Identifying high school risk factors that forecast heavy drinking onset in understudied young adults

Qingyu Zhao, Magdalini Paschali, Joseph Dehoney, Fiona C. Baker, Massimiliano de Zambotti, Michael D. De Bellis, David B. Goldston, Kate B. Nooner, Duncan B. Clark, Beatríz Luna, Bonnie J. Nagel, Sandra A. Brown, Susan F. Tapert, Sonja Eberson, Wesley K. Thompson, Adolf Pfefferbaum, Edith V. Sullivan, Kilian M. PohlDevelopmental 2024 · DOI

TL;DR

This study uses machine learning to identify distinct high school risk factors that predict heavy drinking onset in young adults, differentiating between those who attend 4-year colleges and those who enter the workforce, military, or community colleges. The findings highlight that childhood trauma and extraversion are key predictors for the latter group, while emotion recognition and openness are more relevant for college-bound individuals, underscoring the need for tailored prevention strategies.

What Problem Does This Paper Address?

Heavy alcohol consumption is a significant public health issue, particularly among young adults. Existing research on predicting this behavior has largely focused on individuals attending 4-year residential college programs (4YCPs). However, this group represents a minority of young adults; the majority pursue other paths like entering the workforce, military service, or attending community colleges. These "understudied young adults" (USYAs) have different life trajectories and may face unique risk factors for heavy drinking that have not been well-characterized. This paper aims to bridge this gap by identifying specific high school predictors of heavy drinking onset in USYA individuals, comparing them to their 4YCP peers, and exploring sex-specific differences.

Methods

The researchers employed a machine learning approach, specifically a linear Support Vector Machine (SVM), to forecast the likelihood of initiating heavy drinking after high school. They utilized a comprehensive set of mental health metrics collected during adolescence, including historical factors (family drinking history, childhood trauma), personality traits (impulsivity, openness, extraversion), mood (depressive symptoms), and cognitive abilities (emotion recognition, response times). These 27 factors were averaged across pre-18 years of age. A Leave-One-Out cross-validation strategy was used to train and evaluate the SVM model, maximizing the training sample size for the relatively smaller USYA cohort. To understand the contribution of each factor, SHapley Additive exPlanations (SHAP) values were calculated. Statistical tests, including t-tests and general linear models, were used to compare risk factors between groups and sexes.

Datasets

The study utilized data from the National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA) study. This longitudinal dataset includes psychobiological measures from participants aged 12-21 at baseline, followed for up to 8 years into young adulthood. For this study, 106 USYA participants and 453 4YCP participants who were no-to-low drinkers before age 18 were included. A subset of 106 4YCP participants was matched demographically to the USYA cohort for comparison.

Key Insights and Ideas

  • Divergent Pathways to Heavy Drinking: The most striking insight is that the predictors of heavy drinking onset are not universal. They differ significantly based on whether a young adult enters a 4-year college or pursues other career paths. This challenges the assumption that risk factors are uniform across all young adults.
  • Childhood Trauma as a Major Risk for USYAs: For young adults not attending 4-year colleges, early life experiences like sexual and physical abuse emerged as significant predictors of later heavy drinking. This suggests that unresolved trauma may play a more prominent role in shaping drinking behaviors in this population.
  • Personality and Cognitive Differences: While trauma is key for USYAs, cognitive abilities (specifically, the speed of emotion recognition) and personality traits (openness) are more predictive for 4YCP individuals. This hints at different underlying mechanisms driving drinking behaviors in these distinct groups.
  • Sex-Specific Risk Factors: The study found that risk factors are not only dependent on career path but also on sex. For instance, physical abuse was a stronger predictor for female USYA drinkers, while extraversion was more relevant for male USYA drinkers. This emphasizes the need for gender-sensitive prevention programs.
  • Timing Matters: USYA individuals tend to initiate heavy drinking later than 4YCP individuals, who often start during their college years. This temporal difference likely relates to their different social environments and responsibilities.

Key Results

  • The machine learning model achieved a 78.8% accuracy (BAcc) in predicting heavy drinking onset for USYA participants, which was significantly higher than the 66.8% accuracy for the matched 4YCP cohort.
  • For USYA individuals, the key predictors identified were childhood sexual abuse (significant for both sexes), physical abuse (specifically for girls), and extraversion (specifically for boys).
  • For 4YCP individuals, the key predictors were the ability to recognize facial emotions (specifically, faster response times for happiness and anger) and greater openness (specifically for boys).
  • The classification accuracy for female USYA participants (84.7%) was significantly higher than for male USYA participants (72.9%).

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Size: The relatively small sample size of USYA participants (n=106) limits the ability to further subtype risk factors for more specific career choices (e.g., military vs. workforce).
  • Female Heavy Drinkers: The limited number of female heavy drinkers in the USYA group restricted the investigation of sex differences across a wider range of risk factors.
  • Focus on No-to-Low Drinkers: The study only examined individuals who were no-to-low drinkers before age 18. Future research should explore the drinking trajectories and risk factors for those who already initiated drinking during adolescence.

The authors suggest that future work should focus on replicating these findings with larger datasets and exploring more granular career path distinctions within the USYA group. They also highlight the need for tailored prevention and intervention programs that consider the unique risk profiles of understudied young adults.

Summary updated

Abstract

Heavy alcohol drinking is a major, preventable problem that adversely impacts the physical and mental health of US young adults. Studies seeking drinking risk factors typically focus on young adults who enrolled in 4-year residential college programs (4YCP) even though most high school graduates join the workforce, military, or community colleges. We examined 106 of these understudied young adults (USYA) and 453 4YCPs from the National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA) by longitudinally following their drinking patterns for 8 years from adolescence to young adulthood. All participants were no-to-low drinkers during high school. Whereas 4YCP individuals were more likely to initiate heavy drinking during college years, USYA participants did so later. Using mental health metrics recorded during high school, machine learning forecasted individual-level risk for initiating heavy drinking after leaving high school. The risk factors differed between demographically matched USYA and 4YCP individuals and between sexes. Predictors for USYA drinkers were sexual abuse, physical abuse for girls, and extraversion for boys, whereas 4YCP drinkers were predicted by the ability to recognize facial emotion and, for boys, greater openness. Thus, alcohol prevention programs need to give special consideration to those joining the workforce, military, or community colleges, who make up the majority of this age group.
BibTeX
@inproceedings{zhao2024identifying,
  author = {Qingyu Zhao and Magdalini Paschali and Joseph Dehoney and Fiona C. Baker and Massimiliano de Zambotti and Michael D. De Bellis and David B. Goldston and Kate B. Nooner and Duncan B. Clark and Beatríz Luna and Bonnie J. Nagel and Sandra A. Brown and Susan F. Tapert and Sonja Eberson and Wesley K. Thompson and Adolf Pfefferbaum and Edith V. Sullivan and Kilian M. Pohl},
  title = {{Identifying high school risk factors that forecast heavy drinking onset in understudied young adults}},
  booktitle = {Developmental},
  year = {2024},
  doi = {10.1016/j.dcn.2024.101413},
}
Figure