What is conflict-related mortality?
Direct conflict-related deaths occur when civilians die from events associated with combat, whether or not they were intended to harm civilians (such as bombings).
Indirect conflict-related deaths – from, for example, malnutrition and preventable diseases – occur after the conflict has destroyed transportation networks, health facilities and critical components of the health system, or other infrastructure.
Indirect conflict-related deaths usually outnumber direct conflict-related deaths.
Why measure conflict-related mortality?
Armed conflict results in substantial morbidity and injury, mortality, and displacement of individuals, households and entire communities.
The nature of armed conflicts has changed substantially. Traditionally, regular armies fought wars to capture territory and advance geopolitical or ideological interests. Today state- and non-state actors fight over issues of identity, such as religion or ethnicity. While the number of conflicts and their direct casualties have decreased over time, conflicts last longer, especially in low-income and ethnically divided countries (see Chaudoin et al) . Conflicts often occur in densely populated urban areas in middle-income countries, displacing substantial numbers of people.
Because armed conflict destroys systems and infrastructure and displaces populations, it is difficult for health workers to provide and monitor population health. Routine data collection, using health information and civil registration systems, may not function, and conflict can hamper efforts to maintain surveillance and undertake surveys. Mortality rates measure population health in war-affected societies and guide public health interventions and humanitarian response, and can be used to create a historical record, advocate on behalf of affected populations, and create awareness about the impact of conflicts on mortality.
Measuring the mortality effects of conflict requires multiple data sources and complementary methods of estimation.
How is conflict mortality measured?
Conflict demographers and epidemiologists measure mortality using the following indicators:
Under-five mortality rate (U5MR)
There is much uncertainty associated with measurement of all-cause mortality during conflict. Measurement of cause-specific mortality is particularly challenging – often relying on crude cause of death codes, given that a qualified medical practitioner seldom certifies deaths.
Despite the scale of mortality in conflicts, there are many data gaps and selection biases. Mortality data in conflicts are incomplete and it is rare that a physician records a cause of death. Poor specificity and data quality make it difficult for researchers to draw conclusions and make generalizable inferences. Improving data and methods to measure conflict-related mortality is crucial to advance public health and protect human rights.
Data are usually collected using prospective surveillance systems or retrospective surveys and censuses. Another approach is to rely on referrals from community informants.
Prospective surveillance systems
These systems collect data continuously as soon as possible after deaths occur. Active surveillance is when a staff member visits households regularly to gather information about deaths that occurred in the households. Passive surveillance is when the system relies on a household member or some other person to report deaths. Surveillance systems include:
Civil registration and vital statistics (CRVS) systems
CRVS systems provide the best source of mortality data when coverage is complete, but most low- and middle-income countries do not have complete CRVS even during peacetime.
Demographic and health surveillance systems (DHSS)
DHSS monitor ongoing demographic events occurring in a geographically defined population. So long as a DHSS operates during the conflict, it can provide mortality data for the small population it covers.
Dedicated surveillance systems
When no CRVS or DHSS exists, authorities can set up an active surveillance system to monitor mortality and morbidity. Typically, a community health worker visits households routinely to ask about demographic events and then updates the population figures. Health workers may monitor graveyards, although relying on this method alone may under-report deaths and incur selection bias.
Facility-based health information systems
Facility-based systems record deaths and other health outcomes that occur in health facilities. Because most conflict-related deaths occur outside health facilities, these sources can only provide data to calculate case-fatality ratios, not population-based mortality rates.
Media reports
Media reports are a source that analysts can harvest for counts of deaths. These reports usually undercount deaths and are biased because news agencies tend to cover deaths that are newsworthy, for example when large numbers of deaths occur at one time or when deaths occur in an unusual manner.
Surveys and censuses
When prospective surveillance data are not available, researchers can use retrospective surveys and censuses to collect data on mortality.
Retrospective surveys
Surveys ask about deaths and morbidity that occurred during a time-period. Sampling is a major challenge in surveying conflict-affected populations. It is often impossible to prepare an accurate sampling frame from which to draw a simple random sample or to organize households in a clear pattern for systematic random sampling. Researchers commonly use multi-stage cluster sampling instead. Since households are not randomly selected within a cluster, they may be similar to each other creating excessive homogeneity within clusters. This clustering of risk of death can be more marked in conflict than in peace time settings (see the SMART project).
Population censuses
Censuses involve complete enumeration of households and individuals in a country or a defined geographic area. A census provides population denominator data with which to calculate death rates. Some censuses include questions about past mortality – such as summary birth histories, sibling histories or accounts of household deaths – with which to estimate mortality. Since censuses cover the entire population, researchers can calculate mortality rates for small areas without sampling error. However, censuses occur infrequently, usually every ten years, making timely and up-to-date mortality measurement difficult. Armed conflicts often interrupt the decennial census schedule, making the duration between censuses even longer. Migration during and after conflicts further complicates the use of census data to estimate mortality.
Methods for ascertaining a death in a household
- Past household approach: the interviewer lists the household members who were present at the beginning of the recall period and determines which of them are no longer present. For each person who is no longer present, the interviewer determines whether they have died or migrated.
- Current household approach: the interviewer lists all members of the household on the day of the survey and then asks about any deaths that have occurred since the beginning of the recall period.
- Standardized monitoring and assessment of relief and transitions (SMART) is similar to the current household census, but also asks about in- and out-migration.
- Survival of children/birth histories and survival of parents, and siblings: the interviewer asks questions about children ever-born and surviving birth histories, and survival of parents and siblings.
Mortality data from retrospective surveys and censuses is subject to non-sampling biases. Recall bias occurs when respondents erroneously report or omit deaths. For example, respondents may over-report if the deaths were violent. They may under-report deaths to household members who were not related to the respondent or of young children. Poor recall of date of death can lead to over- or under-reporting of deaths within the recall period. Reporting bias may result from respondents intentionally under- or over-reporting deaths or the number of people living in the household. In conflict settings, there can be many reasons why respondents want to under- or over-report deaths. Survival bias results from interviewing only households with at least one surviving member who can report deaths.
The informant-based method
This is an exhaustive search process to record all deaths occurring in a population within a recent time-period (usually 60 days). It begins with a focus group to identify key informants and locate other sources of death records. During key informant interviews, the informants independently list all deaths within the time-period. Interviewers visit all households which informants identify as having had a death; and ask the next of kin, aged 18 years or older, for the date, cause, and place of death, and to list other deaths in the household or in the community in the time-period. Interviewers visit new households and repeat the process until they have visited all households identified. To estimate mortality rates, the method requires knowledge of the population size (obtained either from existing sources or through estimation).
Ethical issues around confidentiality are important when using this method. Both referral-based sampling and neighbourhood methods reduce costs and staff time in data collection over full-scale surveys but both methods are subject to selection bias.
Relative to retrospective surveys, the method is cheaper, requires less time for data collection, entry and analysis, and reduces respondent time. The method only measures mortality, whereas retrospective surveys can also measure other indicators.
Data analysis
Investigators must check data for quality, consistency and bias, especially given the challenges of data collection in armed-conflict settings.
Estimation of mortality
If investigators have recorded the numbers of deaths and the population size, they can estimate measures of both direct and indirect conflict-related mortality. If data include numbers of children ever-born and those surviving, birth histories, and survival of parents and siblings, they may estimate mortality rates using indirect methods (see United Nations). Investigators can also estimate mortality by applying demographic methods to the age and sex structure of a population before and after a conflict;
Comparison with other estimates
Investigators can compare estimated CMR and U5MR against emergency thresholds to determine the level of response. Thresholds, which vary by region based on baseline mortality, range from a CMR of 0.3 to 0.8 deaths per 10,000 per day and an U5MR of 0.1 to 2.1 deaths per 10,000 per day. Investigators can compare mortality measures from various data sources to determine plausible levels.
Multiple systems estimation
Multiple systems estimation (MSE ) (or capture-recapture) addresses the problem that some deaths are never recorded in armed conflict settings. This statistical approach uses numerous data sources and examines the overlaps among them to estimate the total number of deaths, including those missing from all data sources. Researchers have applied MSE in several countries to estimate the total number of deaths due to human rights violations and armed conflicts.
Emerging opportunities
Opportunities to further advance the measurement of conflict-related mortality include:
- Technologies provide more ways to collect data on conflict deaths in challenging situations. For example, Elamein et al. used the popular messaging application, WhatsApp, to collect, coordinate and triangulate reported attacks on health workers and health facilities in Syria in real-time by members of the health cluster activated as part of the UN humanitarian response. This mobile-based technology provided timely and reliable information and drew attention to violations of international humanitarian law.
- New approaches to data integration increase options for improved estimation and analysis. Data systems are under-utilised in humanitarian settings partly because there are few platforms to share and integrate data from multiple sources. Although MSE is not a new technique, this approach can provide accurate estimates or validate mortality data, as Roberts et al. have done to test the informant method. The UNHCR used MSE to establish a minimum bound for conflict mortality in the early years of the Syria Crisis.
- Health information systems, mortality data systems, and displacement data systems are not linked in many humanitarian settings. Integrating multiple sources of mortality information can correct for under-enumeration in individual data systems. For example, by combining data on internal displacement (such as from the International Organization for Migration’s Internal Displacement Tracking Matrix) and international refugee stocks (such as from UNHCR’s ProGres registration system), analysts can better estimate the population-at-risk and construct more accurate population mortality rates.
- Combining geospatial data with demographic data, Pezzulo et al. have harnessed high-resolution satellite imagery to estimate child mortality at subnational levels. Use of satellite imagery (specifically satellite-based measures of night-time lights and satellite-based vegetation indices) along with socio-demographic covariates make it possible to estimate mortality for smaller geographic units. Although these are modelled estimates based on mortality at larger geographic levels rather than observed data, this linking of data is an innovative way to estimate mortality in small areas.
- Modelled estimates rely on assumptions about the homogeneity of mortality within geographic areas – an assumption that glosses over systems-level perturbations and forced migration effects in armed conflict situations. Decisions about resource allocation and humanitarian response ultimately need to engage local contexts and factor in the variability in population processes and health outcomes that result from armed conflict. Satellite imagery can provide a sampling frame for retrospective mortality surveys. Galway et al. used gridded population data and geographic information systems for first-stage sampling for a cluster mortality survey in Iraq. This meant they could draw a representative sample despite that the last census was outdated and security was tenuous in some parts of the country.
- As new data sources become available and estimation methods more complex, end-users and practitioners require improved standards of reporting and technical guidance. It has been more than ten years since Checchi and Roberts developed their primer on interpreting and using mortality data in humanitarian emergencies. This primer provided an excellent overview of classical approaches to data collection, such as retrospective mortality surveys and prospective surveillance. It also presented fundamental concepts for interpretation of data such as validity, reliability and bias. But the field has evolved substantially.
- New technologies have created alternative sources of data (such as satellite imagery and mobile-based data sources) and facilitated more sophisticated modelling techniques such as MSE and spatio-temporal-regression models. These methods entail data quality issues and simplifying assumptions, which investigators must consider when interpreting or making decisions based on their findings. For example, MSE relies on strong assumptions of independence and error-free matching of data across systems. Methods that use data from satellite imagery and geographic information systems, make assumptions about the relationship between geographic information and the demographic profile of small areas and their health and mortality outcomes. Practitioners and decision-makers need practical guidance on how to responsibly use new types of mortality data and modelled mortality estimates.
Contents
Source chapter
The complete chapter on which we based this page:
Silva R., Mizoguchi N. (2019) Mortality Data in Service of Conflict-Affected Populations. In: Macfarlane S., AbouZahr C. (eds) The Palgrave Handbook of Global Health Data Methods for Policy and Practice. Palgrave Macmillan, London.
Additional resources
United Nations Handbook on the Collection of Fertility and Mortality Data
Interpreting and using mortality data in humanitarian emergencies A primer for non-epidemiologists
Twine is a standardised health information system for refugee settings developed by UNHCR and a web application that facilitates standardised data collection, analysis, and sharing.
proGres (Profile Global Registration System) is an IT case management tool developed by UNHCR. proGres is the main repository in UNHCR for storing individuals’ data.
PRIMES (Population Registration and Identity Management EcoSystem) encompasses all interoperable UNHCR registration, identity management and caseload management tools and applications
Uppsala/PRIO is a data repository from armed conflict settings maintained by the Uppsala University and the Peace Research Institute that contains passive reports of violent war-related deaths from 1900 onward.(44)
The Complex Emergency Database (CEDAT) is a repository of small-scale mortality and nutrition data from field surveys conducted by humanitarian agencies but is no longer publicly available.
The Sphere Project. Sphere handbook: humanitarian charter and minimum standards in disaster response.
United Nations High Commissioner for Refugees. Health information system (HIS) toolkit.
United Nations High Commissioner for Refugees. Twine.
World Health Organization. Outbreak surveillance and response in humanitarian emergencies: WHO guidelines for EWARN implementation.
Checchi F, Roberts L. Interpreting and using mortality data in humanitarian emergencies: a primer for non-epidemiologists.
The SMART Project. Measuring mortality, nutritional status, and food security in crisis situations: SMART methodology version 1.
United Nations, Department of Economic and Social Affairs, Statistical Division. Handbook on the collection of fertility and mortality data.
Working Group for Mortality Estimation in Emergencies. Wanted: studies on mortality estimation methods for humanitarian emergencies, suggestions for future research.
Roberts B et al. A new method to estimate mortality in crisis-affected and resource-poor settings: validation study.
Lacina B, Gleditsch NP. Monitoring trends in global combat: a new dataset of battle deaths.
The complex emergency database: a global repository of small-scale surveys on nutrition, health and mortality.
Latest publications ⇒
Latest news ⇒