Page 17 - IJPS-6-2

Garenne

3.10. Analysis of Case Fatality
The gradients in case fatality did not have a common measure with those of incidence: They were much more concentrated.
In particular, there were few regular gradients, and few groups were different from the average. However, most of the
extreme categories (the most advanced) had a lower than average case fatality, which corresponds to the relationship
observed between incidence and development, or between incidence and demographic transition: Very high population
density (13); very high urbanization (16); older age structure (19); very low fertility (14); heavy air traffic (12); population
concentration in islands (13); and for an average of 23 deaths per 1000 cases. However, this effect was not verified for the
two public health parameters: Very low under-five mortality (30) and very high medical density (19).
A multivariate analysis of case fatality was performed in the same way as that for incidence, using a simple linear
regression model. Results revealed only one significant factor: Under-five mortality (P = 0.003) and a factor at borderline
statistical significance: Geographic concentration of the population (P = 0.084) (Table 5). These two factors explained only
16% of the variance between countries. For the first factor, the relationship was straightforward: The higher the level of
mortality, the higher case fatality could be expected because of failures or defects in the health system. The effect of the
second factor indicates that the more concentrated the population, the lower the case fatality. This could be explained by
better access to health care or by a correlation with another factor not taken into account. However, it should be noted that the
relationship with the geographical concentration of the population was not linear and that the islands were treated separately.
No other factor was significant when introducing under-five mortality into the linear regression equation for case fatality.

4. Discussion
For the continent as a whole, the dynamics of the COVID-19 epidemic appeared slower than in Europe or America and they
rather resembled those in the Indian subcontinent. This observation can be related to the levels of economic development:
The more developed countries have more transport, trade, and travelers, are more urbanized, and more densely populated,
all factors contributing to the rapid spread of the virus. For instance, in Europe, Belgium is one of the most urbanized and
densely populated countries and also one of the European countries most affected by COVID-19. Africa and South Asia
have closer and lower levels of development, so one could expect similar and slower COVID-19 epidemics.
Variations in incidence by country were very large in Africa. The main factors seem to be demographic (density,
urbanization) and economic (GDP, air traffic). The fact that these factors point to the transmission of the disease indicates
that data are probably reliable, which indirectly validates the statistics, although they may be questioned in some countries.
These variations between countries seem to be greater than in Europe. However, in Europe too, there were large variations
in incidence, from 11,363 (Luxembourg) to 475 (Slovakia), that is, a ratio of 24 to 1 among the 48 European countries,
excluding the former USSR (WHO database 2020). These large variations in Europe remain poorly explained, and the
unexplained part could be due, in addition to population density and economic development, to random phenomena,
chaotic dynamics, or quality of surveillance systems.
Variations in case fatality were much smaller, and Africa as a whole appeared fairly homogeneous. These results
again point to good data quality, even if it seems surprising that some countries report so few deaths given the number
of reported cases (Ghana, Guinea, Cote d’Ivoire in West Africa; Burundi, Rwanda, and Uganda in Central Africa; and
Botswana and Namibia in Southern Africa). In Europe, there were also large variations in case fatality between countries,
ranging from 16.3 (France) to 0.5 (Iceland) per 1000 reported cases, that is, a ratio of 32 to 1, variations which remain
largely unexplained (WHO database 2020). Variations in case fatality in Africa could be explained in part by levels of
mortality, by the effectiveness of treatments in the case of severe forms of the disease, by better inclusion of mild cases in
the denominator, and perhaps by notification bias, or by other unidentified factors.
The question of the effect of the age structure of African populations remains open. In univariate analysis, more
COVID-19 was found in countries with an older population, but in multivariate analysis, the reverse was found after

Table 5. Results of multivariate analysis of COVID-19 case fatality, 56 African countries and territories.
Variable X Coefficient B Standard error t-test P-value Significance Net effect
i i
Constant 14.868 5.086 2.923 0.005 23.01
Under-five mortality +0.261 0.084 3.123 0.003 *** +7.86
Population concentration index −0.199 0.113 -1.764 0.084 * −4.44
“***”=P<0.01; “**”=P<0.05; “*”=P<0.10; “NS”=Not significant. Coefficients are raw beta coefficients. The net effect was calculated for one standard deviation of each
independent variable, for constant=mean value. Model: Case fatality=Constant+∑ B ×X i
i
i
International Journal of Population Studies | 2020, Volume 6, Issue 2 11