There was excess mortality in all 12 districts included in the study between January 2020-December 2022, with the most excess deaths (52.6%) in 2021. These results suggest that the impact of the COVID-19 pandemic in Zambia was more widespread than official statistics indicate, and not only restricted to urban centres such as Lusaka (9, 21, 25, 26). Similar findings of excess mortality in other sub-Saharan countries during the COVID-19 pandemic have been observed (27, 28). Globally, the World Health Organisation estimates excess mortality is 2.74 times higher than reported COVID-19 deaths (28).
Modelled estimates of excess mortality during the COVID-19 pandemic in Zambia vary widely (74.3 credible interval (CI) [2.5-147.6](29) vs 228.2 [165.9-322.8] per 100,000 population (30)). Due to the unavailability of publicly available population-level mortality data, these current estimates of excess mortality in Zambia used statistical models to directly predict excess mortality for Zambia(30) or used mathematical models to generate historical and current mortality data and then calculated the excess mortality rate (29). Our national estimate of the excess mortality rate (median 237.5 [(IQR)170.5-282.5] per 100,000 population), which lies within the range of modelled estimates for Zambia, is most likely an underestimate as not all deaths that occur within communities are reported at health facilities, and therefore were not included in this analysis. Our study demonstrates the value of applying mortality surveillance to understand the impact of a major public health event. If done in real-time, it could have also helped inform public health messaging and policymaking in response to the COVID-19 pandemic in Zambia. Further studies need to be conducted to understand the characteristics and explore surveillance strategies to detect and report these otherwise undocumented community deaths. Further studies should also focus on what effect if any, that district-level socio-economic determinants of health during the COVID-19 pandemic such as COVID-19 vaccine coverage and healthcare access had on the observed subnational distribution of excess mortality rates during the COVID-19 pandemic.
The observed excess deaths were most likely due to COVID-19 due to the observed increase in the daily number of deaths during COVID-19 wave periods compared to non-COVID-19 wave periods. To our knowledge, there were no other reported widespread public health emergencies during the study period that could explain the abrupt increase in the number of deaths during the specific wave periods and across all 12 districts almost simultaneously (1). Additionally, when we analysed the different wave periods, we noted that the relative increase in daily deaths during wave periods was associated with the relative case fatality of the predominant strain of SARS-CoV-2 associated with that wave. With waves associated with SARS-CoV-2 strains with a higher case fatality recording more deaths per day than those with a lower-case fatality (4, 9, 25). Further, after an increase in the daily number of deaths during a COVID-19 wave period, we observed a return to baseline pre-pandemic mortality rates between waves and this was consistent across all 12 districts visited and across the different age groups.
COVID-19 mortality has been shown to disproportionately affect the elderly (31–34). We observed an increase in the overall median age at death (44 years vs 41 years, p < 0.001) and in the proportion of deceased persons aged 65 years and older (27.7 per cent vs 23.4 per cent, p < 0.001) during COVID-19 wave periods with a return to baseline between wave periods respectively. However, the exact proportion of these excess deaths that were directly attributable to COVID-19 remains unknown because of limitations in antemortem and postmortem SARS-CoV-2 testing and limited death registration and certification across the country.
There were more community deaths than facility deaths in both wave periods and non-wave periods. Community deaths increased during wave periods, suggesting potential gaps in health services brought on by the COVID-19 pandemic. As not all deaths that occur within the community are brought to health facilities before burial, the actual proportion of total deaths that occur within the community may even be higher. An analysis of places of death in Zambia among adults 15–59 years between 2010–2012 showed that slightly less than half of the adult deaths occurred in the home, factors associated with dying in a health facility included higher educational attainment, urban versus rural residence, and being of female gender (35). The observed increase in the proportion of deaths in the community during COVID-19 waves could be due to barriers to access to health facilities as some facilities were repurposed to serve as specialised COVID-19 treatment centres whilst other facilities scaled down services offered by only offering essential health services or attending to only emergencies (36–38). This could have compromised the quality of outpatient care that chronically ill patients received. Additionally, the myths and misconceptions around COVID-19 could have prevented those in most need of care from seeking health care (39). We recommend risk communication and engagement strategies tailored to increasing demand within the community for seeking health services during public health emergencies. Additionally, surge capacity plans should be developed by the Ministry of Health and implemented during public health emergencies. These could help ensure the continued provision of essential health services as well as provide additional capacity to respond to the public health emergency.
Our study had several limitations. As we investigated excess deaths, we were unable to quantify the proportion of these excess deaths that were due to COVID-19, however, due to the timing and demographic composition of these deaths and the absence of other explanatory events, we believe that most of these deaths could have been due to COVID-19. Only community deaths that were reported at health facilities were analysed during this investigation, as such our findings underestimate the total deaths that occurred within these districts. Study sites were purposively selected, and this could have introduced self-serving bias in our findings. However, we assumed that the findings from the 12 districts was representative of the entire country due to the heterogeneity of the districts selected. Sampling additional districts was not possible because of resource limitations (data abstraction was time-consuming). Our method of determining excess mortality did not consider potential seasonal variations. However, due to the large size of the data set, the consistency of findings across the different districts, the consistency of our findings with current known epidemiological characteristics of COVID-19 and the consistency of our findings with other similar studies, we believe our findings are credible.
There was excess mortality in all 12 districts visited during the COVID-19 pandemic in Zambia with most of these deaths occurring within the community and among the elderly. These findings suggest the impact of the COVID-19 pandemic in Zambia was far greater than implied by reported COVID-19 deaths alone. Strengthening routine and continuous mortality surveillance systems with cause of death ascertainment especially among community deaths could help guide public health decision-making and strengthen risk communication and community engagement during public health emergencies.