부산대학교 도서관

Background. Health economic evaluations of interventions against infectious diseases are commonly based on the predictions of compartmental models such as ordinary differential equation (ODE) systems and Markov models (MMs). In contrast to standard MMs, ODE systems of infectious diseases are commonly dynamic and account for the effects of herd immunity. This is crucial to prevent overestimation of infection prevalence. Despite their computational effort, ODE systems including whole distributions on model parameters are considered the ``gold standard'' in infectious disease modelling. However, the literature mainly contains ODE-based models which only include a predefined value on each model parameter and thus do not account for parameter uncertainty. As a consequence, probabilistic sensitivity analysis, a crucial component of health economic evaluations, cannot be conducted straightforwardly. Methodology. We present an approach to a dynamic MM under a Bayesian framework. The stochastic MM incorporates a probability distribution on each model parameter. We extend a static MM by incorporating the force of infection into the state allocation algorithm. The corresponding output is based on dynamic changes in population prevalence. In contrast to deterministic ODE-based models including a predefined value for each parameter, probabilistic sensitivity analysis can be conducted straightforwardly. The main motivation for our approach was to conduct a cost-effectiveness analysis of human papillomavirus vaccination. Results. We introduce a case study of a fictional sexually transmitted infection. By means of this example, we show that our methodology produces results which are comparable to the ``gold standard'' of an ODE system in a Bayesian framework. When applied to a cost-effectiveness analysis of human papillomavirus vaccination, our method indicates that universal vaccination (including both sexes) is cost-effective. A comparison of universal to female-only vaccination and cervical screening-only results in an Incremental Cost-Effectiveness Ratio (ICER) of €11,600 and €1,500, respectively. Conclusions. The dynamic Bayesian MM is suitable to include a high number of states and age cohorts, which are for example required in conclusive human papillomavirus modelling. In contrast to deterministic ODE systems, the setting is fully probabilistic at manageable computational effort.