Covid-19: What you need to know today

The celebrations can wait. This column has pointed out, more than once, that the trajectory of Covid-19 infections in India is very different from that in any other country. It’s slower; there haven’t been too many deaths; and, at least anecdotally, the proportion of patients requiring ventilator support is lower than that in many other countries. It isn’t clear why. Nor is it clear that this trend will continue (this writer’s hope is that it will). The relatively low number of cases in India — again, as this column has pointed out; when you write a column every day for 30 days, there’s bound to be a bit of repetition — can’t entirely be explained by the country’s low testing. But that stroke of fortune — till science can explain, it is just that — is no excuse not to test more. India has tested 203 people per million of its population. This is lower than even Brazil (296), it is definitely lower than the US (9,866), the UK (6,152), Italy (18,481) and Spain (19,896). A comparative chart put out by the government says that at the time India crossed 5,000 infections and then, 10,000 infections, it had carried out more tests (in absolute terms) than the US, UK, and Italy. That just means that we haven’t learnt from their mistakes. India has been a clear laggard when it comes to testing, just as it has been a leader when it comes to enforcing a lockdown, something that is certain to flatten the curve of infections and also delay the peak. Maharashtra’s aggressive testing, of those at high-risk of infections, is perhaps one reason why the state has the highest number of cases in India — minus a super-event such as the Tablighi Jamaat’s gathering in New Delhi that is responsible for 68% of the city-state’s cases. Since testing resources, like other resources, will always be scarce in India, the protocol will always be skewed towards those at high risk — which will, in turn, mean that at least in the initial phases of testing, states that test more will show more cases (see page 5). Maharashtra’s testing stands at 504 per million of its population as of Thursday (see page 11). India will test more in the coming weeks, though, not just in the so-called hot spots and containment zones writhing hot spots but also in what the health ministry describes as green zones -- parts of the country that have seen either no infections or an insignificant number of them. These tests will be done using rapid testing kits (RTKs) that screen for antibodies and will provide crucial information on the actual spread of Covid-19 in India (see page 1). India has defined a protocol for these tests. A negative test may require a quarantine followed by another antibody test, or an immediate RT-PCR test. A positive test may require isolation and treatment. But countries such as Italy and the UK that want to use the test to decide who gets to return to work will need to do it differently. That’s because of a concept in probability called Bayes’ theorem. Tests for many diseases are defined in terms of sensitivity (probability of an infected person testing positive) and specificity (probability of an uninfected person testing negative). Assuming a sensitivity of 99% and a specificity of 98% (very high, but lower than ICMR’s requirement of 100% on each, or a USFDA certification), and a situation where 10% of the population is infected, it turns out that the probability of a person who tests positive being infected is (drumroll here), not 99% but only 84.61%. This is because the probability is calculated the other way around in this case — starting from a positive test. I will avoid going into the details because I don’t want to make this column any more geeky, but this is the reason why many people worry about using antibody tests as a basis of deciding who gets to be declared immune. Imagine: a person tests positive and thinks she is immune because she is infected; she is declared eligible to return to work; but according to Bayes’ theorem there is only a 84.61% channel of her being infected (and that too only if the test is very good). Many tests have sensitivity and specificity ratings in the mid-90s. The probability that someone who tests positive is actually infected falls sharply in these. I was first pointed in this direction by the erudite twitter feed of Taal Levi, an associate professor of the College of Agricultural Sciences, Oregon State University (@taaltree).