Some thoughts on State-Wise Excess Deaths

If our eminent intellectuals and politicians have taught us anything, it is that one can have opinions on anything.

Introduction

In this article, I shall look into the COVID excess death analysis carried-out by The Hindu. This is my third article on excess deaths. 

As mentioned above, I am no expert and my view can be considered amateurish at best.        

Analysis of The Hindu 

It is creditable that The Hindu does this investigation into excess deaths due to the COVID-19 pandemic state by state. Not only do they do this, they put the monthly death data for the respective state so that the interested people have the scope of deep diving into the numbers.

I have defined excess deaths in a previous article and am not going into its details here. It would suffice to say that excess deaths can be considered equal to total COVID deaths for the respective state.

The below table summarizes most of their investigation till date. As can be observed, they have calculated excess deaths for each state using three methods.

  1. Average of 2018 and 2019 as the baseline number
  2. Average of 2015 to 2019 as baseline number
  3. Linear growth of 2015-19
 


  

Since the table is completely filled for the first scenario, it seems (I may be wrong) that The Hindu believes that this would be the most appropriate number. Is it so? Let's see.    

Issue with averages



    The above graph shows the average annual growth rate of death registrations (the analysis is restricted to those states covered by The Hindu article) from 2015 to 2019. It is pretty obvious that the death registrations have increased for all the states during this period. So, taking average of 2018 and 2019 deaths as baseline is incorrect. 

    One more observation, among those states which have reached near 100% or more of death registration, the annual growth rate is comparatively lower, but still not stagnant. In the graph, the bottom 6 states registered near 100% or more deaths by 2015 or 2016. Their growth rate is less than 3% (except for Karnataka). On the other hand, the top 4 states which did not have near 100% registration by 2016 showed much higher growth rate (with the exception of Himachal Pradesh). Even for states which have achieved 100%+ registrations, the death registrations have not stagnated. So, growth should be factored for even such states when we are computing excess deaths.

    Dr. Shamika Ravi in one of her articles in Hindustan Times says: Second, the level of registration, which is the ratio of registered deaths from CRS to the estimated deaths from SRS, is not uniform across the country and over time. It varied widely from 21% in Manipur to 463% in Chandigarh in 2019. There are several possible reasons for this, perhaps the most important being that people from states with inadequate health facilities (typically rural) travel to regions with better facilities located primarily in urban areas. In some cases, this might result in a casualty, which is registered where it happens and not where the person resided. Therefore, CRS might report a higher incidence of death in one region versus a lower incidence in other areas.

    Anyway, it should be clear by now that using averages to compute excess deaths exaggerates the actual number.

    Adjusting the excess deaths: Step 1

    From various articles of The Hindu on excess deaths in each state, I was able to source the difference between deaths registered in 2018 and 2019. On an approximate basis, for all the above states combined, the 2019 deaths are higher than 2018 deaths by 2.4 lakhs. Additionally, The Hindu does its analysis for a period of 14-15 months (from April 2020 to May-June 2021). After factoring in the same as well, the total excess deaths reduce from 1.22 mn (as per The Hindu) to 1.08 mn. Also, the multiple (excess deaths/ reported COVID deaths) reduces from 6.27 to 5.55. This is without factoring in any growth. 

    As an aside, The Hindu does not consider the backlog deaths added by Maharashtra in June and July 2021 in their reported deaths. This would further reduce the multiple for Maharashtra.     

    Analysis of infection fatality rate

    Infection fatality rate (IFR) is the actual number of deaths divided by total number of cases. The total number of cases can be obtained from Sero Survey data collected.

    While it has been mentioned that State-wise data may not be representative since it covers only 70 districts across the country and does not include most of the hotspots (from a Swarajyamag article). Until we get better state-wise sero surveys, we do not have an option but to rely on the current state-wise sero survey data available with us. 

    Since the Sero Survey was conducted in June-July, I have taken 30 June 2021 as the cut-off date and used the excess death multiple state-wise to arrive at the revised excess deaths as of 30 June 2021 (since most of the data as per The Hindu is available up to May 2021 only, this adjustment should be carried out). For considering the excess death, after adjusting the excess deaths as mentioned above and basis the state-wise sero survey data, the IFR state-wise has been shown below.   


       



    Adjusting the excess deaths: Step 2

    Out of the 10 states, Kerala, HP, Maharashtra and AP are clear outliers for IFR. Other states are close to the median line of 0.31%. This points out another anomaly in The Hindu's excess deaths computation where in 4 states show significant divergence from the median IFR. While Kerala and HP show significantly lower IFR, AP and Maharashtra show significantly higher IFR. It is difficult to believe that AP and Maharashtra with significantly better healthcare facilities compared to rest of India would have an IFR almost 2x compared to the median. Hence, if we adjust the excess deaths for all these 4 states by the Median IFR, the total excess deaths would be 1.14 mn while the reported number is 2.71 lacs. This would translate to a multiplication factor of 4.22 compared to a multiple of 6.27 as per The Hindu (a 33% drop).

    Conclusions

    We have seen that there are some issues with the excess deaths calculated by The Hindu. One is that the baseline number they have used is average of 2018 and 2019 while it is perfectly clear that there has been a growth in the death registrations across all these states from 2015 onwards even if the registration is more than 100%. Even if we just change the base line from 2018-2019 average to 2019, there is a significant drop in the excess death multiple. If we factor growth the multiple would drop further. This has not been done in the current article. 

    Further, the IFR seems off the mark for four of the 10 states. If we adjust for IFR as well, the excess death multiple drops by 33% to 4.22 from 6.27.    

    This article has been written to indicate some issues with the excess deaths computations and its purpose is neither to deny the good work done by the media nor to suggest that there has been no underreporting of COVID deaths.      

    Sources

    https://www.thehindu.com/news/national/excess-deaths-in-maharashtra-were-at-least-3-times-the-official-covid-toll/article35708965.ece

    https://www.hindustantimes.com/india-news/covid19-seroprevalence-in-fully-immunised-people-higher-than-unvaccinated-ones-says-icmr-101626787680366.html

    https://www.hindustantimes.com/india-news/kerala-scores-lowest-in-icmr-s-seroprevalence-survey-madhya-pradesh-highest-explained-101627478969937.html

    https://www.medrxiv.org/content/10.1101/2021.06.02.21258203v1.full.pdf

    https://theprint.in/health/4th-sero-survey-finds-2-of-3-indians-with-covid-antibodies-but-still-avoid-crowds-icmr-warns/699600/

    https://www.hindustantimes.com/opinion/counting-deaths-in-india-is-difficult-101626273326958.html

    https://swarajyamag.com/politics/the-myth-of-the-kerala-model

    Comments

    Post a Comment

    Popular posts from this blog

    Alternate methods of benchmarking consumption expenditure

    Image
    Section 1: Introduction The recently released fact sheet of Household Consumption Expenditure Survey (HCES) 2022-23 (Source: S1) has created quite a stir with one set of academics/ economists saying it is not comparable to the earlier surveys due to change in the methodology and other reasons while there is another set which argued the opposite. Though not comprehensive, some of the concerns have been addressed in an earlier article titled “ Debunking the K-shaped Recovery Theory ” (Source: S2) . However, in this article I shall not directly address the concerns raised on the issue of comparability with the earlier surveys but rather look at alternate indicators which may be used to benchmark the growth in consumption expenditure from 2011-12 to 2022-23. Considering the HCES is a relatively rare occurrence (there is one more going on for the year 2023-24 but it may not happen post that immediately), two benchmark indicators have been used for measuring consumption expenditure using ...
    Read more

    Missing the Rich - Analysis of NSSO Data

    Image
    Summary Some of the Government statistical data published (or leaked) in 2019 created quite a stir among the academic community and even among the general public. One was the Periodic Labour Force Survey of 2017-18 which put India's unemployment rates at 6.1%, the highest in 4 decades and the other was the Consumer Expenditure Survey of 2017-18 (whose results got leaked) which showed a dip in consumption expenditure from 2011-12 to 2017-18. The sample surveys conducted by National Sample Survey Office (NSSO) are generally regarded to provide the most authentic data about the social and economic conditions of the Indian population. These surveys cover a wide range of issues like employment, consumer expenditure, agricultural incomes, spending on education, spending on healthcare, time-use etc. Out of these, the surveys on employment (officially called the Periodic Labour Force Survey (PLFS) from 2017-18 onwards and till 2011-12 it was called Employment-Unemployment Survey (EUS)) and...
    Read more

    Excess deaths in India till May 2021 as per HMIS

    Image
    If our eminent intellectuals and politicians have taught us anything, it is that one can have opinions on anything. Introduction This article is a continuation of the previous article on excess deaths in UP. In this article, I shall be looking at the excess deaths across the country as per National Health Mission's NHM's data. I have taken data from an article on IndiaSpend though my conclusions may be different. Kudos to the journalists who took this data out and more power to them.      As a caveat, I would be using rounded numbers for ease.  What is NHM? As per the HMIS website: National Health Mission (NHM)  is a flagship programme of the Government of India to address the health needs of under-served rural areas and health concerns of the urban poor population. Now coming to HMIS:  Health Management Information System (HMIS) is a Government to Government (G2G) web-based Monitoring Information System that has been put in place by Ministry of Health ...
    Read more