How many dead are there, really?
Delhi’s crematoriums running out of wood, underreported deaths from Bihar and Gujarat and Uttar Pradesh — newspapers and websites (including HT) have been peppered with anecdotal reports over the past few weeks that show that many states are under-reporting the number of coronavirus disease or Covid-19 deaths. In the absence of reliable, timely, and comprehensive death records, at the local level, aggregating up to the national — and the problem is on all three fronts, reliability, timeliness, and comprehensiveness — there has been a lot of speculation, and scientific exploration, some reasoned, others not-so, on the actual number of deaths. Is it twice the reported number, five times, thirty?
This instalment of my column will try and answer that. It’s a question I have sought to answer time and again, over the past 14 months in this column, but never in as much detail as I do here.
There are two parts to the exercise and both involve assumptions.
The first is the actual number of Covid-19 infections in the country. There’s been quite a bit of research on this — in the US, France, Austria, and several other countries. Most suggest that the cases are likely underreported by 75-90%. Or, at the upper bound, the number of actual infections is 10 times the reported number of cases. India reported 18.75 million cases of Covid-19 till Thursday night. That translates into 187.5 million overall infections. What if the multiple is 15? After all, India is a large country, with a huge rural population, and with very poor testing infrastructure in many predominantly rural states. Then the number of infections is 281.25 million.
The second is the infection fatality rate for Covid-19. Again, there is a lot of research on this, including by Murad Banaji (on the pre print server medRxiv, and the paper is yet to be peer reviewed, but it looks solid) on the infection fatality rate in Mumbai. He arrives at a number of 0.23% (median). Interestingly, this number is exactly the same arrived at by a meta-study of data from 51 regions carried out by John P A Ioannidis at Stanford University’s Meta-Research Innovation Centre (and published in a bulletin of WHO). Sure, Ioannidis’s research may be unrepresentative because it was carried out at a time when the world was still learning how to treat Covid-19 (progressive case fatality rates have become better, which implies that progressive infection fatality rates also might have). So, it’s possible that the actual infection fatality rate is lower, say 0.115%.
{{/usCountry}}The second is the infection fatality rate for Covid-19. Again, there is a lot of research on this, including by Murad Banaji (on the pre print server medRxiv, and the paper is yet to be peer reviewed, but it looks solid) on the infection fatality rate in Mumbai. He arrives at a number of 0.23% (median). Interestingly, this number is exactly the same arrived at by a meta-study of data from 51 regions carried out by John P A Ioannidis at Stanford University’s Meta-Research Innovation Centre (and published in a bulletin of WHO). Sure, Ioannidis’s research may be unrepresentative because it was carried out at a time when the world was still learning how to treat Covid-19 (progressive case fatality rates have become better, which implies that progressive infection fatality rates also might have). So, it’s possible that the actual infection fatality rate is lower, say 0.115%.
{{/usCountry}}Now, assuming these two sets of numbers, it’s possible to calculate the actual number of dead in India.
With only one in 10 cases reported, and a 0.115% infection fatality rate, this is 215,625 deaths.
With one in 10 cases reported, and a 0.23% infection fatality rate, this is 430,675.
With one in 15 cases reported and a 0.115% infection fatality rate, this is 323,438.
And with one in 15 cases reported and a 0.23% infection fatality rate, this is 646,875.
That gives us a range of 215,625 to 646,875 overall deaths. My own estimate is that the real number will likely be around the midpoint of this range — 431,250.
India ended Thursday with 208,332 recorded deaths.
This would suggest that around half the deaths in India are unreported, which is plausible. That would also take India’s case fatality rate to 2.3%, which is definitely more believable than the current number.
Reports suggesting the actual number of deaths is 5x or 10x or maybe even more are likely exaggerations and over-exaggerations, although even the 2x number I have arrived at above is worrying — but not surprising.