Validation of the “Gender Affirmation in Healthcare Settings” Scale Among Transgender Women Living with HIV in India: Findings from Exploratory and Confirmatory Factor Analyses

Abstract

Purpose

Gender affirmation (GA) has been associated with improved healthcare engagement and health outcomes. Given the lack of a valid tool to measure GA in healthcare settings (GAHS) among transgender women living with HIV (TWLH) in India, we tested the validity and reliability of a 4-item GAHS scale.

Methods

We used baseline, 3-month, and 6-month data from a longitudinal observational cohort study among 140 TWLH. We conducted exploratory factor analysis (EFA) using baseline data to determine the underlying factor structure measuring GA, and confirmatory factor analysis (CFA) using 3-month data. We assessed group (‘transgender woman’ vs indigenous transfeminine identities) and longitudinal measurement invariance across three time points. Convergent validity was assessed by correlating GA scores with medical GA and social factors, discriminant validity by their correlation with baseline HIV-related stigma, and predictive validity by their correlation with mental health and antiretroviral treatment (ART) outcomes.

Results

EFA indicated a single-factor structure. CFA confirmed a 4-item single-factor model with good fit (χ²=7.66 (2), p=0.02), high reliability (α=0.83), and measurement invariance. Convergent validity was supported by significant correlations between GA and social support, resilience resources, and physician trust. Discriminant validity was supported as GA scores were not correlated with internalized HIV-stigma. Higher baseline GA scores significantly predicted better 3-month ART adherence and lower anxiety, demonstrating predictive validity.

Conclusion

The GAHS scale demonstrated strong psychometric properties, exhibiting good validity and reliability, making it suitable for screening or monitoring GA in healthcare settings and for research purposes. Future studies should evaluate this scale among diverse populations of transgender people.

Introduction

Globally, transgender women face stigma and discrimination from families, the general public, and healthcare institutions, contributing to their high vulnerability to HIV.^1,2 The Gender Minority Stress theory³ suggests that transgender women’s experiences of anticipated discrimination in healthcare can negatively influence health outcomes – for instance, delays in accessing healthcare⁴ or suboptimal HIV treatment adherence.^5,6 Sevelius’ Model of Gender Affirmation (GA),⁷ discusses interdependent domains of GA: social (interpersonal validation – e.g., use of correct names and pronouns in social interactions), psychological (self-affirmation), structural (having legal gender identity documents that are congruent with one’s self-affirmed gender), and medical (e.g., access to gender-affirming hormones and surgeries).⁸ In reality, these various types of GA are interconnected. For example, a lack of a legal gender identity document may prevent transgender women from accessing healthcare services, including free antiretroviral treatment, as they might fear discrimination and misgendering in healthcare settings.^9,10

The GA framework elucidates how structural and institutional discrimination exacerbates health disparities while highlighting the protective effects of gender affirmation on health outcomes. For example, healthcare providers’ lack of recognition of self-affirmed gender identity – misgendering by using dead names (i.e., using birth names rather than trans person’s current name) or inappropriate pronouns – has been found to be associated with HIV care disengagement, suboptimal antiretroviral (ART) adherence, and poor mental health.^11,12

In India, a medical certificate stating that one has undergone gender-affirmative medical procedure(s) is needed to get a legal gender identity document as a man or woman,¹³ while those transgender people who have not undergone medical procedure(s) can get a legal gender identity document as a “transgender” person.¹³ Social GA in healthcare settings may reflect whether one has legal and/or medical GA – as transgender women living with HIV (TWLH) in government ART centers are usually registered in the name appearing in a government-issued legal name/gender identity document,¹² and the name and pronouns by which healthcare providers address transgender women may depend on in which gender or name they are registered.

While scales exist to measure stigma and discrimination faced by transgender women,¹⁴ including a validated scale on transgender identity-related stigma from India,¹⁵ there is a lack of a valid tool to measure transgender women’s experiences of social GA in healthcare settings in India. In USA, Sevelius et al.¹¹ tested a 12-item scale on access to GA in healthcare among TWLH who reported having ever received HIV primary care and found it to have good reliability. Guided by Sevelius’ Model of Gender Affirmation, we adapted four items from Sevelius et al.’s 12-item Access to Gender Affirmation in Healthcare scale¹¹ to create a 4-item Gender Affirmation in Healthcare Settings (GAHS) scale among TWLH in India as part of a study that assessed the impact of multiple stigmas on HIV treatment adherence. The objective of this paper was to assess and validate the dimensionality (factor structure) of the 4-item GAHS and assess its reliability or internal consistency.

Materials and Methods

Participants and recruitment

A prospective observational longitudinal cohort study was conducted in 2020-2021 among TWLH registered for HIV treatment/care at Alliance India’s Care and Support Centers (CSCs). The inclusion criteria were: 1) TWLH diagnosed with HIV in the last 12 months, 2) age ≥18 years, 3) currently identifying as transgender women, hijra, or using an equivalent indigenous term for transgender women or transfeminine persons, 4) able to provide informed consent, and 5) fluent in their native language or English.

Trained interviewers screened potential participants from Alliance’s CSC registry for eligibility by telephone in participants’ preferred language, and enrolled a total of 140 TWLH. Due to the COVID-19 pandemic, all interviews were conducted over the phone for the safety of participants and research staff. Interviewers obtained oral informed consent from all participants and directly entered responses into REDCap software at baseline, 3 months, and 6 months in English or the local languages of the respective states.

To minimize data entry errors, we used built-in logic checks, conditional branching, and range restrictions within REDCap. Additionally, each completed survey in REDCap underwent a thorough review and verification by a team member to ensure data accuracy and completeness. Participants were compensated with INR 200 (US $3) at baseline and INR 250 (US $4) at 3- and 6-month follow-up via e-cash (e.g., e-account transfer, mobile recharge). The study protocol was approved by the institutional review boards of Alliance India (Approval Number: 007/2019) and Albert Einstein College of Medicine, New York (Approval Number: 2019-9915).

Study variables and measures

Sociodemographic characteristics. Participants were asked the following questions. Age: “Please tell me your age in completed years” (open-ended response); Education: “What is the highest level of education you have achieved?” Response options included: no schooling, primary education, elementary, high school (10th grade), higher secondary (12th grade), college, and diploma course; Personal monthly income: “What is your average personal monthly income (in rupees)?” (open-ended response); Employment: “What is your employment status?” Options included: employed full-time (≥30 hours per week), employed part-time (<30 hours per week), unemployed/jobless, disabled, and retired. Gender identity (two questions): “What was the sex assigned to you at birth?” (options: male, female, intersex), and “What gender do you consider yourself to be now?” (options: man, woman, transgender man, transgender woman, genderqueer or gender non-binary, Hijra, Thirunangai, Jogta, Mangalmukhi, Shiv-Shakti, other).

Gender Affirmation in Healthcare Settings scale. A 4-item gender affirmation in healthcare settings (GAHS) scale (Table 1) was adapted from a pre-existing 12-item Access to Gender Affirmation in Healthcare scale with permission obtained from the primary author (Dr. Sevelius).¹¹ We chose the 4 items based on qualitative formative research findings and community consultations, which indicated the need for a concise measure to reduce participant burden while capturing domains encompassing physical environment, staff interactions, communication practices, and perceived provider competence. To measure GA in healthcare settings as a whole and not just GA by healthcare providers, we selected those items that reflected GA in the setting itself (item-1 on welcoming environment), GA by staff (item-2 on perceived respect and item-3 on pronoun use) and perceived confidence in the providers’ knowledge on transgender people’s health issues (item-4). The items were scored on a Likert scale ranging from 1 to 5 or 6. The total GAHS score was calculated by adding the score for each item (range: 4-21). Higher scores indicated better gender-affirming HIV care.

Table 1. Summary of study measures – Key measure and validity-related measures.

Measuresa	Number of items	Example items or content	Range of responses and scoring	Calculation of scoresb or % for binary items	Reliabilityc (Non-standardized alpha)
Key Measure
Gender Affirmation in Healthcare Settings (GAHS)	4d	During your last HIV medical appointment - How welcoming was your medical care provider’s waiting room for Hijra or trans women? How respectful was the front-desk staff towards you? How often were you addressed with your preferred pronoun by staff at the ART centre, including front desk staff, doctor, or counsellor? How confident were you in your medical care providers (at ART center) knowledge of transgender health issues? Higher scores indicated better gender-affirming HIV care.	1 for ‘Not welcoming at all’ to 5 for ‘Extremely welcoming’ 1 for ‘Not respectful at all’ to 5 for ‘Extremely respectful’ 1 for ‘Never’ to 6 for Always’ 1 for ‘Not confident at all’ to 5 for ‘Extremely confident’	Responses were summed (Score range: 4-21)	0.89
Measures for assessing the validity of the GAHS scale
Social support	11e	There is a special person who is around when I am in need There is a special person with whom I can share my joys and sorrows	1 for ‘Very strongly disagree’ to 6 ‘Very strongly agree’	The scores were averaged to create a composite score (range: 1-6)	0.83
Physician/Provider Trust	4f	Sometimes your doctor cares more about what is convenient for them than about your medical needs (reverse coded) Your doctor is extremely thorough and careful in investigations and examinations	1 for ‘Strongly disagree’ to 5 for ‘Strongly agree’	Responses were summed (score range: 4−20)	0.89
Resilience resources	9g	How difficult or easy is it to get help, assistance, or any kind of support from the following people, if you needed it? Friends you can trust Friends who are transgender women/Hijra Doctor you trust A nurse or other healthcare provider Parents Other family members Your guru A peer counselor or outreach worker Another person living with HIV (who is NOT a peer counselor/outreach worker)	‘Very difficult’ (0), ‘Difficult’ (1), ‘Easy’, and ‘Very easy’ (4)	Responses were summed (range 0−36)	0.75
Depression (past two-week)	9 (PHQ-9)	Little interest or pleasure in doing things Feeling down, depressed, or hopeless	0 for ‘Not at all’ to 3 for ‘Nearly every day’	Responses were summed (score range: 0-27)	0.89
Anxiety	2 (GAD-2)	Feeling nervous, anxious, or on edge Not being able to stop or control worrying	0 for ‘not at all’ to 3 for ‘nearly every day’	Responses were summed (score range: 0-6)	0.92
Internalized HIV stigma	9	In the last 3 months, how much have you felt that you should avoid sharing dishes or glasses just in case someone might catch HIV from you?	0 for ‘never’ to 3 for frequently	Scores were averaged to create a composite score (range: 0-3)	0.72
Medical Gender Affirmation	1	Participants reported whether they had undergone gender affirmation surgery	0 = No, 1 = Yes	%	--
Self-reported ART adherence	3	In the last 30 days, on how many days did you miss at least one dose of any of your ART medicine? In the last 30 days, how often did you take your ART medicine in the way you were supposed to? In the last 30 days, how good a job did you do at taking your ART medicine in the way you were supposed to?	0-30 days 1 for ‘Never’ 6 for ‘Always’ 1 for ‘Very poor’ to 6 for ‘Excellent’	Responses for the three items were linearly transformed to a 0–100 scale (0 – worst adherence, and 100 – best adherence	--

^aAll measures except medical gender affirmation are continuous.

^bHigher scores indicate better gender-affirming HIV care, more trust in the physician, higher resilience, higher levels of depression and stigma and better adherence.

^cCronbach’s α >0.70 was considered adequate.

^d4 items were used from the 12-item scale for assessing gender affirmation in healthcare.¹¹

^e10 items were used from the 12-item Multi-dimensional Scale of Perceived Social Support ¹⁶ and added a new item “I know someone else living with HIV that I can talk to about my health”.

^fAdapted from a 5-item scale.¹⁷

^gAdapted from the PLHIV Resilience Scale.¹⁸

ART, antiretroviral treatment; GAD-2, Generalized Anxiety Disorder-2 Scale; GAHS, Gender Affirmation in Healthcare Settings; PHQ-9, Patient Health Questionnaire-9 Scale.

Measures for assessing convergent, discriminant and predictive validity. The measures or scales used for assessing convergent or discriminant validity (Table 1) of the GAHS scale were: Social support¹⁶; Physician/Provider Trust¹⁷; Resilience resources (adapted from the People living with HIV [PLHIV] Resilience Scale)¹⁸; Depression¹⁹; Anxiety²⁰; Internalized HIV stigma²¹; Medical Gender Affirmation; and Self-reported ART adherence.²² The details of how these variables were measured (the number of items, coding and scoring of the items) and their reliability (non-standardized Cronbach’s alpha coefficients) are provided in Table 1. All the measures, except the dichotomous medical gender affirmation measure, were continuous, as indicated by the score range mentioned in Table 1.

Data analysis

Exploratory factor analysis (EFA). We performed EFA on the baseline data (N=140). Before conducting EFA, factorability of the correlation matrix was assessed using Bartlett’s test for sphericity and the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (>0.50).^23,24 EFA with Horn’s parallel analysis (PA) using Robust Maximum likelihood (MLR) estimator in Mplus8,²⁵ was conducted to determine the factor structure (number of factors) underlying the data. We hypothesized that a single factor underlies the data. For this study, items were retained if factor loadings were greater than 0.4.²⁶

Confirmatory factor analysis (CFA). We performed CFA on the 3-month data (N=134, 6 lost to follow-up) to validate the factor structure (relationship of each item with the factor) obtained in the EFA, using Mplus-8. MLR estimation was used to generate robust parameter estimates. Model fit was assessed using chi-square test of exact fit, the Root Mean Square Error of Approximation (RMSEA) (good ≤0.05; acceptable ≤0.08), Bentler’s Comparative Fit Index (CFI) (good ≥0.95; acceptable ≥0.90), the Tucker-Lewis index (TLI) (good ≥0.95; acceptable ≥0.90), and the Standardized Root Mean Square Residual (SRMR) (≤0.08).²⁷

We also tested for group invariance (‘transgender women’ vs. other trans identities [e.g., hijra]) and longitudinal/temporal invariance (based on repeated measurements: baseline, 3-month, and 6-month follow-up) of the GAHS scale. Three-stage testing was used: configural invariance (all parameters were estimated freely), metric invariance (the factor loadings were constrained to be equal), and scalar invariance (factor loadings and the intercepts were forced to be equal). Given that likelihood ratio tests are sensitive to sample size, the value obtained by dividing the chi-square value with the degrees of freedom below 2.5 was considered a perfect fit.²⁸ Additionally, models were compared based on the amount of change in RMSEA and CFI;²⁹ values ≤0.01 were considered nonsignificant, indicating the presence of measurement invariance.³⁰ Missing data were handled using full-information maximum likelihood (FIML), which assumes data are missing-at-random. The extent of missing data was minimal, with 4.3% and 7.8% missingness at the 3-month and 6-month follow-up, respectively.

Scale’s convergent, discriminant and predictive validity, and reliability. CFA model-based average variance extracted (AVE) was used as a measure of convergent validity. Additionally, convergent and discriminant validity were assessed using Pearson’s correlations with listwise deletion of missing data. For convergent validity, we hypothesized that baseline social support, resilience resources, and provider trust would be correlated with baseline GA scores. We also hypothesized that GA scores at 6-months would be correlated with medical GA at 6-months (as data on medical GA were not collected at baseline or 3-month visits). For discriminant validity, we hypothesized that there would be no significant correlation of GA at baseline with internalized HIV stigma at baseline. To evaluate predictive validity, we regressed ART adherence, depression, and anxiety at 3-month and 6-month follow-up on GA scores at baseline by running linear regression models. We hypothesized that higher scores on GA would be associated with higher scores of ART adherence and lower scores of depression and anxiety. In exploring the reliability of the GAHS scale, we computed non-standardized Cronbach’s-alpha for scale items at baseline, 3 months, and 6 months and CFA model-based reliability at 3 months, using FIML.

Results

Participants’ mean age was 31.1±9.7 years and almost a quarter (22.9%) had no schooling (Table 2). Almost three-fourths (71%) were employed, most were unmarried (92%), more than half reported (52.9%) living alone and over one-fifth (21.4%) living with their parents. Most participants self-identified as transgender woman (55%) and hijra (40%).

Table 2. Baseline characteristics of transgender women living with HIV (N=140).

Characteristics	n (%)
Age (Mean±SD)	31.1±9.7
Education
No Schooling	32 (22.9)
Completed Primary education (5th grade)	39 (27.9)
Completed Elementary education (8th grade)	25 (17.9)
Completed High school (10th grade)	26 (18.6)
Completed Higher secondary education (12th grade or higher)	18 (12.9)
Employment status
Employed Full-Time (30 or more hours per week)	57 (40.7)
Employed Part-Time (less than 30 hours per week)	44 (31.4)
Unemployed	29 (20.7)
Retired	10 (7.1)
Marital Status
Married	7 (5.0)
Unmarried	129 (92.1)
Divorced/Separated/Widow	4 (2.9)
Living arrangement
Living with parents or other family members	30 (21.4)
Living with a male partner	10 (7.1)
Living with peers or guru	22 (15.8)
Living alone or homeless	78 (55.8)
Gender identity
Transgender woman (English term)	77 (55.0)
Hijra	56 (40.0)
Others/Genderqueer/Thirunangai/Jogta	7 (5.0)

SD, standard deviation.

Data normality, factorability and sample size adequacy

Table 3 summarizes item-wise descriptive findings. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy value of 0.82 for the model was acceptable, indicating the 4-items were measuring a common factor. A significant Bartlett’s test of sphericity (chi-square=362.1, df=6, p<0.001) suggested that correlations between variables result from common variance. The sample sizes at baseline (N=140) and the follow-up (N=134) with the participant-to-item ratio of 35:1 and 33:1 at baseline and follow-up were adequate for a stable factor solution. Altogether, these findings supported factorability.

Table 3. Correlation coefficients and descriptive statistics of the four items of the Gender Affirmation in Healthcare Settings scale (Baseline data, N=140).

	GA1	GA2	GA3	GA4
GA1	1.00
GA2	0.74***	1.00
GA3	0.58***	0.71***	1.00
GA4	0.67***	0.76***	0.75***	1.00
Mean	2.47	2.51	2.91	2.63
SD	1.18	1.17	1.65	1.30

GA1-GA4: Four items of the GAHS scale.

^***p<0.001.

Exploratory factor analysis: Structure of the scale

The Horn’s parallel analysis with eigenvalue of 1.19 and scree plot based on parallel analysis (Figure 1) indicated a one-factor solution. Hence, based on these diagnostic criteria, a single-factor solution was chosen to model the items. All standardized factor loadings were high (range 0.78 to 0.89) and significant at p <0.05 (Figure 2). The model had an acceptable fit (χ²=3.74 (2), p=0.15, RMSEA=0.079, CFI=0.99, TLI=0.96, SRMR=0.02). The 4-item GAHS scale had good reliability (non-standardized alpha) at all three time points – baseline (α=0.89), 3 months (α=0.83) and 6 months (α=0.80).

Figure 1. Scree plot showing cut-off point for retained factors from parallel analysis indicating one-factor solution.

Figure 2. Exploratory factor analysis at baseline (n=140).

GA, gender affirmation; GA1-GA4: Four items of the Gender Affirmation in Healthcare Settings scale.

Confirmatory factor analysis: Validation of the scale’s structure

CFA indicated that a 4-item single-factor model had an acceptable fit (χ²=7.66 (2), p=0.02, RMSEA=0.14, CFI=0.96, TLI=0.86, SRMR=0.04) and high reliability (α=0.83). The CFA standardized estimates were all significant at p<0.001, ranging from 0.61 to 0.88 (Figure 3). The model-based composite reliability was high at 0.91 (threshold 0.70) and the average variance extracted (AVE) was 0.71 (threshold >0.50), indicating convergent validity. After choosing the single-factor structure, we summed the responses from the 4-items of the GAHS scale to create three composite scores: baseline 10.5±4.6 (range 4-19), 3-months 11.47±4.46 (range 4-20), and 6-months 12.5±3.6 (range 4-21).

Figure 3. Confirmatory factor analysis at 3 months (n=134).

Convergent and discriminant validity

As hypothesized, the GAHS scale was positively and significantly (p<0.05) correlated with social support, physician/provider trust, resilience resources, and medical gender affirmation (Table 4), indicating good convergent validity. Further, as hypothesized, the GAHS scale was not significantly correlated with internalized HIV-stigma, demonstrating evidence for discriminant validity. Baseline GA scores significantly predicted anxiety at 3-month (-1.14; 95% CI -1.57 to - 0.72) and 6-month follow-ups (0.07; 95% CI 0.00 to 0.13) and ART adherence scores at 3-month follow-up (-0.85; 95% CI -1.30 to -0.39), providing evidence for predictive validity. GA scores did not have a statistically significant association with depression scores.

Table 4. Validity of the 4-item Gender Affirmation in Healthcare Settings scale.

Validity	Measures	Correlation coefficienta(95% CI)	P-value
Convergent validity	Social support (score)b	0.17 (0.01 to 0.33)	0.04
	Physician/provider trust (score)b	0.42 (0.27 to 0.55)	<0.001
	Resilience resources (score)b	0.18 (0.02 to 0.34)	0.03
	Medical gender affirmation (Yes)c	0.22 (0.05 to 0.38)	0.01
Discriminant validity	Internalized HIV stigmab	-0.16 (-0.02 to 0.35)	0.07
Predictive validity	Outcome measures (scores)	Regression estimate (95% CI); Standardized beta coefficient	P-value
3-month follow-up	Anxiety	-1.14 (-1.57 to -0.72); -0.42	<0.001
	ART adherence (3 items)	0.07 (0.001 to 0.13); 0.17	0.04
6-month follow-up	Anxiety	-0.85 (-1.30 to -0.39); -0.31	<0.001

^aPearson’s correlation coefficient for continuous measures and Spearman’s rho for categorical measures.

^bGender affirmation at baseline correlated with measures at 3-month follow-up.

^cGender affirmation at 6-month follow-up correlated with measures at 6-month follow-up.

The sample size (N) at baseline, 3-month and 6-month follow-up was 140, 134 and 129, respectively. CI, confidence interval.

Measurement invariance

Group invariance testing (‘transgender woman’ identity vs other transfeminine identity) indicated that all three invariance models (configural, metric and scalar) indicate good fit based on chi-square p-values (all nonsignificant) and other fit indices (Table 5). Model fit indices in the first stage indicated the presence of configural measurement invariance. Comparisons of the metric model against the configural model (second stage testing) and the scalar model against the metric model (third stage testing), indicated that the metric and scalar measurement invariances were achieved - all having nonsignificant chi-square p-values and changes in RMSEA and CFI were <0.01 or close to these values. In summary, group invariance was achieved. Similarly, longitudinal measurement invariance testing (baseline, 3-month and 6-month follow-up) provided evidence for structural invariance over time, with all the models meeting fit criteria (χ²/df < 2.50, CFI >0.90 and SRMR <0.08) and changes in RMSEA and CFI were <0.01 (Table 4).

Table 5. Measurement invariance of Gender Affirmation in Healthcare Settings scale.

Invariance Model	Model fit indicesa							Change in fit indices
	χ2	df	p-value	RMSEA	CFI	TLI	SRMR	χ2	df	p-value	RMSEA	CFI	TLI	SRMR
Group invariance: Transgender identity vs others
Configural	4.148	4	0.38	0.013	0.999	0.997	0.022	-	-	-	-	-	-	-
Metric	6.839	7	0.44	0.000	1	1	0.040	2.244	3	0.52	0.013	-0.001	-0.003	-0.018
Scalar	9.839	10	0.45	0.000	1	1	0.039	3.034	3	0.38	0.000	0.000	0.000	0.001
Longitudinal measurement invariance: Baseline, 3- and 6-month follow-up visits
Configural	86.047	39	<0.001	0.093	0.924	0.871	0.073	-	-	-	-	-	-	-
Metric	90.854	45	<0.001	0.085	0.926	0.891	0.076	4.807	6	<0.001	0.008	-0.002	-0.020	-0.003
Scalar	97.905	51	<0.001	0.081	0.924	0.902	0.078	7.051	6	<0.001	0.004	0.002	-0.011	-0.002

^aχ²/df values were < 2.50 for all models, indicating a good fit for all models.

The sample size (N) at baseline, 3-month and 6-month follow-up was 140, 134 and 129, respectively.

χ², chi-square statistic; CFI, comparative fit index; df, degrees of freedom; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual; TLI, Tucker-Lewis index.

Discussion

Gender affirmation,^11,31 a construct distinct from transgender identity-based discrimination,¹⁵ can influence health outcomes for transgender women. We assessed the dimensionality, validity and reliability of a 4-item Gender Affirmation in Healthcare Settings (GAHS) scale adapted from an existing 12-item scale measuring GA in healthcare in the United States – a first GAHS scale validation study reported from India. Our findings showed the presence of a single factor, confirming that all four items reflect the underlying construct of GA in healthcare settings and had good internal consistency (reliability). The scale was also found to have good construct validity, and satisfactory convergent, discriminant, and predictive validity. This scale could be used in research to test associations between GA in healthcare settings and health outcomes, and as a tool to monitor and evaluate programs or interventions to improve GA practices in healthcare.

From the original 12 items that were used in the scale on GA in healthcare, we chose 4 items based on qualitative formative research and community consultations, as described earlier. While item-3 (pronoun use) is directly related to social gender affirmation, item-1 (welcoming environment) and item-2 (perceived respect) are also associated with intersecting issues such as HIV status and gender identity. Because of this intersectionality, it may be difficult to separate whether the perceived disrespect or non-welcoming environment is because of one’s gender identity or HIV-positive status. Given the poor infrastructure in government hospitals in general and the common waiting rooms for the general population and transgender people in the government ART centers, the perception of a welcoming environment could be influenced by factors other than gender identity status (e.g., lower socioeconomic class, lower educational status). However, the high intercorrelations among the 4-items and identification of a single factor provide evidence that the chosen items measure a single underlying construct – gender affirmation in healthcare settings.

The GAHS scale was positively and significantly correlated with provider trust and having undergone any medical gender affirmation procedure, indicating good convergent validity. If providers are knowledgeable and engage in gender-affirmative practices (using proper pronouns and being respectful), it can increase trust. Similarly, those who have undergone any medical GA procedure are more likely to have legal gender affirmation and be registered as a woman or transgender person in healthcare settings, which in turn may be associated with proper pronoun use and respectful treatment. We found evidence for predictive validity as higher baseline GA scores were associated with higher ART adherence scores at the follow-up visit.

Similarly, baseline GA scores significantly predicted anxiety scores in follow-up; lack of GA in the baseline visit could lead to anxiety because of anticipated discrimination and gender non-affirmation. These findings are consistent with studies that have reported associations between GA and mental health,^31,32 and between GA and ART adherence.^11,12 Additionally, we found that group invariance (‘transgender woman’ identity vs. other transfeminine identities) and longitudinal measurement invariance (across three time points) were achieved, indicating structural equivalence across groups and over time – supporting construct validity.

Limitations

This study has several limitations. First, the participants were recruited through PLHIV care and support centers whose staff support TWLH in navigating the government ART centers. Thus, TWLH who are not affiliated with these centers may have different characteristics and experiences at the government ART centers, and thus the findings may not be generalizable to the broader TWLH population, although participants were from diverse regions of India. Further, the sample was limited to PLHIV. Hence, the findings may not be generalizable to transgender women not living with HIV, highlighting the need for future studies to test this scale among transgender women as well as transmasculine persons not living with HIV. Second, when assessing the convergent and discriminant validity of the GAHS scale, we used previously published scales from India or developing countries, including Zimet’s social support scale and PLHIV resilience scale. Although these scales showed adequate or good internal consistency (reliability) in this study, the lack of prior construct validation of these scales among transgender women in India may limit the robustness of our validity assessments in terms of the strength of association between these scales and GAHS.

Third, as the study was conducted during the COVID-19 pandemic, all interviews were conducted by telephone, which may have precluded participation of TWLH who had concerns about confidentiality or outing of HIV status due to living with peers or family, potentially affecting generalizability. However, potential participants were informed that interviews could be scheduled as per their convenience and they could complete the survey in one or two sessions to support privacy. Fourth, while the sample size was adequate for the CFA, the relatively smaller number of certain subgroups of TWLH – especially gender non-binary persons or persons with indigenous trans-identities such as thirunangai – meant that we could not assess measurement invariance for those subgroups, which warrant evaluation in future studies.

Finally, the tool was administered in five Indian languages (Hindi, Telugu, Bengali, Kannada, Tamil – See Supplementary Appendix SA1); however, we did not have sufficient sample size to check for the validity and reliability of each of these translated versions. We tried ensuring conceptual equivalency of the translated versions with the English version by discussing the whole instrument including the GAHS scale with bilingual research staff, piloting each translated version with a few participants, and revising them based on suggestions from the interviewers and interviewees. Future studies could specifically assess the validity and reliability of the GAHS scale in different Indian regional languages.

Conclusion

This is the first study from India on the validation of a GAHS scale. The GAHS scale was found to have a single-factor structure, good reliability and satisfactory convergent, discriminant and predictive validity; and achieved measurement invariance. The scale can be used by researchers to assess associations between GA and mental and physical health outcomes among TWLH in healthcare settings. It could also be used to assess whether healthcare settings are trans-affirming and to improve GA in healthcare settings by evaluating interventions to improve GA healthcare practices and interactions. Further studies need to test this scale, with appropriate adaptations in the wordings as needed, among diverse subgroups of transgender women, including those who are HIV negative, and gender non-binary people.