Excess mortality during the COVID-19 pandemic: a geospatial and statistical analysis in Mogadishu, Somalia

Abstract

Background While the impact of the COVID-19 pandemic has been well documented in high-income countries, less is known about the health effects in Somalia, where health systems are weak and vital registration is underdeveloped. Methods We used remote sensing and geospatial analysis to quantify burial numbers from January 2017 to September 2020 in Mogadishu. We imputed missing grave counts using surface area data. Simple interpolation and a generalised additive mixed growth model were used to predict actual and counterfactual burial rates by cemetery and across Mogadishu during the most likely period of COVID-19 excess mortality and to compute excess burials. We undertook a qualitative survey of key informants to determine the drivers of COVID-19 excess mortality. Results Burial rates increased during the pandemic, averaging 1.5-fold and peaking at a 2.2-fold increase on pre-pandemic levels. When scaled to plausible range of baseline crude death rates, the excess death toll between January and September 2020 was 3200–11 800. Compared with Barakaat Cemetery Committee's burial records, our estimates were lower. Conclusions Our study indicates considerable underestimation of the health effects of COVID-19 in Banadir and an overburdened public health system struggling to deal with the increasing severity of the epidemic in 2020.

Date of introduction and epidemiologic patterns of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in Mogadishu, Somalia: estimates from transmission modelling of satellite-based excess mortality data in 2020

“…Details of the method for inferring excess mortality from satellite images are described in an accompanying article 18 . Briefly, cemeteries in the Banadir administrative region, which contains Mogadishu, were identified via open-source location data and satellite imagery (OpenStreetMap, Google Earth, GoogleMaps) by a combination of automatic image recognition and manual annotation, in addition to key informant interviews and field visits to identified cemeteries.…”

Date of introduction and epidemiologic patterns of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in Mogadishu, Somalia: estimates from transmission modelling of satellite-based excess mortality data in 2020

“…The analysis code has be deposited on Zenodo 19 . Details of the method for inferring excess mortality from satellite images are described in an accompanying article 18 . Briefly, cemeteries in the Banadir administrative region, which contains Mogadishu, were identified via open-source location data and satellite imagery (OpenStreetMap, Google Earth, GoogleMaps) by a combination of automatic image recognition and manual annotation, in addition to key informant interviews and field visits to identified cemeteries.…”

“…Deaths occur in the model with a gamma-distributed delay 27,28 following the transition from exposed (E) to pre-symptomatic (I p ) state. To account for the lack of intensive care unit (ICU) beds (the available estimate is around twenty ICU beds in the entire Mogadishu area 18 ), we set the mean delay to 15 days, one week less than the estimate used for high-income countries 26 , to reflect the fact that severe cases were unlikely to receive adequate treatment. Other parameters of disease progression were fixed to consensus estimates in the literature (Extended data, SI Table 1 22 ).…”

“…Therefore, it is possible that the real peak of deaths was higher and the excess mortality curve had a sharper exponential rise, but the number of burials underestimates excess deaths during the peak period of the pandemic, perhaps because of out-migration from the city (as observed in India 35 ) during the pandemic or opening of new burial sites not included in our satellite data. While field visits and interviews 18 did not identify new burial sites and we therefore cannot ascertain the veracity of this hypothesis, we approximated it by re-fitting the model to the pre-peak period up to 13 April only. This led to a shift in the date of introduction estimates to the period of mid-November to mid-December for seed sizes of 20 to 100 (Extended data, SI Figure 9 22 ), with median R 0 estimates rising to 1.4-1.7.…”

“…In the current study we used a dynamic transmission model to analyse a time series of excess deaths in Mogadishu (Somalia) inferred from satellite images of the six main cemeteries in the city 18 . Our aim was to estimate the probable date of introduction of SARS-CoV-2, as well as the basic reproduction number (R 0 ) and the effect of non-pharmaceutical interventions.…”