Projected Cost-effectiveness of Pneumococcal Conjugate Vaccination of Healthy Infants and Young Children

Streptococcus pneumoniae is the leading bacterial cause of meningitis, sepsis, pneumonia, and otitis media (OM) among US children.^{1 - 2} Emergence of drug-resistant pneumococci has substantially complicated therapy of these infections and has focused attention on the need for effective prevention strategies.³ New pneumococcal conjugate vaccines designed to stimulate immunity in infants and young children who respond poorly to pneumococcal polysaccharide vaccine could prevent much of this disease.^{4 - 5} The first such vaccine was licensed by the Food and Drug Administration on February 17, 2000.

National recommendations about new preventive health interventions ideally should be based on clinical effectiveness. For many health care policymakers, cost-effectiveness is another important consideration. Our aim was to evaluate the projected health benefits, costs, and cost-effectiveness of routine pneumococcal conjugate vaccination of healthy infants, who are currently being considered as the primary target group for vaccination in the United States. We also analyzed the potential cost-effectiveness of catch-up vaccination of children during the transitional years of vaccine implementation.

We constructed a decision tree (Figure 1) to compare 2 major options. Under "no vaccination," infection with S pneumoniae could cause meningitis, bacteremia, pneumonia, simple or complex OM, or OM with tympanostomy tube placement. Meningitis could lead to death, disability, deafness, or no sequelae. Bacteremia was defined as all other bacteremic infections, including bacteremic pneumonia, and could result in death or no sequelae. Under "vaccination," the incidence of each infection was reduced in proportion to the demonstrated efficacy of the vaccine against that infection.⁶ We assumed that routine vaccination of healthy infants required 4 doses (at 2, 4, 6, and 12-15 months), and that catch-up vaccination of children aged 2 to 4.9 years required only 1 dose.

Our model took into account the fact that the pneumococcal conjugate vaccine currently under consideration includes 7 serotypes, but other pneumococcal serotypes may also cause disease. In addition, pneumonia and OM may be caused by organisms other than pneumococci and children with these conditions do not typically have bacterial cultures performed to confirm origin. For pneumonia and OM, we designed our model to maximize consistency with the best available data on disease incidence and vaccine efficacy from the Northern California Kaiser Permanente (NCKP) randomized trial and from published sources on disease epidemiology.^{6 - 8} The model did not include sinusitis because there are no data to suggest that vaccination prevents sinusitis.

We conducted the analysis from both the societal perspective (medical and nonmedical costs) and the health care payer's perspective (medical costs primarily borne by health plans). We projected pneumococcal disease outcomes from birth to death for a hypothetical US birth cohort of 3.8 million infants. The model used a microsimulation (semi-Markov) approach in which disease incidence was calculated on a monthly basis, and was programmed using statistical software (Data [version 3.5], TreeAge Software Inc, Williamstown, Mass).

The probabilities of events in the decision model (Table 1) were derived from published studies, unpublished data from local and national sources, and expert panel consensus. We convened 9 experts for a 1-day meeting and used a modified Delphi process to derive estimates for selected questions on which the data were sparse or lacking. When estimates or assumptions were equivocal, we chose the one that would bias the results against vaccination.

Age-specific incidence rates of pneumococcal meningitis and bacteremia (Table 1) were from recent data from the Centers for Disease Control and Prevention's Active Bacterial Core Surveillance/Emerging Infections Program Network (Chris Van Beneden, MD, MPH, written communication, September 1999).^{1 ,9} Rates of pneumonia and OM were from analyses of the NCKP population and from other published sources.^{8 ,10} Probabilities of death, deafness, and disability after invasive pneumococcal disease were based on Active Bacterial Core Surveillance/Emerging Infections Program data and published sources.^{1 ,11 - 12}

Estimates of vaccine efficacy were based on the findings of the NCKP trial. Because bacteriologic cultures are not routinely ordered for pneumonia and OM, we used estimates contributed by the expert panel to infer efficacy against pneumococcal vaccine serotypes for these conditions. When making inferences, we always maintained consistency with the trial's intention-to-treat efficacy results for each disease. For example, the trial's intention-to-treat analysis observed 73% efficacy against pneumonia defined as pneumococcal based on chest radiograph findings that met World Health Organization guidelines for this definition. Extrapolating from studies of invasive disease, we estimated that 80% of pneumococcal pneumonia was due to vaccine serotypes. Thus, we inferred that the vaccine was 90% efficacious against pneumococcal pneumonia due to vaccine serotypes.

The expert panel recommended making the conservative assumption that a vaccinated infant would experience reductions in pneumococcal diseases only until his/her fifth birthday. Based on expert panel consensus, in vaccinated infants, vaccine efficacy against invasive disease was assumed to decrease from 100% during ages 0 to 23 months to 93% during 2 to 4.9 years. Vaccine efficacy against pneumonia and simple or complex OM was similarly assumed to wane during ages 2 to 4.9 years.

Medical Care for Pneumococcal-Associated Diseases. The mean medical costs of pneumococcal diseases (Table 2) were derived by analyzing the costs of disease episodes using NCKP's Cost Management Information System. This accounting system uses a step-down method of allocating all fixed and variable costs, including administrative overhead, to units of service (eg, a 10-minute visit to a pediatrician). The cost analysis included all hospital, emergency department, outpatient, and prescription medication use.

We determined the costs of pneumococcal meningitis and bacteremia (including bacteremic pneumonia) by studying invasive pneumococcal disease cases identified by NCKP's active surveillance program.⁵ To determine the cost of pneumonia, we identified an initial set of hospitalized and outpatient cases using International Classification of Diseases, Ninth Revision (ICD-9) codes for pneumonia (486, 481, and 482.9). From this initial set, we followed recent World Health Organization guidelines and defined pneumococcal pneumonia cases as those with chest radiographs read by the original radiologist as having consolidation, air bronchograms, or pleural effusion without other known causes. Because this analysis focuses on routine vaccination of low-risk infants, we used the conservative approach of excluding children with immunodeficiency, cancer, and congenital cardiac or respiratory diseases when making cost estimates.

To determine the cost of OM, we conducted an analysis of approximately 120,000 children who were members of NCKP during 1997; details are reported elsewhere.¹⁰ We defined an OM visit as one with an appropriate ICD-9 code (382.XX and related codes) and an OM-related antibiotic prescription filled. Consistent with a previous study,⁷ we defined a new OM visit as one that was not preceded by another OM visit during the prior 21 days. An OM episode was defined as the period beginning with a new visit and ending with the last OM-related physician visit or medication before the next new OM visit. The costs of follow-up visits (ie, those coded as OM but without antibiotics), audiometry, otorhinolaryngologist visits, antibiotics, surgical procedures (eg, tympanostomy tube placement), and hospitalizations related to OM were counted as associated with the episode during which they occurred. We divided OM episodes into those that included tympanostomy tube placement and those that did not; the latter group were further classified as simple (1 or 2 OM visits during the episode) or complex (≥3 OM visits during the episode).

The costs of long-term medical care and special education for neurologically disabled survivors of meningitis were assumed to be similar to the costs for cerebral palsy.¹³ We assumed that patients with deafness after meningitis would be candidates for a cochlear implant, whose cost was estimated from published sources.¹⁴

Vaccination Program Costs. The manufacturer's list price for the recently licensed version of this vaccine is $58 per dose. We generated estimates of cost-effectiveness for vaccine costs ranging from $1 to $100 per dose, consistent with the price ranges of recently licensed vaccines. The cost of vaccine administration at a routine infant visit was estimated at $10 based on a previous study and on consultation with NCKP administrators.¹⁵ In this analysis, we prorated the cost of vaccine administration to $5 per dose because at least 1 other vaccination would be given at each visit during which pneumococcal vaccine was recommended.

Work-Loss and Other Nonmedical Costs. Lost time from work, valued at the parent's wage rate, was used as a proxy for the value of the time spent tending a child with pneumococcal disease. Wage rates and lost work time were derived from interviews with parents whose children had experienced bacteremia (n = 17), pneumonia (n = 307), and OM (n = 300). There were no cases of meningitis available during the survey period. We assumed that meningitis work-loss costs would equal bacteremia work-loss costs multiplied by the ratio of meningitis medical costs to bacteremia medical costs. Lost productivity for individuals who died or had long-term disability due to meningitis was assigned a monetary value based on the present value of expected future lifetime earnings forgone.^{13 ,16}

The base case analysis was conducted from the societal perspective for a hypothetical birth cohort of 3.8 million US infants. The base case analysis made projections for a fully implemented vaccination program at steady state (ie, it assumed that 100% of infants would receive 4 doses of vaccine). Consistent with the recommendations of a national panel, costs and benefits were discounted at a rate of 3% per year.¹⁷ Cost-effectiveness ratios were calculated as dollars invested in the vaccination program minus dollars saved due to disease episodes averted divided by health benefits, in which the health benefits were life-years saved or episodes of meningitis, bacteremia, pneumonia, or OM averted. Based on the national panel's recommendations, to avoid double counting benefits, productivity losses due to death were not included in calculating dollars per life-year saved. However, when calculating dollars per invasive disease case prevented, productivity losses due to disability were included in the numerator. Likewise, productivity losses were included in calculations of break-even points.

We evaluated how the model's results changed as we varied key assumptions over plausible ranges. Sensitivity analyses included: (1) the incidence of invasive disease was varied from 50% to 200% of baseline; (2) the proportion of invasive disease that was meningitis was varied up to 14%^{18 - 19} ; (3) rates of OM-related use were increased (outpatient visit rates were increased to 125% and tympanostomy tube placement rates were increased to 200% of baseline) based on estimates and inferences from alternative data sources^{20 - 21} ; (4) vaccine efficacy was varied as shown in Table 1, based on the 95% confidence intervals from the NCKP randomized trial; (5) the cost of vaccine administration was increased to $13^{15 ,22} ; (6) the costs of clinic visits and hospitalizations were changed to those from national sources, mainly the Medicare fee schedule^{23 - 25} ; (7) the discount rate was changed to 5% for both costs and benefits; (8) the coverage rate was decreased to 90%²⁶ ; and (9) potential costs of medical use and work loss due to vaccine adverse reactions were incorporated. The base case did not include costs for vaccine adverse reactions because 1 study found that fever and irritability among pneumococcal conjugate vaccine recipients were no higher than among controls, although it must be recognized that controls received a different conjugate vaccine.²⁷ To accommodate this in the sensitivity analysis, we estimated the cost of adverse reactions after pneumococcal vaccination at $5 per dose, consistent with our findings on other vaccines (T.A.L., unpublished data, 1998).

Best-case (most favorable to vaccine using assumptions 1, 2, 3, and 4 above) and worst-case (least favorable to vaccine using assumptions 1, 4, and 5 above) scenarios also were modeled. Unless otherwise noted, sensitivity analysis results were from the societal perspective at a vaccine cost, based on the manufacturer's anticipated list price, of $58 per dose.

When any new infant immunization program is introduced, policymakers must decide whether those persons older than the recommended age should be vaccinated in a supplementary catch-up program.²⁸ We modeled the cost-effectiveness of catch-up pneumococcal vaccination of children aged 2 to 4.9 years. Immunogenicity data suggest that this may require only 1 dose of vaccine for this age group.²⁹ For children, we used vaccine efficacy estimates that the expert panel derived by extrapolation from the infant efficacy data. We assumed that catch-up vaccination would not generate additional clinic visits, but would be offered during routine preventive or urgent clinic visits that already occurred in practice. We made the conservative assumption that a child who received catch-up vaccination would experience reduced pneumococcal disease for 3 years. Based on a recent study of risk factors for invasive pneumococcal disease, children in day care were assumed to have a 2.4-fold elevated risk of all pneumococcal diseases.³⁰ In accord with this study, we defined day care as care of 2 or more unrelated children outside the home for 4 or more hours per week. National statistics suggest that 25% to 30% of all children are in this type of day care.^{31 - 32}

This study was an academic-public health collaboration in which the vaccine manufacturer contributed a grant to support the work of the non-governmental investigators. Representatives of the manufacturer reviewed an interim analysis and a draft of the manuscript, but did not have authority for scientific or editorial decisions.

Table 3 and Table 4 show projected disease outcomes and costs with and without a routine pneumococcal vaccination program for infants. Vaccination would prevent approximately 1 million episodes of OM, 53,000 cases of pneumonia, 12,000 cases of invasive disease, and 116 deaths due to pneumococcal infection for each US birth cohort. Pneumococcal-associated diseases (including OM due to all causes) were estimated to cause $2.5 billion in direct medical costs, and $3 billion in work-loss and productivity costs for each US birth cohort. A fully implemented infant vaccination program was estimated to reduce direct medical costs by $342 million and other costs by $415 million, before accounting for the costs of vaccine doses and administration.

Figure 2 shows how the cost per life-year saved and the cost per OM episode prevented varied depending on vaccine cost. From the societal perspective, pneumococcal vaccination of healthy infants would result in savings if the vaccine cost $46 or less per dose. From the health care payer perspective, the vaccination program would result in net savings if the vaccine cost $18 or less per dose, and would result in $25,000 per life-year saved if the vaccine cost $25 per dose.

Preliminary information suggests the vaccine's nondiscounted list price will be $58. At this price, the costs per life-year saved would be $80,000 (societal) and $176,000 (health care payer). Using alternative measures of health benefit, vaccination at $58 per dose would cost society (or the health care payer) $160 ($550) per OM, $3200 ($11,000) per pneumonia, $15,000 ($50,000) per bacteremia, or $280,000 ($970,000) per meningitis episode prevented. These ratios reflect the net cost of vaccination after accounting for medical costs, work loss, and other costs averted from disease episodes prevented.

Disease Incidence. The results were sensitive to variation in the incidence of invasive disease. As incidence estimates were varied from 50% to 200% of base case assumptions and the vaccine cost was held constant at $58 per dose, the societal cost per life-year saved decreased from $188,000 to $25,000. When the proportion of invasive disease due to meningitis was increased to 14%, the societal break-even vaccine cost increased from $46 to $58 per dose. When health care services use due to OM was increased from base case assumptions (by 125% for outpatient visits and 200% for tympanostomy tube placement), the break-even vaccine costs increased to $58 per dose (societal) and $25 per dose (health care payer).

Other Sensitivity Analyses. As vaccine efficacy assumptions were varied from optimistic to pessimistic over the ranges shown in Table 1, the societal cost per life-year saved increased from $20,000 to $235,000 (as vaccine cost was held constant at $58 per dose) and the break-even cost decreased from $60 to $23. When the cost of vaccine administration was increased to $13, the societal cost per life-year saved increased to $113,000. When alternative costs from national data sources were used, there were negligible changes in the cost per life-year saved ($80,000) and the break-even vaccine cost ($46). At a discount rate of 5%, the societal cost per life-year saved was $144,000. Decreasing the vaccine coverage rate to 90% reduced projected costs and savings by a similar proportion but did not change cost-effectiveness results. When we incorporated a cost for potential vaccine adverse reactions of $5 per dose, the break-even cost decreased by $5.

Best- and Worst-Case Scenarios. From a societal perspective, the best-case scenario resulted in savings and a break-even vaccine cost of $112, while the worst-case scenario resulted in a cost per life-year saved of $593,000 and a break-even vaccine cost of $10.

One dose of catch-up vaccine for all 2- to 4.9-year-old children would result in savings from the societal perspective if vaccine costs were in the $60 range or less. As Figure 3 shows, cost-effectiveness results were sensitive to the relative risk of pneumococcal disease as well as vaccine cost. The break-even vaccine cost also varied depending on children's ages. From the societal perspective, the break-even vaccine cost for children in day care ranged from $85 (for 4- to 4.9-year-olds) to $135 (for 2- to 2.9-year-olds), compared with $32 to $53 for those not in day care. From the health care payer perspective, the break-even vaccine costs for children were $41 to $66 for those in day care and $14 to $25 for those not in day care.

This study found that routine pneumococcal conjugate vaccination of healthy infants is potentially cost-effective, although its projected savings for society are more than double those for health care payers. From the societal perspective, vaccination is projected to reduce pneumococcal disease costs by almost $760 million for each cohort of infants born in the United States each year. However, more than half of the projected savings are from reduced work loss by parents who care for ill children or averted productivity loss due to disability or death caused by pneumococcal disease.

From the health care payer perspective, pneumococcal conjugate vaccination—like varicella, hepatitis B, and other recently recommended vaccines—may not result in net savings but could be relatively cost-effective compared with other health interventions.^{33 - 36} Table 5 illustrates how pneumococcal conjugate vaccination at varying vaccine costs would compare with other health interventions in cost per life-year saved. At its anticipated list price, infant pneumococcal vaccination would have a cost-effectiveness ratio at the high end of the range for current preventive measures. However, the average cost of vaccine will likely be lower because government agencies, which buy approximately 60% of vaccine doses, typically receive discounted prices under negotiated contracts. If our assumptions on the vaccine's efficacy for children hold true, catch-up vaccination of children would be cost-effective compared with infant vaccination, especially when directed toward high-risk groups such as those in day care.

In sensitivity analyses, we found that results changed as key assumptions were varied over plausible ranges. For example, under assumptions unfavorable to vaccination, including low incidence of invasive disease, low vaccine efficacy, or high vaccine administration costs, routine infant vaccination at a cost of $40 per dose would not result in societal cost savings. Conversely, under favorable assumptions about invasive disease incidence, OM use, or vaccine efficacy, the societal break-even vaccine cost increased from $46 to approximately $60. These findings underscore the importance of gathering more robust empirical data on vaccine effectiveness and costs in postlicensure studies.

The current analysis made conservative assumptions that likely biased against a vaccination program. Pneumococcal conjugate vaccination will reduce not only mortality due to invasive disease, but also morbidity due to OM and pneumonia, which account for more than 90% of the disease episodes prevented. We have presented results in terms of cost per life-year saved to enable comparisons with other health interventions. However, this ratio underrepresents the vaccine's value because it gives credit only for reducing mortality, but not morbidity. We have also not attempted to place a value on the psychological costs of pain for the child, and anxiety for the child and parent, due to OM, pneumonia, and invasive pneumococcal diseases. As yet, these psychological costs have not been standardized as utilities (the formal measure of preferences recommended for calculating quality-adjusted life-years) in population-based studies.^{37 - 39} Ideally, however, policymakers evaluating these cost-effectiveness ratios should take into account the benefits of preventing morbidity and suffering due to OM and pneumonia as well as mortality due to invasive pneumococcal disease.

Based on the recommendations of our expert panel, vaccine efficacy was assumed to last only 5 years for infants and 3 years for children. Based on the panel's recommendations, the vaccine was not assumed to have any efficacy against cross-reactive serotypes, which account for at least 5% of pneumococcal infections in young children,⁴⁰ because reliable data on this topic are not yet available.

In the current analysis, we did not evaluate the potential cost-effectiveness of using 23-valent pneumococcal polysaccharide vaccine in 2- to 4.9-year-olds. There is some evidence that this vaccine may be effective at preventing invasive disease in this age group.⁴¹ The polysaccharide vaccine costs $5 to $13 per dose. However, limited evidence exists about its impact on OM,^{42 - 43} which drives the pneumococcal vaccination's cost-effectiveness.

Some potentially important indirect effects of a nationwide pneumococcal vaccination program are unknown. For example, reduced pneumococcal carriage in vaccinated children could lead to reduced transmission to unvaccinated persons. Thus, vaccine effectiveness in a population may be greater than initially observed in randomized trials.⁴⁴ Because Pneumococcus is the leading cause of occult bacteremia, vaccination might also result in a more limited set of clinical circumstances in which blood cultures and empirical antimicrobials are indicated.

In addition, the emergence of drug-resistant S pneumoniae has made treatment of pneumococcal infections more difficult and expensive.³ Pneumococcal conjugate vaccination could reduce the proportion of infections caused by drug-resistant strains and allow clinicians to reduce empiric, broad-spectrum antibiotic prescribing for young children. The end result could be less disease caused by drug-resistant strains and reduced antimicrobial pressure causing selection of new drug-resistant strains. Our analysis might have underestimated the vaccination's health benefits and cost savings because we did not attempt to model these indirect effects.

We did not evaluate the cost-effectiveness of vaccinating of infants and children in developing countries in which pneumococcal disease has higher incidence and impact. For example, the mortality rate due to bacteremic pneumococcal pneumonia is less than 5% in US children but is likely much higher in developing countries.⁴⁵ Although the theoretical model in our analysis is generalizable, alternative assumptions on disease epidemiology and costs would be needed for these settings.

Some experts have raised concerns that although vaccination may lead to a reduction in vaccine serotypes, other disease-causing serotypes may replace them in the nasopharynx of vaccinated children.^{46 - 48} To date, no data indicate increasing rates of disease with nonvaccine serotypes in vaccinated children. Postimplementation research should evaluate the effects of vaccination on disease incidence in unvaccinated persons, use of antimicrobial drugs, and carriage and disease due to nonvaccine serotypes.

We conclude that routine pneumococcal conjugate vaccination of healthy US infants has the potential to be cost-effective relative to other preventive health interventions. Decisions about implementation should rest not only on cost considerations but also on qualitative judgments about the value of preventing mortality due to invasive disease and morbidity due to invasive disease, OM, and pneumonia.