The Development and Piloting of Parallel Scales Measuring External and Internal HIV and Tuberculosis Stigma Among Healthcare Workers in the Free State Province, South Africa

Abstract

Background

The dual burden of tuberculosis and human immunodeficiency virus (HIV) is severely impacting the South African healthcare workforce. However, the use of on-site occupational health services is hampered by stigma among the healthcare workforce. The success of stigma-reduction interventions is difficult to evaluate because of a dearth of appropriate scientific tools to measure stigma in this specific professional setting.

Methods

The current pilot study aimed to develop and test a range of scales measuring different aspects of stigma—internal and external stigma toward tuberculosis as well as HIV—in a South African healthcare setting. The study employed data of a sample of 200 staff members of a large hospital in Bloemfontein, South Africa.

Results

Confirmatory factor analysis produced 7 scales, displaying internal construct validity: (1) colleagues’ external HIV stigma, (2) colleagues’ actions against external HIV stigma, (3) respondent’s external HIV stigma, (4) respondent’s internal HIV stigma, (5) colleagues’ external tuberculosis stigma, (6) respondent’s external tuberculosis stigma, and (7) respondent’s internal tuberculosis stigma. Subsequent analyses (reliability analysis, structural equation modeling) demonstrated that the scales displayed good psychometric properties in terms of reliability and external construct validity.

Conclusions

The study outcomes support the use of the developed scales as a valid and reliable means to measure levels of tuberculosis- and HIV-related stigma among the healthcare workforce in a resource-limited context. Future studies should build on these findings to fine-tune the instruments and apply them to larger study populations across a range of different resource-limited healthcare settings with high HIV and tuberculosis prevalence.

Human immunodeficiency virus (HIV)/AIDS and tuberculosis have merged into a deadly coepidemic in South Africa. In absolute numbers, the country houses the highest number of people living with HIV: In 2013, 6.3 million South Africans were infected with HIV, constituting 18.5% of the global burden of HIV infection [1]. In addition, South Africa has one of the most severe tuberculosis epidemics in the world; the country had the highest incidence of tuberculosis (860/100 000 in 2013) and had 26 023 reported cases of rifampicin-resistant or multi-drug-resistant (MDR) tuberculosis in 2013 [2]. Both epidemics are intricately intertwined, as approximately 73% of patients with tuberculosis are coinfected with HIV and as many as 60% of HIV-infected South Africans have HIV-associated tuberculosis [3, 4].

The dual burden of tuberculosis and HIV also has a severe impact on the South African healthcare workforce. Occupational exposure to tuberculosis constitutes a major health risk for healthcare workers (HCWs), especially in resource-constrained settings with large patient numbers, resulting in overcrowded health facilities and poorly implemented infection control strategies. A 2006 systematic review on low- and middle-income countries reported that the excess incidence of tuberculosis that was attributable to being an HCW ranged from 25 to 5361 cases per 100 000 people per year [5]. A South African study on the extent and impact of nosocomial transmission of tuberculosis among HCWs indicated that these professional categories were up to 3 times more likely to acquire tuberculosis than the general population [6]. In addition, O’Donnell et al demonstrated that HCWs have a 5- to 6-fold increased rate of hospital admission with MDR or extremely drug-resistant tuberculosis compared with non-HCWs [7]. Consequently, tuberculosis is officially classified as an occupational disease. The HIV epidemic is equally affecting the health workforce as the mutually reinforcing epidemiology of HIV and tuberculosis evidently gravely affects this subpopulation. Although HIV predominantly remains a sexually acquired infection, there is also an occupational risk as the preliminary results of a 2012 survey among 513 healthcare workers in 3 hospitals in the Free State province of South Africa revealed that >21.2% of the respondents reported needle-stick injury or unprotected exposure to bodily fluids [8]. The limited epidemiological evidence indicates that HIV prevalence among South African HCWs ranges from 11.5% to 20.0%, demonstrating the impact that HIV/AIDS is having on the workforce [9].

The high HIV prevalence among the healthcare workforce and the above-mentioned risk of tuberculosis (co-)infection render workplace health services for tuberculosis and HIV/AIDS an essential part of any overall health systems strengthening strategy [10]. Research has demonstrated that providing HIV and tuberculosis services to HCWs at work is cost-effective and is an approach preferred by the majority of HCWs, especially when part of a more comprehensive care package [11–14]. Accordingly, a joint World Health Organization/International Labour Organization/Joint United Nations Programme on HIV/AIDS (UNAIDS) policy document on the provision of tuberculosis and HIV prevention and care for HCWs explicitly recommends the on-site availability of occupational health services for the entire workforce so that full access to HIV and tuberculosis prevention, treatment, care, and support for this vulnerable group can be attained [15].

Despite these promising policy recommendations, a recent review article stated that (HIV- and tuberculosis-related) stigma and discrimination act as “key barriers to both the delivery of quality health services by health providers and to their utilization by community members and health providers themselves” [16]. Stigmatization in the healthcare setting can have severe implications for HCWs and health facilities when HIV- and/or tuberculosis-infected HCWs delay or avoid care, causing increased morbidity and mortality and further strain on an overburdened health system [16]. Accordingly, the recent past has been marked by the development and testing of stigma-reduction tools and interventions in healthcare settings [17].

However, the success of the interventions is difficult to evaluate because of a dearth of appropriate scientific tools to measure stigma in this specific professional setting. Uys et al indicated that the majority of articles reporting intervention outcomes did not include a validated instrument to measure change in stigma over time [17, 18]. The limited number of studies that did measure stigma in the healthcare setting focused almost exclusively on (1) HIV (and not tuberculosis) [19]; (2) stigmatizing attitudes of HCWs toward patients or patients toward HCWs (but not among HCWs) [20]; and (3) healthcare professionals (eg, doctors, nurses), but not the entire healthcare workforce. A recent study by Nyblade et al [19] produced the only validated tool identified in the survey of literature to assess HIV stigma among the entire healthcare workforce, but it did not include tuberculosis stigma and it was not executed in a high-HIV-prevalence setting (cleaners, clerks, security personnel, etc) [19, 21–23], rendering a more comprehensive measurement of HIV and tuberculosis stigma in the healthcare setting—based upon a solid theoretical framework—a research priority.

Theoretical Framework

Erving Goffman’s 1963 seminal analyses of stigma as a social phenomenon—which built on Émile Durkheim’s explorations in 1894 [24]—inspired and formed the bases of current research on stigma [25]. Goffman defined stigma as a “mark” or aspect of the self that is socially devalued [25]. People who display an undesirable difference from desirable norms are thought of differently, usually with negative consequences to their emotional well-being and social standing. Goffman’s interest was in people’s lived experience of stigma, but over time and in the field of health, subsequent analyses put increasing emphasis on the causes and consequences of stigma [16, 26, 27]. Also, following the influential work of Link and Phelan, focus was concentrated on stigma-as-process involving, among other things, labeling, stereotyping, rejecting, excluding, and discriminating [28–32].

Although the literature is easy to assimilate in terms of causes, consequences, and processes of stigma, this is less so when analyzing for different types of stigma. What emerges clearly in the literature is a set of 2 overarching types: external stigma and internal stigma [33, 34]. To produce an instrument that is conceptually clear, precise, and economical, we confined the types of stigma to be measured among HCWs to these 2 main categories: external stigma, which is directed by healthcare workers toward other healthcare workers, and internal stigma, which is directed by healthcare workers toward themselves. Rather than presenting new types or groups of stigma, these categories capture different directional dimensions of stigma—that which is externalized and that which is internalized.

The current pilot study was conducted in the context of a multicomponent international research collaboration aiming to improve the health of HCWs [35]. Within this framework, the current study aims to develop and test a range of scales measuring different aspects of stigma—internal and external stigma toward tuberculosis as well as HIV—in the healthcare setting. In this way, we aim to address the above-cited research gaps. First, we aim to develop a series of interrelated instruments that measure the different aspects of stigmatization (internal as well as external stigma) toward both HIV and tuberculosis. Second, we aim to develop scales that measure the stigmatizing attitudes of the healthcare workforce toward their fellow colleagues, as these particular stigmatizing attitudes are likely to be the primary barrier to the optimal use of occupational HIV and tuberculosis services by this at-risk group. Finally, we aim to develop and test instruments that measure stigma in the healthcare setting. We thus have to include the entire healthcare workforce—including cleaning and security personnel, food service workers, etc—as the sum of the attitudes of all people working in the facility constitutes the level of stigma in the healthcare setting.

MATERIALS AND METHODS

Stage I: Instrument Development

The instrument design was based on findings from a comprehensive literature review. Criteria for sourcing and selecting texts for review were that they should match as closely as possible the specific context and aims of our research by focusing predominantly on HIV and/or tuberculosis stigma in public healthcare facilities as perceived and experienced by HCWs. Texts for the review were sourced by searching across databases (eg, EBSCOHost and Web of Science), following up references in the most relevant texts, and using keyword searches in Google and Google Scholar. Keyword combinations included terms such as HIV stigma; TB stigma; discrimination; healthcare worker; health professional; occupational health; measur* stigma; validat* scale; stigma tool; Africa. The following types of texts were sourced: peer-reviewed journal articles, including systematic reviews; reports and other resources from websites of key international, regional, and national institutions as well as networks involved in stigma-focused research, interventions, and evaluations; stigma intervention toolkits; and existing stigma survey tools. In addition, a limited number of texts on stigma theory and on the broader context of HIV and tuberculosis epidemiology, prevention communication, and occupational health were selected based on our estimation of their reach and their relevance to our project.

In accordance with the above-described theoretical framework, different series of items were designed to measure stigma according to 4 scales: HIV external stigma; tuberculosis external stigma; HIV internal stigma; and tuberculosis internal stigma. Reflecting on these items, it became evident that a further differentiation was possible, one that would capture nuances in the main categories of external and internal stigma and which would differentiate the sources from the targets of stigma. This refinement resulted in 6 final scales:

• Others’ external stigma (OES) scales measure perceptions, attitudes, and behaviors that respondents witness being enacted, or perceive as existing, among other HCWs (referred to also as “colleagues”) in the hospital (scale 1 = HIVOES; scale 2 = TBOES).

• Respondent’s external stigma (RES) scales measure respondents’ perceptions, attitudes, and behaviors toward other HCWs in the hospital (scale 3 = HIVRES; scale 4 = TBRES).

• Respondent’s internal stigma (RIS) scales measure the perceptions, attitudes, and behaviors of respondents toward themselves, as well as stigma that they perceive or anticipate as being directed toward them from other HCWs (scale 5 = HIVRIS; scale 6 = TBRIS).

The study aims to develop scales that measure stigma among the entire healthcare workforce. The representatives of the Free State Department of Health (Dr. Kerry Uebel [Centre of Excellence, Pelonomi Hospital, Bloemfontein] and Lucky Nophale [Provincial Occupational Health Unit, Free State Department of Health]) insisted that in a South African context it is understood and well accepted that “HCWs” refers to the entire work-force from top to bottom.

The items of the different scales were informed by the selected literature: Items from existing stigma scales—not targeted at the healthcare workforce—were carefully selected and, if required, adapted to validly measure our 6 concepts. Changes in the wording of questions were sometimes needed so that they would be consistent with the strict definitions for each scale, and also to ensure that there would be enough questions allocated to each scale in order to measure the construct in a reliable and valid manner. Questions on HIV external stigma were taken, and sometimes adapted from, Buregyeya et al [13]; Health Policy Project [36]; Kalichman and Simbayi [37]; Kalichman et al [38]; Nyblade and MacQuarrie [39]; USAID Health Policy Initiative [40]; and Uys et al [17]. Questions on HIV internal stigma were taken, and sometimes adapted from, Buregyeya et al [13]; Feyissa et al [41]; Kalichman et al [38]; Nyblade and MacQuarrie [39]; USAID Health Policy Initiative [40]; and Uys et al [17]. Questions on tuberculosis external and internal stigma were informed by, or taken directly from or adapted from, Bond and Nyblade [26]; Coreil et al [42, 43]; Courtwright and Turner [28]; Daftary [44]; and Van Rie et al [45]. Particular attention was devoted to parsimony as the scales needed to be as short as possible to encourage the healthcare workforce to complete it, and to do so in the context of understaffing and busy schedules.

Stage II: Prepilot and Adaptation

A final, prepilot cross-examination of the instrument was carried out in a meeting of key stakeholders: 4 of the instrument designers; 3 experienced health-services field workers (1 of whom is HIV infected and understands HIV and tuberculosis stigma from an insider perspective); 1 medical doctor specializing in HIV/tuberculosis in local public hospitals; 1 local occupational health nurse who was also the coordinator of occupational health in the Free State province; and 1 Master of Public Health student. Final, small edits followed the group’s inputs.

Stage III: Piloting the Instruments

The work of stages I and II resulted in 6 scales: (1) colleagues’ (by which is meant HCWs other than the respondent) external HIV stigma (HIVOES: 9 items); and (2) colleagues’ external tuberculosis stigma (TBOES: 5 items); (3) respondent’s external HIV stigma (HIVRES: 9 items); (4) respondent’s external tuberculosis stigma (TBRES: 5 items); (5) respondent’s internal HIV stigma (HIVRIS: 8 items); and (6) respondent’s internal tuberculosis stigma (TBRIS: 5 items). The instruments were piloted in this study—a cross-sectional study among the staff of a large public hospital in the Free State province of South Africa—which is the third and final stage of instrument development.

Study Sample and Data Collection

The pilot study was conducted in a large hospital in Bloemfontein in the Free State province of South Africa. All staff members were eligible to participate. Field workers personally recruited 220 participants with the assistance of hospital department managers. Purposive recruitment was organized to be representative of the number of people in each job category: (1) doctors, (2) nurses, (3) allied health professionals, (4) administrative staff, and (5) support staff. The pilot study was granted ethical clearance by the ethics committee of the University of the Free State’s Faculty of Medicine (ECUFS NR 192/2012).

After obtaining written informed consent from all of the participants, trained field workers provided the participants with the standard questionnaires that were completed in a self-administered process. Respondents with low levels of reading literacy were gathered into small groups where trained field workers read through the questions verbally and answered any queries from the group; however, respondents needed to fill in the questionnaires themselves. Field workers scanned each questionnaire very quickly to identify incomplete questionnaires and returned these to respondents for editing. Questionnaires thus completed were then separated from the respondents’ signed consent forms, sealed in an envelope, and from then on treated as anonymous. The sample population interviewed consisted of 220 healthcare staff: 127 persons involved in direct patient care (eg, doctors and nurses), 60 support staff (eg, messengers, porters, cleaners), and 33 administrative staff.

Measures

The entire pilot study questionnaire included 87 questions and/or items and took approximately 15 minutes to complete. Apart from the 6 stigma scales to be tested, the survey included a series of sociodemographic questions (age, sex, occupation, education). Details for different occupations and departments in the hospital came from various health services instruments used, and refined over time, by the Centre for Health Systems Research and Development of the University of the Free State [8]. In addition, the pilot study assessed the HIV- and tuberculosis-related knowledge of the healthcare workforce, as the literature has repeatedly shown a link between knowledge and stigma [46, 47]. The instrument included 5 items testing the respondents’ HIV knowledge (eg, “Can people protect themselves from HIV by only having sex with healthy-looking people?”) and a rapid assessment of their knowledge on tuberculosis symptoms. Finally, previous research [16, 19, 42] clearly indicates that links exist between stigma and confidentiality. As informed by existing scientific evidence [16, 19, 42], the survey included a series of questions on confidentiality in the workplace (eg, “Do you think confidentiality is maintained in your occupational health unit?”).

Analysis

In a first step, and as a theory-testing model, internal construct validity was assessed by a series of separate CFAs using MPlus version 6 [48]. For each stigma scale, we removed items that did not successfully load (>0.40) onto the theoretical stigma domain [49]. The fit indices that were used were the Tucker–Lewis index (TLI), the comparative fit index (CFI), the root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). Following the recommendations of Hu and Bentler’s seminal article [50], 2 of the following 3 criteria had to be met for satisfactory global model fit to be attained: CFI/TLI ≥ 0.95, RMSEA ≤ 0.06, and SRMR ≤ 0.08. Other methodologists have proposed that RMSEA values <0.08 suggest adequate model fit [51], and CFI and TLI values in the range of 0.90–0.95 are indicative of acceptable model fit [52]. Following Brown’s advice, it is thus “especially important to consider the consistency of model fit as expressed by the various types of fit indices in tandem with the particular aspects of the analytic situation; for example, when N is somewhat small, an RMSEA = 0.08 may be of less concern if all other indices are strongly in a range suggesting ‘good’ model fit” [49].

Second, the reliability of the different stigma scales was measured by the Cronbach’s α coefficient. For a stigma scale to be considered consistent, the value of the coefficient has to be >0.70 [53]. We also sought to improve the internal reliability by removing items one by one and reporting the impact on the coefficient.

In a final step, we tested the correlations between the different stigma subscales to assess the subscales’ interrelationships as well as their ability to differentiate between the different types of HIV-and tuberculosis-related stigma. External construct validity was investigated by assessing the relationship between the different stigma subscales and relevant correlates, using structural equation modeling (SEM) [48]. However, the workforce interviewed is rather diverse—both health professionals and support and administrative staff—necessitating us to assess these relationships while controlling for the sex of the respondent, his/her role in the facility (health professional or not), and the level of education attained. It can be hypothesized that (1) the health-related knowledge of the respondents and (2) the level of confidentiality in the workplace are correlated with the different stigma scales. Based on the literature [16, 42], we expected the stigma scales (HIVOES and TBOES) assessing the stigmatization by others to be negatively correlated with the level of confidentiality, as a breach in confidentiality can be a proxy of stigmatizing behavior on the floor. We expected the respondent’s external stigmatizing perceptions, attitudes, and behaviors (HIVRES and TBRES) to be negatively correlated with his/her knowledge of the disease, as this relationship has been repeatedly reported in the literature [46, 47, 54]. Finally, we expect the respondent’s internal stigmatizing perceptions, attitudes, and behaviors (HIVRIS and TBRIS) to be positively correlated with his/her external stigmatizing perceptions, attitudes, and behaviors, as past studies demonstrated that past experiences of external stigma such as blame, rejection, intimidation, name-calling, exclusion, and isolation influence internal stigma [55]. We furthermore expect internal stigma to be negatively correlated with the level of confidentiality in the facility, as a breach of confidentiality could instigate internal feelings of stigma in affected individuals.

RESULTS

Study Population

The mean age of the members of the healthcare workforce in this sample was 40.8 years (SD, 11.4 years). The majority of the respondents were female (67.7%). With regard to the highest level of education achieved, 1.8% had only completed primary education, 12.3% had some secondary education, 39.3% had completed secondary education, and 46.6% had completed tertiary education (university or college). The vast majority of respondents were black (79.4%), whereas 8.3% of the respondents were Coloured, 11.9% were white, and 0.5% were Asian. The majority of respondents were either married (37.7%) or single (40.0%), while the remainder of the healthcare workforce was living together (7.3%), divorced (7.7%), widowed (6.8%), or separated (0.5%).

Presentation of the Items

Table 1 displays the different items ascribed to the different stigma scales as well as the spread of the responses of the healthcare work-force. Although it is not really possible to interpret these data due to a lack of normative data, it is clear that, on average, the level of reported stigma by others (HIVOES and TBOES) is higher than respondents’ own reported stigma (HIVRES and TBRES).

Table 1.

Stigma Scales, Presentation of the Items, and the Distribution of Responses (N = 220)

Stigma Scale Item	Item No.	Strongly Disagree	Disagree	Agree	Strongly Agree
Others’ external stigma toward HIV
6.1. I have noticed that some of my coworkers in this hospital look down on HCWs who they think may be HIV infected	1	65 (30.1)	72 (33.3)	41 (19.0)	38 (17.6)
6.4. I have heard HCWs making negative remarks about the health of coworkers involved in HIV care and treatment	4	80 (37.4)	80 (37.4)	31 (14.5)	23 (10.7)
6.12. I have witnessed HCWs suspected of having HIV being rejected by coworkers in the workplace	12	79 (36.6)	94 (43.5)	32 (14.8)	11 (5.0)
6.13. I have heard that other HCWs in this hospital are afraid of catching HIV from coworkers who care for HIV-infected patients	13	77 (35.6)	92 (42.6)	34 (15.7)	13 (6.0)
6.14. Other HCWs think it is worthwhile for the hospital to invest in the career development of HIV-infected HCWs	14	15 (7.0)	39 (18.2)	96 (44.9)	64 (29.9)
6.15. I have heard about other HCWs in this hospital who give extra support to coworkers with HIV	15	8 (3.8)	41 (19.2)	96 (45.1)	68 (31.9)
6.16. I have heard that some HCWs educate coworkers who stigmatize people living with HIV	16	10 (4.6)	42 (19.4)	101 (46.8)	63 (29.2)
6.17. It has been made clear by people in charge at this hospital that HCWs will not lose their jobs because of being HIV infected	17	1 (0.5)	26 (12.0)	90 (41.7)	99 (45.8)
6.18. I have witnessed other HCWs in this hospital doing something to stop stigma in the workplace	18	19 (8.8)	55 (25.6)	92 (42.8)	49 (22.8)
Respondent’s external stigma toward HIV
6.2. I would feel comfortable being close friends with an HCW who is known to be HIV infected	2	8 (3.7)	14 (6.5)	87 (40.3)	107 (49.5)
6.3. HIV-infected HCWs should have the same chances for promotion as HCWs who are HIV negative	3	4 (1.8)	4 (1.8)	43 (19.8)	166 (76.5)
6.5. I am comfortable having HIV-infected HCWs alongside me in my job	5	6 (2.8)	6 (2.8)	84 (38.7)	121 (55.8)
6.6. I am comfortable sharing a bathroom with HIV-infected coworkers	6	3 (1.4)	9 (4.1)	91 (41.9)	114 (52.5)
6.7. In my opinion, HCWs living with HIV should probably feel shame	7	142 (65.1)	55 (25.2)	11 (5.0)	10 (4.6)
6.8. HCWs with HIV should not feel guilty about it	8	11 (5.0)	8 (3.7)	72 (33.0)	127 (58.3)
6.9. Most HCWs with HIV have had many sexual partners	9	108 (49.8)	87 (40.1)	12 (5.5)	10 (4.6)
6.10. HIV-infected HCWs can be good role models in the workplace	10	6 (2.8)	8 (3.7)	94 (43.1)	110 (50.5)
6.11. Doctors and nurses with HIV should continue to practice medicine	11	3 (1.4)	4 (1.8)	83 (38.1)	128 (58.7)
Respondent’s internal stigma toward HIV
7.1. If I had HIV, I would feel comfortable disclosing to some of my coworkers	19	35 (16.2)	42 (19.4)	88 (40.7)	51 (23.6)
7.2. I would feel ashamed if coworkers knew that someone in my family was HIV infected	20	75 (34.7)	107 (49.5)	26 (12.0)	8 (3.7)
7.3. If I was diagnosed with HIV, I would be afraid that some coworkers might blame me	21	62 (28.6)	80 (36.9)	58 (26.7)	17 (7.8)
7.4. I would not need to feel shame if I was HIV infected	22	8 (3.7)	29 (13.4)	100 (46.3)	79 (36.6)
7.5. If I was HIV infected, I would worry that some coworkers might avoid touching me	23	58 (27.0)	93 (43.3)	52 (24.2)	12 (5.6)
7.6. As an HCW, I would feel it was my fault if I was infected with HIV	24	67 (31.2)	77 (35.8)	48 (22.3)	23 (10.7)
7.7. I have felt stigmatized by coworkers in my workplace because of a health condition	25	72 (33.5)	97 (45.1)	38 (17.7)	8 (3.7)
7.8. If I had HIV, I would avoid making new friends at my workplace	26	93 (43.7)	98 (46.0)	13 (6.1)	9 (4.2)
Others’ external stigma toward tuberculosis
8.1. I have witnessed HCWs who are suspected of having tuberculosis being stigmatized in this hospital	1	75 (35.0)	91 (42.5)	36 (16.8)	12 (5.6)
8.3. I have witnessed some HCWs in this hospital avoiding contact with coworkers who they think may have tuberculosis	3	66 (30.7)	96 (44.7)	42 (19.5)	11 (5.1)
8.6. Some HCWs in this hospital would not want to eat or drink with a coworker who they think has tuberculosis	6	60 (27.8)	73 (33.8)	65 (30.1)	18 (8.3)
8.9. Some HCWs in this hospital are stigmatized when coworkers find out that they have gone for tuberculosis screening	9	68 (31.8)	83 (38.8)	52 (24.3)	11 (5.1)
8.11. I have noticed that some other HCWs in this hospital feel uncomfortable to work near coworkers with tuberculosis	11	55 (25.6)	82 (38.1)	67 (32.1)	11 (5.1)
Respondent’s external stigma toward tuberculosis
8.7. I would feel comfortable working side by side with a coworker after she/he has been on tuberculosis treatment	7	14 (6.5)	23 (10.7)	102 (47.4)	76 (35.3)
8.10. I do not want to work together with coworkers who have tuberculosis	10	102 (47.0)	92 (42.4)	16 (7.4)	7 (3.2)
8.13. I am afraid of coworkers with tuberculosis	13	93 (42.9)	103 (47.5)	16 (7.4)	5 (2.3)
8.14. HCWs with tuberculosis probably also have HIV	14	95 (44.0)	96 (44.4)	19 (8.8)	6 (2.8)
8.15. I do not want to eat or drink in the same room as a coworker who has tuberculosis	15	98 (45.4)	95 (44.0)	16 (7.4)	7 (3.2)
Respondent’s internal stigma toward tuberculosis
8.2. If I was diagnosed with tuberculosis, I would worry that my coworkers may think I have also got HIV	2	52 (24.3)	80 (37.4)	60 (28.0)	22 (10.3)
8.4. If I was diagnosed with tuberculosis, I would not need to feel shame	4	11 (5.1)	24 (11.2)	103 (47.9)	77 (35.8)
8.5. As an HCW, I would feel it was my fault if I was infected with tuberculosis	5	91 (42.5)	90 (42.1)	20 (9.3)	13 (6.1)
8.8. If I was diagnosed with tuberculosis, I would feel comfortable to tell some of my coworkers	8	5 (2.3)	29 (13.4)	105 (48.6)	77 (35.6)
8.12. If I was diagnosed with tuberculosis, I would feel alone in my workplace	12	72 (33.2)	110 (50.7)	29 (13.4)	6 (2.8)

Data are presented as No. (%).

Abbreviations: HCW, healthcare worker; HIV, human immunodeficiency virus.

Confirmatory Factor Analyses: Internal Construct Validity

A series of CFAs were carried out to test the internal construct validity of the 6 stigma scales. Table 2 shows the factor loadings for the different theory-based items onto the different stigma scales. Items were removed from the scale if the loading (ie, the standardized regression coefficient) was <0.40. Table 2 also demonstrates the model fit of the different scales.

Table 2.

Seven-Factor Scales, Factor Loadings (α Reliability Estimate if Item Was Removed), Goodness-of-Fit Indices, and α Reliability Estimates

Itema	HIVOES	HIVOES	HIVFightOES	HIVRES	HIVRIS	TBOES	TBRES	TBRIS
6.1	0.538 (0.560)	0.544 (0.740)
6.4	0.654 (0.544)	0.667 (0.670)
6.12	0.729 (0.535)	0.725 (0.691)
6.13	0.728 (0.550)	0.722 (0.687)
6.14	0.027 (0.659)	/	/
6.15	−0.043 (0587)		0.759 (0.585)
6.16	−0.037 (0.601)		0.550 (0.663)
6.17	−0.228 (0.549)		0.670 (0.634)
6.18	−0.228 (0.583)		0.507 (0.687)
6.2				0.564 (0.756)
6.3				0.457 (0.774)
6.5				0.744 (0.724)
6.6				0.763 (0.723)
6.7				/
6.8				/
6.9				/
6.10				0.644 (0.765)
6.11				0.619 (0.744)
7.1					/
7.2					0.568 (0.742)
7.3					0.572 (0.733)
7.4					−0.483 (0.753)
7.5					0.629 (0.729)
7.6					0.465 (0.760)
7.7					0.640 (0.728)
7.8					0.627 (0.732)
8.1						0.698 (0.820)
8.3						0.780 (0.801)
8.6						0.670 (0.829)
8.9						0.710 (0.817)
8.11						0.763 (0.803)
8.7							/
8.10							0.841 (0.813)
8.13							0.887 (0.795)
8.14							0.602 (0.888)
8.15							0.830 (0.812)
8.2								0.663 (0.530)
8.4								/
8.5								0.453 (0.634)
8.8								−0.451 (0.521)
8.12								0.703 (0.627)
Goodness-of-fit indices
RMSEA	0.159	0.153	0.000	0.000	0.019	0.086	0.000	0.000
CFI	0.492	0.914	1.000	1.000	0.994	0.965	1.000	1.000
TLI	0.323	0.743	1.000	1.000	0.991	0.929	1.000	1.000
SRMR	0.134	0.044	0.012	0.025	0.036	0.033	0.009	0.012
Reliability
α	0.605	0.754	0.706	0.781	0.768	0.846	0.866	0.650

Abbreviations: CFI, comparative fit index; HIV, human immunodeficiency virus; HIVFightOES, colleagues fighting other colleagues’ external HIV stigma; HIVOES, colleagues’ external HIV stigma; HIVRES, respondent’s external HIV stigma; HIVRIS, respondent’s internal HIV stigma; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual; TB, tuberculosis; TBOES, colleagues’ external tuberculosis stigma; TBRES, respondent’s external tuberculosis stigma; TBRIS, respondent’s internal tuberculosis stigma; TLI, Tucker Lewis index.

^aItems listed in Table 1.

The first theory-based scale intended to measure the stigmatizing attitudes toward HIV of the healthcare workforce surrounding the respondent (HIVOES). The CFA resulted in a poor model fit, with 5 items loading insufficiently onto the stigma factor. When looking at the pattern of loadings, it became evident that the scale was measuring different aspects of external stigma by coworkers, namely (1) the stigmatizing attitudes of the colleagues (eg, “I have noticed that some of my coworkers in this hospital look down on HCWs who they think may be HIV infected”) and (2) the actions of colleagues against stigmatizing by other colleagues (eg, “I have heard that some HCWs educate coworkers who stigmatize people living with HIV”). These items are not “positively” and “negatively” phrased items measuring the same construct, as we did not detect such method effect associated with negatively and/or positively worded items using the Marsh and Grayson correlated traits, correlated methods (CTCM) framework [56, 57]. The SEM of the wording effect as a latent trait did not fit the data well, while the loadings of the positively and negatively worded items onto the stigma factor were still too low. Consequently, we created 2 factors (HIVOES and HIVFightOES [reflecting colleagues fighting other colleagues’ external HIV stigma toward HIV]), each consisting of 4 items. The scale measuring the coworkers’ stigma (HIVOES) still only displayed a borderline fit, with the SRMR and CFI displaying acceptable fit and the 2 other measures suggesting poor fit. The scales measuring the actions of the coworkers against stigma displayed an excellent fit to the data. Item 14 (“Other HCWs think it is worthwhile for the hospital to invest in the career development of HIV-positive HCWs”) did not load onto any of the 2 factors and was omitted from any further analyses.

Nine items were piloted for the second scale measuring the respondent’s perceptions, attitudes, and behaviors toward other HIV-infected HCWs in the hospital (HIVRES). The factor loadings indicated that all but 3 items loaded well onto the overarching factor. These items were omitted from further analysis (“In my opinion, HCWs living with HIV should probably feel shame”; “HCWs who have HIV should not feel guilty about it”; and “Most HCWs with HIV have many sexual partners”). The resulting scale, consisting of 6 items, fit the data well.

The final HIV scale measures respondents’ internal stigma (HIVRIS)—the perceptions, attitudes, and behaviors of respondents toward themselves, as well as stigma that they perceive or anticipate as being directed toward them from other HCWs. The original scale, comprising 8 items, displayed a good fit, but 1 item did not load onto the factor (“If I had HIV, I would feel uncomfortable disclosing to some of my coworkers” [λ = 0.230]). This item was omitted from the scale, which resulted in a 7-item factor, displaying a good fit to the dataset.

In accordance with the theoretical framework, the first tuberculosis scale measures perceptions, attitudes, and behaviors that respondents witness being enacted, or perceive as existing, among other HCWs in the hospital (TBOES). The original 5-item structure displayed an acceptable fit. All 5 items loaded well onto the theory-driven factor.

The second tuberculosis scale assessed the respondents’ perceptions, attitudes, and behaviors toward other tuberculosis-infected coworkers in the hospital (TBRES). The original 5-item scale displayed a good fit, but the first item (“I would feel uncomfortable working side by side with a coworker after he/she has been on tuberculosis treatment”) did not load sufficiently onto the factor (λ = 0.161). Therefore, this item was deleted from the any further analyses. The resulting factor, comprising 4 items, displayed an excellent fit to the data.

The final tuberculosis scale to be tested measured the respondent’s internal stigma. The original structure fit the data well, but 1 item (“If I was diagnosed with tuberculosis, I would not need to feel shame”) did not sufficiently load onto the factor (λ = −0.395). After deleting this item, the resulting 4-item solution fit the data well and displayed sufficiently high loadings of the different items onto the factor.

Structural Equation Modeling: External Construct Validity

In a first step, we performed a CFA incorporating all stigma scales—after item reduction—simultaneously. We also assessed the correlations between the different types of stigma as measured by our scales (Table 3). Employing Hu and Bentler’s [50] criteria, the total CFA model fit the data well (RMSEA = 0.049; CFI = 0.888; TLI = 0.876; SRMR = 0.063). All items sufficiently load onto the intended stigma factor (λ > 0.400). The correlations between the different types of stigma are displayed in Table 3. We immediately see significant and moderate to very strong correlations between equivalent scales measuring the same type of stigma for HIV and tuberculosis. Evidently, these theory-based scales contain similar items. However, the research team had explicitly avoided including the identical phrasings across these equivalent scales. Correlating the error variances of such identically phased items to control for a wording effect is thus not needed. Nevertheless, we conducted a χ² difference test to assess whether the very strong correlation between the respondents’ internal tuberculosis and HIV stigma meant that these 2 scales were measuring the same concept: The test revealed a significant difference between the 2 models (2 separate constructs vs 2 perfectly correlated constructs) demonstrating the discriminant validity of the scales. Second, we see very strong correlations between the respondents’ reported internal tuberculosis stigma and the external tuberculosis-related stigma measures. We also see a strong correlation between the respondents’ reported internal HIV stigma and the HIV-related perceptions, attitudes, and behaviors that respondents witnesses being enacted, or perceive as existing, among other HCWs (referred to also as “colleagues”) in the hospital (HIVOES). Each time we demonstrated the discriminant validity using a χ² difference test. Finally, we see that the respondents’ perception of the actions by coworkers against HIV stigma is negatively correlated with almost all other stigma measures, indicating that combating stigma on the work floor is associated with less-stigmatizing attitudes and with lower perceived stigma by others.

Table 3.

Estimated Correlation Matrix for the Latent Constructs

Construct	HIVOES	HIVFightOES	HIVRES	HIVRIS	TBOES	TBRES	TBRIS
HIVOES	1
HIVFightOES	−0.110	1
HIVRES	0.255**	−0.441***	1
HIVRIS	0.654**	−0.394***	0.484***	1
TBOES	0.791***	−0.227*	0.255**	0.854***	1
TBRES	0.334***	−0.269**	0.494***	0.669***	0.653***	1
TBRIS	0.689***	−0.345**	0.459***	0.979***	0.955***	0.862***	1

Abbreviations: HIV, human immunodeficiency virus; HIVFightOES, colleagues fighting other colleagues’ external HIV stigma; HIVOES, colleagues’ external HIV stigma; HIVRES, respondent’s external HIV stigma; HIVRIS, respondent’s internal HIV stigma; TB, tuberculosis; TBOES, colleagues’ external tuberculosis stigma; TBRES, respondent’s external tuberculosis stigma; TBRIS, respondent’s internal tuberculosis stigma.

^*P < .05,

^**P < .01,

^***P < .001.

Reliability Analysis After Item Reduction

The resulting factor solutions were subjected to a reliability analysis. All but 1 of the final scales displayed good reliability, with α ranging between 0.706 for HIVFightOES and 0.866 for TBRES. Only the scale measuring the respondent’s internal stigma toward tuberculosis displayed a Cronbach α of 0.650. In all scales, if final items were deleted, the α for that domain was lowered, demonstrating the significant contribution of each item to the reliability of the scale (Table 2).

External Construct Validity

To test the external construct validity of the 7 selected scales, we tested 7 structural equation models. Each model assessed the correlations between the stigma scale and (1) the level of confidentiality in the occupational health unit, (2) the respondent’s knowledge of HIV transmission, and (3) the respondent’s knowledge of tuberculosis. Each model controlled for the impact of sex, education, and occupational role on stigma; the perceived confidentiality; and the HIV/tuberculosis-related knowledge.

The first stigma scale—measuring HIV-related perceptions, attitudes, and behaviors that respondents witness being enacted, or perceive as existing, among other HCWs (referred to also as “colleagues”) in the hospital (HIVOES)—was negatively correlated with the perceived level of confidentiality (r = 0.364; P < .001). The model fit the data well (RMSEA = 0.066; CFI = 0.922; TLI = 0.868; SRMR = 0.052). As was expected, the scale was not significantly correlated with the knowledge of the respondent. The corresponding scale on tuberculosis (TBOES) displayed a similar pattern; it was weakly and negatively correlated (r = 0.223, P < .01) with the level of confidentiality in the facility after controlling for the education and sex of the respondent (RMSEA = 0.057; CFI = 0.941; TLI = 0.909; SRMR = 0.045). Finally, the related scale measuring the fighting of HIV stigma by coworkers in the facility was positively correlated with the level of confidentiality (r = 0.256; P < .01), while this SEM fit the data well (RMSEA = 0.000; CFI = 1.000; TLI = 1.000; SRMR = 0.036).

The scale measuring respondents’ own HIV-related perceptions, attitudes, and behaviors toward other HCWs/colleagues in the hospital (HIVRES) was, as expected, weakly but significantly correlated with the HIV-related knowledge of the respondent (r = −0.155; P < .05), with better-informed workers displaying less-stigmatizing attitudes than ill-informed colleagues (RMSEA = 0.000; CFI = 1.000; TLI = 1.000; SRMR = 0.028). Similarly, the corresponding tuberculosis stigma scale (TBRES) was also weakly and negatively correlated (r = −0.197; P < .05) with the respondent’s knowledge of HIV transmission (RMSEA = 0.000; CFI = 1.000; TLI = 1.000; SRMR = 0.026). The respondent’s knowledge of tuberculosis symptoms was not significantly associated with this HIV stigma scale. It must be noted that, as in each of the SEMs, the respondent’s educational level had a significant impact on the knowledge of HIV (r = 0.294; P < .001) and tuberculosis (r = 0.259; P < .001).

The SEM assessing the correlations with the respondent’s internal stigma fit the data well (RMSEA = 0.041; CFI = 0.932; TLI = 0.907; SRMR = 0.045). Again, the level of confidentiality in the facility was negatively correlated (r = 0.235; P < .01) with the internal HIV stigma of the respondent (RMSEA < 0.05; CFI = 0.932; TLI = 0.907). Similarly, the internal tuberculosis stigma as reported by the respondents was weakly and negatively correlated (r = −0.235; P < .05) with the level of confidentiality in the facility (RMSEA = 0.000; CFI = 1.000; TLI = 1.000; SRMR = 0.035).

DISCUSSION

The current pilot study reports on the development and psychometric properties of a series of measures of HIV and tuberculosis stigma in the healthcare setting. In accordance with the theoretical framework, the resulting 7 scales assessed (1) colleagues’ external HIV stigma, (2) colleagues’ actions against external HIV stigma, (3) respondent’s external HIV stigma, (4) respondent’s internal HIV stigma, (5) colleagues’ external tuberculosis stigma, (6) respondent’s external tuberculosis stigma, and (7) respondent’s internal tuberculosis stigma. These instruments showed good psychometric properties in terms of internal construct validity, reliability, and external construct validity. The descriptive analysis indicated that the reported levels of respondents’ internal stigma were higher than the perceived stigmatization by colleagues as well as the respondents’ external stigma.

The results of the CFAs demonstrated that the initial scale measuring the colleagues’ external HIV stigma should be divided into 2 separate scales, one measuring the intended concept and one measuring the actions taken to fight HIV-related stigma. The nonsignificant correlation between the 2 constructs confirmed this clear distinction. The internal construct validity of the other 5 stigma scales was confirmed in the separate CFAs. These analyses confirm the distinction made in the literature between internalized and externalized stigma [33, 34, 43, 58, 59]. However, the strong correlations between the colleagues’ external stigma and the respondent’s internal stigma—for both HIV and tuberculosis—indicate that the attitudes and behavior of colleagues is strongly linked to the internalized opinion of the individual toward him/herself. This confirms previous research by Greeff and Phetlhu reporting that past experiences of external stigma—witnessing or experiencing rejection, intimidation, exclusion, and isolation—influence internal stigma [55]. Future studies should further try to disentangle these 2 types of stigma by fine-tuning the different items, as the strong associations render a clear distinction between the 2 concepts difficult.

A review study by Nyblade et al indicated that there is a link between the level of confidentiality in the facility and the potential stigma that staff are expecting to experience [16]. In accordance with these findings, the level of confidentiality in the facility was negatively correlated with our respondents’ perceptions of their colleagues’ externalizing stigma. In addition, respondents’ internalizing stigma was also significantly and positively related to breaches of confidentiality in the facility. These findings support the external construct validity of these scales [16, 42]. The findings are in line with those of a recent study by Khan et al conducted in this same facility, which demonstrated that stigma and confidentiality were the 2 main barriers to the uptake of HIV counseling and testing services within occupational health units [60]. Conversely, various studies have demonstrated that people’s own stigmatizing attitudes toward others (external stigma of the respondents) are related to their knowledge of the illness [46,47]. Accordingly, the developed scales (HIVRES and TBRES) were negatively correlated with the level of HIV-related knowledge of the respondents. No association, however, was found between respondents’ external stigma with our measurement of tuberculosis-related knowledge.

The study outcomes reveal a strong link between tuberculosis-related and HIV-related stigma. The correlations between the HIV scales on the one hand and their corresponding tuberculosis scales on the other hand ranged from moderate to very strong (r > 0.9). Although the results of the χ² difference testing confirmed the discriminant validity of the different scales, it appeared difficult to clearly disentangle the 2 objects of stigma (tuberculosis and HIV). These quantitative findings confirm recent quantitative evidence of Daftary demonstrating that the confluence of the tuberculosis and HIV epidemics rendered “tuberculosis symbolic and symptomatic of HIV,” thus producing a unique, overlapping double stigma [44]. Correspondingly, Bond and Nyblade already stated that in the context of high HIV prevalence, tuberculosis stigma can no longer be thought of separately from HIV stigma [26]. These authors also called for attempts to disentangle the double stigma of tuberculosis and HIV. The current study is exactly this—an attempt to develop and validate instruments to disentangle the different types of HIV as well as tuberculosis stigma in the healthcare setting. However, the results of our analyses indicate that future research is needed to further disentangle these intricately interrelated phenomena, as the developed scales have difficulty discerning stigma toward HIV or tuberculosis among the healthcare workforce.

The strengths of this study include (1) its incorporation of different aspects of the interrelated concepts of HIV and tuberculosis stigma and (2) the application of this comprehensive theoretical framework to a very relevant population and setting: the healthcare workforce active in the facilities fighting HIV and tuberculosis. However, there were some limitations to our study. First, the current study assessed internalized and externalized stigma among the healthcare workforce. However, despite its focus on the healthcare setting, the current study did not incorporate what the Siyam’kela Project [32] called “stigma by association.” This type of secondary stigma is defined as “incidents that describe stigma against people who work or associate with HIV/AIDS-affected people” [32, 59] and is thus a relevant concept in the healthcare setting. However, as the focus of the current study is on developing scales that measure the stigmatizing attitudes of the healthcare workforce toward their fellow colleagues—because these particular stigmatizing attitudes are likely to be the primary barrier to the optimal use of occupational HIV and tuberculosis services—associated stigmas fell outside of its scope [35]. Second, the current study should be considered as a preparatory pilot study, informing future work on these instruments. The sample size was relatively small, and the results may not be applicable to alternative settings. Future research informed by our findings and executed in a range of different high-HIV-prevalence healthcare settings is needed to unarguably validate the instruments. Finally, the study opted to include the entire healthcare workforce (ie, health professionals as well as supporting staff), rendering the respondent group diverse. Future attempts to validate these instruments should incorporate this diversity by performing multigroup confirmatory factor analyses and structural equation modeling. The limited sample size did not allow this within the scope of the current study.

CONCLUSIONS

There is a great need for the use of appropriate scales in the evaluation of interventions to reduce HIV and tuberculosis stigma in the healthcare setting. Although the development of such specific scales for this context is still in its infancy, the current exploratory analyses have both practical and theoretical implications. Theoretically, the distinction between internalized and externalized stigma and the attempt to disentangle the double tuberculosis/HIV stigma can inform the further development of appropriate scales. Future studies should build on our findings to fine-tune the instruments and apply them to a larger study population. Practically, the successful development and piloting of parallel scales measuring different aspects of stigma in the healthcare setting will enable future studies to (1) identify which type of stigma (external/internal) acts as the primary barrier to the use of occupational health services; (2) develop appropriate stigma reduction programs that optimally address these main barriers [61]; and (3) scientifically assess the impact of these programs on the stigma levels as well as the mechanisms through which the programs impact the health of the healthcare workforce.