CASCADE has four specific aims to be delivered over the five years of the project.
To test four theories that have been contended to explain alcohol use behaviors, using multiple secondary datasets in the US from 1979 to 2015.
We will perform quantitative modeling to operationalise four theories of alcohol use (AU) behaviors, encompassing individual decision-making, social interaction and social context.
These theories – Maturing Out, Social Contagion, Cohort-Specific Exposure, and Rational Choice – have been debated extensively in the alcohol research and policy communities over the last 30 years.
Agent-based models (ABMs) will be used to scale each theory to the populations of five US states (California – CA, New York – NY, Tennessee - TN, Minnesota - MN and Texas – TX), the wider population of the US.
The ability of the models to reproduce population-level measures of AU from 1979 to 2015 will be tested. Uncertainty in model structures and parameters will be handled rigorously using a Bayesian model calibration approach.
- Hypothesis 1: Change in population-level measures of AU over time during adolescence and early adulthood can be explained by personality development, coupled with engagement in adult social roles.
- Hypothesis 2: Change in population-level measures of AU over time can be explained by the modeling and control influences of friends and family.
- Hypothesis 3: Differences in population-level measures AU are explained by a cycling of restrictive and permissive norms between generations, coupled with changes in female education and labor market participation.
- Hypothesis 4: Change in population-level measures of AU over time can be explained by changes in the affordability and availability of alcohol to consumers.
To develop a systems-based, mechanism-driven, model of alcohol use behaviors to predict population-level measures of alcohol use in the US from 2016 to 2025.
Based on our findings in Specific Aim I, we will develop a grammar that integrates mechanisms for the individual theories and enables new systems-based accounts to be generated.
Searching over systems-based structures will be undertaken using genetic programming in a Bayesian model calibration framework for period 1979-1998.
The usefulness of the models for policy analysis will be assessed in terms of the tightness of 95% Bayesian confidence intervals for 1999-2015 model outputs to the true population-level trajectories for the US as a whole, CA, NY, TN, MN and TX separately.
The models will be used to predict AU between 2016 and 2025.
- Hypothesis 5 (exploratory): A calibrated systems-based model will produce estimates for population-level measures of AU over time that validate against observed data.
- Hypothesis 6 (exploratory): A calibrated and validated systems-based model will be useful for predicting population-level measures of AU over time.
To combine the systems-based models with the latest epidemiological models relating population-level AU and risk of harmful outcomes to predict the burden of liver cirrhosis and alcohol poisonings in the US from 2016 to 2025.
The models in Specific Aim II will produce outputs that form the main exposure variables for epidemiological models of alcohol-related harms, enabling estimated changes in AU behaviors to flow through into estimated changes in alcohol-related harm.
To perform this integration, we will build new epidemiological models for the US. Model calibration and validation activities will be performed for alcohol poisoning and liver cirrhosis, for which AU dominates as a risk factor.
Calibration will be performed using mortality and morbidity data for 1979-1998, with model outputs validated for 1999-2015. Predictions will be made for period 2016-2025.
- Hypothesis 7 (exploratory): A calibrated and validated systems-based model, when combined with a calibrated epidemiological model, produces morbidity and mortality estimates for liver cirrhosis and alcohol poisoning that validate against observed data.
- Hypothesis 8 (exploratory): A calibrated and validated systems-based model, when combined with a calibrated and validated epidemiological model, will be useful for prediction of morbidity and mortality rates for liver cirrhosis and alcohol poisoning.
To use the new models to retrospectively evaluate minimum legal drinking age and taxation changes 1979-2015 and to appraise prospective policies for minimum pricing and a universal program of screening and brief intervention in the US from 2016 to 2025.
We will demonstrate the usefulness of the proposed approach by building additional mechanisms that link both retrospective and prospective policies and interventions to inputs and parameters in the systems-based model.
We will evaluate the impact of minimum legal drinking age laws (which affect norms and controls) in the US and taxation changes (which affect affordability) in CA, NY, TX, TN and MN, between 1979 and 2015.
These mechanisms will then be exploited prospectively to examine the impact of a program of screening and brief intervention in primary care and introduction of minimum pricing, assuming hypothetical implementation in 2016.
We will also develop open-source software and make this available for all stakeholders in the alcohol debate to use.
- Hypothesis 9 (exploratory): A calibrated and validated systems-model, when combined with a calibrated and validated epidemiological model, will be useful for evaluation of existing interventions.