Investigating Trait and State Risk Factors for Substance Use Among Young Adults in Norway

Abstract

Substance use and substance use disorders (SUD) cause significant mental, physical, and societal costs and suffering. This creates a need for prevention and timely interventions based on appropriate risk models. Cognitive impairment and mental health distress are important risk factors for substance use. These factors vary over time, ranging from stable traits to rapidly fluctuating states. However, traditional data collection approaches do not always capture the necessary variability in state risk factors and substance use outcomes, in or out of laboratory-like, controlled conditions. This hampers the identification of ‘when’ substance use risk is high, not just for ‘who.’ Data collection that captures this variability is thus needed. The current thesis thus aims to explore trait and state risk factors for substance use. It also aims to explore knowledge regarding who is at risk of substance use and when that risk is elevated in young adults from general and clinical cohorts in Norway. Three studies on substance use risk in young adults in Norway were included in this thesis. Study 1 used data from a subsample of the 4th wave of the Trøndelag Health Survey (HUNT4) study that also took part in the digital Memoro substudy, which included self-report data of cognition and mental health and remotely collected web-based neuropsychological test data. This was a crosssectional study on young adults. Studies 2 and 3 included repeated, temporal, prospective data collected either as pen- and paper or on a digital platform in Ecological Momentary Assessment designs (EMA) among patients in substance use treatment. These clinical studies also included baseline mental health assessments, substance use, and cognition. The rationale for selecting these three studies was twofold. First, we sought to assess the differences between the roles of demographic, more static traits, and general risk factors versus momentary, acutely registered state risk factors for substance use. Second, by employing appropriate advanced statistical methods, we sought to explore potential models based on these risk factors that could help detect the immediate risk of future substance use. The overall purpose of exploring such risk models was to help advance the field of substance use prevention and treatment, by finding methods for early detection. Such methods can aid in the timing of interventions to prevent substance use in the future. The three studies generated a total of eight statistical risk models. These models contained both cognitive and mental health variables as significant risk factors for lifetime illicit substance use in the general population. In the clinical populations, variability in cognitive and mental health symptoms and mobile sensor data was significant explanatory variables of craving intensity. Furthermore, only craving intensity was a significant explanatory risk factor for substance use in both clinical studies. The risk models were generally plausible in light of previous research, with few unexpected findings. However, methodological and cohort differences in some areas hamper comparability with previous research in this emerging field. The clinical and public health implications of this research are particularly important in contributing to future research on when individuals are at elevated risk for substance use. Study 2 illustrated a temporal structure among risk factors and outcomes despite its limited sample size and attrition. As such, the findings and conclusions drawn here align with a broader trend in mental health and substance use research, striving towards real-life data collection closer to the patients’ lived lives, to capture and model more of the important variability in risk factors than prevailing methods. This trend aims to inform more timely interventions when risk is elevated. Methodological concerns are related to data quality and the choice of statistical methods that allow for inference from state risk and temporal dynamics of risk factors and outcomes. While the obtained data quality was comparable to similar studies in many respects, there is still a need to improve data quality to provide better statistical power and data resolution for improved risk models.