The status of emergency obstetric and newborn care in post-conflict eastern DRC: a facility-level cross-sectional study

Background Pregnancy-related mortality remains persistently higher in post-conflict areas. Part of the blame lies with continued disruption to vital care provision, especially emergency obstetric and newborn care (EmONC). In such settings, assessment of EmONC is essential for data-driven interventions needed to reduce preventable maternal and neonatal mortality. In the North Kivu Province (NKP), the epicentre of armed conflict in eastern Democratic Republic of the Congo (DRC) between 2006 and 2013, the post-conflict status of EmONC is unknown. We assessed the availability, use, and quality of EmONC in 3 health zones (HZs) of the NKP to contribute to informed policy and programming in improving maternal and newborn health (MNH) in the region. Method A cross-sectional survey of all 42 public facilities designated to provide EmONC in 3 purposively selected HZs in the NKP (Goma, Karisimbi, and Rutshuru) was conducted in 2017. Interviews, reviews of maternity ward records, and observations were used to assess the accessibility, use, and quality of EmONC against WHO standards. Results Only three referral facilities (two faith-based facilities in Goma and the MSF-supported referral hospital of Rutshuru) met the criteria for comprehensive EmONC. None of the health centres qualified as basic EmONC, nor could they offer EmONC services 24 h, 7 days a week (24/7). The number of functioning EmONC per 500,000 population was 1.5. Assisted vaginal delivery was the least performed signal function, followed by parenteral administration of anticonvulsants, mainly due to policy restrictions and lack of demand. The 3 HZs fell short of WHO standards for the use and quality of EmONC. The met need for EmONC was very low and the direct obstetric case fatality rate exceeded the maximum acceptable level. However, the proportion the proportion of births by caesarean section in EmONC facilities was within acceptable range in the HZs of Goma and Rutshuru. Overall, the intrapartum and very early neonatal death rate was 1.5%. Conclusion This study provides grounds for the development of coordinated and evidence-based programming, involving local and external stakeholders, as part of the post-conflict effort to address maternal and neonatal morbidity and mortality in the NKP. Particular attention to basic EmONC is required, focusing on strengthening human resources, equipment, supply chains, and referral capacity, on the one hand, and on tackling residual insecurity that might hinder 24/7 staff availability, on the other.

vulnerability to adverse outcomes related to these complications [15,23].
The Inter-Agency Working Group for Reproductive Health in Crises (IAWG) guidelines place a priority on ensuring optimal provision of EmONC services during post-conflict recovery to address excess mortality in mothers and newborns [24]. A clear understanding of the capacity of the existing health system to respond to the reproductive health needs of affected populations is essential for effective program planning and implementation, especially with regard to safe motherhood [25]. For example, the IAWG advocates the use of a wellestablished Needs Assessment Toolkit [26] that allows for a system of EmONC process/performance evaluation against United Nations (UN) standards, which is instrumental in identifying gaps and guiding programmes.
From 2006 to 2013, state-based conflict (i.e., armed fighting between a state and an opposition group resulting in at least 25 battle-related deaths in a year) disproportionately affected the eastern Democratic Republic of the Congo (DRC) in general and the North Kivu Province (NKP) in particular [27]. During that period, the North Kivu Province witnessed 98 of the 128 conflict events that occurred across the country, opposing government troops against foreign-backed rebel groups [27]. As a reflection of the impact of conflict on public health, the east of DRC recorded higher preventable mortality, including excess neonatal mortality, than in the west [28,29]. In the North Kivu Province, maternal mortality was as high as 790 deaths per 100,000 live births for the first half of 2013 [30]. Recent evidence indicates that proximity to and deadliness and duration of armed conflict in SSA are associated with indirect effects on the survival of mothers and their babies; these effects persist for years after the conflict has ceased [15,31]. This underscores the need for sustained and informed measures in revitalizing essential health services, including EmONC. However, in the North Kivu Province, there is a lack of data on the post-conflict status of EmONC. This situation can be explained in part by poor government stewardship and 'the proliferation of fragmented humanitarian and recovery initiatives', factors previously observed in other post-conflict contexts in SSA (South Sudan and Sierra Leone) and eastern DRC [32][33][34][35]. These often lead to uncoordinated health interventions, with little attention to data-driven plans and policies that are customized to local contexts [36,37]. Therefore, this study aimed to contribute to filling this gap by assessing the availability, use and quality of EmONC in 3 Health Zones (HZs) in the North Kivu Province of eastern DRC. Evidence generated from this study will contribute to informed programming and data-driven interventions, involving local and external stakeholders, as part of the post-conflict effort to improve maternal and neonatal health in the Province.

Study design and setting
This is a cross-sectional survey at health facility level conducted from March to May 2017 in the North Kivu Province (NKP). From 2006 to 2013, this Province accounted for about two thirds of battle-related deaths across the country, reflecting the intensity of conflict events between government troops and two major rebel groups backed by neighbouring countries (the Congrès National pour la Défense du Peuple [CNDP] and the Mouvement du 23 Mars [M23]) [27]. The NKP counts 34 health zones (HZ) out of the 515 nationwide, serving urban, rural or urbano-rural administrative areas with at least 100,000 inhabitants and a maximum diameter of 150 km [38].
The HZ is a decentralized entity that constitutes the operational unit for planning and implementing health services in DRC. It operates as a tiered system consisting of a network of health facilities, supervised by the Central Bureau of the Health Zone (CBHZ), with increasing service capacity along a chain of referral [38]. Health centres (HC) are first-line facilities dedicated to primary health care tasks (including the provision of antenatal care and the management of normal deliveries), delegated to a team of versatile nurses. They refer cases that are beyond their means and competencies to a referral health centre (RHC) or the referral hospital (RH).
Three health zones (HZs), namely Goma, Karisimbi, and Rutshuru, were purposively selected based on their 'red flag' proximity-that is, within a 50 km range [15,31], to most of the deadliest conflict events from 2006 to 2013 [27,39], and based on their position across the urban/rural spectrum in the North-Kivu Province to represent the status of maternal and neonatal health care in urban (Goma), urbano-rural (Karisimbi) and rural (Rutshuru) post-conflict settings.
The HZs of Goma and Karisimbi integrate both stateowned and private not-for-profit (religious or other) referral facilities (RHs and RHCs) into the public health sector. Not-for-profit organizations (NFPO) have entered into a contractual agreement with the state specifying the rights and duties of both partners; however, they remain largely autonomous in terms of management.

Selection of health facilities and sample size
We selected only public health facilities (state-owned and private not-for-profit) providing maternal and neonatal health (MNH) care across the 3 HZs. The rationale for this was that institutional births represent 92% of deliveries in North Kivu province, of which 80% take place in public facilities [40]. A complete list of facilities with maternal services, including HCs, RHCs, and RHs, was provided by the respective CBHZs. Therefore, all 42 public facilities providing MNH services in the 3 HZs were included in the study. Table 2 shows these health facilities per HZ and per level. As per the DRC Ministry of Health (MOH) guidelines, all HCs and referral facilities (RHs and RHCs) are required to provide basic and comprehensive EmONC 24 h a day and 7 days a week, respectively.

Data collection
Data were obtained from medical records and registries for the year 2016, as well as from interviews with relevant in-charges and observation of available equipment and medical supplies, using the EmONC Needs Assessment (NA) toolkit. This toolkit is organised in modules of questionnaires developed and refined over time by the Averting Maternal Death and Disability (AMDD) programme at Columbia University Mailman School of Public Health. The WHO, UNFPA, and UNICEF have adopted these modules to capture key indicators of availability, use, and quality of EmONC services. These indicators are constructed to inform interventions aiming to reduce maternal and neonatal mortality by identifying gaps, monitoring implementation, and measuring progress [26]. Moreover, these modules have been proven useful in conflict-affected settings [41].
Prior to the survey, we held meetings with the Chief Medical Officer (CMO) and the CBHZ team of each HZ. During these meetings, the French versions of the AMDD NA modules were discussed and adapted to suit the local context, guided by the research objectives. Adaptations included editing/deleting some questions and response options. The CMOs granted authorizations to collect data at the surveyed facilities and provided relevant information on public health facilities (population covered, ownership/management).
Four research assistants, all final year students at the Midwifery Section of the Institut Supérieur des Technique Médicales de Goma, were recruited. They received 3-day training to ensure a clear understanding of the objectives/methods of the survey, the content of the modules and how to fill them out. Didactic sessions covered proper interviewer behaviour, relevant ethical considerations, and appropriate identification of direct obstetric complications (ante-and post-partum haemorrhage, severe pre-eclampsia/eclampsia, obstructed labour, ectopic pregnancy, and postpartum sepsis) using the WHO definitions [26].
In each HZ, the Principal Investigator (PI) organized a data collection schedule in consultation with the heads of health facilities. The PI supervised data collection, which followed a top-down order in every HZ (i.e., RHs first, then RHCs and HCs). Data on deliveries, pregnancy-related and childbirth complications, maternal, foetal and early neonatal outcomes, and procedures related to EmONC were extracted from facility registries, procedure reports, and from patient records when necessary. Interviews were conducted with heads of facilities (medical doctors in referral facilities and registered nurses in HCs) and in-charges of units of interest (maternity ward, delivery room, operation theatre, pharmacy, and laboratory), or with a direct collaborator. They focused on the provision of signal functions during the last 3 months, the number and composition of health staff, and the availability of equipment and medical supplies. Facility walk-throughs were carried out to observe the availability of medical supplies, facilities, and equipment.

Data management and analysis
During data collection, a number was attributed to each facility and written at the top of every page of each module, starting with RH, then RHC and HC, in Goma, Karisimbi, and Rutshuru, respectively. EpiData 3.1 was used to create data entry fields with in-build checks and each module had its own file. Data were double entered by 2 trained data entry operators and discordances were sorted out by revisiting the respective questionnaires.
Data keyed in using EpiData were exported to Stata 15 for data handling and analysis. Descriptive statistics including means and medians for continuous variables, and frequencies and proportions for categorical variables were performed as appropriate. Analyses were guided by the Handbook on Monitoring Emergency Obstetric Care

Profile of surveyed health facilities and volume of deliveries
Of the 42 public health facilities surveyed in the 3 HZs, 24 were health centres, all owned and managed by the state, and 18 were referral institutions, of which 5 were RHs and 13 were RHCs (

Provision of signal functions
Only 17% (3/18) of the referral facilities (the faith-based RH and 1 faith-based RHC in Goma, and the RH in Rutshuru) provided the nine signal functions within the last 3 months prior to the survey and thus met the criteria for functioning cEmONC facilities (see Table 3). All the remaining (83%; 15/18) were partial cEmONC facilities (i.e., 8 or fewer signal functions provided). Of these, 20% (3/15) were short of a signal function (AVD) to qualified as functioning cEmONC facilities, which included 1 faith-based RHC and the government RH in Goma, and 1 faith-based RHC in Karisimbi.
Further analyses showed that nearly all the facilities that reported 'no indication' for parenteral administration of anticonvulsants (95.0%; 19/20) didn't provide this signal function even during the past 12 months. Also, 62.5% (10/16) and 50.0% (4/8) of facilities that didn't provide neonatal resuscitation (NR) and AVD due to no indication, respectively, didn't have the necessary equipment to perform these signal functions (i.e., neonatal bag and mask for NR and vacuum extractor or forceps for AVD).

Indicators of EmONC
The existing public health facilities designated to provide EmONC services exceeded the minimum number of EmONC delivery points needed per 500,000 population in the 3 HZs (Table 4). However, none of the health zones achieved the minimum number of functioning EmONC facilities as per the WHO recommendations (at least 5 EmONC facilities for every 500,000 population with at least 1 cEmONC facility), with an unmet need less pronounced in the HZ of Goma (Tables 4 and 5). In other words, the number of functioning EmONC facilities per 500,000 population in the study HZs was 1.5.
Regarding the availability of essential drugs, equipment in the facilities surveyed, magnesium sulphate was the least available drug in the facilities surveyed (45.2%;19/ 42). Vacuum extractors were the least available equipment (21.4%;9/42), followed by filled oxygen cylinders and neonatal intravenous fluid sets (Table 8). These drugs and equipment were more frequently unavailable in HCs located in the HZs of Karisimbi and Rutshuru. There was a pattern whereby essential drugs and equipment for EmONC were more likely to be available in facilities located in the HZ of Goma. In the HZ of Rutshuru, 33% (2/6) of the referral facilities, and 36% (4/ 11) of the HCs did not have an operational laboratory because of a lack of equipment or technicians.

Transport and communication means for referral
The ten facilities with a functional ambulance (car or motorcycle) on-site were all referral facilities, including two thirds of referral facilities in Rutshuru (4/6) and half of the referral facilities both in Goma and Karismbi (Tale 7). They were all charging fees for the use of the ambulance when an institutional referral was requested, except for the referral hospital of Rutshuru where MSF took care of the ambulance operations. Relatedly, all the referral facilities in Goma and Karisimbi and 50% (3/6) of those in Rutshuru had a means of communication (cell phone or two-way radio) for referral purposes, in stark contrast with almost none of the HCs (4.1%, 1/24).

Discussion
According to the World Health Organization, the delivery of all EmONC services over the past 3 months by at least 5 facilities serving 500,000 inhabitants indicates that pregnancy-related complications are managed by trained personnel, medical equipment and supplies are sufficient, and skills are maintained [26]. Our findings show that only 30% of the minimum acceptable number EmONC facilities was met, with an availability gap less marked in the HZ of Goma than in the HZs of Rutshuru and Karisimbi, respectively. Relatedly, the process indicators related to the use and quality EmONC indicate that study HZs fell short of the WHO standards, except for the proportion of births by caesarean section. Resources needed for EmONC and institutional referral were scarce in surveyed HCs.
None of the surveyed HCs could offer EmONC 24 h a day and 7 days a week (24/7). It is conceivable that this finding is related to the impact of residual insecurity that characterize post-conflict settings [42]. This situation impedes patients' admission or referral within the health system and tend to limit the availability of skilled health personnel to provide EmONC 24/7, especially in primary healthcare facilities, as suggested in previous studies in post-conflict SSA [43][44][45]. A high priority needs to be  Expected number of women with MDOC was calculated multiplying 15% with expected number of births in Table 2 c Calculated as number of CS/expected number of births × 100 ‡ Standard to be determined NA: not applicable, _: not calculated placed on further addressing residual insecurity in the 3 HZs, which is paramount in reducing preventable mortality in conflict-affected populations [29]. Very few of the facilities surveyed performed all the 7 basic signal functions in the past 3 months, with AVD being the least frequently performed. This parallels the findings of a previous survey on the provision of key health services in a national sample of health facilities in DRC. It revealed that less than 12% of those offering maternity services were performing all basic EmONC services [46]. Also, the finding that AVD was the least likely signal function to be reported, especially in HCs, is consistent with existing literature on the subject in SSA [47,48].
If we were to ignore AVD, the 3 HZs could meet the target of at least 5 functioning EmONC facilities for 500, 000 people. While most facilities couldn't offer this signal function on policy grounds (i.e., HZ or facility   policies not allowing the performance of a signal function), half of the facilities that reported 'no indication' didn't have a vacuum extractor for instrumental vaginal delivery. Restrictive policies targeted mainly HCs regardless of their location, illustrating a perceived lack of trained operators at that level of care. Moreover, 50% of facilities that owned a vacuum extractor did not indicate any AVD, which might reflect unfamiliarity with this technique. Lack of equipment and limited staff training and exposure are common obstacles to the performance of AVD in resource-poor settings [49][50][51][52]. Similarly, restrictive policies and the lack of equipment/drugs were the main reasons why neonatal resuscitation and parenteral administration of anticonvulsants could not be performed.
Restrictive policies raise concerns as to training and practice opportunities for health providers, especially those working in HCs. These restrictions, along with a lack of equipment, compromise chances for health providers to acquire skills and contribute to maintaining a situation where targeted services remain virtually unavailable at the primary health care level [50]. In line with WHO guidelines [53], interventions prioritising competency-based training for frontline providers and the use of simple and cost-effective equipment have been proved effective [54,55]. Such programming could offer a practical alternative to restrictive policies and promote better access to quality primary health care, which is a key step toward post-conflict health recovery [43,56].
In Karisimbi and Rutshuru, none of the HCs administered anticonvulsants and, very few had stocked magnesium sulphate, the anticonvulsant of choice, shown to avert up to 85% of severe pre−/eclampsia (SPE) related deaths and disabilities [57]. This suggests that it was very likely for a pregnant woman presenting with SPE to be Table 8 Availability of  Communication means for referral ‡ 16 (38.1) 6 (100) 0 (0) 6 (1009 0 (0) 3 [44] 1 [ 9] RH: referral hospital, RHC: referral health centre, HC: health centre, IV: intravenous †Car or motorcycle ‡Cell phone or two-way radio referred without being administered the loading dose of magnesium sulphate (10 mg). In rural Bangladesh, Shamsuddin et al. found better maternal and perinatal prognoses in pregnant women who received the loading dose before referral than in those that did not [58]. Health facilities in Goma outperformed those in Karisimbi and Rutshuru in terms of service provision, overall. The performance of more EmONC signal functions in urban settings has already been reported in SSA [59] and in South-East Asia [60], a disparity that might illustrate an imbalance in the procurement of medical supplies and equipment, in training opportunities, and in the posting and retention of skilled health staff. Remarkably, two of the three functioning cEmONC facilities catered to the urban population, which might imply better access to comprehensive life-saving care for pregnant women in Goma. Of note, the presence of a functioning cEmONC facility in Rutshuru is consistent with a previous study in rural humanitarian settings in the North Kivu province of eastern DRC [61].
The inadequacies of the EmONC services found in state-run facilities in the study HZs are concordant with previous literature in DRC [61][62][63]. They appear to be systemic in nature and might reflect the level of public resource allocation for the health system. Available evidence points to this analysis. The DRC government expenditure on health (USD 12-13 per capita) is one the lowest in SSA [64], with a health budgetary allocation (4-5%) far below the 2001 Abuja declaration target of at least 15% of the national budget to be allocated to the health sector to ensure universal care coverage [65,66]. Also, studies in DRC showed other factors related to persistent fragility that create bottlenecks in the functioning of the health system, including dependence on user fees and fragmented/vertical multi-donor inputs, inefficient budget preparation approaches, unmonitored budget execution, and poor governance [67][68][69]. Previously identified barriers to the delivery of EmONC in post-conflict SSA included systemic and human resource factors such as limited infrastructures and procurements, insignificant and erratic pay and poor living and working conditions of health workers [44].
Our results show that about half of facility births in the HZs of Goma and Karisimbi took place in referral NFPOs and that these facilities hired a higher proportion of health staff in both HZs and performed more signal functions. This tends to reflect how the availability of enough resources translates into adequate functionality of health facilities and points to women's self-referral behaviour to seek maternal care where better quality is perceived, as shown in previous work [70][71][72]. A study by Tabatabai P et al. in southern Tanzania comparing faith-based organisations (FBO) and public hospitals found that maternal health service capacity was more appropriate in FBOs [73]. In addition, the number of EmONC functions performed during a given time period has been shown to have an effect (i.e, for each additional EmONC SF available, bypassing odds decrease substanatially) on bypassing frontline facilities for childbirth care [71,74]. However, bypassing frontline facilities has an incidence on the number of cases of deliveries and pregnancy-related complications primary care health workers can see and manage in a given time period, perpetuating low EmONC functionality [75].
In the rural HZ of Rutshuru, the RH attended over a third of deliveries and four fifths of CS and hired the majority of health workers across the HZ. These findings reflect the impact of humanitarian assistance through financial and technical support that this RH was receiving from MSF. For example, in a rural conflict-affected district in Afghanistan, Lagrou D et al. found a steady increase in caseload at a cEmONC facility run by MSF [76]. In Rutshuru, MSF offered user fees exemption for CS and a maternity-waiting home, measures proven to address inequities in childbirth care [77,78]. However, this assistance comes with potential downsides. First, by absorbing the majority of cases within a health district, humanitarian-assisted hospitals deprive other facilities, particularly primary care ones, of being exposed to case management, undermining the acquisition and maintenance of necessary skills. Also, it raises sustainability concerns, given the time-bound nature of its implementation [79].
The DOCFR exceeded the maximum acceptable level, suggesting poor quality of EmONC. Similar findings were reported in DRC and other SSA countries [50,62,63,[80][81][82]. However, given the observation that all direct obstetric deaths occurred only in the referral facilities, this indicator might also be reflective of the state of referral systems in the study HZs. Inadequate means of transportation and communication to link referring and receiving facilities may have contributed to the delayed arrival of unannounced referrals of women with MDOC at higher levels of EmONC, further compromising the prognosis. Maternal mortality audits in resourceconstrained settings have pointed to a number of barriers to timely access to EmONC, including inappropriate and ineffective referral, inefficient ambulance services, cost of transportation, and long waiting time before care is received at higher-level facilities [83,84].
The population-based CS rate was the only process indicator found to be within an acceptable range in 2 of our study HZs. This is in contrast with findings from an EmONC study in Lubumbashi, the second largest city in DRC, barely affected by the 2006-2013 conflict episodes, that revealed an unmet need for CS deliveries [63]. These CS rates in the study HZs are rather encouraging, reflecting the contribution of humanitarian actors and NFPOs to the provision of public maternal and newborn health care services in the region. Even in the HZ of Karisimbi where none of the facilities qualified as functioning EmONC facility, the population-based CS rate was within the limits set by the WHO.
Since only a little over one-sixth of deliveries occurred in functioning EmONC and given the shortcomings in referral capabilities and unmet minimum acceptable levels of use and quality, one can surmise that most pregnancy-related complications had likely received substandard or delayed care. One systematic review found an inverse correlation between the met need for EmONC and maternal mortality ratio [85], highlighting the impact of better access to quality EmONC on reducing preventable mortality, a key element in postconflict recovery. A study in Nigeria identified factors contributing to maternal mortality, including a dysfunctional referral system and limited intensive care capabilities causing delays in providing EmONC [84].
This study has some limitations. First, only three out of the 34 HZs in the North-Kivu Province were included in a purposive manner and, therefore, our findings and analyses cannot be generalized to the whole Province. The EmONC status in HZs located further to the 2006-2013 conflict events might be different. Second, profit-run health facilities were not selected, which very likely underestimated the provision and capacity of EmONC as well as facility deliveries in the 3 HZs. Our findings can also not be extrapolated to these facilities. Third, we did not assess staff knowledge of EmONC procedures and standards of care, leaving gaps in the actual ability to provide quality care in the study HZs and calling for further research. Fourth, we collected data in 2017 and this might raise concerns as to the current pertinence of our findings. Nevertheless, recent qualitative findings on health services for women, children, and adolescents in North Kivu, highlighting human resource and logistic barriers, might suggest that this study remains relevant [86].

Conclusion
This study gives for the first time a quantitative assessment of the post-conflict status of EmONC in the North-Kivu Province in eastern DRC. By identifying the gaps in the availability, use and quality of EmONC in the study HZs, it raises the prospect for evidence-based policies and programming, as well as coordinated EmONC interventions, which should encourage local and external stakeholders to work together in improving maternal and neonatal health in the province as part of the post-conflict recovery efforts. Despite the fact that most process/performance indicators didn't meet the WHO standards, EmONC services were provided to a degree that appeared to be more advantageous for urban populations and more comprehensive in public referral facilities, especially where the NFPO or humanitarian partners were involved. Special attention to basic EmONC and referral linkage among public health facilities is required, particularly in non-urban HZs, with an emphasis on strengthening human resources, equipment, supply chains, and referral capacity, on the one hand, and on tackling residual insecurity that might hinder 24/7 staff availability, on the other hand. In view of the marked inadequacy between the number of HCs (designated bEmONC points) and the provision of basic signal functions across the study HZs, initial efforts to improve access to quality EmONC could be directed towards upgrading the capacities of a few HCs (i.e., through supplying and equipping them appropriately, as well as through in-service training programmes), starting with those located in non-urban HZs. There is also the need to review the existing policies with respect to restrictions on the provision of EmONC signal functions in primary health care centres, particularly on AVD and neonatal resuscitation, provided appropriate planning and support. This study was conducted before the 2018 Ebola outbreak in North Kivu. The Ebola outbreak disrupted the provision of, and access to, essential health services, including EmONC [87]. Thus, our data can form a baseline for a follow-up survey to update and evaluate the status of EmONC service provision in these HZs.