Introducing the Parameter Outcome Overview Plot: A New Addition to the Sensitivity Analysis Armory for Economic Models
Author(s)
Conor Teljeur, PhD1, Susan Ahern, MSc, PhD2, Mairin Ryan, PhD1.
1Health Information and Quality Authority, Dublin, Ireland, 2Health Information & Quality Authority of Ireland, Cork, Ireland.
1Health Information and Quality Authority, Dublin, Ireland, 2Health Information & Quality Authority of Ireland, Cork, Ireland.
OBJECTIVES: Most economic models are probabilistic, requiring substantial sensitivity analysis to explore decision uncertainty. Sensitivity analyses are frequently deterministic, potentially masking important uncertainty. We present the Parameter-Outcome Overview Plot - analogous to the cost-effectiveness acceptability curve (CEAC) - supporting one-way probabilistic sensitivity analysis and also applicable to threshold analyses.
METHODS: Instead of varying the willingness-to-pay (WTP) threshold as per a CEAC, a single parameter is varied using a moving window approach across a parameter’s distribution to identify simulations where the parameter of interest is within 2.5 percentiles of the window midpoint. Each strategy’s probability of being the most cost-effective strategy (at a WTP of €45,000/QALY) at the window mid-point was calculated using the simulations within the window. The method was applied to four parameters from an economic evaluation of childhood varicella vaccination assessing four strategies: no vaccination, single dose, and two dose strategies (short and long).
RESULTS: The probability of the two-dose long interval strategy being the most cost-effective strategy was negatively correlated with the cost of the vaccine and of administering the vaccine. The switch to a single dose strategy being most cost-effective occurred at the mean value for the parameters. For the probability of hospitalisation, changing the parameter value had a limited impact on which strategy was most cost-effective. For the disutility associated with varicella infection, low parameter values were associated with a greater likelihood of a single dose strategy being the most cost-effective. These relationships were not evident from typical deterministic one-way sensitivity analyses.
CONCLUSIONS: By using a visual language familiar to decision makers through the use of CEACs, we have developed a simple tool to aid presentation of one-way sensitivity and threshold analyses. Importantly, it has the advantage of expressing uncertainty and considering all interventions simultaneously. As it is estimated from the probabilistic model outputs, the computational burden is minimal.
METHODS: Instead of varying the willingness-to-pay (WTP) threshold as per a CEAC, a single parameter is varied using a moving window approach across a parameter’s distribution to identify simulations where the parameter of interest is within 2.5 percentiles of the window midpoint. Each strategy’s probability of being the most cost-effective strategy (at a WTP of €45,000/QALY) at the window mid-point was calculated using the simulations within the window. The method was applied to four parameters from an economic evaluation of childhood varicella vaccination assessing four strategies: no vaccination, single dose, and two dose strategies (short and long).
RESULTS: The probability of the two-dose long interval strategy being the most cost-effective strategy was negatively correlated with the cost of the vaccine and of administering the vaccine. The switch to a single dose strategy being most cost-effective occurred at the mean value for the parameters. For the probability of hospitalisation, changing the parameter value had a limited impact on which strategy was most cost-effective. For the disutility associated with varicella infection, low parameter values were associated with a greater likelihood of a single dose strategy being the most cost-effective. These relationships were not evident from typical deterministic one-way sensitivity analyses.
CONCLUSIONS: By using a visual language familiar to decision makers through the use of CEACs, we have developed a simple tool to aid presentation of one-way sensitivity and threshold analyses. Importantly, it has the advantage of expressing uncertainty and considering all interventions simultaneously. As it is estimated from the probabilistic model outputs, the computational burden is minimal.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
EE560
Topic
Economic Evaluation, Health Technology Assessment, Methodological & Statistical Research
Disease
No Additional Disease & Conditions/Specialized Treatment Areas