Assessing uncertainty in microsimulation modelling with application to cancer screening interventions

Area of Science:

• Computational epidemiology
• Health economics modeling
• Biostatistics

Background:

• Microsimulation is increasingly used for complex health interventions, particularly chronic diseases like cancer.
• Parameter uncertainty is a significant challenge in microsimulation, often unquantified in results.
• Existing methods lack robust approaches to address uncertainty in microsimulation model parameters.

Purpose of the Study:

• To introduce a Bayesian framework for quantifying parameter uncertainty in microsimulation models.
• To develop a method for integrating parameter uncertainty into the analysis of simulation outcomes.
• To demonstrate the application of this method to prostate cancer screening interventions.

Main Methods:

• Designed a simulation experiment for comprehensive parameter space coverage.
• Developed a response surface model to approximate simulation outcomes based on parameters.
• Combined response surface with Bayesian parameter distributions to quantify outcome variability.

Main Results:

• The proposed Bayesian approach effectively quantifies uncertainty in microsimulation outcomes.
• Demonstrated the method's utility in assessing prostate specific antigen (PSA) screening's impact on prostate cancer mortality.
• The response surface approach provides a computationally efficient way to explore parameter uncertainty.

Conclusions:

• The Bayesian framework offers a robust solution for uncertainty quantification in microsimulation.
• This method enhances the reliability and interpretability of simulation studies for health interventions.
• Accurate quantification of parameter uncertainty is crucial for evidence-based decision-making in public health policy.