Visualizing a STAMINA-built model
The model tracks a population across the experiences a person with a substance use disorder may go through. It represents the disease as a finite set of possible circumstances an individual could be in at any given time. Each month, it projects how many people remain where they are and how many move to a new presentation of the disease, represented uniquely in the model. Each unique presentation has corresponding health outcomes and costs, which can be considered from a variety of perspectives depending on project scope.
The model repeats this calculation monthly over many years. Separate projections are run for the status quo and for each intervention, so that evidence of an intervention’s success can be translated into long-term outcomes.
Illustrative example: not calculated results
What a model produces: 1,000 people over their lifetime, with the intervention compared with the status quo
410
Years of life gained
560
Years of healthy life gained
2,100
Years in recovery gained
$21.6M
Cost avoided over a lifetime, in today’s dollars
Annual cost to all budgets, $M
Cost avoided by budget, lifetime, $M
Data related to long-term addiction outcomes is rare
Whether a hospital admission, a social benefit or an incarceration was driven by addiction is often unclear. Which indirect societal costs to include is often subject to debate. And because the course of the disease runs over years, it is harder to track, so there is less concrete data to project from.
The model addresses this by framing assumptions as three questions, answered by the people who know the population.
Published evidence is used to guide ranges where it exists. These figures are combined with the expertise and experience of the clinicians, people in recovery, agencies and budget owners who know the population. The model is flexible to test across ranges of values where opinions differ.
Disease presentations
What set of circumstances could someone with addiction be in?
In the projection, a person has exactly one disease presentation category at a time. These each carry a measure of health, a risk of death and a monthly cost that estimate averages of impacts across a cohort.
Transitions
Of 100 people with the same disease presentation this month, how are they expected to present next month?
Costs
What does each disease presentation cost, and which budgets and perspectives should be included?
Questions a decision-maker may ask
What an intervention is worth over the long term: health gained, deaths avoided and cost avoided, by year, in the budgets that carry them. The model is run once without the intervention and once with it; the difference is the result. A model of this kind supports a decision; it does not make one.
The model’s aim is not to predict the future, but rather to provide a consistent comparison of various potential interventions.
Yes. STAMINA methodology was built with audit and validation in mind. The same model inputs always give the same result, and any number can be traced back to the assumption behind it. Every input carries a source; a named model owner signs off that the model is fit for the decision it supports after independent review.
Yes. A model built for a state or province is reused by a county, or for another substance, by changing the assumptions that differ locally and keeping the rest.
STAMINA is a methodology to develop a model. It defines a structure and process for building addiction state-transition models in partnership with local understanding and diverse expertise. A model exists once a build has been validated against the specific purpose it is intended for.
Ways to engage
Anyone interested in a conversation about the approach, or about a specific application, is invited to reach out.
[email protected]