More than half of patients who are acutely infected with HBV are asymptomatic (Supplementary Fig. 1A). For those who develop symptoms, some recover without hospitalization, whereas others are hospitalized. Some hospitalized patients develop fulminant hepatic failure, which can lead to death, liver transplant, or recovery. Most patients who receive a liver transplant survive, but some die.

Acute infection is a necessary precursor to chronic infection. Approximately 6% of people who experience acute hepatitis B develop chronic hepatitis B (Supplementary Fig. 1B). People with uncomplicated chronic hepatitis can become inactive carriers, develop cirrhosis or hepatocellular carcinoma (HCC), or remain with uncomplicated chronic hepatitis. Inactive carriers can return to the chronic hepatitis state, and they can transmit HBV to others. People with cirrhosis can remain in that state, develop compensated cirrhosis or HCC, or die of hepatitis-related causes. People with decompensated cirrhosis can remain in that state, develop HCC, receive a liver transplant, or die of hepatitis-related causes. People with HCC can remain in that state, receive a liver transplant, or die of hepatitis-related causes. People in the liver transplant state can survive or die of hepatitis-related causes. At any time, people can die of other causes. Other-cause mortality rates are age specific and represent all-cause mortality for people with diabetes.

The HBV infection rates for people with diagnosed diabetes who are susceptible to infection depend on the incidence of HBV infection among all people with diagnosed diabetes, the prevalence of previous HBV infection, existing vaccine coverage, and the efficacy of prior vaccination. The Supplementary Data describes components of the calculation. Incidence rates for susceptible people with diagnosed diabetes increase from 91 per 100,000 people with diabetes 20–24 years of age to a peak of 101 per 100,000 people 35–39 years of age, and then decline to 13 per 100,000 people 70 years of age or older (Supplementary Table 2). Other disease progression probabilities and epidemiologic parameters are listed in Table 1 (10–14).

Efficacy of vaccination is based on a CDC review (S.S., T.M., B. Baack, unpublished data) (the Supplementary Data lists source studies) of vaccine studies in adults (19 studies) and people with diagnosed diabetes (10 studies). Efficacy declines with age, from 95% for 15–29 years of age to 40% for 70 years of age or older (Supplementary Data).

Table 1 contains all cost values and utility values for health states. Direct program costs and direct medical costs are included. All final cost figures were converted to 2010 U.S. dollars using the U.S. Bureau of Labor Statistics’ Consumer Price Index (All Urban Consumers, Medical Care, U.S. City Average). The Supplementary Data describes the calculation of other-cause mortality rates for people with diabetes.

The primary vaccination strategy considered is vaccination with adult hepatitis B vaccine (recombinant). Analyses of routine vaccination of adults 20–59 years of age with diabetes are based on assumptions that vaccination will be offered to all adults 20–59 years of age with diagnosed diabetes; 10% of adults with diagnosed diabetes who are susceptible to hepatitis B and who report no previous vaccination will accept the vaccination (the assumption is based on hepatitis B vaccine uptake for adults with previously recognized risk factors, uptake for other adult vaccines, and manufacturer projections) (Supplementary Data); all people who accept the vaccine will complete a primary series of three doses (1 mL each) given on a 0-, 1-, and 6-month schedule or other approved schedule; immunity will not wane over time for people achieving seroprotection; and vaccination will occur during regularly scheduled patient visits (76% of primary care offices stock hepatitis B vaccine for adults) (15).

The age-group 20–59 years was chosen based on evidence of declining vaccination efficacy in older age-groups and preliminary analyses indicating that vaccination had very high cost-effectiveness ratios among older age-groups. For completeness, the vaccination of adults 60 years of age or older was analyzed as an alternative intervention strategy. The intervention may be viewed as a catch-up strategy because vaccination will be offered to all people who currently have diagnosed diabetes. For purposes of the model, it is assumed that people who refuse vaccination in the first year will not accept vaccination in subsequent years; thus, all vaccination associated with the catch-up program will occur during the first year.

The vaccination of adults with diagnosed diabetes was compared with the status quo of no additional vaccination. Under the status quo strategy, some adults with diabetes have previously been vaccinated but no additional vaccination is assumed to occur.

For each vaccination strategy, incremental cost-effectiveness ratios were calculated as net costs per quality-adjusted life-year (QALY) gained. Net costs of vaccination were estimated as vaccine costs and administration costs minus averted medical costs. The net costs were then divided by the estimated improvement in QALYs resulting from vaccination.

The perspective used is the health care system perspective. The costs included the direct costs to the health care system of providing hepatitis B vaccination and the direct medical costs of hepatitis B–related illness and complications that are averted by preventing HBV infection. The summary measure of effectiveness is QALYs saved by the program. Productivity losses associated with hepatitis-related morbidity or mortality are not included in the analysis because they are implicitly included in the QALY measures (16).

QALYs saved by preventing acute and chronic hepatitis B are estimated for the remaining life expectancy of the target population. The analytical horizon was selected because available data suggest that vaccine-induced immunity against hepatitis B does not wane over time and because chronic hepatitis B can lead to serious health consequences years after an acute infection. All future costs and benefits were discounted at a 3% annual rate.

In one-way sensitivity analyses, model input parameters were individually varied from their low to high values while keeping all other variables fixed. Parameter ranges are based on 95% CIs. When 95% CIs were not available, parameters were varied between the highest and lowest values identified in the literature or by 25% if only one value was available.

In probabilistic sensitivity analyses, parameters were varied simultaneously, and the model was run 1,000 times. For each run, key input parameters were drawn from appropriate distributions. Probability and utility decrements were drawn from beta distributions to ensure values between 0 and 1. Other disease parameters and disease state costs were drawn from lognormal distributions to account for skewness. Costs for small, distinct medical events were varied based on normal distributions. For groupings of clearly associated variables (vaccine efficacy, hepatitis incidence, and utility decrements), correlation between variables was allowed when parameters were drawn for individual runs. We estimated 95% credible intervals of the cost-effectiveness ratio by bootstrapping the results of the probabilistic sensitivity analysis (17).

The private cost of vaccine (which is known) and the cost of administering the vaccine were separately varied by a multiplicative factor between 0.75 and 1.25 for one-way sensitivity analyses and with a corresponding normal distribution in probabilistic sensitivity analyses. The discount factor was varied from 0 to 5% in the one-way sensitivity analysis and was not varied in the probabilistic sensitivity analysis.

In addition to the one-way and probabilistic sensitivity analyses, we considered several alternative scenarios. Separately, the age at vaccination was varied in 10-year increments, the vaccine uptake rate was varied from 5 to 40%, and the vaccine was available at the CDC price.

Projected reductions in acute and chronic health outcomes resulting from vaccination are shown in Table 2. Vaccinating susceptible people with diagnosed diabetes 20–59 years of age with an assumed 10% uptake rate (main analysis) results in vaccination of 528,047 people. This program will prevent 4,271 acute HBV infections over the course of the vaccinated individuals’ lifetimes, 467 hospitalizations for acute infections, 19 fulminant cases, 8 transplants, and 14 deaths from fulminant hepatitis. The program will also prevent 256 cases of chronic HBV infection, thereby preventing 146 cirrhosis cases, 56 decompensated cirrhosis cases, 33 HCC cases, 5 transplants, and 116 deaths from chronic hepatitis.

Vaccinating people with diagnosed diabetes 60 years of age or older and people with diagnosed diabetes 20 years of age or older (the alternative strategies analyzed) would prevent 723 and 4,994 acute HBV infections, respectively. Although a large number of people 60 years of age or older would be vaccinated, few acute and chronic cases would be prevented.

The vaccination program increases net costs and increases QALYs (Table 3). The point estimate of the incremental cost-effectiveness ratio of hepatitis B vaccination for people 20–59 years of age with diagnosed diabetes is $75,094 per QALY. The cost-effectiveness ratio for the alternative strategy of vaccinating adults with diagnosed diabetes 20 years of age or older is $196,557 per QALY; vaccinating only adults with diagnosed diabetes 60 years of age or older produces a cost-effectiveness ratio of $2,760,753 per QALY.

The cost-effectiveness ratios increase with age at vaccination. The factors underlying this increase are discussed below.

Figure 1 shows a tornado diagram depicting the cost-effectiveness results of the one-way sensitivity analyses. The center line of the tornado is the baseline cost-effectiveness ratio of $75,094 per QALY gained. For presentation purposes for certain related groups of parameters (complication costs, chronic hepatitis transition probabilities, acute infection transition probabilities, and complication utilities), we set all parameters in the group to their low and high values. The effects of varying individual parameters within these groups were smaller than the aggregated effects for the group. The parameter labels in Fig. 1 indicate which values were associated with improved cost-effectiveness and which resulted in worse cost-effectiveness. Results were most sensitive to the discount rate, with no discounting improving cost-effectiveness to $27,000 per QALY, and a 5% discount rate resulting in $128,000 per QALY. Results were also sensitive to the hepatitis B incidence ratio for people with diagnosed diabetes relative to people without diabetes, hepatitis B progression rates, and complication costs.

The cost-effectiveness ratio based on the probabilistic sensitivity analysis results is $74,478 per QALY gained, slightly lower than the results using baseline values. The bootstrapped 95% credible interval for the cost-effectiveness ratio ranges from $69,000 to $80,400 per QALY gained.

As uptake rates increase, net costs and QALYs saved increase as more people are vaccinated. The cost-effectiveness ratio does not change because its numerator and denominator increase proportionately (Supplementary Data).

If the hepatitis B vaccine were available at the lower CDC price, the program cost would decline and the cost-effectiveness ratio would decrease to $41,622 per QALY gained (Supplementary Data).