Thread by @HShewade

@BWDDPH @BWDDPG @janisfrayer @pash
I watched the video with interest. I have been following the COVID-19 numbers and here is my take esp on the deaths.

To infer the extent of COVID-19 mortality in India using ‘reported’ COVID-19 deaths is not correct. Let me explain.
(1/n)

@BWDDPH @janisfrayer @pash In a scenario of very high seroprevalence (20%), very low covid case detection rate (3.6%), very poor coverage and quality of routine death surveillance (18% coverage), and low % of deaths in hospitals (34%), for me the reported covid deaths are waay lower
(2/n)

@BWDDPH @janisfrayer @pash Also, There is limited excess deaths data in public domain, there are no post mortem COVID-19 studies (testing all deaths post-mortem in a study area / period for COVID-19), we are not reconciling data from routine death surveillance (however good or bad it is)
(3/n)

@BWDDPH @janisfrayer @pash Undercounting is a major issue. My guestimate is that 2-3 is the factor due to error in assigning cause of death - this has been covered in detail by Indian media. Do note that If not tested before death, we are not testing for COVID-19 post-death
(4/n)

@BWDDPH @janisfrayer @pash undercounting factor should be 5 times if we consider that only 18% deaths in India undergo death registration along with cause of death certification.

UP/Bihar/Jharkhand (30% of Indian pop) registers only 35-60% of its deaths

Overall undercounting factor 6-7
(5/n)

@BWDDPH @janisfrayer @pash Let me give an example from another disease programme. TB is one the most well designed disease control programmes in India. Annual reported deaths in India are 80000. Annual estimated deaths are 450000 (5.6 times). We never use 80000 to compare with other countries.
(6/n)

@BWDDPH @janisfrayer @pash To summarise, cause-specific deaths reported from India are way lower than the true picture and cannot be used to compare the cause-specific deaths reported from countries with near-perfect death surveillance.

This is also a generic rule.
(7/n)

@BWDDPH @janisfrayer @pash Now the disease has spread in rural areas (less dense) in India, where testing, death surveillance and hospital admissions of severe cases is lower than urban areas. As we do not have disaggregated data on testing, cases, death, we just don’t know the extent.
(8/n)

@BWDDPH @janisfrayer @pash 60% of India resides in villages (less dense). Even if the true cumulative covid deaths were 500 per million in rural, that would mean 5 deaths in five large villages(each village 2000 people) over 12 months. Whether reported or not, these will not cause panic
(9/n)

@BWDDPH @janisfrayer @pash Two relevant quotes
Quote 1/2
“Absence of data does not mean absence of event” (with reference to using reported deaths from a country with poor routine death surveillance to infer and compare)

(10/n)

@BWDDPH @janisfrayer @pash “Repeat a lie often enough and it becomes the truth”

(Govt in press conf (again n again n again) saying that we have low covid deaths in India - based on low reported deaths. This is despite knowing the facts shared in tweet 2/n and 3/n)

(11/n)

@BWDDPH @janisfrayer @pash I urge that we take take out time and go through these two tweet threads

Tweet thread 1/2

https://twitter.com/hshewade/status/1356831895380578310?s=21

(12/n)

@BWDDPH @janisfrayer @pash Tweet thread 2/2

https://twitter.com/hshewade/status/1362614016305106947?s=21

(n/n)