Background: Complex humanitarian emergencies are characterised by a break-down of health systems. All-cause mortality increases and non-violent excess deaths (predominantly due to infectious diseases) have been shown to outnumber violent deaths even in exceptionally brutal conflicts. However, affected populations are very heterogeneous and refugees, internally displaced persons (IDPs) and resident (non-displaced) populations differ substantially in their access to health services. We aim to show how this translates into health outcomes by quantifying excess all-cause mortality in emergencies by displacement status. Methods: As standard data sources on mortality only poorly represent these populations, we use data from CEDAT, a database established by aid agencies to share operational health data collected for planning, monitoring and evaluation of humanitarian aid. We obtained 1759 Crude Death Rate (CDR) estimates from emergency assessments conducted between 1998 and 2012. We define excess mortality as the ratio of CDR in emergency assessments over `baseline CDR' (as reported in the World Development Indicators). These death rate ratios (DRR) are calculated separately for all emergency assessments and their distribution is analysed by displacement status using non-parametric statistics. Results: We found significant excess mortality in IDPs (median DRR: 2.5; 95 % CI: [2.2, 2.93]) and residents (median DDR: 1.51; 95 % CI: [1.47, 1.58]). Mortality in refugees however is not significantly different from baseline mortality in the host countries (median DRR: 0.94, 95 % CI: [0.73, 1.1]). Conclusions: Aid agencies report the highest excess mortality rates among IDPs, followed by resident populations. In absolute terms however, due to their high share in the total number of people at risk, residents are likely to account for most of the excess deaths in today's emergencies. Further research is needed to clarify whether the low estimates of excess mortality in refugees are the result of successful humanitarian interventions or due to limitations of our methods and data.
