One of the most interesting features of the COVID-19 outbreak is the stark difference between mortality experience in different countries. No simple and plausible explanations that we are aware of have been advanced. Though various hypotheses have been put forward, some more hopeful than others, many display an element of confirmation bias in attempting to locate all differences in nonpharmaceutical intervention approaches.
For each country put forward as an example, usually in some pairwise comparison and with an attendant single cause explanation, there are a host of countries that fail the expectation. We set out to model the disease with every expectation of failure. In choosing variables it was obvious from the outset that there would be contradictory outcomes in the real world. But there were certain variables that appeared to be reliable markers as they had surfaced in much of the media and pre-print papers. These included age, co-morbidity prevalence and the seemingly light population mortality rates in poorer countries than that in richer countries. Even the worst among developing nations—a clutch of countries in equatorial Latin America—have seen lighter overall population mortality than the developed world. Our aim therefore was not to develop the final answer, rather to seek common cause variables that would go some way to providing an explanation and stimulating discussion. There are some very obvious outliers in this theory, not the least of these being Japan.
We test and find wanting the popular notions that lockdowns with their attendant social distancing and various other NPIs confer protection. Health care quality also fails to display any statistical benefit despite the intuitive appeal it has. Similarly, neither dread respiratory disease (ie TB), nor HIV prevalence, have proven the red flags posited by the medical fraternity. Of course, we would have been remiss had we not tested for other plausible concerns such as smoking, cholesterol, child mortality rates, altitude and so on.
While our results explain roughly half of the inter-country variability, they appear to be far more robust than the current explanations in circulation. We are hopeful that other researchers will identify factors that can improve our model.