Single Mothers and Female Sex Workers in Zambia Have Similar Risk Profiles

Abstract

The aim of the study was to compare reproductive health and high-risk behaviors in female sex workers (FSWs) and single mothers (SMs) in Zambia's two largest cities, Lusaka and Ndola. FSWs were invited from known community hot spots, and sexually active HIV− SMs were referred from infant vaccination services for free and anonymous screening and treatment for HIV and other sexually transmitted infections (STIs) and long acting reversible contraception. A subset completed an interviewer-administered survey. From 2012 to 2016, 1,893 women (1,377 FSWs and 516 HIV− SMs) responded to referrals. HIV prevalence was 50% in Lusaka and 33% in Ndola FSWs. Positive syphilis serology (rapid plasmin reagin) was found in 29%–31% of HIV+ FSWs and 9%–12% of HIV− FSWs and SMs. Trichomonas was more common in Ndola (11%–12%), compared with Lusaka (3%–7%). Antiretroviral therapy (ART) use among HIV+ FSWs was 9%–15%. In all groups, consistent condom use (8%–11%) and modern contraceptive use (35%–65%) were low. Low literacy and reported coercion at first sexual intercourse were common in both FSWs and SMs, as was alcohol use during sex among FSWs. Zambian FSWs and SMs have low condom use and high HIV/STI and unplanned pregnancy risk. Many FSWs and half of SMs are ≥25 years of age, and thus too old for HIV prevention services targeting “adolescent girls and young women” (aged 15–24). Tailored and targeted reproductive health services are needed to reduce HIV, STI, and unplanned pregnancy in these vulnerable women.

Introduction

Zambia in southern Africa has a population of 16.6 million and a high burden of HIV: in 2016, 1.2 million Zambians were living with HIV, and there were 59,000 new HIV infections and 21,000 AIDS-related deaths.¹

As elsewhere in Africa, sex work in Zambia is a high-risk activity for exposure to HIV, sexually transmitted infections (STIs), and unplanned pregnancy.^2–5 In 2015, an integrated Bio-Behavioral Survey was conducted with 1,050 female sex workers (FSWs) in five Zambian districts, including five border cities along truck routes. As expected, HIV prevalence was high (56.4%), and syphilis (21%) and trichomonas (9%) were also common. Condom use was inconsistent, particularly with nonpaying partners,⁶ and poverty was the norm.⁷

The 2017 PEPFAR Country operating plan estimated >24,000 FSWs in Zambia.⁸ A 2001 survey estimated that 2,233 women were conducting sex work in Ndola, Zambia's second largest city, with an HIV prevalence of 69%, syphilis prevalence of 42%, and only 28% reporting use of a condom during their last intercourse with a client.⁹ The literature also includes qualitative investigations,^10,11 a description of family-based services for FSWs and their children in Lusaka,¹² a survey of contraceptive use in FSWs in border towns,² and studies of HIV self-testing¹³ and electronic fingerprinting¹⁴ in Zambian FSW. A 2009 model estimated that 1% of Zambian women are FSWs, 90% would become HIV infected after 21 years of sex work, and 36.4% of infections in Zambian men were acquired from an FSW.¹⁵ A 2008 survey reported that 13% of Zambian men had sex with an FSW in the last year.¹⁶

Risk factors for sex work are generally related to poverty resulting from lack of education and employment opportunities. Single mothers (SMs) are likely to be at high risk of initiating sex work, particularly those with an unplanned pregnancy or those who have been widowed or abandoned.^17–19 SMs are an under-researched population in Africa, and little is known about their HIV, STI, and risk behaviors.

We present the results of a free reproductive health program for FSWs and HIV− SMs in Lusaka and Ndola, Zambia's two largest cities. Prevalence and correlates of HIV, STI, as well as contraceptive and condom use are compared in HIV+ and HIV− FSWs and HIV− SMs, and between Lusaka and Ndola to identify gaps in reproductive health services and inform development of tailored and targeted HIV prevention and treatment efforts.

Materials and Methods

Recruitment

From 2012 to 2016, the Zambia Emory HIV Research Project (ZEHRP) offered free reproductive health services to high-risk single women aged 18–45 in Lusaka and Ndola. The goal of the study sponsors was to identify uninfected women at high risk of HIV for possible inclusion in prevention trials. Two recruitment strategies were used. In the first, SMs who had previously tested HIV negative during antenatal care and were not currently pregnant were referred from infant vaccination services in government clinics. With the second strategy, FSWs were invited from known hot spots, including bars, street corners, and lodges by ZEHRP community workers. To protect confidentiality, recruiters distributed invitations to all interested FSWs without asking their HIV status. Exclusion criteria included age <18 or >45. The printed invitation included the address and operating hours of the ZEHRP clinic, and a description of services (free HIV testing, STI screening, long acting reversible contraception (LARC: copper intrauterine devices [IUDs] and implants), and transportation to and from the clinic).

Integrated Reproductive Health Program

Women were screened for HIV using Alere Determine™ HIV-1/2, with positive and indeterminate results confirmed using Trinity Biotech Uni-Gold™ Recombigen^® HIV-1/2.²⁰ BD Macro-Vue™ rapid plasmin reagin (RPR) (Immunotrep) was used for syphilis serologic screening.²¹ A wet mount was prepared from a self-administered vaginal swab for microscopic diagnosis of trichomonas infection and detection of sperm as a biological marker of recent unprotected sex. Women received family planning counseling with the offer of LARC, which were not readily available in government clinics. Free condoms were also provided. As there could be overlap between FSWs and SMs despite the different recruitment methods, SMs referred from infant vaccination clinics were asked if they exchanged sex for money and classified as FSWs in these analyses if they responded affirmatively.

Demographic and behavioral survey

Women receiving services were assigned a unique numeric participant number, and identifiers were not recorded. A subset of participants provided written informed consent for an interviewer-administered questionnaire assessing demographic and behavioral risk factors for HIV and STI. Questions included age, marital status, sexual and reproductive history, condom and contraceptive use, fertility goals, and alcohol use. We assessed literacy in local languages (Nyanja and Bemba) and English. Nyanja and Bemba are Zambia's two most commonly spoken languages. Lusaka is the centrally located capital and home to migrants from all parts of the country. Nyanja is the accepted language of communication though it is not the first language for most. Ndola is in the northern Copperbelt region where Bemba is universally spoken as a first language. In general, English is not taught until secondary school. Initial questionnaires included demographic information, and were administered to HIV+ and HIV− FSWs and HIV− SMs: these findings are presented in Table 2. Subsequently, more detailed questionnaires including sexual risk behaviors and contraceptive use were administered only to HIV− FSWs and SMs: these findings are presented in Table 3 and Figure 1.

Table 2.

Literacy and Duration of Residence in the City in HIV− Female Sex Workers and Single Mothers in Lusaka and Ndola, Zambia (n = 1,165)

	Lusaka									Ndola
	FSW				p-valueaHIV+ vs. HIV−	SM		p-valueaHIV− FSW vs. SM	p-valueaacross groups	FSW				p-valueaHIV+ vs. HIV−	SM		p-valueaHIV− FSW vs. SM	p-valueaacross groups
	HIV+ (n = 55)		HIV− (n = 202)			HIV− (n = 170)				HIV+ (n = 39)		HIV− (n = 353)			HIV− (n = 346)
	n or mean	% or SD	n or mean	% or SD		n or mean	% or SD			n or mean	% or SD	n or mean	% or SD		n or mean	% or SD
Years living in Lusaka/Ndola	22.7	12.2	18.2	9.7	.0147	16.9	8.7	.1761	.0007	23.8	11.6	16.8	9.0	.0007	17.6	8.5	.2401	<.0001
Literacy
Understands Nyanja
Easily	51	94%	187	93%	.7719	165	97%	.0561	.1622	18	46%	99	28%	.0197	52	15%	<.0001	<.0001
With difficulty/not at all	3	6%	15	7%		5	3%			21	54%	253	72%		294	85%
Reads Nyanja
Easily	7	13%	57	28%	.0185	55	32%	.3864	.0184	7	18%	20	6%	.0113	0	0%	<.0001	<.0001
With difficulty/not at all	48	87%	145	72%		115	68%			32	82%	332	94%		346	100%
Understands Bemba
Easily	39	71%	140	69%	.8188	130	76%	.1229	.2958	39	100%	343	97%	.6078	336	97%	.7867	.5580
With difficulty/not at all	16	29%	62	31%		40	24%			0	0%	9	3%		10	3%
Reads Bemba
Easily	9	16%	41	20%	.5136	23	14%	.0849	.2224	17	44%	126	36%	.3376	202	58%	<.0001	<.0001
With difficulty/not at all	46	84%	161	80%		147	86%			22	56%	226	64%		144	42%
Understands English
Easily	19	35%	108	53%	.0128	109	64%	.0379	.0005	12	31%	99	28%	.7364	43	12%	<.0001	<.0001
With difficulty/not at all	36	65%	94	47%		61	36%			27	69%	252	72%		303	88%
Reads English
Easily	16	29%	103	51%	.0039	95	56%	.3462	.0024	9	23%	69	20%	.6126	18	5%	<.0001	<.0001
With difficulty/not at all	39	71%	99	49%		75	44%			30	77%	282	80%		328	95%

^aTwo-tailed t-test and one-way ANOVA for continuous variables, chi-square test for categorical variables with cell counts ≥5, Fisher's exact test for categorical variables with 20% of expected cell counts <5.

Table 3.

Sexual and Reproductive History and Alcohol Use in HIV− Female Sex Workers and Single Mothers in Lusaka and Ndola, Zambia (n = 924)

	Lusaka					Ndola
	HIV− FSW (n = 150)		HIV− SM (n = 170)		pa	HIV− FSW (n = 276)		HIV− SM (n = 346)		pa
	n or mean	% or SD	n or mean	% or SD		n or mean	% or SD	n or mean	% or SD
Lifetime sexual history
Age at first sexual encounter	16.9	2.9	17.9	2.3	.0007	15.2	2.4	16.3	2.0	<.0001
Used a condom during first sexual intercourse
Yes	55	37%	63	37%	.9785	53	19%	73	21%	.5619
No	94	63%	107	63%		222	81%	272	79%
Pressure or force at first sexual intercourse
Willingly	103	69%	114	67%	.7506	195	71%	266	77%	.0002
Pressured verbally	25	17%	34	20%		32	12%	54	16%
Forced physically	21	14%	22	13%		49	18%	25	7%
Received money/goods/favors at first sexual intercourse
Yes	86	58%	40	24%	<.0001	221	81%	106	31%	<.0001
No	63	42%	129	76%		53	19%	237	69%
Reproductive history and fertility goals
No. of live births	1.6	1.5	1.7	1.3	.4621	1.7	1.4	1.8	1.3	.2237
No. of living children	1.5	1.4	1.7	1.2	.3137	1.6	1.4	1.8	1.2	.1867
Plans for more children
Yes, next 3 years	21	16%	3	2%	<.0001	113	50%	62	18%	<.0001
Yes, >3 years	17	13%	21	13%		70	31%	172	50%
No	95	71%	144	86%		42	19%	110	32%
Alcohol and drug use history
Ever uses alcohol
Yes	109	74%	22	13%	<.0001	229	83%	78	23%	<.0001
No	39	26%	144	87%		46	17%	262	77%
Alcohol use frequency
Daily	39	37%	2	10%	<.0001	130	57%	26	33%	<.0001
Weekly	54	51%	5	24%		84	37%	29	37%
1–3 times a month	11	10%	6	29%		11	5%	14	18%
<1–3 times a month	2	2%	8	38%		4	2%	9	12%
Sex during alcohol use frequency
Always	11	10%	1	5%	<.0001	79	34%	15	19%	<.0001
Most of the time	18	17%	1	5%		55	24%	9	12%
Sometimes	43	41%	4	19%		59	26%	21	27%
Rarely	27	25%	3	14%		33	14%	9	12%
Never	7	7%	12	57%		3	1%	24	31%
Highest formal school education completed
None of the above	14	9%	3	2%	.0092	21	8%	41	12%	.0118
Primary	48	32%	53	31%		156	57%	218	63%
Secondary	84	56%	112	66%		98	36%	86	25%
College	4	3%	2	1%		1	0%	1	0%

^aTwo-tailed t-test for continuous variables, chi-square test for categorical variables with cell counts ≥5, Fisher's exact test for categorical variables with 20% of expected cell counts <5.

FIG. 1.

Reported condom use in the last month among FSWs in Lusaka and Ndola by partner type. FSW, female sex workers.

Treatment

HIV+ women were counseled and referred to their nearest government clinic for antiretroviral therapy (ART) evaluation. RPR-positive women were treated at ZEHRP with injectable penicillin or with erythromycin in the case of allergy. Women with trichomonas on vaginal wet mount were provided with a 2-g oral dose of metronidazole (directly observed therapy). Women reporting vaginal discharge were treated for candida or bacterial vaginosis based on examination of vaginal wet mounts.

Data analysis

Data were entered in Microsoft Access and analyzed using Statistical Analysis Software (SAS) version 9.4. Frequencies of laboratory and survey results were compared in three mutually exclusive groups (HIV+ and HIV− FSW and HIV− SM) and between the two cities (Lusaka and Ndola) using chi-square tests of significance for categorical variables and t-tests and ANOVA for differences in means. Categorical variables that were used in interaction terms with city were identified in crude analyses based on (1) differences in significance (p < .05) of association with the outcome in Lusaka versus Ndola or (2) differences in the direction of the magnitude of association in Lusaka versus Ndola. Bivariate and multivariate logistic regression models were run after assessing collinearity. Multivariate logistic regression models assessed predictors of HIV+ status among FSWs and predictors of being an FSW versus SM among HIV− women. Variables and interaction terms by city significantly associated with outcomes in bivariate models at an alpha of 0.05 were included simultaneously in multivariate logistic regression models. Final multivariate logistic regressions were arrived at by dropping nonsignificant variables and interaction terms through backward elimination. Individual variables comprising the interaction terms were also included in the models to ensure that models were hierarchically well formulated. Contrast statements were used to estimate associations between predictor and outcome by city.

Ethics

Approval was obtained from Office of Human Research Protection-registered Institutional Review Boards in Zambia (i.e., University of Zambia Biomedical Research Committee) and at Emory University. All procedures, including counseling, informed consent, and surveys, were administered in English, Nyanja, or Bemba, the predominant local languages in Lusaka and Ndola.

Results

In Lusaka, 731 FSWs received services, of whom 365 (50%) were HIV+. FSWs in Ndola had a lower HIV prevalence, with 211 (33%) HIV+ FSWs among 646 who received services (p < .0001). All SMs referred from government clinics were HIV−, with 170 receiving services in Lusaka and 346 in Ndola. A subset of 1,165 women completed the demographic and behavioral survey, including 55 HIV+ FSWs, 202 HIV− FSWs, and 170 SMs in Lusaka; and 39 HIV+ FSWs, 353 HIV− FSWs, and 346 SMs in Ndola.

HIV and STI prevalence

Across cities, HIV+ FSWs were older than HIV− FSWs, and HIV− FSWs were older than SMs. HIV+ FSWs were more likely to report being divorced, widowed, or separated than HIV− FSW. Marital status profiles for SMs and HIV− FSWs in Lusaka were not different, while in Ndola SMs were more likely to be divorced or separated compared with HIV− FSWs. HIV− FSWs were slightly more likely to be pregnant than HIV+ FSWs in both cities but the differences were not significant, while none of the SMs were pregnant per referral criteria from postnatal infant vaccination clinics (Table 1).

Table 1.

Demographic, HIV and Sexually Transmitted Infections and Contraceptive Use Among Female Sex Workers and Single Mothers Seeking Reproductive Health Services at the Project Clinic in Lusaka and Ndola, Zambia (n = 1,893)

	Lusaka									Ndola
	FSW				p-valueaHIV+ vs. HIV−	SM		p-valueaHIV− FSW vs. SM	p-valueaacross groups	FSW				p-valueaHIV+ vs. HIV−	SM		p-valueaHIV− FSW vs. SM	p-valueaacross groups
	HIV+ (n = 365)		HIV− (n = 366)			HIV− (n = 170)				HIV+ (n = 211)		HIV− (n = 435)			HIV− (n = 346)
	n or mean	% or SD	n or mean	% or SD		n or mean	% or SD			n or mean	% or SD	n or mean	% or SD		n or mean	% or SD
Age	31.5	6.4	26.0	6.6	<.0001	23.9	5.2	<.0001	<.0001	30.5	6.5	24.3	6.4	<.0001	23.1	4.8	.0015	<.0001
Marital status
Single	141	57%	199	69%	<.0001	118	69%	.2937	<.0001	95	66%	353	84%	<.0001	233	67%	<.0001	<.0001
Divorced/separated	82	33%	87	30%		47	28%			30	21%	58	14%		102	29%
Widowed	24	10%	3	1%		5	3%			20	14%	8	2%		11	3%
RPR for syphilis
Positive	106	29%	44	12%	<.0001	17	10%	.4355	<.0001	66	31%	37	9%	<.0001	35	10%	.5376	<.0001
Negative	257	71%	313	88%		153	90%			145	69%	383	91%		311	90%
Trichomonas on wet prep
Positive	26	7%	11	3%	.0103	6	4%	.7641	.0207	26	12%	49	12%	.8103	39	11%	.8645	.9322
Negative	332	93%	351	97%		164	96%			185	88%	371	88%		307	89%
Sperm on wet prep
Positive	6	2%	8	2%	.6114	1	1%	.2836	.3969	2	1%	10	2%	.2137	14	4%	.1880	.0788
Negative	348	98%	352	98%		169	99%			209	99%	410	98%		332	96%
Self-reported pregnancy
Yes	7	2%	13	4%	.1499					6	4%	17	6%	.3455
No	300	98%	284	96%						154	96%	277	94%
Current contraceptive method (excluding those self-reporting pregnancy)
Any modern nonbarrier method	124	35%	145	41%	.0760	39	19%	<.0001	.0003	99	53%	260	65%	.0066	220	65%	.7962	.0166
None/other/condoms only	232	65%	206	59%		129	81%			88	47%	142	35%		126	35%
Previous HIV test
Yes	215	60%	280	78%	<.0001					131	64%	326	79%	.0001
No	142	40%	78	22%						73	36%	89	21%
Among HIV+ reporting previous testing, ART use
Yes	27	15%								10	9%
No	156	85%								98	91%

^aTwo-tailed t-test and one-way ANOVA for continuous variables, chi-square test for categorical variables with cell counts ≥5, Fisher's exact test for categorical variables with 20% of expected cell counts <5.

ART, antiretroviral therapy; FSW, female sex workers; RPR, rapid plasmin reagin; SD, standard deviation; SM, single mother.

The prevalence of positive RPR was substantially higher in HIV+ FSWs compared with HIV− FSWs and SMs in both cities, and trichomonas was more common in HIV+ FSWs compared with HIV− women in Lusaka. Interestingly, the prevalence of trichomonas on vaginal wet mount was higher in Ndola compared with Lusaka (p < .0001) and was not different in all three Ndola groups. Differences in detection of sperm noted on wet mount, an indicator of unprotected sex in the previous 3 days, were not significant.

Contraceptive use

Women in Ndola were more likely than women in Lusaka to report using a modern contraceptive (p < .0001), with HIV− FSWs and SMs reporting the highest usage. SMs in Lusaka were least likely to report using a modern nonbarrier contraceptive method. When the implant and the IUD were offered at the project clinic, 30 women in Lusaka and 22 in Ndola chose the implant with none choosing the IUD (not shown). An additional five women in Lusaka and one in Ndola requested a LARC method but had a positive pregnancy test and did not qualify (not shown).

Literacy and education

Nyanja was understood by ≥93% of Lusaka respondents though literacy was low. At least two-thirds of Lusaka dwellers also understood Bemba easily though only one in five could read Bemba easily (Table 2).

Nearly all women in Ndola understood Bemba, and Bemba literacy rates were high compared with both Nyanja and Bemba literacy in Lusaka (p < .0001 for both languages in Lusaka). In contrast with Lusaka women who were more likely to be comfortable in both languages, the proportion of Ndola women who understood Nyanja was only 46% among HIV+ FSWs, 28% among HIV− FSWs, and only 15% among SMs, and Nyanja literacy was uniformly low.

Understanding English was far more common in Lusaka compared with Ndola (p < .0001), especially among SMs. English literacy trends paralleled comprehension though at lower levels, with over half of Lusaka respondents able to read English with ease compared with only one in 5 FSWs and one in 20 SMs in Ndola.

Number and type of sexual partners and condom use

SMs reported a mean of 0.9 sex partners in the last year in Lusaka and 1.8 in Ndola (p < .0001) (median 1 in both groups) (not shown). Reported condom use was low with most (70% in Lusaka, 62% in Ndola) reporting never using condoms and only 8%–10% reporting always using condoms (not shown) (p = NS between Lusaka and Ndola).

HIV-negative FSWs were asked about steady, nonpaying partners; repeat clients; and nonrepeat clients in the last month. Half of Lusaka FSWs and one-third of Ndola FSWs reported a steady, nonpaying partner (p = .0008). The number of repeat clients in the last month (mean 4.5/month in Ndola vs. 3.8/month in Lusaka, p = .0130; median 4 in both cities) and the number of nonrepeat clients in the last month (15.8/month in Ndola vs. 7.8/month in Lusaka, p < .0001; median 6 in Ndola and 5 in Lusaka) were higher in Ndola.

Figure 1 shows patterns of condom use by city and by partner type. Only 27% of Lusaka FSWs and 9% of Ndola FSWs reported always using condoms with steady, nonpaying partners (p = .0042). In contrast, in Lusaka condom use was reported much more frequently with paying clients (55% always for repeat and 72% always for nonrepeat clients, p = .0035). Reported condom use among paying clients of Ndola FSWs was not different between repeat and nonrepeat clients (12% always and 49% sometimes vs. 16% always and 42% sometimes, respectively, p = NS), but was substantially lower than reported condom use in their Lusaka counterparts (p < .0001).

Sexual and reproductive history, fertility goals, and alcohol use

These questions were asked only of HIV− FSWs and SMs. In both cities, age at first sexual intercourse was younger for FSWs compared with SMs. When comparing across cities, Ndola had significantly younger ages at first sexual intercourse compared with Lusaka for both FSWs and SMs (p < .0001 for both) (Table 3).

Interestingly, in both cities the proportion of women reporting use of a condom during their first sexual encounter was not different in FSWs and SMs. Lusaka FSWs and SMs were more likely to report using a condom at first intercourse (FSW, SM: 37%) than their counterparts in Ndola (FSW: 19%, p < .0001; SM: 21%, p = .0001).

When asked whether they had their first sexual intercourse willingly, two-thirds to three-quarters of all groups said yes. In Lusaka, the proportion reporting being physically forced to have their first sexual encounter was comparable in FSWs (14%) and SMs (13%), while in Ndola, FSWs were more likely to report forced sex (18% vs. 7% in SM). The frequency of being verbally pressured into their first sexual encounter was not different for SMs and FSWs (20% vs. 17% in Lusaka, 16% vs. 12% in Ndola). FSWs were far more likely to report receiving goods or money for their first sexual experience than SMs in both cities with this report being more common in Ndola FSWs (58% of FSW vs. 24% of SM in Lusaka, 81% vs. 31% in Ndola).

The mean number of live births and living children was not different in Lusaka and Ndola FSWs and SMs. The majority of Lusaka respondents did not plan to have more children (71% of FSWs and 86% of SMs), while most women in Ndola planned to have more children within the next 3 years (50% FSW and 18% SM) or in >3 years (31% FSW and 50% SM). Although modern contraceptive use was more common in Ndola compared with Lusaka, among HIV− SMs and FSWs there was no correlation between stated fertility intent and modern contraceptive use. Similarly, among Lusaka SMs, contraceptive use was not different in women who wanted to have more children and in those who did not. Among Lusaka FSWs however, modern contraceptive use was higher among those not wanting more children (53% vs. 28% of those wanting more children, p < .01).

Alcohol use was reported far more frequently by FSWs (74% in Lusaka and 83% in Ndola) compared with SMs (13% in Lusaka and 23% in Ndola; p < .0001 for both cities). Among FSWs reporting ever using alcohol, 88%–93% reported weekly or daily use. Among SMs who reported ever using alcohol, those in Ndola were over twice as likely to report using alcohol daily or weekly than those in Lusaka (p = .0018). Ndola FSWs were also more likely to report sex during alcohol use always or most of the time than their Lusaka counterparts (34% and 24% versus 10% and 17% in Lusaka, respectively; p < .0001). Ndola SMs were also more likely to report sex during alcohol use always (19%) or most of the time (12%) compared with only 5% of Lusaka SMs though the numbers were small in the latter group (p < .05).

Multivariate predictors of HIV+ status among FSWs

Residing in Lusaka, increasing age, being widowed versus single, having a positive RPR, and not reporting previous testing for HIV remained significant predictors of positive HIV status among FSWs in the adjusted multivariate model. Interaction terms for city were initially included for trichomonas and modern contraceptive use but were not significant in the multivariate model and were removed through backward elimination (Table 4).

Table 4.

Logistic Regression Models of Factors Associated with Testing Positive for HIV at Cohort Screening Among Female Sex Workers in Lusaka and Ndola, Zambia (n = 1,377)

	Bivariate models				Final multivariate model
	95% CI				95% CI
	cOR	LLa	ULb	p	aOR	LLa	ULb	p
City
Ndola	Ref	—	—	—	Ref	—	—	—
Lusaka	2.06	1.65	2.56	<.0001	2.01	1.55	2.83	<.0001
Age (per 1 year increase)	1.14	1.12	1.16	<.0001	1.14	1.11	1.17	<.0001
Marital status
Single	Ref	—	—	—	Ref	—	—	—
Divorced/separated	1.81	1.35	2.41	<.0001	0.72	0.50	1.05	.0859
Widowed	9.35	4.75	18.42	<.0001	3.54	2.35	4.83	.0009
RPR for syphilis
Negative	Ref	—	—	—	Ref	—	—	—
Positive	3.68	2.75	4.92	<.0001	3.37	2.35	4.83	<.0001
Previously tested for HIV
Yes	Ref	—	—	—	Ref	—	—	—
No	2.26	1.77	2.87	<.0001	1.87	1.35	2.60	.0002

The following variables were not significant in bivariate analyses and are not tabled: Trichomonas on wet mount result, sperm on wet mount result, and pregnancy (self-reported). The following variables were not significant in initial adjusted analyses and were removed from the final multivariate model through backward elimination and are not tabled: interaction terms between city and trichomonas on wet mount and city and current contraceptive method, and current contraceptive method.

^aLower limit for 95% CI.

^bUpper limit for 95% CI.

AOR, adjusted odds ratio; CI, confidence interval; COR, crude odds ratio.

Multivariate predictors of being an FSW versus an SM among HIV− women

Increasing age, younger age at first sexual encounter, being single versus widowed, divorced or separated, reading the vernacular Bemba and Nyanja with difficulty or not at all versus reading either easily in Ndola, being forced physically versus engaging willingly at first sexual intercourse in Ndola, receiving money or goods at first sexual intercourse, and ever using alcohol remained significant predictors of being an FSW versus an SM. Interestingly, primary and secondary education versus no education were associated with a lower likelihood of being an FSW in Lusaka, while secondary education or higher was associated with a higher likelihood of being an FSW in Ndola (Table 5).

Table 5.

Logistic Regression Models of Factors Associated with Being a Female Sex Worker Versus a Single Mother Among HIV− Women at Baseline in Lusaka and Ndola, Zambia (n = 942)

	Bivariate models				Final multivariate model
	95% CI				95% CI
	cOR	LLa	ULb	p	aOR	LLa	ULb	p
Age (per 1 year increase)	1.05	1.03	1.08	<.0001	1.13	1.09	1.18	<.0001
Age at first sexual encounter (per 1 year increase)	0.84	0.80	0.89	<.0001	0.84	0.77	0.91	<.0001
Marital status
Single	1.89	1.40	2.57	<.0001	3.90	2.32	6.58	<.0001
Divorced/separated/widowed	Ref	—	—	—	Ref	—	—	—
Reads vernacular language (Bemba or Nyanja) x city
Easily (for either language) versus with difficulty/not at all (for both languages) in Ndola	2.36	1.70	3.26	<.0001	2.07	1.25	3.42	.0045
Easily versus with difficulty/not at all in Lusaka	0.85	0.55	1.33	.4850	0.72	0.37	1.38	.3205
Highest education level completed x city
Primary versus none in Ndola	1.40	0.79	2.46	.2458	1.77	0.76	4.16	.1876
Primary versus none in Lusaka	0.19	0.05	0.72	.0139	0.13	0.03	0.67	.0147
Secondary/college versus none in Ndola	2.22	1.22	4.05	.0091	4.11	1.59	10.62	.0035
Secondary/college versus none in Lusaka	0.17	0.05	0.59	.0058	0.10	0.02	0.52	.0059
Pressure or force at first sexual intercourse x city
Pressured verbally versus willingly in Ndola	0.81	0.50	1.30	.3797	1.06	0.53	2.14	.8686
Pressured verbally versus willingly in Lusaka	0.81	0.46	1.46	.4871	0.49	0.22	1.11	.0866
Forced physically versus willingly in Ndola	2.67	1.60	4.48	.0002	4.41	2.05	9.48	.0001
Forced physically versus willingly in Lusaka	1.06	0.55	2.03	.8693	0.97	0.39	2.42	.9454
Received money/goods/favors at first sexual intercourse
No	Ref	—	—	—	Ref	—	—	—
Yes	6.54	4.91	8.70	<.0001	6.21	4.16	9.28	<.0001
Ever uses alcohol
No	Ref	—	—	—	Ref	—	—	—
Yes	16.14	11.69	22.30	<.0001	13.47	9.05	20.05	<.0001

The following variable was not significant in bivariate analyses and is not tabled: years living in Lusaka/Ndola (continuous). The following variables and interaction terms were not significant in initial adjusted analyses and were removed from the final multivariate model through backward elimination and are not tabled: interaction terms between city and marital status and city and current contraceptive method, and current contraceptive method.

^aLower limit for 95% CI.

^bUpper limit for 95% CI.

Discussion

In Zambia's two largest cities, the prevalence of HIV and STI is high among FSWs, and STIs are common among SMs at high risk of initiating sex work. Many FSWs had not previously tested for HIV and among those who were HIV+, less than one in seven were taking ART. In light of these findings, the low rate of condom use—particularly among SMs—the high rate of verbal or physical coercion at first sexual intercourse for both SMs and FSWs, and the frequent use of alcohol during sex by FSWs are especially alarming. Unplanned pregnancy is an additional risk in these women of childbearing age, and use of effective modern contraception is low even among women who state that they do not want to become pregnant. Substantial differences in HIV and STI prevalence and risk behaviors between the two cities indicate that local characteristics are important to consider in prevention and care programs. While comprehension of the local language in culturally homogeneous areas such as Ndola is universal, cities like Lusaka host migrants from several language groups. Understanding of Nyanja may be adequate for day-to-day communication, but nuanced or complex HIV prevention messages may not be thoroughly understood. This, in combination with limited English comprehension and vernacular literacy, highlights the limitations of many HIV prevention efforts for these key populations to date.

Among FSWs in both Lusaka and Ndola, those with HIV were older, more likely to be widowed, to have a positive RPR, and to report not having been previously tested for HIV than their HIV− counterparts. In both groups, uptake of a LARC method when offered at the research clinic was low. While the prevalence of HIV was substantially higher in Lusaka compared with Ndola, the prevalence of positive RPR was not different. Interestingly, the prevalence of trichomonas was twice as high in Ndola as Lusaka and was not different in both HIV serostatus groups, while in Lusaka the prevalence was higher in HIV+ compared with HIV− FSW.

In sub-Saharan Africa, the proportion of women involved in sex work has been estimated at between 0.7% and 4.3% in the capitals and between 0.4% and 4.3% in other urban areas.²² Local surveys show generally higher proportions, and a hidden number of “part-time” sex workers²³ add to transmission of HIV/STIs. Targeted services have been effective in several settings, including southern,²⁴ western,²⁵ and east-central Africa^26,27 as well as on trans-Africa highways.²⁸ Previous surveys and publications of FSWs in Zambia confirm that high prevalence of HIV and STI and poor condom use has been a feature in FSWs for many years.⁹ As in Kenya, the prevalence of HIV in Ndola has decreased in the last 7 years,^6,29 but remains high relative to other groups. Serial Bio-Behavioral Surveys in Zambian border towns confirm that the proportion of FSWs who have been tested for HIV has increased considerably since 2000.³⁰ Unfortunately, programs have typically been geographically limited and intermittently funded.

Programs targeting adolescent girls and young women (AGYW aged up to 24 years) often include educational and economic components designed to prevent entry into the sex industry.^31,32 Many HIV− SMs are in this group and could benefit from targeted programs. However, the mean age of SMs was 24 in Lusaka and 23 in Ndola, and half would thus be too old for AGYW prevention programs. The profile of SMs in our study resembled that of HIV− FSWs in many ways: the prevalence of positive RPR and trichomonas was not different in the two groups in both cities. Literacy in the vernacular was poor as was English comprehension, both indications that employment opportunities and comprehension of written HIV prevention materials would be limited. A quarter to a third of both groups had been pressured physically or verbally at first sex, and among SMs a comparable proportion reported receiving goods or money for that first sexual exposure in both cities. This problem of abuse at a young age has been reported elsewhere in Africa.³³ While SMs were far less likely to report alcohol use than their FSW counterparts, they were also far more likely to report never using condoms.

Less than two-thirds of HIV+ FSWs had been previously tested for HIV, and few of those—15% in Lusaka and only 9% in Ndola—reported ART use. This is in stark contrast to reports of high ART use in the general population of HIV+ Zambians.¹ Availability of ART was generalized in Lusaka and Ndola government clinics in 2012 and thereafter, though eligibility criteria evolved from CD4 < 200 to <350 to <500 cells/mm³. Key populations including FSWs were not deemed eligible regardless of CD4 count until the Zambia Consolidated Guidelines issued in 2016. There have been many legal, policy, and sociocultural barriers to HIV prevention and care among key populations, including FSWs.³⁴ Stigma—both sex work and HIV related—is a major obstacle to prevention, testing, and care for both FSWs and their children,^2,12,35 and the problem is not limited to Zambia.^36–39 Evidence-based stigma-reduction interventions have been the focus of research for ≥10 years.⁴⁰ Integrated stigma mitigation interventions have shown success in Senegal.³⁷ Violence,⁴¹ discrimination, poor nutrition, and food insecurity are commonly reported and associated with poor linkage to care, retention in care, and ART initiation.⁴²

Fertility goals and modern nonbarrier contraceptive use patterns differed between HIV+ and HIV− FSWs, SMs, and between cities. Lusaka respondents were much less likely to report wanting more children than Ndola respondents, although their number of existing children was not different. Paradoxically however, they were also far less likely to be using a modern contraceptive method. Chanda et al. have published that half of Zambian FSWs surveyed reported an unplanned pregnancy.² A study of Kenyan FSWs showed many using modern contraceptives but reporting irregular condom use.⁴³ In our study, when education about and access to LARC methods that were often not available at government clinics were provided, uptake was <3%. This is in contrast to the high uptake of LARC methods among discordant and concordant couples in Zambia and Rwanda after similar LARC counseling and provision.^44–46 Given the low level of education and literacy, it may be that the myths and misconceptions that surround LARC in Africa are more common among FSWs and SMs compared with couples.⁴⁷ In some cases, as has been reported in other studies, FSWs may desire pregnancy^48,49 but may not share those intentions with partners,⁵⁰ clients,⁵¹ or others, including health care providers. Finally, the practice of sex work may be fluid with temporal transitions and combinations of steady and casual, paying or gifting and nonpaying partners.⁵² Effective messages and tailored services^53–55 are needed to increase dual-method use,^56,57 combining consistent condom use with a nonbarrier method to optimize prevention of HIV/STI and unplanned pregnancy.^58,59

Our study had several limitations. Our recruitment strategy targeted sexually active HIV-negative SMs at infant vaccination services in government clinics who responded positively to a referral to the research site. We were unable to collect risk profile from HIV+ SMs because the sponsor's goal was to enroll women at high risk of incident HIV infection, and we did not have information on HIV− SMs who did not respond to the referral. Some women referred from infant vaccination might have been FSWs but declined to respond positively to the question regarding exchange of sex for money, which may have resulted in some misclassification. Our recruitment strategy for FSWs also did not allow data collection for those who did not respond to the invitation. The study was conducted in two of the largest cities in Zambia. Environmental and socioeconomic factors specific to these cities may affect generalizability of the findings in this study to other areas, for example, border towns where commercial sex work is prominent or rural areas. Despite the study having these limitations, the findings are relevant because they highlight the fact that SMs who have not been identified as a group at significant risk of HIV infection have similar risk profiles to FSWs.

Conclusion

FSWs continue to be at high risk of transmitting and acquiring HIV/STIs and having unplanned pregnancies in Zambia, and prevention, care, and treatment services require urgent strengthening. Similarly, SMs had a high prevalence of STI and low reported condom use. This, combined with poverty and lack of education, makes these women highly vulnerable. Programs for AGYW should consider expanding age ranges to provide income generation options other than sex work for SMs.

Authors' Contribution

Contributions of coauthors included conceptual (all); implementation of the work and data collection (W.K., M.I., T.S., K.M., A.-M. V.); data analysis (K.W., R.P., C.H., M.P., A.T.); and article preparation (S.A., W.K., M.P., K.W.). All coauthors provided careful review and editing of the article.

Author Disclosure Statement

No competing financial interests exist.