Neonatal Mortality Burden and Trends in UNHCR Refugee Camps, 2006-2017: A Retrospective Analysis

Hannah Tappis (  hannah.tappis@jhu.edu ) Johns Hopkins University https://orcid.org/0000-0002-4289-5418 Marwa Ramadan Johns Hopkins University Center for Health Security Josep Vargas United Nations High Commissioner for Refugees Vincent Kahi United Nations High Commissioner for Refugees Heiko Hering United Nations High Commissioner for Refugees Catrin Schulte-Hillen United Nations High Commissioner for Refugees Paul Spiegel Johns Hopkins Center for Humanitarian Health


Background
The day of birth and rst month of life are the riskiest periods for child survival. Deaths among children aged 1 month to 5 years old have declined dramatically in recent decades, but too little progress has been made reducing the preventable deaths of newborns -which accounted for 46% of all under-5 deaths in 2017. 1 Recent estimates suggest that many as 7,000 newborns die every day, and 2.6 million stillbirths occur every year. 1 Substantial evidence has shown that most of these deaths can be prevented. 2 The Sustainable Development Goals established at the 2015 United Nations General Assembly include explicit targets for reducing newborn mortality to as low as 12 per 1000 live births in every country by 2030. 3,4 The United Nations (UN) Secretary General's Global Strategy for Women, Children, and Adolescents' Health 2016-2030 and global networks, including the Every Newborn Action Plan movement and Partnership for Maternal, Newborn and Child Health, recognize that these goals will not be met without urgent action in fragile and humanitarian settings. 5,6 Many countries with the highest neonatal mortality rates are currently or have recently been affected by complex humanitarian emergencies. 7,8 Although the widely referenced 2018 Interagency Field Manual for Reproductive Health in Humanitarian Crises explicitly highlights the importance of focusing on the maternal-newborn dyad, and the Newborn Health in Humanitarian Settings Field Guide summarizes existing evidence-based guidelines for interventions aimed at reducing neonatal mortality, limited information exists on the implementation or outcomes of these interventions in humanitarian contexts. 9,10 Humanitarian contexts are diverse, with varying population dynamics, burden of disease, economic opportunities, and health system capacity. Across the globe, the number of individuals who have been forcibly displaced as a result of con ict, persecution or human rights violations is at an all-time high. The United Nations High Commissioner for Refugees (UNHCR) estimates that by the end of 2018, there were 70.8 million forcibly displaced people including 41.3 million people displaced within the borders of their own county, 25.9 million refugees displaced across international borders, and 3.5 million registered asylum seekers. 11 Nearly four out of ve refugees live in countries neighboring their country of origin, and approximately three out of ve refugees live in urban areas. 11 Globally, the majority of refugees live outside of camps. However, the living conditions and policies governing refugee accommodation, movement, livelihoods, and access to services vary substantially from country to country. There are millions of refugees who continue to live in camps, and host governments continue to create new camps in response to context-speci c needs.. In some countries, such as Bangladesh, Ethiopia, Kenya, Nigeria, and Tanzania, UNHCR reports that the majority of refugees continue to live in camps. 11 Studies have repeatedly shown that health service coverage and quality in UNHCR-supported refugee camps are often higher than that in the immediately host country population. 12 13-15 Previous studies using UNHCR data have shown that refugees living in protracted camps generally have lower crude and under-ve mortality rates than surrounding host communities, and possibly lower maternal mortality rates. 16,17 However, little is known about if or how the newborn morality burden has changed over time. The UNHCR Health Information System (HIS), a standardized system for routine reporting and analysis of public health indicators established in 2006, provides a valuable source of data for addressing these questions. 18 A previous study, conducted as part of the 2014 Interagency Working Group for Reproductive Health in Crisis's Global Evaluation, -examined reproductive health indicators, including neonatal mortality, in UNHCR post-emergency camps 2007-2013 in 10 countries. 19 We build on this study with a more extensive analysis of neonatal mortality data availability, burden and trends over the period from 2006 to 2017 to inform UNHCR strategies and global efforts to improve newborn care for con ict-affected populations.

Methods
Data on refugee camp populations, births and deaths over the period of 2006 to 2017 were extracted from the UNHCR HIS Database. 20 For comparison purposes, national and sub-national neonatal mortality rates (NMR) were extracted from Demographic and Health Surveys and Multiple Indicator Cluster Surveys conducted between 2006-2017 in countries with refugee camps using the UNHCR HIS. 21,22 Analysis included the following variables: country, name of camp, year and month of reporting, total camp population, total number of live births, total number of neonatal deaths, NMR (number of deaths during rst 28 days of life per 1,000 live births), and national/sub-national host country neonatal mortality rates. Additional variables included in exploratory analyses and data quality checks included infant mortality rate (number of deaths during rst year of life per 1,000 live births), stillbirth rate (number of fetal deaths after 22 weeks gestation per 1000 births [livebirths and stillbirths]), and under-ve mortality rate (number of deaths during rst year of life per 1,000 live births). Stillbirth rates were not analyzed alongside NMRs due to high frequency of missing data.
Three UNHCR HIS datasets were merged using camp name and reporting year/month as a unique identi er: a population dataset (12,020 entries), a mortality dataset (12,029 entries) and a maternity dataset (12,032 entries), resulting in a master database with 11,910 entries after dropping of duplicates. Camps with less than six months of data on neonatal mortality in a given year were excluded from the analysis (410 entries). Entries with a NMR > 100, infant mortality rate > 150, under-ve mortality rate > 500 or total livebirths less than 15 per month were also dropped from the analysis as fairly implausible outliers, resulting in the analytic database with 8,310 entries from 150 camps in 21 countries. (Fig. 1) Because the UNHCR HIS automatically calculates a mortality rate of 0% where the number of deaths is left blank, it is impossible to discern whether camps with a NMR of zero in a given month or year actually had zero deaths, or whether data is missing. Camps with months reporting zero newborn deaths in a given month were agged and a sensitivity analysis conducted to calculate annual NMR with and without these months included.
Host country NMRs were matched to 947 entries in UNHCR's camp database based on camp location and the year the Demographic and Health Survey (DHS) or Multiple Indicator Cluster survey (MICS) was conducted, with an understanding that DHS and MICS typically reports NMR for the ve and two year periods prior to data collection, respectively.  Figure 2 shows the global refugee NMR for each year, 2006-2017, as reported in the UNHCR HIS (Fig. 2a), and the aggregate refugee NMR for months where newborn deaths are reported (Fig. 2b). Inclusion/exclusion of reports with zero neonatal deaths changes the magnitude of the neonatal mortality burden in refugee camps from an aggregate NMR of 5.2 deaths to 24.5 deaths per 1,000 live births for the period 2006-2017, but trends in NMR over time remain similar. The analysis that follows excludes months where zero newborn deaths are reported (upper bound of NMR sensitivity analysis presented in Supplementary File 2, resulting in a conservative assessment of newborn health in refugee camps. Average annual and country-speci c refugee camp NMRs are presented in Fig. 3 and Fig. 4, respectively. When refugee camp data from all countries using the UNHCR HIS are pooled, there appears to be little change in the average NMR over time. (Fig. 3) However, pooling data may mask disparities in NMRs across sites. Average NMRs for 2006-2017 ranged from a high of 56 neonatal deaths per 1,000 live births in Burkina Faso refugee camps that reported data over a ve-year period (2013-2017) to a low of 12 neonatal deaths per 1,000 live births in Central African Republic camps that only reported data in 2013. (Fig. 4). Country-speci c rates based on reports from months with one or more neonatal deaths reported in the UNHCR HIS are presented by year in Table 1. In most countries and years where DHS or MICS data was available, refugee camp NMRs as reported in the UNHCR HIS was lower than that of the immediate host community (subnational survey division where camp is located) or national average. Figure 5 shows example of refugee camp NMRs calculated using UNHCR HIS data compared with host population NMR estimated from population-based survey data in selected African and Asian countries. Annex 3 (Supplementary Materials) presents NMRs by location and year for refugee camps and national/subnational host populations, where available; 60% of locations/years in this sub-analysis indicated that the neonatal mortality burden among refugees is likely lower than that of immediately surrounding host communities.

Discussion
Understanding what has worked well and what needs to work better to ensure newborns survive and thrive in humanitarian settings requires stakeholder commitment, reliable data, and sustained funding for both health service delivery and implementation research that includes con ict-affected, forcibly displaced, marginalized and stateless populations often missed by national health systems, surveys and surveillance systems. Although residents of refugee camps constitute a relatively small proportion of con ict-affected, forcibly displaced, marginalized and stateless populations that go uncounted in populationbased surveys and vital registration systems, reporting and analyzing neonatal mortality burden and trends among this population is a critical step in strengthening data use and accountability for neonatal health in humanitarian settings. The UNHCR Health Information System provides insights into neonatal mortality burden and trends among refugee camp residents, as well as issues to consider in design and use of routine health information systems to monitor neonatal health in sub-national populations.
Although NMRs in stable refugee camps are often lower than surrounding host populations, the neonatal mortality burden is still too high. Given the widespread recognition that perinatal outcomes are often misclassi ed and underreported, [23][24][25] and that there are still major gaps in stillbirth data in the UNHCR HIS, it is likely that neonatal mortality is underreported, and the true burden in refugee camp populations may be even higher that the upper bounds of gures presented here.
Even in countries hosting refugees in protracted camp situations, there has been limited change in the neonatal mortality burden over the last decade, although one would expect a decline over time. This may be due in part to limited attention to newborn health in humanitarian settings prior to the development of the Newborn Health in Humanitarian Settings Field Guide and UNHCR's Operational Guidelines on Improving Newborn Health in Refugee Operations in 2014-2015. 9,26 The across and within country variation in NMR, and lack of change over time, suggests that concerted efforts are needed to further understand and address neonatal mortality. UNHCR has already begun this work in some settings, including the establishment of a neonatal mortality audit system in Jordan, and projects focused on improving access to quality health services for women and newborns in Cameroon, Chad, Jordan, Kenya, Niger and South Sudan. [27][28][29][30] Continued investment is needed in these and other refugee settings, including efforts to strengthen data availability and use as part of quality improvement efforts, as outlined in the recently launched Roadmap to Accelerate Progress for Every Newborn in Humanitarian Settings 2020-2024. 31 Our analysis has several limitations. First, incomplete and/or inaccurate reporting of routine data and outdated population size estimates can lead to implausibly high or low mortality estimates. In some cases, high average NMRs over the study period may be explained by low numbers of births reported in certain years. For example, low camp refugee population and live births numbers in Burkina Faso might explain NMRs reported in 2015-2017. Second, although UNHCR guidelines suggest that all deaths occurring within a camp should be reported in the HIS, stillbirths and neonatal deaths taking place outside of health facilities may be underreported. Third, the UNHCR HIS does not differentiate between early or late neonatal deaths, or document cause of death, all of which are critical for identifying and addressing gaps in accessibility and quality of care. Finally, population-based survey data for sub-national divisions where refugee camps are located also have limitations and cannot be directly compared to routine health information system data. 32 Surveys typically use a recall period of up to 2 years prior to survey administration; because availability of refugee camp HIS data varied from year to year, we compared NMRs reported in population-based surveys to NMRs reported in the UNHCR HIS in the year the survey was published, not the years where reported deaths occurred.
Data issues identi ed through this analysis point to opportunities for strengthening the UNHCR HIS to better facilitate monitoring and promote accountability for efforts to improve newborn health. These include increasing attention to the documentation of stillbirths in health facility registers and HIS reports; strengthening community health worker engagement in identi cation and reporting of stillbirths and newborn deaths occurring outside of health facilities; establishing mechanisms to distinguish between missing data and reports of zero deaths in a given month; incorporating alerts to ag potentially implausible ratios among neonatal, infant and child mortality in HIS; and consistently reporting neonatal mortality rates alongside infant and child mortality in public health reports. World Health Organization guidance cautions against calculating case fatality rates for time periods when the number of deaths are too small for a stable calculation. 33,34 It may be advisable for facilities with few deaths to calculate NMR on a quarterly or even annual basis for increased stability of the indicator. Future questions to be researched include at which levels of mortality is a stable rate produced, how frequently indicators should be calculated, and how to account for factors unique to refugee camp settings, including in/out migration, community and health facility consolidation during time periods selected for calculation of NMR, which exacerbate challenges in obtaining accurate denominators for mortality rate calculation. Learning from these efforts may also inform efforts to strengthen routine health information systems and maximize the use of routine data for monitoring and evaluation of health development efforts. [35][36][37] Outside of humanitarian settings, there are efforts to develop practical methods for improving measurements of perinatal mortality (fresh stillbirth and early neonatal mortality) in health facilities, 38 and strengthen maternal and perinatal death surveillance and response systems which may provide models for replication in refugee camps. [39][40][41] Conclusions The perinatal period represents a critical time of vulnerability and risk for newborns. Accurate and reliable data on mortality rates during this period is critical for improving access and quality of care for settings. The UNHCR HIS is a valuable source of information on refugee population health service coverage and outcomes. However, its functionality for tracking and analysis of neonatal mortality is quite limited. Concerted efforts and investments are needed to ensure every neonatal death and stillbirth are captured in routine data collection, monitoring, and reporting, so that no populations are left behind in efforts to improve maternal and newborn health and well-being. Global refugee neonatal mortality rate (NMR), with months reporting zero deaths included and excluded