Positive deviance for promoting dual-method contraceptive use among women in Uganda: a cluster randomized controlled trial

Background Dual-method contraceptive use, or using condoms with highly effective contraceptives, is effective at preventing both unintended pregnancies and HIV infections. Although it remains uncommon among women in long-term relationships in sub-Saharan Africa, some do practice it. The positive deviance approach aims to promulgate practices of such individuals to other members in the community. We examined the effects of a positive deviance intervention on the dual-method use among married or in-union women. Methods We conducted a cluster randomized controlled trial in 20 health facilities in Mbarara District, Uganda, with 960 women aged 18-49 years. The intervention was a combination of clinic- and phone-based counseling and a participatory workshop, which was developed based on the qualitative study of women practicing the dual-method use in the study area. The control group received regular clinic-based counseling and health-related messages via phone. We assessed dual-method contraceptive use at the last sexual intercourse and its consistent use at two, four, six, and eight months after enrollment. Findings More women in the intervention group reported dual-method contraceptive use at the last sexual intercourse at two months (AOR = 4.29; 95% CI: 2.12-8.69; p < 0.001) and at eight months (AOR = 2.19; 95% CI: 1.07-4.48; p = 0.032) than in the control group. Consistent dual-method contraceptive use was also more prevalent in the intervention group at two months (AOR = 13.71; 95% CI: 3.59-52.43; p < 0.001), and the intervention effect remained at four, six, and eight months. Conclusion Dual-method contraceptive use increased significantly among women in the intervention group. The positive deviance intervention can be a potential option for promoting the dual-method use among women in long-term relationships in Uganda.

for other reasons, and (iii) had been using condoms consistently with an HEC in the last two 1 3 4 months before the recruitment. The sample size of 960 was calculated based on the effect size 1 3 5 of 2.43 reported in a dual-method intervention trial in the USA, considering an intraclass 1 3 6 correlation coefficient of 0.006 and a 26% dropout rate.
The power of the study was set 1 3 7 at 80%, and the significance level was set at 5%. 1 3 8 Female research assistants recruited women who visited the family planning sections of the 1 3 9 selected health facilities. They approached every third woman after selecting the first woman 1 4 0 purposively to inform the opportunity to participate in the study. If a woman was interested, 1 4 1 they confirmed HEC use with her family planning client record card and asked questions to 1 4 2 verify eligibility. The process was repeated until the required sample size was reached. All 1 4 3 women received an information sheet for their partners, which included the objectives of the 1 4 4 study and contact information of the researchers for notifications in the case of conflict with 1 4 5 their partners. After two weeks of enrollment, women were invited for a one-day participatory learning 1 8 2 workshop at the same health facility where they were recruited. Participation in the workshop 1 8 3 was voluntary. The four PDs facilitated the workshop with support from one research 1 8 4 assistant. It included role-play exercises to enable women to acquire successful 1 8 5 communication skills of the PDs, practice of male condom use, and group discussions on the 1 8 6 dual risk of unintended pregnancies and HIV/STI infection from their partners. 1 8 7 In addition, women in the intervention group received a bi-monthly telephone counseling call 1 8 8 from the PDs three times (i.e., three, five, and seven months after enrollment). It aimed to 1 8 9 confirm women's dual-method status and challenges, provide reminders regarding the risk of 1 9 0 unintended pregnancies and HIV/STIs, and strengthen their capacity to communicate with 1 9 1 their partners. In addition, the call included brief health education messages on family 1 9 2 planning and HIV/STI based on an existing tool.

2
Each PD provided the same women with 1 9 3 counseling during each call to build rapport and ensure effective counseling. Each counseling 1 9 4 lasted for 15 to 30 minutes. The PDs maintained a counseling record and held a group 1 9 5 meeting after each counseling period to reflect on the problems of women and advice given.
The research assistants facilitated such meetings and answered questions from the PDs. 1 9 7 Women in the control group received family planning counseling, including condom use, 1 9 8 from female research assistants for 10 to 20 minutes using the existing tool on the day of 1 9 9 enrollment. 2 2 However, this group of women did not receive the handout. Furthermore, the 2 0 0 research assistants provided bimonthly health education three times (i.e., three, five, and 2 0 1 seven months after enrollment) by phone. The topics were the same as those for the 2 0 2 intervention group. Each call lasted for approximately ten minutes. 2 0 3 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 1 1 At the selected health facilities, condoms were provided for free regardless of group 2 0 4 allocation. Before providing the intervention, the research assistants received a two-day 2 0 5 training on the contents of the existing counseling tool. In addition, the four PDs received a 2 0 6 one-day training on counseling and ethics, including the confidentiality of their clients. The 2 0 7 PDs joined the intervention as volunteers but received 30,000 UGX (equivalent to 9 USD) 2 0 8 per day when they engaged in the workshop and counseling to compensate for their time and 2 0 9 transportation. 2 1 0 <Insert Table 1 here>  2  1 1 The primary outcome was dual-method contraceptive use, which was defined as the 2 1 3 application of a male or female condom along with an HEC, such as injectables, implants, 2 1 4 intrauterine devices, pills, and female sterilization. 4 It was measured in two timeframes: dual-2 1 5 method contraceptive use at the last sexual intercourse and its consistent use in the last two 2 1 6 months before each follow-up. The former is easier for women to answer accurately than the 2 1 7 latter, which requires to estimate the frequency of condom use in the past.

3
Two questions regarding HEC use and the frequency of condom use were 2 2 0 combined to measure consistent dual-method contraceptive use. The following question was 2 2 1 posed for HEC use: "Apart from condoms, have you been using any other forms of protection 2 2 2 against pregnancy during the past two months?" The frequency of condom use was asked 2 2 3 with an item: "How often did you and your partner use a male or female condom during the 2 2 4 past two months?" Women answered this question using a four-point scale "every time," 2 2 5 "almost every time," "sometimes," and "never." Women using an HEC and a condom every 2 2 6 time were considered practicing consistent dual-method contraceptive use. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 6, 2020. . https://doi.org/10.1101/2020.10.03.20206169 doi: medRxiv preprint The secondary outcome was communication about HIV/STI risk with partners in the last two 2 2 8 months prior to each follow-up. This outcome was assessed using the following item: "Have 2 2 9 you ever discussed HIV/STI risk with your husband/live-in sexual partner in the past two 2 3 0 months?" Another secondary outcome was the self-reported incidence of pregnancy in the 2 3 1 two months before each follow-up regardless of whether the pregnancy was intended or not.

3 2
This outcome was assessed using the following questions: "Have you been told by a 2 3 3 healthcare provider that you got pregnant for the first time in the past two months?" 2 3 4 In addition, the following information was collected at baseline: age, education, religion, condom use self-efficacy, and sexual relationship control power (the Sexual Relationship Power Scale).

6
Several 2 4 0 changes were made to the outcomes after the trial commenced. An outcome for STI incidence 2 4 1 was omitted because we found that the reliability of self-reported STI incidence could be low All research assistants received a two-day training on data collection and ethics before the 2 4 7 baseline data collection. After enrollment, the research assistants interviewed women to 2 4 8 identify their baseline characteristics using a pre-tested structured questionnaire. Each 2 4 9 interview lasted approximately 30 to 45 minutes. 2 5 0 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020.  Table 2 presents the sociodemographic characteristics of 960 women at baseline. The mean 2 9 7 age was 30.1 (SD: 6.7) years. The mean number of children was three (SD: 1.8). Of 960 2 9 8 women, more than 70% completed primary education. In addition, 9% were HIV-positive, 2 9 9 7.6% had an HIV-positive partner, and 84.5% perceived a certain level of risk for HIV/STIs. 3 0 0 Injectables were the most common HEC (51.9%), followed by implants (31.6%).

0 1
Characteristics were similar for the intervention and control groups with a few slight 3 0 2 imbalances. Specifically, women in the control group were more likely to have primary or 3 0 3 Effect of the intervention 3 1 0 Table 3 presents the data of outcomes by group and time. More women in the intervention 3 1 1 than in the control group used dual-method contraception at the last sexual intercourse and 3 1 2 consistently at each follow-up point. These differences were largest at two months (dual-3 1 3 method contraceptive use at last sexual intercourse: 42.6% vs. 13.8%; p < 0.001; consistent 3 1 4 dual-method contraceptive use: 15.5% vs. 1.5%; p < 0.001). The proportion of women 3 1 5 practicing the dual-method decreased for both groups over time. More women discussed 3 1 6 HIV/STI risk with their partners in the intervention than in the control group at each follow-3 1 7 up. The difference was also largest at the first follow-up (83.5% vs. 64.9%; p < 0.001).

1 8
However, the incidence of pregnancy was not significantly different between the groups. 3 1 9 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. . https://doi.org/10.1101/2020.10.03.20206169 doi: medRxiv preprint Throughout the data collection period, 6 and 15 women became pregnant in the intervention 3 2 0 and control groups, respectively. Notably, the result of the chi-squared test of the 3 2 1 accumulated cases of pregnancies in eight months illustrated a significantly lower incidence 3 2 2 of pregnancy in the intervention group (p = 0.047). 3 2 3 <Insert Table 3 here>  3 2 4 Table 4 illustrates the effects of the intervention on the primary and secondary outcomes at 3 2 5 two, four, six, and eight months after enrollment. Model 2 indicates that more women in the 3 2 6 intervention reported dual-method contraceptive use at the last sexual intercourse at two 3 2 7 months (AOR = 4.29; 95% CI: 2.12-8.69; p < 0.001) than in the control group. The 3 2 8 intervention group also reported more dual-method contraceptive use at the last sexual 3 2 9 intercourse at four, six, and eight months, although the difference was statistically significant  Supplementary Tables S3-S17.  3  3  9 <Insert Table 4  based intervention for promoting the dual-method use. 1 1 However, reaching out to male 3 8 2 partners may be more difficult compared to providing education to women visiting family 3 8 3 planning clinics. In the current intervention, women received the handout used in the initial 3 8 4 dual-method contraceptive counseling and were encouraged to discuss condom use with their 3 8 5 partners. A qualitative study on married couples in Uganda found that women were more 3 8 6 likely to initiate discussion and persuade their male partners to use condoms. 2 7 In addition, the 3 8 7 findings of our preliminary study reported that women could talk more comfortably about 3 8 8 sensitive topics, such as condom use, by sharing information they received as a conversation 3 8 9 starter.

6
The majority of women in this study were willing to share the health messages and 3 9 0 discuss HIV/STI risk with their partners. Given that women using HECs visit health facilities 3 9 1 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. . https://doi.org/10.1101/2020.10.03.20206169 doi: medRxiv preprint presumably more frequently than men, educating them on the dual-method use and 3 9 2 encouraging them to share such messages with their partners are effective strategies.

9 3
Despite the increase in dual-method contraceptive use, no significant difference was observed 3 9 4 in pregnancy occurrence between the intervention and control groups at each follow-up point. 3 9 5 In this study, many women started the dual-method use but practiced it inconsistently.

9 6
Inconsistent dual-method contraceptive use may explain the lack of effect on avoiding 3 9 7 pregnancies.

8
It might also be explained by a lack of statistical power. Only 21 women 3 9 8 (about 2% of the participants) became pregnant during the eight-month follow-up. The low 3 9 9 incidence of pregnancy is reasonable because we recruited women using an HEC and who 4 0 0 wanted to avoid pregnancy at baseline. However, the intervention group showed the lower 4 0 1 incidence of pregnancy over time. Thus, a further trial with a larger sample size is 4 0 2 recommended to examine the effect of the intervention on the incidence of pregnancy. in the dual-method use, and adherence to such practice was frequently low.

1
Condom use 4 0 6 is often considered awkward or unacceptable in long-term relationships, especially when 4 0 7 women are using HECs. Moreover, condom use is typically considered a male 4 0 8 responsibility.

9
The positive deviance intervention can be effective in changing such norms. 4 0 9 During counseling and workshop, the PDs shared their experiences to help participants 4 1 0 realize that condom use is normal even among married or in-union women using HECs. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. . https://doi.org/10.1101/2020.10.03.20206169 doi: medRxiv preprint 0 skills necessary for playing a proactive role in negotiation and condom use with their 4 1 5 partners.

1 6
The study has several limitations. First, the study measured outcomes based on self-reports 4 1 7 from the participants. Therefore, it is subject to measurement errors. Especially, dual-method 4 1 8 contraceptive use could have been over-reported given the information provided to conducting interviews by experienced female research assistants. Second, we collected data 4 2 6 on pregnancy incidence during follow-up, but the rate was too low to use as a proxy for the 4 2 7 dual-method use. Other clinical meaningful data, such as the incidence of STI, should be 4 2 8 collected to evaluate interventions for the dual-method use in future research. Third, several 4 2 9 characteristics of the participants were imbalanced between the intervention and control  Therefore, they must participate in the process to identify 4 3 7 their own solutions. Those solutions will be different from the ones identified in this study. 4 3 8 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. . https://doi.org/10.1101/2020.10.03.20206169 doi: medRxiv preprint 1 Further research is recommended to assess the effectiveness of the positive deviance 4 3 9 approach in a given context with careful attention to its process. can be a potential option for promoting the dual-method use by empowering women to play a 4 4 7 proactive role in negotiation and condom use with their partners despite such social norms. revised the manuscript. All authors approved the final version for submission. 4 6 0 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. . https://doi.org/10.1101/2020.10.03.20206169 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. . https://doi.org/10.1101/2020.10.03.20206169 doi: medRxiv preprint  HIV: human immunodeficiency virus; STI: sexually transmitted infection Note: The effect estimates are reported using odds ratio (OR) and adjusted odds ratio (AOR) from multiple logistic regression using the control group as the reference category. *** p < 0.001. ** p < 0.01. * p < 0.05. a Adjusted for the cluster effect and individuals. b Adjusted for cluster effect, individuals, age, education, religion, wealth index, number of children, pregnancy intention, partner's pregnancy intention, history of unintended pregnancy, multiple sex partnership, HEC methods, HIV status, partner's HIV status, HIV/STI risk perception, HIV-related knowledge, condom use self-efficacy, and sexual relationship control power.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted October 6, 2020. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 6, 2020. . https://doi.org/10.1101/2020.10.03.20206169 doi: medRxiv preprint