OBJECTIVES: Probabilistic sensitivity analysis (PSA) is used to characterise uncertainty in cost-effectiveness models. A model was developed using R and Shiny to explore the impact of different parameter correlation structures on PSA outputs.
METHODS: A Markov model was built in R to compare a hypothetical treatment and comparator. Three options were built into the model: no correlation (inputs varied independently); part correlation (correlation within but not between costs, utilities and transition matrices); and full correlation (correlation between all inputs). A Shiny interface allowed users to explore the impact of the correlation options with different model parameters. Features of the Shiny model included preloading base case results, running the Markov model with an arbitrary number of health states and costs, and displaying results for different subsets of correlation options. A scenario analysis was included in the Shiny model to determine the circumstances in which correlation had the largest impact by varying the treatment cost as a proxy for the ICER.
RESULTS: While the ICER was comparable across all correlation options, the likelihood of cost-effectiveness differed substantially from 61% to 93%. In all scenarios, the ‘no correlation’ option displayed the most certain likelihood (closest to either 0 or 1) of cost-effectiveness, while the least certain was produced by the full correlation option. Counterintuitively, correlating inputs increased uncertainty because it allowed for a greater number of ‘extreme’ scenarios to be generated, whereas allowing independent generation of large numbers of inputs tends to lead to a ‘cancelling out’ effect. This effect was most pronounced when the ICER is moderately close to the willingness-to-pay threshold.
CONCLUSION: This analysis demonstrates that input correlation can have a substantial impact on the level of certainty in model outputs, and by ignoring this, the model may be over- or under-stating the true level of confidence.