Can we use HMIS numbers to estimate unrecorded Covid-19 deaths?
On July 9, the Health Management Information System (HMIS), a data dashboard of the National Health Mission (NHM), uploaded death statistics for the months of April and May 2021. An HT story published on July 10 looked at these numbers to argue that the latest HMIS data “provides fresh evidence that India’s official Covid-19 death toll may be an underestimate”. A health ministry statement issued on July 14, has termed comparison of data from Civil Registration System (CRS) and HMIS with official Covid-19 mortality figures as “conjectures and speculation without any substantial basis”. How accurate are such comparisons? Here are four charts which answer this question.
What does HMIS/CRS data tell us about deaths during the second wave?
According to the July 9 update on HMIS, India recorded 792,171 adult and adolescent deaths during the months of April and May 2021. This is at least two times the recorded numbers since 2018. This data is in keeping with the trend of a sharp rise in the numbers of reported deaths recorded in the CRS in some states, which has been reported by various media organisations including The Hindu, Scroll and The News Minute. The nation-wide numbers for deaths recorded in the CRS in 2021 will only be published in 2023. So, a sharp jump in CRS deaths in states has been used to argue that official Covid-19 tally is an underestimate.
The HMIS actually further corroborates this theory because unlike the CRS, it also gives a break-up of deaths by causes. Fever and respiratory problems, symptoms associated with Covid-19, show the largest jump in causes of death in the HMIS data for 2021. It is also noteworthy that infant and child deaths decreased in April-May 2021 compared to April-May 2019 in HMIS data. Children are least likely to die from Covid-19. However, the HMIS numbers are better treated as an indication rather than quantifier of unrecorded Covid-19 deaths. Here’s why.
See Chart 1A and 1B: total deaths and cause-wise deaths under HMIS
HMIS is better used to capture a trend than actual number of unreported Covid-19 deaths
Caution should be observed in using the HMIS data to arrive at an absolute number of Covid-19 deaths (over what has been officially recorded) in the country. Here is why. The HMIS primarily collects statistics from around 200,000 government health facilities in rural areas. This means that it does not capture the entire health-care eco-system of the country. The numbers speak for themselves. Deaths recorded under HMIS are only a fraction of those under the CRS or Sample Registration System (SRS). The latter is based on survey conducted by the Office of Registrar General of India under the home ministry. The HMIS’s ability or lack of it to capture data also varies across states/districts, perhaps a direct result of the use of health facilities and reporting of numbers.
Chart 2: State-wise comparison of HMIS deaths with SRS/CRS in 2019
Suggested undercounting of Covid-19 deaths in the HMIS shows great regional variation
The HMIS data is available by districts. We also have official Covid-19 deaths at the district level. An HT analysis of these two statistics shows that scale of undercounting of deaths due to Covid-19 – the ratio of increase in recorded deaths in the HMIS between April-May 2021 and April-May 2019 and the official Covid-19 death toll in April-May 2021 – varies significantly across districts. Of the 401 districts for which Covid-19 death data is available and which also showed an increase in deaths in April-May 2021 compared to April-May 2019 in the HMIS data, this number was 15.7 or more in 81 districts. In 79 districts with the least scale of undercounting, this number was less than 2.2.
Map 1: District-wise undercounting of Covid-19 deaths using HMIS
But these numbers could be a result of HMIS’s poor coverage rather than better counting of Covid-19 deaths
Because the HMIS does not capture a lot of deaths in the country, the scale of undercounting (or lack of it) of Covid-19 deaths in various regions is likely to depend on the coverage of the HMIS. An HT analysis supports this theory. There are 401 districts which recorded at least one Covid-19 death in April-May 2021 and where HMIS data is available. A comparison of scale of undercounting of Covid-19 deaths; as described above, with the scale of undercounting of overall deaths under the HMIS – defined as HMIS deaths as a percentage of CRS deaths in 2019 – shows that the scale of undercounting of Covid-19 mortality is the lowest in districts where the HMIS has had poor coverage of deaths in 2019 and increases where the HMIS counted more deaths even before the pandemic.
Chart 3: Scale of undercounting of Covid-19 deaths and HMIS’s coverage of deaths
Please sign in to continue reading
- Get access to exclusive articles, newsletters, alerts and recommendations
- Read, share and save articles of enduring value