Simplified Methods for Modelling Dependent Parameters in Health Economic Evaluations: A Tutorial

Area of Science:

• Health economics
• Biostatistics
• Mathematical modeling

Background:

• Model parameters in health economic evaluations are frequently interdependent.
• Existing methods for simulating multivariate normal (MVN) data and estimating Markov model transition probabilities under competing risks are complex for health economists.
• This work addresses the need for accessible techniques to manage dependent parameters in health economic modeling.

Purpose of the Study:

• To provide easily implementable methods for handling dependent parameters in health economic modeling.
• To illustrate these methods with practical examples and code in SAS and R.
• To extend routinely used techniques for broader applicability.

Main Methods:

• Presents analytical proofs and simplified methods for dependent parameter handling in health economic models.
• Demonstrates quantification of covariance and correlation coefficients from summary statistics.
• Illustrates generation of MVN distribution data and use of univariate normal distribution data for population heterogeneity.
• Introduces a conditional probability method for multiple state transitions within a single Markov model cycle.

Main Results:

• Successfully quantifies covariance and correlation coefficients using summary statistics.
• Demonstrates MVN data generation with physician visits and cost data examples.
• Shows effective use of univariate normal distribution data to capture population heterogeneity via regression models.
• Applies conditional probability method to one- and two-way state transitions in Markov models.

Conclusions:

• Proposes extensions to standard methods for handling dependent parameters.
• Offers simplified, easily applicable methods for health economic modelers of varying statistical expertise.
• Facilitates more robust and accurate health economic evaluations through improved parameter handling.