Influenza (vernacular name, flu) is a seasonal viral infection. The infection affects individuals at every age, although for the elderly it is the first cause of hospitalization and the fourth cause of death (Reference Wallinga, Teunis and Kretzschmar25).
In addition, influenza causes a high consumption of resources both from a social and a third-party payer (TPP) perspective. In 1981, the Office for Technology Assessment (United States) estimated a cost for a TPP between 1 and 3 billion dollars and 15 billion dollars for loss of productivity (22). The vaccination against influenza is actually recommended by the World Health Organization (27).
Despite the findings of Jefferson et al. (Reference Jefferson, Smith and Demicheli18), the European Scientific Working Group on Influenza (7) states that the recommendations in favor of a vaccination policy are principally targeted at individuals at high risk of complication. Although there is a general consensus about which categories of individuals should be considered at high risk (Reference Carman, Elder and Wallace3), different guidelines have been published in the European countries. In recent years, several studies have been carried out to provide an economic evaluation of the vaccination strategies against influenza (Reference Aballéa, Chancellor and Martin1;Reference Aballéa, De Juanes and Barbieri2;Reference Turner, Wailoo and Nicholson24). Aballea et al. (Reference Aballéa, Chancellor and Martin1;Reference Aballéa, De Juanes and Barbieri2), estimated the productivity gains in terms of number of working weeks in Italy, Germany, France, and Spain ranging from 77.319 (in Germany) to 151.737 (in France). These results were based on a decision tree focused on the minimization of the cost per quality-adjusted life-years (QALYs). However, the studies did not consider the dynamic effects resulting from the reduction of the transmission power. This should be taken into account to determine the optimal vaccination coverage for the health system. This study considers the impact on contagiousness of alternative strategies against influenza in people aged 50–64. We determine the social savings (in million Euros) of different coverage levels starting from the currently recommended in every country.
METHODS
Study Design
The Influsim 2.0 (Reference Eichner, Schwehm and Duerr6) dynamic model was used to simulate the course of an influenza spread in Italy, France, Germany, and Spain for people aged 50–64 and to estimate the incremental social savings of different vaccination strategies. The incremental social savings were defined as the difference with respect to savings in terms of weeks of absence from work, general practitioner (GP) visits, antiviral drugs, and hospital admissions resulting from the extension of the vaccination of 1 percent.
A social perspective was considered because alternative strategies against influenza largely affect the productive sector in terms of gains/losses of working days.
First, we estimated the resources involved in the implementation of the current recommendations which include vaccination for a high-risk population (20 percent in Spain and 25 percent of the aged 50–64 in the other countries). Second, we considered that the extension of the coverage leads to an increase of costs for vaccine doses which are compensated by a lower consumption of antiviral drugs. Third, we defined the optimal budget allocation as the one that maximized the number of vaccine doses given a stock of antiviral drugs to treat the infected population. Finally, we estimated the additional budget for different coverage levels and the incremental social savings.
Model Specification
Influsim 2.0 has been implemented by the Department of Biometry, University of Tubingen. It is based on a system of 1,081 differential equations to consider every clinical, demographic. and social parameter that is relevant to plan a strategy against influenza (Reference Eichner, Schwehm and Duerr6). We simulated the course for Italy, France, Germany, and Spain using a country-specific population aged 50–64. The effect of contagiousness among age groups was also included in the simulation (Reference Wallinga, Teunis and Kretzschmar25). We considered a population of 10.748.040 individuals for Italy, and respectively 10.630.900, 15.502.340, and 7.064.182 for France, Germany, and Spain (9;10;13;14). To model the contagiousness, we used the basic reproduction number (BRN) defined as the mean number of secondary cases that a typical single infected case will cause in a population with no immunity to the disease in the absence of interventions to control the infection (Reference Chowell, Nishiura and Bettencourt5). Table 1 shows the input parameters used for the structural hypothesis of the model.
Model Specification: Input Parameters
aStandard deviation, alpha, and beta parameters as well as the statistic distribution assumed for probabilistic sensitivity analysis are available in Supplementary Table 1 published in the online version of the manuscript.
HR, high risk of complications; LR, low risk of complications; GP, general practitioner.
Identification, Measurement, and Evaluation of Costs
The basic measure to model costs was given by the length of the infection. The cost drivers were identified in weeks of absence from work, number of GP visits, number of hospital admissions, doses of antiviral drugs, and doses of vaccine (Reference Aballéa, Chancellor and Martin1;Reference Aballéa, De Juanes and Barbieri2).
The costs did not relate to complications. The prices in Euros to value the resources (see Table 2) were extracted from literature and international database (8;11;12;15;16;Reference Medicom20;Reference Salute21;Reference de Salud23;Reference Wezel26). The human capital approach was used to estimate the production losses. Patients’ time off work was measured in terms of hourly wages with the assumption that it reflects productivity. The average annual costs of work in different sectors of activity were estimated according to the data previously published (9;10;13;14).
Prices of the Identified Cost Drivers
Sensitivity Analysis
We performed a one-way probabilistic sensitivity analysis (PSA) to model the uncertainty of the parameters. Supplementary Table 1 (which can be viewed online at www.journals.cambridge.org/thc2010020) shows the values of the alpha and beta parameters used to fit the random distributions consistently with the international guidelines (17). Second, we conducted a Monte Carlo simulation to assess the uncertainty through the different parameters. The heterogeneity of the epidemic spread, was assessed ranging the BRN from 1.68 to 3.
RESULTS
Base Case
Table 3 shows the incremental social savings after the solution of the optimization problem for different coverage levels in Italy, France, Germany, and Spain.
Results: Incremental Social Savings and Additional Level for Different Coverage Levels
Italy
In Italy, the budget to cover 25 percent of the aged 50–64 and to purchase the antiviral drugs is 67 million Euros. The optimal budget allocation suggests the extension of the coverage to 32.75 percent with an incremental social saving of 125 million Euros. The extension of the coverage up to 100 percent always increases the social savings. In conclusion, with a total budget spending of 114 million Euros the entire population aged 50–64 is vaccinated and the total social savings are 600 million Euros.
France
In France, the optimization of the budget available for vaccine doses and antiviral drugs (91 million Euros) suggests the extension of the coverage up to 32.4 percent and an incremental social saving of 117 million Euros. The extension of the coverage up to 100 percent increases the social savings. In conclusion, with a total budget spending of 131 millions of Euro the entire population aged 50–64 is vaccinated and the total social savings are 740 million Euros.
Germany
In Germany, the budget is 71 million Euros and the optimal allocation suggests the extension of the coverage to 38,5 percent and 148 million Euros social savings. However, due to the population composition, the additional spending is higher than the incremental benefit starting from the 80 percent coverage. In conclusion, with a total budget spending of 122 million Euros, the entire population aged 50–64 is immune and the total social savings compared with the starting coverage (25 percent) are 811 million Euros.
Spain
In Spain, the estimated budget is 45 million Euros and its optimal allocation suggests the coverage of 28.3 percent of the population aged 50–64 with an incremental social saving of 129 million Euros. However, the additional spending is higher than the incremental benefit starting from the 80 percent coverage because the whole population is assumed to be immune at this level and social savings are 720 million Euros with a total budget spending of 54 million Euros.
Sensitivity Analysis
One-way PSA shows that the parameters with an higher impact on the increase/decrease of the social savings, are reduction of the absenteeism, percent of severe cases among symptomatic, percent of asymptomatic, and average duration of the convalescence period.
Supplementary Table 2 shows some statistics to resume the results of the Monte Carlo simulations conducted for each country included in the study. The results show the variation in total social savings, for the optimal coverage level and for 50 percent, 80 percent, and 100 percent coverage. The table shows a huge variability of the results. Nevertheless, the confidence interval (95 percent) shows that the results can be considered consistent.
DISCUSSION
In this study, we used a dynamic model to simulate the course of a hypothetic infection caused by influenza. We compared, in terms of incremental social savings, alternative strategies corresponding to different levels of coverage starting from the level currently recommended in Italy, France, Germany, and Spain. In Italy and France, we showed how an optimal budget allocation entails an extension of the coverage level from 25 percent to 32.75 percent and 32.4 percent, respectively. In the analysis, we also considered that the marginal benefits (in terms of social savings) of a coverage expansion tend to decrease because of the progressive reduction in contagiousness.
In Germany and Spain, the optimal coverage would be 38.5 percent and 28.3 percent, respectively. Nevertheless, in these two countries, the total savings decrease starting from an 80 percent coverage level. The population in both cases plays a crucial role. In Germany, the total savings tend to decrease beyond 80 percent coverage level, because 100 percent coverage of a huge population suggests a higher cost for vaccine doses than the incremental social savings.
In Spain, the transmission power would be halted at 80 percent coverage and further extensions would suggest additional costs for vaccination without additional gains.
To test the consistency of our results, we applied the dynamic model with the same assumptions used in Aballéa et al. (Reference Aballéa, Chancellor and Martin1) and (Reference Aballéa, De Juanes and Barbieri2). For example, in Aballéa et al. (Reference Aballéa, Chancellor and Martin1) it is shown that, in Italy, a vaccination strategy involving 67 percent of the high-risk population and 52 percent of low-risk individuals would save 111.981 weeks of work with respect to a strategy involving the coverage for 42 percent of the high-risk population and 17 percent of the low-risk individuals. We replicated the two strategies with our model and the results showed a saving that ranged from 104.000 (BRN = 1,68) to 143.294 (BRN = 3) weeks of work. However, our study shows that the strategies considered in Aballéa et al. (Reference Aballéa, Chancellor and Martin1) and (Reference Aballéa, De Juanes and Barbieri2) are sub-optimal. Some limitations have to be underlined in our analysis as follows: (i) Results are based on a simulation with a hypothetical population and the real evolution of a FLU infection should be observed; (ii) Results are not considered in terms of incremental cost-effectiveness ratio (ICER). The reason is that we aimed to add some evidence to the already known results of (Reference Aballéa, Chancellor and Martin1) and (Reference Aballéa, De Juanes and Barbieri2); (iii) We considered oseltamivir and zanamivir as the antiviral drugs administered to the infected and alternative treatments (both antipyretics and antiviral drugs) should be included; (iv) We did not consider the different types (A-B) of the infection, but we modeled this difference by using the BRN measure.
POLICY IMPLICATIONS
In conclusion, our study was aimed at underlining the need for decision makers to determine optimal vaccination policies with the budget available. This should be made by using analytical supports to model the epidemiological characteristics of the infection, the effects of the extension of the vaccination on the contagiousness and the population composition.
SUPPLEMENTARY MATERIAL
Supplementary Table 1
Supplementary Table 2
CONTACT INFORMATION
Americo Cicchetti, PhD (acicchetti@rm.unicatt.it), Professor, Scientific Health Technology Assessment Unit, Policlinics “A. Gemelli, Matteo Ruggeri, MSc, MA, PhD (mruggeri@rm.unicatt.it), Lecturer, Facoltà di Economia, Università Cattolica del Sacro Cuore, Largo F. Vito 1, 00168 Roma, Italy
Lara Gitto, MSc (Gitto@ceis.uniroma2.it), Research Fellow, Francesco Saverio Mennini, MSc, BA (f.mennini@uniroma2.it), Professor, CEIS Sanità – Tor Vergata, Facoltà di Economia, Università di Roma Tor Vergata, Via Columbia 2, 00133 Roma, Italy
CONFLICT OF INTEREST
All authors report having no potential conflicts of interest.
