Integrating a Markov Chain Monte Carlo Convergence Diagnostics Output Analysis (CODA) Into State Transition Cost-Effectiveness Models in Microsoft Excel: A Stepwise Approach

OBJECTIVES: Probabilistic sensitivity analysis (PSA) allows the evaluation of parameter uncertainty in the outcomes of cost-effectiveness models. Relative effect measures with different levels of clinical response from Network Meta-Analyses (NMAs) are often used to derive transition probabilities and commonly have a correlated distribution. Using these parameters in the PSA of CEMs requires a joint estimation of random samples within a Monte Carlo simulation rather than relying on individual sampling using standard errors of point estimates of treatment effects. Although Health Technology Assessment Agencies (HTA) strongly recommend the use of PSA in economic models with reimbursement purposes, practical guidance on the implementation of a CODA is lacking. This abstract aims to provide guidance on integrating a CODA into state-transition CEMs in Microsoft Excel.

METHODS: A stepwise approach for implementing CODA in CEMs was developed using MS Excel. To identify implementation limitations, a review of previously built state-transition economic models in MS Excel was conducted. A more efficient technical programming in a sample semi-Markov model spreadsheet was developed using simulated efficacy data of an NMA. Six evaluated treatments and three health states was considered.

RESULTS: Decision modelers are encouraged to use outputs from a Bayesian NMA using one CODA value per PSA simulation for all health states and treatments considered in an economic model. Data from each Chain should be arranged so the chains are consecutive in a single column per treatment and health state. The total number of simulations is recommended to be calculated as the product of the number of simulations per treatment per chain and the number of chains. Programming based on INDEX-MATCHING statements is recommended to avoid errors in the use of data.

CONCLUSIONS: Given the paucity of practical guidance on integrating CODA into CEMs in MS Excel, a simplistic instructional approach would allow a more efficient implementation.