatment. • ACCEPTABILITY: Treatment to change the risk factors should align with personal family values. None of these requirements can be assumed to hold true for all families alike. First, what constitutes a risk factor for one family may not constitute a risk factor for another—the relevance of risk factors varies across families. Second, as my previous work has shown, some parenting risk factors are more malleable than others [Publication 9], and such malleability also differs across families. Third, parents vary in how acceptable they find evidence-based strategies to address risk factors. In the proposed project, I will develop an algorithm that predicts the optimal treatment content for individual children with early disruptive behavior problems. First, I will identify clusters of families who show similar profiles in terms of risk factor relevance (Study 1), risk factor malleability (Study 2), and perceived treatment acceptability (Study 3). The family profiles that derive from Studies 1–3 allow for assigning scores to individual families for each risk factor (i.e., distorted parent-child relationship, coercive parenting child interactions, and compromised parental self-efficacy), on risk factor relevance, malleability, and acceptability. These are the explanatory variables in the equation that predicts treatment effects for individual families. I will then empirically test the algorithm in an independent trial (Study 4). Finally, I will translate the algorithm into a decision tool for GGz professionals, and implement and evaluate this tool in various GGz organizations (Study 5). In addition, I will translate the algorithm into teaching modules for institutes that train new generations of GGz professionals. Client panels have laid the basis for this project, and remain part of it throughout. Scientifically, this project aspires to break new ground in our understanding of how risk factor relevance and malleability differ across families. For the GGz, it will provide a clinical decision tool to aid evidence-based personalized treatment for early disruptive behavior problems. The absence of evidence-based personalization tools in the GGz starkly contrasts with areas of medicine (e.g., oncology and cardiology), where algorithms have important roles in clinical decision making, and routinely outperform care as usual. This project will lay the empirical groundwork for the use of algorithms to aid personalized evidence-based treatment in the GGz.