Martin I. Meltzer, Ph.D., Nancy J. Cox, Ph.D., Keiji Fukuda, MD
Background paper
Date of this draft: April 30, 1999
Correspondence to:
Martin Meltzer, Ph.D.
National Center for Infectious Diseases
Centers for Disease Control and Prevention
Mailstop C-12 Ph: 404-639-4643
1600 Clifton Rd. Fax: 404-639-3039
Atlanta, GA 30333 e-mail: QZM4@CDC.GOV
Note: This paper is a companion to the following published paper:
Meltzer MI, Cox NJ, Fukuda K. The economic impact of pandemic influenza in the United States: Implications for setting priorities for intervention. Emerging Infectious Diseases 1999 (in press).
This paper provides additional information on methods, data, and results and is intended to supplement the above referenced published paper.
Table of Contents
Introduction
Methods
Non-high risk groups
High risk groups:
Comparisons with other
data sets
Non-high risk:
High risk:
Comparisons with other
data sets:
Non-high risk
groups:
High risk groups:
Comparisons with other
data sets: Hospitalizations and outpatient visits:
Results
Discussion and conclusions
References
Tables
Table 1: U.S. population estimates
Table 2: Two scenarios of age distributions
Table 3: Variables used
Table 4: International Classification of Diseases
Table 5: Average age of patients
Table 6: Average inpatient and outpatient charges
Table 7: Number and frequency of outpatient visits
Table 8: Average age of patients and costs
Table 9: Average payments (standard deviations) for drug claims
Table 10: Pearson correlation coefficients1
Table 11: Inputs used to calculate the economic impact (direct and
indirect costs) of an influenza pandemic in the U.S. (1995 U.S.$)
Table 12: Two scenarios for the cost of vaccination
Table 13: Two scenarios for vaccine effectiveness
Table 14: Four scenarios
Table R1: Burden of impact
Table R2: Distribution, by age group, of deaths
Table R3: Economic impact (direct and indirect costs) of influenza
pandemic
Table R4: Proportion of costs
Table R5: Proportion of total costs attributable to deaths by risk
category
Table R6: Total net value of vaccinating against an influenza pandemic
(Age distribution scenario A, "high" level of vaccine effectiveness)
Table R7: Total net value of vaccinating against an influenza pandemic
(Age distribution scenario B, "high" level of vaccine effectiveness)
Table R8: Total net value of vaccinating against an influenza pandemic
(Age distribution scenario A, "low" level of vaccine effectiveness)
Table R9: Total net value of vaccinating against an influenza pandemic
(Age distribution scenario B, "low" level of vaccine effectiveness)
Table R10: The annual "insurance premium" for planning,
preparing and practicing to respond to the next influenza pandemic:
Table R11: Setting priorities:
Figure legends
Figure 1: Frequency of outpatient visits per patient for influenza,
pneumonia and acute bronchitis: 1993-95
Figure 2: Impact of influenza pandemic in the United States: Mean,
minimum, maximum, 5th and 95th percentiles of total deaths and hospitalizations for
different gross attack rates
Figure 3: Impact of influenza pandemic in the United States: Mean,
minimum, and maximum of total outpatients and those ill (but not seeking formal medical
care) for different gross attack rates
Figure 4: Sensitivity analysis: Mean net returns to vaccination, by age
group, for different death rates, vaccine effectiveness, and percentage compliance:
Non-high risk patients
Figure 5: Four options to respond to an influenza pandemic: Mean net
economic returns
"A thousand years in the making, the religion of technology has become the common
enchantment, not only of the designers of technology but also of those caught up in, and
undone by, their godly designs. . . . This popular faith, subliminally indulged and
intensified by corporate, government, and media pitchmen, inspires an awed deference to
the practitioners and their promises of deliverance while diverting attention from more
urgent concerns. . . . Pleas for some rationality, for reflection about pace and purpose,
for sober assessment of costs and benefits - for evidence even of economic value, much
less larger social gains - are dismissed as irrational. From within the faith, any and all
criticism appears irrelevant, and irreverent."
Noble, D. The religion of technology. New York, NY: Alfred Knopf, 1997, p.207.
Introduction
Although there has been a number of authors have examined or reviewed the economics of influenza vaccination (6,20,22-26,35) only one previous study (5) appears in the literature that examines the economics of a planned intervention aimed at reducing the impact of an influenza epidemic in the United States. The overall objective of this study is to examine the possible economic impact of the next influenza pandemic in the United States, and then use those results to analyze the costs and benefits of interventions designed to reduce the impact. These estimates can then be used in the development of national and state-level plans to respond to an influenza pandemic.1
The model can also be used to examine the economics of various strategies and options that might be available for the use of influenza vaccines.
Methods
Objectives
The specific objectives of the modeling effort are to:
1) provide a range of estimates regarding the number of deaths, hospitalizations, outpatients, and those ill, but not seeking medical care;
2) provide a dollar estimate of the impacts;
3) estimate the potential net present value of some possible vaccination strategies;(2)
4) evaluate the impact of using different criteria (e.g., death rates, economic returns to vaccination) to set vaccination priorities;
5) assess the economic impact of various scenarios regarding the total number of doses of vaccine administered, and the distribution of vaccine among different age and risk groups; and,
6) calculate an "insurance premium" that could reasonably be spent each year on planning, preparedness and practice (the 3 P's).
The intent is not to provide "the" estimate of impact, but rather to examine the effect of altering a number of variables, and evaluating how the results may effect key decisions. For example, if influenza-related deaths rates among 20-64 years old are assumed to range between 0.0675 - 0.15 per 1,000 persons, would it make economic "sense" (i.e., generate a positive net present value) to vaccinate everybody in this age group if a pandemic had an overall (gross) attack rate of, say, 15 percent? If not, would the results change if, say, the maximum death rate were doubled to 0.30 per 1,000? If doubling still does not generate a positive net present value for a "vaccinate all" strategy, then decision makers would be aware that the decision to vaccinate all 20-64 years of age would rest on a valuation of the intangibles, such as the reduction in fear of death due to influenza, and the value of human life above and beyond the economic value of lost productivity.
Modeling approach
Mathematically modeling the spread of, and numbers affected by, influenza is a difficult task. Differences in virology, lack of understanding of how influenza is actually spread in a community, and lack of adequate population-based data are some of the factors that have hampered efforts to produce realistic estimates of the numbers of cases that may be caused by the next influenza pandemic.8 Therefore, in the face of a great deal of uncertainty regarding the possible impact of an influenza pandemic, we used a Monte Carlo simulation approach. In Monte Carlo simulations, uncertainty is explicitly allowed for by using pre-specified probability distributions to describe the range and frequency of probable values of key variables (9-11). The model is run for several iterations, often 1,000 or more, and during each iteration values for each variable are drawn from their probability distributions. The results from all the iterations are then pooled and descriptive statistics (e.g., average, median, mode, 5th and 95th percentiles) can be calculated. For this paper, the impact of some variables, such as attack rates, cost of vaccine, and numbers effectively vaccinated, were examined at pre-determined intervals over fixed ranges, with values for other variables chosen from pre-determined probability distributions.
Gross attack rates
We defined gross attack rate as the number of clinical cases of illness (i.e., not infections) caused by influenza per unit population. Persons who become infected but show no symptoms or only very mild symptoms, such as a headache or mild nausea, are deemed not to have an economically important case of influenza (although such infections may have important epidemiological consequences). Because nobody can predict with any great certainty the attack rate of a pandemic, we modeled a range of attack rates, from 15 to 35 percent, in steps of 5 percent. The number of cases generated by a given attack rate was distributed among the U.S. population first by age and then by "high risk" status (see later).
Age group distribution of number of cases
The U.S. population (1) for 1997 was categorized into 3 age groups, 0-19 years of age, 20-64 years of age, and 65 years of age and older (Table 1). Using only three age groups simplifies modeling, and the oldest age group matches a defined "target" group for vaccination during inter-pandemic years (2). Since the actual age distribution of cases during an influenza pandemic is unknown, we calculated two age-related distributions of cases, or scenarios (Table 2).
Percentages of high risk cases
There are a proportion of persons who, because they have a pre-existing medical condition, are deemed as being at a higher risk of contracting an influenza-related illness with a serious health outcome (defined later). For the total U.S. population, we used lower and upper age-weighted averages of 15.4 and 24.8 percent (Table 2). These estimates are similar to the 22.5 percent figure quoted by Schoenbaum et al. (5), and the 19.6 percent for 1970-1978 used by the Office of Technology Assessment (OTA) study (6).
For age-specific estimates, we used lower and upper estimates of 6.4 and 11.1 percent for the 0-19 years old and a lower estimate of 14.4 percent for the 20-64 years old (Table 2). These estimates were obtained from the Working Group on Influenza Pandemic Preparedness and Emergency Response (GRIPPE, unpublished data). In comparison, a study in Oregon found that between 2.5 and 6.4 percent of 0-14 years old had one or more high risk conditions (7). The upper limit for the 20-64 years old and the lower and upper estimates for the 65 years and older were obtained from expert opinion (Table 2). Note that both the Advisory Committee on Immunization Practices (ACIP) (2) and GRIPPE (unpublished data) categorized all person 65 years and older as 'high risk." This categorization, however, is more to indicate high priority targets for inter-pandemic vaccination as opposed to describing the numbers of persons in that age group who are at higher risk of contracting an influenza-related illness with a serious health outcome.
Total numbers of cases
For each of the five assumed gross attack rates, the total number of cases were distributed among the age groups, using the two distribution scenarios (Table 2). Then, the number of cases in each age group were divided into "non-high" and "high risk" groups using the lower and upper estimates of age-defined percentages of high risk persons in each age group (Table 2). This strategy resulted in a total of ten different scenarios describing the possible number of cases resulting from an influenza pandemic.
Health outcomes: Four categories
Health outcomes were divided into four categories: deaths, hospitalizations, outpatients, and those ill, but not seeking medical care. The last category included only those whose illness would cause an economic impact, such as staying away from work for a half-day or more. Those who become ill, but do not stop or significantly reduce their daily activities, are not considered to have had an illness of any notable economic impact. For each age and risk group, a range of rates for each of the four outcomes were obtained from the literature or expert opinion
Health outcomes: Deaths
Non-high risk groups: For lower and "most likely" estimates, age-weighted average excess death rates were calculated using Serfling et al.'s data (16). Lower estimates were obtained the lowest age-weighted rate found in the four epidemic periods studied, while the "most likely" estimate was the average of the age-weighted rates studied (16) (Table 3). The same data set also provided a maximum value for the 65 and older age group (Table 3). Note that, Serfling et al.'s estimates are effectively the combination of the rates for both high and non-high risk. Thus, using those rates for non-high risk will result in rates higher than might actually occur, especially in terms of lower limits. The upper limits for the 0-19 years old were the age-weighted average rate for the 0-4 and 5-24 year olds rate per 1,000 cases, assumed by Schoenbaum et al. (5), converted to rates per 1,000 general population (see later for notes on conversion methodology) (Table 3). Data from Serfling et al. were used to set the upper limit for the 20-64 years old (Table 3).
High risk groups: In OregoN (17) death rates for those aged 45-64 years ranged from 0.1 per 1,000 (with one high risk condition) to 5.72 per 1,000 (for those with 2 or more high risk conditions). Those aged 65 and older had rates of 2.76 per 1,000 (one high risk condition) to 5.63 per 1,000 (2 or more high risk conditions). These rates were used as ranges for 20-64 years old and 65 years and older, respectively (Table 3). Data regarding the death rate among 0-19 years old with high risk conditions are scarce. The Office of Technology (OTA) report (6) estimated the risk of pneumonia-related deaths among those 44 years and under with medically attended heart conditions was 1.0 per 1,000 of the general high risk population. Among those 44 years and under, those with a high risk condition (i.e., medically attended heart condition) were calculated to have a relative risk of 9.09 of dying from pneumonia when compared to those without a high risk condition (6). It was assumed, therefore, that the rate of death per 1,000 high risk population among the 0-19 years old was 9 times greater than the rates used for the non-high risk population of the same age.
Comparisons with other data sets: Kavet (20), in examining the impact of epidemics in 1962-63, 1965-66 and 1968-69 (with 1966-67 as the non-epidemic reference year) estimated non-age, non-risk specific mortality rates of 0.14-0.249 per 1,000 general population. Glezen (3) reported similar rates for 65 years and over, and an age-weighted (1) average rate of 6.75 per 100,000 for 20-64 years old. For 45-64 years old, the death rate in Oregon 0.02 per 1,000, and 0.09 per 1,000 for 65 and older (17). In France, from 1980-81 to 1989-90, the average death rates among those aged 75 years and older ranged from 4.81 to 0.28 per 1,000 (23). All of these estimates are encompassed in either the rates used for the non-high or high risk groups (Table 3). When considering the economics of vaccinating elderly patients in nursing homes, Patriarca et al. (25) assumed a probability of death after infection, with no hospitalization, of 0.16. This is approximately 5 times greater than the rate per 1,000 cases for 65 years and older used here (Table 3), and far greater than most death rates reported for the elderly. The difference can probably be attributed to the fact that almost the entire population in a nursing home is likely to have one or more high risk conditions. In 1974-1982, in Houston, Texas, (18) there was a fairly steady age distribution of percentage of total influenza-related deaths. Deaths among the 65 years and older accounted for an average of 62.5 percent (range: 55-75 percent) of all influenza-related deaths. Kavet (20) reported similar percentages for the three epidemics which he studied. The 20-64 years old had 29.6 percent (range: 20-33 percent) and 0-19 years old accounted for 7.9 percent (range: 5-12 percent) (18).
Health outcomes: Hospitalizations
Non-high risk: The excess hospitalizations due to influenza were obtained from Mullooly and Barker (7,12), with data from Schoenbaum et al. (5) used to define upper limits. The latter study used estimates obtained from a Delphi panel, but their estimates appear to include "background," non-epidemic hospitalizations as opposed to excess hospitalizations. The lower limits for the 0-19 years old age group were derived from the combined male and female excess rates for 5-14 years of age (7), while the "most likely" rates are the rates for the entire 0-14 years of age (7) (Table 3). The upper limit for the 0-19 years old were the age-weighted average rate per 1,000 cases (see later for notes on conversion methodology) for the 0-4 years and 5-24 years old age groups as assumed by Schoenbaum et al. (5) (Table 3). The lower limits of excess hospitalizations for the 20-64 years age group were taken from the average 1968-69 and 1972-73 epidemic excess hospitalization rates for the 15-44 and 45-65 year old age groups in Oregon (Table 3) (12). The upper limits for the 20-64 years old were calculated using the data set (5) and methodology employed to calculate the upper limits for the 0-19 years old (Table 3). Mullooly and Barker's data were used to obtain the upper and lower limits for the 65 and older age group (12) (Table 3).
High risk: The same methodology and data sets used to set hospitalization rates for non-high risk patiients were also used to set the rates of hospitalizations for the high risk groups (Table 3).
Comparisons with other data sets: It is noted that the Office of Technology Assessment report (6) used the Mullooly and Barker data sets (7,12). Glezen (3), without categorizing patients by risk, reports rates of hospitalizations for acute respiratory disease as approximately 4.2, 0.8, 0.5 per 1,000 for the <5, 5-9, and 10-19 year age groups, respectively. This produces a population-weighted1 average for the 0-19 year olds of 1.5 per 1,000 population. This rate is lower than the upper limit used here for 0-19 years old with non-high risk (Table 3). From the same data set (3), the age-weighted hospitalization rate for 20-64 years of age was 1.1 per 1,000, and the rate for 65 years and older was 7.5 per 1,000.3 These rates are lower than the upper rates used in this study for those categorized as "high risk" (Table 3). Barker's (21) estimates of national excess hospitalization rates, averaged over 5 epidemics from 1970-78, are within the ranges used here for the "non-high risk" group (Table 3). The only exception is the rate of 3.7 per 1,000 found for 65 years and older, which is higher than given in Table 3 for non-high risk, but is below the range used for "high risk."
The highest recorded estimates of hospitalizations appear in a study by McBean et al (19). They calculated the rate of hospitalizations for influenza and influenza-related conditions for a sample of Medicare patients (i.e., 65 years and older) from 1989-1991. Unlike the studies referenced above, which compared rates from epidemic to non-epidemic years, McBean et al. compared hospitalization rates of influenza "seasons" (e.g., December to March) to an interim period (e.g., May - September). They found excess rates (using their definition of excess) of up to 14.25 per 1,000 for patients hospitalized with influenza and pneumonia as the primary diagnosis, and rates of up to 19.4 per 1,000 for patients hospitalized with influenza or pneumonia as any part of the diagnosis. These rates are approximately 4 times greater than the rates that we used for "non-high risk" 65 years and older, but similar to the rate we used of 13 per 1,000 for those 65 and older categorized as "high risk" (Table 3).
Health outcomes: Outpatient (ambulatory care)
Non-high risk groups: An age-weighted average rate for 0-19 years old was obtained using the rates reported by Glezen (3) for <5, 5-9, 10-14 and 15-19 years old (Table 3). The rate of medically attended illnesses for the 20-64 year old age group were derived from the excess contact rates for 45-64 year age group in Oregon, where excess was defined as the increased visits due to the 1968-69 and 1972-73 epidemics compared to the 1970-71 period (12) (Table 3). The same data source (12) provided the excess outpatient visits for those 65 years and older (Table 3).
High risk groups: Although some studies have considered hospitalization rates by category of risk (see later), there are no studies that consider outpatient (ambulatory) visits by risk category. Therefore, it was decided to calculate the rates by multiplying all the rates used for the non-high risk groups by an arbitrarily defined factor of 1.75. It was found, by trial and error, that any factor noticeably higher than that (e.g., 2.0), resulted in more than 100 percent of the high risk population requiring outpatient care.
Comparisons with other data sets: Hospitalizations and outpatient visits: Kavet (20), in studying the epidemics of 1962-63, 1965-66 and 1968-69 (with 1967-68 as a non-epidemic reference year), found that between 41.7 and 51.3 percent of all cases required some medical attention (e.g., outpatient visit or hospitalization). These percentages were not broken down by age groups or risk categories. However, these total average rates are similar to, or encompassed by, to the sum of outpatients and hospitalizations rates used in this study (Table 3).
Health outcomes: Ill (no formal medical care)
Besides those that will die, or require hospitalization, or require outpatient (ambulatory) visits, there will be a number of persons who will become clinically ill but not seek formal medical care (e.g., a visit to a physicians office). These clinically ill people will, however, generate economic impact, primarily in days lost from work or school, and the use of over-the-counter medications. Note that this definition of clinically ill excludes those that will develop mild symptoms (e.g., nausea, headache, low-grade fever), but essentially still continue their daily activities.
The numbers of persons in each age group who will incur a clinical case that will have economic impact was calculated by using the following formula:
Number illage = (U.S. populationage x gross attack rate) - (number deadage + number hospitalizedage + number requiring outpatient servicesage)
where age = number of persons in each of the three age groups (for U.S. populationage, see Table 1).
Converting from general population rates to rates per clinically ill population
The data reported for outpatient (ambulatory) visits, hospitalizations and deaths are presented in rates per 1,000 general population (Table 3). These summarize the risk of a particular health outcome for any member of the general population, and the rates reflect the impact of a particular attack rate(s). Presumably, a given rate per 1,000 general population changes with attack rate. That is, the rate of, say, deaths per 1,000 non-high risk persons aged 20-64 years will change with changes in attack rate. We are not aware of any published study that contains a statistical analysis of the correlation between attack rate and the rate of a given health outcome. This is lack of data is problematic, as one of the goals of this paper is to examine the potential impact of pandemic influenza assuming a range of pre-set gross attack rates (see earlier). Thus, the data regarding health outcomes must be converted to rates per unit population with clinical illness. That is, for the population that becomes clinically ill, the rates must reflect the probability of outpatient (ambulatory) visits, hospitalizations and deaths.
To convert rates based on a general population to rates based on a clinically-ill population, the following formula was used:
Rate per unit clinically ill population (Xclinically ill) = Rate per unit
general population (Xgeneral)/conversion factor.
Where Xclincially ill is the rate of interest per 1,000 clinically ill
persons (e.g., hospitalizations for influenza related illnesses among 0-19 years old per
1,000 clinically ill 0-19 years old). Xgeneral is the rate for the
outcome of interest per 1,000 population, ill and not ill combined (e.g., hospitalizations
for influenza illnesses per 1,000 population aged 0-19 years).
The key to applying the above equation is determining the conversion factor. One logical candidate for the conversion factor is the attack rate that created the rate per unit general population (Xgeneral). Such data are rare, because it implies that there would be active surveillance among the general population, asking the populace if they have suffered any influenza-like illnesses during a specified time period. In order to disregard illnesses that patients may report as influenza, but are actually due to non-influenza pathogens, a sub-sample of the population would have to be tested for evidence of influenza infection (e.g., take nasal swabs and culture). Further, because the virulence of influenza can change dramatically from year to year (13), such a study would have to be conducted over several years. Lacking such data, we used the equation above to derive rates per clinically ill population using upper and lower limits of attack rates calculated for the general population from Texas for 1976-1984 (3), and Seattle for 1975-1979 (3,14) (Table 3). These data were the closest in time and geography to the data used to calculate the population based data rates (Table 3). Note that we assumed that the conversion calculations was the same for both non-high and high risk groups. For comparison, Kavet (20) provides data allowing the conversion of mortality rates per general population to case fatality rates. His non-age, non-risk specific conversion factors for deaths for the three epidemics studied were 0.11, 0.14, and 0.26 equivalent to 11%, 14%, 16%, respectively). The last two estimates are within the ranges used here (Table 3), whilst the conversion factor of 0.11 is very close to the lower limit of 0.12 used here for the age groups 20-64 years and 65+ years (Table 3).
Resource use: Direct medical costs: Database
The direct medical costs associated with hospitalizations, outpatient visits and drug purchases related to outpatient visits were obtained from the MarketScan, proprietary health insurance claims database (The MEDSTAT Group, Ann Arbor, MI). The database available to us contains health insurance claims data from approximately 4 million covered lives, most of whom obtain coverage as part of their employment with large companies. These data are collected from over 100 health care payers, servicing 45 major metropolitan areas. For 1995, the database has 287,000 inpatient records and 42.8 million outpatient claims. To extract direct medical costs from the database, we used the same International Classification of Disease, ninth revision (ICD-9) codes as used by McBean et al. (19) (Table 4). Simple descriptive statistics describing the extracted data were calculated using either spreadsheet software (Excel, Microsoft, WA) or a statistical package (SAS, SAS Institute, NC).
Resource use: Standardizing year of costs to 1995
To calculate descriptive statistics over the three years of available cost data, data from 1993 and 1994 were inflated to equivalent 1995 prices. The medical care portion of the consumer price index was used to inflate prices to 1995 equivalents. The factor for 1993 was 1.095 and for 1994 the factor was 1.045 (1).
Resource use: Valuing death due to influenza
It was assumed that the average age of death was the mid-point of each of the three age groups (i.e., 9, 35, and 74 years). For each average age of death, a life lost was valued at the average present value, discounted at 3%, of future expected life-time earnings and housekeeping services (27). This average was calculated using wages of full-time employees, weighted for gender composition of the U.S. workforce, and assumed a 1% per annum growth rate in productivity (27). The original estimates were converted from 1990 dollars to 1995 dollars by multiplying by a factor of 1.07, which represents the average increase from 1990 to 1995 (in constant 1992 dollars) in total and per capita personal income (1).
Resource use: Inpatient care associated with death
It was assumed that the majority of deaths would occur after some use of hospital resources (i.e., death would occur after a stay in a hospital). For each of the three age groups the average age, length of stay and net payment were obtained from the medical cost database for each of the predefined ICD-9 codes used to describe influenza and influenza-related illnesses (Table 4). Following the methodology of McBean et al.(19), the data were then subdivided into those whose principal diagnosis was either influenza, pneumonia or acute bronchitis ("Principal diagnosis only"), and those whose had these diseases coded as either the principal diagnosis or one of the first four diseases entered in the coding sheet ("All diagnoses"). Since influenza may also cause a patient to be hospitalized with a pre-existing condition (i.e., influenza "triggers" an underlying problem), the inpatient costs for those with heart-related conditions (Table 4) were also extracted from the data base. A weighted average for each age group of these three categorizations of outcomes was then calculated as follows:
Weighted average age group = (number of claims in a category age group /total claims for all categories age group) x (valueage group)
The results are presented in Table 5, and were then added to the value of a human live lost due to influenza.
Resource use: Hospitalized patients
Costs for those patients who would be hospitalized due to an influenza pandemic, but will not die, were obtained from the medical cost database for all those patients whose diagnoses (i.e., ICD-9 codes) included influenza, pneumonia or acute bronchitis (Table 4). Further, it was assumed that a hospitalized patient, who eventually recovered from a case of influenza, would also generate outpatients visits before and after hospitalization. In order to generate a composite estimate of inpatient and outpatient resources used, for each individual hospitalized, records of outpatient visits and drug claims up to 14 days prior to a admission date, and 30 days after discharge, were identified and added to individual net hospital payments. Length of stay during hospitalization was also obtained from the database, and to this number one day was added to allow for convalescence for the 0-19 and 20-64 years age groups. For the 65 + years age group, two days were added for post-hospitalization convalescence. Results are given in Table 6, and it can be seen that the average length of stay was 4, 7 and 8 days for the 0-17, 18-64, and 65+ year age groups, respectively. Adding the assumed post-hospitalization period of convalescence, the number of work-days lost are: 5, 8 and 9 days for the 0-17, 18-64, and 65+ year age groups, respectively.
The number of days of work lost due to hospitalization extracted from the MarketScan database can be compared to estimates used by Kavet (20). Kavet, using data from the National Health Survey, assumed that average length of hospital stay was 9 days, with 2.9-3.8 days of bed disability per case, and 5.0-6.2 days of restricted activity per case. These loses resulted in 3.2-3.4 days lost from work per case and 2.8-4.7 days lost from school per case. In a small study (22) covering 130 unvaccinated textile factory employees, 93 workdays (average: 0.71 workday/ employee) were lost due to influenza-like illness. Also a factor is the idea that, during a pandemic, the number of people requiring some form of hospitalization will be so great that hospitals will be forced to increase patient turnover by reducing average length of stay. Thus, average length of stay obtained from data collected during inter-pandemic years may over-estimate the amount of hospital resources used per patient during a pandemic.
Resource use: Outpatients
The direct medical resources used by an ambulatory patient includes the average number of physician visits per case of influenza, the co-payment associated with each visit, the net payment made by the insurance company, and any drug claims (see earlier). Records of outpatient claims matching the pre-selected ICD-9 codes for influenza, pneumonia and acute bronchitis (Table 4) were used to calculate the distribution of visits-per-case to a physician for each calendar year in the database (1993-1995) (Figure 3). Then, to compare results from a calender year to those associated with part of an influenza "season," the average number visits per patient for October to December, 1993 were calculated. To allow for errors in the data set (e.g., several patients entered with same identification number), the means were calculated for three different frequencies, each with a pre-specified maximum (truncated) number of visits (< 20, < 7, < 3 visits) (Table 7). The same data also provided the average age of patient making an outpatient visit, the average co-payment, and the average net payment to physician paid by a third party payer for services provided (Table 8). The number of work days lost by a patient who requires outpatient care was based on the numbers used by Kavet (20) (see later).
Resource use: Drug claims related to outpatient visits
Patients who had one or more outpatient visits in either 1994 or 1995 for influenza, pneumonia or acute bronchitis (Table 4), were matched to drug claim data for 1994 and 1995 (Table 7). Some drug claims identified in this manner may not be related to the outpatient visits for influenza, pneumonia or acute bronchitis. It was assumed, however, that the sample sizes were sufficiently large so that the average costs would "reasonably" represent the relevant drug costs associated with influenza and influenza-related outpatient visits. The average number of drug claims per outpatient visit for each age group was calculated by dividing the number of claims by the number of visits (Table 9).
Resource use: Ill, no medical care sought
Patients who have a clinical illness due to influenza but do not seek medical care still use resources, principally in the form of time off work, and perhaps some non-prescription (over-the-counter) medications. The value of non-prescription medications was assumed to be $2 per case, and the number of days lost was based on the number used by Kavet (20) (see later).
Resource use: Correlations between age, length of hospital stay, and net payment
Person correlation coefficients were calculated to test for potentially statistically significant relationships between age, length of stay in hospital, and net payment for hospitalizations. Significant relationships of these variables would indicate the need to incorporate such relationships into the Monte Carlo model. The results (Table 10), however, showed that the only correlation of practical importance was between net payment and length of stay (R = 0.64 or 0.61, P >0.0001). Although the correlation coefficients between age of patient and the other two variables were statistically significant, the coefficients were judged to be too small to be of much practical significance (Table 10).
Calculating the economic impact of an influenza pandemic
Using the data reported in Tables 5 through 9, and data from Kavet (20) regarding workdays lost for a person with an illness requiring outpatient visits (discussed above), economic cost, by age group, was calculated for an individual case of each of the four outcome categories (death, hospitalization, outpatients, ill no medical care sought) (Table 11). For each category of outcome, the cost per individual outcome was then multiplied by the estimated number of outcomes (see earlier), and then summed together to give a total cost of a pandemic, assuming no large-scale effective intervention occurs.
Returns to vaccination: Cost of vaccination
The total cost of vaccinating against an influenza pandemic is dependant upon the cost of vaccination, the target group(s) selected, and compliance rate(s). We divided the U.S. population into the 6 age and risk groups defined earlier, with the cost of vaccinating each group calculated separately. Based on the compliance rate of influenza vaccination among persons aged 65 years or older in the U.S. (29), two compliance rates, 40 and 60 percent, were chosen. It is noted that while the U.S. average rate for persons aged 65 years and older is 58 percent, the average among the states ranges from 44 to 70 percent (29). It was assumed that compliance rates were equal among all age and risk groups, although the model does have the capacity to allow for different compliance rates for each age and risk group.
The cost of vaccination includes the cost of the vaccine, the cost of administering the vaccine, value of time spent by an individual traveling to and from the place of vaccination, travel costs, and the costs of treating adverse side effects. The latter includes costs associated with treating effects of Guillian-Barré Syndrome (GBS). The incidence rates and costs of treatment of an individual case of mild-side effects, anaphylaxis, and GBS were obtained from a previous study (6).
A number of factors will effect the total cost of vaccination, and include the number of doses required,3 the cost of rapid production of a larger-than-usual number of doses, the rapid delivery and correct storage of doses at vaccination sites around the country, the cost of administration, and the length of time each patient may have to wait in order to receive a dose (i.e., long lines of waiting patients. In order to illustrate the impact of different costs upon the net value of vaccination, two costs of vaccination were assumed ($18 and $$59 per vaccinee), to which were added specific estimates of the costs of treating side effects (Table 12). The two total costs of vaccination modeled were $21.26 and $62.26 per vaccinee (Table 12). ). The value of a patient's time was based on the value of an 8 hour day "unspecified" work day, valued at $64 per day (27). The cost of travel ($4) was assumed. As probable costs of vaccination in a pandemic become better defined, the model has the capacity to use any other defined estimate.
Returns to vaccination: Vaccine effectiveness
Economic returns to vaccination are dependent upon vaccine effectiveness to prevent each of the four outcomes modeled (death, hospitalization, outpatient visits and ill with no medical care sought). Although there are some data from controlled trials regarding the effectiveness of influenza vaccination in preventing clinical outcomes, those data invariably refer to experiments conducted during inter-pandemic periods. There are no published data from controlled trials measuring the effectiveness of an influenza vaccine during a pandemic situation. Given that a pandemic will most likely be caused by a new subtype of influenza, vaccine effectiveness under such situations is speculative. Therefore, two scenarios of vaccine effectiveness by age group and health outcome were constructed, labeled "high" and "low" levels of effectiveness (Table 13). Within each age group it was assumed that there would be no difference in vaccine effectiveness between those at "high risk" and those at "non-high risk" (Table 2). The model has the capacity to evaluate the outcomes using different values for vaccine effectiveness for each age group and health outcome.
Returns to vaccination: Net returns of vaccinating against influenza
One important measure of the economics associated with vaccinating against an influenza pandemic is the net returns, in dollars, to vaccination. This was calculated using the following formula for each age and risk group:
Net returnsage, risk group = Savings from outcomes averted in populationage, risk group - cost of vaccination of populationage, risk group
Where:
Savings from outcomes avertedage, risk group = Soutcomes (number with outcomedeath,hospitalization,outpatient,ill,no medical care before interventionage,risk group x complianceage, risk group x vaccine effectivenessoutcome x $value of outcomedeath,hospitalization,outpatient,ill,no medical care prevented)
and;
Cost of vaccinationage, risk group = $cost/vaccinee x populationage,risk group x complianceage,risk group
The numbers of persons within a given age and risk group with one of the four outcomes before any intervention are obtained by running the portion of the model that estimates the impact of a pandemic (Tables 1 -3 ). The variables used to define the value of a given outcome are given in Table 11, the cost of vaccination in Table 12, and vaccine effectiveness in Table 13. Compliance rates were initially set at 40 and 60 percent. As discussed earlier, most of these variables can be adjusted in age or risk or both specific groups. The calculations were also made for the two different age distributions (Table 2).
Returns to vaccination: Sensitivity analyses
Because the model was run using ranges for the attack rates and other input variables (see Tables 1-3, 11, 12, 13), the results can be considered as a set of sensitivity analyses. To illustrate the importance of the death rate in determining economic outcomes, further sensitivity analyses were conducted by altering the death rates (Table 3) . The average death rates for the non-high risk groups was altered (reduced) so that they were 0.25 or 0.50 of the death rates used in the original analyses (Table 3).
Implications for policy: "Insurance premiums" and the three P's
The analyses of the costs and benefits of a vaccination-based intervention during a pandemic implicitly assumes that such an intervention could occur if needed. This paper is part of a planning process whose end product is intended to improve the likelihood of a well-organized and effective response to a pandemic. The plan includes improvements in surveillance systems, ensuring sufficient supply of vaccine to vaccinate high-priority groups (and possibly the entire U.S. population), investigating the feasibility of liability programs for vaccine manufacturers, research to improve detection of new influenza sub-types, research into the acceleration of the availability of new and existing vaccines and antiviral agents, developing a communications network that can rapidly disseminate a wide variety of information in different media forms to different target audiences, and developing emergency preparedness plans to ensure that there will be adequate medical care and that essential community services will be maintained (30). Meeting these goals can be described as enacting the 3 P's (planning, preparedness, and practice).
A question arising from such a list of goals is: What is a "reasonable" amount to spend each year to ensure that these goals are reached? Similar to an earlier study on the economics of intervention programs,31 the annual cost of paying for the 3 P's can be considered as an "insurance premium." The purpose of such an insurance premium is to ensure that the 3 P's have been accomplished and that a planned vaccination intervention can actually take place. An annual "insurance premium" to pay for the 3 P's can be calculated as follows: (32)
Annual "insurance premium" = net returns from an intervention x annual probability of pandemic occurring
The net returns from an intervention are those calculated using the formulae described earlier. The actual returns will be, of course, impacted by a number of variables that cannot be controlled. For example, the gross attack rate before intervention (the percentage of the total U.S. population that would become clinically ill) is beyond human control, as is the probability of a pandemic. Even vaccine effectiveness may have an element of chance. Thus, "premiums" were calculated for a range of gross attack rates (15, 25 and 35 percent), three different probabilities of a pandemic occurring in a given year (1 in 30, 1 in 60, and 1 in 100 years), two different scenarios regarding vaccine effectiveness (Table 13), and two possible costs of vaccination (Table 12).
Implications for policy: Using different criteria to set vaccination priorities
If an influenza pandemic should occur, it is possible that initially there may be a limited supply of influenza vaccine. Even if there were a sufficient supply of vaccine to ensure that the entire U.S. population could be vaccinated, it will take some time to actually administer the vaccine to all, especially if two doses are required to ensure a "satisfactory" immune response. These factors raise the question of who should receive priority for vaccination, at least until vaccine supplies are more plentiful. For example, an argument could be made that all health care workers (e.g., nurses, practicing physicians, paramedics, etc.), and essential service providers (police, fire, phone, electric and gas workers, air traffic controllers, etc.) should be among the first to be vaccinated. However , the logic behind using essential services as "the" criteria for setting priorities will not cover the majority of the population.
To illustrate the implications of using various criteria to set priorities, three different criteria were chosen to create sample priority lists: Net returns to vaccination, percentage of total deaths, and risk of death. A priority list constructed using net returns to vaccination will give the highest priority (top rank) to the age-risk group that "produces" the highest net return to vaccination. Similarly, the highest percentage of total deaths and death rate will give the highest priority to the age-risk group with the largest number of deaths and the highest death rate, respectively. The different criteria imply a different set of values would be used in deciding who receives top priority for vaccination. Society must debate what is the main goal in deciding who is to have top priority during a pandemic plan: Is the goal to prevent deaths, regardless of age and position in society, which implies use of the death rate as the deciding criterion? Or, does society wish to focus on trying to prevent total numbers of deaths (i.e., use percentage of total deaths as the criterion)? Minimizing economic impact and ensuring that society does not collapse implies that net returns to vaccination should be used to set priorities.
Implications for policy: Four scenarios regarding vaccine availability and distribution
The model can be used to compare the potential benefits of different plans each designed to ensure that various amounts of vaccine will be available. To illustrate this capability, and to provide policy makers with some valuation associated with different options, four scenarios or options with different levels of guaranteed amounts of vaccine were constructed. Each option contains assumptions concerning the targeted population (Table 14). Each option was estimated using age distribution scenario A (Table 2), two gross attack rates (15 and 25 percent), and two estimates of the cost-of-vaccination ($21.26 and $62.26, Table 12). Further, for each option, vaccine effectiveness was assumed to result, for all age groups, in a 50 percent reduction in deaths and hospitalizations, and a 40 percent reduction in outpatient visits and illnesses not requiring formal medical care. Obviously, the model can be recalculated using different assumptions, or new scenarios can be constructed. Each option has potentially different implications regarding the level of involvement from Federal, state and local government agencies. The four options are:
Option A: Similar to current Advisory Committee on Immunization Practices (ACIP) recommendations, with production and use similar to current, intra-pandemic recommendations (2). Assumed approximately 77 million vaccinees.
Option B: Number of vaccinees as outlined in Scenario A plus an additional 20 million essential service providers (5 million health care workers + 15 million other service providers).
Option C: Aim to achieve a 40 percent coverage in each age and risk group, regardless of occupation.
Option D: Aim to achieve 60 percent coverage in each age and risk group, regardless of occupation.
Results
Deaths
Figure 2 shows the mean, 5th, 95th, minimum and maximum number of estimated deaths and hospitalizations for the two age distribution scenarios (Table 2). For case-age distribution Scenario A, the number of deaths due to an attack rate of 15 percent was approximately 89,000 (5th percentile = 55,000; 95th percentile = 122,000), and at an attack rate of 35 percent a mean of approximately 207,000 deaths was calculated (5th percentile = 127,000; 95th percentile = 285,000) (Figure 2). At an attack rate of 15 percent, the minimum and maximum deaths were approximately 44,000 and 135,000, respectively. At an attack rate of 35 percent, the minimum and maximum deaths were approximately 102,000 and 315,000, respectively (Figure 2).
For age distribution scenario B, which has the higher percentage of cases in the 65 and older and 0-19 years old age groups (Table 2), the number of deaths at any given attack rate was approximately 43 percent higher than that calculated for Scenario A. For example, at an attack rate of 15 percent, the mean number of deaths was approximately 128,000 (5th percentile = 75,000; 95th percentile = 181,000), with a minimum and maximum of approximately 56,000 and 202,000. At an attack rate of 35 percent, the mean number of deaths was approximately300,000 (5th percentile = 175,000; 95th percentile = 422,000), with a minimum and maximum of approximately 132,000 and 472,000.
Hospitalizations
For age distribution scenario A (Table 2), the calculated number of hospitalizations ranged from a mean of approximately 314,000 (5th percentile = 210,000; 95th percentile = 417,000) at a gross attack rate of 15 percent, to a mean of approximately 734,000 (5th percentile = 441,000; 95th percentile = 973,000) at a gross attack rate of 35 percent (Figure 2). The estimates of hospitalizations for age distribution scenario B were approximately 11 percent than those calculated for age distribution scenario A (Figure 2).
Outpatients
For age distribution scenario A (Table 2), the mean numbers of persons requiring outpatient-based care ranges from approximately 18 million at a gross attack rate of 15 percent, to 42 million at a gross attack rate of 35% (Figure 3). For age distribution scenario B, the number of outpatient cases was approximately 7 percent greater than those calculated for age distribution scenario A. (Figure 3).
Ill (no formal medical care sought)
For age distribution scenario A (Table 2), the mean numbers of those clinically ill who will not seek formal medical care, but still suffer some economic loss, ranges from approximately 20 million at a gross attack rate of 15%, to 47 million at a gross attack rate of 35 percent (Figure 3). For age distribution scenario B, the number of outpatient cases was approximately 6 percent greater than those calculated for age distribution scenario A. (Figure 3).
Burden of impact among the high risk groups
Although those in the high risk groups constitute approximately 15 - 25 percent of the total U.S. population (Table 2), they bear a disproportionate burden of the potential impact of an influenza pandemic. On average, high risk groups will account for approximately 85-90 percent of all deaths (Table R1). By age group, the high risk 20-64 years old will bear the largest proportion of deaths, accounting for approximately 41-43 percent of total deaths (Table R1). It is also important to note that, for each age group, there is a wide range between the 5th and 95th percentiles (Table R1). For example, for age distribution scenario A, although the high risk 20-64 years of age were calculated to account for a mean of 40.9 percent of all deaths, the 5th percentile is approximately 11 percent, and the 95th percentile is approximately 61 percent (Table R1). High risk groups will also account for 38-52 percent of all hospitalizations and 20-30 percent of all outpatient visits (Table R1).
Burden of death by age group
The distribution by age group of total deaths is give in Table R2. For age distribution scenario A, those aged 65 years and older ("non-high" and "high" risk combined) accounted for a mean of approximately 39 percent of all deaths (5th percentile = 28 percent; 95th percentile = 63 percent). The minimum and maximum percentage for this age group were approximately 24 and 76 percent. The 20-64 year age group was calculated, on average, to account for about 5 percent more deaths than the 65+ year old age group, with a mean of 47 percent of all deaths for age distribution scenario A (Table R2). The 0-19 age group accounted, on average, for less than 15 percent of all deaths, although a maximum of approximately 41 percent was calculated for age distribution scenario B (Table R2).
Estimated economic costs due to an influenza pandemic: No intervention
In the absence of a large-scale intervention, the estimates of the total economic impact in the U.S. of an influenza pandemic range from an average of $71.3 billion (5th percentile = $35.4 billion; 95th percentile = $107.0 billion) for a gross attack rate of 15 percent, to an average of $166.5 billion (5th percentile = $82.6 billion; 95th percentile = $249.6 billion) for a gross attack rate of 35 percent (Table R3). At any given attack rate, loss of life due to an epidemic accounts for approximately 83 percent of all economic losses, with outpatients, ill not seeking medical care, and inpatient care accounting for approximately 8, 6, and 3 percent, respectively (Table R3).
Distribution of economic impact: Direct and indirect costs
Except for hospitalizations, indirect losses (primarily lost productivity, see Table 11) account for at least 65 percent of all economic losses (Table R4). In the case of deaths and ill, no medical care sought, indirect costs accounted for 99 percent of all costs (Table R4). Direct costs accounted for 89 percent of all costs associated with hospitalizations (Table R4). Across all health outcomes, indirect costs accounted for a weighted average of 94 percent of all economic costs (Tables R3, R4). Deaths among the high risk groups account for approximately 70 percent (5th percentile = 47 percent; 95th = 83 percent) of all economic losses (Table R5).
Net value of vaccinating against an influenza pandemic
The net value of vaccinating against an influenza pandemic varies substantially by age-risk group, gross attack rate, assumed distribution of cases across age groups (Table 2), assumed vaccine effectiveness, compliance, and cost of vaccination (Tables R6 - R9). At a cost of vaccination of $21.26, and a "high" level of vaccine effectiveness, the mean net value of vaccinating against a pandemic is always positive, regardless of the value of the other variables (e.g., compliance, gross attack rate, distribution of cases by age) (Tables R6-R7). For example, under the stated assumptions, the smallest positive return occurs when vaccinating 40 percent of the 65+ year old age group at "non-high risk," with a net present value of $46 million (5th percentile = $15 million; 95th percentile = $76 million) (Table R7).
However, when assuming a "low" vaccine effectiveness (Table 13), a gross attack rate of 15 percent, and Scenario B for the distribution of cases by age (Table 2), the 5th percentiles are negative (-$20 million) for the age group 65 + years, with non-high risk (Table R9). That is, under those assumptions, there is a 5 percent or greater possibility that vaccinating persons who are non-high risk and 65 years and older will generate a net loss to society.
When the cost of vaccination is $62.26 per vaccinee, at a gross attack rate of 15 percent, vaccinating all non-high risk age groups will generate net losses (Tables R6-R9). Under these assumptions, all but two net return are negative even at the 95th percentile, regardless of level of compliance, vaccine effectiveness, and age distribution of cases. The two exceptions occur in the age group 0-19 years, and for age distribution Scenario B (Table R7). However, for even these two situations, the 85th percentiles are negative (data not shown).
When the gross attack rate is increase to 25 percent, and assuming a "high" level of vaccine effectiveness, the non-high risk 0-19 and 20-64 age groups all have positive mean net returns, although some of the 5th percentiles are negative (Tables R6, R7). At the same gross attack rate, with a "low" level of vaccinee effectiveness, all non-high risk age groups have negative mean net returns, although some 95th percentiles are positive (Tables R8, R9). For "low" vaccine effectiveness, and regardless of other variables modeled, the non-high risk 65 years and older age group still generates negative mean net returns at a gross attack rate of 35 percent (Tables R8, R9). All other age groups, in contrast, generate positive mean net returns at 35 percent gross attack rate, although some 5th percentiles are still negative (Tables R6-R9).
Note that the high risk groups, regardless of age, always generate positive mean net returns, and that, in the majority of scenarios modeled, the 5th percentiles are also positive (Tables R6-R9).
Sensitivity analyses: Impact of reducing death rates
At a vaccination cost of $21.26 per vaccinee, reducing the death rates (Table 3) to 0.50 and 0.25 of the initial values still left positive mean net returns for all non-high risk age groups (Figure 4). However, at a cost of vaccination of $62.26 per vaccinee, reducing death rates to 0.50 and 0.25 of the initial values resulted in negative mean net returns for all non-high risk age groups (Figure 4). Note that, compared to changes in the death rate, the results are relatively insensitive to increases in gross attack rate (in the range of 15 to 25 percent). For example, at a cost of vaccination of $62.26 per vaccinee, and a death rate at 0.50 of the initial rates used, increasing the gross attack rate from 15 to 25 percent still results in a negative net returns for all age groups and levels of vaccine effectiveness (Figure 4).
Implications for policy: "Insurance premiums" to pay for the three P's
The amount of money that could "reasonably" be spent each year to ensure that the U.S. has "adequately" planned, prepared, and practiced (the 3 P's) for the next influenza pandemic ranges from $48 million per year (s.d. $37 million) to $2,184 million per year (s.d. $796 million) (Table R10). The lowest estimate arises by assuming that the probability of a pandemic is equivalent to 1 in 100 years, the gross attack rate will be 15 percent, the cost of vaccination will be $62.26 per vaccinee, a compliance rate of 40 percent, and a "low" level of vaccine effectiveness (defined in Table 13). Conversely, the highest estimate occurs when it is assumed that the probability of a pandemic is equivalent to 1 in 30 years, the gross attack rate will be 35 percent, the cost of vaccination will be $21.26 per vaccinee, a compliance rate of 60 percent, and a "high" level of vaccine effectiveness (defined in Table 13).
The assumed probability of a pandemic is one of the most important variables defining the size of the "premium." For example, the premiums increase by more than 3 times when the probability of a pandemic is increased from 1 in 100 years to 1 in 30 years (for a given gross attack rate, cost of vaccination, and level of vaccine effectiveness).
Implications for policy: Using different criteria to set vaccination priorities
Table R11 provides three lists of priority groups using three different criteria - risk of death, percentage distribution of deaths, and economic returns to vaccinating against influenza. The use of the three different criteria produced notable differences. For example, when risk of death is used to set priorities, persons aged 65 + years will receive top priority (Table R11). But, when mean net economic returns to vaccination are used as the criterion, then persons age 65 + years will receive the lowest priority (Table R11). Regardless of criteria used, the high risk groups for 0-19 and 20-64 years will always receive priority over non-high risk persons from the same age group (Table R13).
Implications for policy: Four scenarios regarding vaccine availability and distribution
While Option A, vaccinating some 77 million persons (see earlier for description of options), will ensure positive mean net returns, vaccinating an additional 20 million essential service personnel will result in increased mean net returns (Figure 5). Changing the vaccination strategy from targeting specific groups (Option B), to aiming to vaccinate 40 percent of the population regardless of occupation results in a decrease in mean net returns (Figure 5). Only Option D, vaccinating 60 percent of the population, will result in higher mean net returns.
An important feature to note in Figure 5 are the relatively wide ranges encompassed by the 5th and 95th percentiles. For example, at a gross attack rate of 15 percent, and a cost of vaccination of $62.26 per vaccinee, the 5th percentiles for all options are less than $1 billion, while the 95th percentiles are all above $14 billion (Figure 5). The minimum value for each option under those specific assumptions is less than $0 (i.e., a net loss) (data not shown).
Discussion and conclusions
Impact of an influenza pandemic
The next influenza pandemic in the U.S. may cause considerable impact in terms of loss of life, hospitalizations, outpatient visits and persons becoming ill but seeking no medical care (Figures 2 and 3). However, it must be clearly understood that there is a great deal of uncertainty associated with any estimate of the potential impact of an influenza pandemic. For example, while the results portray the potential impact for gross attack rates ranging from 15 to 35 percent, there are no data that describe the probability of any of those attack rates actually occurring in the next pandemic. This uncertainty should temper any reaction or policy relating to responses to a potential pandemic. The "Swine Flu Affair" (33) of 1976 clearly demonstrated the perils attached to making the switch from possibility to certainty (i.e., probability = 1.00).
The two other important points to remember when considering the impact of a pandemic are that the high risk groups are likely to bear a disproportion burden of the deaths (Table R1), and that 50 percent or more of the deaths are likely to occur among persons aged less than 65 years (Table R2). This finding is similar to a recent review of the distribution of deaths caused by the influenza pandemics of 1918, 1958, and 1968 (34).
With regard to measuring the economic impact associated with a pandemic, the results clearly illustrate that the greatest economic burden is due to death (Tables R3 and R4). This result is influenced by both the death rates (Table 3) and the economic values accorded to death (Table 11). The value of death used should be considered as a measure of the economic opportunity cost associated with a death. The implication is that the opportunity cost exists regardless of the employment status of anybody who may actually die. It is also important to note that the opportunity cost, as used, is not a measure of any intrinsic value that society may place on a life. Such a valuation is perhaps best left to a full and open debate. The resultant economic impact due to deaths means that, all other things being equal, the largest economic returns will come from the intervention(s) that prevents the largest number of deaths.
Careful attention should be paid to the fact that, beyond the value of a lost day of work (Table 11), the model does not include any valuation for disruptions in commerce and society due to an influenza pandemic. For example, if a large number of long distance truck drivers were all unavailable to drive for, say, two weeks, there may be difficulties in ensuring distribution of perishable items, especially food. Short term disruptions might not cause a lasting economic impact, but it is possible that some persons may suffer long-term consequences because of short disruptions caused by influenza. These "multiplier effects" are not accounted for in this model, mainly because an estimate of an appropriate multiplier will depend on who becomes ill, how many become ill, when they become ill, and for how long they are ill.
Returns to vaccination
All other factors being held constant, the net return to vaccination is sensitive to the combination of price and gross attack rate (Tables R6-R9, Figure 4). Regardless of gross attack rate assumed (in the range 15 - 35 percent), a vaccination price of $21.26 per vaccinee is almost always certain to generate a positive net return, even at the 5th percentile. At a cost of $62.26 per vaccinee, a gross attack rate will almost certainly generate net losses among non-high risk groups (Tables R6-R9).
Perhaps the most important element of the estimates of the net returns are the potentially wide ranges covered between the 5th and 95th percentiles (Tables R6-R9). Some of the 5th and 95th percentile values go from less than 50 percent to more than 70 percent of the means. Given the uncertainty surrounding the values of the input variables (Tables 3, 11, 12, 13), these ranges could be considered by some to be relatively narrow. However, the fact that some 5th percentiles are negative should serve as a warning to any decision maker that many interventions may not guarantee a net positive return.
Relative importance of input variables
The results of the net returns (TablesR6-R9, Figure 4) can be used to rank the input variables in terms of their relative influence in determining the net value of vaccination. Such a ranking is, of course, only valid for the ranges used for the inputs (Tables 3, 11, 12, 13). The variables that cause the largest percentage changes in net returns are: death rate (Figure 4), cost of vaccination per vaccinee, and gross attack rate (Tables R6-R9). The next most important variable is vaccine effectiveness, which is more important than compliance, but compliance is more important than age distribution of cases (Table 2).
Implications for policy: "Insurance premiums" for the 3P's
The size of the "premium" that could be used as a measure of what could logically be spent each year to plan, prepare and practice (3 P's) for the next influenza pandemic depends most on the assumed probability of occurrence (Table R12). The results presented in Table R12 present a cautionary tale of the difference between possibility and probability of an influenza pandemic occurring. Assuming that an influenza pandemic could actually occur, the numbers in Table R12 can be used to argue that it is legitimate for society to spend some money each year on the 3 P's. What cannot be currently stated with any certainty is the degree of probability of such a pandemic occurring, nor the number of people who will succumb to clinical illness. Defining the difference between possibility and probability was a key decision point in the "swine 'flu affair" of 1976-1977 (33).
Implications for policy: Setting priorities
The results presented in Table R11 illustrate that priorities for vaccination depend upon the objective function chosen. If preventing the most number of deaths possible is seen as the most important function of an intervention, then society should ensure that all those in the high risk groups become vaccinated first, followed by those of non-high risk, aged 65 + years (Table R11). However, if maximizing economic returns to vaccination is defined as the objective for setting priorities, then those aged 0-64 years of age, regardless of risk, should be vaccinated first.
The results also illustrate the need to be rather precise in defining the criteria used for setting priorities. For example, stating that "preventing death" will be the criteria used is not sufficiently precise, because different priority lists can be drawn up using death rates versus actual deaths (or percentages of total deaths) (Table R11). It should also be carefully noted that the criteria used to generate the results presented in Table R11 do not define the entire set of possible methods of prioritization. Society may decide to use another criterion or a set of criteria. Thus, rather than be used to absolutely set priorities, Table R11 provides a starting point for debate within society regarding the setting of priorities.
Implications for policy: Four scenarios regarding vaccine availability and distribution
The net returns for the four scenarios modeled (Figure 5) further illustrate the need to clearly set criteria, goals and objectives for a vaccine-based intervention for the next influenza pandemic. Option C and Option D aim for 40 percent coverage, and 60 percent coverage, respectively, regardless of age or risk group. Some may state that these options represent a more egalitarian, or "fair," means of distributing vaccine. However, such egalitarianism would cost society as the net returns to Options C are lower than those from Option B (Figure 5).
Option D does produce higher returns than Option B, but it is questionable if 60 percent of U.S. society could be successfully vaccinated in the appropriate time span (2-3 months?) needed to generate the net returns shown in Figure 5. Achieving 60 percent coverage of the U.S. population will be especially difficult if 2 doses of vaccine are needed to provide "satisfactory" protection. Giving two doses for 60 percent of the U.S. population is equivalent to almost 320 million doses (Table 14). More doses would probably be needed since it is likely that not everybody who will receive a first dose will return for a second dose (i.e., compliance between the first and second doses will decrease). Thus, in order to achieve a 60 percent effective coverage, more than 60 percent of the population will need to receive the first dose. Also, spoilage will also increase the total number of doses that must be produced. It should not be forgotten that, at no point in history, has + 320 million doses of vaccine been delivered and administered to the U.S. population in a 2 - 3 month time period.
Summary of some main points
References
Table 1: U.S. population estimates used to define impact of influenza
Age group | Numbers (millions) |
Percentage of total (%) |
0-19 yrs old | 76.37 | 28.8 |
20-64 yrs old | 155.03 | 58.4 |
65 + yrs old | 33.86 | 12.8 |
Total | 265.26 | 100.0 |
Sources: U.S. Bureau of the Census. Statistical abstract of the United States: 1997 (117th edition). Washington, D.C., 1997.
Table 2: Two scenarios1 of age distributions of cases and percentage of high risk population used to examine the impact of pandemic influenza in the United States
Age group | Distribution A: % of all cases2 |
Distribution B: % of all cases2 |
0-19 yrs old | 40.0 | 46.0 |
20-64 yrs old | 53.1 | 46.7 |
65 + yrs old | 6.8 | 7.3 |
Totals3 | 100.0 | 100.0 |
Assumed percentage at "high risk" | ||
0-19 yrs old | 6.4 | 11.1 |
20-64 yrs old | 14.4 | 25.0 |
65 + yrs old | 40.0 | 55.0 |
U.S. average4 | 15.4 | 24.8 |
Note: 1) The scenarios were calculated using upper and lower estimates of
age-specific attack rates from 1918, 1928-29, and 1957.3
2) The actual number of cases will depend upon the assumed gross attack rate (see text for
further details).
3) Totals may not exactly add to 100 percent due to rounding up.
4) Average is an age-weighted average, using each age group's proportion of the total U.S.
population.
Table 3: Variables used to define distribution of health outcomes of those with clinical cases of influenza
Variable | Rates per 1,000 general population | Rates per 1,000 cases1 | |||||||||
Lower | "Most likely" |
Upper | Probability distribution |
Lower | "Most likely" |
Upper | Probability distribution |
||||
Outpatient visits (per 1,000) | |||||||||||
"Non-high risk" | |||||||||||
0-19 yrs old (rate) | 165 | 230 | Uniform | 471 | 548 | Uniform | |||||
20-64 yrs old (rate) | 40 | 85 | Uniform | 333 | 370 | Uniform | |||||
65 + yrs old (rate) | 45 | 74 | Uniform | 375 | 389 | Uniform | |||||
"High risk" | |||||||||||
0-19 yrs old (rate) | 289 | 403 | Uniform | 825 | 958 | Uniform | |||||
20-64 yrs old (rate) | 70 | 149 | Uniform | 583 | 647 | Uniform | |||||
65 + yrs old (rate) | 79 | 130 | Uniform | 656 | 682 | Uniform | |||||
Hospitalizations (per 1,000) | |||||||||||
"Non-high risk" | |||||||||||
0-19 yrs old (rate) | 0.2 | 0.5 | 2.9 | Triangular | 0.57 | 6.9 | Uniform | ||||
20-64 yrs old (rate) | 0.18 | 2.75 | Uniform | 1.5 | 12.0 | Uniform | |||||
65 + yrs old (rate) | 1.5 | 3.0 | Uniform | 12.5 | 15.8 | Uniform | |||||
"High risk" | |||||||||||
0-19 yrs old (rate) | 2.1 | 2.9 | 9.0 | Triangular | 6.0 | 21.4 | Uniform | ||||
20-64 yrs old (rate) | 0.83 | 5.14 | Uniform | 6.9 | 22.3 | Uniform | |||||
65 + yrs old (rate) | 4.0 | 13 | Uniform | 33.3 | 68.4 | Uniform | |||||
Deaths (per 1,000) | |||||||||||
"Non-high risk" | |||||||||||
0-19 yrs old (rate) | 0.014 | 0.024 | 0.125 | Triangular | 0.041 | 0.07 | 0.30 | Triangular | |||
20-64 yrs old (rate) | 0.025 | 0.037 | .09 | Triangular | 0.21 | 0.31 | 0.41 | Triangular | |||
65 + yrs old (rate) | 0.28 | 0.42 | 0.54 | Triangular | 2.3 | 3.51 | 4.52 | Triangular | |||
"High risk" | |||||||||||
0-19 yrs old (rate) | 0.126 | 0.22 | 7.65 | 0.4 | 0.6 | 21.9 | Triangular | ||||
20-64 yrs old (rate) | 0.1 | 5.72 | 0.8 | 24.9 | Uniform | ||||||
65 + yrs old (rate) | 2.76 | 5.63 | 23 | 29.6 | Uniform | ||||||
Conversion factors2 | |||||||||||
0-19 yrs old | 0.35 | 0.42 | Uniform | ||||||||
20-64 yrs old | 0.12 | 0.23 | Uniform | ||||||||
65 + yrs old | 0.12 | 0.19 | Uniform |
Notes: 1) Cases defined here as those persons with a clinical illness sufficient
to cause an economic impact.
2) Conversion factor used to convert rates per 1,000 general population to rates per 1,000
cases ( Equation 1, text). The factors are equivalent to attack rates within the general
population. Thus, a conversion factor of 0.35 is equivalent to a 35% attack rate.
Sources: See text.
Table 4: International Classification of Diseases, Ninth Revision (ICD-9) codes used to extract medical costs for inpatient, outpatient and related drug claims.
Diagnosis | ICD-9 code |
Pneumonia and influenza | 480-487.8 |
Acute bronchitis | 466-466.1 |
Chronic respiratory disease | 490-496 |
Heart-related diagnoses | |
Acute ischemic heart disease | 410-411.90, 413-413.9 |
Chronic ischemic heart disease | 412, 414.0, 414.8, 414.9, 416-416.9 |
Heart rhythm disorders | 426-427.9 |
Congestive heart failure | 428-428.9 |
Table 5: Average age of patients, length of stay, and net
payments (standard deviations) of an inpatient hospitalization, by type of diagnosis and
age: 1993-95 combined data1,2
Type of diagnosis | Age group | n (# of claims) |
Avg. age of patient (yrs) | Avg. length of stay (days) | Avg. net pay. ($)3 |
Influenza, pneumonia, bronchitis, and chronic respiratory: Principal diag. only4 | 0-17 years | 10,253 | 3.9 (4.8) |
3.5 (3.6) |
4,789 (8,146) |
18-64 years | 22,793 | 50.4 (11.6) |
6.2 (7.9) |
9,691 (18,166) |
|
65 + years | 1,063 | 69.6 (4.7) |
7.9 (9.7) |
11,047 (15,673) |
|
Influenza, pneumonia, bronchitis, and chronic respiratory: All diag.5 | 0-17 years | 12,836 | 4.1 (5.0) |
4.0 (6.1) |
6,007 (14,604) |
18-64 years | 44,457 | 51.6 (11.2) |
7.1 (11.0) |
12,622 (21,577) |
|
65 + years | 2,339 | 69.3 (4.5) |
8.3 (10.6) |
14,139 (20,488) |
|
Heart related conditions: All diag.6 | 0-17 years | 1,150 | 7.1 (6.8) |
8.3 (12.7) |
22,637 (40,285) |
18-64 years | 96,033 | 54.5 (8.1) |
5.3 (7.8) |
15,710 (22,000) |
|
65 + years | 4,705 | 69.1 (4.4) |
6.6 (8.7) |
17,140 (22,392) |
|
Weighted averages7 | 0-17 yrs | 13,986 | 4.2 | 4.0 | 6,481 |
18-64 yrs | 140,490 | 53.8 | 5.8 | 14,359 | |
65 + yrs | 7,044 | 69.2 | 7.1 | 15,677 |
Notes: 1) Illness defined using the following ICD-9 codes in Table 4.
2) 1993 and 1944 cost data were inflated to 1995 using the medical care component of the
consumer price index (Table 752, U.S., Bureau of Census, ref. 1).
3) Net payment made by health care payer to provider(s) of health care services.
4) ICD-9 code (Table 4) for either influenza, pneumonia, acute bronchitis, or chronic
respiratory disease was entered as principal diagnosis (i.e., main diagnosis).
5) ICD-9 code (Table 4) for either influenza, pneumonia, acute bronchitis, or chronic
respiratory disease was entered as either principal diagnosis (i.e., main diagnosis) or
one of the first four diseases coded.
6) ICD-9 code for all heart-related conditions identified in Table 4 was entered as either
principal diagnosis (i.e., main diagnosis) or one of the first four diseases coded.
7) For each age group and each variable, weighted averages were calculated as the sum of
proportion of claims under each "type of diagnosis" times the value of the
variable. In order to reduce overlap, the number of cases of "Influenza, . . :All
diag."was first reduced by removing the number of claims under "Influenza .
."Principal daig. only".
Source: MarketScan database, The MEDSTAT Group.
Table 6: Average inpatient and outpatient charges and drug claims (standard deviations) associated with each inpatient admission for influenza or pneumonia or acute bronchitis:1993-95 combined data1,2
Age group | n (# of admissions)3 |
Avg. net payments to hospitals ($)4 | Length of stay (days) | Outpatient payments4,5 ($) | Drug claims ($) |
0-17 years | 7,421 | 5,439 (12,376) |
4 (6) |
74 (160) |
26 (36) |
18-64 years | 16,873 | 11,217 (12,300) |
7 (8) |
92 (283) |
42 (123) |
65 + years | 702 | 12,793 (18,865) |
8 (10) |
102 (241) |
41 (41) |
All ages | 24,998 | 9,485 (17,553) |
6 (8) |
87 (251) |
37 (105) |
Notes: 1) Illness defined using the following ICD-9 codes in Table 4.
2) 1993 and 1944 cost data were inflated to 1995 using the medical care component of the
consumer price index (Table 752, U.S., Bureau of Census, ref. 1).
3) Number of admissions for which at least one outpatient visit was identified.
4) Net payment made by health care payer to provider(s) of health care services.
5) Outpatient claims linked to inpatients using patient ID numbers. Average includes all
outpatient claims for 14 days prior to admission date and up to 30 days after discharge.
6) Drug claims linked to patient ID numbers. The total number of drug claims identified
was less that the number of identified inpatient admissions. The number of drug claims
identified were: 0-17 years - 3,682 claims; 18-64 years - 8,594 claims; 65 + years - 376
claims (total - 12,652 claims).
Table 7: Number and frequency of outpatient visits for diagnoses of influenza, pneumonia and acute bronchitis:1 1993-95 data
Max. # visits/ patient/ year | # of identified patients | % of identified patients | Total visits | Weighted average of visits/ patient |
1993-95 combined | ||||
< 20 | 687,349 | 100.0 | 1,673,726 | 2.44 |
< 7 | 655,622 | 95.4 | 1,331,868 | 2.03 |
< 3 | 558,360 | 81.2 | 846,327 | 1.52 |
1995 only | ||||
< 20 | 262,720 | 100.0 | 632,456 | 2.41 |
< 7 | 250,905 | 95.5 | 504,905 | 2.01 |
< 3 | 214,391 | 81.6 | 322,571 | 1.50 |
1994 only | ||||
< 20 | 229,225 | 100.0 | 566,118 | 2.47 |
< 7 | 218,135 | 95.2 | 446,591 | 2.05 |
< 3 | 184,939 | 80.7 | 280,577 | 1.52 |
Oct-Dec 1993 only2 | ||||
< 20 | 70,750 | 100.0 | 148,844 | 2.10 |
< 7 | 68,770 | 97.2 | 128,122 | 1.86 |
< 3 | 60,856 | 86.0 | 89,426 | 1.47 |
Notes: 1) Illness defined using ICD-9 codes: 480-487.9, 466-466.1 (Table 4).
2) Outpatient visits that occurred between October 01 - December 31, 1993, only. These
visits represented 36% of all patients, and 31% of all outpatient visits, recorded for all
of 1993 for the ICD-9 codes used to define influenza, pneumonia, and acute bronchitis
(Table 4).
Source: MarketScan database, The MEDSTAT Group.
Table 8: Average age of patients and costs (standard deviations) of an outpatient visit, by age group, for diagnoses of influenza, pneumonia and acute bronchitis: 1994-95 data1,2
Age group | n (# of claims) | Average age of patient (years) | Average co-payment ($) | Average net payment ($)3 |
0-17 years | 357, 071 | 6.5 (5.4) |
4.63 (17.01) |
49.17 (97.05) |
18-64 years | 883,484 | 45.2 (12.0) |
4.49 (21.54) |
37.68 (125.35) |
65 + years | 16,505 | 68.9 (4.4) |
4.19 (18.81) |
50.03 (220.47) |
Notes: 1) Illness defined using the following ICD-9 codes: 480-487.9, 466-466.1
2) 1994 cost data were inflated to 1995 using a factor of 1.045, the medical care
component of the consumer price index (Table 752, U.S., Bureau of Census, ref. 1).
3) Net payment made by health care payer to provider(s) of health care services.
Source: MarketScan database, The MEDSTAT Group.
Table 9: Average payments (standard deviations) for drug claims made by patients with an outpatient visit for influenza, pneumonia and acute bronchitis: 1994-95 data1
Age group | n (# drug claims) |
Average supply (days) | Average co-payment ($) | Average net payment2 | Average number prescript. per visit3 |
1995 | |||||
0-17 years | 173,263 | 14 (13) |
3.41 (6.81) |
24.57 (57.80) |
0.90 |
18-64 years | 870,173 | 24 (21) |
3.67 (5.31) |
36.41 (166.78) |
1.89 |
65 + years | 11,970 | 26 (20) |
3.86 (6.05) |
37.35 (68.87) |
1.37 |
1994 | |||||
0-17 years | 136,954 | 15 (14) |
3.13 (7.86) |
25.08 (86.11) |
0.83 |
18-64 years | 716,199 | 24 (20) |
3.47 (14.57) |
35.04 (94.59) |
1.70 |
65 + years | 10,994 | 27 (20) |
3.14 (2.83) |
35.93 (40.7) |
1.42 |
Notes: 1) Illness defined using the following ICD-9 codes: 480-487.9, 466-466.1
2) Net payment made by health care payer to provider(s) of health care services.
3) Calculated by dividing the number of drug claims (prescriptions) by the number of
outpatient visits for each age group and year.
Source: MarketScan database, The MEDSTAT Group.
Table 10: Pearson correlation coefficients1 for inpatient care related
to influenza, pneumonia, and acute bronchitis: 1994 and 1995 data2
Principal diagnosis only3 (n = 34,118) |
All diagnoses3 (n = 59,642) |
||||||
Age of patient | Length of stay | Net payment | Age of patient | Length of stay | Net payment | ||
Age of patient | 1.00 | 1.00 | |||||
Length of stay | -0.02 (0.000) |
1.00 | -0.01 (0.141) |
1.00 | |||
Net payment | 0.01 (0.032) |
0.64 (0.000) |
1.00 | 0.01 (0.002) |
0.61 (0.000) |
1.00 |
Notes: 1) Probability under Ho: R = 0 in parentheses.
2) 1994 cost data inflated to equivalent 1995 values using a factor of 1.045, the medical
care component of the consumer price index (ref. 1).
3) Principal diagnosis means that the ICD-9 codes (Table 4) for either influenza,
pneumonia, or acute bronchitis must be entered as main diagnosis. All diagnoses means that
the codes can be entered as either main diagnosis or any of the first for diseases in the
code sheet.
Table 11: Inputs used to calculate the economic impact (direct and indirect costs) of an influenza pandemic in the U.S. (1995 U.S.$)
Outcome category Item |
Type of cost | 0-19 years | 20-64 years | 65+ years | Sources |
Deaths | |||||
Average age (years) | 9 | 35 | 74 | Assumed | |
PV earnings lost ($)1 | indirect | 1,016,101 | 1,037,673 | 65,837 | Refs: 1, 27 |
"Most likely" + min or max hospital costs ($)2 | direct | 3,435 + 2,632 |
7,605 + 3,888 |
8,309 + 3,692 |
Marketscan Database®; Ref. 28. (Table 5) |
Sub-total ($)3 | 1,019,536 | 1,045,278 | 74,146 | ||
Hospitalizations | |||||
"Most likely" + min or max hospital costs ($)2 | direct | 2,936 + 2,099 |
6,016 + 2,086 |
6,856 + 3,200 |
Marketscan Database®; Ref. 28. (Table 6) |
"Most likely" + min or max net pay for outpatient visits ($)4 | direct | 74 + 40 | 94 + 70 | 102 + 60 | Marketscan Database®; Ref. 28. (Table 6) |
Avg. co-payment for outpatient visit ($) | direct | 5 | 4 | 4 | Marketscan Database® (Table 8) |
"Most likely" + min or max net payment for drug claims($)5 | direct | 26 + 9 | 42 + 30 | 41 + 10 | Marketscan Database® (Table 6) |
"Most likely" + min or max days lost6 | indirect | 5 + 2.7 | 8 + 4.8 | 10 +5.4 | Marketscan Database®; Ref. 28. (Table 6) |
Value 1 day lost ($)7 | indirect | 65 | 100 or 65 | 65 | Ref. 27 |
Sub-total ($)3 | 3,366 | 6,842 | 7,653 | ||
Outpatient | |||||
Avg. no. visits | direct | 1.52 | 1.52 | 1.52 | Marketscan Database® (Table 7; Figure 3) |
"Most likely" + min or max net payment per visit($)8 | direct | 49 +13 | 38 + 12 | 50 + 16 | Marketscan Database® (Table 8) |
Avg. co-payment for outpatient visit ($) | direct | 5 | 4 | 4 | Marketscan Database® (Table 8) |
"Most likely" + min or max net payment per prescription($)9 | direct | 25 + 18 | 36 + 27 | 36 + 22 | Marketscan Database® (Table 9) |
Avg. prescriptions per visit8 | direct | 0.9 | 1.8 | 1.4 | Marketscan Database® (Table 9) |
Avg. co-payment per prescription ($) | direct | 3 | 3 | 3 | Marketscan Database® (Table 9) |
Days lost | indirect | 3 | 2 | 5 | Refs. 20; 22 |
Value 1 day lost ($)7 | indirect | 65 | 100 | 65 | Ref. 27 |
Sub-total ($)3 | 300 | 330 | 458 | ||
Ill, no medical care sought | |||||
Days lost | indirect | 3 | 2 | 5 | Refs. 20; 22 |
Value 1 day lost ($)7 | indirect | 65 | 100 | 65 | Ref. 27 |
Over-the-counter drugs ($) | direct | 2 | 2 | 2 | Assumed |
Sub-total ($)3 | 197 | 202 | 327 |
Notes: 1) Present value (PV), using a discount rate of 3% of expected future
lifetime earnings and housekeeping services, adjusted to 1995 dollars. Haddix et al.
(Appendix I in Ref. 27) estimated the present value of wages, using 1990 data. Using Ref.
1 (Tables 705, 706), the average increase (in constant 1992 $) in total personal income in
the U.S. and per capita personal income from 1990-1995 was 7%. Thus, data from Ref. 27
were multiplied by 1.07.
2) "Most likely," with + defining the minimum and maximum costs for a
triangular distribution for Monte Carlo analysis. The values were calculated using cost
data from Marketscan Database®, The MEDSTAT Group, and multiplying it by a hospital
cost-to-charge ratio of 0.53. The ratio is a weighted average of the urban and rural
(urban = 0.80, rural = 0.20) cost-to-charge ratios calculated by the Health Care Finance
administration (HCFA) for August, 1996 (Ref. 28).
3) Sub-totals of are the totals for each category of outcome, using the "most
likely" estimates. A distribution of subtotals is constructed by using the minimum
and maximum values provided for some of the items within the outcome category.
4) "Most likely," with minimum and maximum values of net payments for outpatient
visits up to 14 days prior to admission date and up to 30 days after discharge date.
5) Net payment for drug claims associated with outpatient visits up to 14 days prior to
admission and up to 30 days after discharge.
6) "Most likely," with + defining the minimum and maximum days lost due
to hospitalization for a triangular distribution for Monte Carlo analysis. Calculated
using length-of-stay in hospital data from Marketscan Database®, and adding a total of
one additional day for convalescence and pre- and post-hospitalization outpatient visits
for 0-19, and 20-64 years of age. For 65 + years, two additional days were added to
length-of-stay in hospital for convalescence and pre- and post-hospitalization outpatient
visits.
7) For 0-19 and 65+ years age groups, a day lost to influenza was valued as equivalent to
an "unspecified" day (Appendix I in Ref. 27), denoting a value for time lost by
care givers and family members related to taking care of a patient in these age groups.
For 20-64 years of age, 60% of days lost due to hospitalizations and related convalescence
and pre- and post-hospitalization outpatient visits were valued as "day off
work" ($100/day). The remaining 40% of days lost were valued as
"unspecified" days ($65/day). For 20-64 years of age, when patients were not
hospitalized at any point during their illness (i.e., outpatient status), all days lost
were assumed "days off work" ($100/day).
8) "Most likely," with minimum and maximum values of net payments for outpatient
visits without any specified association to hospitalizations.
9) "Most likely," with + defining the minimum and maximum cost per
prescription, with the number of prescriptions per visit.
Table 12: Two scenarios for the cost of vaccinationa during an influenza pandemic, with specific costs assigned to side-effects of vaccination.
Lower cost scenario ($/patient) |
Upper cost scenario ($/patient) |
|||
Item | Probability of side effectb | $/ case of side-effectb | ||
Assumed cost of vaccinationa (excluding side effects) | 18 | 59 | ||
Side effects | ||||
Mildc | 0.0325 | 94 | 3.05 | 3.05 |
GBSd | 0.000002 | 100,800 | 0.20 | 0.20 |
Anaphylaxis | 0.000000157 | 2,490 | 0.01 | 0.01 |
Total costs/ patient | 21.26 | 62.26 |
Notes: 1) The cost of vaccination includes the cost of the vaccine, the cost of
administering the vaccine, value of time spent by an individual traveling to and from the
place of vaccination, and patient associated travel costs. Included in the costs of the
vaccine are any costs associated with the rapid production of a larger-than-usual number
of doses, and the rapid delivery and correct storage of doses at vaccination sites around
the country. For $18, the costs were assumed to be broken down as follows: $10 for vaccine
+ administration, $4 patient time (one-half hour), $4 patient travel costs. For $59, the
costs were assumed to be broken down as follows: $20 for vaccine + administration (this
could include the cost of 2 doses), $32 patient time (2 trips at 2 hours per trip), $7
patient travel costs. Note that, for comparison, a review of 10 published papers found a
range of $5 - $22 per dose of vaccine, with a "medium" [sic] cost of $14 per
dose (10). These breakdowns are illustrations only of what might be deemed
"reasonable" estimates of time and cost. Actual costs might vary substantially,
and will depend on the number of doses needed to achieve a "satisfactory"
protective response, as well as how efficiently vaccinations can be give to millions of
people.
2) Probabilities and average cost of treating each category of side effect were derived
from the Office of Technology report (3).
3) Mild side-effects include sore arms due to vaccination, headaches, and other
"minor" side effects that may require a visit to a physician, or may cause the
patient to miss 1-2 days of work.
4) GBS = Guillian Barré Syndrome.
Table 13: Two scenarios for vaccine effectiveness1 in preventing health outcomes by age group2
Health outcomes |
Vaccine effectiveness in preventing health outcomes1 | ||||||
"High" level of effectiveness3 | "Low" level of effectiveness3 | ||||||
0-19 yrs | 20-64 yrs | 65 + yrs | 0-19 yrs | 20-64 yrs | 65 + yrs | ||
Death | 0.75 | 0.70 | 0.60 | 0.40 | 0.40 | 0.30 | |
Hospitalization | 0.55 | 0.55 | 0.50 | 0.55 | 0.55 | 0.50 | |
Outpatient visits | 0.40 | 0.40 | 0.40 | 0.40 | 0.40 | 0.40 | |
Ill, no medical care sought | 0.40 | 0.40 | 0.40 | 0.40 | 0.40 | 0.40 |
Notes: 1) Vaccine effectiveness is defined as the reduction in the number of
cases in each of the age and health outcome categories.
2) Within a defined age group, it was assumed that there was no difference in vaccine
effectiveness between "high" risk and "non-high" risk sub-groups.
3) The terms "high" and "low" level of effectiveness are subjective
and reflect only a judgement of the levels of effectiveness in the two scenarios relative
to each other.
Table 14: Four scenarios1 regarding vaccine availability and distribution: Percentage and numbers vaccinated2 by age and risk group
Age group | Risk group | Scenario A | Scenario B | Scenario C | Scenario D | ||||
% vaccinated | # (mill) |
% vaccinated | # (mill) |
% vaccinated | # (mill) |
% vaccinated | # (mill) |
||
0-19 yrs | Non-high | 20 | 13.9 | 20 | 13.9 | 40 | 27.9 | 60 | 41.8 |
High | 40 | 2.7 | 40 | 2.7 | 40 | 2.7 | 60 | 4.0 | |
20-64 yrs | Non-high | 20 | 24.9 | 35 | 43.6 | 40 | 49.8 | 60 | 74.7 |
High | 40 | 12.2 | 50 | 15.3 | 40 | 12.2 | 60 | 18.3 | |
64 + yrs | Non-high | 70 | 12.4 | 70 | 12.4 | 40 | 7.1 | 60 | 10.7 |
High | 70 | 11.3 | 70 | 11.3 | 40 | 6.4 | 60 | 9.7 | |
Total patients | 77.4 | 99.2 | 106.1 | 159.2 | |||||
Total doses needed3 | |||||||||
1 dose per patient | 77.4 | 99.2 | 106.1 | 159.2 | |||||
2 doses per patient | 154.8 | 198.4 | 212.2 | 318.4 |
Notes: 1) Scenario A is similar to current Advisory Committee of Immunization
practices (ACIP) recommendations. Scenario B is the same as Scenario A, with an additional
20 million essential service providers. Scenarios C and D aim to achieve 40 and 60 percent
coverage, regardless of age and risk categories. See text for complete description of
assumptions.
2) All scenarios assume vaccination will result, for all age groups, in a 50 percent
reduction in deaths and hospitalizations, and a 40 percent reduction in outpatient visits
and illnesses not requiring formal medical care.
3) In a pandemic, which will be caused by a new subtype of influenza, it may be that
everybody will require 2 doses in order to induce a "satisfactory" antibody
response.
Table R1: Burden of impact, by age group, of an influenza pandemic among high risk groups relative to total impact1
Category of impact | Age group | Age distribution A: Percentage of total cases that are High Risk |
Age distribution B: Percentage of total cases that are High Risk |
|||||
Mean | 5th | 95th | Mean | 5th | 95th | |||
Death | 0-19 yrs | 9.0 | 1.4 | 20.2 | 12.3 | 1.9 | 27.3 | |
20-64 yrs | 40.9 | 11.1 | 60.9 | 43.0 | 12.3 | 63.8 | ||
65 + yrs | 34.4 | 22.7 | 52.1 | 35.5 | 22.8 | 56.2 | ||
Total | 84.3 | 90.8 | ||||||
Hospitalization | 0-19 yrs | 4.6 | 2.1 | 7.9 | 8.1 | 4.0 | 13.0 | |
20-64 yrs | 14.7 | 7.4 | 23.4 | 19.6 | 10.7 | 29.3 | ||
65 + yrs | 18.3 | 11.0 | 27.6 | 23.8 | 15.4 | 33.6 | ||
Total | 37.6 | 51.5 | ||||||
Outpatient | 0-19 yrs | 5.0 | 4.7 | 5.4 | 9.4 | 8.7 | 10.0 | |
20-64 yrs | 10.4 | 9.8 | 11.0 | 14.7 | 13.9 | 15.5 | ||
65 + yrs | 4.0 | 3.9 | 4.2 | 5.5 | 5.3 | 5.7 | ||
Total | 19.5 | 29.6 |
Notes: 1) See Table 2 for distribution of persons at high risk by age group.
Table R2: Distribution, by age group, of deaths due to an influenza pandemic
Age group | Age distribution A: Percentage of total deaths |
Age distribution B: Percentage of total deaths |
|||||||||
Mean | 5th | 95th | Min. | Max. | Mean | 5th | 95th | Min. | Max. | ||
0-19 yrs | 11.5 | 3.2 | 23.1 | 1.2 | 32.7 | 14.3 | 3.3 | 29.3 | 1.1 | 41.3 | |
20-64 yrs | 47.4 | 20.5 | 65.2 | 13.2 | 70.3 | 46.5 | 17.2 | 66.2 | 9.2 | 71.6 | |
65 + yrs | 41.1 | 27.8 | 62.5 | 24.0 | 76.4 | 39.3 | 25.9 | 62.0 | 21.9 | 78.9 | |
Total | 100.0 | 100.0 |
Table R3: Economic impact (direct and indirect costs) of influenza pandemic per gross attack rate:1,2 Deaths, hospitalizations, outpatients, illnesses and total costs (1995 U.S.$)
Cost per gross attack rate1 ($ millions) | |||||
15% | 20% | 25% | 30% | 35% | |
Deaths | |||||
Mean | 59,288 | 79,051 | 98,814 | 118,577 | 138,340 |
5th percentile | 23,800 | 31,733 | 39,666 | 47,599 | 55,532 |
95th percentile | 94,907 | 126,543 | 158,179 | 189,815 | 221,451 |
Hospitalizations | |||||
Mean | 1,928 | 2,571 | 3,214 | 3,856 | 4,499 |
5th percentile | 1,250 | 1,667 | 2,084 | 2,501 | 2,917 |
95th percentile | 2,683 | 3,579 | 4,472 | 5,367 | 6,261 |
Outpatients | |||||
Mean | 5,708 | 7,611 | 9,513 | 11,416 | 13,318 |
5th percentile | 4,871 | 6,495 | 8,119 | 9,742 | 11,366 |
95th percentile | 6,557 | 8,742 | 10,928 | 13,113 | 15,299 |
Ill, no medical care sought3 | |||||
Mean | 4,422 | 5,896 | 7,370 | 8,844 | 10,317 |
5th percentile | 3,270 | 4,360 | 5,450 | 6,540 | 7,629 |
95th percentile | 5,557 | 7,409 | 9,262 | 11,114 | 12,967 |
Grand Totals | |||||
Mean | 71,346 | 95,128 | 118,910 | 142,692 | 166,474 |
5th percentile | 35,405 | 47,206 | 59,008 | 70,810 | 82,611 |
95th percentile | 106,988 | 142,650 | 178,313 | 213,975 | 249,638 |
Notes: 1) Estimates are for age distribution scenario A (Table 2).
2) Gross attack rate refers to the proportion of the total U.S. population that becomes
clinically ill due to influenza such that their illness causes an economic impact.
3) Ill, no medical care" sought is defined as persons who become clinically ill due
to influenza, and that illness results in an economic impact (e.g., half day off work).
However, these persons do not seek health care as an outpatient or an inpatient.
Table R4: Proportion of costs attributable to direct and indirect cost items, by category of health outcome and age group
Percentage of costs attributable to direct or indirect items2 | ||||||
0-19 years | 20-64 years | 65+ years | ||||
Direct (%) | Indirect (%) | Direct (%) | Indirect (%) | Direct (%) | Indirect (%) | |
Deaths | ||||||
Mean | 0.3 | 99.7 | 0.7 | 99.3 | 11.2 | 88.8 |
5th percentile | 0.2 | 99.5 | 0.5 | 99.0 | 8.1 | 85.9 |
95th percentile | 0.5 | 99.8 | 1.0 | 99.5 | 14.1 | 91.9 |
Hospitalizations | ||||||
Mean | 89.3 | 10.7 | 89.6 | 10.4 | 91.1 | 8.9 |
5th percentile | 81.6 | 5.7 | 84.7 | 6.0 | 86.7 | 5.3 |
95th percentile | 94.3 | 18.2 | 94.0 | 15.3 | 94.7 | 13.3 |
Outpatients | ||||||
Mean | 34.9 | 65.1 | 40.3 | 59.7 | 29.0 | 71.0 |
5th percentile | 31.0 | 61.4 | 28.5 | 45.3 | 24.9 | 66.9 |
95th percentile | 38.6 | 69.0 | 55.4 | 71.5 | 35.5 | 75.1 |
Ill, no medical care sought2 | ||||||
Mean | 1.0 | 99.0 | 1.1 | 98.9 | 0.6 | 99.4 |
5th percentile | 1.0 | 99.0 | 0.7 | 98.2 | 0.6 | 99.4 |
95th percentile | 1.0 | 99.0 | 1.8 | 99.3 | 0.6 | 99.4 |
Notes: 1) See Table 11 for description of direct and indirect cost items.
2) For each age group and health outcome, the means of the direct and indirect costs
should add to 100%,. Any difference will be due to rounding the results to the first
decimal place. The 5th and 95th percentiles will not necessarily add
to 100%.
Table R5: Proportion of total costs attributable to deaths by risk category: Non-high and "high"1
Percentage of costs attributable to death (all ages) | |||
Non-high risk (%) |
High risk (%) | All deaths (%) | |
Mean | 10.8 | 70.1 | 80.9 |
5th percentile | 6.0 | 46.7 | 65.7 |
95th percentile | 19.4 | 82.7 | 89.1 |
Notes: 1) "High" risk is defined as those patients who have one or more pre-existing medical conditions (e.g., emphysema) that make them prone to severe outcomes if they have a clinical case of influenza. Non-high risk are all those who are not categorized as "high" risk.
Table R6: Total net value of vaccinating against an influenza pandemic by level of compliance, age-risk group, cost of vaccination, and gross attack rate: Age distribution scenario A,1 "high" level of vaccine effectiveness2
Vaccination compliance Age, risk group6 |
Mean (5th; 95th
percentiles) net value of vaccination ($ millions, 1995 $)3 |
||||||
$21.26 cost of vaccination4 | $62.26 cost of vaccination4 | ||||||
Gross attack rate (%)5 | Gross attack rate (%)5 | ||||||
15 | 25 | 35 | 15 | 25 | 35 | ||
40% compliance | |||||||
0-19 yrs: Non-high risk | 618 (262; 1,122) |
1,426 (836; 2,259) |
2,232 (1,412; 3,395) |
-525 (-884; -31) |
282 (-305; 1,103) |
1,089 (266; 2,235) |
|
High risk | 2,421 (365; 5,338) |
4,074 (643; 8,941) |
5,726 (921; 12,543) |
2,314 (257; 5,261) |
3,968 (537; 8,885) |
5,622 (815; 12,512) |
|
20-64 yrs: Non-high risk | 1,470 (1,043; 1,893) |
3,156 (2,453; 3,847) |
4,842 (3,870; 5,800) |
-653 (-1,109; -194) |
978 (260; 1,663) |
2,609 (1,638; 3,550) |
|
High risk | 11,467 (1,830; 21,097) |
19,285 (3,215; 35,321) |
27,103 (4,612; 49,553) |
10,978 (1,361; 20,636) |
18,804 (2,721; 34,853) |
26,631 (4,060; 49,108) |
|
64 + yrs: Non-high risk | 81 (47; 114) |
236 (188; 283) |
391 (328; 452) |
-209 (-273; -146) |
-54 (-124; 17) |
102 (21; 182) |
|
High risk | 521 (446; 598) |
960 (840; 1,085) |
1,399 (1,232; 1,572) |
259 (164; 353) |
699 (564; 836) |
1,139 (966; 1,320) |
|
60% compliance | |||||||
0-19 yrs: Non-high risk | 927 (389; 1,662) |
2,137 (1,256; 3,365) |
3,348 (2,114; 5,080) |
-789 (-1,327; -45) |
420 (-463; 1,658) |
1,629 (395; 3,360) |
|
High risk | 3,636 (547; 8,087) |
6,117 (970; 13,537) |
8,598 (1,385; 18,982) |
3,471 (383; 7,896) |
5,951 (807; 13,346) |
8,432 (1,226; 18,786) |
|
20-64 yrs: Non-high risk | 2,082 (1,442; 2,706) |
4,529 (3,498; 5,556) |
6,975 (5,551; 8,386) |
-859 (-1,544; -194) |
1,668 (621; 2,709) |
4,194 (2,755; 5,634) |
|
High risk | 17,218 (2,776; 31,688) |
28,957 (4,883; 53,058) |
40,696 (6,955; 74,432) |
16,501 (1,983; 30,967) |
28,264 (4,061; 52,432) |
40,026 (6,162; 73,851) |
|
64 + yrs: Non-high risk | 123 (72; 174) |
357 (287; 426) |
590 (499; 680) |
-315 (-411; -219) |
-82 (-191; 28) |
150 (23; 276) |
|
High risk | 785 (674; 902) |
1,445; (1,267; 1,635) |
2,105 (1,857; 2,368) |
386 (244; 530) |
1,045 (842; 1,250) |
1,703 (1,435; 1,977) |
Notes: 1) Table 2 provides the definition of distribution scenario A.
2) Table 13 provides the assumed levels of vaccine effectiveness, by the four health
outcomes studied, that define "high" level of vaccine effectiveness.
3) Mean, 5th, and 95th percentiles calculated using a Monte Carlo
model incorporating the input variables specified in Tables 3 and 11. Estimates are to the
nearest million $.
4) Elements defining cost of vaccination presented in Table 12.
5) Gross attack rate is defined as the number of clinical cases of illness caused by
influenza that will result in an economic impact.
6) Percentage of population in risk groups, by age category, is given in Table 2.
Table R7: Total net value of vaccinating against an influenza pandemic by level of compliance, age-risk group, cost of vaccination, and gross attack rate: Age distribution scenario B,1 "high" level of vaccine effectiveness2
Vaccination compliance Age, risk group6 |
Mean (5th; 95th
percentiles) net value of vaccination ($ millions, 1995 $)3 |
||||||
$21.26 cost of vaccination4 | $62.26 cost of vaccination4 | ||||||
Gross attack rate (%)5 | Gross attack rate (%)5 | ||||||
15 | 25 | 35 | 15 | 25 | 35 | ||
40% compliance | |||||||
0-19 yrs: Non-high risk | 730 (342; 1,279) |
1,612 (971; 2,520) |
2,494 (1,600; 3,762) |
-412 (-802; 125) |
470 (-171; 1,365) |
1,351 (454; 2,600) |
|
High risk | 4,840 (737; 10,658) |
8,105 (1,267; 17,799) |
11,370 (1,798; 24,945) |
4,736 (634; 10,615) |
8,004 (1,161; 17,814) |
11,272 (1,695; 25,014) |
|
20-64 yrs: Non-high risk | 911 (575; 1,250) |
2,225 (1,674; 2,769) |
3,538 (2,777; 4,292) |
-1,193 (-1,574; -808) |
78 (-494; 638) |
1,349 (571; 2,100) |
|
High risk | 17,507 (2,852; 32,131) |
29,352 (4,939; 53,725) |
41,197 (7,064; 75,333) |
17,024 (2,367; 31,689) |
28,882 (4,398; 53,296) |
40,739 (6,433; 75,003) |
|
64 + yrs: Non-high risk | 46 (15; 76) |
177 (136; 217) |
309 (256; 360) |
-245 (-306; -183) |
-113 (-179; 46) |
19 (-54; 93) |
|
High risk | 830 (722; 941) |
1,475 (1,298; 1,655) |
2,119 (1,872; 2,372) |
569 (444; 692) |
1,215 (1,031; 1,404) |
1,861 (1,609; 2,113) |
|
60% compliance | |||||||
0-19 yrs: Non-high risk | 1,096 (511; 1,895) |
2,418 (1,457; 3,759) |
83,741 (2,397; 5,633) |
-621 (-1,206; 188) |
701 (-263; 2,051) |
2,022 (675; 3,913) |
|
High risk | 7,268 (1,114; 16,134) |
12,170 (1,908; 26,938) |
17,072 (2,697; 37,741) |
7,102 (954; 15,946) |
12,004 (1,746; 26,751) |
16,905 (2,548; 37,549) |
|
20-64 yrs: Non-high risk | 1,273 (764; 1,770) |
3,179 (2,361; 3,985) |
5,086 (3,955; 6,200) |
-1,696 (-2,265; -1,145) |
272 (-571; 1,106) |
2,241 (1,100; 3,375) |
|
High risk | 26,288 (4,323; 48,267) |
44,073 (7,418; 80,742) |
61.858 (10,507; 113,356) |
25,589 (3,527; 47,616) |
43,410 (6,627; 80,142) |
61,231 (9,849; 112,634) |
|
64 + yrs: Non-high risk | 70 (24; 115) |
268 (207; 328) |
466 (390; 541) |
-368 (-462; -276) |
-171 (-273; -67) |
27 (88; 141) |
|
High risk | 1,249 (1,089; 1,419) |
2,218; (1,956; 2,495) |
3,817 (2,817; 3,572) |
849 (664; 1,037) |
1,816 (1,535; 2,102) |
12,782 (2,402; 3,174) |
Notes: 1) Table 2 provides the definition of distribution scenario B.
2) Table 13 provides the assumed levels of vaccine effectiveness, by the four health
outcomes studied, that define "high" level of vaccine effectiveness.
3) Mean, 5th, and 95th percentiles calculated using a Monte Carlo
model incorporating the input variables specified in Tables3 and 11. Estimates are to the
nearest million $.
4) Elements defining cost of vaccination presented in Table 12.
5) Gross attack rate is defined as the number of clinical cases of illness caused by
influenza that will result in an economic impact.
6) Percentage of population in risk groups, by age category, is given in Table 2.
Table R8: Total net value of vaccinating against an influenza pandemic by level of compliance, age-risk group, cost of vaccination, and gross attack rate: Age distribution scenario A,1 "low" level of vaccine effectiveness2
Vaccination compliance Age, risk group6 |
Mean (5th; 95th
percentiles) net value of vaccination ($ millions, 1995 $)3 |
||||||
$21.26 cost of vaccination4 | $62.26 cost of vaccination4 | ||||||
Gross attack rate (%)5 | Gross attack rate (%)5 | ||||||
15 | 25 | 35 | 15 | 25 | 35 | ||
40% compliance | |||||||
0-19 yrs: Non-high risk | 329 (127; 593) |
943 (616; 1,382) |
1,557 (1,102; 2,173) |
-814 (-1,023; -549) |
-199 (-532; 241) |
415 (-46; 1,030) |
|
High risk | 1,313 (217; 2,890) |
2,226 (400; 4,856) |
3,140 (581; 6,820) |
1,203 (106; 2,776) |
2,117 (289; 4,743) |
3,030 (473; 6,703) |
|
20-64 yrs: Non-high risk | 702 (390; 1,013) |
1,876 (1,376; 2,380) |
3,050 (2,358; 3,754) |
-1,339 (-1,669; -982) |
-165 (-703; 365) |
1,009 (287; 1,726) |
|
High risk | 6,554 (1,031; 12,082) |
11,097 (1,883; 20,305) |
15,640 (2,739; 28,532) |
6,053 (521; 11,583) |
10,596 (1,393; 19,790) |
15,139 (2,247; 28,029) |
|
64 + yrs: Non-high risk | 32 (4; 60) |
155 (120; 188) |
277 (234; 319) |
-259 (-321; -199) |
-137 (-202; 73) |
-15 (84; 55) |
|
High risk | 269 (215; 324) |
539 (456; 625) |
809 (696; 927) |
5 (-73; 83) |
275 (174; 378) |
546 (417; 676) |
|
60% compliance | |||||||
0-19 yrs: Non-high risk | 493 (193; 893) |
1,415 (928; 2,077) |
2,336 (1,658; 3,261) |
-1,221 (-1,531; -817) |
-299 (-794; 365) |
622 (-65; 1,545) |
|
High risk | 1,970 (323; 4,334) |
3,340 (592; 7,283) |
4,710 (866; 10,233) |
1,805 (158; 4,162) |
3,175 (432; 7,113) |
4,545 (705; 10,063) |
|
20-64 yrs: Non-high risk | 1,054 (577; 1,531) |
2,814 (2,054; 3,583) |
4,575 (3,520; 5,639) |
-2,008 (-2,553; -1,467) |
-247 (-1,048; 570) |
1,513 (438; 2,619) |
|
High risk | 9,832 (1,529; 18,128) |
16,646 (2,801; 30,478) |
23,461 (4,097; 48,827) |
9,080 (795; 17,367) |
15,894 (2,069; 29,703) |
22,709 (3,329; 42,037) |
|
64 + yrs: Non-high risk | 48 (6; 90) |
232 (180; 285) |
415 (352; 480) |
-389 (-479; -298) |
-206 (-301; -108) |
-22 (-125; 84) |
|
High risk | 403 (324; 487) |
809 (685; 941) |
1,214 (1,044; 1,400) |
7 (-108; 129) |
413 (262; 570) |
818 (629; 1,019) |
Notes: 1) Table 2 provides the definition of distribution scenario A.
2) Table 13 provides the assumed levels of vaccine effectiveness, by the four health
outcomes studied, that define "low" level of vaccine effectiveness.
3) Mean, 5th, and 95th percentiles calculated using a Monte Carlo
model incorporating the input variables specified in Tables 3 and 11. Estimates are to the
nearest million $.
4) Elements defining cost of vaccination presented in Table 12.
5) Gross attack rate is defined as the number of clinical cases of illness caused by
influenza that will result in an economic impact.
6) Percentage of population in risk groups, by age category, is given in Table 2.
Table R9: Total net value of vaccinating against an influenza pandemic by level of compliance, age-risk group, cost of vaccination, and gross attack rate: Age distribution scenario B,1 "low" level of vaccine effectiveness2
Vaccination compliance Age, risk group6 |
Mean (5th; 95th
percentiles) net value of vaccination ($ millions, 1995 $)3 |
||||||
$21.26 cost of vaccination4 | $62.26 cost of vaccination4 | ||||||
Gross attack rate (%)5 | Gross attack rate (%)5 | ||||||
15 | 25 | 35 | 15 | 25 | 35 | ||
40% compliance | |||||||
0-19 yrs: Non-high risk | 415 (195; 704) |
1,087 (730; 1,566) |
1,758 (1,261; 2,430) |
-728 (-953; -439) |
-56 (-418; 426) |
616 (114; 1,287) |
|
High risk | 2,632 (447; 5,776) |
4,424 (782; 9,662) |
6,217 (1,114; 13,557) |
2,522 (335; 5,664) |
4,314 (677; 9,555) |
6,107 (1,009; 13,439) |
|
20-64 yrs: Non-high risk | 320 (69; 571) |
1,239 (841; 1,539) |
2,159 (1,610; 2,715) |
-1,721 (-2,038; -1,414) |
-802 (-1,245; -365) |
118 (-468; 698) |
|
High risk | 10,041 (1,648; 18,431) |
16,908 (2,908; 30,888) |
23,775 (4,166; 43,328) |
9,539 (1,145; 17,921) |
16,407 (2,420; 30,384) |
23,274 (3,676; 42,839) |
|
64 + yrs: Non-high risk | 6 (-20; 33) |
112 (80; 142) |
217 (179; 253) |
-285 (-345; -225) |
-180 (-243; -117) |
-75 (-141; -9) |
|
High risk | 456 (382; 533) |
852 (732; 976) |
1,247 (1,082; 1,421) |
193 (98; 288) |
588 (454; 725) |
983 (809; 1,164) |
|
60% compliance | |||||||
0-19 yrs: Non-high risk | 623 (297; 1,059) |
1,630 (1,099; 2,353) |
2,637 (1,898; 3,646) |
-1,091 (-1,428; -651) |
-84 (-622; 638) |
923 (180; 1,931) |
|
High risk | 3,948 (665; 8,664) |
6,637 (1,168; 14,497) |
9,325 (1,668; 20,331) |
3,783 (505; 8,495) |
6,472 (1,002; 14,335) |
9,160 (1,496; 20,171) |
|
20-64 yrs: Non-high risk | 481 (91; 868) |
1,859 (1,253; 2,477) |
3,238 (2,398; 4,085) |
-2,581 (-3,051; -2,117) |
-1,202 (-1,863; -537) |
176 (-693; 1,067) |
|
High risk | 15,061 (2,446; 27,694) |
25,363 (4,324; 46,391) |
35,664 (6,226; 65,080) |
14,309 (1,707; 26,913) |
24,610 (3,569; 45,626) |
34,911 (5,447; 64,388) |
|
64 + yrs: Non-high risk | 10 (-30; 50) |
167 (120; 214) |
325 (269; 382) |
-426 (-517; -338) |
-270 (-363; -176) |
-112 (-210; -12) |
|
High risk | 684 (574; 803) |
1,278 (1,098; 1,473) |
1,871 (1,621; 2,143) |
289 (149; 435) |
882 (683; 1,092) |
1,475 (1,213; 1,756) |
Notes: 1) Table 2 provides the definition of distribution scenario B.
2) Table 13 provides the assumed levels of vaccine effectiveness, by the four health
outcomes studied, that define "low" level of vaccine effectiveness.
3) Mean, 5th, and 95th percentiles calculated using a Monte Carlo
model incorporating the input variables specified in Tables3 and 11. Estimates are to the
nearest million $.
4) Elements defining cost of vaccination presented in Table 12.
5) Gross attack rate is defined as the number of clinical cases of illness caused by
influenza that will result in an economic impact.
6) Percentage of population in risk groups, by age category, is given in Table 2.
Table R10: The annual "insurance premium"1,2 for planning, preparing and practicing to respond to the next influenza pandemic: Differences due to probability of event, attack rate, vaccine effectiveness, compliance and cost of vaccination
Gross attack rate | Cost of vaccination | Actuarially fair annual premium ($ millions) | ||||||
"Low" vaccine effectiveness3 x 40% compliance | "High" vaccine effectiveness3 x 60% compliance | |||||||
Probability of pandemic | Probability of pandemic | |||||||
1 in 30 years | 1 in 60 years | 1 in 100 years | 1 in 30 years | 1 in 60 years | 1 in 100 years | |||
15% | $21/vaccinee | 306 (122) |
153 (61) |
92 (37) |
872 (341) |
435 (170) |
262 (103) |
|
$62/ vaccinee | 162 (122) |
81 (61) |
48 (37) |
654 (341) |
326 (170) |
196 (103) |
||
25% | $21/vaccinee | 561 (204) |
280 (102) |
168 (61) |
1,528 (569) |
762 (284) |
459 (171) |
|
$62/ vaccinee | 416 (204) |
207 (102) |
125 (61) |
1,311 (569) |
653 (284) |
394 (171) |
||
35% | $21/vaccinee | 815 (286) |
406 (142) |
245 (86) |
2,184 (796) |
1,089 (397) |
656 (239) |
|
$62/ vaccinee | 670 (286) |
334 (142) |
201 (86) |
1,967 (796) |
980 (397) |
591 (239) |
Note: 1) The "insurance premium" is defined here as the amount of
money to be spent each year to plan, prepare and practice (the 3 P's) to ensure that such
mass vaccinations can take place if needed. The premium is calculated as follows:30
Annual "insurance premium" = net returns from vaccination x annual probability
of pandemic occurring.. The net returns to vaccination are given in Tables R6-R9.
2) Although the "premium" is to pay for planning, practicing and preparing a
response to the next influenza pandemic, the mathematically optimal allocation of such
funds requires a separate set of calculations.
3) "Low" and "high" levels of vaccine effectiveness are defined in
Table 13.
Table R11: Setting priorities: Samples of lists to decide which age and risk group should be vaccinated first
Priority | Criteria for prioritization (age - risk group) | ||
Risk of death1 | Percentage of deaths2 | Returns to vaccination3 | |
1 (top) | High risk 65 + yrs | High risk 20 - 64 yrs | High risk 20 - 64 yrs |
2 | Non-high risk 65 + yrs | High risk 65 + yrs | High risk 0 - 19 yrs |
3 | High risk 0 - 19 yrs | High risk 0 - 19 yrs | Non-high risk 20 - 64 yrs |
4 | High risk 20 - 64 yrs | Non-high risk 65 + yrs | Non-high risk 0 - 19 yrs |
5 | Non-high risk 20 - 64 yrs | Non-high risk 20 - 64 yrs | High risk 65 + yrs |
6 (bottom) | Non-high risk 0 - 19 yrs | Non-high risk 0 - 19 yrs | Non-high risk 65 + yrs |
Notes: 1) Priorities set by "risk of death" are set according to
lower-limit estimates of deaths per 1,000 population for each age and risk group (see
Table 3).
2) Priorities set using percentage distribution of total deaths. Data from Tables R1 and
R2.
3) The priorities based on returns to vaccination use the mean net value from vaccination
(Tables R6-R9).
Figure legends
1A complete plan detailing a response to an influenza pandemic should include items such as a definition of a pandemic, "trigger" points that will initiate various steps in the response plan, and details regarding how to actually deploy the intervention. While a U.S. Federal influenza pandemic plan is still being developed, a guide for state and territorial health officials to aid them in developing plans for their jurisdictions is available at the following Internet website (printed copies can be obtained from the author): http://www.cdc.gov/nip/temp/pandemic-flu.htm
2We limited our examination of possible interventions to those involving influenza vaccines. We did not consider the use of anti-viral drugs for influenza prophylaxis. The reasons for the omission include: a) possible lack of supplies; b) first priority for such drugs may be for treatment; and, c) the side-effects from the drugs, particularly amantadine, make then unsuitable for long term prophylaxis for many occupations, such as drivers, heavy construction operators etc.
3Although it is recommended that persons age 9 years or older need only one dose of the currently available vaccine, children under 9 years may require 2 doses to elicit a "satisfactory" immune response (2). In a pandemic, which will be caused by a new subtype of influenza, it may be that everybody will require 2 doses in order to induce a "satisfactory" immune response, such that the vaccinee is protected against disease, or at least has a less severe health outcome.