key: cord-0945641-4eiyj5bs
authors: Fokoua-Maxime, C. D.; Bellouche, Y.; Tchigui-Ariolle, D. N.; Tchato-Yann, T. L.; Choukem, S. P.
title: The spread and burden of the COVID-19 pandemic in sub-Saharan Africa: comparison between predictions and actual data and lessons learned.
date: 2022-05-05
journal: nan
DOI: 10.1101/2022.05.04.22274692
sha: 7480fe4edd129a0ebba21bc96deb7461f7939571
doc_id: 945641
cord_uid: 4eiyj5bs

Introduction Sub-Saharan Africa (SSA) was predicted to be severely affected by the coronavirus disease 2019 (COVID-19) pandemic, but the actual data seem to have contradicted these forecasts. This study attempted to verify this observation by comparing predictions against actual data on the spread and burden of the COVID-19 pandemic in SSA. Methods Focused on the period from March 1 st to September 30 th , 2020, we compared (1) the predicted interval dates when each SSA country would report 1 000 and 10 000 COVID-19 cases, to the actual dates when these numbers were attained, as well as (2) the daily number of predicted versus actual COVID-19 cases. Further, we calculated the case fatality ratio of the COVID-19 infection in SSA, and the correlation coefficient between the weekly average number of confirmed COVID-19 cases reported by each country and the weekly average stringency index of its anti-COVID-19 policy measures. Results 84.61% (33) and 100% (39) of the 39 SSA countries for which predictions were made did not reach a total of 1 000 and 10 000 confirmed COVID-19 cases at the predicted interval dates. The daily number of confirmed COVID-19 cases was lower than the one projected for all SSA countries. The case fatality ratio of the COVID-19 infection in SSA was 3.42%. Among the 44 SSA countries for which the correlation could be estimated, it was negative for 17 (38.6%) of them. Conclusions The natural characteristics of SSA and the public health measures implemented might partly explain that the actual data were lower than the predictions on the COVID-19 pandemic in SSA, but the low case ascertainment and the numerous asymptomatic cases did significantly influence this observation.

The Coronavirus disease 2019 outbreak was classified as a pandemic on March 11 th , 2020, owing we presented the total number of confirmed COVID-19 cases reported at the date this paper was submitted for 113 peer review and publication. Of note, Pearson et al. did not compute predictions for Botswana, Burundi, belong to SSA. In a sake of completeness, we retrieved the total number of confirmed COVID-19 cases reported by these countries on April 30 th and May 31 st , 2020, respectively. These dates were chosen because Person et al.

had predicted that most SSA countries would report a total of 1 000 COVID-19 cases by the end of April and 10 118 000 cases couple of weeks later (1). In line with our primary objective of comparing predictions to actual data, 119 we further scavenged the WHO SITREPs to find the actual dates at which these specific countries reported these 120 numbers of cases.

To complement our research, we also evaluated the impact of the anti-COVID-19 policy measures implemented.

The latter were valued through the Stringency Index (SI). The SI is a metric conceived as part of the Oxford 123 COVID-19 Government Response Tracker (OxCGRT) project and has been previously described in details (13).

In brief, the OxCGRT project is an endeavor undertaken by Oxford University to measure the strictness of the public gatherings, closures of public transport, stay-at-home requirements, public information campaigns, Then, we calculated the case fatality ratio of the COVID-19 infection (proportion of deaths among all the identified confirmed cases of ) in SSA during the period spanning from March 1 st to September weekly average number of confirmed COVID-19 cases at time t + 14 days, along with the 95% CI of the said 146 correlation coefficient. We considered a 14-day lag between the 2 metrics because it takes approximately one 147 incubation period to see the effects of newly implemented anti-COVID-19 policy measures (15). Further, we 148 used weekly averages instead of daily numbers to account for the delays in case ascertainment and reporting.

Finally, to prevent any violation of the assumption of independence required for any correlation analyses, we 

the continental level, the total number of actual confirmed COVID-19 cases was 1 126 341, which is far lower case fatality ratio of the COVID-19 infection in SSA from March to September 2020 was approximately 3.42%. 

Overall, most SSA countries did not report 1 000 and 10 000 cases at the predicted dates, and the actual 184 numbers of COVID-19 cases were lower than those predicted. These results might be explained by the 185 limitations of the statistical models which yielded these predictions. Additionally, specific local population and 186 environmental characteristics as well as the low case ascertainment might have had a mitigating effect.

The prediction model of the MRC Centre for Global Infectious Disease Analysis at Imperial College London 188 was built on estimates of severity obtained from data from China and Europe, and model parameters obtained 189 from data from China and the United Kingdom (11). On the other hand, Pearson et al. considered that the 190 reproductive number R (which is the number of ancillary cases that one case would generate if in contact with a 191 completely susceptible population (18)) would be 2, that the dispersion estimate k (which is the variance of R 192 over the mean of R and quantifies whether a set of observed cases are clustered or dispersed when compared to 193 cases following a standard negative binomial distribution (19)) would be 0.58, and that the serial interval (which 194 is the time that elapses between two consecutive cases of an infectious disease (20)) would be normally . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 5, 2022. 

In SSA, specific population and environmental characteristics also mitigated the epidemic pace of the COVID-and regions (14) . During the last 50 years, the increase in road constructions across SSA significantly improved 255 the connectivity and reduced the travel time between localities (14). However, the daily commute is not custom 256 in SSA. Therefore, the spread of the COVID-19 infection might have been limited to narrow geographic areas.

The transmissibility of the COVID-19 infection is also impacted by the prevalence of severe cases. Indeed, 258 severe cases are more prone to transmit the COVID-19 infection because of their higher viral load, and they are 259 also more prone to die of the infection (37). The case fatality ratio of COVID-19 in SSA was approximately 260 3.42%, which is much lower compared than the case fatality ratio reported on other continents (38-40).

Evidence suggest that the lower susceptibility of SSA populations to COVID-19 might have resulted in most 262 cases being asymptomatic (41), meaning less contagious. Therefore, SSA might have hosted mostly less severe 263 cases, which translated into a lower probability of transmission of the COVID-19 infection in the region.

The actual number of COVID-19 cases in SSA might also be this low because not all cases were reported. 279 Therefore, a cross-sectional study was indicated. Another limitation is the small number of countries (17 out of 280 44) for which there was a negative correlation between the weekly average number of COVID-19 cases and the 281 weekly average SI, which questions the true efficacy of the prevention efforts deployed inside SSA countries.

Of note, the scientists who designed the index cautioned that it does not measure the actual level of 283 implementation of the anti-COVID-19 policy measures which make the index (13). Nevertheless, studies had . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 5, 2022. demonstrated the efficacy of non-pharmacologic measures against COVID-19, and the SI was the sole metric available in that regard for our use. Finally, the vast array of the potential contributors to the spread of COVIDhealth data, point to the necessity to exercise caution in the interpretation of these results which were pooled 288 across several very dissimilar geographic and socioeconomic settings.

Our research does have several strengths. It responds to the long-lasting call for a comprehensive analysis of the 290 differences between the predictions and the actual data on the COVID-19 pandemic in SSA. In addition, our 291 analysis also included the countries for which predictions were not made. Finally, we discussed the population CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 5, 2022. (10 129) 28 Apr 2020 -11 Jun 2020 a Originates from the World Health Organization (WHO) situation reports (SITREPs) and it is the date at which each country reported a cumulative total of 1000 confirmed COVID-19 cases ; in instances where this exact number could not be found is reported the earliest date at which this number was surpassed, and in instances where this number was never reached is reported the date at which this paper was submitted for peer review and publication. b Originates from the publication by Pearson et al., 1 and it is the 95% confidence interval of the date at which each country was predicted to have reported a cumulative total of 1000 confirmed COVID-19 cases. c Originates from the WHO SITREPs and it is the date at which each country reported a cumulative total of 10 000 confirmed COVID-19 cases ; in instances where this exact number could not be found is reported the earliest date at which this number was surpassed, and in instances where this number was never reached is reported the date at which this paper was submitted for peer review and publication. d Originates from the publication by Pearson et al., 1 and it is the 95% confidence interval of the date at which each country was predicted to have reported a total of 10 000 confirmed COVID-19 cases. CAR = Central African Republic. DRC = Democratic Republic of Congo. *Countries which fulfilled the predictions made by Pearson et al. 1 # The specific date at which Tanzania reported 10 000 COVID-19 cases could not be identified. Table 1 : Predictions versus actual data on the spread of COVID-19 in SSA. 1b. Total number of confirmed COVID-19 cases reported on 30 April 2020 and 31 May 2020 by the sub-Saharan African countries which were not included in the publication of Pearson et al., 1 and dates at which these countries reported a cumulative of total of 1000 and 10 000 confirmed COVID-19 cases in the World Health Organization Situation Reports. a Originates from the World Health Organization (WHO) situation reports (SITREPs) and is the cumulative total of confirmed COVID-19 cases reported by each country as of 30 April 2020. b Originates from the WHO SITREPs and it is the date at which each country reported a cumulative total of 1000 confirmed COVID-19 cases ; in instances where this exact number could not be found is reported the earliest date at which this number was surpassed, and in instances where this number was never reached is reported the date at which this paper was submitted for peer review and publication. c Originates from the WHO SITREPs and is the cumulative total of confirmed COVID-19 cases reported by each country as of 31 May 2020. d Originates from the WHO SITREPs and it is the date at which each country reported a cumulative total of 10 000 confirmed COVID-19 cases ; in instances where this exact number could not be found is reported the earliest date at which this number was surpassed, and in instances where this number was never reached is reported the date at which this paper was submitted for peer review and publication. ST . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 5, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 by each country and the stringency index of its anti-COVID-19 policy measures. b Repeated measures correlation coefficient indicating a negative correlation between the weekly average number of confirmed COVID-19 cases reported by each country and the weekly average stringency index of its anti-COVID-19 policy measures. c The repeated measures correlation coefficient between the weekly average number of confirmed COVID-19 cases and the weekly average stringency index anti-COVID-19 policy measures as well as its 95% confidence interval were not estimated because data on the stringency index for these countries are not available in the Oxford Coronavirus Government Response Tracker (OxCGRT)

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 5, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 5, 2022. Generalisability 21 Discuss the generalisability (external validity) of the study results 9

Funding 22

Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based 10 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 5, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 

Group CC 19 working

We thank Professor Martin Tenniswood for his comments and suggestions.

 CDFM