The intersection of sexual orientation with race and ethnicity in cervical cancer screening

Abstract

BACKGROUND:

Cervical cancer screening is recommended for those with a cervix who are 21 to 65 years old, with specific timelines being dependent on individual risk. This study compared rates of ever undergoing Papanicolaou (Pap) testing at the intersection of self-reported sexual minority (SM) status and race/ethnicity.

METHODS:

Data from the National Health Interview Survey (2015 and 2018) were used to examine cervical cancer screening disparities. Natal females without a history of hysterectomy who were 21 to 65 years old and had reported their sexual orientation and Pap testing history were included. Demographic and health characteristics were summarized with descriptive statistics. To adjust for differences in confounding variables between groups, propensity score–based inverse probability of treatment weighting (IPTW) was performed. IPTW-adjusted multivariable logistic regression models estimated odds of ever undergoing a Pap test by sexual orientation alone and with race/ethnicity (non-Hispanic White, non-Hispanic Black, and Hispanic).

RESULTS:

SM persons (n = 877) had significantly reduced odds of ever undergoing Pap testing (odds ratio, 0.54; 95% confidence interval, 0.42-0.70) in comparison with heterosexual persons (n = 17,760). When the intersection of sexual orientation and race/ethnicity was considered, non-Hispanic White SM participants and Hispanic SM participants had reduced odds of ever undergoing Pap testing in comparison with non-Hispanic White heterosexual participants. No significant differences were observed between non-Hispanic White heterosexual participants and participants of non-Hispanic Black SM or Hispanic heterosexual identities.

CONCLUSIONS:

SM participants were significantly less likely to have ever undergone a Pap test in comparison with heterosexual participants, with Hispanic SM participants having the lowest uptake. Future studies should further examine the roles of systemic discrimination and other key drivers of these disparities.

INTRODUCTION

Within the year 2021, the United States anticipates approximately 14,480 new cases of invasive cervical cancer and 4290 related deaths.¹ Screening recommendations are in place to detect premalignant lesions or early-stage cancers and to decrease disease-related morbidity and mortality.² Evidence in the literature, albeit limited, has suggested that sexual minority (SM) individuals (ie, those whose sexual orientation differs from societal norms, including but not limited to those identifying as gay, lesbian, queer, bisexual, pansexual, and beyond³) may be at increased risk for developing gynecologic cancers while also being less likely to undergo cancer screening.^4–14

Barriers to cervical cancer screening among SM people include inadequate patient-provider communication, mistrust in medical providers,¹⁵ fear of discrimination in the clinic, and the belief that Papanicolaou (Pap) testing is not beneficial.^14,16 Additionally, differences in hormonal contraceptive use,¹⁴ the absence of a regular medical provider, gynecology referrals, and cervical cancer screening recommendations may play a substantial role in screening participation.¹⁷ Although general cervical cancer screening disparities among SM persons have been identified,^{5–9,12–14,18} it is not yet known how utilization among subgroups (eg, by race in combination with belonging to an SM group) compares with utilization among White heterosexual individuals with a cervix. One previous study by Agénor and colleagues⁵ assessed disparities in the reporting of Pap testing within 12 months of the survey at the intersection of race/ethnicity and sexual orientation (based on partner sex) among natal females aged 21 to 44 years, and they found significant interactions between the intersecting identities of race/ethnicity and sexual orientation.⁵

Intersectionality theory provides an analytical framework that acknowledges an individual’s multiple social and political identities and examines how power differentials between identities may produce or exacerbate experiences of both privilege and discrimination.^19,20 The application of intersectionality theory enables us to examine discrimination at a deeper level that more accurately captures the experiences of SM patients with multiple oppressed identities. Disparities in cancer screening, incidence, and outcomes by race/ethnicity have been well documented in the past.^21–27 If we consider race alone, members of the Black community face institutional discrimination in comparison with White patients, and this includes not receiving the same clinician referrals, being less likely to be prescribed surgical and therapeutic interventions, and being less likely to have screening recommended by their providers.²⁸ Although limited, studies suggest that discriminatory experiences among people with marginalized intersecting identities may lead to avoidance related to cancer screening.^5,29,30 Thus, more research on systemic discrimination and cancer disparities at the intersection of multiple social constructs (ie, SM identity, race, and social class) is necessary to reduce disparities in cervical cancer screening and prevention.

We sought to expand on the study by Agénor and colleagues⁵ in a US population more inclusive of those eligible for cervical cancer screening per current guidelines³¹ (age range, 21-65 years) to examine cervical cancer screening disparities among people with a cervix who identified as an SM (lesbian/gay, bisexual, or something else/unsure). Specifically, we aimed to examine whether disparities were accentuated among people at particular intersections of sexual orientation and racial/ethnic backgrounds (non-Hispanic White, non-Hispanic Black, or Hispanic any race) while accounting for individual socioeconomic status and other important confounding variables.

MATERIALS AND METHODS

Data Acquisition and Study Population

We used data from the 2015 and 2018 editions of the US National Health Interview Survey (NHIS), which was selected because of its consistency in questions relating to Pap testing and cervical cancer screening as well as sexual orientation. The NHIS is a cross-sectional survey program that collects data on a broad range of health topics through personal household interviews and is representative of the noninstitutionalized civilian population of the United States, with oversampling of persons identifying as non-Hispanic Black, Hispanic, and Asian.³² Interviews for the NHIS are conducted by the US Census Bureau and are available in both English and Spanish.³³ Starting in 2013, the NHIS included a question about sexual orientation with the following options: lesbian or gay; straight, that is, not lesbian or gay; bisexual; something else; and I don’t know the answer (hereby referred to as unsure). All data were downloaded directly from the Centers for Disease Control and Prevention NHIS website.³⁴

The data for this analysis were available from the adult in the household selected to participate in the Sample Adult and Sample Adult Cancer questionnaires. Variables for demographics were also incorporated from the Family and Person Files. The data set was further refined to include those who reported female for assigned sex, had a cervix at the time of the survey response (ie, no history of hysterectomy), and were 21 to 65 years old. This study was reviewed and deemed exempt from IRB oversight by the University of Minnesota IRB.

Study Measures

The outcome of interest in this study was self-report of undergoing Pap testing (ever or never). The exposure was self-reported sexual orientation with the following categories: heterosexual, lesbian/gay, bisexual, and something else/unsure. Because of sample size constraints, these categories were then dichotomized into heterosexual and SM participants. Demographic characteristics of interest for this analysis included the following: age at the time of the survey (years), race and ethnicity (non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, Hispanic of any race, or non-Hispanic of another race), partner status (living with a partner or not living with a partner), annual household income ($0-$34,999, $35,000-$74,999, $75,000-$99,999, ≥$100,000, or unknown), employment status (not employed for pay or employed), education (high school or less, some college, college degree, graduate school, or unknown), place of birth (outside the United States or in the United States), perceived health status (excellent/very good/good, or fair/poor), and health insurance status (having health insurance or not having health insurance).

Statistical Analysis

Distributions of demographic and health/medical-related characteristics were compared between participants identifying as SMs and those identifying as heterosexuals via χ² tests for categorical variables and Wilcoxon rank-sum tests for continuous variables.

Univariate and age-adjusted logistic regression models were used to examine the odds of reporting ever undergoing Pap testing among SM participants versus those who identified as heterosexual. We similarly calculated the odds of ever undergoing Pap testing by specific sexual orientations (lesbian or gay, bisexual, and something else/unsure). To examine the intersection of sexual orientation and race/ethnicity, logistic regression models were used to estimate the odds of ever undergoing Pap testing with the following exposure categories: non-Hispanic White heterosexual (reference group), non-Hispanic Black heterosexual, Hispanic heterosexual, non-Hispanic White SM, non-Hispanic Black SM, and Hispanic SM. Because of the small sample size of those identifying as both an SM and non-Hispanic Asian (n = 30), this group was not included in the intersectional models. To account for the potential selection bias and confounding, we included NHIS sampling and confounding weights in our fully adjusted models as described next.

NHIS Sampling Weights

We incorporated sampling weights into all statistical models to account for the complex multistage probability sampling scheme of the NHIS.³⁵ These weights were used to account for potential selection biases and provide estimates more generalizable to the US population.

Confounding Weights

To account for residual confounding due to demographic differences between comparison groups, we calculated confounding weights by using an inverse probability of treatment weighting model to balance measured covariates and reduce bias in estimates.³⁶ We estimated the propensity scores—the probability of being an SM (or particular sexual orientation and race) given an individual’s characteristics—by using a binary (multinomial) logistic regression model and computed the weight by using the inverse propensity score. The variables used to calculate the propensity scores were age, partner status, education, health insurance status, and annual household income, with race/ethnicity added to the model looking at sexual orientation alone. Weights for each individual were calculated as the inverse of the propensity score and were used in the final regression models to create a pseudopopulation in which measured confounders were balanced across groups. Because of the large differences in the sample sizes of the groups, weights were stabilized. We assessed the weighted distributions of these variables to confirm balance.

Combined Sampling-Confounding Weights

The inverse probability weights were multiplied by the NHIS survey weights and included in the final, fully adjusted logistic regression models.

All analyses were conducted with SAS 9.4 (SAS, Cary, North Carolina). Odds ratios (ORs) and 95% confidence intervals (CIs) are reported unless otherwise noted, and P values < .05 were considered statistically significant.

RESULTS

There were 20,014 NHIS participants in the years 2015 and 2018 who were assigned female at birth, were without a history of hysterectomy, and were between the ages of 21 and 65 years (Supporting Fig. 1). In our eligible cohort, 666 (3.3%) did not report sexual orientation, and a further 711 (3.6%) did not report their Pap testing history (ever or never); this left a final analysis sample size of 18,637. When we compared those who reported sexual orientation (n = 19,348) to those who did not (n = 666), those with an unknown sexual orientation were older (45.7 vs 39.9 years), were more likely to identify as non-Hispanic Black (17.3% vs 12.9%), were less likely to identify as Hispanic (13.0% vs 17.9%), were more likely to report being in fair or poor health (16.1% vs 10.5%), and were less likely to live with a partner (51.7% vs 63.6%; Supporting Table 1). Those with an unknown sexual orientation were also more likely to have other missing data in comparison with those reporting their sexual orientation, such as an unknown annual household income (19.4% vs 8.9%) and an unknown Pap testing history (92.3% vs 5.2%). When we compared those with complete Pap testing data (n = 18,637) to those with an unknown Pap testing history (n = 711), those with an unknown history were slightly younger (39.9 vs 40.2 years) and were more likely to be heterosexual (96.2% vs 95.1%) or report their sexuality as “something else/unsure” (2.5% vs 1.2%), to identify as non-Hispanic Black (18.5% vs 14.0%), and to report an annual household income of less than $35,000 per year (34.5% vs 30.8%; Supporting Table 2). The groups were otherwise similar in demographic characteristics.

Distributions of demographic characteristics differed by intersectional groups (Table 1). Those who identified as an SM were younger than heterosexual individuals, with those identifying as both Hispanic and SM being the youngest on average. Non-Hispanic Black participants were least likely to be living with a partner at the time of the survey and most likely to report an annual household income of less than $35,000 in comparison with the other intersectional groups. Hispanic participants of any sexual orientation were most likely to report less education, be born outside the United States, and not have health insurance.

TABLE 1.

Distribution of Characteristics by Intersecting Racial/Ethnic Identity and Sexual Orientation Status Among 2015 and 2018 NHIS Participants

Characteristic	Non-Hispanic White Heterosexual (n = 10,471)	Non-Hispanic Black Heterosexual (n = 2419	Hispanic Heterosexual (n = 3259)	Non-Hispanic White SM (n = 533)a	Non-Hispanic Black SM (n = 116)a	Hispanic SM (n = 158)a	P b
Age at survey, median (95% CI), y	42.3 (41.7-42.9)	37.9 (37.0-38.8)	37.3 (36.5-38.0)	33.9 (32.5-35.2)	33.0 (30.2-35.7)	31.1 (28.6-33.7)	<.0001
Living with a partner, %							<.0001
No	40.6	74.9	46.5	60.3	80.2	63.9
Yes	59.4	25.1	53.5	39.7	19.8	36.1
Annual household income, %							<.0001
$0-$34,999	23.4	49.3	41.1	38.5	59.0	50.0
$35,000-$74,999	27.0	26.3	29.4	28.4	21.2	30.6
$75,000-$99,999	13.1	7.9	7.9	11.1	8.5	6.8
≥$100,000	30.5	9.4	13.1	16.8	6.1	6.9
Unknown	6.1	7.1	8.5	5.2	5.3	5.6
Employment status, %							<.0001
Not employed for pay	30.7	32.9	37.7	28.9	36.3	38.0
Employed	69.3	67.1	62.3	71.1	63.7	62.0
Education, %							<.0001
High school or less	23.0	35.2	51.0	23.6	37.0	51.0
Some college	31.4	37.5	27.7	31.3	41.2	26.5
College degree	28.3	17.8	15.4	28.0	14.9	16.3
Graduate school	17.3	9.5	5.9	17.0	6.9	6.2
Place of birth, %							<.0001
Outside the United States	5.8	13.7	56.1	1.8	7.6	47.3
In the United States	94.2	86.3	43.9	98.2	92.4	52.7
Perceived health, %							<.0001
Excellent/good/very good	91.5	84.3	87.3	87.3	82.0	80.6
Fair/poor	8.5	15.7	12.7	12.7	18.0	19.4
Health insurance status, %							<.0001
Has health insurance	92.5	88.4	74.5	89.6	86.5	74.7
Does not have health insurance	7.5	11.6	25.5	10.4	13.5	25.3

Abbreviations: CI, confidence interval; NHIS, National Health Interview Survey; SM, sexual minority.

^aSelf-reported SM was defined as lesbian, gay, bisexual, something else, or unsure of one’s sexual orientation.

^bP values were calculated with χ² tests for categorical variables and with Wilcoxon rank-sum tests for continuous variables with survey weights.

When we examined cervical cancer screening disparities by sexual orientation alone, those identifying as an SM had reduced odds of ever undergoing a Pap test in comparison with heterosexual counterparts after adjustments for confounding (OR, 0.54; 95% CI, 0.42-0.70; Table 2). A further breakdown by specific SM identities had similar findings across SM groups, but only those identifying as “something else/unsure” had statistically significant reduced odds of ever undergoing Pap testing in comparison with heterosexual participants (OR, 0.29; 95% CI, 0.19-0.45).

TABLE 2.

Association of Ever Undergoing a Pap Test With Sexual Orientation and Intersectional Identities of Sexual Orientation and Race/Ethnicity

	Sexual Orientation Alone and Odds of Ever Undergoing Pap Testa
	No.	Unadjusted Odds Ratio (95% CI)b	Age-Adjusted Odds Ratio (95% CI)b	Fully Adjusted Odds Ratio (95% CI)c
Heterosexual	17,760	Reference	Reference	Reference
SM	877	0.41 (0.32-0.52)	0.53 (0.42-0.68)	0.54 (0.42-0.70)
	Sexual Orientation Alone and Odds of Ever Undergoing Pap Test by SM-Specific Identity
	No.	Unadjusted Odds Ratio (95% CI)b	Age-Adjusted Odds Ratio (95% CI)b	Fully Adjusted Odds Ratio (95% CI)c
Heterosexual	17,760	Reference	Reference	Reference
Lesbian/gay	313	0.43 (0.30-0.64)	0.44 (0.30-0.66)	0.68 (0.43-1.07)
Bisexual	342	0.59 (0.39-0.90)	0.98 (0.64-1.50)	0.73 (0.47-1.15)
Something else/unsure	222	0.24 (0.16-0.35)	0.26 (0.18-0.39)	0.29 (0.19-0.45)
	Sexual Orientation With Race/Ethnicity and Odds of Ever Undergoing Pap Test
	No.	Unadjusted Odds Ratio (95% CI)b	Age-Adjusted Odds Ratio (95% CI)b	Fully Adjusted Odds Ratio (95% CI)c
Non-Hispanic White heterosexual	10,471	Reference	Reference	Reference
Non-Hispanic Black heterosexual	2419	0.77 (0.60-1.00)	0.89 (0.69-1.15)	1.32 (1.01-1.73)
Hispanic heterosexual	3259	0.44 (0.35-0.53)	0.51 (0.41-0.63)	0.84 (0.68-1.04)
Non-Hispanic White SM	533	0.34 (0.24-0.49)	0.46 (0.32-0.66)	0.62 (0.42-0.91)
Non-Hispanic Black SM	116	0.46 (0.21-1.05)	0.64 (0.28-1.50)	0.64 (0.25-1.60)
Hispanic SM	158	0.15 (0.10-0.24)	0.22 (0.14-0.35)	0.26 (0.15-0.43)

Abbreviations: CI, confidence interval; NHIS, National Health Interview Survey; Pap, Papanicolaou; SM, sexual minority.

^aSelf-reported SM was defined as lesbian, gay, bisexual, something else, or unsure of one’s sexual orientation.

^bIncorporates NHIS sampling weights.

^cThe fully adjusted model incorporates NHIS sampling weights and confounding weights calculated with inverse probability of treatment models that account for age, partner status, education, health insurance status, and annual household income, with race/ethnicity added to the model looking at sexual orientation alone.

In the fully adjusted intersectional model, those identifying as both Hispanic and SM had the lowest odds of ever undergoing Pap testing in comparison with non-Hispanic White heterosexual participants (OR, 0.26; 95% CI, 0.15-0.43). Additionally, those identifying as non-Hispanic White and SM (OR, 0.62; 95% CI, 0.42-0.91) had reduced odds of ever undergoing Pap testing in comparison with non-Hispanic White heterosexual counterparts. When making comparisons with non-Hispanic White heterosexual individuals, we did not observe significant differences in the odds of undergoing Pap testing among those identifying as Hispanic and heterosexual (OR, 0.84; 95% CI, 0.68-1.04) or those identifying as non-Hispanic Black and SM (OR, 0.64; 95% CI, 0.25-1.60). However, those identifying as non-Hispanic Black and heterosexual were significantly more likely to have ever undergone Pap testing than non-Hispanic White heterosexual counterparts (OR, 1.32; 95% CI, 1.01-1.73).

DISCUSSION

Our findings show that those identifying as an SM had reduced odds of ever undergoing Pap testing in comparison with heterosexual individuals. This finding was consistent across specific SM identities, and although sample size limitations prevented us from formally testing differences between specific SM identities, these results point to the heterogeneity in screening behaviors and health care access among people of different SM identities.^18,37,38

Additionally, when we examined these disparities across race and ethnicity, nearly all SM persons (non-Hispanic White and Hispanic) had reduced odds of undergoing Pap testing in comparison with non-Hispanic White heterosexual participants. Taken together, these findings suggest that a large majority of SM persons are less likely to undergo Pap testing for cervical cancer regardless of race and ethnicity.

A recent study by Porsch et al⁶ examined cervical cancer screening disparities among SM women. Similarly to our findings, when identifying SM participants by the anatomy of their partners, they found that those identified as an SM were less likely to be current on cervical cancer screening. Work by Charlton et al has demonstrated that SM women underuse cervical cancer screening services,¹² and these disparities may partly be due to differing health beliefs by sexual orientation¹⁴ and differences in hormonal contraceptive use resulting in gynecologic health resource underutilization. A study by Agénor and colleagues⁵ using the National Survey of Family Growth examined sexual orientation alone and the interaction with race and ethnicity to assess disparities in cervical cancer screening. They found that women categorized as an SM on the basis of their partner’s sex were at reduced odds of undergoing Pap testing, and those who identified as Black might have lower Pap testing uptake. Our study expands on this work: We compared the odds of ever undergoing Pap testing between specific populations and expanded the age group to better represent current recommendations for those eligible for Pap testing (age range, 21-65 years). Rather than modeling intersecting identities as statistical interactions between race/ethnicity and sexual orientation, we modeled these intersectional identities by using unique SM/race/ethnicity categories within 1 statistical model. Our results add to the previous work in that we found significant disparities between those identifying as Hispanic SM individuals and those identifying as non-Hispanic White heterosexual individuals, whereas their analyses (stratified by race/ethnicity after interaction effects were found between sexual orientation and race/ethnicity) did not allow for direct comparisons across ethnoracial groups. Thus, our results bridge a gap in the literature to understand which specific groups are most affected by these disparities.

Our results did not indicate a significant difference in the odds of Pap testing between non-Hispanic Black SM participants and non-Hispanic White heterosexual participants. Furthermore, our results showed that non-Hispanic Black heterosexual participants had greater odds of ever undergoing Pap testing in comparison with non-Hispanic White heterosexual participants. In an article by Garner,³⁹ it is suggested that income and education are most likely to drive cervical cancer screening disparities. Here, we accounted for both of these factors in our analysis, and this may explain the lack of differences in Pap testing between non-Hispanic Black participants and White counterparts. Similarly to our findings, a recent study using 2010 NHIS data that examined differences in cervical cancer screening by race alone among Black and White survey participants found that Black participants reported greater odds of undergoing Pap testing.⁴⁰ However, they did identify disparities in the receipt of follow-up recommendations and care and highlighted areas of systemic racism relating to inadequate cervical cancer prevention despite higher odds of initial screening. Therefore, needs regarding research and clinical models incorporating the intersection of race and ethnicity with sexual orientation remain for this population.

Our study has many strengths, including a large sample size of individuals identifying as an SM, the ability to perform intersectional analyses while adjusting for important confounding variables, and the use of data derived from a nationally representative sample. However, important limitations should also be noted. The inverse probability of treatment weighting approach assumes that all confounding has been adjusted for in the model; although we believe that we have adjusted for all measured confounders, it is unlikely that we have fully accounted for differences between groups. We conducted the analysis via a complete case approach and included only individuals with both sexual orientation and Pap history reported because the amount of missingness was low; however, a nonresponse bias is still possible. Although we were able to identify broad disparities by race and Hispanic ethnicity, we were unable to further examine differences in Pap testing by specific groups within these categories (eg, country of Latin/Hispanic origin) or other racial and ethnic populations because of the small sample size. Similarly, because of data limitations, we had to aggregate all SM identities to assess disparities at the intersection of sexual orientation and race/ethnicity, and we were unable to assess disparities by gender identity because the NHIS does not collect this information. Continued research in this area will be necessary with larger sample sizes to continue to examine and identify underserved populations. We were unable to look specifically at structural oppression, and we instead examined participants at the intersections of their socially constructed identities. More research is necessary to measure the structural oppression, including discrimination, experienced by those with multiple marginalized identities.

In conclusion, our study has demonstrated that most persons with a cervix who identify as an SM, regardless of race and ethnicity, are at reduced odds of ever undergoing any Pap testing in comparison with White heterosexual counterparts, even when adjustments are made for demographic characteristics. Future work should examine disparities in Pap testing through an intersectional lens to not only identify groups at greater risk for not receiving such care but also examine systemic discrimination and identify where the system may be failing, potential interventions in clinical care at the institutional level, and opportunities for community outreach.