Modeling Drinking Behavior Progression in Youth with Cross-sectional Data: Solving an Under-identified Probabilistic...

Area of Science:

• Behavioral Science
• Quantitative Psychology
• Public Health

Background:

• Probabilistic discrete event systems (PDES) offer a method for studying underage drinking dynamics using cross-sectional data.
• PDES model identifiability issues often limit their practical application.
• A novel approach is needed to address these limitations and enhance PDES utility.

Purpose of the Study:

• To develop and test a new method for solving non-identifiable PDES models.
• To apply this method to understand the dynamics of underage alcohol consumption.

Main Methods:

• A PDES model was constructed with four stages: never-drinkers (ND), light/moderate-drinkers (LMD), heavy-drinkers (HD), and ex-drinkers (XD), with 13 transition paths.
• Data from the 2012 National Survey on Drug Use and Health (NSDUH) for individuals aged 12-21 were utilized.
• The Moore-Penrose (M-P) generalized inverse matrix method was employed to solve the PDES model.

Main Results:

• The M-P generalized inverse matrix approach successfully estimated annual transitional probabilities for 13 drinking progression pathways by age group.
• An inverse "J" shaped curve characterized experimental alcohol use patterns from adolescence to young adulthood.
• A significant increase in initiating light/moderate and heavy drinking was observed after age 18, alongside a sharp decrease in quitting.

Conclusions:

• The M-P method provides a unique solution for partially-observed PDES drinking behavior models, aligning with developmental perspectives on underage drinking.
• This approach enhances the utility of PDES for analyzing health behaviors using readily available cross-sectional data.
• The tested M-P approach is a valuable tool for future research on health behaviors.