Uganda is the third-largest refugee-hosting country in the world, with an estimated refugee population of 1.7 million as of May 2024 (28). The majority of refugees in Uganda are from South Sudan (55%) and the Democratic Republic of Congo (31%), with smaller numbers from Somalia, Burundi, Rwanda, and other countries. Furthermore, the majority of this population are women and adolescent girls in need of SRH services. Uganda's progressive legal framework grants refugees' access to social services, including health and education (29). Health services, including SRH, are delivered through public and private facilities, supported by the United Nations High Commissioner for Refugees (UNHCR) and its implementing partners.

The study was conducted in two refugee settlements in Uganda: Kiryandongo and Kyangwali. As of July 2024, Kiryandongo hosted approximately 119,129 refugees, mostly from South Sudan (98%), while Kyangwali had about 139,935 refugees, primarily from the DRC (96%) (28). These settlements were selected in consultation with UNHCR-Uganda and the Department of Refugees in the Office of the Prime Minister (OPM), which represent major refugee groups in Uganda.

The analysis draws on baseline data from a one-year prospective study involving a cohort of randomly selected women and girls aged 15–45 years living in Kiryadongo and Kyangwali refugee settlements. The overall study aims to generate evidence on innovative solutions for addressing unintended pregnancy in refugee settings. The upper age limit of 45 years was chosen to ensure follow-up interviews with the women while within reproductive age.

Sampling was in two stages. First, two zones were purposively selected in each of the two settlements based on predefined criteria relevant to the study objectives—specifically, their differing program implementation contexts, population size (about 3,000 to 5,000 households) and accessibility. Household listing was then conducted in the selected zones to generate the sampling frame from which eligible women and girls were randomly selected. A sample size of 3,019 women aged 15–45 was targeted to detect differences of 20%–50% in two proportions at a 95% confidence level and 80% power, based on a formula developed by Fleiss and colleagues (30). Sample size calculations assumed that both exposure/predictor and outcome variables are dichotomous, for example, assuming 20% in the unexposed group and 40% in the exposed group having a positive outcome, such as the current use of pills or injectables to prevent unintended pregnancy. Interviews were completed with 2,498 women (78% of those sampled).

Data collection was conducted from March to May 2024. Face-to-face computer-assisted personal interviews (CAPI) were conducted by trained female interviewers with sampled eligible participants using a structured questionnaire. Pre-testing of the tool was done to identify and address any issues prior to the actual data collection. Interviews were conducted in English or five other local languages (Acholi, Juba Arabic, Kigegere, Kinyabwisha and Kiswahili) depending on a participant's preference. All data was electronically recorded on the ODK platform.

The questionnaire was adopted from a previous study that examined reasons for unmet need for family planning, with attention to the measurement of fertility preferences in low and middle income countries (31). Information captured includes women's sociodemographic characteristics, reproductive history, contraceptive use, and fertility intentions. Participants were asked if they planned to use contraception within the next 12 months or at any time in the future. Women who responded positively were then asked which method they intended to use. The outcome variable for this study was whether women who were not using a method at the time of the interview intended to use a contraceptive method (implants, injectables or pills) in the future.

All women who had heard of a specific method (i.e., implants, injectables, pills, emergency pills and male condoms) were asked about their beliefs about that method with respect to convenience, effectiveness, and method-related health effects, regardless of whether they were using it or not. In terms of convenience, women were asked to state their opinions on whether the method was easy to obtain and easy to use. Perceived effectiveness was determined by asking whether they considered the method “very effective at preventing pregnancy”. Five items regarding method-related health concerns and safety were assessed: women were asked whether they believed the method was likely to cause (a) serious health problems, (b) unpleasant side effects, (c) disruption to regular menses, (d) long-term infertility, or (e) dangers from long-term continuous use. It is worth noting that beliefs (b) and (c) may be experiential when one is using contraception, while beliefs (a), (d), and (e) are perceptions driven by an expectation of a bodily change which is unsupported by current clinical evidence (32). In addition, the questionnaire asked women opinions about their partner's approval of the method while those who reported past used were asked about their level of satisfaction with the method, which was categorized into: past use and satisfied, or otherwise. Additionally, women were asked if their social network (including friends, relatives, and neighbours) had used the three methods and whether their experience had been satisfactory.

Descriptive statistics was used to examine the characteristics of all interviewed women and crosstabulation with a Chi-square test to assess method-specific beliefs about three popular methods (pills, injectables, and implants) among non-users aware of these three methods. The analysis also assessed which perceived method-specific attributes predict future intentions to use either injectables, implants, or pills among women who were not currently using a method but intended to do so in the next 12 months or at any time in the future and were aware of all three methods.

Alternative specific conditional logit (McFadden's choice model) was used to assess the association between method-specific beliefs and intention to use either implants, injectables, or pills in the future. This model, commonly used in econometric analysis of choice (33), allows the inclusion of two types of independent variables in one regression equation: alternative-specific and case-specific (34). The alternative-specific independent variables (method-specific beliefs) vary among alternatives (methods) as well as cases (respondents). On the other hand, the case-specific independent variables (respondents background characteristics) vary only among cases (respondents). The model relies on the key assumption of Independence of Irrelevant Alternatives (IIA), that is, the relative probability of choosing one alternative over another remains constant in the presence or absence of other alternatives (35, 36). However, this assumption may not hold in some contexts, leading to biased estimates (37). In this paper, each method attribute or belief is represented by a single coefficient indicating its association with future method choice. For example, one might consider whether, after accounting for other factors, women are less likely to choose a method perceived to have specific health effects, such as interfering with menstruation. For respondents' characteristics, arrays of coefficients are provided for effects on two pairwise method choices: implant vs. injectable and pill vs. injectable. A similar approach has been used in published studies (13, 14). A p-value < .05was used to indicate the statistical significance with 95% confidence intervals. Estimation was via maximum likelihood using the asclogit command in Stata version 16.

The Population Council Institutional Review Board (Protocol 1027) and the Mildmay Uganda Research Ethics Committee (MUREC) (REF 0109–2023) granted ethical approval for the study. The research was also granted regulatory approval by the Uganda National Council for Science and Technology (REF SS2085ES). All participants provided consent, electronically recorded on the ODK platform.

Out of 2,264 women who completed the interview, 45% and 55% were from Kiryandongo and Kyangwali refugee settlements, respectively (Table 1). Nearly half of the women were aged 25–34 years (44%) and had no formal education (46%). About 60% were married or living with a partner, 28% did not want to have (another) child and 11% were pregnant at the time of the interview. Only 21% were currently using any method. More than half (55%) of the respondents were originally from the Democratic Republic of Congo (DRC), while 44% were from South Sudan, and the rest were from other countries. A significantly higher proportion of current users than non-users were aged 25–34 years, had primary or higher level of education, and wanted a child but preferred to wait 2–5 years.

Among contraceptive nonusers, 96% were neither infertile nor postmenopausal (Table 2). Of these, 32% intended to use a contraceptive method within the next 12 months or later. Among those intending to use contraception, 39% preferred injectables, 25% implants, 17% pills, 9% other modern methods, 6.6% traditional methods, and 4% were undecided.

Table 3 summarizes the beliefs about pills, injectables, and implants among non-users who are fecund, have heard of all three methods, and intend to use one of these methods in the next 12 months or later (n = 263). The proportion of women who believed that injectables are easy to obtain, effective, and easy to use was significantly higher than for implants and pills (range 76%–89%).

There were significant differences in perceived health-related concerns by type of method. Concerns about unspecified serious health problems were highest for implants (29%) and injectables (18%), compared to pills (10%). Similarly, concerns about interference with menstruation were highest for implants and injectables (both at 58%) compared to pills (38%). Concerns about unpleasant side effects were common across all three methods (range 62%–76%). The majority of women (over 90%) believed that it was unsafe to use implants, injectables, or pills for long-term without taking a break. More than a quarter of the women (26%) believed that implants cause infertility, compared to 17% for injectables and 22% for pills.

The proportion of women reporting knowing someone in their social network who had used the method was significantly higher for injectables (72%) than implants (68%) or pills (54%). Perceived satisfaction among social networks was high across the three methods (range 85%–92%). About 15%–17% of past users of pills and injectables, compared to only 7% of implants, reported satisfaction with the method.

Table 4 presents the results from the conditional logistic regression analysis of non-users who expressed an intention to use either injectables, implants, or pills in the next twelve months or later. In Model I (unadjusted), all attributes, except for non-interference with menstruation, were associated with future intention to use either injectables, implants, or pills. The belief that a method was easy to use or safe for long-term use had the strongest association with the intention to use either injectables, or implants, or pills. This was followed by beliefs such as the method was easy to obtain, use satisfaction within the social network, absence of serious health problems, and effectiveness in preventing pregnancy (OR range 2.90–3.90), as well as absence of long-term fertility impairment and unpleasant side effects (OR range 1.90–2.30). Satisfied past use was associated with increased odds (OR = 3.88; 95% CI = 2.24–6.72) of intending to use injectables, implants, and pills compared to women who had never used any of these methods. If a woman's husband/partner approved the method, she was four times (OR = 4.17; 95% CI = 1.68–10.35) more likely to indicate an intention to use the method compared to her counterparts whose husbands/partners disapproved the method.

In Model II (adjusted), only four attributes remained significantly associated with the intention to use injectables, implants, or pills in the next twelve months or later. Women who perceived the method to be easy to obtain or easy to use had 1.95 (95% CI = 1.03–3.66) times and 4.17 (95% CI = 2.22–7.86) times higher odds, respectively, of indicating an intention to use a method. Similarly, women who perceived the method as safe for long-term use (without a break) had higher odds (OR = 4.51; 95% CI = 1.61–12.64) of indicating an intention to use a method compared to those who perceived otherwise. The odds of choosing that method in future increased if members of the woman's social network were satisfied users of any of the methods (OR = 1.94; 95% CI = 1.18–3.19) compared with women with no or whose social network tried but had an unsatisfactory experience with the method. Additionally, women who had ever used a method and were satisfied had increased odds (OR = 2.87; 95% CI = 1.51–5.46) of choosing the method in future than women who had never used the method.

The lower panels of Table 4 display the effects of respondent characteristics on method choice, first comparing the choice of implant vs. injectable and then the choice of pill vs. injectable. Notably, there were no statistically significant differences in the choice of implants or pills over injectables based on age, fertility preference, and settlement. However, the odds of choosing pills over injectables were reduced for women with incomplete primary education compared to those with no education.

The study has some limitations. The data analyzed is based on women's self-reports, which may introduce reporting biases, such as cognitive and social desirability. The sensitive nature of questions about contraceptive use may lead to underreporting or overreporting of beliefs. Although it would have been ideal to examine the entire spectrum of contraceptive methods, the analysis was limited to three methods (pills, injectables, and implants). The study did not analyse other methods due to insufficient number of cases for meaningful statistical analysis. Despite these limitations, the study is innovative in its detailed measurement of method-specific perceptions that may influence future decisions to adopt or continue using specific methods among women in humanitarian settings. Additionally, there are plans to investigate further how women's opinions or stated intentions about contraceptive choice affect actual method use based on contraceptive calendars to be captured during follow-up interviews.

Intention of contraception among non-users in refugee settlements is low, with widespread negative perceptions of available methods. Future method choice is shaped by perceived convenience, long-term safety of the method, social network satisfaction, and past experiences. These findings highlight the need to improve counseling to counter unfounded negative beliefs and to expand access to a range of contraceptive methods.