| Summary: | During 2009 Mexico experienced an A(H1N1) pandemic with a rapid increase in the number of observed cases. To reduce transmission, the Mexican Government purchased 30 million A(H1N1) vaccines that were under production. There was considerable uncertainty in whether this large expenditure represented value for money. The primary aim of this thesis is to estimate the cost-effectiveness (CE) of vaccination programmes using the information known at the time of the decision. This objective utilised an ordinary differential equations (ODE) approach calibrated via a Markov chain Monte Carlo (MCMC) algorithm. Additional objectives included: assessing whether the observed number of reported cases could also be replicated using discrete event simulation (DES) methodology and documenting the type and prevalence of models used to estimate the CE of an infectious disease vaccine intervention. There was inherent uncertainty regarding the anticipated CE of the vaccine at the time the decision to purchase was made, primarily as no definitive value for the reporting rate (RR), the number of cases that come to clinical attention could be estimated. Three RR values, for the 0-15-year-age group, were explored (0.75, 0.01 and 0.001) with RR in other age groups being estimated through the MCMC calibration. In two of the RRs (0.75 and 0.01), the vaccination programme was cost-effective, for the assumed threshold value for Mexico ($110,000 MXN per QALY gained). In contrast, when a low RR was assumed (0.001) the vaccine was dominated, being more expensive and producing less health due to the adverse events of the vaccine. These results were robust to most sensitivity analyses. When a pessimistic scenario was applied (low vaccine effectiveness, longer time required to apply the vaccines -an additional 55 days compared with the base case-, and vaccine arriving 31 days later) did the vaccine interventions become non-CE assuming an RR of 0.01. For the 0.001 RR scenario, when longer times of latent and infectious periods were assumed the vaccine became CE. As the Mexican Government anticipated an RR of approximately 0.09, it was concluded that the decision to purchase the vaccines would have been considered a cost-effective use of resources. The DES model was found to be an unsuitable approach to predict the pandemic as the calibration attempt was unsuccessful and running times were lengthy. There are clear advantages in using an ODE approach rather than a DES approach in a pandemic setting. The analysis of the papers identified in the literature review has indicated most of the published literature are based on static approaches, although the use of dynamic models has increased over time. Analyses indicated that the year of publication was a significant predictor for the use of dynamic models. The decision to construct a dynamic, rather than a static model, however, was neither influenced by the GDP per capita of the effected country or the location of the lead author.
|
|---|
