Study design and setting
The study compared the relative effectiveness of integrating RFR with PFP (RFR + PFP) in rural households on adolescent outcomes of asset building, school attendance, mental health, experienced stigma, and food security to: 1) households with adolescent participants in RFR only and 2) households with parent participants in PFP only (i.e., no adolescents in the household were in RFR). The study measured the outcomes over a 24-month period and was conducted in the same 10 villages as the PFP effectiveness trial . In addition, a purposive sub-sample of male and female young adolescents participated in in-depth qualitative interviews (at approximately 12 months post baseline) with trained staff on the perceived benefits and challenges of RFR participation. Adolescents were specifically asked to discuss their experience in raising the rabbits, what supports or challenges they may have encountered, how they used any earnings from selling rabbits, and how they perceived RFR affected their status in their community and families. The 10 rural villages for the study were selected based on: (1) feasibility of conducting and managing the study and intervention over a diverse geographic area; (2) village leadership commitment to PFP, RFR, and the study; and (3) existing relationships between the partner Congolese organization and village leadership.
Participant recruitment and eligibility criteria
Community meetings were held first with parents participating in the established PFP program and other adults in the community interested in PFP. The meetings provided details on the RFR and PFP programs and the associated study. The PFP effectiveness study enrolled households with men and/or women (16 years and older) who expressed an understanding and commitment to productive asset/microfinance programs (e.g., credit and repayment of loans), were permanent residents of the village, and were responsible for the household. The eligibility criteria of 16 years and older accommodated rural realities where, due to the conflict, elder siblings are caregivers for younger family members and some women and men are married and have children by 16 years of age. Details on the PFP recruitment/eligibility process are published elsewhere . Male and female adolescents aged 10–15 years of age were eligible for the study if: (1) their parent/guardian was enrolled in the PFP program, but was in the delayed control group and had not yet received the pig asset loan; (2) youth and their parent/guardian expressed an interest and commitment to the program (e.g., willing to build a rabbit hutch, attend meetings, repayment of rabbit offspring and transfer of asset (rabbit) to other adolescent participants); and (3) were permanent residents of the participating village. We expanded the eligibility age to 15 years for the study at the request of parents and our study partner. Although young adolescents are typically defined as age 10–14 years, parents felt 15 years was a vulnerable age and that engaging them in RFR may prevent them from leaving the village to find work in the mines or marrying early. Parents/guardians provided consent for their eligible child to participate in RFR and the study prior to children being approached for assent. As the RFR study builds off of the sample of households randomized into the PFP trial, randomization of youth was not conducted for RFR. An eligible male or female adolescent in each participating PFP household was selected. If the household had more than one eligible adolescent, the parents selected the adolescent to participate (see Fig. 1). Parents with eligible adolescents in the household enrolled in the first PFP delayed control group comprised the PFP only group, which did not include a loan of a rabbit to the youth during the study period. Parents with eligible adolescents enrolled in the second PFP delayed control group comprised the PFP + RFR group. Parents for the RFR only group were recruited through outreach to village households with the help of local leaders to identify eligible and interested adults and young adolescents. Households that were recruited for RFR only had not previously participated in the microfinance/asset transfer program.
RFR microfinance/asset transfer program
RFR works with young male and female adolescents (10–15 years of age) that are interested and committed to raising rabbits, including building a rabbit cage, providing care for the rabbit, participating in a training program and meetings with support by our skilled RFR implementation team, and repaying the loan to the project in the form of two female rabbits when the original rabbit gives birth. These rabbit loan repayments are then given to other adolescents in the project but not in the study. Similar to PFP, the original rabbit loan and remaining offspring are for the RFR members to continue to raise, breed, eat, and sell as decided by the adolescent in collaboration with their parent/caregiver and mentorship from our team. While continuing to breed and raise rabbits, the adolescents have an option to sell some of their rabbits at local market rates (approx. 10 USD value) to pay for school fees, purchase food or medicine, or invest in other income generating activities. The RFR + PFP implementation team facilitated meetings and home visits with parents in PFP and adolescent members in RFR to discuss challenges and identify solutions related to participation, guide members in raising livestock/animals, and encourage timely repayment of pig and rabbit loans. Our team also facilitated regular visits by a project veterinarian technician to review animal care protocols and provide vaccinations as appropriate and needed.
The main outcomes of this study focused on young adolescent asset building, school attendance, mental health, experienced stigma, and food security. The study survey was developed after a review of existing and validated tools that have been used with young adolescents in similar low-resource and conflict-affected settings and through multiple field tests with adolescents not living in study villages. Survey refinement was an iterative process that lasted several months and included translation and back translation from English to French and from French to local languages (Swahili, Mashi) for administration by Congolese team members trained to conduct interviews with young adolescents.
The asset building measure was developed for the study to examine young adolescent perceived resources that will move them towards economic well-being now and in the future. The measure consisted of six items (e.g., I am learning how to better earn cash or acquire income generating assets for myself and family, I have more ways now of financially supporting myself and family than I did 1 year ago, I am acquiring the skills I need to earn the income I desire). Participants responded yes or no to each item. The score is the proportion of items endorsed. Cronbach’s alpha = 0.81 in this sample.
Adolescents enrolled in school answered questions about the number of full days of school that they missed in the past one-month. The number of days missed was measured on a 1–4 scale (0, 1–2 days, 3–5 days, and 6 or more days missed from school in the past month). Since the distribution was extremely skewed, this variable was dichotomized to adolescent report that he/she missed two or more days in the last month versus missed 0–1 days in the last month. Participants that were not enrolled in school were assigned a value of 1 (equivalent to missed 2 or more days of school in the last month).
Adolescent mental health
The reduced Acholi Psychosocial Assessment Instrument (APAI) was developed for use with young adolescents living in rural, post-conflict Northern Uganda [21, 22]. The Internalizing and Prosocial subscales of the APAI were used to assess mental health outcomes. Items are assessed on a 4-point response scale (i.e., never, sometimes, often, always) over the past 7 days. The 19-item Internalizing subscale reflects anxiety and depression (e.g., has constant worries, thinks she/he is of no use, cries when alone) and had a Cronbach’s alpha = 0.70 in this sample. The 8-item Prosocial subscale reflects positive social behaviors (e.g., listens to others and elders, plays together with others, helps others) and had a Cronbach’s alpha = 0.84 in this sample.
We adapted a measure of everyday discrimination [23, 24] to look at “chronic, routine and relatively minor experiences of unfair discrimination”  or stigma experienced by adolescent participants. Participants answered eight questions (e.g., people act as if they are afraid of you, people treat you with less respect than others, you are called names or insulted) on a 3-point scale (i.e., never, sometimes, always) about the frequency of different types of experienced stigma occurring in their day-to-day life. Cronbach’s alpha in this sample was 0.79.
The Household Dietary Diversity Scale (HDDS)  assessed the total number of food groups (range: 0–12 items) consumed by the household members in the previous day and night as reported by the adolescent. Food security as measured by the HDDS is also “used as a proxy measure of the socio-economic level of the household” .
The baseline survey with young adolescents took place after recruitment but prior to RFR program training that included details on health, nutrition, and well-being of the rabbit, building the hutch, composting waste, asset loan distribution, and repayment. Assenting adolescents were informed at the beginning of the survey and during the discussion of sensitive topics (e.g., mental health) and that they could stop or refuse to answer or skip questions at any point without consequence to their participation in RFR. Surveys were conducted after the school day was complete or on the weekend in a private place (i.e., in home or outside the home) identified with the adolescent. Trained Congolese staff used a tablet with the pre-programmed study survey. Use of the tablet was successful in our previous PFP trial in multiple ways: (1) reduced logistical burden of printing and managing the paper questionnaires; and (2) ensured real-time access to the data to monitor data quality and identification of issues so that they could be remedied between surveys. Additionally, participants expressed confidence and comfort when answering questions with the use of the tablet as compared to a paper-based survey where staff write down responses. Adolescents that completed the survey received approximately 1.50 USD for their time (45–90 min), an amount recommended by study partners and consistent with previous studies. Data recorded on tablets are encrypted. Once uploaded to a central US-based server, the data are automatically erased from the tablet. Follow-up interviews were conducted with adolescents at 12 and 24-months post-baseline.
In addition to the surveys, a purposive sub-sample of 30 young male and female adolescents across age ranges in both the RFR only and RFR + PFP groups in each village were selected by the team and invited to participate in in-depth qualitative interviews. Each team member selected three young adolescents that had been active in RFR as well as those that appeared to struggle in the program to learn about the benefits and challenges of participating in RFR and their recommendations for revisions to the programs. We strived to select and recruit equal numbers of girls and boys stratified by age. The in-depth qualitative interviews were conducted at about 12 months from baseline interviews. The in-depth interviews were conducted in private after parental consent and adolescent assent, and lasted on average for 60 min. Participants were reimbursed approximately 1.50 USD for their time.
Study identification codes and names were recorded during one-on-one surveys and in-depth interviews. All data recorded through the tablet-based program and audio recorder were uploaded to a password-protected server managed by the study team. Names were centrally removed and stored in a separate file.
The Johns Hopkins Medical Institute (JHMI) Institutional Review Board (IRB) approved this protocol. As at the time there was no local IRB functioning in South Kivu province of Eastern DRC, an ad hoc committee of respected Congolese scholars at the Université Catholique de Bukavu and community members reviewed the research and intervention protocols before giving approval for the study. A letter of approval from the Congolese scholars was submitted to the Hopkins IRB. With approval from the local experts and Hopkins IRB, all surveys and in-depth interviews were conducted after a parent/guardian provided informed oral consent for their child to participate and the adolescent provided voluntary, informed oral assent to our skilled research team members. Oral consent was chosen as the majority of parents/guardians (60%) had never been to school and were unable to write their name or the name(s) of their children. Oral consent and assent allowed for no names to be linked to survey or in-depth interview data.
Sample size and power
Power for the study is based on a sample size of 480 young adolescents (160 per group), power of 0.80, and α level of 0.05. The study can detect a significant difference between groups if the change over time in APAI [21, 22] scores is 2.67, 2.82, and 2.98 greater in one group for ICCs of 0.001, 0.005, and 0.01, respectively.
Differences between the three groups (RFR + PFP, RFR only, PFP only) on baseline characteristics were compared with generalized linear models with robust standard errors to account for the nesting of adolescents within the 10 villages. Three-level mixed models were used for the main analyses. Time (baseline, 12, and 24 months) was nested within adolescents and adolescents were nested within villages. Group, time, and the group by time interaction were included in the model. Normal Gaussian distribution models were used for all analyses except for school attendance (missing two or more days of school in past month) which used a logistic model. All adolescents (N = 542) were included in the analyses. Mixed models do not require complete data at all time points so all available data were included in the analyses. Analyses were intention-to-treat with adolescents in the group as assigned even if they did not receive a rabbit asset loan (e.g., were unable to build a suitable hutch for the rabbit). Exploratory stratified mixed models by age group (10–11, 12–13, and 14–15 years) and sex were conducted to determine if the differences between groups varied across age and sex. Since the study was not powered for these analyses, effect sizes were examined and compared across strata.
All qualitative interviews were recorded and transcribed by local team members from the local languages (Swahili or Mashi) into French. A code list was then developed by co-author LMJW to identify adolescent responses relating to benefits and challenges of participating in the RFR program by study outcomes (e.g., asset building, health, and school attendance). The code list was reviewed with Congolese team members for accuracy and applied to the French transcripts. All coded statements were then extracted, translated to English, and aggregated for analysis using qualitative descriptive methods. To understand the relative frequency of coded statements, a count of the number of interviews in which each theme was discussed was also tabulated. As a final step, a sample of statements from each of the emergent themes was selected for presentation to characterize the voice of the young adolescent participants in this article. All quoted statements were labeled by gender, age, and village.