Results of the multivariate sensitivity analysis. Shown are the standardized regression coefficients for the variables included in the cost models. A coefficient of 0 indicates that there is no significant relationship between the input variable and output variable. A coefficient of 1 or −1 indicates a 1- or −1-SD change in the output for a 1-SD change in the input. WL indicates work loss; VE, vaccine efficacy. (See Table 5 for a listing of the variables along with their base-case values and ranges of values used in the cost model.) Shown are all variables from the model that had a standardized regression coefficient with an absolute value of 0.1 or greater.
Nichol KL. Cost-Benefit Analysis of a Strategy to Vaccinate Healthy Working Adults Against Influenza. Arch Intern Med. 2001;161(5):749-759. doi:10.1001/archinte.161.5.749
Copyright 2001 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2001
Influenza is a major cause of illness, disruption to daily life, and work absenteeism among healthy working adults aged between 18 and 64 years. This group is not included among the traditional priority groups for annual vaccination. Immunization rates remain low.
To assess the economic implications of a strategy for annual vaccination of this group.
Using the societal perspective, this cost-benefit analysis included the direct and indirect costs associated with vaccination as well as the direct and indirect costs prevented by vaccination. Clinical and economic variable estimates were derived primarily from the published literature. For this model, it was assumed that vaccination occurred in efficient, low-cost settings such as at the work site. Monte Carlo simulation was used to calculate the mean net costs or savings along with the 95% probability interval, and sensitivity analyses explored the sensitivity of the cost model to different values of the input variables.
Vaccinating healthy working adults was on average cost saving, with mean savings of $13.66 per person vaccinated (95% probability interval: net savings of $32.97 to net costs of $2.18), with vaccination generating net savings 95% of the time. The model was most sensitive to the influenza illness rate, the work absenteeism rate due to influenza, and hourly wages. In the worst-case scenario vaccination was not cost saving. Vaccination also generated net costs to society during years with a poor vaccine–circulating virus strain match. In all of the other sensitivity analysis scenarios, vaccination was cost saving.
Influenza vaccination of healthy working adults on average is cost saving. These findings support a strategy of routine, annual vaccination for this group, especially when vaccination occurs in efficient and low-cost sites.
INFLUENZA IS a major cause of morbidity and mortality affecting up to 25% of the population each year.1 Elderly and other high-risk persons are especially vulnerable to the serious complications of influenza such as secondary bacterial pneumonia and exacerbations of underlying medical conditions. In these high-risk groups influenza and its associated complications account for approximately 100 000 to 300 000 excess hospitalizations,1,2 20 000 to 40 000 excess deaths,3,4 and billions of dollars in health care costs.5 Thus elderly and other high-risk persons are specifically targeted for annual vaccination against influenza6 and, in fact, vaccination has been shown to be highly cost-effective and even cost saving for these groups.7,8
Healthy working adults traditionally have not been included among the priority groups targeted for annual influenza vaccination.6 It is not surprising that fewer than 25% of the persons aged between 18 and 64 years received an influenza vaccination during 1997.9 Nevertheless, the effect of influenza on this group is also substantial. Among healthy working adults, the major impact of influenza is related to the disruption of daily life caused by the symptoms of influenza. The typical case of influenza may be characterized by the abrupt onset of fever, sore throat, nonproductive cough, myalgias, headache, and malaise. Symptoms usually last 5 to 6 days. Each episode of illness may be associated on average with 3 to 4 days of bed disability and 3 days of school or work absenteeism.10 Approximately half of all influenza illnesses may result in a health care provider visit.1,10 Data from the National Health Interview Survey11 for persons aged 18 to 64 years indicate that influenza was responsible for more than 200 million days of restricted activity, 100 million days of bed disability, 75 million work absenteeism, and 22 million health care provider visits in 1995.11
The results of a cost-effectiveness analysis conducted in the early 1980s suggest that influenza vaccination of healthy working adults could be highly cost-effective although it did not find vaccination to be a cost saving for adults younger than 65 years.7 That study, however, did not incorporate indirect costs associated with productivity losses averted. A placebo-controlled trial12 that assessed the benefits of influenza vaccination during the 1994-1995 season among healthy working adults found that vaccination was associated with both direct and indirect cost savings based on the illness rates and vaccine-effectiveness rates observed in the study. Another trial13 conducted during the 1997-1998 and 1998-1999 seasons found that vaccination was associated with health benefits but did not find that vaccination was associated with economic benefits. The generalizability of the findings from these recent trials to other influenza seasons and to the general working adult population is uncertain given the demographic characteristics of the study populations and the year-to-year variability of illness rates, vaccine-effectiveness rates, and other parameters that influence overall clinical and economic benefits.
This study represents a cost-benefit analysis of influenza vaccination of healthy working adults. While such a vaccination strategy advocates annual immunizations, this analysis assesses the average, expected net costs or savings for a single season. To explore the potential benefits of a nationwide strategy that targets this group for vaccination, the model uses nationally representative parameter estimates derived from observations spanning multiple seasons whenever possible. It also accounts for the year-to-year variability in illness rates and vaccine-effectiveness rates and considers plausible ranges for costs and other model parameters. Both direct and indirect costs were included in the analysis.
This cost-benefit analysis took the societal perspective and assessed the net costs or savings associated with the use of inactivated influenza virus vaccine in healthy working adults compared with no vaccination. Both direct and indirect costs were included in the model. The basic cost model was as follows:
The costs of vaccination represented the sum of the direct costs of vaccination (vaccine and its administration), indirect costs for work absenteeism to be vaccinated, direct costs for side effects (medical care costs including health care provider visits, tests, and medications), and indirect costs for work absenteeism due to side effects. The costs averted due to vaccination represented the sum of the direct costs of medical care for influenza illness prevented (health care provider visits and hospitalizations), indirect costs of work absenteeism averted, indirect costs associated with future lifetime earnings preserved due to fewer deaths, and the indirect costs of reduced work effectiveness averted. The cost variables and detailed derivation equations are given in Table 1.
Up to 25% or more of the general population may develop influenza each year, although illness rates vary by age group.1 Illness rates are highest among school-aged children, at times reaching 30% or more. In adults annual laboratory-confirmed influenza illness rates may be as low as 5% as reported for the placebo groups in 2 recently published clinical trials14,15 assessing the effectiveness of neuraminidase inhibitors for the prevention of influenza. In another controlled trial,16 however, the laboratory-confirmed illness rate among unvaccinated participants was 13%. The 8-year average influenza illness rates in 1 longitudinal study17 spanning the years from 1976 through 1984 ranged from 19.7% for persons aged 18 to 24 years to 14.8% for persons 35 years or older. These rates were similar to the interpandemic influenza illness rates among adults that were reported for the 1928-1929 season.18 These studies are summarized in Table 2. In a systematic review of the prevention and treatment of influenza in healthy adults, the mean absolute risk differences for clinical influenza observed in trials assessing the efficacy of influenza vaccination in healthy adults were 5% for studies in which the vaccine studied was the vaccine recommended by the World Health Organization for that season and 7% when the vaccine studied was well matched to the predominant circulating virus strains for those seasons.19 Given the average vaccine efficacy rates in these studies (68% and 72%, respectively), the underlying clinical influenza illness rates among unvaccinated persons would have been 7.4% and 9.7% for the 2 groups of studies. For this cost-benefit analysis, 5% was selected as the base-case illness rate with a range of 5% to 15%. It was also estimated that approximately 45% of illness episodes would be medically attended.1,10,20
Hospitalization rates for acute respiratory tract disease during the influenza season among adults aged from 20 to 44 years have been reported to be in the range of 4.6 per 10 000 population to 13.5 per 10 000 population while the hospitalization rates for adults aged from 45 to 64 years have been reported to be 6.8 per 10 000 population to 25.2 per 10 000.4,21 Among persons without cardiopulmonary disease, rates of 6.6 and 13.6 per 10 000 population for persons aged 20 to 44 years and 45 to 64 years, respectively, have been observed while the rates among persons with cardiopulmonary disease were been almost twice those of the low-risk group at 12.5 and 20.6 per 10 000 population.17 Other investigators have reported excess hospitalization rates for persons aged 15 to 44 years of 3.9 per 10 000 population and rates of 11.2 per 10 000 for persons aged 45 to 64 years.2 In a large observational study22 of women enrolled in the Tennessee Medicaid program spanning the years 1974 through 1993, excess hospitalization and death rates for low-risk women were 4 per 10 000 population for person aged 15 to 44 years and 6 per 10 000 population for persons aged 45 to 64 years. Most of these events probably represented hospitalizations as only about 8% of the persons hospitalized with an influenza-associated complication die.23 The results of these studies are summarized in Table 3. For this cost-benefit study, the base-case excess hospitalization rate was 4 per 10 000 population with a range of 1 to 7 per 10 000 population.
Excess pneumonia and influenza mortality rates during influenza seasons for persons of all ages have been reported at rates of 2.5 to 4.1 per 100 000 population during the 1940s and 1950s while excess mortality rates for deaths from all causes ranged from 14.5 to 24.4 per 100 000 population.24 Kavet10 reported excess all-cause mortality rates of 10.5 per 100 000 persons to 24.9 per 100 000 persons of all ages for epidemic years occurring in the 1960s. Excess pneumonia and influenza mortality rates for adults up to age 44 years have been reported to be in the range of 0.1 per 100 000 population to 2.7 per 100 000 population while excess pneumonia and influenza mortality rates for persons aged 45 to 64 years have ranged from 0.8 per 100 000 population to 8.9 per 100 000 population.25,26 In a recently published study26 assessing the excess pneumonia and influenza death rates spanning the years 1968 through 1992, it was noted that excess pneumonia and influenza deaths represented only 25% of all excess deaths observed. Thus, the study findings of excess pneumonia and influenza deaths of 0.1 and 0.8 per 100 000 for persons aged 15 to 44 years and 45 to 64 years would correspond with excess deaths from all causes of 0.4 and 3.2 per 100 000 population, respectively. Table 4 summarizes these studies. For this analysis, an excess all-cause mortality rate of 1 per 100 000 population (range, 0.5-2 per 100 000 population) was used as the base-case estimate.
Based on data from the National Health Surveys, Kavet10 estimated that influenza illness episodes are, on average, associated with 3.2 to 3.4 days of work loss. Findings from a more recent observational study27 of industrial workers in the United Kingdom indicated that workers lost 2.8 days (SD = 2) of work for each occurrence of influenzalike illness. In another quasi-experimental study28 that assessed the benefits of influenza vaccination among working adults, employees with influenza illness missed an average of 4.9 days of work. However, in a recent clinical trial13 of healthy working professionals, the average number of working days lost per episode was 0.79 days. The base case used for the present study was 2 work absenteeism days per episode of influenza illness (range, 0.75-4 work absenteeism days).
Employees with influenza may continue to work or return to work before they are fully recovered. Acute influenza has been shown adversely to affect the ability to perform reaction time tasks to a degree similar to that seen with alcohol consumption or work at night.29 Illness has also been associated with lower levels of work effectiveness. For workers in the United Kingdom who had influenzalike illnesses, among those who were still symptomatic at work their effectiveness at work was reported to be at a level of 4.6 on a scale of 1 to 10 (1, totally ineffective; 10, totally effective).27 In that study,27 workers who experienced an influenzalike illness reported a mean duration of illness of 3.4 days with 2.8 days of work loss. Therefore, the number of days of working at reduced effectiveness was about 0.7 days. However, for influenza illness episodes of average or longer duration (eg, ≥5 days), if workers miss only 3 days of work, they may have up to 2 or more days per illness episode of working while at reduced effectiveness. In the present study, it was assumed for the base case that employees would work only 0.7 days at reduced levels of effectiveness per illness episode (range, 0.5-0.9) with an average level of effectiveness of 50% (range, 30%-70%).
A recent double-blind, placebo-controlled trial30 showed that influenza vaccination with inactivated vaccine is not associated with higher rates of systemic symptoms than are seen following placebo injection among healthy working adults. Thus, the most likely estimate for work absenteeism and health care provider visits due to side effects from vaccination is close to 0. However, in that trial 4.5% of vaccine recipients vs 3.5% of placebo recipients missed work during the week following vaccination (P = .58). In both groups, 73% of people missing work missed only 1 day of work during the week following vaccination (K.L.N., unpublished data, 2000) For this study, 1 additional work absentee day per 100 persons vaccinated due to possible side effects was used as the base-case estimate, with half of these episodes being associated with a health care provider visit (base case of 0.5 health care provider visits due to side effects per 100 persons vaccinated with a range of 0-1 per 100 persons vaccinated [0-10 per 1000 persons]).
The 1976 swine influenza vaccine was associated with an increased frequency of Guillain-Barré syndrome, at an absolute excess rate of about 1 case per 100 000 persons vaccinated.6 Evidence for a causal relationship of subsequent vaccines with Guillain-Barré syndrome has been less clear. However, in a study31 of the 1992-1993 and 1993-1994 seasons, the overall relative risk for Guillain-Barré syndrome among influenza vaccine recipients during the 6 weeks following vaccination was 1.7 (95% confidence interval, 1.0-2.8; P = .04), with an absolute risk of about 1 per 1 million persons vaccinated. For this study, a risk of Guillain-Barré syndrome of 1 per million persons vaccinated was used for the base case (range, 0.5-2 per 1 million persons vaccinated).
During years with a good match between circulating viruses and the corresponding vaccine strains, vaccine efficacy for reducing laboratory-confirmed influenza illness has generally been 70% to 90% (Table 5).16,32- 35 For this study, the base-case vaccine efficacy rate for years with a good match was 75% (range, 60%-90%). During years with a poor match, vaccine efficacy in several previous studies32- 34 has been observed to be 30% to 60%. The base-case efficacy during years with a poor match for this cost-benefit analysis was selected to be 35% (range, 0%-50%). Over the past decade, for 8 of 10 years there was a good to excellent match between the predominant epidemic virus strain and the corresponding vaccine strain (Nancy J. Cox, PhD, Centers for Disease Control and Prevention, unpublished data, November 1999. Thus 80% (range, 72%-88%) was used as the likelihood of a good match for any given year.
Among elderly persons living in nursing homes, influenza vaccine is more effective in preventing the serious complications of influenza such as hospitalization or death than in preventing influenza illness itself.6 Similar findings have not been demonstrated for younger, healthy populations. Accordingly, for the purposes of this study it was assumed that the efficacy of influenza vaccination was the same for all clinical outcomes.
For this study it was assumed that vaccination would occur in work site–based clinics, community health department clinics, or at other nontraditional sites such as public clinics in drug stores and grocery stores. A survey of 61 such sites for the 1998-1999 season showed a mean charge for vaccination of $9.64 (median, $10; range, $5-$15) (Mari Drake, MPH, American Lung Association of Minnesota, written communication, October 1999). The base-case cost for the vaccine and its administration used for the present analysis was $10 (range, $5-$15). This value for vaccine cost is higher than the Medicare reimbursement rate for influenza vaccine and its administration which was $8.50 in 1998.36
As part of the economic analysis of a placebo-controlled trial assessing the benefits of influenza vaccination in healthy adults, the cost of a health care provider visit including procedures and medications was estimated as $69.51 in 1994 dollars12 or $80 when inflated to 1998 dollars. This estimate was based on information from the American Medical Association's Socioeconomic Monitoring System Core Survey for physician's fees, a list of diagnostic tests and medications associations with physicians visits for upper respiratory tract infections from the National Ambulatory Medical Care Survey, payments to physicians for diagnostic tests from the Physician Payment Review Commission Multipayor Database, and costs of generic medications from local branches of national franchise pharmacies.12 Another estimate of the cost of an initial outpatient visit for influenza illness was $112 in 1995 dollars or $123 in 1998 dollars including the costs of physicians' fees as well as medications and patient co-payments.37 This estimate was based on data from the proprietary MEDSTAT database (The MEDSTAT Group, Ann Arbor, Mich) that contains health insurance claims data from approximately 4 million insured persons. For the present study, the mean in 1998 dollars of these 2 previous estimates, $102 (range, $80-$123), was used as the base-case value for the models.
The costs per hospitalization for influenza-associated complications were estimated using the national Healthcare Cost and Utilization Project data for 1996.38 This project which is sponsored by the Agency for Healthcare Research and Quality represents a federal-state-industry partnership designed to develop a multistate health care database for health services research and policy analysis. The uniform data make possible comparative studies of health care services and costs of hospital care. The Nationwide Inpatient Sample: Healthcare Cost and Utilization Project is the largest all-payor inpatient care database in the United States, containing inpatient stay data from a national sample of 900 hospitals representing 6.5 million hospital stays. For this study we used the weighted average of the 1996 hospital charges for persons aged 15 to 44 years and for persons aged 45 to 64 years who had a discharge diagnosis of influenza or possible influenza complications including pneumonia, acute bronchitis, other respiratory tract infections, chronic obstructive lung disease, or asthma adjusted by the 1996 Medicare hospital cost-charge ratio.39 For the base case, the estimate standardized to 1998 dollars was $5669 (range, $3669-$7669).
The medical care costs per episode of vaccine-associated Guillain-Barré syndrome were previously estimated to be $100 800 in 1995 dollars.37 These costs were inflated to 1998 dollars ($110 674) and used in the cost model for this study.
As has been recommended, the human capital approach was used to estimate the economic value of worker productivity.40 Hourly wages were used to estimate the value of a workday. The mean hourly wage for full-time, year-round workers weighted for age and sex was $19 in 1998, while the median hourly wage for all full-time, year-round workers weighted for sex was $15.41 In a recent clinical trial13 of influenza vaccination in healthy working adults, the mean hourly wage for study participants was $29.39. For the present study, the median hourly wage of $15 (range, $10-$30) for all US workers was used as the base-case value. The economic cost of a premature death was estimated as the present net value of future lifetime earnings and housekeeping services foregone for someone dying between the ages of 20 and 64 years42 weighted for sex and expected age at death. The base-case and best-case estimate used a 3% discount rate while a 5% discount rate was used for the worst-case scenario. These discount rates were selected to be consistent with current guidelines.43
For the analyses, the mean values for the output variables (eg, net costs or savings) associated with vaccinating healthy, working adults were calculated using Monte Carlo simulation (@risk, Windows Version, July 1997; Palisade Corp, Newfield, NY).44 For these multivariate probabilistic simulations, the uncertain parameters in the cost model were varied simultaneously across predefined probability distributions to generate a probability distribution for the likely values of the output variable. Triangular probability distributions were created for the model variables that were defined by a minimum, maximum, and most likely or base-case value for each variable. Up to 10 000 iterations were conducted to ensure convergence of the models. The standardized regression coefficients for the variables in the simulation models were calculated to demonstrate the sensitivity of net costs to each of the input variables and their probability distributions while controlling for the other variables in the model. In addition, sensitivity analyses were conducted to evaluate the effect of fixing individual model parameters at their highest or lowest values on the mean net costs or savings. The worst-case and best-case scenarios (ie, those cases in which all model variables were fixed at the values least favorable and most favorable to vaccination) were also analyzed.
Recently published data exploring the association of pneumonia and influenza hospitalization and death rates have demonstrated that the 2 are highly and positively (but not perfectly) correlated with a correlation coefficient of 0.8.45 Likewise, data from cohort studies over 6 consecutive seasons that evaluated the health and economic benefits of influenza vaccinations among elderly members of a large managed care organization suggest that hospitalizations for all acute and chronic respiratory tract diseases and death rates from all causes during influenza seasons are also highly and positively correlated.(K.L.N., unpublished data, 2000) Therefore, a rank-order correlation coefficient of 0.8 was included in the models to account for this degree of correlation or relationship between hospitalization and death rates. It was assumed that there would be a similar association between influenza illness rates and hospitalization rates. It has also been observed that, among persons with influenza illness, the number of days of sick leave may differ according to profession, with managers taking fewer days of sick leave than secretarial or administrative staff.27 Accordingly, a correlation coefficient of −0.8 was included in the model to describe the inverse relation between hourly wages and days of work absenteeism due to influenza.
The most likely (ie, base-case) values and ranges for the variables included in the cost model are listed in Table 6. All costs were standardized to 1998 dollars using the appropriate component of the Consumer Price Index.46
The mean reductions in clinical outcomes among vaccinated persons as calculated in the Monte Carlo simulation model are summarized in Table 7. Vaccination in the model prevented an average of 12.3 workday absenteeisms and 2.5 physician visits per 100 persons vaccinated. Vaccination was also estimated to prevent 2.6 hospitalizations per 10 000 persons vaccinated and 0.77 deaths per 100 000 persons vaccinated. The weighted mean vaccine efficacy rate in the model was 66% (95% probability interval, 55%-76%).
The mean costs per person vaccinated are summarized in Table 8. Vaccination resulted in net savings of $13.66 for each person vaccinated. The best-case scenario showed even greater savings of $174.32; while in the worst-case scenerio, vaccination generated net costs of $21.27 per person vaccinated (Table 9). The best-case and worst-case scenarios, however, were substantially more extreme than the range of plausible values derived from the multivariate Monte Carlo simulation. The 95% probability interval for the average cost savings ranged from a net savings of $32.97 to net costs of $2.18 per person vaccinated. According to the probability distribution generated for the net costs, vaccination resulted in net savings 95% of the time.
The cost model was most sensitive to the influenza illness rate, work absenteeism due to influenza, and hourly wages (Figure 1). In addition, vaccine efficacy, the cost of vaccination, the death rate due to complications of influenza, the likelihood of missing work to be vaccinated, and the amount of work absenteeism to be vaccinated were also important contributors to the model. The model was insensitive to rates of health care use or the costs of health care (health care provider visits and hospitalizations) resulting from either influenza or possible side effects due to vaccination.
The results of the sensitivity analyses demonstrating how the average costs or savings varied when the values of single input variables were fixed at their lowest or highest values are given in Table 9. Vaccination generated net costs in the worst-case scenerio and also when there was a poor match between vaccine and circulating virus strains. In each of the other scenarios, vaccination was associated with net savings.
To further assess the robustness of the model, the Monte Carlo simulation was repeated after changing the most likely values used for the probability distributions for the 8 input variables to which the model was most sensitive (ie, those shown in Figure 1) to the value least favorable to vaccination (ie, the worst-case value). The results of the Monte Carlo simulation under these conditions demonstrated that vaccination would still generate cost savings 57% of the time with a mean net savings of $2.02 per person vaccinated.
Influenza can have a substantial effect on healthy working adults and their employers, resulting in tens of millions of workday absenteeisms and health care provider visits each year. In this cost-benefit analysis, influenza vaccination of healthy working adults was on average cost saving. These findings were robust; vaccination saved money in each of the sensitivity analysis scenarios except in the worst-case scenerio and when there was a poor match between vaccine and circulating virus strains. According to the multivariate probabilistic sensitivity analysis, vaccination would generate net costs only about 5% of the time, ie, vaccination should be cost saving about 95% of the time.
Clinical studies including trials12,13 as well as quasi-experimental47 and observational studies (Wayne Lednar, MD, written communication, October 1998)48,49 that have assessed both the health and economic benefits of influenza vaccination among working adult populations have found vaccination to be associated with health benefits and often cost savings, although the magnitude of economic benefits varies from study to study. In one recently published study28 in which participants experienced low illness rates, vaccination was associated with net costs that exceeded the benefits. In another trial13 that spanned 2 seasons (1 of which was characterized by a poor match between vaccine and circulating virus strains), vaccination was associated with health but not economic benefits. However, the participants in that study reported lower than anticipated average rates of influenza-related work absenteeism. While the reasons for the low work absenteeism rate are unclear, the study population was preponderantly male with a high income and with a mean age over 40 years. Male sex, higher income, and increasing age have all been associated with lower rates of work absenteeism due to influenza and other respiratory tract illnesses.11 Thus, the study findings may have limited applicability to the general US working population. Given the inherent variability of the key parameters that determine the overall net costs or savings associated with vaccinating healthy working adults against influenza, it is not surprising that individual studies have provided somewhat differing results. The present cost-benefit analysis clarifies the net cost savings that might be expected when averaged over multiple seasons.
While the findings of this cost-benefit study demonstrate cost savings, an earlier economic analysis of the benefits of influenza vaccination conducted by the Office of Technology Assessment found that vaccination was not cost saving but was, nevertheless, highly cost-effective, with estimated costs per year of healthy life gained of $64 for persons aged 25 to 44 years in the general population (or about $250 in 1998 dollars) and $23 per year of healthy life gained for persons aged 45 to 64 years (or about $90 in 1998 dollars).7 That study did not include indirect costs in the model. Nevertheless, the findings of such low cost-effectiveness ratios provide powerful evidence suggesting that routine vaccination of all persons aged 25 to 64 years might be beneficial even in the absence of expected cost savings.
The findings of the present study are also consistent with a recently published analysis of the economic impact of pandemic influenza in the United States.37 In that study both direct and indirect costs were included in the models. The authors found that when considering total economic benefits from the societal perspective net returns to society for vaccinating low-risk adults aged 20 to 64 years in the setting of a pandemic influenza would exceed those from vaccinating high-risk elderly persons and would most probably generate cost savings.
The major economic effect of influenza among healthy working adults is the impact on worker productivity. In the present cost-benefit analysis, the economic value of the indirect costs or productivity losses avoided represented about 78% of all costs prevented because of vaccination. Thus, how workers and employers actually value work time and productivity will substantially influence the actual net costs or savings of a strategy to vaccinate all working adults. If the costs to the employer of a lost workday do not equal the value of a day's wages, then the net savings associated with vaccination will be lower than what has been outlined in this study. In a systematic review of the prevention and treatment of influenza in healthy adults it was found that vaccination on average reduced absenteeism by about 0.34 days per person vaccinated (about 34 days per 100 persons vaccinated).50 However, productivity losses were felt to have no economic value in that study, and vaccination was not found to be cost saving.19 However, if one worker's absenteeism results not only in the loss of that person's productivity, but also reduces the productivity of other workers, then the economic benefits associated with vaccination may be even higher. Savings due to reduced work absenteeism have been significant contributors to the net cost savings reported in recent economic analyses of strategies to promote the use of other vaccines as well.51,52
For this study it was assumed that vaccination occurred in convenient and efficient settings such as work sites, community-based walk-in clinics, or other nontraditional settings. If working adults received vaccinations in more costly settings, then the costs of vaccination would be higher and the net savings would be lower. National data suggest that increasing numbers of people are being immunized in nontraditional settings such as those assumed for this study. For the 1998-1999 influenza vaccination season, about 60% of vaccinated adults younger than 65 years were immunized at the work site, a store, or other such setting (James A. Singleton, MS, Centers for Disease Control and Prevention, unpublished data, June 2000). In the model used for this cost-benefit analysis, if 60% of persons vaccinated incurred average costs as listed in Table 5, the break-even costs for the other 40% of persons receiving vaccinations in higher cost settings would be $48 per person vaccinated (K.L.N., data not shown, 2000). Receipt of vaccine in highly inefficient settings would also increase vaccination costs and decrease the cost savings associated with vaccination. For example, in one study when vaccination required 2 hours of lost working time per person vaccinated, vaccination was not cost saving.28
This study probably underestimates the magnitude of the costs of influenza and savings due to vaccination. The economic value of leisure time or the value that individuals place on avoidance of suffering was not included in the model. Furthermore, the economic value of a human life, represented in this study by the present value of future lifetime earnings foregone in the event of a premature death, does not represent the full value of that life. This study also did not consider the potential benefits that might be realized due to reduced severity of illness among vaccinated persons who might develop influenza or due to the benefits to be derived from the prevention of complications of influenza that do not require hospitalization such as bronchitis or other secondary infections.
This study has several limitations. The findings are valid only insofar as the underlying assumptions and values used for the individual variables included in the model are valid. Every attempt was made to select representative and even conservative estimates for the uncertain parameters included in the model, and the results of the sensitivity analyses suggest that the findings were robust. The average vaccine-efficacy rate as well as reductions in clinical influenza rates and workday absenteeisms from a recent systematic review of influenza vaccines in healthy adults19 are similar to the corresponding rates in the cost model used for this study, suggesting that the model functioned fairly well at least for these key clinical variables. Nevertheless, the results should be interpreted with some caution. Furthermore, because the model assumed that the persons targeted for vaccination were full-time, year-round workers, the findings may not apply to part-time or seasonal workers. Thus, the results may be most applicable to the 94 million full-time, year-round workers between the ages of 18 and 64 years in the United States. How the findings might apply to the other 72 million persons in this age group who may be part-time or seasonal workers or who may not be employed is less clear.
An additional caution is noteworthy when considering the results of this study. In times of vaccine shortage or delay in the availability of vaccine as occurred during the 2000-2001 vaccination season, priorities for targeting groups for vaccination will most likely be based not on economic considerations but on the relative risks of certain groups for the serious complications of influenza that might result in hospitalization or death.53 Thus, the findings from this study are best interpreted under conditions of adequate and timely availability of vaccine for all persons wishing to be immunized.
Recently the Advisory Committee on Immunization Practices for the Centers for Disease Control and Prevention and the American Academy of Family Practice lowered their age-based recommendations for influenza vaccination from 65 to 50 years of age.54,55 The results of this cost-benefit analysis suggest that substantial health and economic benefits might be realized from vaccinating all working adults against influenza, especially when immunization occurs at the work site or other efficient and low-cost settings.
Accepted for publication August 24, 2000.
I thank Martin I. Meltzer, PhD, Centers for Disease Control and Prevention, Atlanta, Ga, and Yvonne Jonk, PhD, Center for Chronic Disease Outcomes Research, Veterans Affairs Medical Center, Minneapolis, Minn, for their helpful comments and suggestions.
Corresponding author: Kristin L. Nichol, MD, MPH, MBA, Medicine Service (111), Veterans Affairs Medical Center, One Veterans Drive, Minneapolis, MN 55417 (e-mail: email@example.com).