OBJECTIVES:

Uncertainty surrounding the decision to adopt a new intervention is formally considered by health technology assessment bodies. Conventionally probabilistic sensitivity analysis (PSA) is conducted to assess parameter uncertainty, with results presented in the form of cost-effectiveness acceptability curves (CEAC). Nonetheless CEACs are subject to limitations, including: ambiguity in interpretation and being unhelpful in characterizing the key factors contributing to uncertainty. Robust decision making (RDM) explores how the uncertainty around parameter values would affect the decisions and helps determine what would need to be true to discard one strategy in favour of another. We compare traditional methods in representing uncertainty with methods proposed by RDM using the decision to adopt a screening program for a cardiac disease as a case study application.

METHODS:

A Markov model was developed to evaluate the cost-effectiveness of a screening program for a cardiac disease as compared to no screening. Probability distributions were assigned to each parameter in the model, which were then sampled and recorded over 5,000 simulations to generate CEACs and scatterplots. Using PSA results we estimated the distribution of “regret” for each screening strategy versus no screening in terms of net monetary benefit across simulation runs. Decision-trees were then fit to the simulation data to identify the parameter values that need to hold for the screening strategy to be the optimal choice.

RESULTS:

The CEAC suggested that at a common willingness to pay threshold the probability of the screening program being cost-effective would be 50%. The mean % regret however was very low (0%-3%), highlighting that the likelihood of a much higher net benefit without screening would be low. Screening was found to perform particularly well in clusters where the cardiovascular event risks and the relative treatment effect were high.

CONCLUSIONS:

Use of RDM may improve understanding of the uncertainty surrounding decisions on health care interventions.