Target population

The model includes cohorts of children and adults aged 6 months and older stratified by age and risk of complications. Age groups were: 6–23 month, 24–35 month (2 yrs), 3–4 years, 5–11 years, 12–17 years, 18–49 years, 50–64 years, and 65 years and older. Each age group was then divided into two risk-based groups, higher risk and lower risk, except for age 65 years and older who are all assumed to be at higher risk for complications. Higher risk groups were defined using conditions identified by CDC as placing individuals at higher risk for medical complications of influenza illness.[6]

Natural history of influenza

Probabilities of hospitalization and death following pH1N1 illness were derived from emerging data for pH1N1 in spring/summer 2009.[11](Table 1) Input values for probabilities of other influenza-related outcomes were based on previously established estimates for seasonal influenza, such as probability of seeking medical attention during an episode of influenza illness, other complications treated on an outpatient setting, and long-term sequelae following hospitalization.[12], [13] The range of possible values for probability of pH1N1 influenza illness during a single season was varied from 5% to 30% to reflect the limited data on possible overall pH1N1 illness attack rates with intermediate probabilities of 7.6%, 15%, and 21%. One intermediate probability, 21%, represents the most recent estimate for the pH1N1 pandemic derived from Shrestha et al (2011).[14] The probability of 7.6% represents an average non-pandemic influenza season.

Costs of influenza-related health events

Costs included direct medical costs for influenza events, including physician visits, over-the-counter and prescription medications, diagnostic tests, hospitalizations, and long-term sequelae, based on established costs for seasonal influenza.[12], [13], [15], [16] Direct medical costs were adjusted to 2009 dollars using the medical component of the Consumer Price Index.[17]

Vaccination assumptions

Pandemic H1N1 vaccine effectiveness was assumed to have similar effectiveness as for seasonal influenza vaccine based on preliminary studies of 2009 pH1N1 vaccine immunogenicity which shows immune responses comparable to seasonal influenza vaccine.[18], [19], [20], [21] Base case (range) vaccine effectiveness was assumed to be 69% (40%–90%) for individuals aged 6 months to 17 years, 69% (30%–90%) for those 18 to 64 years and 60% (30%–90%) for those 65 years and older. (Table 2) Vaccination-related adverse events were assumed to be consistent with rates for seasonal influenza vaccine. Injection site reactions, systemic reactions, anaphylaxis, and Guillain Barré syndrome varied by age and are listed in Table S1. Incidence of Guillain Barré syndrome was based on data from seasonal influenza vaccine[22] and varied in sensitivity analyses to reflect rates observed in the 1976 swine flu vaccination program.[23], [24], [25], [26], [27], [28], [29], [30], [31], [32].

Vaccination-related costs included cost of pH1N1 vaccine doses, administration fees, and time costs (for parents or patients, depending on the age of the vaccinee).(Tables 2, S1, S2) We assumed full vaccination required 2 doses for individuals aged 6 months to 9 years and 1 dose for individuals aged 10 years and older. Administration costs varied by setting: lower for mass vaccination clinic and higher for the physician office setting.[33] For the physician office setting, the proportion of persons vaccinated at an existing or vaccination-specific visit varied by age. Assumptions for mid-range costs use data on vaccination setting by age from seasonal influenza vaccination for adults and assume a mix of vaccination settings.[34] The proportion of individuals receiving vaccination in a mass vaccination or physician office setting varied by age and risk group. Assumptions for mid-range costs for children also vary by age. 75% of school-aged children were assumed to be vaccinated in a school-located setting. Children younger than 5 years were assumed to be vaccinated in the physician office setting. Costs of adverse events were based on costs of adverse events associated with inactivated influenza vaccine.[15] Time costs for vaccination time for adults and parents of vaccinated children were also included.

Health outcomes and quality adjustments

The primary health outcome for the analysis is the quality-adjusted life year (QALY). A QALY attempts to measure a patient's physical health and well being including, among other factors, the ability to engage in “normal,” everyday activities. QALYs lost to a disease or condition, therefore, measure the overall reduction in a patient's well being, or health-related quality of life, due to an episode of disease and its consequences (which may last a life time). We obtained the loss in QALYs associated with each influenza-related health event from published studies and primary data.[35], [36], [37] In these studies, respondents were asked how much of their own lifetime they would be willing to trade in order to avoid a case of influenza-related illness or a vaccination-related adverse event (i.e., a time-tradeoff valuation). (Tables S1, S2)

Cost-effectiveness Analysis

The main endpoint for the study was the incremental cost-effectiveness ratio (ICER) calculated by dividing the net costs by the net health benefits (as measured via QALYs) for vaccination compared to no vaccination. An intervention is defined as cost-saving if the intervention decreases dollar costs and also results in an increase in QALYs. An intervention is defined as cost-effective if it results in an increase in both costs and QALYs and the resulting ratio is less than a determined threshold.[38] The primary analysis explored a range of values for pH1N1 illness rates and vaccination costs due to uncertainty regarding the severity of the pH1N1 influenza season and preferred vaccination settings. All costs and health effects lasting more than 1 year were discounted at 3% per year.

Considerable uncertainty exists regarding the most likely setting for pH1N1 vaccination (i.e., mass vaccination clinics compared with physician offices). Therefore the primary analysis includes a mass vaccination setting, a physician office setting, and a “mid-range” cost setting that assumes a proportion of individuals is vaccinated in each setting. This proportion varies by age. The primary analysis assumes that some individuals will receive pH1N1 vaccination at an existing physician visit which would be associated with lower administration costs. An existing visit is defined as a visit previously-scheduled for a purpose other than pH1N1 vaccination. In the sensitivity analysis, we include a more conservative scenario in which all individuals are vaccinated in the physician office setting and require a vaccine-specific visit.

Sensitivity analyses explored changes in key variables including the number of doses required for vaccination, costs of vaccination, vaccination-related adverse events, and influenza-related hospitalization rates. A sensitivity analysis explored the change in cost-effectiveness ratios if only one dose were required for children younger than 10 years. Some studies have suggested higher costs of influenza vaccination in the physician office setting[39], [40], therefore, we also conducted a sensitivity analysis that assumed higher administration costs. Due to concerns that swine influenza vaccination in 1976–1977 may have been associated with Guillain-Barré syndrome, sensitivity analyses included varying the probability of Guillain-Barré syndrome following vaccination over a wide range of possible values, from rates that may occur with seasonal influenza vaccine (1 per million) to rates observed with 1976 swine influenza vaccine (1 per 100,000).

The primary analysis assumed timely pH1N1 vaccination prior to the start of the outbreak. A scenario analysis evaluated initiation of vaccination after the start of a hypothetical influenza season. Each week after the start of the season that vaccination was initiated was assumed to reduce the protective effect of the vaccine according to the expected distribution of cases over a hypothetical 16-week influenza season assuming peak at 9 weeks and 70% of cases occur between weeks 7–10.(Table S3) [41]

Policy Implications

This analysis differs from dynamic models, which model the indirect effects of vaccination. One such model suggested that vaccinating school-aged children and adults between the ages of 30 and 40 would be most cost-effective.[7] Dynamic models simulate the transmission of infection among individuals and estimate the reduction in infections as a result of reduced transmission among unvaccinated age groups (e.g., vaccinating school children will reduce transmission to individuals of other ages and result in indirect effects of reducing illness and deaths in infants and the elderly). The current analysis intentionally excludes possible indirect effects of vaccination and restricts the primary analysis to the costs and health benefits to the vaccinated individual. Required coverage rates to generate herd effects for pH1N1 are unknown and substantial uncertainty exists for the required minimum coverage level for seasonal influenza[49], [50], [51]. Even if higher coverage rates are attained or pH1N1 vaccination compared to seasonal influenza vaccination, the effect of herd immunity is uncertain.[49], [50], [51] Another rationale for restricting the analysis to individual-level benefits relates to societal preferences. A policy of vaccinating school children to prevent illness in other age groups assumes that public preferences are consistent with trading off the health and well-being of school-aged children (in the form of risk for vaccination-related adverse events) to protect people in other age groups. Evidence suggesting that societal preferences may be more consistent with prioritizing child health over adult health could clearly support the vaccination of children if the expected benefits to the child outweighed the potential costs and risks but the decision to vaccinate a child may only consider benefits and risks to the vaccinated child.[37] Given that the inclusion of herd effects would result in an increase in the health benefits associated with vaccination, an extension of the current analysis to include herd effects would result in more favorable cost-effectiveness results.

The costs of vaccination have a substantial impact on cost-effectiveness results. Uncertainty exists as to the proportion of individuals likely to get vaccinated in each type of setting. Since costs of vaccination will vary with setting, these are key assumptions for the analysis. For higher attack rate scenarios, cost-effectiveness ratios vary about 10% across settings, however, if most people are vaccinated in the physician office setting and require a vaccine-specific visit, vaccination costs will be higher and associated cost-effectiveness ratios will be less favorable. On the other hand, if only one dose is required for vaccinating children, the associated costs will be lower and cost-effectiveness ratios will be more favorable. Recent data from the 2009–2010 pH1N1 vaccination season indicate that a mix of settings was used for vaccination against pH1N1 influenza.[52] Results of sensitivity analysis varying the costs of vaccination could provide useful information for future decision making given the sensitivity of the results to this input parameter.

Initiation of vaccination after the start of the season will affect the cost-effectiveness of vaccination depending on the timing of availability relative to the start, duration, and intensity of influenza activity in a community. For children who require two doses, vaccination may not be cost-effective if vaccine is delivered such that full protection is not achieved until after the 8^th week (the peak) of a hypothetical influenza season. For adults and children requiring only one dose, results are similar but the timing of vaccination required will differ since only one dose is assumed to be required for full protection. Cost-effectiveness results would differ if additional pandemic waves caused by a similar virus were to occur within the same vaccination year, or if vaccination later during a single pandemic wave provided some beneficial immunologic priming for subsequent vaccination against a drifted influenza virus, or if the pattern of disease during the pandemic wave does not conform to our model of a hypothetical season.

Comparison to Other Economic Studies of pH1N1 Vaccination

Estimates of the economic impact of vaccination are available for the US and other countries. In the US, Beigi et al. evaluate the economic value of vaccinating pregnant women, a very high risk group not included in our analysis, and report favorable cost-effectiveness ratios for maternal influenza vaccination.[53] Khazeni et al. (2009) evaluate the cost-effectiveness of vaccination against pandemic influenza A (H1N1) for a major US metropolitan city and early vaccination to be cost-saving.[54] Lee et al. (2010) estimate averted lost productivity costs for an employee population but do not report results using an economic metric such as an incremental cost-effectiveness ratio making it difficult to compare these results with those from our study.[55] Sander et al (2010) find the cost-effectiveness of a mass immunization program for pandemic H1N1 to be favorable but this study does not evaluate the cost-effectiveness of individual age and risk groups.[56] Baguelin et al. (2010) use a dynamic model to evaluate vaccination against pandemic influenza A(H1N1) in England.[57] Our analysis complements other published analyses both for the US and abroad by evaluating the cost-effectiveness of individual age and risk groups and considering explicitly the effects of delay in vaccination on cost-effectiveness. While it is difficult to directly compare cost-effectiveness analyses across countries due to differences in costs of services and level of intensity of care, results from other countries were consistent with ours in estimating favorable cost-effectiveness for mass vaccination in England and Canada. Our analysis provides additional information to the previously published studies by providing incremental cost-effectiveness ratios for separate age and risk groups relevant to the US setting. In addition, we explored sensitivity analyses relevant to the US decision maker perspective for costs of vaccination, influenza illness rates, and delays in vaccination using the best available information at the time of the pandemic.

None of the published studies reviewed above appear to have accounted for costs or health consequences potentially associated with H1N1 vaccination, except for the Khazeni et al and Beigi et al. studies. The exclusion of vaccination-related adverse events would yield more favorable results for vaccination compared with a more comprehensive analysis that included adverse events. Both costs and health effects of potential vaccine adverse events are explicitly accounted for in our analysis. Additionally, outside the US, adjuvanted vaccine was typically used and this formulation could result in a different risk profile than non-adjuvanted vaccine. The study by Khazeni et al. assumed the use of adjuvanted vaccine for a US setting and is not directly comparable to our study due to assumed differences in effectiveness and side effect profile between the two formulations.

Conclusions

Vaccination for pH1N1 influenza for children and young adults is cost-effective compared to other preventive health interventions under a wide range of scenarios. Delayed availability of pH1N1 vaccine results in less favorable cost-effectiveness results. A vaccination program for pH1N1 influenza for target groups can be justified from an economic perspective when indirect benefits are not considered and assuming that vaccine supplies are sufficient. Additional economic and health benefits beyond direct benefits would only add to the cost-effectiveness of pandemic influenza vaccination.