Exploring the Association Between Depression and Social and Biobehavioral HIV Risk Factors Among Female Sex Workers in Nelson Mandela Bay Municipality,South Africa

Abstract

The aim of this study was to estimate the prevalence of depression among female sex workers (FSW) in an urban coastal city in South Africa, and to explore the relationship between depression and HIV-related social and biobehavioral determinants. A cross-sectional respondent-driven sampling study was conducted with FSW (n = 410), including a sociobehavioral questionnaire, PHQ-9 (Patient Health Questionnaire-9) based assessment of depression, and biological testing for HIV and syphilis. The prevalence of HIV in the sample was 64.1%. The estimated prevalence of depression was 28.8%. Depression was associated with social vulnerability such as living alone [adjusted prevalence ratio 1.82, 95% confidence interval (CI) 1.15–2.90] and food insecurity (aPR 2.19, 95% CI 1.42–3.38). A positive syphilis test result (aPR 1.46, 95% CI 1.02–2.09) and self-reported sexually transmitted disease symptoms (aPR 1.78, 95% CI 1.29–2.46) was associated with depression, but self-reported condom use and HIV status was not. FSW were also less likely to disclose their occupational status to health care providers (aPR 0.61, 95% CI 0.42–0.89) or undergo sexually transmitted infection screening in the last 12 months if they are depressed (aPR 0.64, 95% CI 0.43–0.95). The results demonstrate that the prevalence of depression is high among FSW and that depressive symptoms are associated with social covariates and biobehavioral HIV risk factors.

Introduction

HIV disproportionately affects sex workers due to intersections between biological and behavioral risks as well as the social and structural conditions that contextualize these risks.¹ Female sex workers (FSW) globally have an estimated HIV prevalence of 12%, whereas the estimated prevalence in Sub-Saharan Africa is three times higher at 36%.² The burden of HIV is further concentrated among FSW within South Africa, with recent estimates of HIV prevalence ranging from 40% to 72%.³

Studies on sex workers have established consistent relationships between HIV and biobehavioral risk factors, such as condom use, drug use, and sexually transmitted infections (STIs), and sociostructural drivers of HIV, including stigma, police harassment, and exposure to violence.^{1

–10} However, there has been a limited study of depression and its association with these established social and biobehavioral correlates among FSW, especially within the South African context. Depression is generally high among FSW with a recent meta-analysis reporting a pooled depression prevalence of 62% [95% confidence interval (CI) 59.1–65.7].¹¹ Only two estimates of depression among sex workers in South Africa were found. A cross-sectional study using respondent-driven sampling (RDS) in the urban township of Soweto, in Gauteng Province, found that 69% of FSW had depression, while another cross-sectional study in KwaZulu-Natal Province reported a prevalence of 48%.^12,13 This is comparatively higher than the adult population in South Africa, with an estimated 10% of adults ever experiencing depression in their lifetime,^14,15 and 5% of young people living in South African informal settlements currently experiencing depression.¹⁶

Outside of Sub-Saharan Africa, there has been emerging evidence linking depression and HIV risks. In India, a cross-sectional study demonstrated that FSW who were depressed were more likely to engage in condomless sex.¹⁷ Similarly, a study in Canada found that FSW who were depressed were also more likely to use drugs and engage in condomless sex.¹⁸ In other populations, including gay men and other men who have sex with men and adolescents, the link between depression and HIV has been well established.^19

–23

Collectively, there is an emerging understanding of the intersections of depression and social and biobehavioral HIV risks among FSW in other parts of the world, but limited data from Sub-Saharan Africa and South Africa specifically. In response, the objective of this study was to estimate the prevalence of depression among FSW in the Nelson Mandela Bay Municipality, South Africa, and to explore the relationship between biobehavioral determinants, social factors, and depression.

Materials and Methods

Study context, population, design, and procedures

This cross-sectional RDS study took place between October 2014 and April 2015 in Port Elizabeth and the broader Nelson Mandela Bay Municipality in the Eastern Cape Province. The Eastern Cape has the third highest HIV prevalence in South Africa among adults 15–49 years of age, estimated at 25.2% (95% CI 19.8–31.5).²⁴ The Eastern Cape is a mostly rural province, with large industrial urban coastal cities. The study was conducted through a partnership between the Human Science Research Council (HSRC), TB HIV Care, a nonprofit South African organization, and the Johns Hopkins School of Public Health. Antiretroviral treatment (ART) initiation and management in South Africa is primarily managed by public clinics within communities. Nongovernment organizations, such as TB HIV Care, support the existing health system through the provision of sex worker focused peer-led outreach, human rights support, health screening, family planning, STI treatment, pre-exposure prophylaxis, condom distribution, and in some cases ART initiation and management.^3,4

The methods used for this study have been described in detail previously.^25,26 Participation in the study was restricted to current cisgender FSW (defined as making the majority of their income through sex work), who resided in the municipality and who were 18 years or older. RDS has been broadly described, but briefly, study coupons were distributed to a diverse set of nine well-connected seeds or initial recruits in the FSW community, who were each given three coupons to distribute to other FSW in their network, who, upon participation, each also received three coupons for distribution. This process continued until the proposed sample size estimation and outcome equilibrium were met. More details on RDS and its appropriateness in the study of key populations may be found elsewhere.^27,28 The sample size was set to allow for the estimation of HIV prevalence within a +/−5% margin of error—assuming that the prevalence was 40%–70%. Participants were remunerated ZAR50 (South African Rand) for redeeming their coupons and ZAR25 for each additional participant successfully recruited. After an interviewer-administered questionnaire was conducted, a study nurse administered HIV tests as well as a rapid plasma regain (RPR) test for syphilis for all study participants. The initial HIV test was performed either using the First Response Rapid HIV Test or the Advanced Quality HIV Test (Xiamen, China). All participants with a reactive test result had a confirmatory test using the Abon-Alere Rapid HIV Test (Hangzhou, China). Each participant provided written informed consent.

Ethics review

The protocol was approved by the HSRC Review Board (REC 3/18/06/14) and the JHU School of Public Health Review Committee (No. 00005829). Approval for this analysis was received from the University of Cape Town Human Research Ethics Committee (IRB0001938, REC-210208-007).

Outcome measure

Depression was the primary outcome of this analysis, which was measured with the Patient Health Questionnaire-9 (PHQ-9) instrument. This is an established instrument that is used within clinical settings as a screening and diagnostic tool and is based on the diagnostic criteria for major depressive disorder within the Diagnostic Manual of Mental Disorders, fifth edition (DSM-V).²⁹ The instrument consists of nine questions with a possible score of 0 to 3 for each of the questions and a total maximum score of 27. The PHQ-9 instrument has been validated within the South African context among HIV-positive adults.³⁰ A cutoff of >9 (10–27) was used as an indication of clinically significant depression, the threshold at which psychotherapy and/or medication is indicated. This threshold of interest was used in similar studies among key populations.^31,32 A lower threshold of >4 (5–27) could have been used, but this would have included those with only mild depressive symptoms, which was outside the scope of this study.²⁹ There were two participants with missing values on some of the PHQ-9 instrument items. The mean score on the rest of the PHQ-9 items was used to impute their missing values.

Measures of covariates

All social covariates and exposures were self-reported. Social stigma (three-items), health care stigma (five-items), and police harassment (six-items) indices were based on a previously validated comprehensive stigma index covering many types of stigma.³³ The scores on these stigma measures were transformed into categorical variables based on variable distribution. The social stigma and police harassment index scores were divided into tertiles and coded “low,” “medium,” and “high,” whereas the health care stigma index score was dichotomized into “exposed” and “not exposed.” The Cronbach's alpha was used to select variables with the highest internal consistency for these indices. One item was dropped from the social stigma index, namely “are there safe spaces for you to socialize in the area?” as it reduced the internal consistency of the index. The safe space variable was kept as an individual correlate of interest. Income was transformed from a numeric variable (ZAR) to a binary variable based on the median income value. The rape variable was composed from self-reports on when last a participant experienced rape and if they experienced rape as a child: “No exposure,” “Exposed >12 months prior, but never as a minor,” “Exposed >12 months prior and exposed as a minor,” “Exposed within the last 12 months, but never as a minor,” and “Exposed within the last 12 months and exposed as a minor.”

All participants received a screening and confirmatory HIV test and were asked about their HIV status and ART status before the administration of the HIV test. The HIV status variable was derived from biological testing combined with self-reporting of known HIV-positive status and ART status: “HIV negative,” “newly diagnosed HIV positive,” “Known HIV positive, never initiated on ART,” “Known HIV positive, initiated on ART, not on ART,” and “Known HIV positive on ART.” We did not have information on prior HIV diagnosis for nine participants, so they were not included in the analyses. The syphilis test outcome variable was based on the RPR test result. Apart from the syphilis test result, participants were also asked if they have experienced any symptoms of an STI in the last 12 months and if they had received any STI screening in the last 12 months. Condom use was self-reported based on condom use within the last 12 months. For drug use, the participant could answer: “no,” “yes, within the last 12 months,” or “yes, but not within the last 12 months.” Binge drinking was defined as those who drink six or more standard drinks per occasion—at least once a week. An interaction term between self-reported binge drinking and drug use within the last 12 months was created to assess if those who drink excessively and have used drugs recently are more at risk of depression.

Statistical analyses

RDS-adjusted weights were applied to extrapolate population parameter estimates and 95% bootstrapped confidence intervals (BCI) for the demographic and depression prevalence statistics in Table 1. Data were imputed for variables with missing values using the mean of the sample. Data were analyzed in Stata 14.2 (College Station, TX) with the Stata RDS add-on package.³⁴ Adjusted prevalence ratios (aPR) for the multivariate analysis were estimated with a robust Poisson regression model.³⁵ We provide descriptive statistics for the social covariates and biobehavioral covariates in Tables 2 and 3, respectively. In the first adjusted multivariate analysis in Table 4 we focus on the association between depression and social covariates and in the second model in Table 5 we focus on the association between depression and biobehavioral risk factors. In both of the multivariate adjusted analyses, we controlled for demographic factors: age, income, and number of years practicing sex work. The Cronbach's alpha was used to measure the reliability of the PHQ-9 instrument as well as the police harassment and the social and health care stigma indices.

Table 1.

Participant Characteristics (n = 410)

	Crude (sample statistics)		RDS population estimate
	n	%	% (BCI)^a	Homophily
Age, median (IQR)	28 (24–33)
18–24	122/410	29.8	37.7 (3.0–45.5)	0.14
25–34	205/410	50.0	43.9 (37.0–50.7)	0.15
35+	83/410	20.2	18.4 (12.7–24.1)	0.15
Education
Grade 1–7	50/410	12.2	12.2 (8.0–16.4)	0.04
Grade 8–11	300/410	73.2	72.4 (66.6–78.1)	0.08
Grade 12 and higher	60/410	14.6	15.4 (10.7–20.2)	0.05
Dependents
None	150/410	36.6	35.8 (28.9–42.7)	0.12
One or more	260/410	63.4	64.2 (57.3–71.1)	0.07
Income monthly, median, SD (IQR)	2,100, 3,297 (1,680–3,780)
Under ZAR2100	151/410	36.8	37.8 (31.1–44.4)	0.13
Over ZAR2100	259/410	63.2	62.2 (55.6–68.9)	0.15
HIV status^b,c
Unknown HIV⁺	43/401	10.7 (8.0–14.2)
Known HIV⁺	214/401	53.4 (48.4–58.2)
HIV⁻	144/401	35.9 (31.3–40.7)
PHQ-9 depression categories,^d median, SD (range)	4, 6.29 (0–27)
Minimal (0–4)	218/410	53.2	55.9 (49.0–62.8)	0.11
Mild (5–9)	69/410	16.8	15.6 (11.0–20.1)	0.08
Moderate (10–14)	74/410	18.0	16.5 (11.3–21.7)	0.06
Moderately Severe (15–19)	38/410	9.3	8.8 (5.3–12.4)	0.05
Severe (20–27)	11/410	2.7	3.3 (0.1–6.6)	-0.03
Depressive symptoms^e
Minimal to mild (0–9)	287/410	70.0	71.2 (64.7–77.5)	0.04
Moderate to severe (10–27)	123/410	30.0	28.8 (22.4–35.2)	0.06

Bootstrapped confidence interval = 1,000 reps.

RDS extrapolation to population was not possible, as there were too many missing values.

HIV status derived from HIV test result and self-report of HIV status before this HIV test.

Missing value imputed for participants using mean.

Classification of depressive symptoms used for analysis (PHQ-9 cut-off >9).

IQR, interquartile range; PHQ-9, Patient Health Questionnaire-9; RDS, respondent-driven sampling; SD, standard deviation.

Table 2.

Social Covariates

	N (%)	Moderate-to-severe depression, n (%)
Social vulnerability
Homeless
No	398/410 (97)	115/398 (28.8)
Yes	12/410 (3)	8/12 (66.7)
Living arrangements
Living with family	243/410 (59.3)	59/243 (24.3)
Living with friends/colleagues	135/410 (32.9)	48/135 (35.6)
Living alone	32/410 (7.8)	16/32 (50)
Currently engaged in sex work to feed family
No	120/410 (29.3)	18/120 (15.0)
Yes	290/410 (70.7)	105/290 (36.2)
Safe space in community
No	119/410 (29.0)	47/119 (39.5)
Yes	291/410 (71.0)	76/291 (26.1)
Stigma and harassment
Health care stigma^a
No	340/410 (83.0)	89/340 (26.2)
Yes	70/410 (17.0)	34/70 (48.6)
Tell health care provider about sex work status
No	298/409 (72.8)	99/298 (33.2)
Yes	111/409 (27.2)	24/111 (21.6)
Police harassment^b
Low	204/410 (49.8)	58/204 (28.4)
Medium	158/410 (38.5)	44/158 (27.8)
High	48/410 (11.7)	21/48 (43.7)
Social stigma^c
Low	269/410 (65.6)	69/269 (25.6)
Medium	99/410 (24.2)	38/99 (38.3)
High	42/410 (10.2)	16/42 (38.0)
Sexual and physical violence
Raped
No exposure	253/410 (61.7)	66/253 (26.1)
More than 12 months but never as minor (<18)	46/410 (11.2)	10/46 (21.7)
More than 12 months and/or as minor (<18)	25/410 (6.1)	11/25 (44.0)
Within last 12 months but never as minor (<18)	69/410 (16.8)	24/69 (34.8)
Within last 12 months and as minor (<18)	17/410 (4.1)	12/17 (70.6)
Physical violence
No exposure	157/410 (38.3)	40/157 (25.5)
>12 months ago	94/410 (22.9)	24/94 (25.5)
Within 12 months	159/410 (38.8)	49/159 (30.8)

Health care stigma index, have you ever: felt afraid to go to health care services; avoided health care services; been denied health services; felt that you were not treated well in a health center; heard a health care worker gossip about you?

Police harassment index, have you: felt the police refused to protect you; avoided carrying condoms because of police; heard of police confiscating condoms; ever been arrested; have a bad relationship with the police; been harassed by the police?

Social stigma index: ever felt excluded from family gatherings; ever felt family members made discriminatory remarks or gossiped about you because; ever felt rejected by friends, because you practice sex work?

Table 3.

Biobehavioral Covariates

	N (%)	Moderate-to-severe depression, n (%)
HIV status
HIV negative	144/401 (35.9)	47/144 (32.6)
HIV positive newly diagnosed	43/401 (10.7)	13/43 (30.2)
HIV positive not initiated on ART^a	95/401 (23.7)	27/95 (28.4)
HIV positive initiated ART, not on ART currently	17/401 (4.2)	7/17 (41.2)
HIV positive on ART	102/401 (25.4)	26/102 (25.5)
Syphilis
No	315/396 (79.5)	86/315 (27.3)
Yes	81/396 (20.5)	29/81 (35.8)
Condom use with clients
Never/Almost never	18/406 (4.4)	4/18 (22.2)
Sometimes	163/406 (40.1)	46/163 (28.2)
Almost always/Always	225/406 (55.4)	70/225 (31.1)
STI screening in last 12 months
No	300/410 (73.2)	97/300 (32.3)
Yes	110/410 (26.8)	26/110 (23.6)
STI symptoms in last 12 months
No	255/410 (62.2)	62/255 (24.3)
Yes	155/410 (37.8)	61/155 (39.4)
Drug use in last 12 months
No	315/410 (76.8)	87/315 (27.6)
Yes	95/410 (23.2)	36/95 (37.9)
Binge drinking (current)
No	347/410 (84.6)	105/347 (30.3)
Yes	63/410 (15.4)	18/63 (28.6)
Drug use and binge drinking (interaction)
No	396/410 (96.6)	115/396 (29.0)
Yes	14/410 (3.4)	8/14 (57.1)

South African ART policy at the start of the study had an ART eligibility criteria of CD4 ≤ 350, which during the study changed to ≤500 January 1, 2015.

ART, antiretroviral treatment; STI, sexually transmitted infection.

Table 4.

Association Between Depression and Social Covariates

	PR (95% CI)	aPR (95% CI)^a
Social vulnerability
Homeless
No	Ref.	Ref.
Yes	2.30 (1.5–3.54)	2.16 (1.38–3.40)
Living arrangements
Living with family (husband/children/relatives)	Ref.	Ref.
Living with friends/boyfriends/colleagues	1.46 (1.06–2.01)	1.38 (1.0–1.91)
Living alone	2.05 (1.36–3.10)	1.82 (1.15–2.90)
Currently engaged in sex work to feed family
No	Ref.	Ref.
Yes	2.41 (1.54–3.80)	2.19 (1.42–3.38)
Safe space in community
No	Ref.	Ref.
Yes	0.66 (0.50–0.89)	0.77 (0.56–1.05)
Stigma and harassment
Health care stigma
No	Ref.	Ref.
Yes	1.86 (1.38–2.51)	1.35 (0.94–1.93)
Tell health care provider about sex work status
No	Ref.	Ref.
Yes	0.65 (0.45–0.97)	0.61 (0.42–0.89)
Police harassment
Low	Ref.	Ref.
Medium	0.97 (0.70–1.36)	0.99 (0.71–1.37)
High	1.53 (1.04–2.26)	1.24 (0.85–1.81)
Social stigma
Low	Ref.	Ref.
Medium	1.49 (1.08–2.06)	1.34 (0.96–1.88)
High	1.48 (0.96–2.29)	1.28 (079–2.07)
Sexual and physical violence
Raped
No exposure	Ref.	Ref.
More than 12 months but never as minor (<18)	0.83 (0.46–1.49)	0.75 (0.40–1.40)
More than 12 months and/or as minor (<18)	1.69 (1.02–2.75)	1.43 (0.85–2.41)
Within last 12 months but never as minor (<18)	1.33 (0.90–1.95)	1.03 (0.70–1.50)
Within last 12 months and as minor (<18)	2.70 (1.86–3.92)	1.73 (1.01–2.94)
Physical violence
No exposure	Ref.	Ref.
>12 months ago	1.0 (0.64–1.55)	0.92 (0.61–1.39)
Within 12 months	1.45 (1.04–2.03)	1.02 (0.70–1.7)

Adjusted model controls for age, years practicing sex work, income, and all other social covariates.

aPR, adjusted prevalence ratio; CI, confidence interval.

Table 5.

Association Between Depression and Biobehavioral Covariates

	PR (95% CI)	aPR (95% CI)^a
HIV status
HIV negative	Ref.	Ref.
HIV positive newly diagnosed	0.92 (0.55–1.54)	0.99 (0.60–1.63)
HIV positive not initiated on ART	0.87 (0.58–1.29)	0.82 (0.53–1.26)
HIV positive initiated ART, not on ART currently	1.26 (0.68–2.33)	1.44 (0.74–2.79)
HIV positive on ART	0.78 (0.51–1.17)	0.92 (0.58–1.46)
Syphilis
No	Ref.	Ref.
Yes	1.31 (0.93–1.85)	1.46 (1.02–2.09)
Condom use with clients
Never/Almost never	Ref.	Ref.
Sometimes	1.26 (0.52–3.18)	1.38 (0.55–3.43)
Almost always/Always	1.4 (0.57–3.39)	1.30 (0.54–3.13)
STI screening in last 12 months
No	Ref.	Ref.
Yes	0.73 (0.50–1.06)	0.64 (0.43–0.95)
STI in last 12 months
No	Ref.	Ref.
Yes	1.62 (1.20–2.16)	1.78 (1.29–2.46)
Drug use in last 12 months
No	Ref.	Ref.
Yes	1.37 (1.00–1.88)	1.10 (0.76–1.61)
Binge drinking at least weekly
No	Ref.	Ref.
Yes	0.94 (0.62–1.44)	0.63 (0.34–1.19)
Drug use and binge drinking (interaction)
No	Ref.	Ref.
Yes	1.97 (1.22–3.18)	2.49 (1.02–6.05)

Adjusted model controls for age, years practicing sex work, income and all other biobehavioral covariates.

Results

Demographics and HIV status

Table 1 presents participants' demographic characteristics and HIV status. A total of 410 participants were recruited through 9 initial seeds (participants). The median number of referral waves was 6 with an interquartile range (IQR) of 4–9, which met the requirements of convergence. The median network size of a study participant was 10 (IQR 6–15). Homophily was low (<0.2), which indicates that the groups are networked. The median age was 28 (unadjusted IQR 24–33). Most participants only had high school level education (RDS 72.4%, 95% BCI 66.6–78.1). More than 60% of participants had one or more dependent (RDS 64.2%, 95% BCI 55.6–68.9). The median income for the participants was ZAR2,100 (IQR 1,690–3,780). A total of 64.1% (n = 257) tested HIV positive.

Depression

The PHQ-9 instrument was found to have a high internal consistency with a Cronbach's alpha of 0.87. The median PHQ-9 score was four with an IQR of 0–11. Over 50% of the participants did not have depression (RDS 55.9%, 95% BCI 49–62.8)—as indicated by a PHQ-9 score of <5. Approximately 30% of FSW in the sample and 28.8% (95% BCI 22.4–35.2) of the FSW population in Nelson Mandela Bay are estimated to have clinically significant depression based on a PHQ-9 cutoff of >9.

Social covariates

Table 2 provides descriptive statistics on the social covariates. The social covariates are organized in three categories, namely (1) social vulnerability, which includes living arrangements and food security, (2) stigma and harassment, and (3) violence, which includes both physical and sexual violence.

The majority of participants were living with their families, indicating some level of familial connectedness and housing stability, while a small number of participants (7.8%, n = 32) were living alone. Homelessness was very low at 3% (n = 12). Food insecurity was high with 71% (n = 290) of participants saying that they practice sex work to feed their families. Most of the participants (71%, n = 291) say they have access to spaces within their communities.

There were high levels of exposure to social stigma and police harassment reported, with 34.4% (n = 141) of participants reporting medium to high levels of social stigma and 50.2% (n = 206) indicating medium to high levels of police harassment. Health care stigma was low with only 17% (n = 70) exposed. The Cronbach's alpha for the social stigma index was 0.78, whereas the reliability for the police harassment index (0.68) the perceived health care stigma index (0.69) was relatively lower.

Exposure to physical and sexual violence was high. In total, 157 (38%) of the sample had ever been raped. Of these, 10% (n = 42) had experienced rape as a minor (<18 years old) and a further 20.9% (n = 86) experienced rape within the last 12 months. Exposure to physical violence was high at 61% (n = 253) reporting ever being exposed and 38% (n = 159) reporting that they were exposed within the last 12 months.

Biobehavioral covariates

Table 3 presents descriptive statistics on the biobehavioral covariates. The majority of the participants were HIV positive (64.1%, n = 257) and of those, 83.2% (n = 214) knew their status and 39.6% (n = 102) were currently on ART. Very few of the participants reported never or almost never using a condom (4.4%, n = 18). The prevalence of STIs were high with 20.5% (n = 81) testing positive for syphilis and 37.8% (n = 155) reporting that they had experienced symptoms of an STI within the last 12 months. However, only a minority reported being screened for any STI within the last 12 months (26.8%, n = 110). Less than a quarter (23.2%, n = 95) of the participants reported drug use within the last 12 months, whereas only 15.4% (n = 63) reported binge drinking.

Association between depression and social and biobehavioral covariates

Table 4 presents the association between the depression and the social covariates. Participants who were living with people other than direct family members such as friends, boyfriends, or colleagues (aPR 1.38, 95% CI 1.0–1.91) and those staying alone (aPR 1.82, 95% CI 1.15–2.90) were more likely to be depressed than those who lived with their families (children/husband/other relatives). Being homeless was a strong predictor of depression (aPR 2.16, 95% CI 1.38–3.40). Participants citing food insecurity as their motivation to practice sex work were more than twice as likely to be depressed (aPR 2.19, 95% CI 1.42–3.38), compared with those who did not cite food insecurity.

Both stigma and police harassment were associated with depression. Those who reported high levels of perceived health care stigma were more likely to be depressed (aPR 1.35, 95% CI 0.94–1.93) compared with those who reported low levels of stigma, but this was only significant in the unadjusted model. Those who were depressed were also less likely to have disclosed their occupational status to a health care provider (aPR 0.61, 95% CI 0.42–0.89). Participants reporting high police harassment (aPR 1.24, 95% CI 0.85–1.81) as well as those reporting high social stigma (aPR 1.28, 95% CI 0.79–2.07) were more likely to be depressed compared with those reporting low exposure but this was not statistically significant at the 95% level of confidence.

Sexual violence was associated with depression whereas physical violence was only associated with depression in the unadjusted model. Participants who were raped when they were minors as well as within the last 12 months (n = 17) were 70% (aPR 1.73, 95% CI 1.01–2.94) more likely to be depressed compared with those who have never been raped.

Table 5 provides an overview of the association between depression and biobehavioral risk factors of HIV. There was no statistically significant association between condom use and depression. HIV status and ART status also had no relationship with depression. A positive syphilis test result (aPR 1.46, 95% CI 1.02–2.09) as well as self-reported STIs within the last 12 months (aPR 1.78, 95% CI 1.29–2.46) were, however, associated with depression. Those who were depressed were also less likely to have undergone STI screening within the last 12 months (aPR 0.64, 95% CI 0.43–0.95). There was no association between depression and binge drinking or drug use, but those who reported both drug use and binge drinking were almost two and a half times as likely to be depressed compared with those who indicated neither or only one (aPR 2.49, 95% CI 1.02–6.05).

Discussion

In general, we found high levels of depression in our sample. We also found a strong association between depression and many of the included social covariates. While there was no association between depression and HIV status, there was a relationship between depression and some of the included biobehavioral risk factors, most notably self-reported STIs and laboratory-confirmed syphilis results. Of particular interest is the overlap among depression, living arrangements, and food security. Social and economic well-being is expectedly linked with mental well-being. Overall, there is evidence that FSW in South Africa are exposed to interrelated HIV risks, social vulnerability, and mental health issues.

Our estimate adds to the growing evidence that depression is prevalent among FSW in South Africa. While the prevalence of depression among FSW (29%) in our study was three times higher compared with HIV-positive females attending an ART clinic in South Africa (10%),³⁶ our results were notably lower than the two other South African studies focusing on FSW. It was ∼40% lower than the study in Soweto, Gauteng, and 20% lower than the study in KwaZulu-Natal.^12,13 The same measurement instrument (PHQ-9) and cutoff value was used to compare our results with the KwaZulu-Natal-based study. The instrument used in the Gauteng study was, however, different to the one used in our study, and the cutoff used would include those experiencing mild depression, unlike our cutoff, which excludes mild depression. This may potentially explain some of the variance, but likely not all of it. The social milieu would also be very different among the urban township of Soweto, Gauteng, and the smaller coastal municipality in which our study is based.

Our study reported positive associations between depression and biobehavioral risk factors, whereas the other two South African studies did not report any such association. No association was found among depression, self-reported condom use, or HIV status. Many other similar studies, however, found a relationship between depression and consistent condom use among key populations.^17,18,37 This study did observe a positive association between depression and syphilis, as well as self-reported STIs symptoms as an indirect measure of condom use. We also found that FSW were less likely to screen for STIs or tell their health care providers they are engaging in sex work if they were depressed. The presence of depressive symptoms could potentially reduce the likelihood of STIs being diagnosed and treated.

Social factors were also linked with depression. FSW living arrangements appear to greatly influence the likelihood of depression, most notably food insecurity and isolation appear to increase the risk of depression. We also found high levels of exposure to violence and sexual assault and rape within the sample, which concurs with other studies among FSW in South Africa¹² and elsewhere.^37,38 A risk reduction framework alone might not be suitable in dealing with much of the aftermath of sexual and physical violence on risk behavior and depression. Therapeutic interventions for survivors of physical and sexual trauma may also need to be considered.

As this study was cross-sectional, we cannot determine causality. This study did not provide an in-depth analysis of the causal mechanism through which depression may act as a mediator or moderator between biobehavioral and social risk factors and HIV outcomes. It is also possible that STI and HIV outcomes influence mental health outcomes, in turn affecting behavioral and social risk factors, highlighting the need for longitudinal work in this space. We also did not collect any information on prior or current mental health issues or treatments. The inclusion of this could have further elucidated the nature of the mental health needs of sex workers. This study did not find a relationship between depression and condom use, although this has been observed elsewhere. Sex workers may have overstated their use of condoms as a result of social desirability bias. Based on the prevalence of syphilis, HIV, and self-reported STIs in the sample, it is likely that the self-reported condom use measure is biased, thus obscuring any association with depression. There are also limitations with the methods used. While RDS is one of the few available methods to extrapolate population estimates for key populations, there are valid concerns around whether the method ultimately results in an unbiased population proportion estimate. However, the RDS approach offers value in the absence of other probability-based sampling methods.³⁴ Furthermore, the PHQ-9 instrument is a short tool used to screen for depression—so there is a risk of misclassification. However, the PHQ-9 has been validated in many settings, including South Africa and has a reported high sensitivity and specificity when compared with clinical diagnosis by mental health professionals.^29,30 Lastly, depression, as measured in this study and other studies, varies across settings—even within a country. Thus, generalizing our findings to other sex worker groups without an appreciation of the variability in the available literature is not suggested.

Conclusion

This study demonstrates that depression is high among FSW in the Nelson Bay Municipality Area and that depression is associated with HIV risk factors as well as social covariates, notably food security, and social circumstances. There is evidence that these factors may work together to cause and maintain the increased risk of HIV among FSW. To provide effective HIV prevention and treatment services to sex workers, mental health may need to be considered within the package of care. However, mere integration of mental health components into existing interventions may not be effective in isolation—if such interventions do not deal with the real social and economic stressors that may cause and maintain risk behavior and depressive symptoms. The predominant social and biobehavioral risk factor framework could be significantly improved by including mental health factors. Such inclusion could support more holistic focused research and programmatic public health interventions for sex workers.

Footnotes

Authors' Contributions

J.R. conducted the secondary analysis of the study and drafted this article. H.H. (TB HIV Care) was a PI on the original study and reviewed and contributed toward the article. M.M. (TBHIV Care) was involved in the original study and reviewed and contributed toward the article. K.Y. (TB HIV Care) reviewed the article and supported the lead author in drafting it. S.B. (Johns Hopkins University) was a PI on the original study and reviewed and contributed toward the article. S.S. (Johns Hopkins University) was a PI on the original study and reviewed and contributed toward the article. A.R. (Johns Hopkins University) reviewed the article and provided support in drafting the article and carrying out the analysis.

Acknowledgments

This study was funded by the MAC AIDS Foundation. The authors are grateful to the study team for their efforts and are particularly grateful to the women who participated in this study.

Data Sharing

The data are jointly owned by the partner institutions. Requests for collaboration or utilization of the data should be sent to Sheree Schwarz at sschwartz@jhu.edu.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was funded by the MAC AIDS Foundation and conducted by the Human Sciences Research Council (HSRC) and TB HIV Care in collaboration with the Johns Hopkins University. Support for secondary analysis for Stefan Baral and Sheree Schwartz were provided through grant number NR016650 (NIH-NINR).

References

Scorgie

, Chersich

, Ntaganira

, Gerbase

, Lule

, Lo

: Socio-demographic characteristics and behavioral risk factors of female sex workers in sub-Saharan Africa: A systematic review. AIDS Behav, 2012; 16:920–933.

Baral

, Beyrer

, Muessig

, et al.: Burden of HIV among female sex workers in low-income and middle-income countries: A systematic review and meta-analysis. Lancet Infect Dis, 2012; 12:538–549.

UCSF, Anova Health Institute

WHRI

. South African Health Monitoring Study (SAHMS), Final Report: The Integrated Biological and Behavioural Survey Among Female Sex Workers, South Africa 2013–2014. San Francisco UCSF, San Francisco, 2015.

Bekker

, Johnson

, Cowan

, et al.: Combination HIV prevention for female sex workers: What is the evidence?. Lancet, 2015; 385:72–87.

Chow

, Tung

, Tucker

, et al.: Behavioural interventions improve condom use and HIV testing uptake among female sex workers in China: A systematic review and meta-analysis. Sex Transm Infect, 2016; 29:454–459.

Footer

KHA

, Silberzahn

, Tormohlen

, Sherman

: Policing practices as a structural determinant for HIV among sex workers: A systematic review of empirical findings. J Int AIDS Soc, 2016; 19(Suppl 3):20883.

Kerrigan

, Fonner

, Stromdahl

, Kennedy

: Community empowerment among female sex workers is an effective HIV prevention intervention: A systematic review of the peer-reviewed evidence from low- and middle-income countries. AIDS Behav, 2013; 17:1926–1940.

Shahmanesh

, Patel

, Mabey

, Cowan

: Effectiveness of interventions for the prevention of HIV and other sexually transmitted infections in female sex workers in resource poor settings: A systematic review. Trop Med Int Health, 2008; 13:659–679.

Shannon

, Strathdee

, Goldenberg

, et al.: Global epidemiology of HIV among female sex workers: Influence of structural determinants. Lancet, 2015; 385:55–71.

10.

Shushtari

, Hosseini

, Sajjadi

, Salimi

, Latkin

, Snijders

TAB

: Social network and HIV risk behaviors in female sex workers: A systematic review. BMC Public Health, 2018; 18:1020.

11.

Yuen

WWY

, Tran

, Wong

CKH

, Holroyd

, Tang

CSK

, Wong

WCW

. Psychological health and HIV transmission among female sex workers: A systematic review and meta-analysis. AIDS Care, 2016; 28:816–824.

12.

Coetzee

, Buckley

, Otwombe

, Milovanovic

, Gray

, Jewkes

: Depression and post-traumatic stress amongst female sex workers in Soweto, South Africa: A cross sectional, respondent driven sample. PLoS One, 2018; 13:e0196759.

13.

Poliah

, Paruk

: Depression, anxiety symptoms and substance use amongst sex workers attending a non-governmental organisation in KwaZulu-Natal, South Africa. South African Fam Pract, 2017; 59:116–122.

14.

Stein

, Seedat

, Herman

, et al.: Lifetime prevalence of psychiatric disorders in South Africa. Br J Psychiatry, 2008; 192:112–117.

15.

Tomlinson

, Grimsrud

, Stein

, Williams

, Myer

: The epidemiology of major depression in South Africa: Results from the South African Stress and Health study. South African Med J, 2009; 99(Pt 2):367–373.

16.

Gibbs

, Govender

, Jewkes

: An exploratory analysis of factors associated with depression in a vulnerable group of young people living in informal settlements in South Africa. Glob Public Health, 2018; 13:788–803.

17.

Patel

, Saggurti

, Pachauri

, Prabhakar

: Correlates of mental depression among female sex workers in southern India. Asia-Pacific J Public Health, 2015; 27:809–819;17:133.

18.

Puri

, Shannon

, Nguyen

, Goldenberg

: Burden and correlates of mental health diagnoses among sex workers in an urban setting. BMC Womens Health, 2017; 17:133.

19.

Miltz

, Rodger

, Cozzi Leri

, et al.: Investigating conceptual models for the relationship between depression and condomless sex among gay, bisexual and other men who have sex with men: Using structural equation modelling to assess mediation. AIDS Behav, 2020; 24:1793–1806.

20.

Ahaneku

, Ross

, Nyoni

, et al.: Depression and HIV risk among men who have sex with men in Tanzania. AIDS Care, 2016; 28(Suppl 1):140–147.

21.

Tucker

, Liht

, De Swardt

, et al.: Homophobic stigma, depression, self-efficacy and unprotected anal intercourse for peri-urban township men who have sex with men in Cape Town, South Africa: A cross-sectional association model. AIDS Care, 2014; 26:882–889.

22.

Lehrer

, Shrier

, Gortmaker

, Buka

: Depressive symptoms as a longitudinal predictor of sexual risk behaviors among US middle and high school students. Pediatrics, 2006; 118:189–200.

23.

Nduna

, Jewkes

, Dunkle

, Shai

NPJ

, Colman

: Associations between depressive symptoms, sexual behaviour and relationship characteristics: A prospective cohort study of young women and men in the Eastern Cape, South Africa. J Int AIDS Soc, 2010; 13:44.

24.

HSRC. The Fifth South African National HIV Prevalence, Incidence, Behaviour and Communication Survey, 2017, HIV Impact Assessment Summary. HSRC Press, Cape Town, 2018.

25.

Schwartz

, Lambert

, Phaswana-Mafuya

, Kose

, Mcingana

, Holland

: Engagement in the HIV care cascade and barriers to antiretroviral therapy uptake among female sex workers in Port Elizabeth, South Africa: Findings from a respondent-driven sampling study. Sex Transm Infect, 2017; 93:290–296.

26.

Rao

, Baral

, Phaswana-Mafuya

, Lamber

, Mcingana

, Holland

: Pregnancy intentions and safer pregnancy knowledge among female sex workers in Port Elizabeth, South Africa. Obstet Gynecol, 2016; 128:15–21.

27.

Heckathorn

: Respondent-driven sampling: A new approach to the study of hidden populations. Soc Probl, 2012; 44:174–199.

28.

Goel

, Salganik

: Assessing respondent-driven sampling. Proc Natl Acad Sci U S A, 2010; 107:6743–6747.

29.

Kroenke

, Spitzer

, Williams

JBW

: The PHQ-9: Validity of a brief depression severity measure. J Gen Intern Med, 2001; 16:606–613.

30.

Cholera

, Gaynes

, Pence

, et al.: Validity of the patient health questionnaire-9 to screen for depression in a high-HIV burden primary healthcare clinic in Johannesburg, South Africa. J Affective Disorders, 2014; 167:160–166.

31.

Armstrong

, Jorm

, Samson

, et al.: Association of depression, anxiety, and suicidal ideation with high-risk behaviors among men who inject drugs in Delhi, India. J Acquir Immune Defic Syndr, 2013; 64:502–510.

32.

Maclean

, Lancaster

, Lungu

, et al.: Prevalence and correlates of probable depression and post-traumatic stress disorder among female sex workers in Lilongwe, Malawi. Int J Ment Health Addict, 2018; 16:150–163.

33.

Grosso

, Ketende

, Baral

: Stigma and Discrimination Indicators Related to Distal Determinants. Johns Hopkins University, Baltimore, MD, 2014.

34.

Schonlau

, Liebau

: Respondent-driven sampling. Stata J, 2012; 12:72–93.

35.

Barros

AJD

, Hirakata

: Alternatives for logistic regression in cross-sectional studies: An empirical comparison of models that directly estimate the prevalence ratio. BMC Med Res Methodol, 2003; 3:21.

36.

Magidson

, Saal

, Nel

, Remmert

, Kagee

: Relationship between depressive symptoms, alcohol use, and antiretroviral therapy adherence among HIV-infected, clinic-attending patients in South Africa. J Health Psychol, 2017; 22:1426–1433.

37.

Surratt

, Kurtz

, Chen

, Mooss

: HIV risk among female sex workers in Miami: The impact of violent victimization and untreated mental illness. AIDS Care, 2012; 24:553–561.

38.

Štulhofer

, Sinković

, Božić

, Baćak

: Victimization and HIV risks among Croatian female sex workers: Exploring the mediation role of depressiveness and the moderation role of social support. Violence Against Women, 2017; 23:67–88.