COVID-19: What is Turkey's true case count?
As noted 8 days ago, Turkey had revised its COVID-19 reporting criteria on November 25 to include asymptomatic cases, increasing its daily reported cases by a factor of 3.6 literally overnight. With cases now averaging 32,000/day and rising, and 925,000 officially reported cases (ranked 16th overall, behind Peru) the country is three days from crossing the million-case threshold, officially.
But we know that this is a spectacular undercount. The question arises, what is the likely actual number of cases?
I'll argue that this is 2.1 million (6th ranked) with a true ncidence of at least 3.5 million, possibly as high as 5 million, based on prior undercounts and observed mortality.
Data are all based on Worldometers unless otherwise noted.
Underreported diagnoses
We can look at the change in reported new daily cases from 24 and 25 November 2020. On the 24th, Turkey reported 7,831 new cases, on the 25th, 28,351. Turkey ranked 24th among countries on 24 November.
Case progression before and after both point was nearly identical --- for the three days preceding the 24th, increasing by an average of 583 new cases/day, by 584 for the three days following the 25th. Expected progression from the 24th to the 25th would have been about 584 cases. Actual was 20,520. The actual value was 3.55 times greater than expected.
This suggests Turkey had previously only been counting 28% of diagnosible cases, or omitting 72% of cases that would be classsified under post-November 25 criteria.
We can adjust the Nov 24 cumulative case total by multiplying the reported 460,916 cases by 3.55, giving about 1.63 million cases. Adding the additional 432,714 actually reported cases since, the current total as of 9 December using current diagnostic criteria would be 2.07 million. With this revised value, Turkey ranks 6th in reported cases overall, following France.
Estimations based on mortality
Another approach is to estimate total cases using a presumed incidence fatality rate (IFR) and average duration from diagnosis to death. This makes several additional assumptions, overstates the number of cases compared to typical diagnostic and testing practices, typically by a factor of 6--8 or more, and may be inaccurate. It remains a useful sanity check.
A key assumption is that Turkey has been consistently reporting all actual COVID-19 deaths. I've some questions about this as there is also a sharp uptick in these numbers, though beginning slightly earlier 17 November. I'm assuming this reflects ground truth and not a reporting change.
With a 9 December report of 15,531 deaths, an assumed IFR of 0.5%, and a two-week average duration from diagnosis to death, my calculation is that the true incidence count as of November 25 was 1.7 million cases. Adding the known 432,714 diagnosed cases since, the true total incidence on December 9 is likely at least 3.5 million cases.
Given that testing still underreports actual incidence by a factor of 6--8, as noted, the total could be approaching 5 million actual infections.
As few countries even approach comprehensive testing, assigning a rank is inappropriate. France, previously mentioned, is highly suspect given a case fatality rate (CFR) of 25%, some fifty times greater than expected, stronly suggesting underreporting of actual cases. I'll not that by total reported deaths, Turkey still ranks only 20th overall, trailing Indonesia, Belgium, and Chile.
Summary
Turkey currently reports 925,342 COVID-19 cases (16th overall by country) where the actual reported value based on curent reporting criteria should be 2.1 million cases (6th overall), and the underlying incidence of all infections is likely 3.5--5 million.
Turkey isn't the only ... data turkey
Whilst there's clearly something rotten in Ankara, Turkey is not the only country whose reported cases fall well short of expected incidence. I may revisit this later.
