Preferences for herpes zoster vaccination among adults aged 50 years and older in the United States: results from a discrete choice experiment

ABSTRACT

Background

Most adults, and disproportionately fewer African-Americans, have not received herpes zoster (HZ) vaccination despite current recommendations. This study (GSK study identifiers: 208677/HO-17-18066) assessed HZ vaccination preferences among adults aged ≥ 50 years.

Research design and methods

In this discrete choice experiment, respondents chose among a ‘no vaccine’ option and two HZ vaccine profiles, characterized by seven attributes, in a series of choice questions. Random-parameters logit results were used to predict likely vaccine uptake. Subgroup and latent class analysis of African-American’s preferences were performed.

Results

The preference weight for choosing HZ vaccines over no vaccine was statistically significant among the 1,454 respondents (71.9% whites; 25.2% African-Americans). Out-of-pocket (OOP) cost and vaccine effectiveness (VE) were the most important attributes. The African-American and the non-African-American subgroups had statistically significant differences in preferences (χ² = 59.91, p < 0.001), mainly driven by OOP cost and VE. Latent class analysis identified three groups of African-American respondents with systematically different preferences; two comprised likely-vaccinators, with one being more cost sensitive at lower price thresholds, and one likely non-vaccinators.

Conclusions

For all respondents, HZ vaccine choices were most sensitive to total OOP cost, followed by VE.

PLAIN LANGUAGE SUMMARY

What is the context?

• Herpes zoster, or shingles, is a viral disease characterized by a painful, localized skin rash. It affects approximately 32% of US citizens at least once in their lifetime.

• The risk of contracting shingles increases with age.

• Most American adults over 50 years have not received the shingles vaccine, and vaccination rates are especially low for African-Americans. What is new?

• This is the first study to evaluate what drives shingles vaccination decisions among US adults ≥ 50 years of age. We also assessed the differences between African-American and non-African-American adults, and inside the African-American group.

• In this choice experiment, 1,454 people ≥ 50 years completed a survey of 8 choice questions, as well as questions on their previous experiences with vaccines, socioeconomic, and demographic characteristics. Seven factors were evaluated.

• We found that American adults preferred to get vaccinated, and the most influential factors were costs and vaccine effectiveness while location of vaccination was the least important. There were differences in preferences between African-American and non-African-American adults, mainly driven by costs and vaccine effectiveness. 3 different groups of African-American adults with systematically different preferences could be identified; two were likely to vaccinate, with one being more cost sensitive at lower price thresholds, and the third was unlikely to vaccinate.

What is the impact?

• Decisions on shingles vaccination appear to be mostly driven by costs, which could be a barrier to those who do not have appropriate insurance, especially among some African-Americans.

• However, healthcare professionals should continue to educate patients on other vaccine characteristics, as they also influence vaccination decisions.

Graphical abstract

KEYWORDS:

• Discrete choice experiment
• herpes zoster
• older adults
• stated preferences
• vaccination

1. Introduction

One million people in the United States (US) are estimated to develop herpes zoster (HZ) every year, with about 32% of US population developing HZ once in their lifetime [1]. Additionally, some will suffer from HZ more than once, with an estimated cumulative recurrence increasing by 1% each year after incident HZ episode [2]. The risk of HZ increases with age, especially after the age of 50 years [1,3,4]. Autoimmune diseases and other comorbidities are associated with a higher risk of HZ [5]. In the US, estimates of the annual economic burden associated with HZ are between $782 million [6] and $5 billion [7]; it is therefore a significant source of health care resource consumption. This vaccine-preventable disease [8] burden can be reduced with case-avoidance strategies [9] that would advocate for policies encouraging vaccination.

A single dose Zoster Vaccine Live (ZVL; Zostavax, Merck & Co., Inc) was available for use in the US from 2006 [10] to November 2020 [11]. However, only one HZ vaccine is currently available: the two-dose adjuvanted Recombinant Zoster Vaccine (RZV; Shingrix, GSK) licensed in 2017 [11,12]. With licensure of the new vaccine in 2017, the Advisory Committee on Immunization Practices (ACIP) revised its HZ vaccine recommendations for older adults in the US so that all adults aged ≥ 50 years should receive the preferentially recommended RZV vaccine regardless of prior HZ vaccination status [13].

During a decade of availability with the first HZ vaccine (ZVL), vaccination coverage (VC) remains low, increasing from 6.7% in 2008 [14] to 34.9% in 2017 [15] among adults aged ≥ 60 years. In addition, VC among adults aged ≥ 60 years in 2017 was also associated with an ethnicity disparity in coverage (39.3% VC in whites, 17.1% in African-Americans, and 19.9% in Hispanics) [15]. This specific discrepancy in HZ prevention may result in disparities in the growing public health and economic burden of HZ as the population aged ≥ 50 years increases over time [16].

Very little is known about what drives individual HZ vaccination decisions. To date, only one published study has examined HZ vaccine preferences and acceptance among older adults [17]. However, the focus of this Dutch study evaluated the relative importance of HZ vaccination decision-making relative to pneumococcal, pertussis, and influenza vaccines and disease characteristics, and did not consider the relative importance of the characteristics of HZ vaccines themselves.

The objective of this study was to explore the extent to which different HZ vaccine and respondent characteristics influenced personal vaccination decisions; it further evaluated systematic differences in preferences between African-American respondents and non-African-American respondents and preference heterogeneity within African-American respondents. Finally, it explored whether and how these characteristics influenced respondents stated intent to receive the vaccination and likely compliance with the second dose.

2. Patients and methods

2.1. Survey design

A cross-sectional discrete choice experiment (DCE) survey developed using good practices [18] and administered in January 2019 to adults aged ≥ 50 years in the US (GSK study identifiers: 208677/HO-17-18066).

The survey consisted of: a) DCE questions; b) questions about compliance with a two-dose vaccine; and c) measures of respondents characteristics (Supplementary File 1). Respondents were assigned to one of two survey versions based on their age-specific risk of HZ: a lower risk for those aged 50–64 years and higher risk for those aged ≥ 65 years [1,19,20].

Respondents were presented with a series of choices, each with two hypothetical HZ vaccines (Vaccine A or Vaccine B) and a no-HZ-vaccine option. The selection of seven attributes and attributes levels were informed by the features that could both characterize and differentiate the HZ vaccines: vaccine effectiveness (VE), duration of protection, location of vaccination, number of doses, days with flu-like symptoms (FLS), injection-site reactions (ISR) severity, and total out-of-pocket (OOP) cost (Table 1). The ranges for the levels were based on the product labels for the available HZ vaccines and on published clinical data [10,12,13,21,22].

Table 1. Attributes and levels for the DCE

Attribute	Levels for experimentally designed vaccines		Levels for the no vaccine option
Number of HZ cases prevented in the next 5 years^a	Low Baseline Risk Prevents 10 out of 33 cases in 1,000 people (a 30% reduction) Prevents 17 out of 33 cases in 1,000 people (a 50% reduction) Prevents 32 out of 33 cases in 1,000 people (a 97% reduction)	High Baseline Risk Prevents 16 out of 54 cases in 1,000 people (a 30% reduction) Prevents 27 out of 54 cases in 1,000 people (a 50% reduction) Prevents 53 out of 54 cases in 1,000 people (a 97% reduction)	Prevents 0 out of [33,54] cases in 1,000 people (a 0% reduction)
Number of years until you need to get the vaccine again	2 years 12 years 20 years		None
Where you can go to get the vaccine	Doctor’s office only Doctor’s office or pharmacy		None
Number of vaccine doses	1 dose 2 doses		None
Number of days with FLS^b	No days 1 day 3 days		No days
Severity of ISR^c	None Mild Moderate-to-severe		None
Total OOP cost for all vaccine doses	$0 for all doses $15–$25 for all doses $280–$300 for all doses		$0 for all doses

^aRespondents were presented with different baseline risks of contracting HZ in the next 5 years based on their age. Respondents aged 50–64 years (i.e. the commercial health plan population) were shown a baseline HZ risk of 33 out of 1,000 people in the next 5 years. Respondents aged ≥ 65 years (i.e. the Medicare population) were shown a baseline HZ risk of 54 out of 1,000 people. Therefore, VE was communicated to respondents as the percent reduction in the number of HZ cases from baseline. The percentage point change in VE remained the same between the two versions.

^bLevels of FLS were labeled differently depending on the number of vaccine doses shown in each profile. No days, 1 day, and 3 days were the labels when the number of vaccine doses was one; however, for vaccine with 2 doses, the text ‘2 days total’ was added in parentheses after 1 day and ‘6 days total’ was added in parentheses after 3 days.

^cLevels of severity of ISR were labeled differently depending on the number of vaccine doses shown in each profile. When the level for the number of vaccine doses attribute was one, the labels for the ISR attribute were: no reactions, 1 mild reaction, and 1 moderate-to-severe reaction. When the level for the number of vaccine doses attribute was two, the labels for the ISR were no reactions, 2 mild reactions, and 2 moderate-to-severe reactions.

Abbreviations: DCE, discrete choice experiment; FLS, flu-like symptoms; HZ, herpes zoster; ISR, injection-site reactions; OOP, out-of-pocket; VE, vaccine effectiveness.

2.1.1. DCE questions

The vaccine profiles in each choice question were determined by an experimental design using the attributes and levels in Table 1. DCE applications commonly use a D-optimal experimental design to reduce the number of paired comparisons to the smallest number for efficient estimation of the preference weights [23–26]. The full design included 72 unique choice questions. However, there is a limit to the number of questions each respondent can answer before becoming fatigued. A review of empirical DCE studies in health care showed that the number of questions presented to each respondent is 12 or fewer in most studies [27]. In keeping with this practice, the current study was designed in nine blocks, each consisting of eight DCE questions. Each respondent was randomly assigned to one of the nine blocks and the order of the questions was randomized across respondents to limit the influence of question order or fatigue on choice question’s responses. An example choice question is shown in Figure 1. Before answering the DCE questions, the vaccine features were defined one-by-one and respondents were shown partial DCE questions to familiarize them with the question content and format. There were also a number of questions to assess respondent comprehension of the survey content. For example, for the vaccine feature ‘number of HZ cases prevented in the next 5 years’, the respondents read a description of the attribute and a description of the risk pictograms. Questions were asked to assess respondent comprehension of the pictograms and upon successfully testing, the survey respondents proceeded (see the survey in the Supplementary File 1 for details).

Figure 1. Sample DCE question from the high-baseline-risk (aged ≥ 65 years, Medicare population) version

Notes: a) Each respondent answered eight DCE questions; the vaccine profiles and pairs were determined by an experimental design (as described in the text). b) ‘Number of HZ cases prevented in the next 5 years’: Risk pictograms were used to describe vaccine effectiveness. There are 1,000 dots in each risk pictogram. Each dot represents one person who has received an HZ vaccine. The gray dots show how many people will not get HZ in the next five years. The blue dots show how many people will get HZ in the next five years. The orange dots show the number of people that would have gotten HZ without the vaccine, i.e. the number of HZ cases prevented by the vaccine. Abbreviations: DCE, discrete choice experiment, FLS, flu-like symptoms; HZ, herpes zoster; ISR, injection-site reactions; OOP, out-of-pocket.

2.1.2. Compliance questions

Immediately following the DCE questions, respondents were presented with a shorter series of stated-preference questions to assess compliance with a two-dose vaccine. Each question described a two-dose HZ vaccine and the side effects and OOP cost experienced upon completing the first dose of the vaccine. In particular, respondents were provided with hypothetical vaccine profiles requiring two doses, which were meant to be given 12 months apart, were 97% effective, and would last for 12 years. These levels were similar to current ACIP vaccination recommendations and other published data [13,21,22]. Respondents were then presented with three pairs of questions in which each pair described a specific ‘first-dose’ experience in terms of FLS and ISR, where the levels were determined by a full factorial experimental design, and an OOP cost of $8–$13. The first question in each pair asked respondents whether they would have a second dose if the experience was the same as the first in terms of FLS, ISR, and OOP cost. If the respondent said that they would definitely or probably have the second dose, then the question was repeated with a higher OOP cost ($140–$150 per dose) and constant FLS and ISR levels. If the respondent said that they would definitely or probably not have the second dose based on their ‘first-dose’ experience, the question was repeated with a lower OOP cost ($0 per dose).

2.1.3. Measures of respondent characteristics

The final online survey also contained questions on previous experience with vaccines and vaccination history, socioeconomic and demographic characteristics, as well as questions designed to provide information about respondents’ perceptions of vaccines and awareness of HZ in order to examine whether these factors influenced vaccination preferences. To further explore the potential disparities in vaccine preferences between African-American adults and non-African-American adults, the survey included questions designed to measure respondents’ perceptions of the fairness of policies in general as well as health care services, health, and vaccination recommendations.

2.1.4. Pretests

The draft survey instrument was finalized based on the findings from pretest interviews with 15 adults aged ≥ 50 years. The purpose of the interviews was to assess whether the survey was well-understood, the appropriateness of the descriptive information related to HZ, and the difficulty of the DCE questions (details in Supplementary File 2).

2.2. Study population

Eligible respondents were aged ≥ 50 years, living in the US, and able to read and understand English. Dynata (https://www.dynata.com/), a health-related market research firm, recruited survey respondents from their general consumer panel. As compensation for their time, respondents received loyalty points for completing the survey, which could be accumulated and be exchanged for cash or gifts. All respondents provided informed consent electronically before completing the survey. The survey was granted an exemption from review by RTI International’s institutional review board (Federal Wide Assurance #3331).

2.2.1. Stratification and sample size

The study population was stratified such that half of the participants should have received an influenza vaccine at least once in the last two years – a subgroup which was considered a proxy for patients who tend to get vaccines. Approximately 25% of the participants were to be African-American, and approximately 20% of the participants were to be at increased risk for HZ due to condition or treatment for select conditions (i.e. rheumatoid arthritis, ulcerative colitis, Crohn’s disease, chronic obstructive pulmonary disease, type 2 diabetes, lupus, and psoriasis). Participants aged between 50–64 years old were likely to be covered by commercial health plans and those aged ≥ 65 years were likely to be covered by Medicare.

Most published DCE studies have a sample size between 100 and 300 respondents [27]. Therefore, a sample size of 1,450 respondents was expected to be sufficient to estimate preference weights for each attribute level included in the study.

2.3. Statistical methods

The vaccine-choice data from the DCE survey were analyzed using a random-parameters logit (RPL) model (details in Supplementary File 3) including all attribute levels shown in Table 1.

The regression parameters from an RPL model are preference weights, or utilities of attribute levels, indicating the relative strength of preference for each attribute level. More-preferred outcomes have higher preference weights. Conditional relative importance of an attribute was calculated on the difference between the preference weight for the attribute level with the highest preference weight and the preference weight for the level of the same attribute with the lowest preference weight. This difference represents the maximum change in utility achievable with any attribute, conditional upon the levels chosen for the attributes in the study. The RPL parameters were also used to predict likelihood of choosing a specific HZ vaccine profile. Finally, the RPL results were used to estimate the minimum acceptable benefits (MAB) for various changes in vaccine-related features (details in Supplementary File 3).

Differences in preferences were assessed for four pairs of mutually exclusive respondent subgroups. This paper describes the results of one of the subgroup analyses – the analysis of the African-American and non-African-American subgroups. The remaining subgroup analyses are available in the supplementary materials. To evaluate differences in preferences, a dummy-coded variable identifying respondents belonging to a subgroup was interacted with each level of the study attributes. The interaction terms can be interpreted as the difference in preferences between a baseline group (not represented by the identifier dummy-coded variable) and a subgroup of interest. Thus, a chi-squared (χ²) test of joint significance of all interaction terms provided the necessary information to determine whether preferences were systematically different between these subgroups. The χ² tests were performed separately for each set of interaction terms involving a particular dummy-coded interaction term.

To further explore preference heterogeneity among the African-American subgroup, a post-hoc analysis using latent class (LC) models (details in Supplementary File 3) [28] was conducted to identify groups within the African-American respondents with distinct preferences. The preference weights from the LC models were used to estimate the conditional relative attribute importance for each group. The results from the LC model were also applied to HZ vaccine profiles with fixed effectiveness, duration of protection, and doses as well as where patients can get the vaccine, varying levels of cost, ISR, and FLS to predict likely uptake, or the proportion of African-Americans within each LC group that would select a vaccine with certain characteristics over no vaccine.

Finally, responses to the DCE questions were analyzed using logit regression to explore whether and how vaccine and respondent characteristics influenced stated willingness to get vaccinated. For these analyses, the dependent variable was 1 if the respondent chose one of the hypothetical vaccine profiles and 0 if no vaccine was chosen. Descriptive statistics were conducted using SAS version 9.4 (SAS Institute Inc, Cary, NC), while the analysis of the DCE data were performed with NLOGIT (Econometric Software, Plainview, NY).

3. Results

3.1. Sample characteristics

In total, the final survey was accessed by 8,524 individuals of whom 3,353 were eligible and gave consent to participate; 1,454 (43%) of these completed the survey. Only completed surveys were included in the analysis. Table 2 summarizes respondents’ characteristics. The mean age was 64.4 years (standard deviation 7.8). Half of the respondents had received an influenza vaccination in the past two years and 25.2% were identified as Black or African-American.

Table 2. Respondent characteristics (N = 1,454)

Characteristic	Responses, n (%)
Aged 50–64 years	712 (49.0)
Aged ≥ 65 years	742 (51.0)
Female	852 (58.6)
Highest level of education^a
Less than high school/some high school but no degree	26 (1.8)
High school or equivalent	254 (17.5)
Some college but no degree/technical school/associate’s degree (2-year college degree)	580 (39.9)
4-year college degree	319 (21.9)
Some graduate school but no degree	58 (4.0)
Graduate or professional degree	217 (14.9)
Employment status^a
Employed full-time	278 (19.1)
Employed part-time	115 (7.9)
Self-employed/homemaker	149 (10.2)
Student	3 (0.2)
Retired	743 (51.1)
Disabled/unable to work	106 (7.3)
Unemployed	60 (4.1)
Race or ethnicity^b
American Indian or Alaska Native	17 (1.2)
Asian Indian	2 (0.1)
Black or African-American	366 (25.2)
Chinese	8 (0.6)
Filipino	2 (0.1)
Japanese	7 (0.5)
Korean	2 (0.1)
Native Hawaiian	2 (0.1)
Other Asian	4 (0.3)
Other Pacific Islander (for example: Fijian, Tongan, etc.)	0
White	1,046 (71.9)
Other	21 (1.4)
I prefer not to say	7 (0.5)
Had influenza vaccination in the past 2 years (since 2016)	728 (50.1)
Health insurance^b
I do not have health insurance	63 (4.3)
Private insurance that I pay for myself (including Medicare supplemental)	235 (16.2)
Private insurance that my employer or my spouse’s employer pays for	357 (24.6)
Medicaid	195 (13.4)
Medicare	749 (51.5)
Veterans’ Health Insurance	89 (6.1)
Other	54 (3.7)
Don’t know or not sure	8 (0.6)
Total household income before tax and other deductions in 2017^a
Less than $25,000	301 (20.7)
$25,000 to $49,999	405 (27.9)
$50,000 to $74,999	259 (17.8)
$75,000 to $99,999	170 (11.7)
$100,000 to $149,999	142 (9.8)
$150,000 to $199,999	58 (4.0)
$200,000 or more	40 (2.8)
Do not know or not sure	2 (0.1)
Prefer not to answer	77 (5.3)
Among respondents who identify as Black or African-American (n = 366)
Ethnic background/origin^b
Africa	175 (47.8)
The Caribbean	34 (9.3)
Central, Latin, or South America	7 (1.9)
Don’t know or not sure	153 (41.8)
I prefer not to say	12 (3.3)

^aSums of percentages may not sum to 100.0% due to rounding.

^bThe percentages may sum to more than 100% because respondents had the option to select more than one response category.

3.2. Preference weights and conditional relative importance for overall sample

As shown in Figure 2, a value of −3.3 on the alternative specific constant indicates that, in general, adults ≥ 50 years had a strong and statistically significant preference for choosing HZ vaccination rather than opting out of HZ vaccination. The estimated preference weights for all attributes were consistent with the natural ordering of the levels – that is, better outcomes or features were preferred to worse outcomes or features (Figure 2). Therefore, on average, respondents preferred the following:

Figure 2. Preference weights and conditional relative importance of attributes of HZ vaccines for the full sample (N = 1,454)

Notes: a) The dark dots in the figure represent the point estimates of relative mean preference weight for attribute levels. b) The vertical bars surrounding each mean preference weight denote the 95%CI about the point estimate. c) The asterisks indicate the segments or differences in attribute levels that were significant at the 95%CI. d) The blue rectangles and associated numbers indicate the conditional relative importance of HZ vaccine attributes. With the exception of ‘where you can go to get the vaccine’, the conditional relative importance weights for all other vaccine attributes were statistically significantly different from zero. In addition, the differences between the conditional relative importance weights for each attribute were statistically significantly different from one another at the 95%CI, with the exception of the following: duration of effectiveness and FLS, duration of effectiveness and ISR, FLS and ISR. Abbreviations: CI, confidence interval; FLS, flu-like symptoms; HZ, herpes zoster; ISR, injection-site reactions; OOP, out-of-pocket.

• Vaccines that prevented more HZ cases in the next 5 years

• Vaccines with one dose (rather than 2)

• Vaccines with fewer days of FLS and less-severe ISR

• Vaccines with lower total OOP cost

The preference weights of all adjacent levels of the same attribute were statistically significantly different from one another (p < 0.05) with the following exceptions:

• Respondents did not differentiate between 12-year and 20-year duration until the vaccine would need to be administered again.

• Respondents did not differentiate between the levels of where you can go to get the vaccine (doctor’s office only or doctor’s office or pharmacy).

Figure 2 also shows the conditional relative importance estimate for each attribute. Preferences were strongest for total OOP cost, choosing vaccine alternatives over no vaccine, and VE. As an example, an increase in VE from 30% to 97% had a relative importance of approximately 2.5 while a decrease of total OOP cost from $280–$300 to $0 had a relative importance of 3.4; thus, this change in cost was therefore 1.4 times (= 3.4 ÷ 2.5) more important than the change in VE.

The MAB for various changes in vaccine-related features are presented in Supplementary Table 1.

3.3. Subgroup analyses

The results of the χ² test for joint significance indicated that there were statistically significant differences in preferences between African-Americans and non-African-Americans (χ² = 59.91, p < 0.001). The systematic differences in preferences were largely driven by cost, VE, and the utility associated with the non-vaccine alternative.

OOP cost and getting a vaccine rather than not getting a vaccine were the most important for the African-American and the non-African-American subgroups, but OOP cost was slightly more important to the African-American subgroup than it was to the non-African-American subgroup, though the different was not statistically significant (Figure 3). The African-American subgroup placed less importance on VE than the non-African-American subgroup.

Figure 3. Conditional relative importance weights of attributes of HZ vaccines for African-American subgroups

Note: The vertical bars surrounding each mean preference weight denote the 95%CI about the point estimate. The asterisk indicates that the conditional relative attribute importance was statistically significantly different between subgroups at the 95%CI. Abbreviations: CI, confidence interval; FLS, flu-like symptoms; HZ, herpes zoster; ISR, injection-site reactions; OOP, out-of-pocket.

The results of the χ² test for joint significance indicated that there were statistically significant differences in preferences between all mutually exclusive subgroups, with the exception of respondents who were autoimmune compromised or had other at-risk conditions compared with respondents who were not autoimmune compromised or did not have other at-risk conditions; therefore, no conditional relative importance weights were calculated for this subgroup. The conditional relative importance results for the remaining subgroups are reported in Supplementary Figures 1 and 2.

3.4. Preference heterogeneity among African-Americans

The LC analysis identified three distinct groups (or classes) of African-American respondents with systematically different vaccine preferences. Respondents had a 30% probability of being in Group 1, a 21% probability of being in Group 2, and a 49% probability of being in Group 3. To better illustrate the differences in the disutility associated with the opt-out alternative among the three groups, Figure 4 displays the conditional relative importance weights for changes that result in a net disutility as negative values. Figure 4 shows that the differences in preferences across these groups lie in the importance placed on getting an HZ vaccine over no vaccine, total OOP cost, and the duration of VE:

Figure 4. Conditional relative importance weights of attributes of HZ vaccines for latent groups within the African-American subgroup (N = 366)

Note: The vertical bars surrounding each mean preference weight denote the 95%CI about the point estimate. Abbreviations: CI, confidence interval; FLS, flu-like symptoms; HZ, herpes zoster; ISR, injection-site reactions; OOP, out-of-pocket.

• Group 1 respondents can be described as cost-sensitive likely vaccinators. The most important attribute to this group was OOP cost, relative to the other attributes and levels included in the study, and these respondents were more likely to vaccinate rather than opt out of vaccination.

• Group 2 respondents were likely non-vaccinators. Their most important attribute was duration of VE relative to the other attributes and levels included in the study; however, group 2 respondents were unlikely to opt for vaccination.

• Group 3 respondents were the cost-balancing likely vaccinators. Choosing HZ vaccine over no vaccine was their most important attribute relative to the other attributes and levels included in the study, and respondents in this group were most sensitive to changes in total OOP cost.

Figure 5 illustrates the differences in predicted choice probabilities (likelihood of choosing a vaccine) between the three groups given specific changes in OOP cost (panel A), ISR (panel B) and FLS (panel C). For example, panel A shows how predicted uptake would change for specific changes in OOP cost, holding ISR and FLS constant, while panel B shows how predicted uptake would change for specific changes in ISR, holding OOP and FLS constant. The results in Figure 5 can be interpreted in the following ways:

Figure 5. Predicted choice probabilities or uptake for HZ vaccines with different attributes among latent groups within the African-American subgroup (N = 366)

Note: All other vaccine characteristics are held constant for these comparisons. Abbreviations: FLS, flu-like symptoms; HZ, herpes zoster; ISR, injection-site reactions; OOP, out-of-pocket; vs, versus.

• Panel A shows that Group 1 respondents were the most sensitive to changes in total OOP cost, as predicted uptake was 78.2% when the vaccine cost was $0 and 45.8% when the cost was $280–$300. In contrast, Group 3 respondents are not as sensitive to cost, with predicted uptake only going down by approximately 5 percentage points when the cost of the vaccine goes from $0 to $280–$300. Group 2 respondents were unlikely to vaccinate, even when the vaccine had no cost (14.8% of predicted uptake).

• Panel B shows that Group 1 is somewhat sensitive to changes in ISR, with predicted uptake going up by approximately 8 percentage points when ISR goes from mild-to-severe to none. In contrast, as with OOP cost, Group 3 is not sensitive to changes in ISR with uptake varying very little if at all from 99%, and Group 2 respondents were unlikely to vaccinate even with no ISR (14.7% of predicted uptake).

• Panel C shows that, as for changes in ISR, Group 1 is somewhat sensitive to changes in FLS, with predicted uptake going up by approximately 5 percentage points when the days with FLS go from 3 to none. As for OOP cost and ISR, Group 3 was also the least sensitive to changes in FLS, with stable predicted uptake of 99.4%, while Group 2 remains unlikely to vaccinate even when the vaccine has no days with FLS (14.7% of predicted uptake).

3.5. Stated willingness to get an HZ vaccine

The logit regression analysis results indicate that respondents who trust health care providers (HCPs) vaccine recommendations were more likely to choose having HZ vaccination than not having HZ vaccination (Supplementary Figure 3). Respondents who believed that HZ is a serious disease and that they would likely contract HZ in the next five years were also more likely to choose HZ vaccination rather than not vaccinating, as were respondents with recent influenza vaccination experience, experience with a two-dose vaccine, and those who agreed that vaccines were safe. In contrast, the following respondent characteristics were associated with a lower likelihood of choosing HZ vaccination over not vaccinating: had never experienced any ISR, had ever chosen not to get a recommended vaccine, and thinking that it was unlikely or unsure that the HZ vaccine would work well for them.

The logit regression results also indicated (Supplementary Table 2) that increasing VE, or its duration, was associated with a higher likelihood of choosing HZ vaccination. Vaccines with two doses, FLS for three days, moderate-to-severe ISR, and total OOP cost of $15–$25 or $280–$300, were associated with a lower likelihood of choosing HZ vaccination over not vaccinating.

3.6. Stated compliance with a two-dose vaccine

Approximately 75% of the sample would opt for a second dose of a two-dose, 97% effective HZ vaccine that lasts 12 years if the vaccine costs $8–$13 per dose (Table 3). Of those, 51.2% would still have a second dose at the higher cost of $140–$150, holding the FLS and ISR constant. The majority (> 80%) of the respondents who would not have a second dose at the fixed cost of $8–$13 would still not have a second dose even if it costs $0.

Table 3. Frequency of responses to the second dose compliance questions for the full sample (N = 1,454)

Question and response	First-dose vaccine			Average across all sets (%)
	Set 1, n (%)	Set 2, n (%)	Set 3, n (%)
Given your experience with the first dose, would you get the second dose if it costs $8–$13?	N = 1,454	N = 1,454	N = 1,454
Definitely yes or probably yes	1,116 (76.8)	1,053 (72.4)	1,127 (77.5)	75.5
Definitely no or probably no	338 (23.2)	401 (27.6)	327 (22.5)	24.4
Would you get the second dose if it costs $140–$150?	n = 1,116	n = 1,053	n = 1,127
Definitely yes or probably yes	572 (51.3)	545 (51.8)	571 (50.7)	51.2
Definitely no or probably no	544 (48.7)	508 (48.2)	556 (49.3)	48.8
Would you get the second dose if it costs $0?	n = 338	n = 401	n = 327
Definitely yes or probably yes	64 (18.9)	79 (19.7)	61 (18.7)	19.1
Definitely no or probably no	274 (81.1)	322 (80.3)	266 (81.3)	80.9

Notes:

a) The hypothetical vaccine profiles in these questions were described as requiring two doses, which were meant to be given 12 months apart, 97% effective, and lasting for 12 years.

b) The FLS and ISR varied. The results in the table combine responses across scenarios describing varying levels of FLS and ISR: 9 combinations i.e. 3 levels of FLS (0, 1, or 3 days with FLS) × 3 levels of ISR (none, mild, moderate-to-severe). Each of these nine possible combinations of FLS and ISR were grouped into sets of three different first-dose vaccine scenarios. Each of the nine possible combination appeared an equal number of times and was combined with other combinations an equal number of times across the three pairs of second-dose compliance questions. This led to 84 sets, where each set included three combinations of FLS and ISR. Each respondent was randomly assigned to one set.

c) Sums of percentages may not sum to 100.0% due to rounding.

Abbreviations: FLS, flu-like symptoms; ISR, injection-site reactions.

4. Discussion

This DCE study on preferences for HZ vaccines in the US found that adults aged ≥ 50 years preferred to get an HZ vaccine over no vaccine and that the total OOP cost was the most important vaccine characteristic conditional on the attributes and levels shown in the study; VE was the next most important characteristic. Local and systemic reactions associated with HZ vaccines and number of doses were of lesser importance; VE was approximately five times more important than reducing the number of days with FLS and the severity of ISR, and ten times more important that decreasing the number of doses from two to one. The least important characteristic was the location where the vaccine would be administered. Most respondents would complete a series of vaccination doses at the lowest suggested fixed price for the second dose ($8–$13), but nearly half would not do so if the second dose costs $140–$150.

To our knowledge, this is the first study designed to comprehensively assess US adult preferences for HZ vaccines. Other studies, conducted on either general or specific (other than HZ) vaccine preferences among adults, have also shown that VE and OOP cost are key factors in vaccination decision-making [29–37]. Not all DCE studies have considered the same set of vaccine attributes. Two studies conducted in the Netherlands showed that VE, longer duration of protection, disease risk, and mortality rates were the determining factors in vaccination decision-making [17,30]; cost of the vaccine was not included in the vaccine attributes considered in these studies. A study from South Africa showed that the most favored vaccine profile would have 90% VE [31]. Interestingly, the decision to be vaccinated was positively associated with increasing VC, i.e. respondents were more likely to choose vaccination if more people around them were getting vaccinated [31]. A study on preferences for hepatitis B vaccines in China showed that protection rates and protection duration were important attributes, while location for receiving the vaccination and number of doses were less important [29]. In China, those born before 1995 must pay for hepatitis B vaccination [29], and this study showed that lower OOP cost was preferred; however, respondents were willing to pay more for increased VE. Thus, they were willing to pay a two times higher price for protection lasting 20 instead of 10 years, and a two and a half times higher price for a 10% increase in protection (from 89% to 99%) [29]. In the US, a study using a stated preference binary choice experiment showed that VE was the most important factor influencing vaccination decisions, followed by OOP cost and recommendation from a primary care physician [32]. The risk of side effects and vaccination location were not crucial factors [32]. The study included 989 adolescents and adults, of whom adults were mainly whites (74.6%) and college-educated [32].

Our analysis showed that, within the African-American subgroup, there were three distinct groups of respondents with systematically different preferences suggesting heterogeneity in vaccine decision-making. More specifically, these three groups expressed different willingness to vaccinate and sensitivity to total OOP cost. Likely non-vaccinators were in one group (group 2) and likely vaccinators were in two groups (groups 1 and 3). The probability of belonging to the likely non-vaccinator group was one out of five and was lower than the probability of belonging in the other two groups. For the two likely vaccinator groups, respondents placed different levels of importance on OOP cost, where respondents in group 3 were less sensitive to changes in cost and had a predicted uptake that remained high even when OOP cost was at the highest levels. In contrast to group 3, predicted uptake for group 1 dropped by approximately 10% and 32% when total OOP cost increased from $0 to $15–$25 and $280–$300 respectively. The likely non-vaccinators had a predicted uptake of about 15%, even when the vaccine costs $0, vaccination had no ISR, and no days with FLS.

The concerns about OOP cost elicited in our findings and the other studies highlight a key issue for payers and policymakers as it appears that VC is sensitive to vaccine price, even when the vaccine reduces the incidence of a disease by nearly 100%. Cost burdens faced by the patients could be limited through minimization or elimination of cost sharing [38,39]. High patient copay was a barrier to vaccinations among Medicare Part D patients, as shown in an analysis of 2012–2014 US administrative claims data [40]. Moreover, retrospective examination of a US pharmacy’s 2014 claims data reported that 38.9% of HZ vaccine prescription fills were not dispensed to the respective individuals because the amount of cost sharing involved was higher than expected [41]. Providing vaccines at low or no copay may improve vaccination rates in these adults The implementation of zero cost sharing provisions of the Affordable Care Act in 2010 increased by one million the number of women aged 19–25 years who initiated three-dose series (≥ $390 total cost) of human papillomavirus vaccination [42]. However, our findings show that although the cost was a concern for the majority of respondents, a small but nevertheless considerable number of respondents would not complete a two-dose vaccination series even if the second dose was offered free of charge. Therefore, for some people, price is not the driving factor when deciding whether to complete a vaccination series; these people may be more sensitive to safety and side effects, or less likely to seek vaccination in general [43]. Other people may believe that a second dose is unnecessary, or they may be more familiar with a single dose regimen. More data are needed to understand such attitudes. For these populations, interventions would need to challenge attitudes toward and beliefs about vaccination by providing education. Such interventions would require significant time and effort from HCPs [43].

The logit regression model of vaccine choice helped to explain some of the potential factors that may support improvement of VC in adults aged ≥ 50 years. The results revealed that respondents who have favorable opinions or attitudes related to vaccines, those who are more likely to be pro-vaccine, those who think HZ is a serious disease, and those who have experience with vaccines with more than one dose were more likely to choose a HZ vaccine over no vaccine. In addition, respondents who had received a recommendation to get the HZ vaccine from a nurse, pharmacist, or doctor, and those who trusted their HCP’s vaccine recommendation were also more likely to choose an HZ vaccine over no vaccine. Therefore, HCPs should continue to talk to their patients about the HZ disease burden and the importance of protecting against it through vaccination, as HCP recommendations appear to have a positive impact on the decision to vaccinate.

Several limitations must be considered when interpreting the results presented. The respondents evaluated hypothetical vaccines and their choices among these vaccines, which may not have the same significance as choices involving actual vaccination decisions. Actual vaccine choices may depend upon a number of contextual factors that were beyond the scope of this study. Similarly, some treatment features selected in this survey may not be relevant to all respondents. Another potential limitation is that the sample might not allow generalization. This study used a web panel to recruit study respondents and their preferences may not be representative of the overall population of adults aged ≥ 50 years in the US. Despite these limitations, this study has a number of strengths derived from the use of good practices [18]. In particular, the survey was pretested using in-depth interviews with adults aged ≥ 50 years, used an experimental design developed using good research practices [44], and followed best practices in statistical analysis [45].

5. Conclusions

Adults aged ≥ 50 years preferred getting an HZ vaccine to opting out of vaccination. However, the preferences for vaccination and stated likelihood of completing the two-dose series were sensitive to the total OOP cost. Furthermore, statistics have shown that VC is lower among racial minorities, including African-Americans. The results from this study revealed that two of the three LC groups within the African-American respondents preferred getting a HZ vaccine over no vaccine; however, the analysis identified a LC group that preferred to opt out of vaccination. In all three groups, respondents placed varying levels of importance on total OOP cost, with groups 1 and 2 being the most sensitive to changes in cost. Therefore, costs associated with the vaccine might present a barrier to some African-American adults. VE was the next most influential factor in HZ vaccination decision-making.

Authors contributions

BJ Patterson, K Myers, A Stewart, EM Hillson, and C Poulos were involved in the design of the study; BJ Patterson, K Myers, A Stewart, B Mange, EM Hillson, and C Poulos collected or generated the data; and BJ Patterson, K Myers, A Stewart, B Mange, and C Poulos analyzed and/or interpreted the data. BJ Patterson, K Myers, and C Poulos developed the manuscript outline. All authors participated in the development of this manuscript and in its critical review with important intellectual contributions. All authors had full access to the data and gave approval before submission. All authors agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The work described was carried out in accordance with ICMJE recommendations for conduct, reporting, editing, and publications of scholarly work in medical journals.

Declaration of interest

BJ Patterson was an employee of the GSK group of companies and held shares in the GSK group of companies at the time of the study conduct and of the manuscript development. K Myers and C Poulos are employees of RTI Health Solutions, and B Mange was an employee of RTI Health Solutions at the time of the study conduct. RTI Health Solutions was contracted by the GSK group of companies to design and implement the present study. A Stewart reports personal fees from the George Washington University, from the Immunization Action Coalition, Take a Stand Project, and from the Advisory Commission on Childhood Vaccines, Health Resources and Services Administration, U.S. Department of Health & Human Services, outside of the present work. EM Hillson was an employee of the GSK group of companies and held shares in the GSK group of companies at the time of the study conduct. The authors declare no other financial and non-financial relationships and activities.

Reviewer disclosures

A reviewer on this manuscript has disclosed that outside this manuscript, they have the following potential conflicts of interest to report: grants from Sanofi Pasteur MSD, the GSK group of companies, Pfizer, Sanofi Pasteur, MSD Italy, Emergent BioSolutions and Seqirus for taking part to advisory boards, expert meetings, for acting as speaker and/or organizer of meetings/congresses and as principal investigator and chief of O.U. in RCTs. This reviewer does not have any competing interest directly related to this paper. Peer reviewers on this manuscript have no other relevant financial or other relationships to disclose.

Previous presentations

IDWeek 2019; 2–6 October 2019, Washington, DC, United States

Trademarks

Shingrix is a trademark owned by or licensed to the GSK group of companies.

Zostavax is a trademark of Merck & Co., Inc.

Acknowledgments

The authors would like to thank Business & Decision Life Sciences platform for editorial assistance and manuscript coordination, on behalf of GSK. Grégory Leroux coordinated manuscript development and editorial support. Athanasia Benekou (Business & Decision Life Sciences, on behalf of GSK) provided medical writing support. The authors also thank Kimberly Moon (RTI Health Solutions) for project management support and Lecia Brown (GSK) for topical discussion about the manuscript content.

Supplemental data

Supplemental data for this article can be accessed here.

Additional information

Funding

GlaxoSmithKline Biologicals SA funded this study (GSK study identifiers: 208677/HO-17-18066) and was involved in all stages of study conduct, including analysis of the data. GlaxoSmithKline Biologicals SA also paid all costs associated with the development and publication of this manuscript.