Bayesian models for semicontinuous outcomes in rolling admission therapy groups.

Area of Science:

• Biostatistics
• Psychology
• Public Health

Background:

• Group therapy is common for substance abuse treatment.
• Rolling enrollment in group therapy creates complex participant outcome correlations.
• Existing statistical models handle normally distributed outcomes but not semicontinuous ones common in substance abuse research.

Purpose of the Study:

• To extend statistical models for group therapy to accommodate semicontinuous outcomes.
• To analyze the impact of a group-based intervention on alcohol consumption.
• To account for temporal dependencies in multivariate session effects.

Main Methods:

• Developed statistical models for semicontinuous outcomes, combining continuous and discrete distributions.
• Utilized Bayesian statistical methods for efficient estimation of nonstandard models.
• Applied models to data from a substance abuse and depression intervention, focusing on average drinks per day.

Main Results:

• The intervention decreased the probability of any alcohol use.
• No significant change was found in the average amount of alcohol consumed, conditional on drinking.
• The developed Bayesian methods efficiently estimated complex, nonstandard models.

Conclusions:

• New statistical models effectively handle semicontinuous outcomes in group therapy research.
• The intervention shows promise in reducing overall alcohol consumption initiation.
• Further research may be needed to address conditional drinking amounts in substance abuse interventions.