A new statistical model for binge drinking pattern classification in college-student populations

Judith André; Momar Diouf; Margaret P Martinetti; Olivia Ortelli; Fabien Gierski; Frederic Fürst; Olivier Pierrefiche; Mickael Naassila

doi:10.3389/fpsyg.2023.1134118

A new statistical model for binge drinking pattern classification in college-student populations

Front Psychol. 2023 Jul 17:14:1134118. doi: 10.3389/fpsyg.2023.1134118. eCollection 2023.

Authors

Judith André¹, Momar Diouf², Margaret P Martinetti^{1

3}, Olivia Ortelli³, Fabien Gierski^{1

4

5

6}, Frederic Fürst⁷, Olivier Pierrefiche¹, Mickael Naassila^{1

5

6}

Affiliations

¹ INSERM UMR 1247, Groupe de Recherche sur l'alcool et les Pharmacodépendances, GRAP, Université Picardie Jules Verne, Amiens, France.
² Biostatistics Unit, Clinical Research Department, Amiens-Picardie University Hospital, Amiens, France.
³ Department of Psychology, The College of New Jersey, Ewing, NJ, United States.
⁴ Cognition, Health, Society Laboratory (C2S - EA 6291), University of Reims Champagne Ardenne (URCA), Reims, France.
⁵ Fédération Hospitalo-Universitaire A2M2P, Améliore le Pronostic des Troubles Addictifs et Mentaux par une Médecine Personnalisée, Paris, France.
⁶ GDR CNRS 3557 Psychiatrie-Addictions, Institut de Psychiatrie, Paris, France.
⁷ Laboratoire MIS (Modélisation, Information et Système) UR 4290, Université Picardie Jules Verne, Amiens, France.

Abstract

Background: Binge drinking (BD) among students is a frequent alcohol consumption pattern that produces adverse consequences. A widely discussed difficulty in the scientific community is defining and characterizing BD patterns. This study aimed to find homogenous drinking groups and then provide a new tool, based on a model that includes several key factors of BD, to assess the severity of BD regardless of the individual's gender.

Methods: Using the learning sample (N1 = 1,271), a K-means clustering algorithm and a partial proportional odds model (PPOM) were used to isolate drinking and behavioral key factors, create homogenous groups of drinkers, and estimate the probability of belonging to these groups. Robustness of our findings were evaluated with Two validations samples (N2 = 2,310, N3 = 120) of French university students (aged 18-25 years) were anonymously investigated via demographic and alcohol consumption questionnaires (AUDIT, AUQ, Alcohol Purchase Task for behavioral economic indices).

Results: The K-means revealed four homogeneous groups, based on drinking profiles: low-risk, hazardous, binge, and high-intensity BD. The PPOM generated the probability of each participant, self-identified as either male or female, to belong to one of these groups. Our results were confirmed in two validation samples, and we observed differences between the 4 drinking groups in terms of consumption consequences and behavioral economic demand indices.

Conclusion: Our model reveals a progressive severity in the drinking pattern and its consequences and may better characterize binge drinking among university student samples. This model provides a new tool for assessing the severity of binge drinking and illustrates that frequency of drinking behavior and particularly drunkenness are central features of a binge drinking model.

Keywords: alcohol; binge drinking; classification tool; college students; drunkenness; gender; high-intensity binge drinking.