Effects of Patient Sexual Orientation and Gender Identity on Medical Students’ Decision-Making Regarding Pre-Exposure Prophylaxis (PrEP) for HIV Prevention: A Vignette-Based Study

Abstract

Introduction:

Pre-exposure prophylaxis (PrEP) is a pillar of our national strategy to end the HIV epidemic. However, one of the largest obstacles to realizing the effectiveness of PrEP is expansion of prescription to all patients at risk for HIV. In this vignette-based study, we sought to investigate medical students’ decision-making regarding PrEP by presenting fictional patients, all of whom had HIV risk-factors based on sexual behavior.

Methods:

We systematically varied patients’ sexual orientation or gender identity (heterosexual female, gay male, bisexual male, transgender male, transgender female, gender nonbinary person). We assessed the medical students’ willingness to prescribe PrEP to the patients as well as their perceptions of the patients’ HIV-risk and behavior.

Results:

A total of 670 U.S. medical students completed the study. The heterosexual female patient was least frequently identified as a PrEP candidate, was viewed as least likely to adhere to PrEP, and the most likely to engage in condomless sex if prescribed PrEP; however, was considered at lower overall HIV-risk. Lower perceived HIV-risk and anticipated PrEP adherence were both associated with lower willingness to prescribe PrEP. Willingness to prescribe PrEP was highest for the gay male patient and lowest for the heterosexual female.

Conclusion:

These analyses suggest that assumptions about epidemiological risk based on patients’ gender identity or sexual orientation may reduce willingness to prescribe PrEP to heterosexual women, ultimately hindering uptake in this critical population.

Summary:

In this study of PrEP decision-making, we found that medical students were less likely to prescribe PrEP to heterosexual women and transgender people relative to gay men, despite identical HIV risk-factors, due to perceived sexual risk-taking.

Introduction

Over 30,000 new diagnoses of HIV are reported annually in the United States.¹ Men who have sex with men (MSM) experience a disproportionate burden of HIV, accounting for approximately two-thirds of new diagnoses.¹ Another quarter of diagnoses are attributed to heterosexual contact.¹ Transgender women are also significantly affected by HIV, with an estimated 40% prevalence in this community.² Daily, oral Pre-Exposure Prophylaxis (PrEP) with tenofovir/emtricitabine (TDF/FTC) has the potential to significantly reduce HIV incidence in the U.S.^3,4 When taken daily, TDF/FTC as PrEP achieves up to 99% efficacy in preventing HIV.^3,5 The U.S. Preventive Services Task Force and Centers for Disease Control and Prevention recommend PrEP for all people at risk for HIV, including MSM, transgender people, and heterosexual people with HIV-risk-factors.^4,6

Despite these recommendations, prescription of PrEP in the U.S. has not reached all patients at risk for HIV.^7–10 Presently, MSM make up a majority of current PrEP users in the U.S.⁹ While HIV incidence has decreased among MSM since the 2012 Food and Drug Administration approval of TDF/FTC for PrEP, a large percentage of MSM still are not prescribed PrEP.¹¹ Furthermore, compared to gay men, bisexual men are less likely to use PrEP, even though bisexual men are also at increased risk for HIV.^12–15 In addition, transgender people are also under-prescribed PrEP despite high HIV-risk, and a similar phenomenon is present among heterosexual, cisgender women.^2,11,16–19 A number of barriers to wider PrEP use have been reported, including clinician knowledge of PrEP, stigma regarding PrEP, and failure of clinicians to identify which patients are PrEP candidates.^20–22

A recent review found that a patients’ anticipated adherence to PrEP was a factor influencing clinicians’ willingness to prescribe PrEP.²⁰ Fears of non-adherence to PrEP are part of broader concerns about PrEP leading to increases in sexual risk-taking, or so-called ‘risk compensation,’ based on lower HIV risk among patients taking PrEP.²³ Risk-taking may include an increase in number of sexual partners or an increase in condomless sex.^24,25 Prior research focused on concerns of risk compensation and intentions to prescribe PrEP has largely focused on MSM patients.^20,26,27 However, these assumptions may affect PrEP decision-making for a broader range of patients, creating additional barriers to PrEP prescription.

Addressing these disparities is a public health priority. As such, medical education must prepare future physicians with the knowledge and skills to counsel all patients at risk for HIV about PrEP. However, gaps in medical education about PrEP and HIV-risk have been identified.^28,29 Furthermore, previous studies have found that medical training is lacking in curriculum related to transgender health in general, highlighting the need to understand medical students’ attitudes and behaviors toward transgender patients to improve training about HIV prevention.^30−33s

Previous work has investigated the role of patient race in medical students’ decision-making regarding PrEP prescription and their assumptions about patients’ sexual behavior, including increased sexual risk-taking if prescribed PrEP.^26,27 These vignette-based studies have focused exclusively on MSM, finding that medical students were less willing to prescribe PrEP to patients they assumed would increase frequency of condomless sex if prescribed, and that explicit heterosexist attitudes predicted lower willingness to prescribe PrEP.^26,27 To the best of our knowledge, no previous studies have investigated differences in willingness to prescribe PrEP to patients of different sexual orientations or gender identities. This represents a critical gap in our understanding of PrEP decision-making among diverse patient populations.

The present study used a novel, vignette-based approach to investigate the roles of patient sexual orientation or gender identity in medical students’ decision-making regarding PrEP for HIV prevention. Specifically, we examined: 1) whether their perceptions of patients’ HIV-risk and related behaviors and willingness to prescribe PrEP differed by patient sexual orientation or gender identity; 2) whether their demographic characteristics and attitudes towards lesbian, gay, bisexual, and transgender (LGBT) patients influenced their perceptions of patients’ HIV-risk and related behavior or willingness to prescribe; 3) whether there were indirect effects of patient sexual orientation and gender identity on medical students’ willingness to prescribe PrEP mediated by perceptions of patients’ HIV-risk and related behavior; and 4) whether confidence caring for the presented patient differed between conditions.

Methods

Participants and Procedure

An online study instrument was distributed to a national sample of medical students in the U.S between February and May 2020. A total of 12 medical schools (10 allopathic, 2 osteopathic) participated, and administrators from these schools distributed the study link to their student email lists with a single reminder one week after the initial message. The study was administered via QualtricsXM (Provo, UT). After completing the study, participants were given the option of entering into a drawing for multiple $25 Amazon.com gift cards. Participants were sent a debriefing message when data collection was complete.

Instrument Development

The study instrument consisted of a vignette describing a fictional patient presenting to a primary care physician requesting PrEP for HIV prevention. The patient reported having multiple sexual partners of unknown HIV-status over the past year with intermittent condom use. The patient’s sexual orientation or gender identity was systematically varied between six experimental conditions (heterosexual female, gay male, bisexual male, transgender male, transgender female, gender nonbinary person). All other vignette details were identical across conditions. Participants were randomly assigned to one condition using the Qualtrics® randomization program which evenly distributed participants to the six experimental conditions utilizing a simple randomization method. The full text of the vignette is in Box 1.

“The patient is a 23-year-old [heterosexual/gay/bisexual/ transgender/gender nonbinary] [male/female/person] presenting to the primary care physician for a yearly physical. The patient is in good health. Family, medical, and surgical history are non-contributory. The patient reports social alcohol use, occasionally drinking 2–3 beers with friends on the weekends. [He/she/they] denies[y] illicit drug use and tobacco use. The patient reports having had 4 sexual partners of unknown HIV status in the past year, with intermittent condom use. [He/She/They] mention[s] seeing a commercial for PrEP and ask[s] if this would be something [he/she/they] should consider for protection from HIV. The patient takes no medications other than a multivitamin. Blood work ordered before this appointment is unremarkable, including appropriate kidney function, electrolyte balances, and CBC values that are within normal limits and an HIV test was non-reactive. The patient has health insurance through [his/her/their] school.”

Vignette-Related Items

After reading the vignette, participants completed a series of follow-up items assessing decision-making regarding PrEP prescription, confidence caring for the patient, and attitudes towards LGBT people. Items were modeled on previous similar studies.^26,27,34s First, participants were asked if the presented patient was an HIV PrEP candidate (yes/no). Next, participants were asked: 1) “How likely would you be to prescribe PrEP to this patient?”; 2) “If prescribed PrEP, how likely would the patient be to have condomless sex?”; 3) “If prescribed PrEP, how likely would the patient be to adhere to the regimen?”; and 4) “How high would you rate this patient’s risk of contracting HIV?” All items were rated on a 10-point Likert scale (1 = lowest, 10 = highest). Then, participants were asked to indicate their confidence caring for the patient, taking a history from the patient, and performing a physical examination on the patient. All confidence items were rated on a 10-point Likert scale (1 = not at all confident, 10 = very confident). Responses to the three confidence items were averaged to create an aggregate confidence score (α = 0.82, 95% Confidence Interval [CI]: [0.79–0.84]).

After completing the items related to the clinical vignette, participants completed the previously published Attitudes Towards LGBT People Scale (ATLPS).^35s Examples of items on the ATLPS include: “Marriage should be equally available to both heterosexual and same-sex couples” and “If I found out a friend was changing sex, I could no longer be his or her friend.” All items on the ATLPS were rated on a 7-point Likert scale (1 = completely disagree, 7 = completely agree). Responses were averaged across items after reverse-scoring appropriate items (α = 0.88, 95%CI[0.86–0.89]). Higher ATLPS scores reflected more negative attitudes towards LGBT people.

Demographics

The final section of the study instrument captured demographic information, including participants’ age, sexual orientation, gender identity, race, and year in training. We also inquired about participants’ degree of religiosity due to its relevance in clinicians’ willingness to care for LGBT patients.^36s,37s Religiosity was rated on a 5-point Likert scale (1 = not religious, 5 = very religious). Finally, participants self-reported how comprehensive their training about LGBT health was during their medical education (1 = minimal training, 10 = comprehensive training).

Statistical Analysis

First, descriptive statistics were calculated for all of the variables. Categorical variables were compared using Pearson’s chi-squared test (χ²). Continuous variables were compared using Analysis of Covariance (ANCOVA) with Bonferroni post-hoc comparisons. ANCOVAs controlled for participants’ sexual orientation, gender identity, and year-in-training due to their conceptual relevance to study variables.

Hierarchical linear regressions were utilized to evaluate primary outcome of willingness to prescribe PrEP to the presented patient. With each additional variable block, model fit and model fit change statistics (Δr², ΔF) were evaluated. To evaluate the indirect effects of patient condition on willingness to prescribe PrEP via the mediators of anticipated adherence, anticipated condomless sex, and perceived HIV-risk, bootstrapping analysis was performed. Indirect effect analyses were conducted utilizing Hayes’s PROCESS macro for SPSS (v3.5) to generate 10,000 bootstrapped samples from which 95% confidence intervals were calculated.^38s Patient condition was the independent variable, with willingness to prescribe PrEP as the dependent variable. For regression and mediation analyses, the gay male patient condition was designated as the reference category due to their high HIV-burden and PrEP prescriptions.^1,10,39s,40s

All analyses were completed using IBM SPSS v26 (Armonk, NY). A P-value < .05 was considered statistically significant.

Results

A total of 670 medical students completed the study. Based on the total enrollment of the schools that participated in the study, the response rate was 14.2% (670/4,702). The number of participants randomized to each condition ranged from 107–115.

Demographics

The majority of the participants were in allopathic training programs (n = 551, 82.2%), and in their fourth-year of training (n = 208, 31.0%). Nearly two-thirds of participants were female (n = 427, 63.7%), and the majority identified as heterosexual (n = 543, 81.0%). Mean participant age was 25.7 years (Standard Deviation [SD] = 3.0). Full demographic information is provided in Table 1. Demographics of students randomized to each patient condition were compared (Supplemental Table 1). The ATLPS had an overall mean of 1.79 (SD = 0.92). The mean of the religiosity item was 2.67 (SD = 1.39).

Table 1.

Sample Demographics (N = 670).

Gender Identity	n	%
Male - Cisgender	228	34.0
Female - Cisgender	427	63.7
Gender Diverse	15	3.3
Race/Ethnicity
African-American (Black)	28	4.2
Caucasian (White)	426	63.6
Hispanic/Latino	37	5.5
Native American	3	0.4
Asian	134	20.0
Other	42	6.3
Sexual Orientation
Heterosexual (straight)	543	81.0
Homosexual (gay)	42	6.3
Bisexual	54	8.1
Other	31	4.6
Year of Training
M1	185	27.6
M2	123	18.4
M3	154	23.0
M4	208	31.0
Type of Training
Medicine (allopathic-MD)	551	82.2
Medicine (osteopathic-DO)	119	17.8

PrEP Candidacy

The percentage of participants indicating the presented patient was a candidate for PrEP for HIV prevention differed between patient conditions (${\chi }_5^2$ = 99.2, P < .001). The greatest percentage of students indicated the gay male patient (93.8%) was a PrEP candidate. The lowest indicated candidacy for the heterosexual female patient (53.3%), compared to the bisexual male (90.8%), transgender male (90.4%), transgender female (88.5%), and the gender nonbinary person (90.4%; all P < .001). No additional pairwise comparisons were statistically significant (all P > .05).

Perceptions of HIV-Risk and Sexual Risk Behaviors

Across all conditions, we found that anticipated adherence (F[5,665] = 3.60, P = .003), frequency of condomless sex (F[5,665] = 5.44, P < .001), and overall HIV-risk (F[5,665] = 17.2, P < .001) differed. Overall mean (M) anticipated adherence to PrEP was 7.09 (SD = 1.66). In ANCOVA analysis, the highest anticipated adherence was found for the transgender female condition (M = 7.51, 95%CI[7.21–7.81]) and the lowest was found for the heterosexual female condition (M = 6.60, 95%CI[6.29–6.91], P = .001). (Figure 1)

Figure 1.

Comparison of overall perceived HIV risk and specific risk compensation Items between the 6 experimental conditions. AM = Mean Difference.*P < .05, **P < .01, ***P < .001.

The overall mean of anticipated increased frequency of condomless sex if prescribed PrEP was 6.46 (SD = 2.04). Heterosexual females (M = 7.20, 95%CI[6.81–7.59]) were viewed as most likely to increase frequency of condomless sex if prescribed PrEP as compared to the bisexual male (M = 6.08, 95%CI[5.71–6.46], P = .001), transgender male (M = 6.23, 95%CI[5.87–6.60], P = .006), transgender female (M = 6.30, 95%CI[5.93–6.67], P = .01), and gender nonbinary (M = 6.15, 95%CI[5.78–6.51], P = .002) conditions. (Figure 1)

The overall perceived HIV-risk of the patient was 6.24 (SD = 1.86). The gay male patient condition was viewed as having the highest HIV-risk (M = 6.89, 95%CI[6.56–7.21]), while the heterosexual female was viewed as having the lowest HIV-risk (M = 4.92, 95%CI[4.59–5.26]). (Figure 1) Perceived HIV-risk of the gay male patient was higher than that of the heterosexual female (M = 4.92, 95%CI[4.59–5.26], P < .001) and the gender nonbinary person (M = 6.06, 95%CI[5.74–6.28], P = .006). The heterosexual female patient condition was perceived as being at lower HIV-risk compared to all other conditions (all P < .001).

Confidence

The overall mean of the confidence scale was 7.94 (SD = 1.60). Confidence differed between conditions (F[5,665] = 3.51, P = .004). Students reported the lowest confidence caring for the heterosexual female patient seeking PrEP (M = 7.69, 95%CI[7.39–7.98]). Only the comparison between the heterosexual female and bisexual male conditions was significant (M = 8.38, 95%CI[8.09–8.66], P = .02).

Willingness to Prescribe PrEP

Willingness to prescribe PrEP varied between patient conditions (F[5,665] = 28.4, P < .001). Overall, mean willingness to prescribe PrEP was 7.04 (SD = 2.51). Participants indicated the highest willingness to prescribe PrEP to the gay male patient (M = 7.82, 95%CI[7.40–8.24]) and the lowest willingness to prescribe to the heterosexual female patient (M = 4.66, 95%CI[4.22–5.10]). Willingness to prescribe PrEP was higher for all patient conditions relative to the heterosexual female condition (all P < .001). No differences in willingness to prescribe PrEP were identified when comparing the gay male to the bisexual male (M = 7.55, 95%CI[7.12–7.98]), transgender male (M = 7.29, 95%CI[6.87–7.70]), transgender female (M = 7.58, 95%CI[7.16–8.00]), or gender non-binary patients (M = 7.34, 95%CI[6.93–7.76]; all P > .05).

Regression analyses evaluating willingness to prescribe PrEP are shown in Table 2. In the first model, significant effects were found for the heterosexual female patient condition (b = −3.17, P < .001). Addition of assumptions about HIV-risk and related behavior (block 2) improved model fit (Δr² = 0.16, ΔF = 51.2, P < .001) with significant effects of anticipated adherence to PrEP (b = 0.41, P < .001) and perceived HIV-risk (b = 0.40, P < .001). In the third model, addition of participants’ demographics and training characteristics did not significantly impact model.

Table 2.

Hierarchal linear regression analysis predicting willingness to prescribe PrEP to presented patient.

	Model 1					Model 2					Model 3					Model 4
	Unstandardized		Standardized			Unstandardized		Standardized			Unstandardized		Standardized			Unstandardized		Standardized
	b	SE	Beta	t	P	b	SE	Beta	t	P	b	SE	Beta	t	P	b	SE	Beta	t	P
Patient Presentation (gay male)
Heterosexual Female	−3.17	0.31	−0.46	−10.20	<.001	−2.16	0.30	−0.32	−7.26	<.001	−2.21	0.30	−0.32	−7.42	<.001	−2.18	0.30	−0.32	−7.40	<.001
Bisexual Male	−0.29	0.31	0.31	−0.95	0.34	−0.22	0.28	−0.03	−0.79	0.43	−0.18	0.28	−0.03	−0.66	0.51	−0.28	0.28	−0.04	−0.99	0.32
Transgender Male	−0.55	0.30	−0.08	−1.81	0.07	−0.45	0.28	−0.07	−1.64	0.10	−0.49	0.28	−0.07	−1.77	0.08	−0.47	0.27	−0.07	−1.71	0.09
Transgender Female	−0.27	0.30	−0.04	−0.87	0.38	−0.40	0.28	−0.06	−1.46	0.15	−0.45	0.28	−0.07	−1.63	0.10	−0.39	0.27	−0.06	−1.42	0.16
Gender Non-Binary Person	−0.42	0.30	−0.06	−1.39	0.17	−0.16	0.28	−0.02	−0.58	0.57	−0.20	0.28	−0.03	−0.74	0.46	−0.24	0.28	−0.04	−0.88	0.38
Risk Compensation
Anticipated Condomless Sex						−0.07	0.04	−0.06	−1.61	0.11	−0.08	0.04	−0.06	−1.84	0.07	−0.07	0.04	−0.06	−1.60	0.11
Perceived HIV Risk						0.40	0.05	0.30	8.48	<.001	0.41	0.05	0.31	8.56	<.001	0.43	0.05	0.32	8.91	<.001
Anticipated PrEP Adherence						0.41	0.05	0.27	8.18	<.001	0.38	0.05	0.25	7.52	<.001	0.34	0.05	0.22	6.55	<.001
Training & Demographics
Participant Gender (cisgender)											0.32	0.65	0.02	0.49	0.63	0.29	0.64	0.02	0.45	0.65
Participant Sexual Orientation (heterosexual)											0.35	0.22	0.06	1.62	0.11	0.28	0.22	0.04	1.28	0.20
Moderate LGBTQ+ Training (minimal)											−0.36	0.27	−0.07	−1.33	0.18	−0.39	0.27	−0.07	−1.46	0.14
Adequate LGBTQ+ Training											−0.34	0.27	−0.07	−1.29	0.20	−0.36	0.26	−0.07	−1.35	0.18
Comprehensive LGBTQ+ Training											−0.07	0.38	−0.01	−0.17	0.86	−0.17	0.38	−0.02	−0.44	0.66
Somewhat Religious (not religious)											−0.06	0.23	−0.01	−0.25	0.80	−0.06	0.22	−0.01	−0.29	0.77
Very Religious											−0.18	0.19	−0.04	−0.96	0.34	−0.17	0.20	−0.03	−0.86	0.39
Year in training - M2 (M1)											0.07	0.25	0.01	0.27	0.79	0.11	0.25	0.02	0.45	0.65
Year in training - M3											0.17	0.24	0.03	0.73	0.47	0.12	0.24	0.02	0.50	0.62
Year in training - M4											−0.06	0.24	−0.01	−0.24	0.81	−0.18	0.24	−0.03	−0.74	0.46
Participant Age											0.07	0.03	0.08	2.24	0.03	0.06	0.03	0.07	2.01	0.05
Attitudes & Confidence
Attitudes towards LGBT People																−0.11	0.10	−0.04	−1.05	0.29
Confidence																0.20	0.06	0.12	3.55	<.001
	Model Summary Statistics					Model Summary Statistics					Model Summary Statistics					Model Summary Statistics
	F	P	r2	adj r2	SE	F	P	r2	adj r2	SE	F	P	r2	adj r2	SE	F	P	r2	adj r2	SE
	28.4	<.001	0.18	0.17	2.26	41.1	<.001	0.34	0.33	2.04	18.3	<.001	0.35	0.33	2.03	17.6	<.001	0.37	0.35	2.01
	Model Change Statistics					Model Change Statistics					Model Change Statistics					Model Change Statistics
	Δr2	ΔF	df1, df2	P		Δr2	ΔF	df1, df2	P		Δr2	ΔF	df1, df2	P		Δr2	ΔF	df1, df2	P
	0.18	28.4	5, 652	<.001		0.16	51.2	3, 649	<.001		0.02	1.44	11, 638	0.15		0.02	7.89	2, 636	<.001

For regression analysis, religiosity was collapsed into ‘not religious’ [1&2], ‘somewhat religious’ [3], and ‘very religious’ [4&5]. Comprehensiveness of training was collapsed into four categories: ‘minimal training’ [1&2], ‘moderate training’ [3–5], ‘adequate training’ [6–8], and ‘comprehensive training’ [9&10]. We also dichotomized participants’ sexual orientation into either ‘heterosexual’ or ‘sexual minority,’ and participants’ gender identity into ‘cisgender’ or ‘gender diverse.’

In the final regression model, the addition of the confidence measure significantly improved model fit (Δr² = 0.02, ΔF = 7.89, P < .001). The heterosexual female patient (b = −2.18, P < .001) was less likely to be prescribed PrEP relative to the gay male patient. The effects of anticipated adherence to PrEP (b = 0.34, P < .001) and perceived HIV-risk (b = 0.43, P < .001) remained significant. Confidence was also significantly associated with willingness to prescribe PrEP (b = 0.20, P < .001). LGBT-attitudes (ATLPS) were not significantly associated with willingness to prescribe PrEP. The final regression model explained 35% (adj. r² = .35) of variance in willingness to prescribe PrEP.

Bootstrapped Mediation Analysis

Mediation models (Figure 2) were performed twice, once without covariates and once adjusting for the relevant covariates. Unadjusted results largely mirrored adjusted results (Supplemental Table 2), and only the latter are discussed in the text. Significant, indirect effects of patient condition on willingness to prescribe PrEP were identified via the mediator of perceived HIV-risk for all patient conditions except the transgender female and transgender male conditions. (Table 3) This indicates that participants perceived heterosexual females, bisexual males, and gender nonbinary people to be at lower HIV-risk, and thus were less willing to prescribe PrEP.

Figure 2. Caption:

Conceptual model of the indirect effects of patient condition on willingness to prescribe PrEP mediated by anticipated condomless sex, perceived HIV risk, and anticipated adherence. Coefficients for all pathways are presented in Table 3. Analyses controlled for participant characteristics, including year in training, sexual orientation, gender identity, and attitudes towards LGBT patients (ATLPS). Analysis was completed using PROCESS model témplate #4.

Table 3. Mediation analysis of indirect effects.

The coefficients for the mediation analysis of indirect effects of patient condition on willingness to prescribe PrEP. For all analyses, the gay male patient condition was the reference group. Indirect effects were considered statistically significant at the P < .05 level if the confidence interval did not include zero. Boldface font denotes statistical significance. All analyses were adjusted for participants’ sexual orientation, gender identity, year in training, and attitudes towards LGBT people (ATLPS).

Indirect Effects

Anticipated Condomless Sex (X)

Perceived HIV Risk (Y)

Anticipated PrEP Adherence (Z)

Willingness to Prescribe PrEP (D; Direct Effects)

Effect

95%CI low

95%CI hi

P

Effect

95%CI low

95%CI hi

P

Effect

95%CI low

95%CI hi

P

Effect

95%CI low

95%CI hi

P

Het. F

−0.02

−0.08

0.02

> .05

−0.80

−1.10

−0.54

< .05

−0.18

−0.38

−0.01

< .05

−2.17

−2.74

−1.59

<.001

Bisexual M

0.04

−0.03

0.13

> .05

−0.21

−0.41

−0.03

< .05

0.08

−0.09

0.25

> .05

−0.18

−0.73

0.36

0.51

Transgender M

0.03

−0.03

0.11

> .05

−0.16

−0.36

0.02

> .05

0.02

−0.13

0.19

> .05

−0.42

−0.96

0.11

0.12

Transgender F

0.03

−0.02

0.10

> .05

−0.09

−0.30

0.09

> .05

0.20

0.03

0.36

< .05

−0.36

−0.89

0.18

0.19

Gender NB

0.04

−0.03

0.13

> .05

−0.36

−0.58

−0.16

< .05

0.03

−0.13

0.19

> .05

−0.17

−0.71

0.37

0.54

Constituent Pathways

Anticipated Condomless Sex (A)

Perceived HIV Risk (B)

Anticipated PrEP Adherence (C)

Effect

95%CI low

95%CI hi

P

Effect

95%CI low

95%CI hi

P

Effect

95%CI low

95%CI hi

P

Het. F

0.37

−0.15

0.90

0.16

−1.95

−2.41

−1.48

<.001

−0.45

−0.89

−0.02

0.04

Bisexual M

−0.77

−1.29

−0.24

<.001

−0.50

−0.96

−0.04

0.03

0.20

−0.23

0.63

0.37

Transgender M

−0.62

−1.14

−0.11

0.02

−0.39

−0.84

0.06

0.09

0.06

−0.36

0.48

0.78

Transgender F

−0.59

−1.11

−0.08

0.02

−0.23

−0.68

0.23

0.33

0.49

0.07

0.91

0.02

Gender NB

−0.75

−1.26

−0.24

<.001

−0.86

−1.31

−0.41

<.001

0.08

−0.35

0.50

0.72

Indirect Effect of Anticipated Condomless Sex

Indirect Effect of Perceived HIV Risk

Indirect Effect of Anticipated PrEP Adherence

Effect

95%CI low

95%CI hi

P

Effect

95%CI low

95%CI hi

P

Effect

95%CI low

95%CI hi

P

−0.05

−0.13

0.03

0.23

0.41

0.32

0.51

<.001

0.40

0.30

0.50

<.001

Additionally, there was a significant indirect effect via anticipated PrEP adherence for the heterosexual female and transgender female patient conditions. Specifically, participants anticipated heterosexual females would be less adherent to PrEP, thus they were less willing to prescribe PrEP to these patients. Conversely, participants viewed transgender female patients as more adherent to PrEP if prescribed, which was associated with higher willingness to prescribe PrEP. The direct effect of patient condition on willingness to prescribe PrEP remained significant for only the heterosexual female patient condition. No indirect effects were identified via anticipated condomless sex. (Table 3)

Discussion

Scale-up of PrEP prescription is a key component of the federal Ending the HIV Epidemic plan.^41s Achieving these ambitious public health goals requires physician training to recognize risk-factors for HIV, and to initiate appropriate counseling about HIV-risk reduction, including PrEP. Education must highlight identification of HIV-risk-factors among patients of all sexual orientations and gender identities.

We found participants were much less likely to identify the heterosexual female patient as a PrEP candidate, even though the behavioral risk factors for HIV were identical across experimental conditions. This may be influenced by the overwhelming marketing of PrEP to MSM, as well as the historical and present, disproportionate burden of HIV among MSM.¹ Medical students also may not be aware of all the indications for PrEP, which include heterosexual people with HIV risk-factors, such as the patient we presented in our vignette. Participants may have also made decisions about HIV risk and PrEP candidacy based on patients’ sexual orientation or gender identity rather than actual behavioral risk factors. While heterosexual women may be considered at lower HIV-risk relative to MSM overall, heterosexual women still account for nearly a fifth of all new HIV diagnoses in the U.S.¹ In addition, reducing risk assumptions to epidemiologic trends ignores individual-level risk factors that drive inequities in HIV incidence, particularly among Black women.^1,42s These findings were further underscored in the multivariable analyses, in which perceived lower HIV-risk was associated with decreased willingness to prescribe PrEP to the heterosexual female, bisexual male, and gender nonbinary patient conditions.

An additional factor that may have influenced participants’ willingness to prescribe PrEP to the patients in the vignettes was their knowledge of epidemiological trends in the prevalence of HIV among different demographic groups. For example, they may have been less willingness to prescribe PrEP to the heterosexual female patient because they may have assumed that her sexual partners would be heterosexual men, a group for whom the overall prevalence of HIV is relatively low (compared to MSM).¹ In turn, this understanding may have influenced subsequent assumptions about the patient’s behavior if prescribed PrEP. Importantly, all of the vignettes presented a patient who was actively seeking PrEP, which is a strong indicator that the patient viewed themselves as being at risk for HIV. It will be important for future research to consider medical students’ assumptions about patients’ sexual partners because their assumptions may be inaccurate (e.g., a heterosexual woman’s sexual partners may be heterosexual or bisexual), in which case basing decisions on assumptions could be problematic.

Furthermore, the assumed non-adherence to PrEP of the heterosexual female condition was in turn associated with lower willingness to prescribe PrEP. Similar findings have been identified in studies of physicians, citing concerns about non-adherence as justification for not prescribing PrEP to patients with HIV-risk-factors.^20,43s−46s Non-adherence is also problematic due to reduced efficacy of TDF/FTC as PrEP with less-frequent dosing.^5,47s While non-adherence to PrEP is also concerning due to the potential for HIV resistance to TDF/FTC, evidence has accrued that suggests that the benefits of HIV prevention with PrEP outweigh this risk.^48s

The heterosexual female was also perceived as most likely to increase the frequency of condomless sex if prescribed PrEP relative to the other patient conditions. This connects with previously reported and documented concerns of sexual disinhibition following PrEP prescription, including increased condomless sex and increased numbers of partners.^{43s−46s,49s−52s} However, it is important to note that this phenomenon has largely been studied among MSM and transgender women, who have comprised the majority of patients in PrEP clinical trials and subsequent open-label extensions.^49s As the heterosexual female patient was perceived as being at lower overall HIV-risk compared to the other patient conditions, participants may have assumed that this patient would be more likely to engage in potential sexual-risk behaviors (e.g., non-adherence, condomless sex) because risk would remain generally low. In contrast, for populations perceived as being at higher risk for HIV (e.g., gay men), participants may have believed that they would be less likely to engage in potential risk behaviors because their risk would remain generally high. However, we did not identify significant indirect effects of patient condition on likelihood to prescribe PrEP via anticipated condomless sex, which has previously been implicated as a reason for students’ hesitation to prescribe PrEP.²⁷

Participants also reported the lowest confidence caring for the heterosexual female patient, although this was only significantly different from the bisexual male patient. As HIV is commonly associated with gay/bisexual men and transgender women, participants may have felt less confident caring for patients at risk for HIV who were not a member of one of those groups. In multivariable analyses, lower confidence was associated with lower willingness to prescribe PrEP. Importantly, we did not identify any significant effects of LGBT-attitudes or of participants’ demographics on willingness to prescribe PrEP.

Implications

These findings have important implications for medical training going forward. First, medical education must include training about PrEP and the full scope of patients who are at risk for HIV. Previous research has identified gaps in medical education regarding PrEP and HIV-risk factors.²⁹ The current findings suggest that this training gap may have implications for subsequent PrEP decision-making. We also did not identify any significant effects of self-reported training about LGBT health on willingness to prescribe PrEP, which may be due to infrequent presentation of heterosexual or transgender people as candidates for PrEP.²⁹

Our findings are aligned with previous work showing heterosexual women are often not identified as candidates for PrEP by clinicians.^18,19 Other studies have found that heterosexual women may also have low knowledge of PrEP or stigmatized attitudes towards PrEP, further limiting scale-up in this population.^{18,19,53s−55s} An underlying factor contributing to heterosexual women not being identified as candidates for PrEP and having low knowledge of PrEP is their exclusion from much of the early research and marketing regarding PrEP and subsequent scale-up efforts.^56s Our findings that heterosexual women were only correctly identified as a PrEP candidate 53.3% of the time and were viewed as less likely to adhere to PrEP and more likely to increase condomless sex if prescribed PrEP may represent a downstream effect of the exclusion of heterosexual women from PrEP research and marketing. Finally, the absence of a significant association between attitudes towards LGBT people and willingness to prescribe PrEP suggests that disparities in the identification of patients as PrEP candidates may be rooted in knowledge deficiencies rather than biased judgements based on patients’ identities.

These knowledge disparities must be addressed during medical education. Recent work has shown that courses within medical education that include content on HIV risk factors often do not include content on PrEP for HIV prevention.^57s This indicates that additional innovation is needed in medical education regarding training about HIV, potentially including patient case discussions, standardized patient encounters, and systematic evaluation of current teaching about HIV to ensure that up to date information about HIV in the biomedical era of prevention is being taught.

Limitations

There are several limitations that should be considered when interpreting these study results. The first is the limitation of data collection utilizing an anonymous, online study instrument and the potential effect on the representativeness of our sample to medical students more generally. The proportion of LGBT medical students in our sample was slightly higher than nationally reported figures (AAMC/AACOM), but we adjusted for participant sexual orientation and gender identity in our analyses.^58s,59s Importantly, the racial composition of our sample was similar to that of medical students at the national level, however we also controlled for this variable to account for differences as well.^58s,59s

A second limitation is the use of an explicit measure of LGBT-attitudes (ATLPS), which introduces the potential for socially desirable responding. This may be one reason we were unable to identify an effect of LGBT-attitudes on willingness to prescribe PrEP. Thus, it is possible that the effect of students’ LGBT-attitudes may be stronger than we identified in the present study, which may become apparent with the use of an implicit LGBT-attitudes measure.

A third limitation is that our vignette explicitly described a patient asking for PrEP rather than requiring the healthcare provider to bring up PrEP during the encounter. Previous research has found that many conversations about PrEP are initiated by patients rather than providers.²¹ Other reports have identified patients who subsequently contracted HIV despite accessing primary care due to healthcare providers’ failure to initiate PrEP counseling.^60s,61s Future work would benefit from assessment of clinicians’ initial decision-making about PrEP based on patient presentation, rather than patient request. Our vignette was also limited in that we did not specify sexual orientation for the transgender and non-binary patient conditions. While all conditions were presented as having “multiple sexual partners of unknown HIV status,” there may also be differences in perceptions of HIV-risk based on the gender of the patient and their partner(s). In addition, we did not specify that the heterosexual female, gay male, and bisexual male were cisgender, although clinicians tend to assume that people are cisgender unless specified.^62s

Finally, the use of medical students as a study population is a limitation. Medical students were selected for their proximity to medical training. However, future work is needed to explore whether the trends we have identified here are also present among practicing physicians. However, identification of these gaps and errors in decision-making among students provides targets for medical education to ensure they are not propagated when students reach their own, independent practice.

Conclusion

Ending the HIV epidemic in the U.S will require prescription of PrEP to a broader array of patients with HIV-risk-factors. In this vignette-based study, we investigated medical students’ decision-making regarding PrEP for a diverse group of fictional patients. We found that even when presented with identical HIV-risk-factors, heterosexual females were less frequently indicated as candidates for PrEP, and students were less willing to prescribe PrEP to these patients. Assumptions about adherence and perceived overall HIV-risk mediated this relationship, pointing out specific targets for education and training about PrEP.