Implementation Outcomes
Adoption:
The uptake of interventions from the programmatic perspective, otherwise known as adoption, was assessed within each case (Additional File 1) to understand key constructs including governance structures, policy adoption, implementation processes and implementation readiness.
In 4 of 5 cases, UN agencies had primary or leading roles in initiating, supporting, and coordinating implementation of MPDSR interventions. Implementation was mainly executed by international and local NGOs under UNFPA and UNHCR leadership, with involvement of WHO, United Nations Children’s Fund (UNICEF), UNRWA (Palestine), and International Organization for Migration (IOM; Cox’s Bazar), and support from multi-partner coordination groups in select cases. Recent MPDSR policy and guidelines were only in place in Uganda (27). MDSR pilots in Yemen were guided by 2013 Yemen Maternal Mortality Audit Guidelines (28), although limited in scope. UNHCR, WHO, and UNFPA technical guidelines were used for partner-specific interventions or to inform development of national procedures.
Implementation processes for facility-based surveillance were largely aligned with the WHO four-step MPDSR cycle (7). Community-based processes followed similar steps including notification of death by community health workers (CHW) or other community actors, verbal autopsy by a midwife or medical team 1–3 weeks later, and response activities. In particular, the community-based pregnancy surveillance system implemented by UNRWA in Palestine comprised of follow up by primary health care centers in the postpartum period, with subsequent investigations if a maternal death was reported.
Establishment of review committees occurred in each setting. All cases, with the exception of Yemen, have or were in the process of establishing a national or centralized committee tasked with coordinating implementation of the system, reviewing cases, and/or mounting responses to identified issues. Sub-national committees were also reported in Uganda, South Sudan, and Palestine. All but CXB reported the establishment of review committees or informal teams at the facility level.
Supporting data systems and tools varied widely by case and implementing partner, with parallel reporting systems identified in Uganda, South Sudan, and Palestine. In CXB, UNHCR tools were adapted and WHO’s Early Warning Alert and Response System (EWARS) was adopted for facility-based reporting of maternal deaths and community-based reporting of deaths of women of reproductive age. In Uganda, the MOH has integrated an MPDSR event tracker with active death notification and review forms into the District Health Information System 2 (DHIS2), which is further supported by a pilot of UNICEF and Centers for Disease Control and Prevention (CDC) monitoring tools. In addition, maternal and perinatal deaths were reported through the UNHCR system. In South Sudan, tools and reporting systems varied by implementing partner; however, parallel reporting to the MoH and partners often employed DHIS2, the Integrated Disease Surveillance and Response (IDSR) systems, UNHCR systems, and other partner-specific systems. In Palestine, each siloed intervention reported through their respective partner-specific system. In particular, UNRWA had implemented a sophisticated electronic health reporting and record system at primary health facilities with a patient smart phone app to support pregnancy surveillance and maternal death investigation and reported through their system and to the MoH. In Hadhramaut, Yemen, a customized electronic system was developed, however its use was discontinued due to funding shortages. In Sana’a, Yemen, UNFPA’s Reproductive Health (RH) logistic management system was linked to maternal death reporting in the absence of a national health information system.
Training and capacity building on MPDSR interventions was limited in most cases. Initial workshops were held at the launch of all MPDSR interventions; refresher training and/or intermittent offerings of capacity building were partner dependent. In Palestine and Uganda, on-the-job training and mentoring were also implemented by expert trainers within select health facilities to expand capacity of the health workforce.
Fidelity:
Implementation adherence to applicable guidelines, quality of reporting and review of deaths, and leadership and implementing actor responsiveness were assessed as signs of fidelity. In CXB, the MDSR system had strong community-based reporting of maternal deaths and had seen gradual improvement in the reporting of facility-based maternal deaths (i.e., more partners accepting to report deaths in their facilities through the system and increased sensitivity of surveillance at facilities). Nonetheless, notable delays ranging from days to months in conducting death reviews and verbal autopsies have been reported. Conversely, within the Ugandan refugee settlements, implementing partners reported timely notification and review of facility-based maternal deaths, yet experienced delays associated with community-based surveillance. Perinatal death notification and review was not consistently implemented and strongly partner dependent in Uganda. In South Sudan and Yemen, MPDSR fidelity was weak. Maternal and perinatal deaths were not reliably reported or reviewed with many established facility-, sub-national-, and national-level review committees having never convened in practice. However, the pilot in Hadhramaut, Yemen had good fidelity with facility-based reporting of maternal deaths and ad hoc implementation at the community level. In Palestine, MPDSR interventions had fairly high fidelity marked by a strong maternal and neonatal death notification and reporting system within the Ministry of Health and UNRWA health systems. Nonetheless, extensive delays, often months long, were reported in collection of case information, review, and analysis of maternal and neonatal deaths. National review committees were also not yet functional. Across all case contexts, fidelity to the “response and action” step of the MPDSR cycle was low; only Uganda had ongoing monitoring mechanisms in place at the national level to ensure coordination and accountability of the MPDSR system.
Overall, the quality of reporting and review of deaths was low across all sites. Incomplete data, poor documentation of patient care, and limited access to records at higher-level facilities was pervasive. Family members often did not want to provide supplementary information, which contributed to underreporting, misreporting, and high proportion of cases with unknown cause of death. In 3 of 5 cases (Uganda, South Sudan, and Palestine), misclassification of stillbirths and neonatal deaths was also a recognized challenge. Nonetheless, some partner-specific MPDSR interventions demonstrated higher quality implementation, namely UNHCR-led systems with comprehensive reporting and review of maternal and perinatal deaths in refugee camps (Uganda and South Sudan), and UNRWA’s family team approach to primary health care with comprehensive pregnancy surveillance and maternal death investigation of registered patients (Palestine).
Across all cases, overburdened and under-paid health providers were often reluctant to participate in the notification and/or review of mortality cases, especially when costs were incurred to travel to a health facility. In fact, review committee responsiveness (i.e., high attendance and fidelity to established meeting frequency) typically decreased as they became more decentralized, with the exception of the national MPDSR committee in Uganda. Facility-based review committee members were most active, yet frequency and participation in these committees varied by facility and their implementation partner(s). In CXB, Uganda, and select sites in South Sudan, CHWs also actively reported community-based deaths and served as the communities’ and families’ link to facility-based MPDSR processes.
Penetration
Penetration refers to the integration of MPDSR interventions within health systems. In humanitarian settings, constructs of penetration include the scale of implementation (e.g., facility and community coverage, representativeness of reporting, and the depth of the review process), its positionality within the health system, and interoperability with other surveillance and health information systems. The scale of MPDSR interventions within each case differed (Fig. 1) but underreporting of maternal and perinatal mortality occurred across all cases, albeit to varying degrees. Of note, perinatal mortality and community-based deaths (both maternal and perinatal) were more likely to be both underreported and not reviewed compared with facility-based maternal deaths. The implementation scale was largest in Uganda, where facility-based maternal death reporting and review was well-established and many communities reported and conducted maternal death verbal autopsies, but neonatal death and stillbirth reporting and review only occurred in some facilities and communities. In comparison, the scale of implementation in South Sudan was considerably smaller than in the other four settings: only some facilities were covered by the various reported MPDSR interventions, and limited maternal and no perinatal verbal autopsies were being conducted. In Palestine, where strong national MDSR and NDSR reporting systems were established, most facility-based deaths were reported; however, many Safe Motherhood Emergency Centers located in vulnerable areas in West Bank had not adopted any MPDSR interventions due to limited resources.
While different approaches were undertaken to adopt and scale MPDSR, stakeholders in CXB, Uganda, and some partners in South Sudan aligned or situated implementation of MPDSR interventions within quality improvement (QI) initiatives. Positionality within QI was reported to improve buy-in of MPDSR, increase implementing actor responsiveness, and/or provide inherent accountability mechanisms to mount appropriate responses to identified recommendations to improve care. Alignment with other systems was limited, with the exception of successful integration of humanitarian-led interventions within the national government-led MPDSR system in Uganda. MDSR in CXB had no linkage with Bangladesh’s national MPDSR system, and established national systems did not exist in South Sudan and Yemen. Palestine was also marked by the absence of a harmonized system; attempts to integrate MPDSR and related death review interventions led by the public health system, private sector, NGO sector, and UNRWA under one unified system had yet to be fruitful. Interoperability with other surveillance and health information systems was also rare across the five case studies. Only in CXB and Uganda were MPDSR interventions linked to or integrated with other surveillance systems; WHO’s EWARS and the Integrated Disease Surveillance and Response (IDSR) systems, respectively. In Palestine and Yemen, mortality reporting through the Civil Registration and Vital Statistics systems were noted as opportunities for integration, albeit no established linkages had been reported to date.
Sustainability
The extent to which MPDSR interventions are institutionalized within a health system or humanitarian programming was conceptualized into three primary constructs: sustained funding streams, local ownership of MPDSR interventions, and institutionalized capacity. While assessment of sustainability is most salient for contexts in mid- to large-scale implementation stages, related themes were identified across all cases.
Resources for sustained implementation were not available in any case. While refugee camps and settlements had comparatively more financial and human resources for MPDSR interventions compared to other humanitarian contexts and UNHCR-led programs were able to sustain programming for longer periods of time, all partners reported funding limitations, without dedicated resources for training, capacity building, and response to review findings. MPDSR interventions also did not escape the stark health workforce realities in humanitarian settings; severe shortages and high attrition of health workers, in particular skilled birth attendants, have challenged efforts to achieve institutionalized capacity to implement MPDSR interventions. Across all cases, such implementation (and much of the health sector programming) was dependent upon external humanitarian support and assistance. Given this strong reliance on external support, local ownership of MPDSR interventions was rare, with the exception of MDSR and NDSR systems led by the Palestinian Ministry of Health.
Cross-case synthesis of implementation complexities
Across the five cases, implementation of MPDSR interventions was often affected by complex implementation climate and system dynamics—characterized by an environment of variable prioritization, buy-in, engagement, and trust from actors at all levels of the system. In most cases, stakeholder prioritization of MPDSR was perceived to be very low given the numerous competing health priorities in these resource-starved contexts.
“We have a silent killer where the surveillance system has not been customized to document the trends or the cases of deaths as they occur… Prioritization is a problem, and we are all to blame. Everyone needs to do their best to ensure that this comes up on the health agenda.” – South Sudan key informant
Across the cases, the level of leadership engagement in MPDSR also varied. In Palestine and Uganda, strong leadership from the MOH was bolstered by collaboration with UN agencies, INGOs, and local partners to improve implementation. In CXB, implementation of MDSR was first confronted with resistance by humanitarian agencies working within the camps, but engagement and coordination by the various UN-led working groups had strengthened implementation within the complex web of implementing partners. Government support to MPDSR interventions in South Sudan and Yemen has only been reported within the past few years, so leadership has been largely dependent on UNFPA and WHO to sensitize key stakeholders on the value of MPDSR and renew efforts via pilot MPDSR programming.
Regardless of the context, external influence from donors or bi-lateral agencies has been the reported impetus for MPDSR implementation in the study cases, giving legitimacy to the issue and nudging governmental authorities and local stakeholders towards buy-in and support. As a result, the MOHs in many cases were actively developing and/or renewing MPDSR policies, guidance, curricula, and implementation plans. However, partners across the cases expressed disquietude in the conflicting donor priorities and metrics for success – namely, the need to demonstrate active implementation of MPDSR via increased reporting and review of maternal and perinatal deaths to their donors, which simultaneously denoted higher maternal and perinatal mortality, an indicator of poor performance of their health service programming. Reliant upon short-term funding that generally did not allow for long-term sustained systems improvements, many partners, especially local organizations, admittedly did not fully implement MPDSR interventions for this reason. In addition, UN agency influence and mandate often drove MPDSR implementation in often conflicting or disparate approaches.
It depends on which agencies are leading, if WHO is leading, they will be asking only ‘What is the maternal mortality ratio?’ because they want to bring the data. But if you ask UNFPA, they will be looking into the actions, ‘What are the actions? What are the recommendations?’ because they want to improve the services there. If you ask UNHCR, they will be only talking about the third delay, quality of care, quality of care, quality of care. So, it’s very contextualized…
Dynamics between implementing actors and perceived implications of the MPDSR intervention also interplayed within the implementation climate. In all contexts, blame culture was pervasive; health providers feared getting blamed, shamed, fined, and/or fired due to their involvement in a maternal or perinatal death case. South Sudan and Uganda, in particular, have were reported to have had long histories of politicizing maternal deaths and criminalizing health providers for perceived negligence. In South Sudan, ethnic discord also exacerbated the mistrust in the health system. On numerous occasions, family or community members have been reported to threaten and/or harm providers as a means of retribution for their loss.
“Sometimes people, if they hear that their relative or a death is because of negligence or delay from the health facility, they will attack and carryout revenge killing in the health facility… A mother … was brought to the hospital due to ruptured uterus…[and] needed a blood transfusion, but there was no blood to transfuse, so the mother passed away. The husband, who was a soldier came with a gun and shot staff. He killed two health workers and injured three. So, in most cases health workers are afraid to give accurate information on what happened to defend themselves from all sides.” – Key informant in South Sudan
In addition, power dynamics between facility personnel often compelled the omission of facts and secrecy of events, especially when provider errors and mismanagement of a patient occurred. Breached confidentiality during death reviews further fueled mistrust between health providers and death review committee members and bred blame within communities. Buy-in from facility administration was essential to combat blame and mistrust and reorient the implementation climate; they were often gatekeepers for MPDSR with power to set the tone (i.e., supportive learning environment for QI vs punitive process) and ensure consistent participation.
Dynamics between communities and health and humanitarian actors were often sensitive, according to participants across the five cases. When community members, and in particular refugees, relied on humanitarian aid, many would not report deaths due to fears of ration reductions or negative consequences impacting access to services.
“There are sometimes when a community has concealed from reporting perinatal deaths because of the benefits it has, for example when a child is born into a family, they get registered into the family and so it increases the family size and therefore an increase in the amount of food rations they are entitled to from UNHCR or from the UN, and so, if you declare that probably someone has passed away, it kind of communicates that the family size has reduced, and therefore the benefits in terms of the food rations and the supplies they get will also be a bit lower.” – Key informant in Uganda
Yet, conversely, when strong community-based services were established enabling frequent interactions with the health system, community trust had been reported to bolster and improve functionality of community-based surveillance (reporting, participation in facility death reviews, and verbal autopsies). For example, in South Sudan, some partners were supporting community-based health management committees, which served as intermediaries between the health system and community; they encouraged community member participation in reviews and verbal autopsies, and functioned as feedback mechanisms to share issues and recommendations identified through QI and MPDSR efforts. In contexts with extensive community reach and/or robust CHW programs, community-based surveillance strategies have also been employed as a validation mechanism for facility-based surveillance, effectively identifying deaths that may not have been reported by health facilities.
Overall, the complexity of fragmented health systems and programming limited the communication, functionality, and reach of MPDSR interventions. Disruptions were reported due to obstacles accessing information about death cases referred in or out of a system’s jurisdiction, or migration of patients across administrative or service catchment areas. These challenges undermined cross-partner collaboration and coordination due to inherent political dynamics of claiming and blaming responsibility and accountability of deaths.
“Here now it is about almost impossible [to get data on women transferred out of the camp]. Last year we had one attempt where we talked to the national MPDSR committee of the facilities. But we couldn't go any farther because the thing I just mentioned, if you’re from a prestigious INGO or from the UN agency or any with a good reputation, then they give you some help, otherwise they don't. And as we're working for implementation partners… it is really very difficult for us to reach them and collect information from them.” – Key informant in CXB
“So sometimes women come to receive services from the Ministry of Health, and they received services from another clinic, or from an UNRWA clinic… There is no coordinated system between national providers. And this is one of the main challenges that we are facing in Palestine. If she received services at a UNRWA clinic, her file will remain there with all the information. So, these multiple providers is a main problem in reporting maternal cases.” – Key informant in Palestine