Study design:
This was a prospective single cohort study, nested in a randomized controlled trial (RCT) which assessed the effect of peer counselling, compared to routine contraceptive counselling on acceptance of modern contraceptives among female refugee adolescents (21). This study was carried out in Palabek refugee settlement in northern Uganda, from May 2019 to January 2020. Participant recruitment occurred from May to July 2019, and participants were then followed for a period of six months till January 2020. The reason for this short follow up period was because of the mobility of the refugees; they are always moving in and out of the settlement at will, either to go back to their native country or to look for employment opportunities outside the camp. Therefore, a longer follow up period would result in increased loss to follow up. The follow ups were done at one, three and six months after receiving a contraceptive method.
Study setting:
Palabek refugee settlement was established in April 2017 and is one of the newest refugee settlement in Uganda. It is situated at the border of Uganda and South Sudan. It hosts close to 55,000 refugees from South Sudan, 85% of whom are women and children (22). Palabek refugee settlement is burdened with adolescent pregnancy among girls aged 17 to 19 years with a prevalence of over 50% (6), and the settlement has four health centres which provide contraceptive services free of charge. The family planning services provided include contraceptive counselling, provision of condoms (both male and female), oral contraceptives, injectables, and insertion and removal of implants and intra-uterine devices (IUDs).
Study participants: A total of 272 female refugee adolescents aged 15 to 19 years who were new starters of reversible modern contraceptives participating in the RCT were consecutively enrolled into this study. The reversible modern contraceptives available were oral contraceptives (both combined and progestin only), the depo medroxyprogesterone acetate (DMPA) injectable, subcutaneous single rod contraceptive implants and copper IUD. Participants were excluded if they could not comprehend either English, Acholi or Arabic (the languages commonly used within the settlement), or were physically or mentally unable to consent or answer interview questions.
Sample size estimation:
The sample size formula for comparing means in two proportions by Hulley (23) was used. An error of 0.05 was assumed, with a power of 80%. The proportion of adolescents who discontinued on a long acting reversible contraceptives (LARCs) was taken at 35% (24) while the proportion of adolescents who discontinued on other contraceptives was taken at 60% assuming a 25% difference between the two groups. The proportion of refugee adolescents on LARCs was estimated to be 12.6% (13). After factoring these into the formula, we got a sample size of 272 participants. Assuming a 10% loss to follow up ((0.1 x 272)+272), the adjusted sample size would be 300 participants. Therefore, a total of 300 female refugee adolescents would be necessary to answer this objective.
Data collection: As part of the RCT (21), baseline information like social demographics, sexual and reproductive history, type of contraceptive received, and partner’s characteristics was collected at the health centre where the participants received their contraceptives. This information was obtained with the help of an interviewer administered questionnaire. A contact locator was developed which had the participants’ address, telephone numbers and contacts of their ‘significant other’ who could be a relative, spouse or friend. Participants were also given a small card with their particulars and an appointment date for their next follow up. Those who did not come to the health centre at the given appointment were traced with the help of the contact locator. In the event of a missed appointment, they were contacted by phone or home visited within two weeks of the missed appointment. A respondent was considered lost to follow up if she was not traced within the six months of the study. Information collected during follow up included continued use of the contraceptive method (adherence) or discontinuation, the reason for discontinuation, switching to other contraceptive methods, reasons for switching, and side effects experienced. Two trained research assistants conducted the interviews.
Study variables:
The outcome variable was “adherence to modern contraceptives” defined as the use of any contraceptive method in an ongoing and consistent manner as prescribed (25). For adherence to be achieved, both continuation and correct use of the contraceptives are required. Hence, a participant was termed as non-adherent if: i) the participant became pregnant during the previous month (and the pregnancy not attributed to method failure as judged by research assistants ii) missed her appointment and iii) missed taking three or more oral contraceptives during the month according to the Guttman scale (26). Self-reported non-adherence was used for all the other contraceptive methods. For the LARC methods, participants were regarded as adherent for the six-months follow up unless they reported removal of the method. Participants who self-reported pregnancy had to undergo an assessment to determine whether the pregnancy was due to method failure or non-adherence.
The predictor variables included socio-demographics like age, highest level of education attained, marital status, occupation, sexual and reproductive history like age at first sex, number of children alive, partner’s characteristics and type of contraceptive given.
Statistical methods: Data were analysed using STATA version 14.0. Continuous variables were summarised as means and standard deviations if they were normally distributed, and as medians and ranges if skewed. Categorical variables were summarised as percentages and proportions. Adherence to contraceptives was analysed as a binary variable, with adherence coded as “1” and non-adherence as “0” at the different time points. Percent adherence at the different follow up times was then obtained. Predictors of adherence were assessed using Generalised Estimating Equations (GEE), with the exchangeable correlation matrix, family binomial and link logit. Bivariate analysis was done by fitting a model for each independent variable with the outcome. All the variables that gave a p-value ≤0.2 at the bivariate analysis were considered for multivariate analysis. Also considered for multivariate were variables which are known to be of significance to adherence from previous research even though they had p-values greater than 0.2.
At the multivariate analysis, GEE was run for all the variables which met criteria for it, and the non-significant ones (p>0.05) were dropped one at a time. Those which retained statistical significance were used to assess for interaction and confounding. Two-way product terms were formed between the variables and used in the assessment of interaction using the chunk test. A variable was considered a confounder if it caused a greater than or equal to 10% change in the odds ratio of adherence to contraceptive use. Odds ratios along with their 95% confidence intervals were reported, and statistical significance reported at p<0.05. Robust standard errors were used in the GEE analysis. To deal with those who were lost to follow up, both intention to treat (ITT) and per protocol analyses were conducted.
Sensitivity analysis:
This was conducted to assess if the participants who were lost to follow up had an impact on our study findings. We first grouped the lost to follow under non-adherent, and then re-analysed them under adherent, and conclusions were checked if they remained similar.
Ethical considerations:
Ethical clearance was obtained from the School of Medicine Higher Degrees Research Ethics Committee (REC REF 2018-059), and from Uganda National Council of Science and Technology (SS 4876). We ensured that all participants gave their written informed consent before enrolling them into the study, and guaranteed their confidentiality with the use of number codes instead of participant names. For participants below 18 years, we did not seek for parental consent because it is stated in the National Policy Guidelines and Service Standards for Sexual and Reproductive Health and Rights, that all individuals seeking family planning services are eligible to receive the services regardless of their age, without need for verbal or written consent from parent, guardian or spouse, as long as they have received prior education or counselling about the contraceptives (27). All interviews were held in private places where discussions could not be over heard.