Development and validation of a risk predictive model for student harmful drinking-A longitudinal data linkage study.

Area of Science:

• Public Health
• Higher Education
• Data Science

Background:

• Student harmful drinking is a significant concern on U.S. university campuses.
• Emergency department (ED) visits and campus incidents are indicators of harmful alcohol use.
• Predictive modeling can help identify at-risk students for targeted interventions.

Purpose of the Study:

• To develop and validate a predictive model for quantifying student risk of harmful drinking.
• To utilize routinely collected student data for risk assessment.
• To support proactive campus health and safety initiatives.

Main Methods:

• Linked six years of student enrollment data with harmful drinking events (ED visits, campus incidents).
• Developed a multivariable logistic regression model using a 3-year cohort (n=93,289).
• Validated the model in a separate 3-year cohort (n=85,876).

Main Results:

• Harmful drinking incidence was 2.8% in the derivation cohort and 3.1% in the validation cohort.
• Significant predictors included demographics, academic factors, Greek life, prior diagnoses (depression, injury), and violence involvement.
• The model demonstrated strong predictive accuracy (C-statistics=0.86) with excellent calibration in both cohorts.

Conclusions:

• A robust, validated predictive model for student harmful drinking risk was developed.
• The model effectively quantifies absolute risk using readily available student data.
• This tool can assist health educators and clinicians in implementing timely, targeted interventions for students at elevated risk.