Secondary Logo

Journal Logo

Supplement Article

Preventing Unintended Pregnancy and HIV Transmission

Effects of the HIV Treatment Cascade on Contraceptive Use and Choice in Rural KwaZulu-Natal

Raifman, Julia SM*,†; Chetty, Terusha FCPHM, MBChB, DCH, MMed; Tanser, Frank PhD; Mutevedzi, Tinofa MSc; Matthews, Philippa MBBS, DCH, FRCGP; Herbst, Kobus MBChB, MSc, FFCH; Pillay, Deenan MBBS, PhD†,‡; Bärnighausen, Till MD, ScD*,†

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: December 1, 2014 - Volume 67 - Issue - p S218-S227
doi: 10.1097/QAI.0000000000000373



All women have the reproductive health rights “to decide freely and responsibly on the number and spacing of their children and to have access to the information, education, and means to enable them to exercise these rights.”1 For all women, the ability to freely choose the method of contraception that best fulfills her individual reproductive health needs and wants is an essential component of these rights.2 Among women living with HIV, prevention of unintended pregnancy is an effective approach to prevent mother-to-child transmission of the virus.3 In making contraceptive choices, women living with HIV have to consider a number of risks that are different from those that HIV-uninfected women are facing. Compared with HIV-uninfected women, HIV-infected women are at greater risk of morbidity and mortality during pregnancy and motherhood4 and are at increased risk of severe illness from sexually transmitted infections (STIs) other than HIV.5–7 HIV-infected women also face the risk of superinfection with a second strain of HIV, which may cause more rapid disease progression and limit treatment options.8,9 Finally, women living with HIV are at risk of transmitting HIV to their uninfected partners.

Male and female condoms can provide dual protection against unintended pregnancy as well as acquisition and transmission of STIs, including HIV. Other methods of contraception, such as oral and injectable contraceptive drugs and male or female sterilization can prevent unintended pregnancy but do not serve the additional purpose of protecting against STI acquisition and transmission. Although condoms alone provide dual-protection, dual-method contraception using both condoms and another method is more effective for preventing unintended pregnancies than condom use alone.

In the following, we will use the term single-method dual protection to indicate condom use alone and dual-method dual protection to indicate concurrent use of condoms and at least one other contraceptive method. We will use the term single protection to indicate contraception without condoms. Although women who are HIV infected can use all of the same contraceptive methods as women who are HIV uninfected, WHO recommends that HIV-infected women use dual protection, and ideally dual-method dual protection to maximize effectiveness in preventing both pregnancy and STI acquisition and transmission.10

The large-scale use of antiretroviral treatment (ART) has changed what it means to live with HIV and to live in one of the communities in sub-Saharan Africa that are severely affected by the HIV epidemic.11 ART substantially reduces HIV-related mortality12 and can dramatically improve life expectancy in communities with high HIV prevalence.11 By reducing the concentration of HIV in body fluids, ART can also substantially decrease the risk of HIV transmission from an infected to an uninfected partner.13 Although these biological effects of antiretroviral medication are well established, our knowledge of the behavioral effects of ART programs is limited.

The HIV patient pathway from infection to long-term treatment, the “HIV treatment cascade,”14 can be divided into several steps. First, an HIV-infected woman learns about her positive HIV status in an HIV testing and counseling session, which typically conveys information about HIV infection, options for long-term care and treatment, the importance of disclosure of HIV status to family members and sexual partners, as well as approaches to prevent onward transmission of HIV.15 Next, the HIV-infected woman can enroll in pre-ART programs for regular review of ART eligibility, prevention of opportunistic infections, contraceptive counseling, and counseling to prevent onward transmission of HIV.16–18 At some point after becoming eligible for HIV treatment, the HIV-infected woman may initiate ART; ART initiation is usually preceded by treatment education including information on the importance of ART adherence, disclosure, and practicing safe sex behaviors.19 Finally, as the HIV-infected woman remains enrolled in the ART program, she will regularly visit ART clinics for assessment of treatment success and continued counseling, including on contraception and prevention of HIV transmission. Progression through the HIV treatment cascade is thus associated with ongoing counseling and knowledge gain on contraceptive choices for HIV-infected women and prevention of HIV transmission. Additionally, the repeated interactions with the health system along the cascade can imply access to contraceptive methods. HIV testing and counseling centers and ART clinics commonly provide male and female condoms, and they are often located close to other health care facilities, such as primary care clinics and family planning centers,20 where contraceptive methods are available. It is thus plausible that HIV-infected women will increasingly use contraception as they advance from one stage in the cascade to the next. However, the type and magnitude of any such effects is unknown.

Here, we use a rare data opportunity—ART program data that has been linked to population-based surveillance data—to examine whether progression through the HIV treatment cascade affected contraceptive use among the HIV-infected women in a community in rural KwaZulu-Natal with high HIV prevalence21 and incidence.22 A previous study in the same community found that ART coverage of HIV-infected populations protected HIV-uninfected individuals from acquiring HIV.23 In addition to the biological effect of ART, one of the potential causal mechanisms underlying this effect of ART coverage on HIV acquisition could be effects of the ART scale-up on dual protection. To elucidate this possible behavioral pathway from ART scale-up to HIV incidence, we estimate the effects of progression through the HIV treatment cascade on single- and dual-protection contraception.


Setting and Data Collection

We use data collected by the Wellcome Trust Africa Centre for Health and Population Studies (Africa Centre). Since 2000, the Africa Centre has operated a longitudinal Health and Demographic Surveillance System, covering the entire population living in a 438 km2 surveillance area (about 100,000 individuals) in the rural uMkhanyakude district in northern KwaZulu-Natal, South Africa.24 HIV prevalence in the adult population in this community was 29% in 201121 and incidence has been around 3 per 100 person-years for the last decade,22 with a slight decline in recent years.23 ART coverage of all HIV-infected adults in the community has risen from 0% in 2003 to 31% in 2011.21 The surveillance includes longitudinally linked annual HIV testing but the HIV test results are not provided to the surveillance participants. The people living in the surveillance area can test for HIV free of charge at public-sector HIV testing and counseling centers and primary care clinics. They can also test for HIV in private-sector physician practices and pharmacies. During the individual surveillance interviews, all respondents are asked whether they know their HIV status and all women are asked about their contraceptive use. The surveillance also includes linked longitudinal data on demographic, social, and economic factors. To determine progression through the HIV treatment cascade, we linked the data on clinic visits and ART initiation collected in the local public-sector ART program to the population-based surveillance data, using the South African identification number, first name, last name, and birth dates for linkage.25

In 2004, the South African Department of Health in collaboration with the Africa Centre started the Hlabisa HIV Treatment and Care Programme with support from the Presidential Emergency Fund for AIDS Relief (PEPFAR). The program delivers ART through the 17 public-sector primary care clinics in Hlabisa subdistrict. The program provides free HIV testing and counseling, ART, and male and female condoms; it also includes an active pre-ART component enrolling patients for ongoing counseling and monitoring of CD4 count, HIV disease progression, and health status to determine ART eligibility. Before ART initiation, all patients participate in three group sessions and individual counseling. After initiation, patients with ART make monthly visits to the program to see a nurse and a counselor and to participate in group and individual counseling sessions.20 Because women who intend to be pregnant or are “not on reliable contraception” should receive a different first-line ART regimen than other patients according to the South African national ART guidelines,18 contraception and fertility intentions should be part of the conversation that health care providers have with their ART patients during the monthly clinic visits.

Study Population

The study population included all women who met the following eligibility criteria: they were of reproductive age (15–49 years); they had either tested HIV positive in the Africa Centre HIV surveillance or were enrolled in the Hlabisa HIV Treatment and Care Programme; they reported being sexually active within the past year; and they reported on their contraceptive use in the Women's General Health Survey. We used the latest report on contraceptive use available for each woman who met these eligibility criteria. We used data beginning in 2005 because the Hlabisa HIV Treatment and Care Programme started enrolling patients only at the end of 2004.20 We use data from the observation period 2005–2012.

Contraceptive Use Variables

Until 2009, the field workers in the surveillance inquired about contraceptive use in the individual interviews by asking “Are you currently doing anything, or using any contraceptive method, to delay or avoid getting pregnant?” If a woman answered yes, interviewers asked her to specify which methods she was using; the precoded answer options included the “pill,” “intrauterine device,” “Depo-Provera injection,” “Nur-Isterate injection,” “male condom,” “female condom,” “female sterilization,” “male sterilization,” and “other.” The question changed slightly in 2009. The surveillance interviewers now asked “Have you ever used contraception?” Women who answered yes were then asked “Which method are you currently using?” and could choose among the precoded answer options: “none,” “male condom,” “female condom,” “female sterilization,” “male sterilization,” “injections,” “pill,” and “other.” One reason for this change in the question was that hardly any women interviewed before 2009 had reported intrauterine device use. As defined above, we categorized the different contraceptive methods into single protection, single-method dual-protection, and dual-method dual-protection contraception (Table 2).

Explanatory Variables

We captured progression through the HIV treatment cascade with dummy variables indicating knowledge of HIV status, enrollment in pre-ART, and having received ART for 0–1, 1–2, 2–4, and 4–7 years. In addition, in the multivariable regression analysis, we controlled for variables that have been found to determine contraceptive use in other studies26–28: age, education, relationship status, parity, current pregnancy, self-reported health status, the distance from a woman's place of residence to the nearest primary and the nearest secondary road, household wealth, and calendar year. With the exception of the ART program data (pre-ART and time on ART), all other information, including awareness of HIV status, was collected by the Africa Centre surveillance. We used school grade attainment data to capture education. We coded women as being married or in a marriage-like relationship if she reported that she was married, engaged, or cohabitating. We coded age in years and included age squared to capture nonlinear age relationships with contraception use. Following a previous study in this community,29 we created wealth quintiles based on the ranking of individuals on the first principal component obtained in a principal component analysis of information on 27 household assets, such as vehicles, stoves, beds, and livestock. We included the distances from a woman's place of residence to the nearest primary and the nearest secondary road to capture geographical access to health care, because car ownership is rare in this community and people usually walk to the nearest road to fetch a mini bus to drive to a health care facility.30


Our primary research question here was whether progression through the HIV treatment cascade affected single- and dual-protection contraception. To answer this question, we chose the bivariate probit model, because the two binary decisions—whether or not to use single-protection contraception (ie, any contraceptive method except for condoms) and whether or not to use single-method dual-protection contraception (ie, condoms)—are likely dependent. We except dependency of the two decisions based on both economic theory (the two contraceptive approaches are imperfect substitutes) and the previous empirical literature.31,32 In addition to the bivariate probit regression coefficients (Table 3), we estimated average marginal effects (AMEs) for not using any contraception, using single protection, using single-method dual protection, and using dual-method dual protection (Table 4). Conceptually, the AME for a dummy variable, such as one of the variables representing a stage in the HIV treatment cascade, is the average across all the individual marginal effects for that dummy variable for each person in the data set. These individual marginal effects are obtained by computing each person's probability of having the outcome when the dummy variable is set to zero and when it is set to unity, in both cases keeping the values of all the other explanatory variables to the values given for that person.33 The AMEs in Table 4 represent the change in the probability of having the outcome when a certain stage of the cascade is reached, compared to the stage that is the reference category. The AME are shown in percentage points (pp). For instance, a woman who has been on ART for 4–7 years is 21.6 pp more likely to use single-method dual protection compared with an HIV-infected woman who does not know her HIV status.


There were 7443 HIV-infected women aged 15–49 years who participated in the Africa Centre Health and Demographic Surveillance between 2005 and 2012. Of these women, 5510 (74.0%) reported on their sexual activity at least once, and 4625 (83.9%) of the women who reported on their sexual activity had been sexually active within the past year. Among the 4625 women who had been sexually active, data on all variables for the multiple regression analysis were available for 3169 (68.5%). Here, we present the complete-case analyses of this sample of 3169 women.

Table 1 describes the characteristics of the 3169 women in this sample. The majority of the HIV-infected women had not yet enrolled in the ART program (55%). Among the remaining women, half were enrolled in pre-ART and half were on ART. More observations occurred in the latter half of the observation period (68% in 2009–2012) than in the earlier half (32% in 2005–2008). Table 2 shows the distribution of contraceptive methods across the women in this sample of sexually active HIV-infected women; 54% used contraception, and 32% used either single- or dual-method dual protection. Figure 1 shows descriptively contraceptive choice through the HIV treatment cascade. Overall, contraceptive use increased steadily across the stages of the cascade from <40% among HIV-infected women who did not know their status to >70% among women who had received ART for 4–7 years. The increase in contraceptive use occurred largely due to an increase in the use of dually protective methods.

Sample Characteristics
Distribution of Contraceptive Use
Progression through the HIV treatment cascade and contraceptive use.

These trends were even more pronounced when we estimated the effect of progression through the HIV treatment cascade on contraceptive use in bivariate probit analysis, controlling for age, education, partnership status, pregnancy status, parity, health status, household wealth, distance to the nearest primary and secondary roads, and calendar year. The coefficient ρ, which measures the correlation between the error terms of the two regressions that we jointly estimated in the analysis, was negative (−0.292) and highly significant (P < 0.0001). This correlation confirms that the two contraceptive choices should indeed be jointly estimated because of a relationship between the choices that is not found in the observed explanatory variables. Table 3 shows the regression coefficients and Table 4 the AME from this analysis. Compared with HIV-infected women who were unaware of their positive HIV status, the likelihood of single-method dual protection increased by 4.6 pp when women became aware of their HIV status (P = 0.030), by 10.3 pp when they initiated ART (P = 0.003), and by 21.6 pp when they had received ART for 4–7 years (P < 0.001). The likelihood of dual-method dual protection increased by 3.5 pp when women became aware of their HIV status (P = 0.001), by 5.2 pp when they initiated ART (P = 0.007), and by 11.2 pp when they had received ART for 4–7 years (P < 0.001).

Effects of Progression Through the HIV Treatment Cascade on Contraception: Bivariate Probit Regression Coefficients Effects
TABLE 4-a:
Effects of Progression Through the HIV Treatment Cascade on Contraception: AME
TABLE 4-b:
Effects of Progression Through the HIV Treatment Cascade on Contraception: AME

As robustness checks of the findings presented here, we repeated the analyses with HIV-uninfected women also included in the sample and after multiple imputation of missing covariates among women who reported being sexually active. The findings from the analysis that includes HIV-negative women are described in detail in the online Appendix (see Supplemental Digital Content,, including the full tables with the descriptive statistics and the regression results. This additional analysis has several advantages (large sample size, ability to compare contraceptive choice by HIV status) but it may also suffer from reverse causality bias because contraceptive choice is an important determinant of HIV status. However, the findings based on the sample including both HIV-infected and HIV-uninfected women are essentially the same as those based on the smaller sample including only HIV-infected women. The findings from the multiply imputed analysis are also similar to those of the main analysis, suggesting that the missing data do not cause significant bias. Finally, changes in the coding and functional forms of the explanatory variables did not lead to any significant changes in the results. We present these additional findings in the online Appendix (see Supplemental Digital Content,, where we also describe and interpret how the other explanatory variables affected contraceptive choice.


We examine for the first time the effect of progression through the HIV treatment cascade on contraceptive use. Among HIV-infected women, dually protective methods of contraception can prevent unintended pregnancies, HIV transmission, and the acquisition of other STIs. In a poor, rural community in KwaZulu-Natal, South Africa, we find that both overall contraceptive use and dual-protection contraception increased significantly as HIV-infected women moved from earlier to later stages in the treatment cascade. Descriptively, the probability of contraception increased from <40% among HIV-infected women who did not know their status to >70% among HIV-infected women 4–7 years on ART. Controlling for a wide range of potential confounders of the relationship between the stages of the treatment cascade and contraceptive use, we find that progression through the cascade significantly increased the overall probability of contraception as well as single- and dual-method dual protection.

Although significant increases in dual protection occurred across the entire cascade, these increases were substantially larger after ART initiation compared with the stages in the cascade when women learnt of their HIV status or were enrolled in pre-ART care. The large ART-associated increases are plausible based on several mechanisms. First, in preparation for ART women receive intensive counseling, including on methods to prevent transmission of HIV to sexual partners. Second, women on ART are likely to discuss their contraceptive behaviors and fertility intentions with ART health workers during the routine ART follow-up visits. Health workers should routinely initiate such discussions, because based on the South African national ART guidelines18 the ART regimen needs to be changed when a woman intends to become pregnant or stops using reliable contraception. These discussions offer repeated opportunities for education on the benefits of contraception with condoms. Third, the ART clinics in this community provide male and female condoms free of charge, so that ART clinic visits imply access to dually protective contraceptives. It is possible that the availability of free condoms in the ART program is a reason for the larger increases observed in single-method dual protection than in dual-method dual protection. Last, the ART clinics are located on the premises of the primary-care clinics and thus in close proximity to family planning and reproductive health services, where contraception information and condoms are available. Future research needs to elucidate whether information and counseling or condom availability is responsible for the large effect of ART on dual protection observed in this study. It will also be important to explore whether more intensive counseling and increased condom availability in HIV testing centers and pre-ART clinic visits could increase use of dually protective methods early in the HIV treatment cascade.

Our results have several important implications for policy and research. First, the ART effects on contraception with condoms could enhance biological treatment-as-prevention effects.13,23 The effects of progression through the HIV cascade found here could also counteract increased sexual risk taking among HIV-uninfected populations in response to ART scale-up. Such “risk compensation”34 behavior has been hypothesized because the availability of ART decreases both the risk of contracting HIV through unprotected sex as well as the expected health losses after contracting HIV. However, the evidence on “risk compensation” and its potential consequences for HIV incidence is weak, and it is possible that counteracting factors, such as ART-associated behavior change in HIV-infected people has prevented its manifestation.

Second, despite the significant and large increases in dual protection across the HIV treatment cascade, in all cascade stages large proportions of HIV-infected women continued using only single-protection contraception. Although there are significant effects of learning about one's positive HIV status on both overall contraceptive use and contraception with condoms, these effects are small compared with the effects of ART. Future intervention research is needed to determine how HIV counseling and testing can be enhanced to achieve larger dual-protection effects than currently.

Third, although dual-method dual protection increased as women progressed through the HIV treatment cascade, these increases were small relative to the increases in the use of single-method dual protection, which is not as effective as dual-method dual protection in preventing unintended pregnancy. Future research needs to establish what interventions—for example, targeted provision of contraceptives, new types of contraceptives, or stronger incentives to use contraceptives—can lead to additional condom use among women who currently use other contraceptives and the addition of other contraceptives among women who currently use condoms.

Our study has several strengths but it also has important limitations. One strength of this study is that information about contraception is elicited in the community and not in patient interviews after HIV counseling or visits to an ART clinic, where previous studies have elicited this information.35−39 Although we cannot rule out social desirability biases, such biases seem much less likely when questions about contraception are asked in patients' homes and as part of an interview on a wide range of issues rather than in clinics after patients have just been counseled on a range of ART-related issues, including on prevention of HIV transmission. Home-based interviews are removed from the social norm-setting context associated with ART and HIV counseling. Additionally, unlike in patient interviews, the fieldworkers conducting home-based interviews are unaware of the HIV status of their interviewees; social norms related to HIV status are thus unlikely to affect responses.

Other strengths of this study include the large sample size and the fact that we could here for the first time directly compare the effects of different important stages across the HIV treatment cascade, including gaining HIV status knowledge, pre-ART, and ART initiation. An important limitation is that we cannot rule out that unobserved confounders have biased the observed relationships between the stages of the cascade and contraceptive choice. One important unobserved factor that could have confounded our results is fertility intention. Fertility intention may decrease when a woman learns about her positive HIV status;35,36 in this case, the estimated effect of HIV awareness on contraceptive choice found in this study may be an overestimate of the true effect. Conversely, fertility intentions may increase after ART initiation as a woman's health and future outlook improves;37–39 in this case, the effects of ART on contraception and dual protection found in this study may be underestimates of the true effects. Follow-up studies need to establish causal effects more firmly. Because we cannot randomly assign individuals to different stages in the cascade, quasi-experimental studies will be the only option to strengthen causal inference about the effects of the cascade on contraceptive choice. Examples of quasi-experimental approaches that could be feasible for this purpose include instrumental variable approaches (using, for instance, distance to the nearest ART clinic as an instrument for ART initiation) or regression discontinuity designs using the fact that ART is initiated in patients by applying a threshold rule to the continuous variable CD4 count.40,41


Progression through the HIV treatment cascade significantly increased the likelihood of contraception in general and contraception with condoms in particular. The largest increases in contraception with condoms occurred after ART initiation. Future integration of HIV and reproductive health services can build on these achievements to further increase the use of dual-protection contraception, especially in the early stages of the HIV treatment cascade. Our results further suggest that ART programs contribute to HIV prevention through the behavioral pathway of changing contraception uptake and choice.


1. The United Nations' Division for the Advancement of Women. Convention on the Elimination of All Forms of Discrimination Against Women. New York, NY: United Nations; 1979.
2. Center for Reproductive Rights and United Nations Population Fund. The Right to Contraceptive Information and Services for Women and Adolescents. New York, NY: Center for Reproductive Rights; 2010.
3. Wilcher R, Cates W. Reproductive choices for women with HIV. Bull World Health Organ. 2009;87:833–839.
4. Lieve V, Shafer LA, Mayanja BN, et al.. Effect of pregnancy on HIV disease progression and survival among women in rural Uganda. Trop Med Int Health. 2007;12:920–928.
5. Luchters SM, Vanden Broeck D, Chersich MF, et al.. Association of HIV infection with distribution and viral load of HPV types in Kenya: a survey with 820 female sex workers. BMC Infect Dis. 2010;10:18.
6. Jamieson DJ, Duerr A, Klein RS, et al.. Longitudinal analysis of bacterial vaginosis: findings from the HIV epidemiology research study. Obstet Gynecol. 2001;98:656–663.
7. Kissinger P, Amedee A, Clark RA, et al.. Trichomonas vaginalis treatment reduces vaginal HIV-1 shedding. Sex Transm Dis. 2009;36:11–16.
8. Smith DM, Richman DD, Little SJ. HIV superinfection. J Infect Dis. 2005;192:438–444.
9. Redd AD, Mullis CE, Serwadda D, et al.. The rates of HIV superinfection and primary HIV incidence in a general population in Rakai, Uganda. J Infect Dis. 2012;206:267–274.
10. WHO. HIV and Hormonal Contraception. Geneva, Switzerland: WHO; 2012.
11. Bor J, Herbst AJ, Newell ML, et al.. Increases in adult life expectancy in rural South Africa: valuing the scale-up of HIV treatment. Science. 2013;339:961–965.
12. Mills EJ, Bakanda C, Birungi J, et al.. Life expectancy of persons receiving combination antiretroviral therapy in low-income countries: a cohort analysis from Uganda. Ann Intern Med. 2011;155:209–216.
13. Cohen MS, Chen YQ, McCauley M, et al.. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med. 2011;365:493–505.
14. UNAIDS. Access to Antiretroviral Therapy in Africa. Geneva, Switzerland: UNAIDS; 2013.
15. WHO. Guide for Monitoring and Evaluating National HIV Testing and Counseling (HCT) Programmes: Field-test Version. Geneva, Switzerland: WHO; 2011.
16. WHO HIV/AIDS Department. Priority Interventions: HIV/AIDS Prevention, Treatment and Care in the Health Sector. Geneva, Switzerland: WHO; 2009.
17. WHO HIV/AIDS Programme. Essential Prevention and Care Interventions for Adults and Adolescents Living with HIV in Resource-limited Settings. Geneva, Switzerland: WHO; 2008.
18. South African Department of Health. The South African Antiretroviral Treatment Guidelines 2010. Pretoria, South Africa: Department of Health; 2010.
19. Interagency Task Team on Education. HIV and AIDS: Treatment Education. Geneva, Switzerland: UNAIDS; 2006.
20. Houlihan CF, Bland RM, Mutevedzi PC, et al.. Cohort profile: Hlabisa HIV treatment and care programme. Int J Epidemiol. 2011;40:318–326.
21. Zaidi J, Grapsa E, Tanser F, et al.. Dramatic increase in HIV prevalence after scale-up of antiretroviral treatment. AIDS. 2013;27:2301–2305.
22. Bärnighausen T, Tanser F, Newell ML. Lack of a decline in HIV incidence in a rural community with high HIV prevalence in South Africa, 2003–2007. AIDS Res Hum Retroviruses. 2009;25:405–409.
23. Tanser F, Bärnighausen T, Grapsa E, et al.. High coverage of ART associated with decline in risk of HIV acquisition in rural KwaZulu-Natal, South Africa. Science. 2013;339:966–971.
24. Tanser F, Hosegood V, Bärnighausen T, et al.. Cohort profile: Africa Centre Demographic Information System (ACDIS) and population-based HIV survey. Int J Epidemiol. 2008;37:956–962.
25. Bor J, Bärnighausen T, Newell C, et al.. Social exposure to an antiretroviral treatment programme in rural KwaZulu-Natal. Trop Med Int Health. 2011;16:988–994.
26. Ainsworth M, Beegle K, Nyamete A. The impact of women's schooling on fertility and contraceptive use: a study of fourteen sub-Saharan African countries. World Bank Econ Rev. 1996;10:85–122.
27. Kapiga SH, Lwihula GK, Shao JF, et al.. Predictors of AIDS knowledge, condom use and high-risk sexual behaviour among women in Dar-es-Salaam, Tanzania. Int J STD AIDS. 1995;6:175–183.
28. Hendriksen ES, Pettifor A, Lee SJ, et al.. Predictors of condom use among young adults in South Africa: the reproductive health and HIV research unit national youth survey. Am J Public Health. 2007;97:1241–1248.
29. Bärnighausen T, Hosegood V, Timaeus IM, et al.. The socioeconomic determinants of HIV incidence: evidence from a longitudinal, population-based study in rural South Africa. AIDS. 2007;21(suppl 7):S29–S38.
30. Tanser F, Gijsbertsen B, Herbst K. Modelling and understanding primary health care accessibility and utilization in rural South Africa: an exploration using a geographical information system. Soc Sci Med. 2006;63:691–705.
31. Rossier C, Leridon H. The pill and the condom, substitution or association? An analysis of the contraceptive histories of young women in France, 1978–2000. Population. 2004;59:387–414.
32. Gray Collins E, Hershbein B. The Impact of Subsidized Birth Control for College Women: Evidence From the Deficit Reduction Act. Report 11-737. Ann Arbor, MI: University of Michigan Population Studies Center; 2011.
33. Bartus T. Estimation of marginal effects using margeff. Stata J. 2005;5:309–329.
34. Cassell MM, Halperin DT, Shelton JD, et al.. Risk compensation: the Achilles' heel of innovations in HIV prevention? BMJ. 2006;332:605–607.
35. Hoffman IF, Martinson FE, Powers KA, et al.. The year-long effect of HIV-positive test results on pregnancy intentions, contraceptive use, and pregnancy incidence among Malawian women. J Acquir Immune Defic Syndr. 2008;47:477–483.
36. Heys J, Kipp W, Jhangri GS, et al.. Fertility desires and infection with the HIV: results from a survey in rural Uganda. AIDS. 2009;23(suppl 1):S37–S45.
37. Homsy J, Bunnell R, Moore D, et al.. Reproductive intentions and outcomes among women on antiretroviral therapy in rural Uganda: a prospective cohort study. PLoS One. 2009;4:e4149.
38. Myer L, Carter RJ, Katyal M, et al.. Impact of antiretroviral therapy on incidence of pregnancy among HIV-infected women in sub-Saharan Africa: a cohort study. PLoS Med. 2010;7:e1000229.
39. Schwartz SR, Mehta SH, Taha TE, et al.. High pregnancy intentions and missed opportunities for patient-provider communication about fertility in a South African cohort of HIV-positive women on antiretroviral therapy. AIDS Behav. 2012;16:69–78.
40. Moscoe E, Bor J, Bärnighausen T. Regression discontinuity designs in medicine, epidemiology, and public health: a review of current and best practice. J Clin Epidemiol. 2014. In press.
41. Bor J, Moscoe E, Mutevedzi P, et al.. Regression discontinuity designs in epidemiology: causal inference without randomized trials. Epidemiology. 2014;25(5):729–737.

unintended pregnancy; HIV; AIDS; reproductive health; contraception; condoms; HIV transmission

Supplemental Digital Content

© 2014 by Lippincott Williams & Wilkins