Wuhan Pandemic, Fatality Rate and Statistical Determinants

The post “Investigating Wuhan Pandemic Spread and Determinants” discussed the statistical determinants of the Wuhan pandemic. The analysis used the death incidence as the dependent variable and tested its susceptibility to changes in different independent variables. A couple of interesting pointers emerged. The first was the significant differences in death incidence between advanced economies and emerging or underdeveloped economies. One of the possible explanations was the probability of deaths being attributed to several other causes rather than Chinese virus induced Wuhan flu. There was a possibility that cases were being under-reported and deaths occurring were delinked from the COVID-19.

Secondly, contrary to arguments by experts, there was no significance of BCG vaccine programmes on the death incidence. This was little puzzling given the opinion of experts. In the meantime there are anecdotal notes about the impact of temperature on the incidence of coronavirus. It is believed that higher temperatures reduce the spread of the disease. Yet there were some reports, rather than higher temperature it is the sunshine that is critical to the spread of the disease or lack of it. Higher sunshine countries are likely to be less susceptible to the disease.

There was also the question about the linkages of incidence to the population. Higher densities are expected to increase the spread of the disease in contrast to those regions with low population densities. In Italy, mortality was linked to the old age population in the country. Higher incidence of old age population coupled with the co-morbidities they usually carry caused the higher death rate.

These additional factors need investigation. Hence there was a refined analysis incorporating these new variables. There was further need to refine the death incidence to account for differences in advanced and emerging economies. Hence death incidence was replaced by case fatality ratio. The case fatality ratio (CFR) is the ratio of deaths to total positive patients. The data was collected for the latest date available in each country from databases like John Hopkins and others. This was tested to its susceptibility to factors enumerated above in addition to the factors tested in the previous analysis, the link of which is indicated above.

The variable CFR was used to denote case fatality ratio and thus the dependent variable. The following variables were used to denote the independent variables.

ECOTYPE- a dummy to denote an economy is advanced or not

MALREG- a dummy to indicate a presence or lack of it of malaraial region

BCGREG- dummy to indicate countries which have active BCG vaccine programs

BCGMAL- an interaction variable to test the anecdotal belief that countries with malarial region are not impacted by the virus given the prevalence of use of anti-malarial drugs which are now being recommended as cure or prophylaxis for Wuhan flu.

TESTINCI- number of tests conducted per million of population

POPINCI- how many square kms per million people in a country. The inverse of the population density.

TEMP- average temperature of a country

AGE- median age of the country

SUNSHINE- dummy indicating the country is sunshine rich or lack of the same

The data for the above was sourced from multiple databases. The age was sourced from UN and World Bank databases on population and validated. Average temperature of the country was sourced again from different databases. The data for sunshine rich countries was sourced from International Solar Alliance database and its membership was used as proxy for many countries to indicate they are sunshine rich. The population incidence was calculated after getting the data on density again from sources like the UN population database. The database on testing pattern in different countries was taken from Worldometer which tracks the cases of coronavirus from different sources.

The regression statistics are indicated in the below table.

Regression Statistics
Multiple R	0.387466
R Square	0.15013
Adjusted R Square	0.085854
Standard Error	0.035907
Observations	129

R-square is at 0.15 which is small but either way, some pointers can be obtained by looking the significant p for different variables.

	Coefficients	P-value
Intercept	0.050423	1.43E-05
ECOTYPE	0.018445	0.056282
MALREG	-0.00657	0.543714
BCGREG	-0.02711	0.004179
BCGMAL	0.023	0.107224
TESTINCI	-7.7E-07	0.002444
POPINCI	-2.2E-08	0.585027
TEMP	0.004278	0.679093
AGE	0.000274	0.975632
SUNSHINE	-0.00348	0.715449

The table above indicates p-significance for TESTINCI which is perhaps obvious. What is however interesting is the significance of ECOTYPE which is at 95% level compared to 99% level before. The co-efficient too is small compared to the data discussed in the previous post.  There is p-significance for BCGREG at 99% which is quite interesting. There is no significance for the rest of the factors.

As one observes the findings. It is clear that there seems to be no evidence at least on the first look that the rising temperatures or the presence of sunshine will slow the down the spread of the disease as measured by the fatality rate. There might be more to investigate but it appears to be a myth at this moment. The death incidence might be high in the advanced economies but that could be due to the higher case reporting. The fatality rate is less significant when compared between developed and developing countries. Yet, what cannot be ruled out is the disease has struck the rich countries more than the developing ones. It is also perhaps due to inbuilt viral immunity in the developing world thanks to the living conditions that might have created some differences in fatality rates. However, the difference is not wide as was being in reported in the previous analysis.

BCG vaccine seems to play a role in reducing the fatality rate. While the death incidence might not be susceptible to the BCG vaccine programmes, the vaccine programmes certainly reduces the fatality rate. As the countries grapple to contain the fatality rate especially in the US, Italy, Spain, France, Britain among others, it perhaps would make sense to have a BCG vaccination up and running at least in the high risk groups to reduce the possible fatality rates. The study might be a back of envelope mode or exploratory model but the significance level of 99% cannot be discounted. The coefficient is small, yet there would be perhaps a benefits relatively more than the costs as one evaluates the BCG programme.

There is also no significance of age or population density on the fatality rate. Therefore it can be safely set aside the impact of population and its characteristics on the fatality rate. However, there might be a need to decompose deeper into those micro-characteristics to generate some additional insights. The contribution of this analysis is to examine further and dig little deeper the statistical models to predict the trends in coronavirus.