Modelling Individual and Parental Predictors of Adolescent Substance Use in Scotland

A large national study of Scottish teenagers has revealed that depression, parental conflict and the quality of the parent–child bond each carry distinct weight in predicting whether adolescents smoke, vape, drink or try drugs — and that mothers and fathers appear to play meaningfully different roles in shaping these risks. The research, published in the International Journal of Mental Health and Addiction, analyzed data from 2,943 fourteen- and fifteen-year-olds and their parents drawn from the Growing Up in Scotland (GUS) birth cohort, one of the most comprehensive longitudinal datasets on child development in the United Kingdom.

The findings arrive at a moment of shifting patterns in youth substance use. While smoking and cannabis use have declined substantially across Scotland and the wider UK in recent decades, alcohol use remains widespread — more than half of the adolescents in the study reported having consumed alcohol at least once — and the rise of e-cigarette use among teenagers has introduced a new and still poorly understood public health concern. Although e-cigarettes are often marketed as harm-reduction alternatives to cigarettes, accumulating evidence suggests they may carry their own health consequences for young people and may act as a gateway to nicotine addiction, although the strength of that gateway effect remains contested in the literature.

The research team, led by Shuya Xie of the University of Edinburgh’s Department of Clinical and Health Psychology, framed their investigation within Bronfenbrenner’s bioecological model of human development, which holds that an adolescent’s health outcomes emerge from the interaction of individual characteristics with nested layers of environmental context — with the family representing the most proximal, or “microsystem,” influence. Few studies have applied this ecological lens systematically to substance use, and even fewer have separated the contributions of maternal and paternal relationship factors, a gap the current study was designed to address.

The methodological backbone of the study was a set of multivariate logistic regression models, run separately for maternal and paternal factors to allow direct comparison between the two sets of influences. Lifetime substance use was measured across six binary outcomes: ever having tried a cigarette, an e-cigarette or vaping device, an alcoholic drink, having been drunk, cannabis use and use of other drugs. These items were adapted from well-established instruments including the Health Behaviour in School-aged Children survey and the Scottish Adolescent Lifestyle and Substance Use Survey. Depression risk was classified using three screening items from the short form of the World Health Organization’s Composite International Diagnostic Interview, while attachment security was measured with six child-reported items adapted from the People in My Life scale, and conflict was assessed through four parent-reported items adapted from Growing Up in Australia. The attachment and conflict scales showed strong internal consistency, with reliability coefficients ranging from 0.76 to 0.91.

To guard against confounding, the models controlled for a wide array of demographic and familial variables: child gender and birth weight, maternal age at birth, parental adverse childhood experiences, parental ethnicity, education and employment, family type, household income, number of children, urban versus rural residence and area-level socioeconomic deprivation measured by the Scottish Index of Multiple Deprivation. Predictors were entered hierarchically into the models, a technique used because logistic regression coefficients can shift depending on which other variables are included and how large the sample is.

The results were striking in their consistency for maternal factors. Adolescents classified as at elevated risk for depression were significantly more likely to report lifetime use of every substance examined, with odds ratios ranging from 1.86 to 4.39 — meaning that depression risk roughly doubled to more than quadrupled the likelihood of having used a given substance. Higher mother–youth conflict predicted increased likelihood of substance use across the board, with odds ratios between 1.19 and 1.81, while stronger mother–youth attachment security was consistently protective, associated with odds ratios of 0.73 to 0.81 for each substance use behaviour.

The paternal picture was more nuanced. Higher father–youth conflict predicted increased likelihood of smoking, drinking and other drug use — with an odds ratio of 1.76 for other drug use — but showed no significant association with vaping, drunkenness or cannabis use. Better father–youth attachment, by contrast, was protective for all outcomes except other drug use, with odds ratios between 0.73 and 0.85. Adolescent depression risk remained a significant predictor for all six substances in the paternal models as well, with odds ratios ranging from 1.93 to 3.96. Notably, the magnitude of the odds ratios for other drug use was noticeably higher than for other substances in both model sets, suggesting that depression may be an especially strong signal for the most problematic forms of substance involvement.

The authors interpret the depression findings through the lens of several competing theoretical frameworks. The self-medication hypothesis proposes that adolescents with depressive symptoms use substances to alleviate negative affect, but this account is complicated by evidence that depressed smokers are often motivated to quit. An alternative shared-vulnerability hypothesis holds that common underlying risk factors — genetic, temperamental or environmental — predispose individuals to both depression and substance use. Regardless of the causal mechanism, the practical implication is that depression screening in adolescence may serve as an early identification tool for substance use risk, a window of opportunity that current clinical systems often miss.

The divergence between maternal and paternal effects is perhaps the study’s most provocative contribution. The authors suggest several possible explanations. Mothers are generally perceived by adolescents as more responsive, accepting and supportive, while fathers tend to be viewed as adopting more authoritarian parenting styles — differences that could translate into different mechanisms of influence on risk behaviour. Time investment may also play a role; research cited in the study indicates fathers spend roughly seven hours less per week on household and parenting activities than mothers. The authors also note broader social change: fathers are now more routinely involved in day-to-day parenting than in previous generations, giving paternal relationship quality greater developmental weight than older literature might suggest. These interpretations align with a growing body of work on paternal influences on child mental health, which has historically received far less research attention than maternal factors.

Beyond the family, the study confirmed several established demographic risk patterns. Male gender was associated with greater likelihood of e-cigarette and cannabis use, lower parental education predicted smoking and vaping, only children showed elevated risk for smoking, vaping and drunkenness, and rural residence was linked to higher likelihood of smoking, vaping and drinking. In an unexpected finding, maternal “other” ethnicity — that is, mothers from non-white backgrounds — emerged as a protective factor against adolescent drinking, a result the authors link to the so-called migrant paradox, whereby first-generation migrant adolescents show lower rates of substance use than native-born peers despite facing socioeconomic disadvantages that typically elevate risk.

The study has limitations the authors are careful to acknowledge. Conflict was reported by parents rather than adolescents, and prior research suggests adolescents’ perceptions of parental relationships may correlate more strongly with their own health outcomes than parental reports do. Parental current and lifetime substance use — a well-established proximal risk factor — could not be controlled because the data were not collected. Attachment was measured as a single continuous dimension of security rather than distinguishing attachment styles, avoidance or anxiety. The survey also overlapped with the COVID-19 pandemic, forcing a shift from face-to-face to remote data collection that may have influenced social desirability bias in responses to sensitive questions.

Even so, the practical implications are substantial. The findings lend support to family-based prevention approaches: interventions such as the Strengthening Families Program, tested in twelve randomized controlled trials, have achieved roughly fifty percent reductions in substance misuse among high-risk children, with mechanisms of change operating through reduced conflict and improved family relationships. The current results suggest such programmes may be strengthened by attending deliberately to both parents — not only mothers — and by integrating depression identification into substance use prevention pathways. Reducing parent–youth conflict, bolstering attachment security into adolescence and treating depression in young people, the authors conclude, could yield downstream benefits for substance use behaviours, positioning the family not merely as a context of risk but as an active, modifiable lever for prevention.