COVID-19 Cases in India: is there an exponential increase?
The front-page news has a time-series graph of COVID-19 cases in India, from March 15th to June 27th, showing total cases increasing in an exponential way from 110 to 508,953. The headline reads: “400,000 to 500,000 in a week”. Frightening!
Has the number of people infected with COVID-19 increased from 400,000 to 500,000 in a week?
What they fail to mention is that tests have increased, and cases track increased testing. By not presenting the test data, it is removed from readers’ cognition. The misleading default assumption is that nothing else changed (that there was the same amount of testing every week). Such suppression of critical information by the analyst is a tool of disinformation.
Does anyone believe that nobody is missed out from this count of COVID-19 cases? It strains my credulity, and I cannot locate anyone who thinks that. Everybody knows that the actual number of cases is higher.
What is the true number of COVID-19 cases in India?
Let’s look at the percentage of positive tests. To reduce errors caused by sampling differences on a daily basis I have smoothed tests and cases with a fifteen day moving average. The percentage positive has risen from four percent to eight percent in the past two months.
It is possible that we have conducted enough testing that the percentage of positive tests truly represents the population. A random-sample test strategy would yield that result. Nobody in India claims that we’re doing random sampling for COVID-19 tests. By an accumulation of offsetting sampling errors over several locations and times we could have inadvertently arrived at roughly the same spot as random sampling would provide.
If the test positive percentage matches population infected percentage, then we have 111 million cases (8.0% of India’s 1,380 million population). Not half a million.
Let’s say we don’t think the test positive percentage matches population infected percentage. Some argue we over-test the infected, so the population infected percentage is lower than the test positive percentage. Others insist that we over-test the uninfected, so the population infected percentage is higher than the test positive percentage. The truth is not known because there are no published results of random testing. Here are three scenarios for infection percentage and test cases:
- Infected Population = Test Positive (8%): 111 million cases
- Infected Population = ½ of Test Positive (4%): 55 million cases
- Infected Population = 1½ times of Test Positive (12%): 166 million cases
We see a gigantic misunderstanding of true counts of COVID-19 cases: the estimate of 508 thousand is off by orders of magnitude. This mistake leads to errors in judgment. I urge rationality. And remember that the term “rationality” has its roots in our ability to use ratios and proportions.
The correct headline is far less arresting: “India’s COVID-19 case positive rate increased from 7.5% to 8% last week. This could have been caused by a changed testing strategy, so we’re not sure what it signifies.”