IDIs

All interviews were conducted in May 2015 with three general categories of informants: 1) policymakers (national, state, and county), 2) program implementers (national, state, or county), and 3) health workers in PHCCs and PHCUs primary health care centers (referral facilities) ( Table 1 ). Respondents were purposively selected based on their ability to provide varied perspectives on research questions. Separate interview guides were developed for each type of informant.

FGDs

All FGDs were conducted in Payinjiar in May 2015 with the following groups: caregivers, CBDs, CBD supervisors, and community leaders. Ten CBDs from each of the targeted payams were randomly selected. Caregivers were mobilized by iCCM CHD officers, county Ministry of Health (MoH) officials, and CBD supervisors in each selected area on the day of the FGD. The caregiver FGDs were comprised of female participants; eligible participants had to have utilized CBD services in the recent past and have children under five years of age to ensure caretakers would be positioned to address research questions focused on how the emergency affected available CBD services, given their previous knowledge and experiences. Community leader participants were recruited by the iCCM Officer and CBD supervisors from a central area in the selected payams where leaders congregate. The categories of respondents and topics for both IDIs and FGDs are summarized in Table 1 .

Data collection

Guides were composed in English and translated into Nuer, then back-translated to check accuracy. The guides were subjected to several rounds of pre-testing and revision. Two Nuer-native translators from Payinjiar County and three expatriate researchers from Forcier Consulting, a private research firm, served as qualitative data collectors.

The IDIs and FGDs were conducted in Payinjiar by two teams, each consisting of a researcher from Forcier Consulting and a Nuer interpreter. Two IRC staff provided technical oversight of the FGDs for quality assurance and only attended FGDs that did not feature IRC-funded staff. The third Forcier Consulting researcher conducted all IDIs with national-level informants in Juba in English. IRC representatives did not attend any IDIs.

Due to staff concerns regarding the sensitivity of recording conversations in a conflict context, all the Nuer-language interviews were translated verbally in real-time. Notes were taken by hand by the Forcier researcher and later typed. For English-language interviews, notes were taken in English by the researcher. Each day, the team debriefed on any challenges during data collection.

Informed consent was obtained from all respondents. Ethical approval was received from the Institutional Review Board of South Sudan MoH.

Data analysis

Data analysis was conducted using a combination of deductive and inductive methods. Initial codes were developed to represent higher-level themes based on the research questions: 1) background, 2) challenges of implementing iCCM prior to the crisis, 3) impact of the conflict, 4) response to the challenges created by the conflict, and 5) recommendations regarding response to conflict or general health service provision. Within each of higher-level theme, sub-themes were identified inductively from the data. The data were coded by hand, once by an IRC focal point and once by a UNICEF focal point independently, and organized into an Excel matrix (row for theme, column for each interview or FGD). The primary author used both sources to summarize the major themes.

Data collection

The routine monitoring data collection system was as follows. CBDs entered data into a Patient Register that captures data on name, sex, age (2 to <11 months vs 12 to <59 months), assessment and classification of illness, treatment given, and whether the child was referred to a health facility. CBD Supervisors used monthly reporting tools that aggregated the aforementioned information, along with drug stock availability at the CBD level. The aggregated data were reviewed and verified by iCCM Program and Monitoring and Evaluation Officers and entered into the program’s District Health Information System (DHIS) database. No changes to the data collection system or tools were implemented in response to the crisis or following the emergency phase. Supervisors and program staff continued to follow the standard protocols and procedures for monthly and quarterly reporting periods. The data elements extracted for this study are listed in Supplemental Table 1. Population estimates for host and IDP communities were taken from the Initial Rapid Needs Assessment mechanism (example report is cited here [14]) and data from World Food Program and the South Sudan National Bureau of Statistics [15].

Other indicators (see Table 2 ) were calculated using the data elements from Supplemental Table 1 to describe utilization, access, and coverage. All data points were exported into Microsoft Excel and pivot tables were created for each indicator and summarized by month. Some indicators were compared against relevant DHIS data from health facilities in Payinjiar County that were supported by the IRC at the time. Monthly data on IRC’s Payinjiar County warehouse stock of artemisinin-based combination therapy, oral rehydration salts (ORS) and amoxicillin and CBD stockout rates were also available. Based on previous programmatic experience however, the quality of the stockout data is questionable.

Population displacement

Payinjiar County experienced a significant influx of IDPs beginning in December 2013 and primarily lasting through the first quarter of 2014. The influx nearly doubled the estimated county population from 44 224 to a total of 83 433 [16]. Some Payinjiar County residents were displaced as well.

Caregivers, CBDs, and CBD supervisors shared painful accounts from the acute emergency. One caregiver recounted, “I moved with the healthy children when the fighting came. We left the lame and sick children behind. We all lost family, people starved to death, things were burned, we fled our homes. We left with nothing, now we have nothing.” Another caregiver shared, “People died from hunger, even if you managed to run with your child, many [later] died from hunger. There are also other dangers, for example from wild animals. Many people died.”

Additionally, during 2014 and most of 2015, Payinjiar County was classified at critical levels of food insecurity at phase 3 (crisis) or phase 4 (emergency) [16]. CBDs were reported as having spent more time outside the home in search of food and water for their own households. An IRC program staff member reported, “Issue of food security was drastically affected by conflict. People were not able to cultivate, had to travel farther to find food. CBDs are not at home.” CBDs recalled an increase in the number of malnourished children. One caregiver noted, “Children eat without milk, [which is] causing a lot of diseases...Another [is] we don’t have enough food. [We are] now eating grass.” CBD supervisors faced challenges finding CBDs for regular supervision visits as they were more frequently absent.

Continued care through displacement

Displaced CBDs were occasionally tracked down by their CBD supervisors and home community members to notify them when it was safe for them to return. This often required supervisors from different catchment areas to travel to and communicate with each other if a CBD was displaced to another supervisor’s area of supervision. CBD supervisor reported, “The community worried a lot about their children when the CBD ran away… No iCCM activities continued until they came back. Only CBDs know how to treat a child.” A respondent within the national MoH recognized, “The availability of the iCCM services was not affected so much by the conflict. CBDs tend to move with their communities when these move and they take their boxes with them. Wherever the community settles, the CBDs continue to provide services to their communities.”

The reporting rate of CBDs to their supervisors dropped from 93% in December 2013 to the lowest value of 77% in February 2014, but a majority of the CBDs reported activity throughout the crisis period, and the rate only recovered from there (Figure S2 in Online Supplementary Document(Online Supplementary Document) ).

CBD utilization

Using the host population as the denominator, an under-five child had on average 5.7 contacts with a CBD per year in December 2013. This dropped to the lowest value of 3.3 contacts in February 2014, but recovered to 5.6 contacts by May 2014. When including the estimated influx of IDPs in the denominator, the rate was lowest in February 2014 at 1.4 contacts per child per year. The number of contacts with CBDs was consistently higher than with health facilities ( Figure 2 ).

Figure 2.  Average number of contacts per under-five child per year in four payams of Payinjiar County, Dec 2012-Dec 2014. The area between the vertical lines roughly represent the period of acute crisis.

CBD treatment

The total number of treatments provided by CBDs is displayed in Figure 3 . While a drop in the number of treatments was seen around February 2014, the total number recovered to pre-crisis levels within a few months. The number consistently remained higher than health facilities.

Figure 3.  Total number of treatments provided to under-five children by CBDs in four study payams and in the nearest health facility (encompassing greater catchment area), Dec 2012-Dec 2014. The area between the vertical lines roughly represent the period of acute crisis.

Treatment rates (number of treatments provided per under-five child per year) are displayed in Figure 4 (panels A-C) by illness, plotted with a horizontal line representing the expected incidence per child. Given that South Sudan-specific estimates were not available, the Uganda national estimate of 2.9 cases per child per year was used as a proxy for fever suspected as malaria [17], and the Sudan estimates of 3.4 cases [18] and 0.5 cases [19] per child per year were used for diarrhea and pneumonia respectively.

Figure 4.  A) Presumptive malaria treatments per under-five child per year, provided by CBDs and health facilities. B) Diarrhea treatments per under-five child per year, provided by CBDs and health facilities. C) Pneumonia treatments per under-five child per year, provided by CBDs and health facilities. The area between the vertical lines roughly represent the period of acute crisis.

Assuming a catchment area only consisting of the host population, CBDs’ treatment rate of fever suspected of being malaria dropped below the expected incidence of fever from December to March 2014, but otherwise recovered. When accounting for the influx of IDPs, CBDs presumed treatment rates per child per year were below the incidence of fever suspected of being malaria. ( Figure 4 , panel A).

A similar pattern was observed for diarrhea (treatment defined as receiving ORS, with or without zinc) ( Figure 4 , panel B). Assuming a catchment area only with the host population, CBDs’ treatment rate of diarrhea dropped below the expected incidence around February 2014 and again around July 2014 (perhaps due to ORS stockout, see Supplemental Table 2 in Online Supplementary Document(Online Supplementary Document) ), but otherwise remained above the expected incidence. When accounting for the IDP influx, CBDs did not provide enough treatments to cover the expected incidence.

For pneumonia, almost no treatments were provided immediately after the crisis. The average number of treatments per child per year provided by CBDs before December 2013 was 0.74, above the expected incidence. From January-July 2014, this number dropped to 0.21 when accounting only for the host population or 0.09 when including IDPs. For August 2014-January 2015, the CBDs treatments recovered to 1.0 when accounting for the host population and 0.4 when including IDPs. The drop in pneumonia treatments between February and August also coincided with national-level stockout of amoxicillin. Treatment rates began to rise in August, after new stock was received ( Figure 4 , panel C).

For all three illnesses, treatment provided by CBDs remained higher than health facilities.

Preference for CBD care

Caregivers preferred seeking care from CBDs over risking an insecure journey to a potentially ill-equipped or unstaffed nearby PHCU for care; for many communities, the only PHCC was a half day’s travel by foot. Many caregivers believed that MoH facilities were non-functioning; one caregiver mentioned, “the (referral facility) has been closed, the health workers all left, they did not want to be on the front lines.” Based on available programmatic data, this was not entirely accurate. Three PHCUs located within the study payams had closed in July 2013, recently but before and unrelated to the crisis. The other facilities remained open even if at lower capacity with an additional mobile PHCU deployed in the study payam of Thornom (location of the largest IDP settlement) as part of IRC’s emergency response efforts in July 2014. A majority of caregivers believed that CBDs were their best option for care. Both community leaders and caregivers, as well as CBDs themselves, reported pressure on CBDs to provide treatment for the population beyond the age of five during the emergency phase. A CBD stated, “only challenge is drugs [are given only for children] under five years downward. Big challenge is 16 [years] and above, [they] come to us when sick, [saying] ‘Why are you not giving us [drugs]?’” Similarly, a caregiver criticized that “the CBD can help, but they only help children, not us older people.”

CBD workload

The average number of children seen by each reporting CBD dropped to approximately 10 children per month around the time of the onset of the conflict. The lowest caseload occurred in February 2014. The number increased from then, with roughly 21 children being seen per month by each active CBD in January 2015 ( Figure 5 ).

Figure 5.  Average number of under-five seen per CHW per month. The area between the vertical lines roughly represent the period of acute crisis.

Supply chain

Several interviewees reported the negative impact of the crisis on drug stock. One CBD supervisor reported that drugs used to be stored in large quantities in Ganyiel, but they no longer felt secure doing so with the threat of damage or looting. Many CBD supervisors maintained a practice of storing the stock for their catchment areas in the local health facilities or in their homes where they had easy access. Supervisors perceived other storage options as riskier, although the IRC’s drug warehouse in Payinjiar County was not looted during the study period. Drug stockouts at the warehouse level occurred occasionally due to the program’s inability to receive resupplies from Juba. According to warehouse data, the stock for ACT and ORS recovered within a few months, while zinc and amoxicillin stock remained low or out (Supplemental Table 2).

Numerous interviewees reported an increase in demand for CBD services with the influx of IDPs, which subsequently led to increased pressure on supplies. Several CBD supervisors indicated that stockouts were more frequent among CBDs working in areas with IDPs. Several CBDs, CBD supervisors, and caregivers indicated that stockouts were not a major concern prior to the start of the crisis. With increased frequency of CBD stockouts, some CBDs referred children to other CBDs who had drugs left.

Supervision

CBD supervisors faced several difficulties in continuing supervision, due to concerns of physical security and CBD displacement. CBD supervisors attempted to continue supervision by utilizing their networks to track down displaced CBDs and to assess the security situation prior to making supervisory visits. Several CBDs and community leaders validated this claim.

Health facility services

Health facilities were severely affected, with looting and damage to equipment and supplies. A Unity State MoH staff reported that “…many health centers lost all equipment and facilities and some were burnt down. Human resources were strained, many health centers lost staff as people fled the area.” Even facilities that remained opened or eventually reopened as part of emergency response experienced major stockouts. The pre-positioned drugs for the rainy season were also largely looted or destroyed.

Caregivers appeared less likely to complete referrals because of their concern that the referral facility would not be operating fully. A caregiver mentioned that some health facilities remained without drugs even after reopening and that “we find out by walking there that there are no drugs. That is the only way to be sure.” There were concerns expressed by caregivers of leaving other children behind in an insecure situation while taking the sick child to the facility.

Funding

Two major donors funded the consortium that IRC operated its iCCM program under, one of which also funded emergency relief efforts following the start of the conflict. IRC staff reported that at the onset of the crisis, one donor communicated hesitancy to continue funding large scale development-focused programming, such as iCCM, since resources could potentially be reprioritized to support emergency response interventions. Uncertainty of the funding stream caused the IRC to limit iCCM programmatic spending and rely heavily on unrestricted private funding to cover implementation costs. The security of iCCM funds was reestablished 2-3 months later. The potential for redistribution or cancelation of funds never came to fruition.

Staff from another implementing agency indicated that one donor “was a very flexible donor to context, allowing the program to continue and trusting the implementing partners’ judgement to get things done. There were no other donors [to the iCCM consortium] which helped in allowing this flexibility.” Another implementing agency indicated that “funding did not significantly change despite the increased program needs and overburdening in some cases. It didn’t stop, which allowed the program to continue, but it also did not increase given the increased demand.” Multiple donor representatives were contacted for interviews, but none were available.