Article Text

Download PDFPDF

Original article
Decisions, decisions: the importance of condom use decision making among HIV sero-discordant couples in a prospective cohort study in Uganda
  1. Katherine A Muldoon1,2,
  2. Putu K Duff1,2,
  3. Josephine Birungi3,
  4. Moses H Ngolobe3,
  5. Jeong Eun Min1,
  6. Rachel King4,5,
  7. Maureen Nyonyintono3,
  8. Yalin Chen1,
  9. Kate Shannon1,2,6,
  10. Sarah Khanakwa3,
  11. David Moore1,6
  1. 1British Columbia Centre for Excellence in HIV/AIDS, St Paul's Hospital, Vancouver, British Columbia, Canada
  2. 2School of Population and Public Health, University of British Columbia, Vancouver, British Columbia, Canada
  3. 3The AIDS Support Organisation, Kampala, Uganda
  4. 4University of California, San Francisco, California, USA
  5. 5Karolinska Institute, Stockholm, Sweden
  6. 6Department of Medicine, University of British Columbia, St Paul's Hospital, Vancouver, British Columbia, Canada
  1. Correspondence to K A Muldoon, BC Centre for Excellence in HIV/AIDS, St Paul's Hospital, 608-1081 Burrard Street, Vancouver, British Columbia, Canada V6Z 1Y6; kmuldoon{at}


Objectives We examined sexual decision making type among couples in HIV sero-discordant partnerships in Uganda, and investigated how sexual decision making type changed over time and its effect on condom use.

Methods Data were drawn from a longitudinal cohort of HIV sero-discordant couples, recruited through the AIDS Support Organisation in Jinja, Uganda. Sexual decision making was measured using the Sexual Relationship Power Scale, and couples’ individual self-report answers were matched to assess agreement for decision making type and condom use. Generalised linear mixed effects modelling was used to assess statistically significant differences in time trend of sexual decision making type, and to investigate the independent association of decision making type on condom use status over time.

Results Of the 533 couples included in this analysis, 345 (65%) reported using condoms at last sex at study enrolment. In the time trend analysis of decision making, the proportion of couples who decided together increased over time while the proportions of couples who reported that one partner decided or no one decided/did not use condoms, decreased over time (overall p<0.001). Compared with couples who decided together, those who disagreed (adjusted OR=0.42, 95% CI 0.28 to 0.64) and those where one partner decided (adjusted OR=0.20, 95% CI 0.12 to 0.34) had significantly lower odds of condom use at last sex, even after controlling for confounders.

Conclusions Couples who disagreed on decision making, or agreed that one partner decides alone, had significantly lower odds of reporting condom use compared with couples who decided together. HIV counselling interventions that encourage joint sexual decision making may improve condom use within this population.

  • HIV
  • Condoms
  • Communication Skills
  • Gender
  • Sexual Behaviour
View Full Text

Statistics from


Within sub-Saharan Africa, structural factors known to contribute to the HIV epidemic include chronic poverty, weakened health and civil infrastructure, food insecurity1–3 and gender norms of male dominance.4 These structures can create social environments that promote gender inequities, which can then manifest within intimate relationship as sexual coercion,5 ,6 reduced sexual negotiating power,7 transactional sex8 ,9 or inter-generational partnering,10 all risk factors for HIV infection. Although Uganda has made significant progress towards reducing the national burden of HIV, it continues to face national drug shortages, insufficient staffing and an overburdened healthcare system. It is estimated that only 46% of 540 000 people eligible to receive antiretroviral therapy (ART) are currently accessing ART.11 Within Uganda, the HIV prevalence has stabilised at 7.3%12 but it is estimated that approximately half of people living with HIV/AIDS are cohabitating with a sexual partner who is HIV negative.13 HIV negative partners are at very high risk for acquiring HIV infection from their HIV positive spouses, particularly in the absence of viral suppression through regular access and adherence to ART. This heightened risk for sero-discordant couples has led to a need for a deeper understanding of decision making for condom use within this context.

There has been a proliferation of research on sero-discordant couples within the context of HIV prevention, particularly the role of HIV counselling and testing on condom use14–16 but few studies have investigated the dynamics surrounding couple level decision making on condom use.

Power imbalances in intimate relationships have been shown to be associated with a higher prevalence and incidence of HIV infection in women.17 Several studies have shown that women who report low control in their relationships are more likely to report inconsistent condom use.18–20 Harrison et al4 conducted a study investigating sexual risk among both young men and women in South Africa. Among male participants, a factor positively associated with condom use included male dominance. Among female participants, the relationship feature most strongly associated with condom use was more egalitarian style relationships. Effectively, these stratified analyses demonstrate that gender acts as an effect modifier between relationship power and condom use. Both members of the same couple were not interviewed in this study but this analysis highlights gender differences in perception about relationship power dynamics and how they influence condom use.

Assessing couple agreement on self-report measures has been used as a statistical technique to reduce measurement error, and has been used in studies measuring inter-partner violence,21 perception of partnership type (casual, steady, etc),22 sexual history and condom use,23 ,24 and fertility intentions.25 ,26 A sero-discordant couples cohort presents an opportunity to understand sexual decision making within couples while simultaneously assessing reported agreement on HIV risk behaviours within a group at high risk for HIV transmission.

The objectives of this analysis were to explore sexual decision making among couples in sero-discordant partnerships and how they may change over time with regular access to HIV counselling and testing, and investigate how decision making influenced condom use over time.



Data were drawn from a prospective cohort of heterosexual HIV sero-discordant couples, known as the Highly Active Antiretroviral Treatment as Prevention (HAARP) study.27 Couples were recruited through The AIDS Support Organisation (TASO) in Jinja town, eastern Uganda, between June 2009 and June 2011, with follow-up until December 2011. All seropositive participants were receiving HIV treatment and care through the TASO-Jinja site. Couples were eligible to participate in the HAARP study if both were ≥18 years, in a stable intimate partnership of ≥6 months, lived together and were sexually active (defined as more than two episodes of sexual intercourse in the past 3 months). A detailed explanation of the study has been published elsewhere.28

Every 6 months, couples completed a detailed questionnaire collecting behavioural, medical and laboratory information. Every 3 months, the HAARP study provided HIV discordant couples counselling, including condom provision, and HIV counselling and testing for the seronegative partner.

All study activities and services were available in three of the region's most common languages: Lusoga, Luganda and Kiswahili. Each member of the couple was interviewed separately prior to counselling to reduce social desirability bias. The HAARP study received scientific and ethics approval in Uganda from the Uganda Virus Research Institute and the protocol was registered at the Ugandan National Council for Science and Technology. The study was also approved in Canada by the Research Ethics Board of the University of British Columbia.


Condom use at last sex (yes vs no) was chosen as the primary outcome of interest because it is the most reliably populated measure of condom use, both at baseline and follow-up. Couples where both partners agreed and reported using a condom were coded as ‘yes’. Couples where both partners agreed and reported that they did not use condoms were coded as ‘no’. There were some couples who disagreed about condom use at last sex but to avoid excluding this sample, they were conservatively coded as ‘no’.

Condom use decision making was the primary explanatory variable of interest and was derived from couples’ answers to the question “who decides whether or not to use a condom”, an item on the Sexual Relationship Power Scale (SRPS),29 with five possible answers: male decides, female decides, they decide together, no one decides or that they do not use condoms. The couples’ answers to this question were compared to assess agreement. The final decision making variable included six categories, with couples agreeing to each of the five possible answers to the decision making variable, and a sixth category including couples that disagreed.

Confounding variables were included in the analysis based on hypothesised and a priori knowledge of their influence on the association between a couple's decision making and condom use based on the literature and included: the male partner’s age, couples’ fertility intentions, male HIV status, couple ART status (defined as one of the partners being on ART), polygynous couple status (one male with more than one female spouse) and number of follow-up visits to control for time.

Statistical analysis

Descriptive and bivariable analyses

The analytic sample included couples with complete answers to the primary outcome of interest (condom use) and explanatory variable (decision making type) at baseline. The sample of sero-discordant couples was characterised using frequencies and proportions for categorical data and medians and interquartile ranges (IQRs) for continuous variables. We compared differences between couples by condom use using the χ2 test or Fisher's exact test for categorical variables, and Wilcoxon rank sum tests for continuous variables. Analyses were conducted using SAS V.9.3 (SAS Institute Inc, Cary, North Carolina, USA).

Longitudinal analyses

Time trend analysis

We used a mixed effects multinomial logistic regression model with random intercept to assess temporal changes in decision making type over 18 months of follow-up, with the hypothesis that the proportion of couples deciding together to use a condom would increase over time in accordance with couples counselling offered at each session. A variable for time has the value of 0 (enrolment) to 1 (6 months), 2 (12 months) or 3 (18 months), and it was entered into a linear observation level term in the model. When we ran this analysis using the decision making variable with six categories, the model did not converge. We collapsed couples who reported that the male or female partner decided into one category of couples who agree that one partner decides. Few couples reported that no one decided or that they do not use condoms, so these categories were also collapsed.

Confounder modelling

To estimate the relationship between decision making type and condom use over time, the sample was restricted to couples with ≥12 months of follow-up. We then used a mixed effects logistic regression model (random intercept) with unadjusted ORs and adjusted ORs and 95% CIs to estimate the relationship between decision making type and condom use at last sex over time. Due to a lack of heterogeneity in the outcome after combining couples that reported that no one decides or they do not use condoms (all responded ‘no’ to condom use at last sex), this group was removed from the final analysis.


A total of 586 sero-discordant couples were recruited into the HAARP study; 533 had complete responses and were included in this analysis. Table 1 displays the descriptive statistics of 533 couples at baseline and the bivariable analysis by condom use status. At baseline, 345 couples (65%) reported using a condom the last time they had sex. The median age of males was 42 years (IQR 36–48) and females 35 years (IQR 30–40). One hundred and twenty-seven couples (24%) were in polygynous relationships. Lusoga was the predominant language (63%) followed by Luganda (19%). In 305 (57%) couples, the male partner was living with HIV. In 239 couples (45%), the HIV positive partner was receiving ART. Couples who reported using a condom at last sex were: significantly older; more likely to decide together to use condoms; more likely to have a male HIV positive partner in the sero-discordant partnership; and less likely to have fertility intentions (plan to have more children).

Table 1

Baseline sociodemographic and other characteristics for 533 enrolled HIV sero-discordant couples (TASO Jinja site), Uganda

At baseline, 345 couples (64%) had used a condom at last sex, 188 couples (36%) had not and 76 couples (22%) disagreed on condom use at last sex and were coded as not using a condom. In total, 159 couples (30%) decided together on condom use and 214 couples (40%) disagreed on decision making type. In 71 couples (13%) the male decided, in 31 couples (6%) the female decided, making for a combined total of 102 couples (19%) reporting that one partner decided. Only six couples (1%) reported that no one decided and 52 couples (10%) reported that they do not use condoms (when combined making 58 couples (11%)). From the six decision making types, reported condom use at last sex was highest among those who decided together (87%), followed by those that disagreed on decision making type (70%) and those where the male decided (58%).

Couples who used a condom at last sex did not differ significantly on language or ART status, but did differ significantly by male age (p=0.008), female age (p=0.032), sex of the index partner (p=0.008), fertility intentions (p<0.001) and decision making type (p<0.001) at baseline.

A mixed effects multinomial logistic model was then run to assess statistical significance in time trends in reported sexual decision making type at baseline, 6 months, 12 months and 18 months (table 2).

Table 2

Longitudinal mixed effects analysis to assess statistical significance in time trend in reported sexual decision making type at each time point

When compared with couples who decided together on condom use, the linear time effect was significantly lower than couples reporting that one partner decided (OR=0.64, 95% CI 0.54 to 0.76) or no one/do not use condoms (OR=0.34, 95% CI 0.23 to 0.48), indicating a decreasing trend over time. There was no statistically significant trend over time for reporting disagreeing responses versus reporting decided together.

To determine the association between decision making type and condom use over time, a confounder model was run using a mixed effects logistic regression analysis. Table 3 shows that, compared with couples who decided together, those who disagreed (adjusted OR=0.42, 95% CI 0.28 to 0.64) and those where one partner decided (adjusted OR=0.20, 95% CI 0.12 to 0.34) had significantly lower odds of condom use, even after controlling for known confounders. Figure 1 shows that the predicted probability of condom use increased over time for all decision making type groups, with the ‘together’ group having the highest probability of condom use throughout the study.

Table 3

Unadjusted and adjusted OR for condom use at last sex from a mixed effect model, among 475 HIV sero-discordant couples (1308 observations)*

Figure 1

Average predicted probability of condom use at each visit from the mixed effect model.


Compared with couples that decided on condom use together, couples where one partner decided or where the couples disagreed on decision making type had significantly lower odds of using condoms over time. Our analysis used statistical techniques that combined the responses of both partners to create a couple level variable that limited measurement bias and improved construct validity of measuring decision making dynamics and condom use over time. Our analysis supports the role of couple level sexual decision making and how this can improve condom use and reduce risks associated for HIV infection.

At enrolment, 65% of couples reported using condoms at last sex, 30% of couples in our study reported deciding on condom use together, 13% reported that the male decided and 6% reported that the female decided. Given prevailing male dominant gender roles in Uganda, this finding was unexpected. It was hypothesised that couples where the male decided would be the largest group. It is possible that the high level of joint decision making, and female deciders in this sample, is not representative of the general population. All seropositive participants in this study were clients at TASO and it is possible that some were exposed to counselling before the study. However, many of the couples only discovered their discordant status at study enrolment and were therefore unlikely to have received discordant couples counselling on a quarterly basis offered during the HAARP study.

This analysis showed changes in condom use decision making over time, with fewer couples who disagreed or reported that only one partner makes decisions regarding condom use with each visit where couples counselling was offered. However, we noted that the proportion of couples that disagreed on decision making type remained relatively high throughout the study period. This may indicate a limitation of couples counselling and a limitation of self-reported data on decision making type since members of a couple may (and likely do) perceive who decides to be different. However, there is much less disagreement on actual reported condom use at last sex. This differs from a comparable study of Rwandan couples documenting that men were more likely to report condom use compared with their female partners.26 Hageman et al also found substantial disagreement when reporting fertility desires, sexual risk behaviours and perceptions of risk. Reporting differences between couples highlight the utility of using matched couple level data to reduce measurement error when examining condom use patterns.

The existing literature on HIV risk and sexual decision making has largely focused on women's reports of decision making dominance in their relationships. Our results supports previous research among pregnant women in Haiti and South Africa that linked higher levels of sexual communication with consistent condom use using the SRPS.30 ,31 Another study from South Africa observed a positive association between sexual power and consistent condom use, although these studies did not explicitly focus on the decision making dominance subscale of the SRPS, where our decision making variable was drawn.20 Campbell et al32 argued that the decision making dominance subscale of the SRPS may be a more accurate predictor of condom use, as it captures women's relative power to her partner.


This analysis has several limitations. The main explanatory variable, condom use decision making, relied on one item from the SRPS because the entire scale was not administered in the study. All couples in this study were recruited through TASO, where some may have had previous exposure to counselling, possibly introducing sampling bias that would bias the results towards the null. Additionally, it is important to recognise that 65% of couples in this study did report using a condom at last sex and the predicted probability of using a condom did increase for all groups over time. Despite levels of condom use far higher than the general population, reported condom use continued to increase with follow-up, suggesting that even if this bias was present, the effect of the couples counselling offered during the study was still detectable. While this analysis is restricted to one male partner and his primary discordant female partner, men in polygynous partnerships were included in this sample. It is possible that the male spouse could have been referring to condom use and decision making with a different spouse but this is unlikely because of study protocols in place to ensure that data were collected on the primary discordant spouse before other spouses in partnership. Previous analyses have demonstrated that polygynous males exhibit a higher degree of dominance in sexual decision making,28 and therefore we felt they were important to include. Finally, although this study aimed to improve response validity by using couples’ responses, self-reported condom use may be subject to social desirability bias.


This analysis has shown that joint couple level sexual decision making consistently predicts higher levels of condom use over time among sero-discordant couples in Uganda. Couples who disagree on decision making, or agreed that one partner decides alone, reported significantly lower odds of condom use compared with couples who decide together. These results show that while many couples disagree on who decides to use condoms, over a third of couples decide together and show higher and consistent rates of condom use over time. These results suggest that regular couples counselling that improve collaborative decision making should be included in HIV treatment programmes in Africa as part of improved efforts to identify HIV negative sexual partners of positive participants and reduce their risk of acquiring HIV.

Key messages

  • Couples counselling and HIV testing are potential mechanisms to improve gender equity and improve condom use through shared sexual decision making.

  • Couples who make sexual decisions together have the highest probability of condom use over time.

  • HIV prevention programming should focus on couple dynamics in addition to actual condom use.


The authors wish to thank the participants from the HAARP Study, the staff at TASO and Hong Wang for her statistical support.


View Abstract


  • Handling editor Jackie A Cassell

  • Contributors KAM and PKD developed the analysis plan and wrote the first draft. JEM, MN and YC were responsible for data analysis and management. JB, MHN, SK and RK provided critical feedback and supported the final manuscript. All authors made significant contributions to the conception and design of the analyses, interpretation of the data and drafting of the manuscript. DM had primary responsibility for the final content.

  • Funding KAM is funded through a doctoral fellowship from the Canadian Association for HIV/AIDS Research. KS is supported by US National Institutes of Health, Canadian Institute for Health Research (CIHR) and Michael Smith Foundation for Health Research (MSFHR). DM is funded through a CIHR New Investigator Award and a Scholar Award from the MSFHR. The HAARP Study was supported by CIHR, grant No MOP-89730.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval The study received scientific and ethical approval in Uganda from the Uganda Virus Research Institute, and the protocol was registered at the Ugandan National Council for Science and Technology. The study was also approved in Canada by the Research Ethics Board of the University of British Columbia.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.