Bayesian Methods in HTA
Bayesian methods help update what we believe about a treatment's effect by combining existing evidence with new study data. They are useful for making decisions in health technology assessment when considering real-world impacts and uncertainties.
At a glance
Use when
When integrating diverse evidence sources, making predictions, conducting value-of-information analyses, designing adaptive studies, or supporting decision-making under uncertainty in HTA.
Avoid when
When there is strong resistance to subjective inputs, lack of defensible prior information, or when regulatory/policy frameworks require strictly frequentist interpretations.
Inputs
Prior distribution (e.g., expert opinion, historical data), likelihood function (from study data), optionally a loss function for decision-making.
Outputs
Posterior distribution, probabilistic statements about parameters, predictions, decision recommendations (if using decision theory).
How it works
Bayesian methods use Bayes's theorem to formally combine a prior distribution (representing pre-existing beliefs about a parameter, such as treatment effect) with the likelihood (from observed data) to produce a posterior distribution. This approach supports probabilistic inference, prediction, evidence synthesis, and decision-making under uncertainty. When integrated with loss functions, it enables full decision-theoretic frameworks for study design, monitoring, and policy decisions. Sensitivity analysis across priors is recommended due to the non-uniqueness and subjectivity of prior distributions.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation, Patient and Social Aspects
- Categories
- AppraisalEvidence SynthesisHeterogeneity
- Assumptions
- The prior distribution adequately represents existing knowledge; the likelihood model is correctly specified; Bayes's theorem applies to the updating process; any loss function reflects true preferences over outcomes.
- Strengths
- Explicitly incorporates external evidence and expert judgment,Provides direct probabilistic interpretation of results,Supports sequential learning and evidence synthesis,Enables decision-theoretic approaches to design and policy,Facilitates sensitivity analysis on assumptions
- Limitations
- Requires specification of prior distributions, which can be subjective,Subjectivity may be perceived as reducing scientific objectivity,Results can be sensitive to choice of prior,Computationally more complex than frequentist methods in some cases,Limited adoption and standardization in practice
- Also known as
- Bayesian analysis, Bayesian inference, Bayesian approach in HTA, Bayesian decision theory
Questions this answers
- › How should new evidence change our current beliefs about a technology's effectiveness?
- › What is the probability that a treatment is cost-effective given current evidence?
- › How can multiple sources of evidence be quantitatively combined in an HTA?
- › What are the expected benefits and risks of adopting a new technology under uncertainty?
- › How should clinical trials or studies be designed to maximize decision-making value?
- › How can subjective expert opinion be formally incorporated into HTA?
References & sources
Related methods
Similar by meaning
- Bayesian random-effects meta-analysis with empirical heterogeneity priors for HTA
- Patients' Perspectives in HTA: A Route to Robust Evidence and Fair Deliberation
- IMPACT-HTA multicriteria value framework
- Key Principles for the Improved Conduct of Health Technology Assessments for Resource Allocation Decisions
- HTA appraisal framework for rare disease treatments
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.