The IRC began developing tools and a protocol that do not require literacy or numeracy for treating uncomplicated SAM cases in 2015, using a human-centered design approach. The first round of development occurred in Chad, Mali, and South Sudan, where local CHWs and IRC’s nutrition staff provided input. The additional rounds were used to refine and test generalizability in other contexts: round 2 (a design workshop in Mali, with additional attendees from Chad, Niger, and South Sudan), round 3 (India), round 4 (South Sudan), and round 5 (South Sudan). The last round was completed in November 2016. More detailed description of this process is available elsewhere [10]. The process resulted in five adapted tools: 1) patient register, 2) modified mid-upper arm circumference (MUAC) tape, 3) weight scale decal to identify daily dosage of treatment, 4) weekly dosage calculator, and 5) visual counselling cards (Figure 1) [9]. Detailed descriptions of each tool and the treatment protocol are available in Appendix S1 of the Online Supplementary Document.

Figure 1.  Low-literacy-adapted job aids and tools. Reproduced from [9], with permission.

The pilot was conducted in Aweil South County, Northern Bahr El Ghazal State, South Sudan, between March-September 2017. CBDs selected for study participation had worked for an IRC-supported integrated community case management (iCCM) of childhood illness program and were already trained to treat children under-five for uncomplicated pneumonia, diarrhea, and malaria. The CBDs received 100 USD per month as an incentive through the iCCM program, per national policy. Each CBD served 35 or more households. We note here that the iCCM program became inactive a month before study implementation due to a lapse in operations when the program was being taken over by another organization from the IRC.

During the peak lean season in July 2017, the study area reported a global acute malnutrition (GAM) prevalence of 17.7% [11]. Four out of eight payams, an administrative unit in South Sudan, in the county were purposively selected, focusing on areas that were accessible during the rainy season. As part of the existing iCCM program, CBDs were already trained to treat children under five in their communities for uncomplicated pneumonia, diarrhea, and malaria. They had also been instructed to refer children who present with medical complications to the nearest health facility, as well as children who meet the MUAC cutoff for SAM (<115 mm) or moderate acute malnutrition (MAM) (115 to <125 mm) to the nearest static clinics running an Outpatient Therapeutic Program (OTP, where SAM children are treated) or Targeted Supplementary Feeding Program (TSFP, where MAM children are treated) respectively.

Of the 397 iCCM CBDs active in the four selected payams, CBDs with the following characteristics were excluded from eligibility for this study (one person may contribute to multiple categories): male (n = 1), received any education (n = 2), lives within 5 km of the nearest OTP clinic (n = 171), CBD’s house was not accessible during rainy season (n = 9), participated in previous training or field testing of the protocol and tools (n = 19), and had travel time greater than 60 minutes to the farthest household in the catchment area (n = 1). Also, to assure a sufficient volume of SAM cases, we only included CBDs who reported serving 35 (the median value among all CBDs) or more households. This left 106 CBDs eligible across the four selected payams.

Out of those CBDs who met the inclusion criteria, fifteen CBDs per payam were randomly selected for a six-day training. The research staff obtained informed consent from CBDs to collect sociodemographic data and data on their performance following the training. Based on a performance assessment conducted at the end of the training, 11 highest performing CBDs from each payam who met a pre-determined minimum performance score of 80% were then deployed to deliver SAM treatment in the community. More details on the composition and the results of the performance score are described elsewhere [9]. The total number of CBDs deployed for the pilot was based on limitations of supervisory capacity of the Research Officers.

The CBDs received an additional incentive of 25 USD per month from the project to compensate them for opportunity costs they faced to provide SAM treatment in their communities. Four Research Officers hired for the project supervised the CBDs through biweekly visits. During the supervision visits, the Research Officers observed the CBDs providing SAM treatment. Based on their observations, they filled out a checklist to assess the CBDs’ performance. The Research Officers provided feedback to the CBDs during each supervision.

CBD supervisors, a pre-existing cadre from the iCCM program, were tasked with the distribution of ready-to-use therapeutic foods (RUTF) from centralized distribution points to the homes of the CBDs. Every month, IRC staff prepositioned RUTF at centralized storage facilities in each administrative unit (payam) the IRC operated the study in. Every week, the CBD supervisors retrieved RUTF at the storage facility closest to the CBD homes and replenished the RUTF stored at the CBD level. This way, considering the high value of RUTF in food-insecure South Sudan, the quantity of RUTF stored at the CBD home was kept to a minimum. They received an incentive of 50 USD, and an additional 15 USD for those who requested financial support for transporting the RUTF.

In addition to the incentives mentioned above, the following project inputs were provided: equipment (locked metallic boxes for RUTF storage at CBD homes and at centralized storage facilities, one set of treatment tools for each CBD), drugs (RUTF, amoxicillin, albendazole), and human resource (four Research Officers for supervision and data collection).

After conducting community sensitization meetings informing community members about the availability of malnutrition treatment from CBDs, CBDs passively identified cases among children who arrived at the CBDs’ home for iCCM and/or malnutrition treatment. They measured the child’s MUAC using the modified MUAC tape with the following color bands: dark red (<90 mm), red (90 to <102.5 mm), pink (102.5 to <115 mm), yellow (115 to <125 mm), and green (≥125 mm) (Figure 1). Children were eligible for treatment if their MUAC was in the red or the pink zones (90 to <115 mm), and if they had no medical complications, a good appetite, and weighed more than 4 kg. Children who had MUAC in the dark red zone (<90 mm), failed the appetite test, or weighed less than 4 kg were immediately referred to the OTP as a danger sign out of concern of severity.

The treatment protocol consisted of danger sign screening, MUAC measurement, administration of an appetite test, weight measurement, determination of daily and weekly RUTF dosage based on weight, drug provision (Amoxicillin on week 1, Albendazole on week 2), and counselling. (Full description available in Appendix S1 of the Online Supplementary Document.) Caregivers were instructed to return on a weekly basis for up to 16 weeks, until reaching a treatment outcome: recovered (two consecutive weeks with MUAC≥125 mm, or the green zone on the MUAC tape), default (three consecutive missed visits), non-response (had not recovered by 16 weeks), death, or referred. Referrals were made if the child presented with clinical danger signs or MUAC dropped below 90 mm. Also, without numeracy, CBDs are unable to differentiate slow regression, static, or slow progression unless MUAC values cross MUAC tape color bands. Thus, we also added a conservative protocol to refer children who stayed in the same color zone for four consecutive weeks. The sample size calculation for how many incident SAM cases to enroll in the study was based on a recovery rate of 75%, precision of 10%, alpha of 5%, and loss-to-follow-up of 10%, resulting in a sample size of 80 children per payam, for a total of 320 children. CBDs were asked to stop enrolling new children after the study met the required sample size.

The unreliable phone network made it difficult for the study staff to guarantee presence during the enrollment visit to administer the informed consent process. Thus, in consultation with the IRC Institutional Review Board, a two-stage consent process was developed. The CBDs were asked to receive oral consent from the caregiver for their eligible child to receive treatment from the CBD, then the study staff followed up at the caregivers’ homes to receive oral informed consent to participate in the research study and to collect sociodemographic data on the child and the household.

The study was approved by the Research Ethics Committee of the South Sudan Ministry of Health and the IRC Institutional Review Board. Informed oral consent was obtained from participating CBDs and caregivers.

The CBD-delivered acute malnutrition treatment model showed high recovery rates among uncomplicated SAM cases, despite the underlying food insecurity and the low literacy of the CBD cadre. These data, combined with the high CBD performance scores for protocol adherence reported in a separate publication [9], show promise for deploying CHWs to provide acute malnutrition treatment, regardless of their literacy levels. The pilot was not powered or intended to capture longer-term health outcomes or changes in incidence or prevalence of acute malnutrition. In a low-income, conflict-affected context like South Sudan, a large proportion of under-five deaths is still driven by the preventable infectious causes treated by iCCM [12]; given the well-documented relationship between infection and malnutrition [13], a program design that treats both infection and malnutrition in the community may have potential to more effectively reduce incidence of both.

The children enrolled by the CBDs had lower MUAC at admission than those at the nearest OTPs, and a large majority of the former reported that they had not received any malnutrition treatment in the last four months. While we did not have the resources to assess change in treatment coverage attributable to this pilot, this point, along with the selection of CBDs residing more than 5 km from the nearest OTP for our study, suggests that the model may have successfully reached those who were previously not receiving care rather than replacing OTP services. There is very limited evidence from studies with designs that allow for strong causal inference on the impact on coverage [8]; a previous study conducted in Mali demonstrated a statistically significant change in coverage, comparing intervention and control (intervention arm coverage from 43.9% to 86.7%, control arm coverage from 43.8% to 41.6%, P < 0.05) [14]. The same study also showed cost-effectiveness [15]. Further exploration is necessary to understand the net effect of introducing a CBD-delivered malnutrition treatment model at scale, and the cost-effectiveness of such a model, taking into account collateral outcomes beyond change in coverage and treatment outcomes.

One challenge to developing a low-literacy protocol was how to monitor whether cases are stationary, slowly regressing, or slowly progressing in treatment, given that numeric MUAC values could not be captured by our CBDs. To address this, we created smaller MUAC color zones (separating the traditional MUAC tape red zone of <115 mm to red zone of 90 to <102.5 mm and pink zone of 102.5 to <115 mm) and set a safeguard for referral after four consecutive weeks in the same color band. Based on the larger than expected proportion of referrals for this reason, we will have to conduct further research to determine whether this is driven by program performance vs a protocol that was too conservative. Further exploration is needed to adjust this safeguard.

Treating malnutrition in the community introduces several operational challenges in addition to those that affect facility-based programs. First, the community acceptability of CHWs treating acute malnutrition is unclear. In qualitative data collected for this pilot, community members expressed general acceptability, but also expressed suspicions that CBDs may be operating on favoritism since CBDs turned away some children from treatment. CBDs in response noted that they clearly explained to the caregivers using the MUAC tape that only children whose MUAC landed in the red or pink zone qualified for treatment from them [9]. Second, RUTF needs to be stored in the CHWs’ homes. While no major incidents were reported during the course of the study, a few CHWs expressed initial anxiety of keeping supply in their house (Unpublished data). Another challenge is that the supply chain needs to reach the CHWs’ homes (see Figure 2 for supply chain model used). CBD supervisors, who were responsible for RUTF delivery had difficulty with moving the physical bulk of the supply, despite the fact that one CBD supervisor had no more than 3 CBDs to stock. The iCCM model seeks to provide physically proximate care for all children, but given the difference in bulk of the commodity comparing iCCM treatment (amoxicillin, ORS, ACT) to malnutrition treatment (RUTF) and the lower incidence of SAM compared to iCCM conditions, the CBD-delivered malnutrition treatment model may need to be less saturated, or in other words, one CBD may need to serve a larger number of households specifically for acute malnutrition. We also note here that the level of supervision available to the CBDs in this study was more than a standard iCCM program affords to the CBDs; iCCM CBDs receive supervision at a monthly frequency, but often less frequently due to logistical constraints, and there is a possibility that treatment outcomes will deteriorate when the nutrition program receives less supervision than what was provided in our study.

Figure 2.  Supply chain system for the study.

There are several limitations to the study. The original objective of the study was for iCCM and malnutrition treatment to be delivered at the same time, hence the existing iCCM CBD cadre was recruited for the study. However, the iCCM program funding transitioned from IRC to another organization in April 2017, and the subsequent organization did not start up the iCCM program operations until after September 2017. It is unclear if and how the treatment outcomes were impacted by this. Our pilot project provided the CBDs with amoxicillin and albendazole per the malnutrition treatment protocol, but the CBDs reported stock-outs in ORS and ACTs due to this pause in iCCM programming. Treatment outcomes may have been worse with a more complicated service delivery model and larger workload for the CBDs with the iCCM program operating, but it is also possible that treatment outcomes may have been better than what was seen here, with successful co-treatment of iCCM conditions. Given that this was a feasibility study, we had a small sample size of CBDs, purposively selected to ensure physical access of the research study team over the duration of the study. We cannot ascertain that the capacity of the CHWs selected and those who were not are comparable, and may see different results in a scaled program.

The high recovery rate for SAM children enrolled in acute malnutrition treatment by low-literate CBDs shows promise that deploying CBDs to treat SAM in areas with high prevalence and low treatment access may lead to higher recovery, better continuity of care in the transition between severe and moderate acute malnutrition, and shorter treatment time. Proper adaptations of tools and protocols can empower CHW cadres with no formal education to successfully complete life-saving tasks. Key questions of scalability and cost-effectiveness remain.