OBJECTIVES: In this study, we assess how temporal uncertainty evaluation can be applied in Cost-Effectiveness models to help decision making when assessing new health technologies

METHODS: A Partitioned-Survival Model with simulated data was used to estimate incremental NMB (iNMB) comparing a new drug against 8 simulated comparators. A Probabilistic Sensitivity Analysis with 1,000 iterations was run to obtain a measure of uncertainty in the outcome of interest. Outcomes for different time horizons were computed within each PSA iteration. Absolute growth in the mean outcome and in uncertainty were computed to get a sense of how uncertainty and mean estimates change with time. Growth in uncertainty was measured as growth in the sum of the absolute 95% quantiles of the PSA simulations. Stability in terms of convergence of the mean outcome and uncertainty were compared between treatments and within treatment. Implications in terms of decision making when several different comparators were discussed.

RESULTS: Mean iNMB and uncertainty growth behave differently within a comparison and between comparisons. Within a comparison, mean iNMB can stabilize several years earlier/later than iNMB uncertainty, which may depend on how symmetric the growth of uncertainty is around the mean iNMB. Between comparisons within a setting, data maturity of each comparator can determine how much and how fast uncertainty grows, and total uncertainty can vary substantially between comparators. This could affect decision makers with different preferences to risk and risk growth, and could have consequences when comparing technologies in other settings with very different levels of uncertainty.

CONCLUSIONS: Decision makers should take into account the size and stabilization of both the mean outcome of interest and of total uncertainty, especially in a context of multiple comparators with different data maturity, and where the decision maker has preferences in terms of risk and risk growth