Vaccination is unlikely to result in societal monetary savings.

West Nile virus (WNV) was first detected in the Western Hemisphere in 1999 in New York City. From 1999 through 2004, >16,600 cases of WNV-related illnesses were reported in the United States, of which >7,000 were neuroinvasive disease and >600 were fatal. Several approaches are under way to develop a human vaccine. Through simulations and sensitivity analysis that incorporated uncertainties regarding future transmission patterns of WNV and costs of health outcomes, we estimated that the range of values for the cost per case of WNV illness prevented by vaccination was US $20,000–$59,000 (mean $36,000). Cost-effectiveness was most sensitive to changes in the risk for infection, probability of symptomatic illness, and vaccination cost. Analysis indicated that universal vaccination against WNV disease would be unlikely to result in societal monetary savings unless disease incidence increases substantially over what has been seen in the past 6 years.

West Nile virus (WNV) was first detected in the Western Hemisphere in 1999 during an outbreak of encephalitis in New York City (

WNV is transmitted to humans primarily through the bite of infected mosquitoes, but transmission through blood transfusion, through organ donation, and from mother to child have been described (

The decision tree used to evaluate the cost-effectiveness of vaccination compared with no vaccination is shown in the

Decision tree for vaccination program. WNV, West Nile virus.

We assumed that a single dose of live-attenuated WNV vaccine would provide immunity for >10 years, as is true for the currently licensed yellow fever vaccine (

Although the time horizon for risk for illness, protection from the vaccine, and cost of vaccination was 10 years, we used estimated lifetime costs of disease outcomes in our model. Thus, we modeled the difference in lifetime costs of illness that would be incurred by society during a 10-year period under an immediately implemented universal vaccination strategy compared with no vaccination.

The probabilities of outcomes and costs modeled are average probabilities for the entire population, regardless of age. Our analysis therefore estimates ranges of average societal costs and outcomes prevented when all people in the society are vaccinated, regardless of the age at vaccination or illness. A more detailed analysis of the effect of vaccinating certain age groups would require estimates of age-specific risks and costs of outcomes, which are not readily available for most outcomes in the model.

The overall cost of WNV illness per person at risk was calculated as the sum of the average costs for each health outcome weighted by the probability of occurrence of each outcome (

Asymptomatic infection was assumed to have no cost. Estimates of the cost of uncomplicated febrile illness due to WNV infection were not available so we assumed a cost of US $1,000 per case, based on 5 days of lost productivity at $165 per day (

The average societal cost due to death from WNV disease was estimated by using productivity loss tables (

Since no human WNV vaccine was licensed at the time of our evaluation, vaccine costs were not available. Based on charges in the United States for yellow fever vaccine (≈$85 per dose), hepatitis A vaccine (≈$75 per dose), Japanese encephalitis vaccine (≈$315 for a 3-dose series), and the previously available Lyme disease vaccine (≈$150 for a 3-dose series), we assumed a total baseline vaccination cost of $100 to include both the actual cost of the vaccine and the cost of administering the vaccine. For the sensitivity analysis focused on the cost of vaccination, we assumed minimum and maximum vaccination costs of $10 and $150, respectively (see below).

Several seroepidemiologic surveys have estimated the proportion of North American populations who were infected with WNV during epidemic transmission. The highest seroprevalence published to date is 2.6% (^{(–0.0016 × 10)} = 0.016. We therefore estimated the baseline probability of infection as 0.016. For sensitivity analysis focused on probability of infection, we assumed for the minimum risk for infection that a person would encounter only 1 year of WNV transmission, yielding a cumulative risk of 0.0016 over the 10-year period. For the maximum risk, we assumed that the risk would be that of yearly epidemic transmission such that 2.6% of the population would be infected each year over the 10-year period, yielding a 10-year cumulative risk of 0.23. Further details regarding sensitivity analysis are described below.

We assumed that symptoms of WNV illness will develop in 20% of infected persons and that neuroinvasive disease will develop in 3.6% of them, which is equivalent to 1 neuroinvasive case for every 140 infections previously described (

Precise data on long-term outcomes from WNV illness are limited. A study of 19 patients with neuroinvasive WNV disease found that 2 (11%) died, and of the 17 survivors, 7 (41%) had recovered fully at the time of discharge, 6 (31%) were discharged without full recovery, and 4 (24%) were discharged to a long-term care facility (

To incorporate uncertainties regarding the values of all input variables, we assigned uniform probability distribution to all variables, allowing 25% variability around the baseline values (

Variable | Lower limit | Baseline | Upper limit |
---|---|---|---|

Probability of infection | 0.012 | 0.016 | 0.02 |

Probability of symptomatic illness | 0.15 | 0.20 | 0.25 |

Probability of symptomatic illness after vaccination† | 0.03 | 0.04 | 0.05 |

Probability of neuroinvasive disease, given symptoms | 0.027 | 0.036 | 0.045 |

Probability of death, given neuroinvasive disease | 0.07 | 0.09 | 0.11 |

Probability of disability, given neuroinvasive disease | 0.26 | 0.35 | 0.44 |

Cost of neuroinvasive disease | $20,625 | $27,500 | $34,375 |

Cost of death (direct and indirect financial losses) | $150,000 | $200,000 | $250,000 |

Cost of lifelong disability | $158,000 | $210,000 | $263,000 |

Cost of uncomplicated WNV febrile illness | $750 | $1,000 | $1,250 |

Cost of vaccination | $75 | $100 | $125 |

*Upper and lower limits are calculated as ±25% of the baseline values and rounded up. †Baseline vaccine effectiveness is assumed to be 80%.

Using baseline values of all input variables, without accounting for uncertainties, the average cost per case of WNV illness prevented would be ≈$34,200. At a cost of $8.7 billion in a hypothetical population of 100 million people, vaccination would prevent 256,000 cases of WNV illness, including 9,216 cases of neuroinvasive disease, 2,935 cases of lifetime disability, and 829 deaths during a 10-year period. Under these assumptions, universal vaccination would yield societal savings if the cumulative incidence of WNV infection over a 10-year period were >0.13 (≈1.4% of the population infected each year), the cost of vaccination were <$12.8, or the cost of lifelong disability were >$3.2 million (≈15 times higher than the baseline estimate).

The simulation results accounting for uncertainties in all input variables are shown in

Statistic | ACER† |
---|---|

5th–95th percentile range, $ | –59,000 to –20,000 |

Mean, $ | –36,000 |

Median, $ | –35,000 |

Mode, $ | –33,000 |

Probability of savings, % | 0 |

*According to the distribution provided in

To identify the sensitivity of the output to all input distributions, we used @Risk sensitivity analysis with a regression in which the dependent variable was the output variable, i.e., ACER, and the independent variables were the input variables presented as @Risk uniform distribution functions (

Rank | Input variables | Regression coefficient† |
---|---|---|

1 | Probability of symptomatic illness | 0.65 |

2 | Probability of infection | 0.51 |

3 | Vaccination cost | 0.50 |

4 | Probability of symptomatic illness after vaccination | –0.14 |

5 | Probability of neuroinvasive disease, given symptoms | 0.05 |

6 | Cost of lifelong disability | –0.03 |

7 | Probability of disability, given neuroinvasive disease | 0.03 |

8 | Cost of neuroinvasive disease | –0.02 |

9 | Cost of uncomplicated WNV febrile illness* | –0.01 |

10 | Cost of death (direct and indirect financial losses) | –0.01 |

11 | Probability of death, given neuroinvasive disease | 0.00 |

*WNV, West Nile virus. †@Risk analysis software runs a regression where the dependent variable is the output variable, i.e., ACER, and the independent variables are the input variables presented as @Risk uniform distribution functions. Each iteration represents an observation for the regression. The coefficient calculated for each input variable measures the sensitivity of the output to that particular input distribution. For example, a coefficient of 0.65 indicates that a 1–standard deviation (SD) increase in probability of symptomatic illness increases the ACER by an SD of 0.65.

The results of the sensitivity analysis focused separately on risk for infection and vaccination cost are shown in

Statistic | Infection rate | Vaccination cost | ||||
---|---|---|---|---|---|---|

0.0016 | 0.016 | 0.23 | –150 | –100 | –10 | |

5th–95th percentile range, $ | –585,000 to –261,000 | –54,000 to –22,000 | 343 to 3,846 | –86,000 to –36,000 | –56,000 to –23,000 | –1,400 to 2,900 |

Mean, $ | –400,000 | –36,000 | 2,096 | –57,000 | –36,000 | 860 |

Median, $ | –386,000 | –34,000 | 2,098 | –55,000 | –35,000 | 920 |

Mode, $ | –373,000 | –30,000 | 1,500 | –54,000 | –36,000 | 740 |

Probability of savings, % | 0 | 0 | 98 | 0 | 0 | 76 |

*All other variables were allowed to vary according to their specified uniform distributions provided in

The economic impact of a vaccination strategy is a determinant of the public health decision regarding whether or not to recommend vaccination, but it is certainly not the only determinant of sound public health vaccination policy. It is also not imperative that a vaccination program result in monetary savings for it to be cost-effective compared with other public health interventions. Societies and people are willing to pay for preventing disease, as indicated by the implementation of preventive interventions that do not result in economic savings, and most relevant, the willingness to pay for expensive vaccines (

Our analysis indicates that a universal vaccination program to prevent WNV disease would be unlikely to result in societal monetary savings unless the incidence of the disease increases substantially over what has been seen in the past 6 years, or the cost of vaccination were <$12 per person vaccinated. The risk for WNV infection, probability of symptomatic illness after infection, and cost of vaccine appeared to have the greatest influence on the cost-effectiveness outcome. Within the range of possible values used in our model, variations in vaccine effectiveness, cost of WNV illness, and probabilities of various health outcomes did not lead to considerable change in the cost-effectiveness.

The future patterns of WNV transmission in North America cannot be accurately predicted. The virus was first detected in North America in 1999, and the epidemiology of WNV illness in the Western Hemisphere continues to evolve. The antigenically related flaviviruses St. Louis encephalitis virus (SLEV) and Japanese encephalitis virus (JEV) demonstrate different patterns of transmission that WNV could assume; SLEV is sporadically transmitted in North America with intense epidemics separated by years of low-level transmission, while JEV occurs in Asia with annual epidemics of intense transmission. If WNV assumes a transmission pattern in North America similar to that of JEV in Asia, then vaccination is likely to be a much more appealing public health prevention strategy and is likely to be more cost-effective than if WNV transmission follows the pattern of SLEV. As WNV spreads southward into Latin America, increased incidence may be seen with less protection from mosquitoes provided by air conditioning and screens (

WNV infection may cause severe untreatable neurologic disease. While the risk is highest in the elderly, severe disease does occur among young adults and children (

Our results provide a general assessment of the likely economic implications of universal vaccination against WNV and an indication of which parameters have the greatest influence on the cost-effectiveness of vaccination. A safe and effective vaccine may prove to be the most effective, and perhaps the most cost-effective, strategy to prevent severe WNV illness. The economic impact of vaccination will depend mostly on the risk for WNV infection, probability of symptomatic illness after infection, and the cost of vaccination.

We used the following formula to calculate the cost per case of West Nile virus (WNV) illness prevented:

_{V}is the cost of vaccination, and

Discounting is an economic notion that even in a world of zero inflation, a dollar today would be of higher value to a person than a dollar in the future. The premium placed on benefits today versus the future is reflected in the rate at which a person is willing to exchange present for future costs and benefits. This quantitative measure of time preference is called the

The average societal cost due to death from West Nile virus (WNV) disease was estimated by using productivity loss tables (

As a proxy for lifetime disability costs due to WNV illness, because of insufficient data, we used the lifetime costs of stroke available from the literature (

To make the lifetime disability costs and the death costs comparable, we used the ratio of the 3% discounted death cost ($200,000) and the 5% discounted death cost ($170,000) as a multiplier for adjusting the disability cost discounted at a 5% ($180,000) to a disability cost discounted at 3%. The result was $210,000 in 2004 dollars, which we used as an estimate for a 3% discounted disability cost.

We thank Roy Campbell for his thoughtful review of this work and the anonymous reviewers for their valuable comments and suggestions.

This study was supported by the National Center for Infectious Diseases, CDC.

Cost category | Average values in 2004 dollars* | Probability of incurring given cost† |
---|---|---|

Inpatient treatment | 19,197 | 1.00 |

Inpatient rehabilitation treatment | 14,977 | 0.14 |

Outpatient hospital cost | 316 | 0.32 |

Outpatient medical visits | 426 | 1.00 |

Outpatient physical rehabilitation | 3,859 | 0.22 |

Outpatient occupational rehabilitation | 3,822 | 0.07 |

Outpatient speech therapy | 556 | 0.01 |

Nursing home | 8,113 | 0.04 |

Productivity losses, temporary | 8,442‡ | 0.40 |

Productivity losses, caregiver | 2,406 | 0.26 |

Transportation cost | 64 | 1.00 |

Miscellaneous | 1,534 | 0.14 |

*Estimated from Zohrabian et al. (

Dr Zohrabian is a health economist with the Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, CDC. Her research interests are in cost-effectiveness and cost-benefit analysis, risk analysis, and summary measures of population health.