Analytical approach:
The analysis was based on a Markov model, with a lifetime horizon. The authors stated that it took the perspective of society.
Effectiveness data:
Meta-analyses were used as sources wherever possible. The epidemiological data were from US mortality tables and other local databases. The vaccination rate (coverage) was the key input for the model and the programme was assumed to increase this rate to a maximum of 70% for elderly African-American and Hispanic people, to match the vaccination rate for elderly Caucasian people.
Monetary benefit and utility valuations:
The utility values were from published sources and were applied to influenza-related conditions, for the duration of the 10-year programme.
Measure of benefit:
Quality-adjusted life-years (QALYs) were the summary benefit measure and they were discounted at an annual rate of 3%.
Cost data:
The economic analysis included the costs of vaccination (dose and administration), programme promotion, over-the-counter medications, time spent seeking and receiving vaccination and care, out-patient care, antiviral drugs, hospitalisation, and deaths. The cost of time was based on average hourly wages. The drug costs were estimated using average wholesale prices. Other costs were from published sources that were based on Centers for Medicare and Medicaid Services fee schedules and Healthcare Cost and Utilization Project data. These costs were assumed to have been constant over the 10 years of the programme, except for programme promotion costs, which grew at an annual rate of 3%. All costs were in US $. A 3% annual discount rate was applied and the price year was 2009.
Analysis of uncertainty:
One- and two-way sensitivity analyses were carried out on all the inputs for the model, using published ranges of values where possible. The perspective was restricted to the third-party payer for an alternative scenario, which included only the direct medical costs incurred by the patients. A probabilistic sensitivity analysis was carried out, in which all the parameters were simultaneously varied over their assumed distributions. Gamma distributions were assumed for the cost and duration parameters, and beta distributions were used for the probability and utility variables.