Corona Daily 468: With a Pinch of Salt

Are the numbers of infected and dead that we check every morning overstated or understated?

First, let me confirm both Johns Hopkins and Worldometer are doing their job professionally. They are independent, the staff competent and hard working, the processes automated, data transparent (both mention sources, Hopkins includes raw data). They cover the world as best as they can.

We must remember, though, there is a human being at the other end. A Mafiosi killing could be mistaken for a Covid-19 death. A patient dying alone at home may not go reported. It may simply be a typo.

The number of total cases is not the total number of infected. Testing is low. Thousands currently under house arrest may be positive if tested. This figure is certainly understated.

Reporting recovery in cases is not mandatory anywhere. People recovering in hospitals routinely go home with nobody closing the case. (Closed cases= recovered + dead). This number is grossly understated, giving a ridiculously high (19%) mortality percent in closed cases.

Deaths are the figure we believe in. There is something definitive about death. Little space for doubt, you would think. But that is not so. With resource constraints, autopsies are rarely performed. Covid-19 cases are classified into three categories: Suspect, Probable and confirmed. Suspect are those dying with Covid-like symptoms but no tests. Probable are those without tests but were in contact with someone who tested positive. Confirmed are those who tested positive. The USA allowed probable cases to be included along with confirmed cases as late as April 14.

Add to that the jolts researchers receive. On 17 April, China reported 1290 Feb/March deaths from Wuhan with retrospective effect. A day earlier New York City reported a decrease of 145 deaths. On 3 April, the French ministry reported 17,827 additional cases and 532 additional deaths from nursing homes. They had forgotten about nursing homes. With few resources, the solitary researcher somehow makes the necessary adjustments.

There is always a time lag. A doctor diagnoses a case. He submits the report to the local health department. The health department must enter the data in the computer. Most of them don’t work weekends. (A seasoned statistician knows why Monday is the most fatal day). That data then travels to the state ministry and finally to the federal ministry where it is aggregated. The ministry then sends it to WHO.

The new cases or new deaths that your TV reported this morning may have happened any time during the last four weeks.

Finally, for a global effort, the researcher must know the numbering conventions. The number 9,876,543.21 (comma: UK, US) looks like 9.876.543,21 (full stop in most of Europe), 98,76,543.21 (India, two spaces), 9’876’543.21 (apostrophe: Switzerland) and 987,6543.21 (Mandarin, four spaces). India counts in lakh (100,000) and crore (10 million). An inexperienced researcher can easily kill ten times more or fewer Indians.

*****

In short, we need another cross-check to answer the question. I will do that tomorrow by discussing the notion of ‘excess mortality’.

Ravi