Josephine van Dongen

Chapter 3 64 Cost-effectiveness and risk-benefit analyses The model was built in Microsoft Excel with the add-in software @Risk, version 7.5 (Palisade). For all three vaccination strategies (“no vaccination,” “targeted vaccination,” and “universal vaccination”), the model estimates the number of rotavirus cases in the population, GP visits, hospitalizations, rotavirus-related deaths, QALYs, and life years.The model fur ther estimates the number of vaccine-induced IS cases and associated QALYs. Net costs (i.e., net social costs and net healthcare costs), life years gained (LYG), and QALYs gained were calculated by summing all costs, life years, and QALYs over the 20-year time horizon. For each simulation, 5000 runs were conducted using Monte Carlo sampling, accounting for the uncer tainty of the model parameters ( Table 1 ). The incremental cost-effectiveness ratios (ICERs) were calculated by dividing the net cost differences between each vaccination strategy, compared to no vaccination, by either QALYs gained or LYG. Our primary perspective was societal (i.e., including non-healthcare costs such as caregiver work loss), and the healthcare payer perspective was included in the sensitivity analysis. Costs are expressed in 2016 euros. A discount rate of 3% was used for both costs and effects 39 . Risk-benefit ratios were calculated by dividing the number of severe outcomes aver ted by vaccination, which included rotavirus hospitalizations or rotavirus fatal cases, by (1) the estimated number of vaccine-induced IS cases and (2) the estimated number of vaccine- induced complicated IS cases.The calculated ratios were used to obtain the benefit per vaccine- induced IS case and per vaccine-induced complicated IS case, respectively. Risk-benefit ratios were calculated both for the total population and for each risk group, since the risk of severe outcomes due to rotavirus differs between children with and without medical risk conditions. Fig. 2 Rotavirus outcome tree and different healthcare paths considered in model. With permission from Bruijning-Verhagen et al. 11