MICROSIMULATION MODEL CALIBRATION USING INCREMENTAL MIXTURE APPROXIMATE BAYESIAN COMPUTATION.

Area of Science:

• Computational epidemiology
• Health policy modeling
• Statistical inference

Background:

• Microsimulation models (MSMs) are crucial for policy analysis, simulating individual health trajectories and policy impacts.
• Model calibration, a complex process, is essential for ensuring MSMs accurately predict real-world outcomes.
• Existing calibration methods can be computationally intensive and may not fully capture parameter uncertainty.

Purpose of the Study:

• To introduce Incremental Mixture Approximate Bayesian Computation (IMABC), a novel method for calibrating MSMs.
• To provide a robust framework for estimating the posterior distribution of MSM parameters.
• To enhance the reliability of policy predictions derived from MSMs.

Main Methods:

• IMABC employs a two-stage approach: an initial rejection-based Approximate Bayesian Computation (ABC) step followed by iterative refinement using a mixture of normal distributions.
• The method adaptively samples the parameter space, accepting parameter sets that yield simulated outcomes close to observed targets.
• Posterior estimates are derived by weighting accepted samples to correct for the adaptive sampling strategy.

Main Results:

• IMABC successfully calibrated the CRC-SPIN 2.0 microsimulation model for colorectal cancer (CRC).
• The method generated a simulated sample from the posterior distribution of model parameters.
• The calibrated model demonstrated improved accuracy in predicting CRC-related outcomes and the impact of screening policies.

Conclusions:

• IMABC offers an efficient and effective approach for calibrating complex MSMs.
• This method enhances the accuracy and reliability of policy-relevant predictions from health simulation models.
• The successful application to CRC screening highlights IMABC's potential for informing public health policy.