Determinants of Condom Use Among Female Sex Workers in Kenya: A Case-Crossover Analysis

Abstract

Background:

We evaluated predictors of consistent condom use among female sex workers (FSWs), a core group for controlling the spread of HIV.

Methods:

In an analysis of data collected in 2004–2005 from 140 Kenyan FSWs who completed questionnaires administered during a baseline study visit and three bimonthly follow-up visits, we used a case-crossover design to identify predictors of consistent condom use during all coital acts in the preceding 2 weeks, overall and by partner type.

Results:

Participants (n=140) completed the baseline visit and 390 bimonthly follow-up visits. Alcohol use during sex was negatively associated with consistent condom use with helping partners (defined as regular sex partners to whom the woman could go for help or support if needed) (adjusted odds ratio [AOR], 2.6, 95% confidence interval [CI] 1.0-6.5) but not associated with condom use with other partners. Coital frequency was associated with condom use with other partners only. Women who reported 1–5 (AOR 11.0, 95% CI 4.3-28.3) or 6–9 recent coital acts (AOR 3.8, 95% CI 1.7-8.8) with other partners were more likely to report consistent condom use with those partners than were women who reported ≥10 acts. Having a recent partner delay payment was inversely associated with consistent condom use with helping, other, or all partners.

Conclusions:

Correlates of consistent condom use differed by partner type. By using a case-crossover design, we were able to identify potentially modifiable factors associated with consistent condom use by FSWs who used condoms consistently with a given partner type during some periods but not others.

Introduction

Male condoms continue to have an important role in reducing users' risk for HIV. Although male circumcision has been shown in randomized controlled trials to decrease HIV incidence by 38%–66% among heterosexual men,¹ the procedure does not appear to extend protection to women.² Despite recent, promising results on the effectiveness of tenofovir gel, no microbicide against HIV, to date, has been approved for use,^3,4 and development of an effective vaccine against the virus appears to be still far in the future.⁵ However, consistent use of male condoms by couples with discordant HIV status has been associated with an 80% reduction in risk for HIV among uninfected partners,⁶ indicating that male condoms are the most effective method of protection against HIV for sexually active women.^7,8 Condoms also provide protection against a variety of other sexually transmitted infections (STIs).^7,9

Although evidence suggests that the prevalence of HIV among female sex workers (FSWs) in Nairobi, Kenya, has decreased since 1985, this population remains at high risk for infection, with an estimated prevalence of 47% in 2005.¹⁰ FSWs in Kenya often work from public areas, including bars, restaurants, hotels, bus stop, discos, and streets, or from their homes.^11
–13 A study of FSWs working along the trans-Africa Highway in Kenya found that only 27% reported using condoms with every act.¹⁴ In another study of part-time sex workers in a suburb of Mombasa, 47% reported condom use at last sex act with a client.¹² Condom use with nonclients appears to be less common.¹⁵

The use of a case-crossover study design could improve our understanding of factors that lead women to use condoms consistently with some partners but not with others.^16
–18 With this design, variables reported by a woman during study visits after periods for which she reported consistent condom use could be compared with variables reported by the same woman during study visits after periods for which she reported using condoms inconsistently or not at all. In this way, a single woman could serve as both a case and control subject, thus eliminating the effect of time-independent factors, such as socioeconomic status or race. This is especially important for controlling for factors that might be unknown or difficult to measure, such as perceived risk for infection and social support and self-efficacy for condom use. We used a case-crossover analysis to evaluate potential predictors of consistent condom use among FSWs in Kenya.

Materials and Methods

We analyzed data from a 6-month prospective cohort study of the acceptability and safety of diaphragm use among FSWs in Nairobi, Kenya, that was conducted in 2004–2005.¹⁹ Ethical review committees at the University of Nairobi, the University of Washington, the University of California, San Francisco, and the U.S. Centers for Disease Control and Prevention (CDC) approved the research. To be eligible for the study, women had to have been ≥18 years of age and to have engaged in sex for money or gifts in the prior 2 weeks. Women were excluded from the study if they were pregnant; planned to become pregnant in the next 6 months; had a latex allergy, history of toxic shock syndrome, or gynecological contraindications to diaphragm use; or reported using a condom during every coital act in the previous 2 months. At baseline, participants received a pelvic examination and syndromic treatment for STIs consistent with the current standard of care and were fitted for and given a diaphragm and lubricant to facilitate its insertion. Because of the unknown efficacy of the diaphragm against acquisition of STIs, participants also were instructed to use a male condom for every coital act. At baseline and at follow-up visits scheduled 2, 4, and 6 months afterward, women received a sufficient supply of male condoms (usually 60 per visit but more if requested) for protection during all coital acts during the ensuing follow-up interval.

At the baseline visit and each bimonthly follow-up visit, study staff administered questionnaires to collect data on demographic characteristics, reproductive history, and sexual behaviors. Participants also were asked to use stickers to complete self-administered, pictorial questionnaires on their frequency of coitus, condom use, and diaphragm use and the number of sex partners they had had in the previous 2 weeks. Drawing on prior formative research with the target population, we asked about sexual behaviors with two types of partners on both the interviewer and self-administered questionnaires: helping partners (defined as regular sex partners to whom a woman could go for help or support if needed) and sexual behaviors with other partners (defined as any other partner). Questions were asked about sexual behaviors during the 2-week period preceding the study visit, which we defined as recent behaviors.

We defined the outcome measure, recent consistent condom use, as use of a condom during all episodes of sexual intercourse in the prior 2 weeks as reported by participants during study visits. We assessed determinants of recent consistent condom use separately for helping partners and other partners, as well as for all partners combined. All analyses were based on reported use of condoms with the relevant partner type by women who reported engaging in coitus at least once with that partner type. The analytic population was further restricted to women who reported using condoms consistently with the relevant partner type on at least one study visit and who reported not doing so on at least one study visit.

Using conditional logistic regression, we fitted individual models for the bivariable analyses to assess the relationship between each potential predictor and consistent condom use with the given partner type. We did not include time-independent factors as possible predictors because these factors could not vary, given that individual study participants served as both case subjects and matching control subjects. For the multivariable analyses, we fitted full models with all potential predictors and used manual, backward elimination to eliminate factors not significantly associated with consistent condom use based on an alpha of 0.05.²⁰ Potential predictors were selected either for their hypothesized relationship with condom use or because of their identification as correlates of condom use in the literature. These factors included number of sex partners, new sex partner, coital frequency, use of an oil-based lubricant during sex, having a partner delay in paying for sex, and participant or partner being under the influence of alcohol or bhang during sex.

Results

Of 180 women who were screened for participation in the study, 140 were deemed eligible and enrolled in the study. Of these, 126 (90%) completed questionnaires at the 6-month visit. Participants completed questionnaires at a total of 140 baseline visits and 390 bimonthly follow-up visits. The study population has been described in more detail elsewhere.¹⁹ Briefly, participants' median age was 30 years (range 18–55), and 73% were cohabiting, divorced, or widowed. At baseline, 63% of participants reported having used condoms for contraception, and 36% reported using condoms as their primary method of contraception.

The percentage of women who reported using condoms consistently increased over the course of the study: from 18% to 36% with all partners, from 18% to 34% with helping partners, and from 57% to 74% with other partners (among women reporting ≥1 coital act with the relevant partner type). Over the course of the study, women reported consistent condom use with all, helping, and other partners at 26%, 28%, and 61% of their study visits, respectively.

The case-crossover analysis of women's consistent condom use with all partners was based on responses to questionnaires by 64 women during 243 study visits. Results of the bivariable analysis showed that two factors were associated with consistent condom use with all sex partners: not being under the influence of alcohol during recent sex with a helping partner and not having recent partners delay payment for services (Table 1). These factors remained the only significant predictors in the multivariable analysis. Women who reported that they were never under the influence of alcohol during recent sex with a helping partner had an odds of using condoms consistently with all partners 2.8 times (95% confidence interval [CI] 1.0-7.7) that of women who reported sometimes having been under the influence. Women who did not report having had a recent partner delay payment for sex had an odds of consistent condom use with all partners 4.1 (95% CI 1.6-10.2) times that of women who reported having had a partner delay payment.

Table 1.

Associations Between Selected Factors and Consistent Use of Condoms with All Partners in Previous 2 Weeks

			Bivariable model		Multivariable model ^a
	Control visits No.	Case visits No.	OR	(95% CI)	AOR	(95% CI)
Number of sex partners (any type)
1–5	82	80	2.4	(0.9–6.2)
≥6	60	21	Referent
New sex partner
Yes	111	72	Referent
No	31	29	1.5	(0.6–4.1)
Coital frequency with any partner
1–7 acts	28	48	2.0	(0.6–6.3)
8–12 acts	52	28	1.9	(0.7–4.9)
≥13 acts	62	25	Referent
Used an oil-based lubricant during sex
Yes	19	10	Referent
No	123	91	2.0	(0.6–6.4)
≥1 partner delayed paying for sex
Yes	78	43	Referent		Referent
No	64	58	4.0	(1.6–9.7)	4.1	(1.6–10.2)
Under the influence of alcohol during sex with a helping partner
≥1 time	57	18	Referent		Referent
Never	60	48	2.6	(1.0–6.8)	2.8	(1.0–7.7)
A helping partner under the influence of alcohol during sex
≥1 time	76	40	Referent
Never	41	26	1.8	(0.5–7.3)
Under the influence of bhang during sex with a helping partner
≥1 time	11	2	Referent
Never	106	64	4.4	(0.5–36.7)
A helping partner under the influence of bhang during sex
≥1 time	25	12	Referent
Never	92	53	0.8	(0.2–3.1)

Findings from conditional logistic regression model based on 101 case visits (i.e., visits with recent consistent condom use) and 142 control visits (i.e., visits with recent inconsistent or noncondom use) completed by 64 women with at least one case and one control visit.

Adjusted for all variables in column.

AOR, adjusted odds ratio; CI, confidence interval; OR, odds ratio.

The analysis of consistent condom use with helping partners was based on responses to questionnaires by 50 women during 175 visits. Results of both the bivariable and multivariable analyses showed that not being under the influence of alcohol during recent coitus with a helping partner and not having a partner delay payment were the only variables associated with consistent condom use with helping partners (Table 2).

Table 2.

Associations Between Selected Factors and Consistent Use of Condoms with Helping Sex Partners in Previous 2 Weeks

			Bivariable model		Multivariable model ^a
	Control visits No.	Case visits No.	OR	(95% CI)	aOR	(95% CI)
Number of sex partners (any type)
1–5	62	52	1.4	(0.6–3.3)
≥6	39	22	Referent
New sex partner
Yes	77	53	Referent
No	24	21	1.2	(0.5–3.1)
Coital frequency with helping partner
1–3 acts	37	34	1.0	(0.5–2.1)
≥4 acts	64	40	Referent
Used an oil-based lubricant during sex
Yes	15	4	Referent
No	86	70	2.4	(0.8–7.6)
≥1 partner delayed paying for sex
Yes	53	24	Referent		Referent
No	48	50	2.9	(1.3–6.2)	2.8	(1.3–6.2)
Under the influence of alcohol during sex with a helping partner
≥1 time	52	24	Referent		Referent
Never	48	48	2.5	(1.0–6.3)	2.6	(1.0–6.5)
A helping partner under the influence of alcohol during sex
≥1 time	69	48	Referent
Never	31	24	1.3	(0.4–4.2)
Under the influence of bhang during sex with a helping partner
≥1 time	7	3	Referent
Never	93	69	2.3	(0.5–11.4)
A helping partner under the influence of bhang during sex
≥1 time	17	11	Referent
Never	83	60	1.2	(0.3–4.0)

Findings from conditional logistic regression model based on 74 case visits (i.e., visits with recent consistent condom use) and 101 control visits (i.e., visits with recent inconsistent or noncondom use) completed by 50 women with at least one case and one control visit.

Adjusted for all variables in column.

The analysis of consistent condom use with other partners was based on responses to questionnaires by 73 women during 266 visits. Results of the bivariable analysis showed that number of sex partners, coital frequency with other partners, and not having a recent partner delay payment were associated with consistent condom use (Table 3). Only the latter two of these factors, however, were associated with consistent condom use according to multivariable analysis results. Women who reported engaging in 1–5 (AOR 11.0, 95% CI 4.3-28.3) or 6–9 recent coital acts (AOR 3.8, 95% CI 1.7-8.8) with other partners were more likely to report consistent condom use than those who reported engaging in ≥10, and those who reported that their recent partners did not delay paying them were more likely (AOR 2.0, 95% CI 1.1-3.9) to report consistent condom use than those who reported delays in payment.

Table 3.

Associations Between Selected Factors and Consistent Use of Condoms with Other Sex Partners in Previous 2 Weeks

			Bivariable model		Multivariable model ^a
	Control visits No.	Case visits No.	OR	(95% CI)	AOR	(95% CI)
Number of sex partners (any type)
1–5	65	99	2.1	(1.1-4.1)
≥6	58	44	Referent
New sex partner
Yes	105	118	Referent
No	18	25	1.7	(0.7-4.1)
Coital frequency with other partner
1–5 acts	24	61	10.5	(4.2-26.2)	11.0	(4.3-28.3)
5–9 acts	40	46	3.5	(1.5-7.8)	3.8	(1.7-8.8)
≥10 acts	59	36	Referent		Referent
Used an oil-based lubricant during sex
Yes	21	18	Referent
No	102	125	1.3	(0.6-2.7)
≥1 partner delayed paying for sex
Yes	75	73	Referent		Referent
No	48	70	2.0	(1.1-3.5)	2.0	(1.1-3.9)

Findings from conditional logistic regression model based on 143 case visits (i.e., visits with recent consistent condom use) and 123 control visits (i.e., visits with recent inconsistent or noncondom use) completed by 73 women with at least one case and one control visit.

Adjusted for all variables in column.

Discussion

We found different factors associated with recent consistent condom use by Kenyan FSWs according to partner type. Prior studies have similarly shown that the frequency of condom use and the determinants of condom use can differ by type of partner.^5,18,21,22 These findings highlight the need to tailor counseling messages to women according to their types of partners. Our use of a case-crossover design allowed us to identify potentially modifiable factors associated with consistent condom use that are unlikely to be the result of uncontrolled confounding related to time-independent characteristics of study participants.

Results from previous research among FSWs in Kenya demonstrated that binge drinking was associated with engaging in unprotected sex, experiencing partner violence, and testing positive for STIs.²³ However, although we found alcohol consumption before sex to be negatively associated with consistent condom use with helping partners, we did not find it to be associated with consistent condom use with other partners. This difference by partner type could reflect differences in FSWs' motivations to consume alcohol around the time of intercourse. For example, they might consume relatively small quantities of alcohol before sex with other partners as a way of coping with the demands of their work without imbibing enough to prevent them from using condoms, whereas they might consume more when drinking with helping partners (with whom they presumably feel safer) and, as a result, be more likely to engage in disinhibited behavior, including unprotected sex. The frequency of sex, on the other hand, correlated with women's consistent use of condoms with other partners but not with their consistent use of condoms with all or helping partners. This finding could be explained, for example, if women have a set level of trust in their helping partners. Having more acts with these partners might not affect condom use. In contrast, with a less regular, less reliable other partner, as the frequency of sex increases and the partner becomes more regular, they might be less likely to use a condom. Finally, we found an inverse relationship between having a recent partner delay payment for sex and consistent condom use. Women who have other sources of income might be able to reduce their exposure to unprotected sex.²⁴ The consistency of the association between delayed payment and consistent condom use across partner types may reflect FSWs' lack of power in their relationship with these partners and a need for societal-level interventions to make social norms more supportive of condom use.

The main limitation of this study was its reliance solely on self-reported data from study participants. A recent analysis found that risk factors of engaging in unprotected sex differed according to whether the determination of condom use was based on self-reports of participants or on the identification of a biomarker for semen in vaginal fluid.²⁵ These differences suggest that self-reported data on the occurrence of unprotected sex might not correlate well with exposure to semen. Furthermore, the present study was conducted among FSWs in Nairobi, Kenya, who were supplied with condoms; study findings might not be generalizable to other populations. Finally, the factors identified in this analysis might be correlates of consistent condom use without having a causal effect on the consistent use of condoms. In this case, interventions to change these factors among FSWs would not necessarily improve their rate of condom use. Given the inability to randomize women to these risky behaviors, however, the case-crossover study might represent the best feasible source of data.

The assessment of the determinants of condom use separately by partner type and the focus on FSWs were strengths of the study. Promoting consistent male condom use among FSWs and other core groups (i.e., those with a high frequency of sexual activity with multiple concurrent partners) is considered a prerequisite to controlling the spread of HIV and other STIs in the general population,^26
–28 even in generalized epidemic settings.²⁹ A better understanding of condom use among FSWs will help in devising interventions to promote condom use among members of a subgroup with high rates of and at high risk for HIV.

An additional strength of the research was the use of the case-crossover analysis to reduce the effects of residual confounding. Although the case-crossover design can only be used to assess time-dependent variables, these factors, by definition, are precisely those that are potentially amenable to intervention. In contrast, time-independent variables, such as race or age, are immutable and cannot be influenced even when associated with the study outcome. The correlates of consistent condom use identified in the present analysis consisted of variables that varied within an individual woman across visits. Interventions should focus on determining and building on the strengths that allowed the woman to achieve correlates of condom use at some of the visits. Thus, the use of the case-crossover design is a methodological advancement in the evaluation of time-dependent determinants of condom use and should be considered by researchers designing future studies on this topic.

Footnotes

Acknowledgments

This study was funded by the U.S. Centers for Disease Control and Prevention through an interagency agreement with the U.S. Agency for International Development and CONRAD. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Disclosure Statement

No competing financial interests exist.

References

Siegfried

, Muller

, Deeks

et al. Male circumcision for prevention of heterosexual acquisition of HIV in men. Cochrane Database Syst Rev, 2009; 2:CD003362.

Wawer

, Makumbi

, Kigozi

et al. Circumcision in HIV-infected men and its effect on HIV transmission to female partners in Rakai, Uganda: A randomised controlled trial. Lancet, 2009; 374:229–237.

Abdool Karim

, Abdool Karim

, Frohlich

et al. Effectiveness and safety of tenofovir gel, an antiretroviral microbicide, for the prevention of HIV infection in women. Science, 2010; 329:1168–1174.

Grant

, Hamer

, Hope

et al.

Whither or wither microbicides?

Science, 2008; 321:532–534.

Day

. AIDS expert doubts vaccine will be found in near future. BMJ, 2007; 334:1133.

Weller

, Davis

. Condom effectiveness in reducing heterosexual HIV transmission. Cochrane Database Syst Rev, 2002; 1:CD003255.

Holmes

, Levine

, Weaver

. Effectiveness of condoms in preventing sexually transmitted infections. Bull WHO, 2004; 82:454–461.

National Institute of Allergy and Infectious Diseases. Workshop summary: Scientific evidence on condom effectiveness for sexually transmitted disease (STD) prevention. Bethesda, MD: National Institutes of Health, National Institute of Allergy and Infectious Diseases, 2001.

Warner

, Stone

, Macaluso

et al. Condom use and risk of gonorrhea and chlamydia: A systematic review of design and measurement factors assessed in epidemiologic studies. Sex Transm Dis, 2006; 33:36–51.

10.

Kimani

, Kaul

, Nagelkerke

et al. Reduced rates of HIV acquisition during unprotected sex by Kenyan female sex workers predating population declines in HIV prevalence. AIDS, 2008; 22:131–137.

11.

Elmore-Meegan

, Conroy

, Agala

. Sex workers in Kenya, numbers of clients and associated risks: An exploratory survey. Reprod Health Matters, 2004; 12:50–57.

12.

Hawken

, Melis

, Ngombo

et al. Part time female sex workers in a suburban community in Kenya: A vulnerable hidden population. Sex Transm Infect, 2002; 78:271–273.

13.

Luchters

, Chersich

, Rinyiru

et al. Impact of five years of peer-mediated interventions on sexual behavior and sexually transmitted infections among female sex workers in Mombasa, Kenya. BMC Public Health, 2008; 8:143.

14.

Morris

, Morris

, Ferguson

. Sexual behavior of female sex workers and access to condoms in Kenya and Uganda on the Trans-Africa highway. AIDS Behav, 2009; 13:860–865.

15.

Voeten

, Egesah

, Varkevisser

et al. Female sex workers and unsafe sex in urban and rural Nyanza, Kenya: Regular partners may contribute more to HIV transmission than clients. Trop Med Int Health, 2007; 12:174–182.

16.

Maclure

. The case-crossover design: A method for studying transient effects on the risk of acute events. Am J Epidemiol, 1991; 133:144–153.

17.

Warner

, Macaluso

, Austin

et al. Application of the case-crossover design to reduce unmeasured confounding in studies of condom effectiveness. Am J Epidemiol, 2005; 161:765–773.

18.

Tassiopoulos

, Kapiga

, Sam

et al. A case-crossover analysis of predictors of condom use by female bar and hotel workers in Moshi, Tanzania. Int J Epidemiol, 2009; 38:552–560.

19.

Bukusi

, Gallo

, Sharma

et al. Adherence to diaphragm use for infection prevention: A prospective study of female sex workers in Kenya. Infect Dis Obstet Gynecol, 2009; 2009article 420196.

20.

Kleinbaum

, Kupper

, Muller

et al. Applied regression analysis and multivariable methods. New York: Duxbury Press, 2007.

21.

Foss

, Hossain

, Vickerman

et al. A systematic review of published evidence on intervention impact on condom use in sub-Saharan Africa and Asia. Sex Transm Infect, 2007; 83:510–516.

22.

Macaluso

, Demand

, Artz

et al. Partner type and condom use. AIDS, 2000; 31:537–546.

23.

Chersich

, Luchters

, Malonza

et al. Heavy episodic drinking among Kenyan female sex workers is associated with unsafe sex, sexual violence, and sexually transmitted infections. Int J STD AIDS, 2007; 18:764–769.

24.

Odek

, Busza

, Morris

et al. Effects of micro-enterprise services on HIV risk behaviour among female sex workers in Kenya's urban slums. AIDS Behav, 2009; 13:449–461.

25.

Gallo

, Steiner

, Hobbs

et al. Predictors of unprotected sex: Comparison of semen biomarker and self-reported data. AIDS Behav, 2010; 14:1279–1286.

26.

Boily

, Lowndes

, Alary

. The impact of HIV epidemic phases on the effectiveness of core group interventions: Insights from mathematical models. Sex Transm Infect, 2002; 78:i78–90.

27.

Lowndes

, Alary

, Meda

et al. Role of core and bridging groups in the transmission dynamics of HIV and STIs in Cotonou, Benin, West Africa. Sex Transm Infect, 2002; 78:i69–77.

28.

Plummer

, Nagelkerke

, Moses

et al. The importance of core groups in the epidemiology and control of HIV-1 infection. AIDS, 1991; 5:S169–S176.

29.

Morris

, Ferguson

. Estimation of the sexual transmission of HIV in Kenya and Uganda on the trans-Africa highway: The continuing role for prevention in high-risk groups. Sex Transm Infect, 2006; 82:368–371.