Model averaging for structural uncertainty in health economic decision models (Jackson et al.)
This method helps account for uncertainty in health economic models by combining results from multiple plausible model structures, rather than relying on a single model. It improves decision-making by considering different ways a disease or treatment might work.
At a glance
Use when
There is uncertainty about the correct model structure in a health economic evaluation, such as which covariates to include or how to model disease progression
Avoid when
Only one clear, well-justified model structure exists, or when computational resources are limited and model space is large
Inputs
Set of competing health economic model structures (e.g., different covariate models), data for model fitting, and a criterion for weighting (e.g., AIC, BIC)
Outputs
Weighted average of model predictions, quantification of uncertainty due to model structure, improved cost-effectiveness estimates
How it works
The method addresses structural uncertainty in health economic decision models by applying model averaging techniques, where competing model structures (e.g., different covariate specifications) are combined using weights derived from model assessment criteria. The paper compares asymptotically consistent criteria like the Bayesian Information Criterion (BIC) with predictive ability measures like Akaike's Information Criterion (AIC), advocating for the latter due to its better performance in predicting complex health processes such as disease progression and treatment response. The approach is demonstrated in a case study comparing two surgical techniques for abdominal aortic aneurysm repair.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation
- Assumptions
- The set of candidate models includes plausible representations of the underlying process; model weights reflect relative plausibility or predictive performance; models are fitted to the same data and outcome
- Strengths
- Formally accounts for structural uncertainty, improves predictive accuracy, enhances transparency in model choice, supports more robust health technology assessments
- Limitations
- Requires specification of all plausible models, computationally intensive, results depend on the choice and quality of candidate models and weighting criteria
- Also known as
- Model averaging, Structural uncertainty model averaging, Jackson et al. model averaging method
Questions this answers
- › How can structural uncertainty in health economic models be formally addressed?
- › What are the differences between using BIC and AIC for weighting models in averaging?
- › Why is predictive ability more important than model consistency in health economic modeling?
- › How does model averaging improve decision robustness in cost-effectiveness analysis?
- › What are the practical implications of structural uncertainty in comparing surgical techniques?
- › How can multiple plausible model structures be combined in a single analysis?
References & sources
Similar by meaning
- Framework for Addressing Structural Uncertainty in Decision Models (Bojke et al.)
- Model Parameter Estimation and Uncertainty
- Iterative Decision-Making Framework for Model-Based Decision Making
- Value of Information Analysis in Models to Inform Health Policy (Kunst-Heath review)
- A Framework for Developing the Structure of Public Health Economic Models
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.