Vaccination differences among U.S. adults by their self-identified sexual orientation, National Health Interview Survey, 2013–2015

Introduction Very few studies have explored the associations between self-identified sexual orientation and comprehensive vaccination coverage. Most of the previous studies that reported health disparities among lesbian, gay and bisexual populations were not based on a nationally representative sample of U.S. adults, limiting the generalizability of the findings. Starting in 2013, the National Health Interview Survey (NHIS) included questions to ascertain the adult’s self-identified sexual orientation that allowed national level vaccination estimation by sexual orientation. This study examined associations of self-reported vaccination coverage for selected vaccines among U.S. adults by their sexual orientation. Methods We analyzed combined data from 2013–2015 NHIS, a nationally representative probability-based health survey of the noninstitutionalized U.S. population ≥18 years. For vaccines other than influenza, weighted proportions were calculated. Influenza coverage was calculated using the Kaplan-Meier procedure. Multivariable logistic regression models were used to calculate adjusted prevalence differences for each vaccine overall and stratified by sexual orientation and to identify factors independently associated with vaccination. Results Significant differences were observed by sexual orientation for self-reported receipt of human papillomavirus (HPV), hepatitis A (HepA), hepatitis B (HepB), and influenza vaccination. Bisexual females (51.6%) had higher HPV coverage than heterosexual females (40.2%). Gay males (40.3% and 53.6%, respectively) had higher HepA and HepB coverage than heterosexual males (25.4% and 32.6%, respectively). Bisexual females (33.9% and 58.5%, respectively) had higher HepA and HepB coverage than heterosexual females (23.5% and 38.4%, respectively) and higher HepB coverage than lesbian females (45.4%). Bisexual adults (34.1%) had lower influenza coverage than gay/lesbian (48.5%) and heterosexual adults (43.8%). Except for the association of having self-identified as gay/lesbian orientation with greater likelihood of HepA, HepB, and influenza vaccination, sexual orientation was not associated with higher or lower likelihood of vaccination. Health status or other behavioral characteristics studied had no consistent relationship with vaccination among all populations. Conclusion Differences were identified in vaccination coverage among the U.S. adult population by self-reported sexual orientation. This study is the first to assess associations of sexual orientation with a comprehensive list of vaccinations. Findings from this study can serve as a baseline for monitoring changes over time. All populations could benefit from improved vaccination.


Introduction
Very few studies have explored the associations between self-identified sexual orientation and comprehensive vaccination coverage. Most of the previous studies that reported health disparities among lesbian, gay and bisexual populations were not based on a nationally representative sample of U.S. adults, limiting the generalizability of the findings. Starting in 2013, the National Health Interview Survey (NHIS) included questions to ascertain the adult's self-identified sexual orientation that allowed national level vaccination estimation by sexual orientation. This study examined associations of self-reported vaccination coverage for selected vaccines among U.S. adults by their sexual orientation.

Methods
We analyzed combined data from 2013-2015 NHIS, a nationally representative probabilitybased health survey of the noninstitutionalized U.S. population �18 years. For vaccines other than influenza, weighted proportions were calculated. Influenza coverage was calculated using the Kaplan-Meier procedure. Multivariable logistic regression models were used to calculate adjusted prevalence differences for each vaccine overall and stratified by sexual orientation and to identify factors independently associated with vaccination.
Limited research has evaluated the associations of sexual orientation with vaccination uptake [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29]. Starting in 2013, the National Health Interview Survey (NHIS), which routinely conducts surveillance of vaccination histories, included questions to ascertain adult sexual orientation. Using data from the 2013, 2014, and 2015 NHIS cycles, this study examines vaccination differences among U.S. adults by their self-identified sexual orientation and explores factors associated with vaccination for selected vaccines. We hypothesized that selfidentified gay/lesbian or bisexual populations will have higher coverage for select vaccines compared with heterosexual adults and there will be differences in coverage for select vaccines among sexual minorities. Because of greater use of health services among gay/lesbian adults and the generally higher acceptability of vaccines in these groups [29] and because HPV, hepatitis A (HepA), and hepatitis B (HepB) vaccinations protect against sexually transmitted infections, persons at higher risk of acquiring these infections because of risky sexual behaviors might be expected to have higher coverage for those vaccines.

Material and methods
The NHIS was approved by the Research Ethics Review Board (ERB number, 2009-16) of the National Center for Health Statistics, CDC and designated as "Public Health Non-Research" during the determination for applicability of human subjects' regulations, because the activity is not intended to include applicable research, but to access the implementation, coverage, performance, and/or satisfaction with an existing public health program, service, function, intervention or recommendation. Data security was addressed and written informed consent was sought.

Survey description
The NHIS is a nationally representative probability-based health survey of the civilian noninstitutionalized U.S. population conducted continuously throughout the year by the National Center for Health Statistics of the Centers for Disease Control and Prevention [38]. The NHIS uses a complex sampling design involving stratification, clustering, and multistage sampling. Results obtained from samples are weighted to the adult civilian noninstitutionalized U.S. population to minimize non-response bias. Starting in 2013, questions allowing self-identification of sexual orientation were added to the survey. The adult questionnaire is completed by one randomly selected adult aged �18 years within each family in the household. This questionnaire included questions on receipt of recommended vaccines. Three years of data from 2013, 2014, and 2015 NHIS were combined to get an adequate sample size of lesbian, gay, and bisexual adults. The combined sample for this analysis was 101,091 adults aged �18 years. There are no unique personal identifiers in the NHIS that would allow determination of repeat respondents in this study across data recruitment waves. In 2013, 2014, and 2015, the final response rates among adults were 61.2%, 58.9%, and 55.2%, respectively [39][40][41].

Sexual orientation assessment
To ascertain the adult's self-identification of sexual orientation, four cascading questions were asked of all sample adults aged �18 years. The first question asked, "Which of the following best represents how you think of yourself?" The response options for males were: gay; straight, that is, not gay; bisexual; something else; and I don't know the answer. The response options for females were: lesbian or gay; straight, that is, not lesbian or gay; bisexual; something else; and I don't know the answer. In the 2013 and 2014 NHIS, respondents who answered "something else" or "I don't know the answer" were asked one or more follow-up question(s) to gather additional information on their sexual orientation, while no follow-up question(s) were asked of respondents in the 2015 NHIS. Data from the 2013 and 2014 NHIS follow-up questions were not used in this study. Respondents answering "something else" (0.2%), "I don't know the answer" (0.5%), and refused (0.5%) were classified as missing and excluded from the analysis.

Sociodemographic and health behavior variables
Sociodemographic and health behavior variables included: demographic characteristics (age, sex, race/ethnicity, education, employment, poverty status); access to healthcare (having health insurance, having a usual place for healthcare, physician contact within the past year); health behaviors (smoking status, consuming five or more alcoholic drinks in one day at least once in the past year, ever being tested for HIV, meeting federal guidelines for aerobic physical activity, self-selected health status); selected health conditions (asthma, diabetes, obesity, cardiovascular disease, chronic liver disease, disability); and travel status (ever traveled outside of the United States to countries other than Europe, Japan, Australia, New Zealand or Canada, since 1995 where HAV and/or HBV infections are endemic). Poverty status was defined according to the ratio of a family's total income to the federal poverty threshold. Federal poverty thresholds are updated annually by the U.S. Census Bureau (https://www.census.gov/data/tables/time-series/ demo/income-poverty/historical-poverty-thresholds.html).

Increased risk for pneumococcal, hepatitis A, and hepatitis B infections
Pneumococcal disease. The presence of selected conditions that increase risk for pneumococcal disease was determined by responses to questions in the NHIS. Adults were considered at increased risk for pneumococcal disease if they had ever been told by a doctor or other health professional that they had diabetes, emphysema, chronic obstructive pulmonary disease, coronary heart disease, angina, heart attack, or other heart condition; had a diagnosis of cancer during the previous 12 months (excluding nonmelanoma skin cancer); had ever been told by a doctor or other health professional that they had lymphoma, leukemia, or blood cancer; or they had been told by a doctor or other health professional that they had chronic bronchitis or weak or failing kidneys during the preceding 12 months or had an asthma episode or attack during the preceding 12 months; or they were current smokers.
For HepA and HepB vaccination, data were collected on selected respondent characteristics that increase the risk for infection (travel to countries in which HAV infections are endemic and having chronic liver disease; travel to countries in which HBV infections are endemic, and having diabetes or chronic liver disease, respectively) [3,5].
Hepatitis A. All adults were considered at high-risk behavior for HAV infection if they reported any one of the following risk factors: ever traveled outside of the United States to countries other than Europe, Japan, Australia, New Zealand or Canada, since 1995 or if they had been ever told by a doctor; or other health professional that they had any kind of chronic, or long-term liver condition.
Hepatitis B. All adults were considered at high-risk behavior for HBV infection if they reported any one of the following risk factors: if they had ever been told by a doctor or other health professional that they had diabetes; ever traveled outside of the United States to countries other than Europe, Japan, Australia, New Zealand or Canada, since 1995; or if they had been ever told by a doctor or other health professional that they had any kind of chronic, or long-term liver condition.

Data analysis
SUDAAN statistical software (Software for the statistical analysis of correlated data, Research Triangle Institute, Research Triangle Park, NC, version 11.0.1) was used to calculate point estimates and 95% confidence intervals for specified outcome variables. T-tests were used to determine differences between demographic and other characteristics stratified by sexual orientation. Statistical significance was defined as p<0.05. For vaccines other than influenza, the weighted proportion of respondents who reported vaccination was calculated. To better assess  [42][43][44].
Adjusted prevalence differences were estimated for each vaccine overall and stratified by sexual orientation using predictive marginals under multivariable logistic regression models. Adjusted estimates for the 'overall' model were produced by adjusting for age, sex, race/ethnicity, education, employment status, poverty status, physician contact within the past year, usual place of healthcare, health insurance status, self-reported smoking status, five or more alcoholic drinks in one day at least once in past year for men/four or more alcoholic drinks in one day a least once in the past year for women, HIV test status, self-reported health status, obesity status (defined as having body mass index �30), history of asthma, diabetes status, cardiovascular disease status, chronic liver disease status, disability status, met Federal guidelines for aerobic physical activity (using 2008 Guidelines criteria [45], respondents were classified as aerobically active if they reported �150 minutes/week of light-to moderate-intensity activity, 75 minutes/week of vigorous-intensity activity, or an equivalent combination of the two), and sexual orientation, and in addition, travel status for HAV and HBV vaccination coverage estimates. These variables were adjusted for in the multivariable logistic regression models, because these variables have been shown to be associated with vaccination.
Being the primary exposure of interest, sexual orientation was not controlled for except in the "overall" multivariable regression model(s). Adjusted estimates for the 'stratified by sexual orientation' model were produced by controlling for all variables adjusted for in the 'overall' model except the sexual orientation variable, which allowed reporting of adjusted vaccination coverage levels among the study population stratified on their sexual orientation. The adjusted prevalence differences (controlling for all variables included in the 'overall' model) were used to identify factors independently associated with vaccination, which allowed assessment of sexual orientation as an independent factor of vaccination. Table 1 presents the percent distribution of sexual orientation by selected demographic, access to healthcare, health behavior, and health condition characteristics. Among adults aged �18 years, 97.6% self-identified as heterosexual, 1.6% as gay/lesbian, and 0.7% as bisexual. Respondents identifying as gay/lesbian and bisexual were more likely to report current smoking, consuming five or more alcoholic drinks in one day at least once in the past year, testing for HIV, and to have asthma compared with respondents identifying as heterosexual. Respondents identifying as gay/lesbian and bisexual differed from respondents identifying as heterosexual on a number of other demographic and access characteristics (Table 1). Table 2 presents the unadjusted self-reported vaccination coverage of selected vaccines among adults aged �18 years, overall and by sexual orientation. Overall HPV vaccination   Table 2). Adjusted vaccination coverage for each vaccine class is shown in Table 3, overall and stratified by each sexual orientation category. Overall adjusted HepA vaccination coverage among gay/lesbian and bisexual adults was significantly higher compared with heterosexual adults. Adjusted HepA vaccination coverage was significantly higher among gay males compared with heterosexual males. Overall adjusted HepB vaccination coverage among gay/lesbian and bisexual adults was significantly higher compared with heterosexual adults. Adjusted HepB vaccination coverage was significantly higher among gay males compared with heterosexual males. After adjustment, overall influenza vaccination coverage was significantly higher among gay/lesbian compared with heterosexual adults, but significantly lower among bisexual compared with gay/lesbian adults. There were no significant differences by sexual orientation for Tdap, pneumococcal (among adults aged 18-64 years with high-risk conditions and those aged �65 years) or shingles vaccination (Table 3). Table 4 presents the results of the multivariable logistic regression models by selected vaccines. Except for the association of having self-identified as gay/lesbian orientation with greater likelihood of HepA, HepB, and influenza vaccination, sexual orientation was not associated with higher or lower likelihood of vaccination. Health status or other behavioral characteristics studied had no consistent relationship with vaccination among all populations. After controlling for sexual orientation, other characteristics that are generally associated with vaccination were still independently associated with likelihood of vaccination (Table 4).

Discussion
Significant differences were observed by sexual orientation for HepA, HepB, and influenza vaccination. Gay/lesbian orientation was associated with a greater likelihood of influenza vaccination. Gay male sexual orientation was associated with greater likelihood of HepA and HepB vaccination.
This study found that HepB vaccination was higher among men who self-identified as gay compared with heterosexual, consistent with findings reported previously [29,46] but lower than reported from studies involving smaller samples of gay males from localized geographical areas and settings [47,48]. Gay/lesbian and bisexual adults also had a higher likelihood of receiving HepA and HepB vaccination. The HepA vaccination coverage observed among gay males (40.3%) in this study was lower than previously reported (69.0%) [48]. Despite recommendations for HepA and HepB vaccination of MSM due to their increased risk for HAV and Vaccination differences by self-identified sexual orientation among U.S. adults Adults were considered at increased risk for pneumococcal disease if they had ever been told by a doctor or other health professional that they had diabetes, emphysema, chronic obstructive pulmonary disease, coronary heart disease, angina, heart attack, or other heart condition; had a diagnosis of cancer during the previous 12 months (excluding nonmelanoma skin cancer); had ever been told by a doctor or other health professional that they had lymphoma, leukemia, or blood cancer; or they had been told by a doctor or other health professional that they had chronic bronchitis or weak or failing kidneys during the preceding 12 months or had an asthma episode or attack during the preceding 12 months; or they were current smokers.
https://doi.org/10.1371/journal.pone.0213431.t003 Vaccination differences by self-identified sexual orientation among U.S. adults Vaccination differences by self-identified sexual orientation among U.S. adults  Vaccination differences by self-identified sexual orientation among U.S. adults   States to countries other than Europe, Japan, Australia, New Zealand or Canada, since 1995 or if they had been ever told by a doctor; or other health professional that they had any kind of chronic, or long-term liver condition. b Adults were considered at high-risk for hepatitis B virus (HBV) infection if they reported any one of the following risk factors: ever traveled outside of the United States to countries other than Europe, Japan, Australia, New Zealand or Canada, since 1995; or if they had been ever told by a doctor or other health professional that they had any kind of chronic, or long-term liver condition; or if they had ever been told by a doctor or other health professional that they had diabetes. c Adults were considered at increased risk for pneumococcal disease if they had ever been told by a doctor or other health professional that they had diabetes, emphysema, chronic obstructive pulmonary disease, coronary heart disease, angina, heart attack, or other heart condition; had a diagnosis of cancer during the previous 12 months (excluding nonmelanoma skin cancer); had ever been told by a doctor or other health professional that they had lymphoma, leukemia, or blood cancer; or they had been told by a doctor or other health professional that they had chronic bronchitis or weak or failing kidneys during the preceding 12 months or had an asthma episode or attack during the preceding 12 months; or they were current smokers. HBV infection [3,5], many gay males and bisexual males have not been vaccinated. More gay men than bisexual men have been reported to disclose to their primary care providers about their sexual encounters with men [29], thus may provide healthcare providers more opportunity to recognize the need for and recommend HepA and HepB vaccination to gay men. This may help explain the observed higher HepA and HepB vaccination coverage in this population. Having self-identified as gay/lesbian orientation was associated with greater likelihood of influenza vaccination in our multivariable analysis, with bisexual adults having lower influenza coverage than gay/lesbian and heterosexual adults. That bisexual adults were less likely to have health insurance or a usual place for health care, and more likely to be below poverty in this study might be contributors to this finding. Because HepA, HepB, and HPV vaccinations protect against sexually transmitted infections, persons at higher risk of acquiring these infections because of risky sexual behaviors (e.g., MSM) might be expected to have higher coverage for those vaccines. Though HepA and HepB vaccination coverage among gay males and bisexual females was significantly higher than heterosexual males, HepA and HepB vaccination coverage among high-risk adults remains suboptimal and most remain at risk for HAV and HBV infection. As new HBV infections continue to occur with cumulative exposure over time and transmission among MSM continues during adulthood [49], until the vaccinated young adults age, which over time may increase the vaccine-induced population immunity, "catch-up" HepA and HepB vaccination campaigns among the unvaccinated, at-risk young adult population [3,5] might help improve coverage. Additional strategies to improve HepA and HepB vaccination among adults at increased risk might include: encouraging healthcare providers to identify candidates for HepA and HepB vaccination and ensuring that all adults at risk for HAV and HBV infection or who seek protection from HAV and HBV infection are offered HepA and HepB vaccines [3,5,50]; healthcare providers providing environments that facilitate both gay and bisexual men disclosing their sexual behaviors and other risk factors, possibly by posting non-discrimination statements in their clinics, encouraging openness in patient-provider discussions, using gender-neutral language about sexual partners, and discussing sexual health issues openly using nonjudgmental questions about sexual practices and behaviors [51][52][53]; and routine provision of HepB vaccine possibly at settings serving gender minorities, such as HIV counseling and testing sites that are able to provide vaccination.
This study used multivariable logistic model(s) to identify factors associated with vaccination. For HepA and HepB vaccinations, the findings from the multivariable models were driven mainly by males. A significantly higher percentage of gay men had received HepA and/ or HepB vaccinations than their heterosexual counterparts, possibly reflecting differential implementation of recommendations for vaccination of individuals at increased risk due to sexual behavior versus travel to countries in which HAV or HBV are endemic or having diabetes or chronic liver disease [2], differential recall of vaccination by the at risk population, as well as additional unmeasured confounding factor(s). A significantly higher percentage of gay/lesbian adults had received influenza vaccination than heterosexual adults, but not Tdap, pneumococcal or shingles vaccination. This observation is not well understood given the greater use of health services among gay/lesbian adults and the generally higher acceptability of vaccines in these groups [29]. In the multivariable models, however, no specific causal model or hypothesis was posed for the relationship between vaccination (for each specific vaccine) and the demographic, behavioral, health status, access to care characteristics, and sexual orientation considered in this report. Thus, although multiple factors were identified as having an independent association with a higher or lower likelihood of vaccination, the relative importance of these factors with respect to their association with vaccination cannot be determined from this analysis. The models were successful, however, in identifying factors generally associated with vaccination. Similar to other reports, characteristics associated with greater likelihood of vaccination included higher education, having health insurance, having had at least one or more physician contacts within the past year, and having a usual place to go for healthcare, even when controlling for sexual orientation [46,48].
Overall self-identified sexual orientation estimates from this report were similar to that reported from state-level and national surveys [17,18,54]. The estimates found in this report were lower compared with sexual orientation estimates reported elsewhere [55][56][57]. The differences in point estimates might be due to differences in sampling design, mode of the surveys, or other survey attributes. With the inclusion of questions on sexual orientation for the first time in the NHIS in 2013, it became possible to examine differences in self-reported receipt of selected vaccines by self-reported sexual orientation in a sample of U.S. adults from a national probability-based survey. Further, we also observed other important differences in respondent characteristics by sexual orientation, similar to those previously reported [1,2,[15][16][17][18][19]31,[52][53][54][58][59][60].
Several limitations should be considered in interpreting the results of this study. First, the information on vaccination was self-reported and may be subject to recall bias. However, adult self-reported vaccination status has been shown to be �70% sensitive in one or more studies for influenza, pneumococcal, tetanus toxoid-containing, herpes zoster, and HepB vaccines and �70% specific in one or more studies for all except tetanus and HepB vaccination [61][62][63][64][65]. Second, adult HepA and HepB vaccination are recommended for all unvaccinated adults at risk for HAV and HBV infection and for all adults requesting protection from HAV and HBV infection with the acknowledgment that a specific risk factor should not be a requirement for vaccination [3,5]. Because the NHIS does not collect information on all risk conditions for HAV and HBV infections, we were unable to identify all adults who were at increased risk for HAV and HBV infection and report vaccination estimates for these groups. Also, as the NHIS does not include measurement tools to assess whether or not prevention methods were used by the study sample while engaging in risky sexual behaviors, we could not conduct additional analysis for HepA and HepB vaccination by stratifying on adults self-identifying as gay/lesbian or bisexual that practiced sexual behaviors that put them at high risk for HAV and HBV infection. Third, the response rates for the three survey years in the report were 61.2%, 58.9%, and 55.2%. Nonresponse bias can result if respondents and nonrespondents differ in their vaccination rates and rates of other characteristics. Fourth, the differences observed in findings across studies could be a result of the differences in survey design, the mode of the surveys, question wording between NHIS and other data sources, and possible differential nonresponse by sexual orientation. Fifth, respondents with responses of "something else" (0.2%), "I don't know the answer" (0.5%), and refused (0.5%) for the sexual orientation questions were classified as missing and excluded from the analyses. Although a small proportion of the overall sample, these exclusions create a small potential for bias and prevent assessment of the health indices and vaccination status of less populous groups (e.g., persons who identify as transgender). Sixth, the Tdap estimate is subject to considerable uncertainty. Respondents who reported a tetanus vaccination but were unable to say whether Td or Tdap was used during 2005-2015 were excluded from estimations of Tdap coverage (36.7%), creating a potential for bias. Seventh, despite combining three years of data for a larger sample size, the number of adults selfidentifying as gay/lesbian, or bisexual was still relatively small when stratified across multiple covariate categories, which might cause some estimates to be unstable. The results pertaining to gay/lesbian or bisexual groups should be interpreted with caution. As additional years of data become available, more stable estimates could be generated. Eighth, the NHIS is a general population survey. The weighting is intended to represent the general population, not specifically populations defined by sexual orientation. Therefore, the estimates among gay, lesbian or bisexual respondents in this study might not be generalizable to the entire gay, lesbian or bisexual population [56,57]. Additionally, as noted, the sample size of persons in the NHIS who identified as gay, lesbian or bisexual is relatively small, which limited the ability to report coverage estimates by individual survey years. Ninth, as there are no unique personal identifiers in the NHIS that would allow determination of the proportion of respondents that might be survey repeaters across the years included in this study, it would be difficult to know if participants from each year were unique from other years. When the sociodemographic and healthrelated factors of the survey respondents from individual years (2013 vs. 2014, 2013 vs. 2015, 2014 vs. 2015 [data not shown]) were compared, similar patterns of similarity and differences were observed across the years, suggesting participants from each were unique from other years. Finally, though we controlled for potential confounders in this study, there might be unknown confounders that were not controlled for that might have biased the study estimates, the extent of which is hard to estimate.
Despite these limitations, study strengths include findings from a national probabilitybased survey that included noninstitutionalized U.S. adults who self-identified as gay/lesbian and bisexual that allowed national level vaccination coverage estimation by sexual orientation. This study is the first to assess associations of sexual orientation with a comprehensive list of vaccinations and could serve as a baseline for monitoring vaccination coverage changes over time.

Conclusions
This study helps document differences in self-reported vaccination coverage by self-identified sexual orientation among U.S. adults aged 18 years and over and is the first to assess associations of sexual orientation with a comprehensive list of vaccinations. Findings from this study can serve as a baseline for monitoring vaccination coverage changes over time, and can assist the development of targeted strategies to improve the health status of lesbian, gay, and bisexual populations. Annual data on sexual orientation from the NHIS can be useful in monitoring the impact of policies and interventions directed at improving the health of those who identify as gay, lesbian and bisexual.