The Cost-effectiveness of Adolescent Hepatitis A Vaccination in States With the Highest Disease Rates FREE

IN THE United States, an estimated 160,000 persons were infected with hepatitis A virus (HAV) in 1997, and 80,000 had symptomatic hepatitis A.¹ There is marked regional variation in hepatitis A incidence, with residents of western states facing more than twice the disease risk of other Americans.² Despite the relatively large number of persons infected with HAV each year, more than 80% of Americans younger than 20 years remain susceptible to infection.³ Preexposure immunization with hepatitis A vaccine affords high levels of protection^{4 - 5} and the potential to substantially reduce disease incidence. Kinetic models of antibody decline suggest protection may persist for 20 years or more.^{6 - 7}

The annual economic consequences of hepatitis A were estimated at $200 million in 1987⁸ and $489 million in 1997.⁹ Hepatitis A costs are disproportionately composed of lost work productivity rather than medical treatment.^{8 - 9} The cost-effectiveness of hepatitis A vaccination is highly dependent on the population attributable risk of infection. Although routine childhood vaccination is estimated to reduce costs in communities with annual disease incidence greater than 150 cases per 100,000 population,¹⁰ vaccination of medical school graduates would cost $58,000 per year of life saved (YOLS).¹¹ Studies considering travelers to countries with high rates of HAV infection suggest vaccination is more cost-effective than preexposure administration of immune globulin,¹² especially when travel frequency exceeds once every other year.¹³

Recently, routine hepatitis A vaccination was recommended for children residing in areas where disease incidence has been higher than the national average.¹ Although preference was not given to any specific age group, it is likely that vaccination activities will focus on preschool children, meaning more than a decade may pass before a generation of adolescents is largely protected from infection. Adolescents are an appropriate group for hepatitis A vaccination, since they are entering a 15-year period with the highest risk of symptomatic HAV infection,² and hepatitis A vaccine could be incorporated into current adolescent immunization schedules. We therefore investigated whether hepatitis A vaccination of adolescents would meet conventional standards of cost-effectiveness.

A decision model¹⁴ was used to examine risks and outcomes of HAV infection with and without routine vaccination of adolescents by modeling the experiences of a single birth cohort. The model considered 15-year-old residents of the 10 states with the highest rates of reported hepatitis A between 1990 and 1997: Arizona, California, Idaho, Missouri, Nevada, New Mexico, Oklahoma, Oregon, Utah, and Washington. A decision node separated vaccination and no vaccination branches, and a chance node separated the vaccination group into those who are fully vs partially vaccinated (Figure 1). The strategy of screening for hepatitis A immunity before vaccination was considered, but based on the relatively low prevalence of antibody to HAV among US adolescents³ and the relatively high cost of screening vs vaccination, this strategy was clearly dominated by the "vaccinate all" approach. The structure of the decision tree is identical for those not vaccinated, partially vaccinated, or fully vaccinated, but vaccinated groups are assigned vaccination costs and reduced risks of infection.

The risk of HAV infection is examined in annual intervals, and infected persons are assigned age-specific risks of symptomatic infection, hospitalization, transplantation due to acute liver failure, and hepatitis A mortality. Persons who are not infected by HAV and do not die of an unrelated cause are monitored for HAV infection risk until their actuarial death from any cause. Since fewer vaccinated persons are infected each year, more are assessed for HAV infection in the later years of the model.

Published information concerning HAV infection outcomes is limited and often conflicting. Several data sources were used to obtain the most reliable information for model parameters: (1) databases monitoring disease incidence, (2) a case series study, and (3) sequential surveys of an expert panel. When the case series and expert panel provided divergent estimates of the same parameter (eg, hospitalization rate, work loss duration), we selected the estimate least favorable to vaccination.

The age-specific incidence of reported hepatitis A for 1990-1997 was obtained from the Centers for Disease Control and Prevention's National Notifiable Diseases Surveillance System. Data from this 8-year period were used to minimize the effect of local variation due to community-wide outbreaks on incidence rates. To adjust for underreporting, we assumed a 3:1 ratio of the actual number of symptomatic cases to reported cases in each age group.^{1 - 2 ,8 ,15} The number of asymptomatic infections was derived from expert panel estimates of the age-specific ratios of asymptomatic to symptomatic infections, which ranged from 0.37:1 for persons aged 15 to 29 years to 0.11:1 for those aged 70 years or older. Annual HAV infection rates were calculated using Bureau of Census population data and were assumed to remain constant throughout the analytic period.

Adolescent and adult patients with serologically confirmed hepatitis A treated since January 1995 were identified by a sample of general and family practitioners, internists, gastroenterologists, and infectious disease specialists practicing in Arizona, California, Nevada, and Oklahoma.⁹ A total of 42 physicians provided data on 254 patients with hepatitis A, representing all eligible patients from their practices. The patient sample was predominantly male (66%) and white (66%), and their mean (SD) age was 33 (12) years. Data collected included the following: duration of symptoms and work loss; hospital length of stay; number and type of medical interventions, including physician visits, imaging studies, and serologic, biochemical, coagulation, hematologic, and microbiologic tests; dose and duration of prescription and over-the-counter drugs; and administration of immune globulin to personal contacts.

We convened a panel of 6 hepatitis experts and conducted 3 sequential surveys using the Delphi method.¹⁶ The first survey was accompanied by literature summaries and requested the following: age-specific estimates of the probability of symptomatic infection,^{17 - 19} hospitalization,^{2 ,8 ,11 - 12} acute liver failure,^{20 - 22} liver transplantation,²³ and mortality^{2 ,8 ,24 - 25} ; estimates of long-term vaccine efficacy following 1 or 2 doses^{4 - 7} ; and profiles describing the number of physician visits, hospital days, imaging studies, laboratory tests, and drugs that would be ordered for a typical patient treated as an outpatient or inpatient. The second survey included a summary of first-round estimates and required panelists to reconsider their initial responses in light of their peers' estimates. Responses were used to calculate means and 95% confidence intervals (CIs) for each item. Consensus was considered to have been achieved when the 95% CI was ±15% of the mean. The third survey was limited to items with larger CIs.

Vaccination costs were estimated from the perspective that 80% of adolescents would receive at least 1 vaccine dose. Further, we assumed 80% of vaccinated adolescents would receive the second scheduled dose based on the experience with adolescent hepatitis B vaccination.²⁶ In the United States, the Vaccines for Children (VFC) program covers children from birth through 18 years of age. Based on VFC data, 70% of adolescents would be eligible for hepatitis A vaccine at $11.08 per dose under the current federal contract, with state-specific reimbursements for vaccine administration ranging from $3.00 to $15.19 per dose (Centers for Disease Control and Prevention, unpublished data, February 1999). For adolescents not covered by VFC, we estimated hepatitis A vaccine costs at $23.08 per dose, with state-specific maximum allowable administration fees, established by VFC, of $13.70 to $17.56. We did not include the cost of additional physician visits because, by VFC policy, providers are not compensated beyond vaccine acquisition and administration.

Hepatitis A treatment costs were estimated by applying Medicare reimbursement rates to services documented through the case series. Payments for hospital stays were based on mean 1997 reimbursements for diagnosis related groups 205 and 206 (diseases of the liver, except malignancy, cirrhosis, and alcoholic hepatitis).²⁷ Payments for physician visits and outpatient procedures were derived by applying Medicare national fees²⁸ to the CPT level utilization data obtained from the case series. Drug costs were estimated by reducing average wholesale prices²⁹ by 20% and assuming the lowest cost generic was used when available.

Days missed from paid employment, obtained from the case series, were valued based on age- and sex-specific estimates of workforce participation and earnings.³⁰ We assumed 55% of infections occur in males² and valued each lost work day at one-two hundred and fiftieth of annual earnings. The value of lost housekeeping services was estimated by applying age-, sex-, and employment status–specific values^{30 - 31} and assuming a 50% reduction in housework while hepatitis A symptoms are present. Earnings and housekeeper production values were adjusted to 1997 levels using the Employment Cost Index for the civilian workforce.

Estimates of HAV infection outcomes are provided in Table 1. Persons aged 20 to 29 years are at greatest risk of HAV infection, with the risk declining by at least 25% in subsequent decades. Other parameters reflect a strong relationship between age at infection and disease outcome. The probability of receiving a liver transplant for hepatitis A–associated acute liver failure was calculated as the product of the risk of acute liver failure and chance of obtaining a donor organ, with only patients younger than 60 years considered transplant candidates. The durations of work loss and disease symptoms were not stratified by decade of age because of limited sample size.

Hepatitis A treatment and work loss costs are summarized in Table 2. The case series yielded lower estimates of both outpatient and inpatient costs than the expert panel, and these were used in the model. The lifetime cost of liver transplantation was estimated at $302,900 for the first year and $21,900 in subsequent years.³² Per-patient costs were estimated assuming a median survival of 10 years,³³ with follow-up costs discounted to the year of transplantation. Indirect costs are largely composed of lost wages and benefits, and are highest for persons between 20 and 64 years of age.

Long-term efficacy following a complete (2-dose) vaccine series has been estimated at 93% to 95% from models of antibody decay for persons vaccinated as young adults.^{6 - 7} The expert panel estimated protective efficacy following full vaccination at 95%, 90%, 81%, 74%, and 68% through 10, 20, 30, 50, and 70 years, respectively. Corresponding estimates for those partially vaccinated are more problematic, since models of antibody decline are unavailable. Nevertheless, expert panel estimates for adolescents receiving a single vaccine dose were 62%, 42%, 32%, 21%, and 13%.

The health system costs of hepatitis A were defined as the sum of vaccine acquisition costs, vaccine administration costs, and hepatitis A treatment costs. Societal costs of hepatitis A include these costs and the value of work loss resulting from hepatitis A morbidity. Work loss costs due to hepatitis A mortality are not considered. The base year of the analysis is 1997, and all costs were converted to 1997 values using a 3% annual discount rate.³⁴ When health system or societal costs were positive (cost reductions were less than vaccination costs), the value of vaccination was expressed as costs per YOLS. Consistent with prior vaccine cost-effectiveness studies,^{26 ,35 - 36} results are presented with YOLS discounted to the base year and with YOLS undiscounted.

Results are presented for the 10 individual states and for the pooled 10-state region. Several sensitivity analyses were conducted on the pooled population: (1) the discount rate was varied between 0% and 5%; (2) vaccine protective efficacy was varied between 95% CIs around expert panel estimates; (3) lower 95% CIs were used to estimate protective efficacy through 20 years, but protective efficacy was assumed to be 0% thereafter; (4) case fatality rates from the Viral Hepatitis Surveillance Program^{2 ,37} were substituted for expert panel estimates; (5) estimates of hospitalization and liver transplantation were varied between 95% CIs around expert panel estimates; (6) estimates of work loss and symptom duration were varied between 95% CIs around case series estimates; and (7) infection rates for the total United States were substituted for the 10 states with the highest infection rates.

During 1990-1997, 106,536 cases of hepatitis A were reported to the National Notifiable Diseases Surveillance System from the 10 states with the highest average rate of disease. Of these, 74,773 cases were among persons 15 years and older, and most of these could have been prevented by an adolescent vaccination program. Each year, 854,300,15-year-old residents would be eligible for adolescent vaccination, of whom 683,440 are predicted to receive at least 1 vaccine dose (Table 3). Vaccination would prevent 18,130 symptomatic infections, representing an 80% reduction in the lifetime risk of infection among those vaccinated, from 3.3% to 0.7%. Under base case assumptions, the numbers of hepatitis A–related hospitalizations, liver transplantations, and deaths would decline by 75%, 71%, and 68%, respectively. Vaccination would save 2117 years of life or 1219 years of life once discounted to present values.

Routine vaccination of adolescents in these states would cost $30.9 million annually or $45.27 for each 15-year-old resident projected to receive at least 1 vaccine dose. In return, hepatitis A treatment costs (discounted) are projected to decline by $14.2 million, while work loss costs (discounted) are projected to decline by $23.8 million. Most of the decline in treatment costs is expected from fewer hospital admissions. The decline in work loss costs would primarily accrue from fewer days missed from paid employment vs unpaid housekeeping. From a societal perspective, vaccination is dominant, providing improved health outcomes and net economic savings of $7.1 million. Health system costs would total $16.7 million, with cost-effectiveness ratios of $7902 per YOLS or $13,722 per discounted YOLS.

State-specific cost-effectiveness ratios (Table 4) are strongly influenced by HAV infection rates and modestly affected by vaccine administration reimbursement rates. In 9 of the 10 states, vaccination is a dominant strategy from the societal perspective, providing a net savings and improved health outcomes. In Utah, which had the highest 1990-1997 HAV infection incidence among older individuals, economic benefits of vaccination are nearly twice as great as vaccination costs. In California, reduced treatment and work loss costs offset 96% of vaccination costs. From the health system perspective, costs per YOLS range from $3090 to $11,431, while costs per discounted YOLS range from $5495 to $20,569.

Base case assumptions were varied in a series of sensitivity analyses, and those with the greatest influence on cost-effectiveness ratios are presented in Table 5. Results are most sensitive to the discount rate, assumptions regarding long-term vaccine protective efficacy, and HAV infection incidence. From the societal perspective, vaccination remained dominant in 12 of 14 sensitivity analyses conducted on the 10-state region. Exceptions were when the discount rate was increased to 5% and when vaccination is assumed to offer no protective efficacy after 20 years. Under the latter assumption, more HAV infections would occur after the age of 35 years with vaccination, because protection conveyed between the ages of 15 and 34 years would cause more persons to be susceptible thereafter. From the health system perspective, no sensitivity analysis increased cost-effectiveness ratios for the 10-state region above $20,000 per YOLS or $25,000 per discounted YOLS. Because the Viral Hepatitis Surveillance Program has consistently observed higher case fatality rates from hepatitis A than the expert panel, substituting these higher rates makes vaccination more attractive. From the health system perspective, cost-effectiveness ratios decline from $7902 to $4773 per YOLS and from $13,722 to $7592 per discounted YOLS. Other sensitivity analyses, including varying estimates of work loss and symptom duration, vaccine efficacy, hospitalization, acute liver disease, and transplantation between their 95% CIs affected cost-effectiveness ratios by less than 10%. Vaccinating adolescents against hepatitis A nationally would be considerably less cost-effective than focusing on states with high disease burdens. From the health system perspective, a national program would cost $27,700 per YOLS or $54,191 per discounted YOLS.

Childhood vaccinations are among the most cost-beneficial medical interventions. From the societal perspective, immunizations to prevent pertussis,^{38 - 39} measles, mumps, and rubella,^{40 - 41} polio,⁴¹ varicella,^{36 ,42} and hepatitis B^{26 ,43} are all cost saving. The present analysis indicates that adolescent hepatitis A vaccination in states with the highest disease rates would reduce societal costs as well. In the 10 states analyzed, projected savings in treatment and work loss costs would be 23% greater than the costs of vaccination, and this strategy would remain dominant unless the price per dose were increased by $5.73, more than 50% above the current federal contract price. Technologies costing the health system $25,000 to $50,000 per YOLS are typically described as cost-effective,^{44 - 45} and the $50,000 per YOLS standard is being increasingly applied to economic evaluations of vaccines.^{43 ,46} With outcomes discounted to present values, varicella vaccine costs the health system $16,000 per YOLS.³⁶ Similar estimates for hepatitis B vaccine range from $20,619²⁶ to $57,197 per YOLS⁴⁷ for infants and from $26,000⁴³ to $97,296 per YOLS³⁵ for adolescents. In states with high disease rates, we estimate cost-effectiveness ratios for adolescent hepatitis A vaccination ranging from $5495 (Utah) to $20,569 per YOLS (California), with a population-weighted mean of $13,722 per YOLS for the 10-state region. When disease incidence rates for the United States as a whole were applied to our model, costs exceeded $50,000 per YOLS. However, a 7% reduction in vaccination costs would make a national initiative cost-effective by this standard.

This analysis was conducted from the perspectives of the health system and society,⁴⁸ using a discount rate consistent with economic growth,³⁴ and considered several populations at varying risks of HAV infection. Although hepatitis A vaccine has been shown to be cost-effective in groups at high risk of infection, these studies are not easily applied to routine vaccination because disease incidence^{10 ,12 ,49 - 50} and work loss costs¹¹ differ significantly. Furthermore, studies not reporting costs per YOLS or quality-adjusted life-year¹² make comparison with other technologies difficult. In the present study, risks of symptomatic illness, hospitalization, acute liver failure, and death were estimated on age-specific bases, whereas many prior studies assumed identical rates for persons of all ages.^{10 - 12} Since adverse outcomes are less common in younger patients, who are most likely to be infected, age-specific estimates provide more accurate assessment of outcomes.

Our cost-effectiveness estimates should be considered conservative for several reasons. When multiple parameter estimates were available, we consistently chose those least favorable to vaccination. Hepatitis A treatment costs were based on Medicare reimbursements, which typically pay less than other third-party payers. We did not consider the effect of vaccination on reducing HAV transmission to unvaccinated individuals. Incorporating a 1:1 secondary attack rate into the model would approximately double the benefits of vaccination. In addition, although immune globulin administered to personal contacts represented $44 of the estimated cost of each infection, we did not consider other costs associated with hepatitis A outbreaks. For instance, one outbreak caused $689,413 in disease control costs borne by health departments, health maintenance organizations, and hospitals.⁵¹ Finally, although mortality is an uncommon hepatitis A outcome, extended and substantial morbidity is the norm. We did not quantify the benefits of reduced morbidity, because utility measures are unavailable for patients with hepatitis A, but an assumption of 50% reduced utility for 1 month per symptomatic infection would increase savings from 2117 YOLS to 2872 quality-adjusted life-years.

The present analysis is limited by several factors common to modeling exercises. First, vaccination will be less valuable if an adolescent moves from an area of high incidence to an area of low incidence. Second, we only examined the cost-effectiveness of statewide vaccination initiatives. Within individual states, there is often marked variation in HAV incidence over time and between geographic areas. Targeting vaccination to areas undergoing outbreaks or to those with the highest disease incidences for a long duration would improve cost-effectiveness ratios. These strategies, however, leave most adolescents unprotected and have little effect on national or regional disease rates.¹ Third, expert panel estimates of protective efficacy are speculative, and sensitivity analyses showed that estimates of long-term protective efficacy have an important impact on cost-effectiveness ratios. The longest follow-up trials of protective efficacy are still less than 10 years after the primary schedule. Although a mathematical model estimated the duration of protection at 25 years,⁷ this study must be validated. It is unclear whether protection will be extended further by immunologic memory.⁵²

Case estimates of the long-term efficacy of hepatitis A vaccine may be difficult to obtain if widespread vaccination is undertaken and rapidly lowers HAV transmission. In contrast to studies of cohorts vaccinated against hepatitis B, an infection persistent in the individual and the population, the long-term efficacy of hepatitis A vaccination will most likely be determined by antibody persistence and/or booster studies in immunized persons who have lost antibody. However, even when pessimistic estimates of protective efficacy are incorporated in our model, results support the cost-effectiveness of vaccinating adolescents in states with high disease burdens.

Accepted for publication February 14, 2000.

This study was partially funded by an unrestricted research grant from SmithKline Beecham Pharmaceuticals, Philadelphia, Pa.

We are grateful to the members of the expert panel on hepatitis A outcomes: Jules Dienstag, MD, Massachusetts General Hospital, Boston; Ira Goldman, MD, Cornell University Medical College, New York, NY; Emmet Keeffe, MD, Stanford University Medical Center, Palo Alto, Calif; David Sack, MD, Johns Hopkins University, Baltimore, Md; Maria Sjögren, MD, Walter Reed Medical Center, Washington, DC; and Robert Weinstein, MD, Cook County Hospital, Chicago, Ill.

Reprints: R. Jake Jacobs, MPA, Capitol Outcomes Research Inc, 6188 Old Franconia Rd, Alexandria, VA 22310 (e-mail: Jake.Jacobs@capitoloutcomesresearch.com).