Tsai, Alexander C.a; Subramanian, S.V.b
Structural violence, including unequal gender relations and gendered structural constraints, has been recognized as a major factor compromising women's health and facilitating the spread of HIV among women [1–6]. This body of work suggests that the proximate context of violence against women and unequal gender-based norms about violence in the context of intimate relationships creates a ‘culture of fear’  or a ‘culture of terror’  that undermines women's ability to engage in HIV-risk reduction behaviors such as negotiating condom use in sexual relationships. However, previous studies have measured violence against women at the individual level [9–13], somewhat limiting their interpretation as providing evidence of contextual effects. Few studies have examined how the proximate context of violence against women or norms about violence against women constrains their ability to negotiate condom use in sexual relationships [14,15]. We undertook this study to determine whether the proximate context of norms about violence against women is associated with lack of condom use among sexually active women in sub-Saharan Africa.
For this analysis we used data from 22 Demographic and Health Surveys (DHS) conducted between 2003–2009 that were administered in continental sub-Saharan Africa (Table S1, http://links.lww.com/QAD/A189). Details on staff training, pretesting, and other survey procedures are detailed in the DHS final reports for each country, which are publicly available from the ICF Macro web site (http://www.measuredhs.com). The primary outcome of interest was self-reported condom use at last sexual intercourse, which was only obtained from participants who reported that they were sexually active. We therefore further restricted our analyses to sexually active women only. To measure individual beliefs about violence against women, we created a scale consisting of five questions about whether the respondent agreed with the appropriateness of a husband beating his wife under five different hypothetical scenarios (Table S2, http://links.lww.com/QAD/A189). Responses were summed across the five questions to generate an ordinal scale ranging from zero (wife beating not appropriate in any scenario) to five (wife beating appropriate in all scenarios). We employed exploratory factor analysis to identify the number of factors. Internal consistency was estimated using the Cronbach's α statistic in the pooled sample and for each country separately. To measure the proximate context of norms about violence against women, we created a derived variable at the level of the primary sampling unit (PSU) by averaging the absolute value of the scale across all women in the PSU.
To estimate the association between condom use and our explanatory variables of interest, we fit a multivariable logistic regression model to the pooled data, with country-level fixed effects and cluster-correlated robust estimates of variance [16–18]. This modeling approach appropriately recognizes that variables measured at the level of the PSU have a smaller effective sample size, correcting the standard errors for potential correlations between participants who live in the same PSU, whereas maintaining the population-average interpretation [19,20]. Our estimates of association were adjusted for potential confounding by the following sociodemographic variables: age, domestic partnership status, household headship, fertility preferences, urban residence, professional occupation status , educational attainment, within-country quintiles of household asset wealth , and religion. A Wald-type F-test was used to test the joint statistical significance of the country-level fixed effects. Multivariable regression models were refit to the data from each country sample separately. All analyses were conducted using the Stata statistical software package (version 11.0, StataCorp L.P., College Station, Texas, USA).
To assess the robustness of the observed associations, we conducted several sensitivity analyses. First, we assessed for interactions between contextual exposures and domestic partnership status and educational attainment. Second, because of the highly negative meanings that may be attached to condom use in the context of marital or regular sexual relationships [2,22], we disaggregated condom use by domestic partnership status and by relationship status of the last sexual partner . Third, to address possible measurement error, we used a more stringently defined dependent variable, consistent condom use . Fourth, we considered potential confounding by alcohol use at last sexual intercourse . Fifth, to more closely compare our findings to previous research [9–13], we included a variable for personal history of intimate partner violence [26,27]. Sixth, because norms about violence against women have been found to differ between men and women in many sub-Saharan African countries , we constructed an alternative contextual-level variable on the basis of men's responses to the questions about wife beating. Seventh, we considered potential confounding by two contextual-level variables: the percentage of women who were in professional occupations , and the percentage of women in the PSU who had attained at least a secondary education.
The data collection procedures for the DHS were approved by the ICF Macro institutional review board as well as by the relevant ethical review boards in the host country for each survey. All participants provided oral informed consent. This analysis was reviewed by the Harvard School of Public Health Office on Human Research Administration and was considered exempt from full review.
The response rates for the country surveys ranged from 85.6% (Zimbabwe) to 97.9% (Rwanda and Lesotho) (Table S1, http://links.lww.com/QAD/A189). The number of PSUs for the country surveys ranged from 275 to 886 (Table S3, http://links.lww.com/QAD/A189). Of the 241 524 women who completed interviews, our analytic sample included data from 198 806 women (82.3%) who reported being sexually active. Among the women included in the analytic sample, 15 654 (9.4%) reported condom use during last sexual intercourse (Table S4, http://links.lww.com/QAD/A189). Exploratory factor analysis indicated a single dimension to the scale measuring beliefs about violence against women, with a single factor that had factor loadings ranging from 0.61 to 0.76. The scale was internally consistent (Cronbach's α = 0.84), and item-to-scale correlations did not suggest item redundancies. Country-specific internal consistency coefficients ranged from 0.70 (Rwanda) to 0.88 (Nigeria).
In univariable regression models fit to the pooled sample, condom use had a statistically significant association with both individual beliefs and contextual norms about violence against women (Table 1). In the multivariable pooled regression model, however, only the effect of gender-unequal contextual norms remained statistically significant [adjusted odds ratio (AOR) = 0.88; 95% confidence interval, 0.85–0.92; P < 0.001]. This effect was large in magnitude: evaluated at the mean of the other covariates, as PSU-level gender-unequal norms about violence against women increased from 0.79 (25th percentile) to 2.33 (75th percentile), this was associated with a 16.5% relative decrease in the predicted probability of condom use, from 4.2 to 3.5%. A Wald-type F-test rejected the null hypothesis that the country-level fixed effects lacked joint statistical significance (P < 0.001).
In the country-specific analyses, the unadjusted odds ratios for the association between condom use and contextual norms about violence against women were less than one in 20 of 22 countries (Table S5, http://links.lww.com/QAD/A189). Adjustment for individual-level sociodemographic characteristics reduced the statistical significance of these associations, but the AORs were less than one in the majority of countries (Fig. 1). None of the sensitivity analyses substantively altered our primary findings (Table S6, http://links.lww.com/QAD/A189).
Using data from 198 806 sexually active women living in 22 sub-Saharan African countries, we found that women's condom use at last sexual intercourse was negatively associated with the PSU-level of agreement about the appropriateness of wife beating. The association was statistically significant, large in magnitude, and robust to statistical adjustment by known sociodemographic correlates. Further, we found evidence that these contextual effects were greatest among women with no education and among married women in the context of sexual intercourse with nonmarital partners.
Prior research has shown that women's personal experience of intimate partner violence is associated with lack of condom use during sexual intercourse [9–13]. A critical limitation of these studies is that the explanatory variables used in these studies were measured at the individual level, thus limiting their interpretation as contextual effects. Two studies have shown that the contextual prevalence of intimate partner violence is associated with noncondom use and early pregnancy among South African female youth  and unintended pregnancy among women in Colombia . We extend these findings by using data from diverse settings throughout sub-Saharan Africa to directly measure the proximate context of gender-unequal norms about violence against women, and we demonstrate that these contextual norms are negatively associated with women's condom use at last sexual intercourse. Our interpretation of these data is that community-wide norms about the acceptability of wife beating may contribute to a pervasive sense of ‘everyday violence’  which, in turn, undermines women's self-efficacy with regards to negotiating condom use in sexual relationships .
Several limitations should be considered when interpreting our findings. First, we were unable to examine measures of primary behavior change or risk avoidance as outcomes, including delay of sexual debut and/or partner reduction [31,32]. Second, the direction of causality is generically uncertain with data of a cross–sectional nature. Furthermore, the cross–sectional design precludes exploration of richer longitudinal patterns of association between norms and behaviors that have been described in other work . Third, self-reported condom use at last sexual intercourse may have been measured with error . However, random measurement error in the dependent variable would have biased our estimates toward the null. In addition, the estimated effects on consistent condom use (a variable that is likely measured with less error ) in the sensitivity analysis sample were qualitatively similar to the effects on condom use at the last sexual intercourse estimated with the full sample. Fourth, ICF Macro did not provide sampling weights for analyses restricted to sexually active women, so our estimates are not nationally representative. However, it is the largest study of its kind to date, suggesting broad generalization across diverse sociodemographic groups and 22 countries in sub-Saharan Africa. Fifth, although our study sheds light on how gender-unequal contextual norms influence women's HIV risk reduction behavior at the individual level, generalizing these findings to the population level may be premature. The fact that, at the PSU level, women in Swaziland agreed with the fewest scenarios under which wife beating was appropriate yet HIV prevalence rates in Swaziland are among the highest in sub-Saharan Africa  suggests caution in generalizing our findings to the population level .
In summary, this study presents evidence from across sub-Saharan Africa that the proximate context of gender-unequal norms about violence against women is associated with lack of condom use among women. These findings suggest that interventions aimed at modifying the proximate context of gender-unequal norms may enhance women's ability to engage in HIV risk reduction behaviors . However, randomized evaluations of interventions designed to address different aspects of the social construction of gender have yielded mixed results [38,39]. Further research will be needed in order to either narrow the focus of future studies following this thread of inquiry or identify new contextual targets for intervention. Behavioral interventions targeted toward women are likely to be less than optimally effective if their proximate context is not taken into account.
Data from the Demographic and Health Surveys are available for download from the ICF Macro web site (http://www.measuredhs.com). We thank Emily Shafer for helpful comments.
Conflicts of interest
The authors received no specific funding for this study. A.C.T. receives salary support through the Robert Wood Johnson Health and Society Scholars Program. S.V.S. receives salary support through a National Institutes of Health K25 Career Development Award (NHLBI K25 HL081275) and a Robert Wood Johnson Foundation Investigator Award in Health Policy Research. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
1. Schoepf BG. Women, AIDS, and economic crisis in Central Africa. Can J Afr Stud 1988; 22:625–644.
2. Worth D. Sexual decision-making and AIDS: why condom promotion among vulnerable women is likely to fail. Stud Fam Plann 1989; 20:297–307.
3. Ankrah EM. AIDS and the social side of health. Soc Sci Med 1991; 32:967–980.
4. McGrath JW, Schumann DA, Pearson-Marks J, Rwabukwali CB, Mukasa R, Namande B, et al. Cultural determinants of sexual risk behavior for AIDS among Baganda women. Med Anthropol Q 1992; 6:153–161.
5. Weiser SD, Leiter K, Bangsberg DR, Butler LM, Percy-de Korte F, Hlanze Z, et al. Food insufficiency is associated with high-risk sexual behavior among women in Botswana and Swaziland. PLoS Med 2007; 4:1589–1597.discussion 1598.
6. Tsai AC, Leiter K, Wolfe WR, Heisler M, Shannon K, Phaladze NA, et al. Prevalence and correlates of forced sex perpetration and victimization in Botswana and Swaziland. Am J Pub Health 2011; 101:1068–1074.
7. O’Donnell G. La cosecha del miedo. Nexos 1983; 6:51–60.
8. Taussig M. Culture of terror: space of death. Roger Casement's Putumayo Report and the explanation of torture. Comp Stud Soc Hist 1984; 26:467–497.
9. Wingood GM, DiClemente RJ. The effects of an abusive primary partner on the condom use and sexual negotiation practices of African-American women. Am J Public Health 1997; 87:1016–1018.
10. Shannon K, Strathdee SA, Shoveller J, Rusch M, Kerr T, Tyndall MW. Structural and environmental barriers to condom use negotiation with clients among female sex workers: implications for HIV-prevention strategies and policy. Am J Public Health 2009; 99:659–665.
11. Van Horne BS, Wiemann CM, Berenson AB, Horwitz IB, Volk RJ. Multilevel predictors of inconsistent condom use among adolescent mothers. Am J Public Health 2009; 99 (Suppl 2):S417–S424.
12. Dunkle KL, Jewkes RK, Brown HC, Gray GE, McIntryre JA, Harlow SD. Gender-based violence, relationship power, and risk of HIV infection in women attending antenatal clinics in South Africa. Lancet 2004; 363:1415–1421.
13. Jewkes RK, Levin JB, Penn-Kekana LA. Gender inequalities, intimate partner violence and HIV preventive practices: findings of a South African cross-sectional study. Soc Sci Med 2003; 56:125–134.
14. Speizer IS, Pettifor A, Cummings S, Macphail C, Kleinschmidt I, Rees HV. Sexual violence and reproductive health outcomes among South African female youths: a contextual analysis. Am J Public Health 2009; 99 (Suppl 2):S425–S431.
15. Pallitto CC, O’Campo P. Community level effects of gender inequality on intimate partner violence and unintended pregnancy in Colombia: testing the feminist perspective. Soc Sci Med 2005; 60:2205–2216.
16. Froot KA. Consistent covariance matrix estimation with cross-sectional dependence and heteroskedasticity in financial data. J Fin Quant Analysis 1989; 24:333–355.
17. Williams RL. A note on robust variance estimation for cluster-correlated data. Biometrics 2000; 56:645–646.
18. Rogers WH. Regression standard errors in clustered samples. Stata Tech Bull 1993; 13:19–23.
19. Zorn C. Comparing GEE and robust standard errors for conditionally dependent data. Pol Res Quarterly 2006; 59:329–341.
20. Subramanian SV, O’Malley AJ. Modeling neighborhood effects: the futility of comparing mixed and marginal approaches. Epidemiology 2010; 21:475–478.discussion 479–481.
21. Filmer D, Pritchett LH. Estimating wealth effects without expenditure data or tears: an application to educational enrollments in states of India. Demography 2001; 38:115–132.
22. Tavory I, Swidler A. Condom semiotics: meaning and condom use in rural Malawi. Am Sociol Rev 2009; 74:171–189.
23. de Walque D, Kline R. Variations in condom use by type of partner in 13 sub-Saharan African countries. Stud Fam Plann 2011; 42:1–10.
24. Hearst N, Chen S. Condom promotion for AIDS prevention in the developing world: is it working?. Stud Fam Plann 2004; 35:39–47.
25. Weiser SD, Leiter K, Heisler M, McFarland W, Percy-de Korte F, DeMonner SM, et al. A population-based study on alcohol and high-risk sexual behaviors in Botswana. PLoS Med 2006; 3:e392.
26. Kishor S, Johnson K. Profiling domestic violence: a multicountry study. Calverton: ORC Macro; 2004.
27. Straus MA. Measuring intrafamily conflict and violence: the conflict tactics (CT) scales. J Marriage Fam 1979; 41:75–88.
28. Rani M, Bonu S, Diop-Sidibe N. An empirical investigation of attitudes towards wife-beating among men and women in seven sub-Saharan African countries. Afr J Reprod Health 2004; 8:116–136.
29. Shannon K, Kerr T, Allinott S, Chettiar J, Shoveller J, Tyndall MW. Social and structural violence and power relations in mitigating HIV risk of drug-using women in survival sex work. Soc Sci Med 2008; 66:911–921.
30. Casey MK, Timmermann L, Allen MH, Krahn S, Turkiewicz KL. Response and self-efficacy of condom use: a meta-analysis of this important element of AIDS education and prevention. South Comm J 2009; 74:57–78.
31. Stoneburner RL, Low-Beer D. Population-level HIV declines and behavioral risk avoidance in Uganda. Science 2004; 304:714–718.
32. Green EC. Rethinking AIDS prevention: learning from successes in developing countries. Westport: Praeger Publishers; 2003.
33. Huebner DM, Neilands TB, Rebchook GM, Kegeles SM. Sorting through chickens and eggs: a longitudinal examination of the associations between attitudes, norms, and sexual risk behavior. Health Psychol 2011; 30:110–118.
34. Gallo MF, Behets FM, Steiner MJ, Hobbs MM, Hoke TH, Van Damme K, et al. Prostate-specific antigen to ascertain reliability of self-reported coital exposure to semen. Sex Transm Dis 2006; 33:476–479.
35. Gouws E, Stanecki KA, Lyerla R, Ghys PD. The epidemiology of HIV infection among young people aged 15–24 years in southern Africa. AIDS 2008; 22 (Suppl 4):S5–S16.
36. Rose G. Sick individuals and sick populations. Int J Epidemiol 1985; 14:32–38.
37. Dunkle KL, Jewkes R. Effective HIV prevention requires gender-transformative work with men. Sex Transm Infect 2007; 83:173–174.
38. Pronyk PM, Hargreaves JR, Kim JC, Morison LA, Phetla G, Watts C, et al. Effect of a structural intervention for the prevention of intimate-partner violence and HIV in rural South Africa: a cluster randomised trial. Lancet 2006; 368:1973–1983.
39. Jewkes R, Nduna M, Levin J, Jama N, Dunkle K, Puren A, et al. Impact of stepping stones on incidence of HIV and HSV-2 and sexual behaviour in rural South Africa: cluster randomised controlled trial. BMJ 2008; 337:a506.
© 2012 Lippincott Williams & Wilkins, Inc.