COVID-19 pandemic in the African continent: forecasts of cumulative cases, new infections, and mortality

Yohannes Kinfu; Uzma Alam; Tom Achoki

The data on prevalence of HIV and asthma, as well as on socio-demographic index and the index of health care access and quality, were from Institute of Health Metrics and Evaluation (IHME). (21) (22) (23) The data on urbanization were obtained from the Population Reference Bureau, (24) while average A f r i c a a n d t h e C O V I D -1 9 p a n d e m i c household size, age profile, and data on the total population used for the calculation of rates were all from the UN Population Division. (25) The index of adherence to international health regulations was accessed from the World Health Organization (WHO), while the number of confirmed cases was from the John Hopkins database of COVID-19. (26 -27) The 11 variables selected for analyses were chosen out of an initial set of over 20 variables that included the proportion of households with at least one member aged 60 years or over, the proportion of households with at least one member aged 65 years or over, the World Bank's governance index, as well as the prevalence of TB, diabetes and malaria. STATA's Lasso software for model selection was used to narrow down the list of covariates. (28) Furthermore, out of the global database covering the 193 countries that we captured as an input in our analyses, for a few countries, data were unavailable for selected variables, namely air traffic (25 countries), adherence to international health regulation (20 countries), and household size (24 countries). To bridge the gap, we performed multiple imputation procedures, particularly by running 1000 imputations and 99 000 iterations using STATA's multiple imputation routine. (29-30) The resulting median values from the imputation exercise were subsequently used to estimate the prediction model specified earlier. We transformed all the regressors as well as the dependent variables to a log-scale before they were put into the model.