Introduction

Members of sexual and gender minority (SGM) communities are less likely to access health care, have higher prevalence of chronic diseases, and subsequently worse health outcomes compared to the general population [1,2]. To improve the health of individuals in SGM communities, it is crucial to conduct further research to better understand the disparities in access, utilization, and outcomes. This knowledge can inform quality improvement initiatives and health policies, such as coverage of services used primarily by SGM communities or implementing payment models tied to reducing disparities for this community [3]. However, the lack of data that identifies members in SGM communities remains a grave limitation and further exacerbates the dearth of literature around the impact of health crises, such as COVID, on SGM populations [4].

For over a decade, the Federal government has issued reports and policies to encourage the standardized collection of sexual orientation and gender identity (SOGI) information from patients. In 2011, the Institute of Medicine and the Joint Commission highlighted the need to collect standardized SOGI data in electronic health records to better understand this population [2,5]. Subsequently, the Office of the National Coordinator for Health Information Technology required electronic health records to enable collection of SOGI to meet Meaningful Use Stage 3 criteria in 2015, and the Health Resources and Services Administration required Federally Qualified Health Centers to collect SOGI data from adult patients in 2016 [6]. Most recently, in 2022, the National Academies published recommendations for collecting and measuring sex, gender identity, and sexual orientation through standardized language [7]. Despite these efforts, and calls from providers, researchers, and advocates, there are currently no national requirements across providers and health care organizations to collect standardized patient SOGI information [8–10].

While research has shown that patients are willing to share this information, especially in outpatient settings, it is unknown if they would be willing to report SOGI data through digital health applications [11–14]. Digital health is a broad term that encompasses the use of technology in medicine and the healthcare field, and includes products and services such as mobile health applications, virtual health care delivery via telemedicine, wearables, and general wellness applications [15]. With the digital health market projected to grow, and specifically the increased use of virtual health care delivery since the pandemic, digital health applications may provide an effective means to collect SOGI data from a broad group of patients [16,17].

The purpose of this study was to determine the likelihood of adults to provide their sexual orientation and gender identity during the enrollment process of a digital health application. It also sought to test our hypotheses that SOGI response rates are lower among members in older age groups, members who reside in the southern and central regions of the US, and are not White [11,18–20].

Methods

This cross-sectional study utilized de-identified data from a digital health application registration and eligibility database, ensuring that the authors did not have access to any identifiable information about the members. As part of the registration process, each member agrees to our data privacy policy, which includes a provision that their de-identified data may be used for research purposes. The WIRB-Copernicus Group (WCG) IRB considered this study exempt from institutional review board approval in accordance with 45 CFR §46.102. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines (S1 File).

Optional questions during registration

During registration, the member was required to provide their name and a unique identifier, such as their birth date or health insurance member ID, to confirm eligibility for the application. If they chose to continue, they were presented with four optional fields to ’personalize their care’: race and ethnicity, pronouns, sexual orientation, and gender identity (Table 1). The SOGI questions and response options were informed by validated instruments and expanded for greater inclusivity [10,21,22].

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Optional questions and response options during digital health application registration.

https://doi.org/10.1371/journal.pone.0292739.t001

Results

Member demographics and responses to optional fields when registering for the digital health application

Overall, 113,064 new members registered during the study period and 41,677 (36.8%) chose to continue to the optional questions. Table 2 includes the responses to the questions and the member demographic characteristics among those presented with the optional questions.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Demographic characteristics of newly registered members who continued to optional questions and member responses to race and ethnicity, sexual orientation, gender identity, and pronoun questions during digital health application registration between September 9th and December 31, 2022.

https://doi.org/10.1371/journal.pone.0292739.t002

Over 67% of the members who were prompted to complete the optional fields were between the ages of 27 and 45 years old and almost half lived in the Pacific region of the United States.

Approximately 42% of members identified as White, 29% as Asian, 6% as Black or African American, 7% as Hispanic, 4% selected multiple races and ethnicities, and 1% responded ‘my race is not listed’. The race and ethnicity nonresponse rate was 11%; 5% chose ‘I prefer not to answer’ and 6% had a null response.

Fifty-one percent of members selected he/him/his pronouns, 39% chose she/her/hers, and 1.1% selected multiple responses. The nonresponse rate was 7.7%; 0.25% chose ‘I prefer not to answer’ and 7% had a null response.

Over 79% of members’ gender identity was cisgender, 1.2% of members’ gender was not listed, and 1.4% identified as transgender, agender, genderfluid or non-binary. The nonresponse rate was 18.7%; 5.3% chose ‘I prefer not to answer’ and 13.4% had a null response.

Almost 73% of members identified as heterosexual, 3.4% as lesbian or gay, 2.2% as bisexual, and 1.4% identified as either aromantic, asexual, pansexual, or queer. Over 18% did not provide their sexual orientation; 7.9% chose ‘I prefer not to answer’ and 10.6% had a null response.

We found that members who had a nonresponse to race and ethnicity were more likely to have a nonresponse to sexual orientation and gender identity (p<0.001). Similarly, members who had a nonresponse to sexual orientation were more likely to have a nonresponse to gender identity (p<0.001).

Differences in sexual orientation and gender identity nonresponse by member demographics

Table 3 displays nonresponse rates by member demographics and compares characteristics of sexual orientation and gender identity respondents to nonrespondents. Respondents who provided information about their sexual orientation and gender identity exhibited differences from nonrespondents in terms of race and ethnicity, age, and geographic region (p<0.001). SOGI nonresponse rates were higher among older members, members residing in the central and southern U.S regions, and varied across different races and ethnicities.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Characteristics of gender identity and sexual orientation respondents and nonrespondents, and nonresponse rates by characteristic between September 9th and December 31, 2022.

https://doi.org/10.1371/journal.pone.0292739.t003

Among members, sexual orientation nonresponse rates ranged from 16% (ages 27–35) to 26% (ages 56–64), while gender identity nonresponse rates ranged from 15% (ages 27–35) to 33% (ages 65+). Nonresponse rates for sexual orientation and gender identity were lowest in the Middle Atlantic (15%) and the Pacific (17% and 15%, respectively), but were higher in central and southern regions, ranging from 21–25% for sexual orientation and 25–29% for gender identity.

SOGI nonresponse rates differed across races and ethnicities. Members who selected multiple races and ethnicities had the lowest nonresponse rates for sexual orientation and gender identity at 9% and 10%, respectively. Among those who provided race and ethnicity information, American Indian or Alaska Native members had the highest nonresponse rates with 19% for sexual orientation and 27% for gender identity. For sexual orientation, Native Hawaiian or Other Pacific Islander members also had a high nonresponse rate of 19%.

Recognizing that motivations could vary between members who responded ’I prefer not to answer’ and members with null responses, we conducted a sensitivity test. This test excluded members who chose ’I prefer not to answer’ and considered only ’null responses’ as nonresponses. The results were consistent to those presented above; differences between respondents and nonrespondents remained statistically significant (S1 Table).

Discussion

To our knowledge, this is the first study to examine response rates to optional SOGI questions within a digital health application. Our study sample represents a diverse group of over 41,000 commercially-insured adults who used a digital health application for virtual care and care navigation services. Adults were located in all 50 states with less than half identifying as White. We found that almost 82% of this diverse group of adults were willing to share their sexual orientation and gender identity during the registration process. These results contribute to the growing body of literature that adults are willing to share their sexual orientation and gender identity with health care entities, including through a digital health application.

We found that the proportion who chose ‘I prefer not to answer’, 7.9% for sexual orientation and 5.3% for gender identity, were within the range of other self-administered surveys. For example, one study found that 12% of patients at urology or cancer clinics refused to answer questions about their sexual orientation, and 7% refused to answer questions about their gender identity [12]. Another study of 293 patients at outpatient clinics had a 5% refusal rate for sexual orientation and 0.3% for gender identity [11]. While our findings are consistent with these studies in terms of nonresponse rates, the magnitude of nonresponse was closer between the two questions in our study. Drivers of this difference are unknown, but two possible explanations may be differences in study samples and study time periods. For example, our study sample was much larger and more diverse in terms of demographic characteristics. Additionally, societal attitudes towards sexual orientation and gender identity have changed over time, which may have contributed to similar rates of nonresponse in our study compared to earlier studies.

Our study found that SOGI nonresponse rates varied by demographic characteristics, similar to previous research. We discovered that gender identity refusal rates increased with age, aligning with Rullo et al.’s findings [11]. Moreover, we observed higher nonresponse rates among individuals residing in states where legislators are approving or introducing anti-LGBTQ+ bills [24]. This finding is consistent with a prior study, which found that SOGI data collection was higher in cities with protective SOGI nondiscrimination policies [18]. The reasons behind the higher nonresponse rates among people residing in areas of the country where state legislators are introducing or approving anti-LGBTQ+ bills are unclear. It may be due to fear of self-identification among members of SGM communities or antagonism towards the SGM community. In our data, the proportion of respondents who identified as a member of a SGM community was similar across geographies, indicating fear of self-identification may not differ by geography. However, further research is necessary to fully understand the underlying factors.

Our study also identified that nonresponse rates varied by race and ethnicity, with the highest nonresponse rates among Black, Native Hawaiian or Other Pacific Islander, and American Indian or Alaska Native adults. White adults had lower nonresponse rates for sexual orientation, but higher nonresponse rates for gender identity compared to Asian adults and those whose race was not listed. Although no studies have compared SOGI nonresponse rates by race and ethnicity from self-administered surveys, unadjusted results from two studies based on the SOGI module of the Behavioral Risk Factor Surveillance System (BRFSS) survey produced mixed findings. One study, based on seven states, reported that non-Hispanic White adults had the lowest gender identity nonresponse rate, but adults of multiple races or non-Hispanic other race had lower nonresponse rates than White adults for sexual orientation [19]. The other study, based only on Washington BRFSS, found that Hispanic and Asian adults had the highest sexual orientation nonresponse rates, whereas non-Hispanic White adults had the lowest [20].

This study is not without limitations. First, results lack generalizability beyond commercially-insured adults who used a digital health application provided through their employer. Additionally, the response rates and distributions may differ if these questions were asked of all members who registered instead of only those who opted to continue to additional questions. The study is also limited by the lack of complete demographic data on the eligible population, which limits our understanding of differences between respondents and nonrespondents. Another limitation is that while the SOGI questions used in the study were informed by validated instruments, the two-step gender identity question was not used. Instead, we asked about current gender identity to reduce member friction, which may have increased comprehension issues with the response options presented. However, we included definitions of each response option within the application to increase member understanding. Finally, the expansive and ever-changing nature of SOGI categories may have increased nonresponse among members whose identity is not included. However, the option to select ‘my identity/gender is not listed’ likely minimized this limitation.

To increase the collection of SOGI data in digital applications, survey structure is a crucial consideration. We identified three potential changes for future consideration and research. First, the survey should be presented to as many members as possible to reduce the drop in responsiveness due to flow design. We found that 63% of potential respondents were lost before they even saw the questions. Second, including a ‘don’t know/not sure’ field provides a response option for members who may not understand the options presented. Finally, to counteract concerns around confidentiality and data use highlighted in the literature, the digital application could prominently display the purpose and secure storage of the data being collected [12,25].

Conclusion

This study highlights the potential of digital health applications to collect SOGI data from a diverse group of commercially-insured adults. Further research is needed to identify strategies to increase SOGI response rates and understand the underlying drivers of nonresponse. As digital health applications become more prevalent, they can be leveraged as a tool to collect critical SOGI data for delivering equitable care and enabling future research and improvement initiatives.

Supporting information

Acknowledgments

Alani Douglas for providing guidance on data access and interpretation and for feedback on the manuscript; Colin Yee for assistance with data analysis; and Ami Parekh, Vibin Roy, and Todd Thames for review and feedback on the manuscript.