Setting up child health and mortality prevention surveillance in Ethiopia

Background: Mortality rates for children under five years of age, and stillbirth risks, remain high in parts of sub-Saharan Africa and South Asia. The Child Health and Mortality Prevention Surveillance (CHAMPS) network aims to ascertain causes of child death in high child mortality settings (>50 deaths/1000 live-births). We aimed to develop a “greenfield” site for CHAMPS, based in Harar and Kersa, in Eastern Ethiopia. This very high mortality setting (>100 deaths/1000 live-births in Kersa) had limited previous surveillance capacity, weak infrastructure and political instability. Here we describe site


Introduction
Child survival has improved in recent years 1 . However, great inequalities remain worldwide and child mortality is disproportionately high in sub-Saharan Africa and South Asia. Causes attributed to deaths are often non-specific and, despite large studies [2][3][4] , the infectious causes of severe disease and death have not been fully established. To accelerate progress, we need a comprehensive understanding of cause-specific mortality to direct public health interventions. The Child Health and Mortality Prevention Surveillance (CHAMPS) network was initiated to describe the causes of child mortality (under-5 years) and stillbirth, in sites in sub-Saharan Africa and South Asia through population-based surveillance, clinical data, verbal autopsy, and advanced post-mortem diagnostic approaches, including minimally invasive tissue sampling (MITS) 5 . Based on more comprehensive data, expert panels are able to determine the most likely underlying, co-morbid and immediate conditions leading to each death. Implementing this requires expertise and infrastructure to support community engagement, rapid notification of deaths, and detailed pathological investigation.
Ethiopia is the most populous landlocked country in the world, with a population of 108,000,000 (July 2018) 6 . Despite reductions since the conception of the study, overall under five child mortality is still 55/1000 live births 7 , representing 181,250 deaths in 2019, and Ethiopia had the fifth highest number of stillbirths in the world (97,000) in 2016 8,9 . Given the burden of mortality, including Ethiopia within the CHAMPS network provided an important opportunity to better understand child deaths. Emory University, which leads the CHAMPS network, requested applications for potential sites across sub-Saharan African and Asian countries. However, there was no location with existing capacity for all CHAMPS activities in Ethiopia. A CHAMPS presence would depend on creating capacity at a new or so-called "greenfield" site. The London School of Hygiene & Tropical Medicine (LSHTM) was commissioned by the Bill & Melinda Gates Foundation (BMGF) to investigate options, following a request from Emory University.
A small team from LSHTM visited Ethiopia in June and August 2015, engaging with national stakeholders, to assess support for the initiative and explore potential partnerships. We also visited universities and their demographic surveillance systems (DSS), selected based on their mortality data; we aimed to visit DSSs with an under-5 child mortality rate (U5MR) of >50/1000 live births in the 2-5 years prior to the visit 10 .
From these visits, we reported information on five potential locations back to Emory University and BMGF, comparing our findings against specific selection criteria. We also included considerations shared by national stakeholders; for example, representatives of the Federal Ministry of Health suggested recently established DSS sites could offer more opportunities for CHAMPS to study in a new area which had not included people studied for some time. Amongst the DSSs visited, only Kersa had a sufficient population size (>100,000 with >3000 births/year and >12,000 children under-5 years) to be included in CHAMPS. An additional advantage was the potential of including Harar DSS, an area of urban surveillance encompassing Hiwot Fana Specialized University Hospital (HFSUH), the regional referral hospital in Harar. Kersa was an emerging DSS which had been set up and expanded more recently, and demonstrated a well-organised DSS platform with readily available data. Emory University and BMGF reviewed the report and recommendations and a representative from BMGF visited the proposed site. Finally, Kersa/Harar was selected and confirmed as the most suitable location. We describe the initiation and development of the CHAMPS site in Ethiopia below.

Site specifics
Country. Ethiopia is landlocked covering 1,104,300 km 2 in the horn of Africa, with borders to Eritrea, Somalia, Kenya, South Sudan, and Sudan. It has ten regional states and two selfgoverning administrative cities, Adis Abeba (Addis Ababa) and Dire Dawa 11 . Ethiopia was ruled by a Monarchy until 1974 when the Derg (military council) deposed Emperor Haile Selassie. The Ethiopian People's Revolutionary Democratic Front took power in 1991. In recent times there has been intermittent instability, with states of emergency declared between 2016 and 2019. Despite this, Ethiopia has the fastest growing economy in the region (averaging 9.9% per year from 2007/08 to 2017/18), attributed to increases in construction and services. However, the economy is predominantly agricultural, with a per capita income of $790 and a quarter of the population living in poverty (2016) 6 . Ethiopia is multi-cultural and multi-ethnic; the major religions are the Ethiopian Orthodox church (44%) and Islam (34%) 11 .
Site general characteristics. CHAMPS in Ethiopia is based in Harar (urban) and Kersa (rural). Harar is on a hilltop (1885m) around 500km from Addis Ababa. Harar is the capital of East Hararghe, a zone of Oromia Region, and the capital of the Harari Peoples Regional state, a small independent city region. Kersa is a rural district around 44km from Harar. The district population is around 173,000 (2007 census), of whom 6.9% live in towns; the population density is low (372 people/km 2 ) 10 . Kersa contains highlands and lowlands, and the area is semi-arid and prone to drought, with scarcity of food production. In rural areas water is sourced from springs and wells; there is no supply of electricity. The towns have water mains and are supplied with electricity, but both are frequently interrupted, often up to weeks at a time. There is farming in two seasons --Kiremt or Meher (June to August) and Belg (Septemberto November)--of wheat, barley and vegetables in the highlands; sorghum, maize and potatoes are farmed in the lowlands. Khat, a stimulant, is the dominant cash crop.

Demographic Surveillance. Kersa Health and Demographic
Surveillance System (HDSS) was established in September 2007 10 , including 12 Kebeles (10 rural and two within small towns), the smallest administrative units in Ethiopia (Figure 1). In 2008, the HDSS catchment covered 50,607 person-years; since then 12 Kebeles have been added, for a total of 136,505 person years in 2019. Under-5 child mortality is high at 58-170/1000 live births (Table 1). Harar HDSS was set

Main diseases endemic in the area, according to previous surveillance or research.
Despite high mortality in Kersa, there were no prior comprehensive studies of the causes of severe illness in children either in hospital or in the community 12 , and cause of death data were from verbal autopsy alone. Verbal autopsy data suggested that severe malnutrition, gastro-intestinal disease and acute lower respiratory tract infections were the leading causes of deaths in children (5-14 years, 2008-13) 13 , and sepsis, birth asphyxia and prematurity were leading causes in neonates (0-27 days) 12 . Malnutrition, diarrhoea and outbreaks of cholera and vaccine preventable diseases, such as measles, are common. A mini demographic and health survey for Ethiopia estimated 19% of children (12-23 months) had not had any vaccines in 2019 7 .

Research capacity in the site
Existing capacity Haramaya University (HU) was set up in 1954 as a University of Agriculture in a large rural campus between Harar and Kersa 10 . HU now has a range of education programmes including College of Health and Medical Sciences campus in Harar, established in 1996. Academic staff from HU and Hiwot Fana Specialized University Hospital (HFSUH) contribute to CHAMPS from a range of disciplines, including demography, social science, microbiology, pathology, paediatrics and obstetrics.
Ethiopia has primary, secondary and tertiary levels of health care. Primary care includes health posts and health centres, with each health post serving 3000-5000 people 14 . Two health extension workers are assigned to each health post to link into the community. Around five health posts work with a health centre, which serves around 25,000 people 14 . Within Kersa HDSS there are currently 19 health posts and six health centres. Secondary care is provided by general hospitals; for Kersa, these are  Despite difficulties in attracting staff to an unknown project in a remote Ethiopian setting, we recruited a small number of international researchers to provide support in social science, paediatrics and microbiology. In addition, we employed national staff through HU. One of the challenges in this was the reluctance of skilled scientists to relocate to Harar. We developed project management support between LSHTM and HU, and established basic operations for communications, finance, human resource management and operational procedures for communication, procurement, transport, and security. These processes complemented and extended policies and procedures of HU and LSHTM. In our first year of initiation (2017-18) the site team grew rapidly to around 40 staff.
In parallel, we developed infrastructure at the site. At HU's College of Medical and Health Sciences, we restructured the hospital's pathology and university's microbiology laboratories; we also installed three septic tanks, a generator, two large water tanks and built an incinerator. We followed this with internal renovation and equipment installation, and we negotiated service contracts for five years where possible. We developed CHAMPS-specific specimen workflows and trained staff to ensure samples would be safely received, appropriately processed and securely stored. In the hospital we renovated rooms to perform MITS procedures and counsel families. We overhauled the mortuary to provide safe disposal of water and refrigeration to store the bodies of the deceased children and neonates. In addition, we arranged for repair and maintenance of basic clinical equipment such as patient scales or incubators to support inpatient care. We also strengthened collection and recording of routine clinical data, specifically in maternity and paediatrics. For our project team, we created office space through division of classrooms not in use at that time. We started to solve communication problems, which included intermittent and low-strength internet connection, and limited facilities for electronic data collection, data management and storage. We worked on these with the IT lead at the College of Health and Medical Sciences with additional expertise and advice from KEMRI-Wellcome Trust Research Programme, Kilifi, Kenya. We initiated infrastructure development and national applications for a satellite licence with a view to establishing a secure server room and a back-up satellite connection, as well as project-specific systems to support data collection, management and analysis which are now in place.
In Kersa HDSS we recruited staff and developed facilities. We trained two nurses in Kersa Health Centre on CHAMPS and identified ~200 community members drawn from community extension workers, religious leaders, traditional birth attendants, local militia, "Qeerroo" (youth representing the Oromo movement in Ethiopia), other community leaders and HDSS data collectors or supervisors willing to be trained and support timely reporting of deaths. At Kersa Health Centre we built additional rooms for counselling, refrigeration of bodies and MITS. In due course similar facilities will be developed in a health centre in Water (also in Kersa District) and in Haramaya district Hospital.

Description of CHAMPS methods.
We investigated child deaths using: (i) social science, to investigate community acceptability, practicality and feasibility of CHAMPS and its methods 15,16 , (ii) clinical surveillance, to support data capture, mortality notification and MITS, (iii) pathology, for histological assessment [17][18][19] , (iv) microbiology 20 to test samples using conventional microbiology, GenXpert and TaqMan Array Cards, and (v) verbal autopsy 5,15-24 . These data were analysed by an expert Determination of Cause of Death "DeCoDe" panel who assign a cause of death. Results will inform strategies to improve public health through a data-to-action plan 21 .
Social science work started first. We had recruited ten people to the social science team by May 2017. Following training, we undertook a Participatory Inquiry into Community Knowledge of Child Health and Mortality Prevention Surveillance (PICK-CHAMPS, July-August 2017). This included 20 workshops in Kersa and Harar (437 participants) and aimed to: (i) develop a social history of the community and to hear about the ways in which the community had successfully met earlier challenges; (2) understand community perceptions, beliefs, and practices related to child death, CHAMPS activities, and the MITS procedure; and (3) build relationships and identify common ground with the community through ongoing, two-way communication. From this we developed a programme of community engagement, alongside formative research 16 . This helped to overcome fears about MITS in the community, for example beliefs that the procedure would lead to mutilation of the body or unacceptable delays in burial.
For mortality and clinical surveillance, we recruited doctors, nurses, public health officers, health assistants, field workers (FWs), and laboratory technicians, to set up and support structures for clinical data capture and mortality notification. These processes were initially developed in HFSUH in Harar, starting in the neonatal and maternity ward (in February 2019) with extension to the paediatric ward (in May 2019). The mortality surveillance established in HFSUH included a death notification system with initial notification of deaths by hospital staff to a central call centre. Following notification CHAMPS staff conducted an eligibility assessment, requested informed written consent from the legal guardians of the child, conducted tissue sampling via MITS, transport the body, and arrange follow-up (including verbal autopsy). Families received support and counselling from trained counsellors throughout the process. In Kersa (which started death notification and MITS in February 2019), mortality surveillance was conducted slightly differently. Death notification and eligibility assessment were undertaken by health care workers in Kersa Health Centre or by notifiers in the community, usually by telephone. Counselling and obtaining informed consent was led by a counsellor, together with members of the community MITS team, with community members usually joining this process. After obtaining formal written consent, and with consensus, the body was moved to Kersa for MITS, and the specimens taken to Harar for testing. Local pathways were developed for sample transport, sample processing in laboratories, and feedback of results, including notification of notifiable diseases to relevant authorities. The first DeCoDe panel 21 , set up to determine the cause of death with multidisciplinary expertise, was held on the 18-19 th September 2019.
Results from the DeCoDe panel aim to inform public health action -"data to action plans" which have been developed in partnership with the Ethiopian Public Health Institute and operate at subnational, national and international levels. The subnational plan includes sharing of results with families, the hospital, and community health care workers. To support policy-making at national level, we share aggregate data with the Ethiopian Public Health Institute and we plan to share non-aggregated data in the future. Internationally, data from Ethiopia are submitted in real time to the CHAMPS network database, which includes data from all sites.

Results
By February 2020 we had received 1173 unique death notifications (4 th February 2019 -3 rd February 2020),1043/1173 (89%) from HFSUH and 880/1173 (75%) from outside of the HDSS catchment area. In the first year of surveillance, families for 59/99 of eligible cases gave consent for MITS (60%) and were included. These included 24 stillbirths, 21 neonates and 14 infants or children (>28 days to <5 years); 35 were from Harar and 24 from Kersa. Among the Kersa deaths, 14 died in the community and 10 died within a health facility. After assignment of cause of death from our expert panel, information was fed back to families, the hospital and community health care workers.

Early learnings from the site
We successfully implemented CHAMPS at a greenfield site, overcoming considerable challenges, and this process generated lessons for future greenfield site development. We based our assessment of potential sites largely on specific scientific criteria, and logistics in terms of travel and communications. We gave less consideration to security and staff living conditions, which, whilst they may not have ultimately influenced the choice of site, have presented us with considerable difficulties. A national state of emergency was declared repeatedly throughout 2016-2018 whilst we were initiating the programme. At these times, it was not always possible to travel or communicate effectively. Recruitment and retention challenges were thus considerably increased.
We recognised when initiating CHAMPS that success would need engagement and collaboration at all levels. Early engagement with national stakeholders, sustained through our scientific advisory committee, has given us political guidance and national support, helping us work at this level. For example, with religious community members, we approached the Ethiopian Islamic Affairs Supreme Council for a fatwah (a ruling on Islamic law) to conduct MITS on members of the Muslim community, which after careful consideration was granted and gave us permission to proceed with MITS activities in the study sites. Through our Community Advisory Boards (CABs) we have learned of, and addressed, specific issues in the community around body preparation and burial, providing opportunities for families to ensure religious practices are followed. To broaden engagement, we have broadcast a regular radio program, and invited community members to visit the research programme.
The large programme of community engagement resulted around in just under two thirds of those eligible providing consent for MITS and within this we recognised that consent was lower in Kersa than in Harar, potentially due to the need to transport bodies from Kersa to Harar for MITS at the start of recruitment. To mitigate this, we developed a facility to undertake MITS in Kersa Health Centre.
Substantial challenges remain. There is an understandable view from the community, that research should involve care for living children and those who are ill, rather than simply investigating those who have died. As a greenfield site, CHAMPS has been the focus of almost all our research work, and although study members have given support to the community where possible, for example in providing COVID testing, and giving clinical advice and health education sessions, we have not yet been able to initiate a significant research programme that investigates sick children, not just deceased children. This issue is an ongoing and major concern, and tackling it, by broadening the research work and securing funding to do this, is also vital for the long-term sustainability of the CHAMPS surveillance at the site.

Conclusions
CHAMPS is a complex multi-faceted project where each component must be executed to high standards to accomplish the whole. Achieving this goal in a new setting, at facilities with modest prior research experience and limited laboratory capacity, has been challenging. It has been made possible by the sustained investment of the funder, collaboration with the CHAMPS network, the determination of the investigating team, and the support of local, regional and national partners in Ethiopia. The project is now at a critical juncture. To sustain momentum, we need to develop young Ethiopian scientists who can carry the project forward over its anticipated 20-year life-span. We also need to broaden the research programme beyond post-mortem studies to tackle health research questions of local priority. This is essential to build understanding and acceptance among the community with whom we work and to attract and retain researchers who can generate the culture of scientific excellence required to sustain CHAMPS.
To tackle these challenges, we have established an overarching entity, the "Hararghe Health Research Partnership" which combines commitments from HU and LSHTM to generate a locally-relevant, sustainable programme of research centered around, but not restricted to, the CHAMPS project. The initial focus of this entity will be to improve health and reduce mortality in children and their mothers in a setting where health and child survival are currently poor. However, the underlying rationale for this entity is to achieve and sustain the ambitious goals of CHAMPS, through a self-sustaining research centre in an area of high disease burden that is able to address health issues of relevance locally, nationally and internationally.

Data availability
Underlying data Summarized data are publicly available through the CHAMPS website https://champshealth.org/data/enrolled-populationsummary/. Requests for further detailed data, for research and evaluation purposes, can be made through https://champshealth. org/data/. This paper is an overview of the background and start of important research in eastern Ethiopia. The project is called the Child Health and Mortality Prevention Surveillance (CHAMPS) network and aims to ascertain causes of child death in the high child mortality setting. The study is a collaboration between the Haramaya University and the London School of Hygiene and Tropical Medicine and builds on the previously established Kersa demographic surveillance site, which has also been expanded to include urban areas.

Open Peer Review
I have read and reviewed several of the publications from this demographic site during the past years. A weakness I recently found, is that the number of events such as births and deaths occurring over the years differ in the various publications and submitted papers I have read or reviewed. I, therefore, question the validity of the database of the actual demographic site that they use. To get sound data, this demographic site should authenticate their database so that they don't produce conflicting results. Similar problems have in the past occurred with other demographic sites in Ethiopia, and they needed to do comprehensive audits to establish the most probable and valid data.
The authors explain why they selected the studies site. The size of the study population seems to have been a factor in this decision. Nevertheless, a weakness of their study site selection is described in this paper as they were not allowed to do autopsies on many of the children who died. The authors ascribe this to the society being predominantly Muslim, and they also describe ways that they try to mitigate this challenge. However, the Harar area is an old and traditional society following strict Muslim faith, and the organisers of this study should have known that this would bring about a selection bias of their study, which makes the findings from this particular area not generalisable to other parts of Ethiopia. But it would be relevant for the eastern parts of the country.
The aim of the project is to develop a programme for monitoring causes of child deaths in Ethiopia. The project plans to last at least 20 years. It is, therefore, surprising that the study doesn't highlight the need to have Ethiopian scientists as leaders of this project right from the start. I mention this controversial issue because the paper states that the implementation was delayed because the project was not capable of recruiting adequately trained personnel to do the tasks. Having many years of working experience in Ethiopia and having evaluated many foreign supported health and research projects, it is important to recognize that sustainability in Ethiopia can only be achieved if the leadership and the management of the project are done by the Ethiopians themselves. In my experience, it is possible to get trained personnel from the country and regions or educate personnel to do necessary tasks. Expatriate personnel can serve in supportive functions.
The programme seemed to have had delays and interruptions due to the social unrest that has occurred in the country the past four to five years. In my experience over the last 45 years, it is most important that projects are not dependent on the presence of expatriates to run projects. I have not experienced projects stop more than a few weeks because of social unrest or civil war when the leadership is national. And -that is another reason to re-evaluate the ownership and leadership model of this project.

If applicable, is the statistical analysis and its interpretation appropriate? Not applicable
Are all the source data underlying the results available to ensure full reproducibility? Partly Are the conclusions drawn adequately supported by the results? Partly