Karwatowska, Lucy; (2024) Risk factors for disruptive behaviour: Triangulating evidence from causal inference methods. Doctoral thesis (Ph.D), UCL (University College London).

Text Karwatowska_10191417_thesis_sigs_removed.pdf Access restricted to UCL open access staff until 1 May 2025. Download (14MB)

Abstract

Disruptive behaviour disorders (DBDs) are a common set of diagnoses affecting around 5.7% of children and adolescents globally. Due to the high personal, social, and economic costs associated with DBDs, there has been a long-standing interest in understanding the environmental, physiological, and genetic factors that may underlie or cause such behaviours. However, until recently, current research on risk factors has been limited due to a lack of generalisability, a failure to account for the complexities of various risk factors, and a reliance on classical correlational analyses, making it hard to establish causation. This thesis will leverage the recent availability of evidence from large pre-existing datasets, such as genome-wide association studies and registry-based administrative databases. It will use modern causal inference methods to examine putative risk factors for DBDs. The first study in this thesis is a systematic review of research using causal inference methods. The findings highlighted which risk factors have been studied using causal inference methods and which methods have been used to answer causal questions on the development of DBD. The results informed which risk factors and methods were employed in the current thesis. In the second study, I examined a well-studied environmental risk factor, parenting practices, that could inform preventative interventions for DBDs. I conducted a meta-analysis of 41 studies using causal inference methods. The pooled estimate suggested evidence of a small, causal effect of negative parenting practices on offspring DBDs (r = 0.142; 95% CI = 0.104, 0.180). I estimated that a hypothetical intervention which reduced negative parenting could lead to a 0.11% reduction in the prevalence of DBDs, the equivalent of 3,614,337 school-aged children worldwide no longer exhibiting clinical levels of DBD symptoms. In the third study, I investigated whether resting heart rate (RHR), a putative physiological risk factor for disruptive behaviour rarely studied using causal inference methods, is causally related to antisocial behaviour (ASB), a common symptom of DBDs. Using two-sample Mendelian randomisation (MR) and linkage disequilibrium score regression analyses, I found no evidence of causal associations (BIVW = –0.0004; 95% CI = -0.004, 0.004, NSNPS = 278) or genetic correlations (rg = 0.057, 95% CI = -0.025, 0.139) between RHR and ASB. I discuss how MR can be an effective tool to assess risk factors for DBDs quickly and efficiently, especially for those with a strong genetic basis, provided certain assumptions are met. In the fourth study, I considered a set of risk factors, exposure to early life adversity, that could help inform public health initiatives for DBDs. I used administrative records from over 1.9 million children and identified four latent trajectory groups of adversity from birth to age six years. I then examined whether group membership was associated with diagnoses of CD or dissocial personality disorder (DPD; n = 6,502) or convictions of sexual and violent crimes (n = 35,036) before the age of 25 years. The rates of diagnoses and convictions were higher for individuals who experienced early life adversity than individuals in the low adversity group. I also estimated the average treatment effect of a hypothetical intervention that assigned individuals to the lowest adversity group and predicted that it could lead to a two- to three-fold decrease in the probability of diagnoses (males: ATE = 2.54, 95% CI = 2.27, 2.80; females: ATE = 3.12, 95% CI = 2.59, 3.68) and convictions (males: ATE = 2.25, 95% CI = 2.15, 2.35; females: ATE = 3.02; 95% CI = 2.71, 3.39) by the age of 25 years. This thesis triangulates evidence on risk factors for DBDs by synthesising findings from previous causal inference studies and adopting several novel causal inference methods that rely on different information sources and assumptions.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item