The COVID-19 pandemic, caused by the SARS-CoV-2 virus, has had a significant global impact, resulting in millions of confirmed cases and deaths despite the implementation of non-pharmaceutical interventions (NPIs). However, the reported impact of COVID-19 in many African countries, including Zambia, has been notably low compared to global averages. Several hypotheses have been proposed to explain this perceived low impact, including a younger population structure, pre-existing immunity from other coronaviruses, climate-related factors, effective NPIs, and genetic factors. These hypotheses either suggest reduced transmission or lower severity of the disease. However, it's widely acknowledged that reported case and death numbers significantly underestimate the true burden of the disease due to limitations in testing capacity and surveillance systems. This raises the crucial question: once these ascertainment biases are accounted for, to what extent do these alternative hypotheses hold? Understanding whether there was an "Africa paradox", where perceived impact was lower than expected, and identifying the factors contributing to this discrepancy are critical for drawing accurate conclusions and informing future pandemic preparedness strategies. Data on SARS-CoV-2 spread during the initial phases of the pandemic (2020) in Africa remain scarce, hindering a comprehensive understanding of the true impact. This research addresses this gap by leveraging data from Lusaka, Zambia, to investigate the age-dependent patterns of COVID-19 severity and compare them to global patterns, explicitly accounting for the protective effect of the city's relatively young population structure. The study uses a unique dataset combining burial registration data, SARS-CoV-2 prevalence from population surveys and post-mortem samples from the University Teaching Hospital (UTH), the largest morgue in Lusaka, to estimate excess mortality, transmission dynamics, and the infection fatality ratio (IFR). The aim is to determine whether the patterns of excess mortality in Lusaka during the first wave of the pandemic align with those observed in other countries, and to assess the validity of previously proposed explanations for the seemingly low reported COVID-19 impact in Africa.

Existing literature highlights the significant challenges in accurately assessing the true burden of COVID-19 in low-income settings, particularly in Africa. Studies have shown that reported cases and deaths significantly underestimate the actual number of infections and COVID-19-related deaths due to limited testing capacity and weak surveillance systems. Several hypotheses have been proposed to explain the seemingly low COVID-19 impact in Africa, including the younger population structure, pre-existing immunity from other coronaviruses, climatic conditions, implementation of NPIs, and genetic factors. Some studies suggest reduced transmission rates due to factors such as lower population density and outdoor lifestyles, while others postulate that high transmission rates were accompanied by lower severity due to pre-existing immunity. However, the lack of robust data, particularly from the early stages of the pandemic, makes it challenging to assess the validity of these hypotheses. Seroprevalence studies have been conducted to estimate the true extent of infection, but they are limited in geographical coverage and the majority of available data comes from high-income countries. Data from excess mortality has proven very challenging to obtain from Africa. There was a lack of sufficient mortality data in the WHO’s global study of excess mortality, making robust comparison with other world regions very difficult. Studies in specific locations have shown stark differences in prevalence levels compared to official reports, with some reporting over 90-fold differences between reported cases and true infection rates. While some studies have examined the impact of COVID-19 on specific regions of Africa, a comprehensive analysis integrating mortality data and infection prevalence data to assess the true burden of the disease is still limited, especially regarding age-dependent severity.

This study used a multi-faceted approach combining several data sources and statistical modeling techniques to estimate the impact of COVID-19 in Lusaka, Zambia, during 2020. The key data sources included: 1. **WHO Excess Mortality Estimates:** Data from the World Health Organization (WHO) on excess mortality in Zambia during 2020 were used to provide a global context for the Lusaka findings. These estimates were adjusted for demographic vulnerability using age-specific IFRs from Brazeau et al. (2022), weighted by the population age structure. A demographic vulnerability-weighted impact (DVWI) measure was calculated to standardize excess mortality by the population's susceptibility to severe disease. 2. **Burial Registration Data:** Age-stratified weekly all-cause burial registration data from Lusaka, spanning from mid-2017 to mid-2021, were collected from official registries. These data were used to estimate excess mortality during the pandemic by comparing age-specific registration patterns during 2020-2021 to pre-pandemic (2018-2019) patterns. A statistical model was developed to account for the inherent volatility in burial registration data. This model used the weekly registration rates of children under five years (U5) as a proxy for underlying registration changes. By modeling the age-distribution of deaths within the registered deaths and comparing it to the observed data, researchers could estimate the excess number of registrations in older age groups (5+). A scaling factor based on the relative difference between reported U5 weekly mortality and the median U5 registration rate during 2018-2019 was then applied to account for potential registration process disruptions. Sensitivity analyses were conducted using different registration capture rates (80%, 90%, 100%). 3. **Post-Mortem PCR Prevalence Data:** Weekly post-mortem polymerase chain reaction (PCR) prevalence data from UTH during June-October 2020 were used to provide insights into COVID-19-related deaths. 4. **Population-Level PCR Prevalence and Seroprevalence Survey Data:** Data from a population-based survey conducted in July 2020, which provided PCR prevalence and seroprevalence rates, were also incorporated. **Statistical Modeling:** A Bayesian inferential framework was used to fit an age-structured SARS-CoV-2 transmission model to the combined data sources (excess mortality, post-mortem PCR prevalence, and population survey data). The model was parameterized with Zambia’s demographic structure and a social contact matrix from a nearby region in Zimbabwe. A time-varying reproduction number (R0(t)) and effective reproduction number (Reff) were estimated. The model was also used to infer key epidemiological parameters, such as the cumulative attack rate and IFR. Sensitivity analyses were conducted to assess the impact of different assumptions, including the age-gradient and overall level of IFR, the duration between infection and death, and potential underlying changes in U5 mortality. The model was fitted using Metropolis-Hastings Markov chain Monte Carlo (MCMC) sampling techniques.

The study's key findings indicate a substantial impact of COVID-19 on mortality in Lusaka during 2020. 1. **Excess Mortality:** The study estimated 3212 excess deaths (95% CrI: 2256–4393) in 2020, representing an 18.5% increase relative to pre-pandemic levels. This equates to 1256.2 excess deaths per million. These figures were robust to the assumptions made regarding burial registration rates. The estimates were higher compared to WHO estimates of excess mortality when taking into account Lusaka's relatively young population. Using the DVWI, Lusaka's excess mortality was far greater than those in Europe and the Americas. 2. **SARS-CoV-2 Transmission Dynamics:** The dynamical model indicated high transmissibility during May-June 2020 (R0 significantly above 2), followed by a decline to below 1 by late July, suggesting the epidemic peaked around then. The cumulative attack rate was estimated to be between 15-30% by October 2020, suggesting substantial undetected spread. The decline in transmissibility happened despite low population immunity. 3. **COVID-19 Severity:** The model's fit to the data indicated that age-specific COVID-19 severity in Lusaka was consistent with estimates from other countries, without requiring exceptional explanations for the low reported figures. Sensitivity analyses showed that the model's fit was more sensitive to variations in the age-gradient of IFR than to changes in the overall IFR level. The results were found to be relatively robust across various sensitivity analyses, although this did not exclude the possibility of relatively small variations in IFR values when compared to those seen elsewhere. Overall IFR estimates were found to be within the range of 0.088 and 0.183%. 4. **Data Limitations:** The study acknowledged the inherent volatility and potential underreporting in burial registration data. However, the approach used to address the volatility was found to be robust and the estimates remained highly robust to differences in assumptions around death capture rates.

This study's findings challenge the prevailing narrative of low COVID-19 impact in Lusaka, Zambia. By combining diverse data sources and using a sophisticated statistical framework, it provides compelling evidence for significant, yet largely unreported, excess mortality. This underscores the importance of using multiple data sources to capture a fuller picture of the pandemic's true burden in data-poor settings. The study's results strongly suggest that the age-specific patterns of COVID-19 severity in Lusaka were similar to those observed in other parts of the world. This contrasts with hypotheses proposing unique protective factors in Africa that significantly reduce COVID-19 severity. The significant excess mortality in Lusaka highlights the importance of improving surveillance and data collection systems in low-income settings to inform more equitable decision-making during future pandemics. The study's insights on transmission dynamics and IFR provide valuable information for pandemic preparedness and response efforts. The finding that a substantial portion of the population remained immune-naïve at the end of 2020 highlights the vulnerability to future waves. The need for improved surveillance and data collection is crucial to ensure that appropriate measures are taken to mitigate future pandemic waves.

This study provides strong evidence that the first COVID-19 wave in Lusaka had a substantial direct impact, resulting in significant excess mortality comparable to that observed in high-income countries. The findings demonstrate the importance of using multiple data sources, including mortality data, to accurately assess the true burden of disease in low-income settings, where underreporting remains a significant challenge. The study's results challenge hypotheses suggesting a unique "Africa paradox" and highlight the need for improved data collection and surveillance systems to ensure more equitable pandemic planning and response. Future research should focus on expanding similar analyses to other regions in Africa and refining methods for estimating excess mortality in data-scarce environments.

The study acknowledges several limitations that could affect the interpretation of results. The inherent volatility and potential underreporting in burial registration data are key limitations. Although statistical methods were used to address this volatility, some uncertainty remains. The reliance on a social contact matrix from a nearby region (Zimbabwe) rather than locally collected data could also introduce some uncertainty into the transmission model. Another limitation is the uncertainty in the true population coverage rate of burial registration, although this uncertainty was assessed through sensitivity analysis. The IFR estimates used in the analysis were derived from global data, and their direct applicability to Lusaka might have some limitations. The study focused on the first COVID-19 wave, which does not necessarily represent subsequent waves.