Cover Image

HARDBACK
$52.95



View/Hide Left Panel

5
Review of the Analytic Model

This chapter reviews in greater detail the components of the committee’s analytic model, which were outlined in Chapter 4.

UNIT OF ANALYSIS

The basic analysis included 27 separate cases. Each case represents a specific combination of pathogen or condition, a vaccine candidate, and a population targeted to receive the vaccine. The committee examined 26 candidate vaccines, but included two distinct and alternative target populations for one candidate, thus 27 separate cases. As explained previously, the conditions selected for analysis are a mix of infectious diseases, cancers, and autoimmune disorders of health significance in the United States, and are conditions for which the committee judged that an adequate science base for vaccine development existed. Targeted populations were defined on the basis of factors such as age, health status, or risk of exposure (e.g., geographic region). Among the target populations are infants, adolescents, pregnant women, travelers, and immigrants to specific geographic areas. For some conditions, two complementary target populations (e.g., infants and regional immigrants for borrelia) were defined. For others, alternative vaccination strategies targeting different populations were considered, and each was examined as a separate case. For group B streptococcus, for example, one case is based on adolescent girls as the target population for immunization and a second case considers immunization of pregnant women. The target population influences factors such as the number of doses of vaccine used per year and the time interval between average age at the time of vaccination and average age at the time of onset of a condition.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking 5 Review of the Analytic Model This chapter reviews in greater detail the components of the committee’s analytic model, which were outlined in Chapter 4. UNIT OF ANALYSIS The basic analysis included 27 separate cases. Each case represents a specific combination of pathogen or condition, a vaccine candidate, and a population targeted to receive the vaccine. The committee examined 26 candidate vaccines, but included two distinct and alternative target populations for one candidate, thus 27 separate cases. As explained previously, the conditions selected for analysis are a mix of infectious diseases, cancers, and autoimmune disorders of health significance in the United States, and are conditions for which the committee judged that an adequate science base for vaccine development existed. Targeted populations were defined on the basis of factors such as age, health status, or risk of exposure (e.g., geographic region). Among the target populations are infants, adolescents, pregnant women, travelers, and immigrants to specific geographic areas. For some conditions, two complementary target populations (e.g., infants and regional immigrants for borrelia) were defined. For others, alternative vaccination strategies targeting different populations were considered, and each was examined as a separate case. For group B streptococcus, for example, one case is based on adolescent girls as the target population for immunization and a second case considers immunization of pregnant women. The target population influences factors such as the number of doses of vaccine used per year and the time interval between average age at the time of vaccination and average age at the time of onset of a condition.

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking IMPLEMENTING THE ANALYSIS Basic Model The essential components used to calculate the cost-effectiveness ratio for each candidate vaccine are the costs of vaccine development, the costs of administering the vaccine to the target population, the reduction in the cost of care expected with the use of the vaccine, and the expected gain in health benefits. The basic calculation can be represented by the following equation: CEV=(CD+CI−CC)/Q, (1) where CEV is the cost-effectiveness ratio for case V, which is the analysis for a specific combination of health condition or pathogen, vaccine type, and target population; CD is the cost of vaccine development; CI is the annual cost of immunizing the target population; CC is the annualized costs of care averted by use of the vaccine; and Q is the annualized health benefit from use of the vaccine. Each of the components of this basic equation must be discounted to account for the time lag between the present and when the cost or health benefit will be realized. The anticipated health benefits and changes in the cost of care must also be reduced to reflect the estimated limits to the efficacy and use of the vaccine. The costs of immunization are also reduced to reflect less than universal use of the vaccine. This more complete characterization of the model can be represented as follows: CEV=[[rCD+{[CI /(1+r)T(use)] • U} −{[Cc/(1+r)T(use) + T(lag)] • Eff • U}]]/ {[Q/(1+r)T(use)+T(lag)] • Eff • U}, (2) where r is the discount rate; T(use) is the time until steady-state vaccine use, which includes the time until licensure plus the time to adoption at the assumed rate of use; U is the assumed rate at which the target population will use the vaccine; T(lag) is the time between the use of the vaccine and the realization of health benefits; and Eff is the assumed efficacy of the vaccine. CEV, CD, CI, Cc, and Q are as defined above for Equation 1. Each of the components of the analysis is reviewed in more detail in subsequent sections of this chapter. Performing the Analysis The elements of Equation 2 were operationalized through a multipage spreadsheet template developed with the Excel spreadsheet package, version 5

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking (Microsoft Corporation, 1984–1994), operating on a personal computer. Using the basic template, a separate file was created for each case. The following data were entered: age-specific incidence and death rates, average age at immunization, morbidity scenarios and associated quality-adjustment weights, typical health care provided for the condition and its associated costs, size of the target population, number of vaccine doses required, cost per dose, estimated vaccine efficacy, anticipated steady-state vaccine utilization rates among the target population, remaining development costs, expected time until licensure, and time from licensure until anticipated utilization rates are reached. The basic template was modified to accommodate variations among the cases in the patterns of illness (e.g., distinctive characteristics by age or sex), in the features of the morbidity scenarios (e.g., numbers of health states included and mix of acute and chronic health states), and in the types of care required. Detailed descriptions of the data and calculations used for each case are provided in the Appendixes. CALCULATION OF HEALTH BENEFITS Health benefits were measured using quality-adjusted life years (QALYs). Described below is the multistep process followed by the committee to estimate vaccine-related health benefits and calculate QALYs. Quality-Adjusted Life Years QALYs reflect the combined impact of morbidity and mortality on the health-related quality of years of life lived. To calculate QALYs, a quality-adjustment weight is applied to each period of time during which a person experiences a changed health state due to a particular condition. These quality-adjusted periods can be summed over a person’s expected lifetime (or some other specified period of time). This can be illustrated in a simplified form as Q=W1t1+W2t2+W3t3, (3) where Q is the total QALYs experienced by the individual, W1 is the quality-adjustment weight associated with health state 1, and t1 is the amount of time spent in that health state. Thus, the lifetime QALYs for an individual who lives for 70 years in perfect health, experiences six months of impaired health from condition 1, and dies five years sooner than the average life expectancy can be represented as Q=(1.0 • 70)+(W1 • 0.5)+(0.0 • 5.0).

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking Quality-Adjustment Weights Based on the Health Utilities Index The committee considered two basic options for determining the quality-adjustment weights used to represent the impact of morbidity associated with the conditions under study. One option was for the committee, on the basis of its judgment and that of other experts, to assign quality-adjustment weights to each health state in each morbidity scenario associated with each condition. The other option was to use an existing generic health status assessment tool to characterize each health state. The committee chose to use a standard assessment tool to promote the comparability of the assessments for each condition. This approach also allows others who use the committee’s work to use the same instrument to make their own assessments of these or other conditions. The committee selected the Health Utilities Index (HUI) Mark II (Feeny et al., 1995: Torrance et al., 1995). The HUI Mark II characterizes morbidity by using seven health attributes (sensation, mobility, emotion, cognition, self-care, pain, and fertility), each of which is divided into three, four, or five levels. Each level has a fixed quantitative score between 0 and 1.0 representing the strength of the “preference” for that level of morbidity relative to full health (1.0) or death (0). A health state is described by assigning to it a specific level from each attribute. The HUI quality-adjustment weight for the health state is then derived from the following formula: U=1.06 (b1 • b2 • b3 • b4 • b5 • b6 • b7)−0.06, (4) where U is the utility of the health state (i.e., the quality weight), and bx is the score for the level assigned for attribute x (Torrance et al., 1995). U corresponds to Wi in Equation 3. Although HUI Mark II was originally developed for a study of childhood cancer survivors, it has been adapted for use with adult populations. It has also been used with the Ontario Health Survey (Berthelot et al., 1992; Roberge et al., 1995) and the Canadian National Population Health Survey (NPHS) (Catlin and Will, 1992; Wolfson, 1996) to develop provisional estimates of age-specific health status at the population level.* *   Final estimates for the NPHS will be based on the scoring system under development for the HUI Mark III, a revised HUI with eight component attributes (Torrance et al., 1992; Boyle et al., 1995).

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking Steps in the Calculation of Anticipated Health Benefits from Vaccine Use The calculation of health benefits from vaccine use for each condition under study is a multistep process. The steps are listed in Box 5–1 and are described in the sections that follow. Develop Morbidity Scenarios For each condition studied, the committee, with the advice of outside experts, developed morbidity scenarios to describe the characteristic sequences of acute or chronic health states, the duration of each health state, and the proportion of persons with the condition who experience each scenario. The scenarios also capture the premature mortality associated with a condition but which is delayed 1 or more years beyond the onset of the condition. For most of the conditions included in the committee’s analysis, several scenarios were required to depict the associated morbidity. Some pathogens or conditions affect specific subpopulations in distinctive ways. For example, the consequences of infection with chlamydia, a sexually transmitted disease, are very different in men and women. For those conditions, separate morbidity scenarios were defined for each appropriate subpopulation. The subpopulation analyses were combined in the final stages of the calculation of the cost-effectiveness ratio. Calculate Quality-Adjustment Weights Once the morbidity scenarios were developed, the committee reviewed each health state and assigned a level in each of the seven attributes of HUI Mark II. To better reflect the range of morbidity represented by some health states, the committee estimated the proportion of cases of illness that should be assigned to each level of a specific health utility attribute. The preference score for the attri bute was calculated as a weighted average of the scores for each level. The quality-adjustment weight for the health state was obtained by combining the scores for each attribute according to Equation 4. The calculations were performed using an Excel spreadsheet template. Calculate Discounted Quality-Adjusted Life Expectancies The quality-adjustment weights for each condition’s morbidity scenarios (described above) were used to measure QALYs without the vaccine intervention. Quality-adjustment weights reflecting the average health status of the

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking BOX 5–1 Steps Used in Calculation of Vaccine-Related Health Benefits for a Specific Condition Develop morbidity scenarios, including estimate of percentage of cases represented by each scenario. Calculate quality-adjustment weights for each health state in each scenario. Calculate discounted quality-adjusted life expectancies. Establish age-specific incidence and death rates for the condition. Calculate average age at onset and average age at death. Calculate average interval between vaccination and onset of illness. Calculate average interval between age at onset of illness and age at condition-related death. Calculate life expectancy at average age at onset and average age at death. Calculate a baseline quality-adjustment weight for the population. Adjust the health state weights to reflect the population baseline. Calculate discounted QALYs for each health state with and without the impact of condition-related morbidity. For each morbidity scenario, sum the QALYs for the component health states. For each morbidity scenario, subtract QALYs with the condition from QALYs without the condition (i.e., calculate the condition-related QALYs lost). Weight the condition-related QALYs lost in each scenario by the percentage of cases represented by that scenario and sum across all scenarios. Multiply the sum of QALYs lost by the total number of cases for the condition. population were used to measure QALYs with the intervention. To measure the impact of mortality and lifetime impairment, standard life table values were “quality adjusted” for the average health status of the population and discounted to their present value. Described here are the population-level quality adjustments used in the committee’s analysis and the calculation of the discounted quality-adjusted life table values. Population-Based Quality-Adjustment Weights The analysis takes into account the underlying average health status of the population, independent of a specific condition or use of a candidate vaccine. Although an individual might be considered to experience periods of perfect health, represented by a quality-adjustment weight of 1.0, the health status of a population will reflect a range of individual levels of health status and should not be represented by a quality-adjustment weight set at 1.0. This average health status in the population is assumed to be the maximum health status level that can be achieved by use of any of the vaccines considered in the committee’s analysis. To represent the average health status of

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking the U.S. population, the committee adopted quality-adjustment weights representing the HUI Mark II-based age- and sex-specific health status results from the Canadian NPHS (Wolfson, 1996). These weights reflect population-based values calculated from self-reported features of health status. They range from .93 for men aged 15–24 to .64 for women aged 85 and older. The separate NPHS health status estimates for men and women were combined as a weighted average based on the age-specific proportions of men and women in the U.S. population. Since NPHS did not produce health status estimates for the population younger than 15 years of age, the committee chose, in the absence of other data, to apply the value for ages 15–24 to the younger age groups. See Table 5–1 for the quality adjustment weights for health status of the general population. Life Expectancies The discounted quality-adjusted life expectancies used in the committee’s analysis are based on the 1993 abridged U.S. life tables for the total population and for males and females separately (National Center for Health Statistics, 1993). These adjustments were incorporated through calculations that use the life table stationary population in each age interval (the nLx values, in life table notation), which can also be interpreted as the number of person-years lived from ages x to x+n. The quality adjustment was incorporated by multiplying the number of person-years by the corresponding age- and sex-specific quality-adjustment weight for the population (Sullivan, 1971; Erickson et al., 1995): nLx*=wx • nLx. (5) The quality-adjusted person-years in each age interval were then discounted for the period between age at the onset of a condition and the midpoint of the age interval. Because the average age at onset for the conditions under study varies from infancy to older than 70 years, a series of life expectancy calculations were made for selected ages at onset (0, 1, 5, 10, 20, 30, 40, 50, 60, 70, 80, and 85 years of age). The discounting period was defined as the difference between the midpoint of the age interval and the age at onset. For example, the discounting period for person-years lived in the age interval 60–65 with disease onset at age 20 was 42.5 years (62.5–20=42.5). Once the discounted person-year values were obtained, standard life table calculations were used to calculate discounted quality-adjusted life expectancies at the selected ages of onset,* with life expectancy at age x designated as ex*. *   To calculate life expectancies, the person-years lived are summed from the oldest to the youngest ages. For each age interval, a cumulative total of all person-years lived in that interval and at all older ages is obtained. Dividing the cumulative total for age interval x (designated Tx) by the number of persons in the life table population alive at the beginning of that age interval (lx) gives the life expectancy (ex) at age x. See Erickson et al. (1995) for additional discussion and illustration of life table calculations incorporating quality adjustments.

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking TABLE 5–1 Population-Based Age-Specific Health Status (quality adjustment weights) Age Group Males Females Combined <1     0.92 1–4     0.92 5–14     0.92 15–24 0.93 0.92 0.92 25–34 0.93 0.92 0.92 35–44 0.92 0.91 0.92 45–54 0.89 0.87 0.88 55–64 0.87 0.86 0.86 65–74 0.85 0.83 0.84 75–84 0.81 0.76 0.78 ≥85 0.71 0.64 0.66 NOTE: Values are based on the overall population (both household and institutionalized); combined values calculated as weighted average of male and female values (weights from distribution of U.S. population); no direct estimates were made for population younger than age 15, values for ages 15–24 were assumed to apply. SOURCE: Wolfson, 1996. Establish Age-Specific Incidence and Death Rates Estimates of current age-specific incidence and mortality for each condition were assembled. Rates and numbers of cases and deaths were estimated for the following age groups: under 1 year, 1–4, 5–14, 15–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84, and ≥85. Although some of the conditions included in the analysis (e.g., Lyme disease [caused by Borrelia burgdorferi], gonorrhea, shigellosis, and tuberculosis) are designated as “reportable” and numbers of reported cases are published by the CDC (1994), the completeness of reporting varies by condition, reflecting both undiagnosed cases and incomplete reporting of diagnosed cases. Separate surveillance programs by CDC and others for conditions such as tuberculosis are the basis for some estimates of incidence. In the absence of surveillance programs, some estimates have been based on data from state- or community-level studies. Others rest largely on expert judgment. The specific sources of data for each condition are described in the Appendixes. It is important to emphasize that the analysis uses incidence data, that is, the number (or rate) of new cases that would be expected during 1 year. For chronic illnesses such as multiple sclerosis, these data will differ from the prevalence estimates that are often reported.

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking Calculate Time Intervals for Discounting Future Health Benefits As noted above, the timing of the health benefits expected from the use of the potential new vaccines included in the committee’s analysis will vary depending on the intervals between the typical age at immunization and the age at onset of an illness or the age at death. The intervals calculated for the analysis are the following: To, the time from vaccination to the average age at onset of illness, and Td, the difference between the average age at onset of illness and average age at the time of illness-related death. Td is of interest for those acute conditions for which the age at death from a condition differs markedly from the overall age of patients with that condition. The time interval related to premature death following a period of chronic illness is accounted for separately. The distribution of cases and deaths by age was used to estimate the average age at the time of onset of the condition and the average age at the time of death. Specifically, the midpoint of each age group was weighted by the proportion of cases or deaths occurring in that age group: A=∑(ag • Pg), (6) where A is the average age of onset (Ao) or death (Ad), a is the midpoint of age group g, and P is the proportion of cases or deaths in age group g. Age at vaccination, AV, was determined by the vaccination strategy and the target population specified by the committee. Most cases fall into one of the following categories: Target Population Age at Vaccination Infants 6 months 6 months Adolescents 12 years Pregnant women Average age of mothers at first births, minus 2 months (24.7 years) New cases (therapeutic vaccines) Age at diagnosis (assumed to equal average age at onset) See the Appendixes for detailed information on the age at vaccination used for each case. Thus, the interval from vaccination to onset of illness is calculated as To=Ao−Av, and the difference between age at onset of illness and illness-related death is calculated as Td=Ad−Ao. Calculate Condition-Related Life Expectancies The discounted quality-adjusted life expectancies for selected ages were used to calculate the life expectancy in the population at the average age at onset

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking of a condition and at the average age of condition-related death. Specifically, the discounted and adjusted life expectancies were weighted by the proportion of cases or deaths occurring in the appropriate age group: e*=∑(eg* • Pg), (7) where e* is the discounted quality-adjusted life expectancy in the population at the average age at onset (eo*) or at the average age at condition-related death (ed*); Pg is the proportion of cases or deaths in age group g. For example, the life expectancy at age 20 years was weighted by the proportion of cases (or deaths) in the age group 15–24. Adjust for the Underlying Health Status of the Population HUI Mark II-based quality-adjustment weights for health states were calculated from attribute scores assigned by the committee without reference to the underlying health status of the general population. Because that underlying health status declines with age, the committee made an adjustment to reflect the differences in the age patterns of the conditions that it considered. First, a population “baseline” quality-adjustment weight (w') for a condition was calculated as the weighted average of the age-specific quality-adjustment weights for the general population. The age-specific values were weighted by the proportion of cases at each age and were summed to produce the baseline weight for that condition. In general, one population baseline weight was calculated for each condition under study. If separate analyses were performed for specific subpopulations, a baseline weight was calculated for each subpopulation. Then, the original quality-adjustment weight for each health state was multiplied by the baseline weight for the condition to produce a baseline-adjusted quality weight: wi=hi • w', (8) where wi is the baseline-adjusted quality weight for health state i, hi, is the quality adjustment weight for health state i calculated from the committee’s HUI Mark II scores; and w' is the population baseline quality-adjustment weight. Calculate Discounted QALYs for Each Health State Morbidity For each health state, two sets of QALYs—QALYs with the condition under study and QALYs without the condition under study—were calculated. QALYs with the condition were derived by multiplying the baseline-adjusted quality weight for the health state by the duration of the health state. A

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking parallel calculation in which the population baseline quality-adjustment weight was multiplied by the duration of the health state determined the QALYs that the general population would experience without the condition under study. Both sets of QALYs were then discounted to adjust for the time from average age at vaccination to average age at onset. Calculation of QALYs when the condition under study is present can be represented as where qi is the QALY value for health state i, wi is the baseline-adjusted quality weight for the health state, ti is the duration of the health state, r is the discount rate, To is the time from average age at vaccination to average age at onset, and tm is the duration of health states that intervene between the onset of the condition and health state i. If related health states that persist for at least 1 year intervene between the onset of the condition and health state i, the discount period must be increased by tm. To calculate q'i, the corresponding QALY value for the general population, wi is replaced by w', the population baseline quality-adjustment weight. Calculation of QALYs for two kinds of health states required alternative approaches. First, for states that persist for a specified multiyear period rather than a period of days or weeks, allowance must be made for discounting the stream of QALYs associated with the health state. In these health states, QALYs discounted to the beginning of the health state were calculated by using the Excel present-value function. This value is then further discounted for the period To+tm. Second, for health states that persist for the normal remaining lifetime, the discounted quality-adjusted life expectancy at onset (eo*) was used as the value of the QALYs that the general population would experience without the condition under study. To estimate QALYs with the condition, eo* was multiplied by the committee’s quality-adjustment weight for the health state hi (Adjustment for the baseline health status of the population is already reflected in the quality-adjusted life expectancy.) Additional discounting for the period from vaccination to onset (To+tm) is performed. Mortality The impact of mortality on QALYs was estimated as described above for morbidity associated with lifetime health states. The death rates obtained for each condition reflect deaths that occur within a short time following onset of the condition. The discounted quality-adjusted life expectancy at the average age at death (ed*) provides the estimate of QALYs lost due to deaths, which must be discounted for the period To+Td, where To is the time from average age at vaccination to average age at onset and Td is the difference between average age at onset of illness and average age at illness-related death. If the age pattern of deaths is similar to the age pattern of illness, Td=0. If deaths are more common among younger or older patients, Td≠0.

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking For some conditions, deaths occur following a period of chronic impairment. These have been assumed to be additional condition-related deaths beyond those reflected in the age-specific death rates. The age at death is calculated as the average age at onset of the condition plus the duration of health states experienced between onset and death. The discounted quality-adjusted life expectancy at the age at death is used to calculate the QALYs lost due to these deaths. Additional discounting for the period To plus the interval between onset and death must be included. Life expectancy at the time of these deaths was estimated by interpolation from the values calculated for specific ages of onset of illness (see above). Calculate Total QALYs Gained with Vaccine Use The QALYs associated with each health state were combined. For each morbidity scenario, the state-specific QALYs were summed, and the QALYs lived with the condition under study were subtracted from the QALYs for the general population without the condition. This provides an estimate of the QALYs that could be gained in each scenario with vaccine use. The scenario-specific QALYs to be gained were multiplied by the proportion of cases of illness experiencing that scenario and were summed across all scenarios. This total was multiplied by the number of cases, or by the number of deaths for the QALYs associated with mortality, to calculate the overall benefit that the use of a vaccine for the condition under study would be expected to have in the population. If subpopulations were used in the analysis, the results were calculated for each subpopulation and the subpopulation results were summed to produce an estimate of the total health benefit. Q;s=[∑j (qj • Pj)] • Ns, (10) where Q;s is the total QALYs to be gained with vaccine use across all subpopulations, qj the individual QALY gain in scenario j; Pj is the proportion of cases of illness experiencing scenario j, and Ns is the total number of cases of illness in subpopulations. COST FACTORS As shown in Equation 1, the cost components of the numerator of the cost-effectiveness ratio include the costs of vaccine development (CD), the cost of vaccine use (CI), and the reduction in health care and related costs (CC) that would be expected with vaccine use. All cost estimates are presented in constant dollars.

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking Vaccine Development The costs of future research and clinical trials that would be needed to complete development of a vaccine and have it licensed, designated CD, are a mix of public- and private-sector expenditures. They were assumed to fall at one of six levels: $120 million, $240 million, $300 million, $360 million, $390 million, or $400 million. A development cost was assigned to each vaccine on the basis of the committee’s assessment of the current stage of the vaccine’s development. In terms of the analysis, the cost of vaccine development is treated as an amortized fixed cost. The committee also assigned each candidate vaccine to one of three development intervals: 3, 7, or 15 years. Discounting incorporated this development interval to adjust for the differences in the times when the associated costs and benefits of the vaccines will be realized. Vaccine Use The cost of vaccine use is a function of the cost per dose of the vaccine, the cost to administer the vaccine, the number of doses each person must receive to be fully immunized, and the size of the population targeted to receive the vaccine: CI=(d+a) • Dose • Pop, (11) where CI is the annual cost of immunizing the target population, d is the cost per dose of the vaccine, a is the cost of administering a dose of vaccine, Dose is the number of doses each person must receive, and Pop is the size of the target population. A vaccine’s cost per dose is represented by the purchase price rather than the marginal cost to manufacturers of producing a single dose. The cost of prophylactic vaccines was assumed to be either $50 or $100. The cost of therapeutic vaccines was set at $500. The marginal cost of administering a dose of vaccine was set at $10. For most vaccines, it was assumed that three doses would be needed to achieve full immunization. In specific cases (e.g., influenza vaccine), the number of doses required was altered to match the available evidence. The committee defined a target population for each vaccine considered, and the size of that target population was determined by using current estimates of the U.S. population by age. For vaccines intended for use in infants or adolescents, the target population was assumed to equal the birth cohort or the population at age 12, respectively. For vaccines intended for use in special subpopulations (e.g., travelers, residents in a specific geographic region, persons with chronic illness) the basis for the estimate of the size of the target population is described in detail in the Appendixes. It was assumed that 5 years would be required following licensing to achieve stable, maximum levels of use of preventive vaccines. For therapeutic

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking vaccines, this interval was assumed to be 2 years. The appropriate interval was incorporated into the discounting calculations. Cost of Care To estimate the cost of health care associated with each condition, the committee developed a set of “clinical scenarios” that specified the health services required, including hospitalizations, procedures, medications, office visits, rehabilitation services, and long-term institutional care. An appropriate unit of service (e.g., hospital days, doses of medication) was specified, and the amount of care received was defined in terms of those units. Costs, represented by charges for specific services, also were specified in terms of units of service. For inpatient hospitalizations, hospital costs were estimated by using diagnosis-related group average national payments by the Health Care Financing Administration (HCFA) (St. Anthony’s Publishing, Inc., 1995). Outpatient costs and inpatient physician visits were estimated from HCFA data as well (HCFA, 1995). For these costs, the committee estimated general categories of costs (outpatient physician visit with and without tests, etc.) and applied these to the morbidity scenarios. See Table 5–2 for examples of unit costs. In addition, for each form of care, the committee specified the proportion of patients within the scenario that received that form of care. It was assumed that all costs of care associated with the condition under study would be averted with vaccine use. The total cost of each form of care was calculated as cc=(uc • nc) • pc • ps • N, (12) where cc is the total cost of type of care c for health state i, uc is the unit cost of this form of care, nc is the number of units of care received, pc is the percentage of patients within the scenario that received this form of care, ps is the percentage of all patients that experience this scenario, and N is the total number of patients with the condition under study. The cost cc was then discounted to adjust for To, the time from average age at vaccination to average age at onset of the condition, plus tm, the duration of intervening health states that persist for at least 1 year. For continuing care required for a specified multiyear period rather than a period of days or weeks, it was necessary to allow for discounting of the stream of future costs. As in the QALY calculations, these costs were discounted to the beginning of the health state by using the Excel present-value function. For some health states, care continues for the remaining lifetime. The length of the remaining lifetime was estimated from the unadjusted 1993 life table life expectancy value (NCHS, 1993) at the average age at onset of the health state.

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking TABLE 5–2 Examples of Health Care Cost Estimates Used Outpatient Costs Physician A $50 Physician B (specialist) $100 Physician C (in hospital) $150 Medication A (nonprescription) $10 Medication B (inexpensive prescription) $50 Medication C (expensive prescription) $150 Diagnostic A $50 Diagnostic B $100 Diagnostic C $500 Hospitalization Costs (per admission; based on hospitalization) Multiple Sclerosis $3,000 Pneumonia $3,000 Viral Meningitis $3,000 Tuberculosis $6,000 Cellulitis $3,000 Amputation $7,000 Cirrhosis $5,000 Ectopic Pregnancies $3,000 Ulcer $3,000 Digestive $4,000 Gastroenteritis $2,000 Complicated delivery (additional cost) $2,000 Infectious myocarditis $3,000 Diabetic complications $3,000 Melanoma $4,000 The total discounted cost of care that would be averted with vaccine use, CC in Equation 1, was obtained by summing the costs across all health states. As was done for health benefits, calculations for separate subpopulations were summed to produce an estimate of total costs. VACCINE EFFICACY AND UTILIZATION The analysis includes adjustments for incomplete efficacy and use of the candidate vaccines, either of which will reduce the expected health benefits and savings in the cost of care. A lower utilization rate will also reduce the costs ssociated with vaccinating the target population. These adjustments were made by multiplying the QALY and cost measures by the assumed efficacy (Eff) and utilization rates (U) (see Equation 2).

OCR for page 93
Vaccines for the 21st Century: A Tool for Decisionmaking It was assumed that preventive vaccines would achieve an efficacy level of 75%. The efficacy of therapeutic vaccines was assumed to be 40%. A utilization rate of 10%, 30%, 50%, 60%, or 90% was assigned to each vaccine. COST-EFFECTIVENESS RATIOS For each condition, cost-effectiveness ratios were calculated at three stages in the analysis. The first ratio examines the potential impact of the vaccine on morbidity and costs under the assumption that the vaccines are available immediately without any additional cost or time for development and that they are fully efficacious and are used by the entire target population. This comparison focuses attention on what might be considered an ideal vaccine benefit. The second cost-effectiveness ratio factors in the adjustments for incomplete efficacy and use, which tend to increase the cost of achieving the anticipated health benefit. The final ratio, which corresponds to Equation 2, shows the impact of the time and money needed to develop these vaccines. Some vaccines that promise substantial benefit require longer and more expensive periods of development, whereas others that offer smaller benefits are expected to be available more quickly and cheaply. In general, the committee found that the adjustments for efficacy and utilization had a more substantial impact on a vaccine’s cost-effectiveness than the additional time and cost needed for development.