NDTV last month reported on a surge in registered deaths in Bihar during January-May 2021. In this period, around 82,500 more deaths were registered in the civil registration system than expected from past data.

We can go back a little further and use the data shared by NDTV to estimate the toll of the first wave of the pandemic in Bihar. There were competing narratives in 2020 about the impact of the epidemic on the state. Was Bihar spared the worst, as some reports based on its official case and fatality data suggested? Or were such claims founded on weak surveillance and death underreporting?

Given that Bihar is a state where the registration of deaths is incomplete, to interpret data from the civil registration system or CRS, we need to take account of possible trends in death registration. This process has many uncertainties, discussed in this supporting document. In a scenario that is neither very optimistic, nor very pessimistic, we find that:

  • There was around 1 excess death per 1,000 people in Bihar during the first wave. This would make first wave excess mortality in Bihar comparable to the level seen in Tamil Nadu, higher than in Madhya Pradesh, and lower than in Andhra Pradesh.
  • The numbers are consistent with what we know of Covid-19 spread in Bihar, and international data on Covid-19 fatality rates.
  • The disparity between excess deaths and recorded Covid-19 deaths is huge – higher than in any other state for which such estimates are currently available.

The data supports the view that Bihar was fairly badly hit during the first wave, and that recording of Covid-19 deaths was extremely poor.

Deaths registered in the civil registration system

Estimating excess mortality is complicated by the fact that the civil registration of deaths in Bihar is incomplete, and somewhat unstable. According to the 2019 CRS report, only around 31% of deaths in the state were registered in 2018. This rose to around 52% in 2019.

Monthly data shows rapidly improving registration during 2018, followed by slower and more uneven improvement in 2019. March to May 2020 saw a sharp decline in registered deaths, a trend seen in several states, and likely connected with disruptions during national lockdown. After June 2020, registered deaths fluctuated close to or above 2019 levels until a huge spike in May 2021.

Excess mortality during the first wave

For the purposes of studying mortality, we will take the “first wave period” in Bihar to begin in June 2020 and end in March 2021. This avoids the most obvious lockdown-related disruption to registration, and covers the variations in mortality which might be connected with the first wave. Detailed discussion of this choice, and how it might affect estimates, are in the accompanying document.

To estimate excess registered deaths, we need to know how many registered deaths to “expect”. In a very optimistic scenario, registration would have continued to follow the improving trend seen in 2019. This scenario ignores the clear disruption to registration in the early part of 2020.

In a very pessimistic scenario, registration would have followed the falling trend seen in the year prior to June 2020, which included the period of disruption. An intermediate scenario supposes that registration was disrupted, but recovered to 2019 levels over the first wave period.

The next plot shows registered deaths during the first wave, along with lines indicating expectations in these three scenarios.

The intermediate scenario avoids the extremes, and correctly predicts deaths during March and April 2021, when we might expect conditions to be close to baseline following the first wave surge, but prior to the second wave surge. In this case, we find around 64,000 excess registered deaths in the first wave period. With registration of 51%, this would correspond to around 127,000 excess deaths.

Bihar’s excess mortality in context

Given the uncertainties, is 127,000 excess deaths a plausible figure? In a population of roughly 120 million, this amounts to excess mortality of around 0.1%, or a little over 1 excess death per 1,000. This is similar to estimates of first wave excess mortality in Tamil Nadu; somewhat higher than in Madhya Pradesh; and somewhat lower than in Andhra Pradesh.

Madhya Pradesh, Tamil Nadu and Andhra Pradesh are states with more reliable death registration than Bihar; so it adds some confidence to Bihar’s estimate that it lies in the range of values seen in these states.

What would this excess mortality look like on the ground? Over a ten month period, total deaths in Bihar would have been roughly 20% above normal: a typical crematorium in the state would have seen one extra cremation for every five. Of course, there may have been periods and localities with much higher excess mortality.

Excess toll consistent with Bihar’s Covid-19 epidemic

We have already seen that Bihar’s excess mortality estimates lie in the range seen in other states. But are Bihar’s figures consistent with what we know of the COVID-19 epidemic in the state?

Reliable data from Bihar on the spread of SARS-CoV-2, the virus which causes COVID-19, is very limited. In six districts surveyed in August 2020 as part of the second national serosurvey around 16% had been infected. This was considerably higher than the national average at the time and, remarkably, there was wider spread in more rural districts surveyed in Bihar.

Based on these results and trends in Bihar’s reported coronavirus deaths we can estimate that 30% or more of the state’s population might have been infected by the end of the first wave. That would amount to over 36 million infections.

How many infections might result in death? Based on international data, we would expect somewhere between 18 and 30 out of every 10,000 infections in Bihar to result in death. These values are relatively low because of the youthful population of the state. By comparison, Andhra Pradesh, with a larger elderly population, might expect 30 to 51 out of every 10,000 infections to result in death.

If, indeed, Bihar saw around 36 million infections during the first Covid-19 wave, then at 30 deaths per 10,000 infections, we would expect around 110,000 Covid-19 deaths. This is fairly close to the estimated 127,000 excess deaths.

What is remarkable is that Bihar’s estimated first wave excess deaths were 85 times its 1,500 recorded Covid-19 deaths. For comparison, Andhra Pradesh’s first wave excess deaths were around 15 times its recorded Covid-19 deaths, while Tamil Nadu’s first wave excess deaths were around 6 times its recorded Covid-19 deaths.

Conclusions

Bihar’s first wave excess mortality appears to be consistent with wide spread of disease, and with Covid-19 fatality rates from international data. It was probably not the worst hit state in India; but nor did it escape lightly. The data, though limited, reinforces the view that stories of Bihar’s success at containing its first Covid-19 wave were unfounded.

We cannot be sure how much of Bihar’s excess mortality was from Covid-19, and what part was played by other factors such as economic distress or disruptions to healthcare. But the divergence between excess deaths and officially recorded Covid-19 deaths is huge.

What about the second wave? The large rise in deaths in May 2021 is consistent with the rapid and deadly second wave of disease seen across the country. But this part of story is still incomplete. It will be some time before we know the full impact of the pandemic on Bihar.

Detailed calculations and discussion of all the estimates are in an accompanying document.

Murad Banaji is a mathematician at Middlesex University, London.