Josephine van Dongen

Chapter 3 72 Our study has several limitations. The model input was largely based on epidemiological data as well as healthcare and non-healthcare cost estimates from the Netherlands. Differences in rotavirus endemicity, population demographics, caregiver employment, and cost prices may limit the generalizability of our findings to other high-income settings. However, we have performed extensive sensitivity analyses to evaluate the robustness of our ICER estimates and the most influential parameters. As targeted vaccination remained cost-saving under all scenarios tested, we are confident that this strategy will be cost-saving to other high-income settings.The ICER for universal vaccination, however, may be more variable, and for some high-income countries it may be better represented by one of the alternative scenarios from our sensitivity analysis. Our model did not include dynamic simulation of herd effects following introduction of universal vaccination. Given the unusual pre-vaccination biennial rotavirus pattern in the Netherlands, observations on herd-protection levels from other countries may not be representative. Therefore, we chose to lower the herd-protection estimates extracted from studies in Europe and Nor th America by 50% for our analysis.We considered this the most likely scenario, but the accuracy of these adjusted estimates remains uncer tain. Our sensitivity analysis showed that a 50% change in herd effects from baseline would result in a 15% change in ICER. Another limitation of our static, rather than a dynamic model, is that we could not explore how universal rotavirus vaccination affects the timing and pattern of rotavirus epidemic peaks. Sudden spikes in incidence put additional pressure on hospital capacity, and this may be especially relevant if these coincide with circulation of respiratory viruses in winter months. The periodicity and timing of rotavirus epidemics may therefore be impor tant for bed capacity planning. Available rotavirus dynamic models so far suggest that high-coverage rotavirus vaccination in temperate climates results in a biannual pattern and a shift of the epidemic peak to April/May 45 . Finally, it is currently uncer tain whether the biennial rotavirus pattern in the Netherlands will be sustained in future years. If conditions affecting rotavirus epidemiology change in the future, disease levels could return to those pre-2014. Naturally, this would change the ICERs for the different vaccination strategies analyzed and the threshold for cost-saving vaccine prices. Conclusion While universal infant rotavirus vaccination results in the highest reductions in the population burden of rotavirus, targeted vaccination should be considered as a cost-saving alternative with the most favorable risk-benefit ratio for high-income settings where universal implementation is unfeasible for reasons of budget restrictions, low rotavirus endemicity, and/or public acceptance.