Methods: We conducted a systematic literature review (Medline and Embase, MeSH "influenza" AND "cost and cost analysis/ or models, biological/ or models, economic/ or models, theoretical/ or models, immunological/ or models, statistical") of influenza models published to date. We evaluated the level of clinical detail, inclusion of herd immunity effects, the model outputs generated, the validation techniques used, the adherence to economic evaluation guidelines, the estimation of cost-effectiveness ratios, and the sensitivity of the cost-effectiveness ratio to important variables.

Results: Twenty four dynamic models and seventeen decision analytic cohort models were included. Dynamic simulation models typically model one influenza season for a defined community and focus on factors related to disease transmission, such as cross-immunity, the circulation of more than one strain, mixing patterns and geographical spread. Outcomes are measured in terms of number of infected cases and mortality. Disease outcomes, such as complications, hospitalizations, deaths, and quality of life are not considered. If investigated, the impact of intervention strategies is mostly applied to the entire population. None of the studies were economic evaluations. In contrast, decision-analytic cohort models of influenza were all economic evaluations, and considered vaccination and antiviral medications. Most models were for specific population groups and usually compare either prevention or treatment to 'do nothing'. Few cohort models have compared prevention with treatment strategies. Even the most well-executed decision-analytic models do not integrate findings into a broader policy and budgetary context, where costs and benefits are considered from a population perspective, nor have combinations of prophylactic and therapeutic approaches been subjected to comprehensive evaluation. Disease transmission effects such as herd immunity are generally not included, potentially underestimating the cost-effectiveness of interventions.

Conclusions: Two research streams in influenza modeling appear to have developed independently with little overlap. While each of the approaches has made valuable contributions to understanding influenza dynamics, health outcomes and costs, we argue that, in order to address questions of policy relevance, it will be necessary to converge the two approaches and to expand dynamic simulation models using decision-analytic tools to evaluate the cost-effectiveness and health outcomes of various influenza management options.