'Herd Immunity A Moving Target As People Start To Mingle More'

Based on seroprevalence data, IDFC has advised state governments on policy to control the spread of the disease, and now on vaccine allocations. He spoke to Rukmini S. about innovations in collecting COVID-19 data in India, the limits of sero-surveys, and how herd immunity thresholds for the disease can change.

Edited excerpts:

It became evident quite early that sero-surveys were going to be very important in understanding the COVID-19 pandemic. How was IDFC able to pull off multiple sero-surveys across the country? What were the logistics of that like?

The first thing I want to say is India is actually remarkable. If you were to list countries by the number of COVID-19 sero-surveys or population-level surveys that they did, India would be close to the top of that list. If you adjust that list by income and healthcare capacity, I think India probably outperformed most governments, other than some in East Asia perhaps. A lot of different organisations have done many sero-surveys in India. And it doesn't stop there. There are also prevalence surveys, a lot of scientific research, surveys of what's happening to incomes. It's remarkable. After we finish blaming a bunch of people, we should just say there are ways in which India was quite successful.

Importantly, IDFC realised the need to work with governments, to adapt to the circumstances, and that speed was key. We immediately identified state governments that were willing to allow or undertake COVID-19 sero-surveys for their particular public policy needs. If there were specific questions that government officials had, the sero-survey would be tailored around that. IDFC worked to get funding for these through non-profits or NGOs in India, or sometimes directly from the government, because we had to come together to assemble a sort of private task force to help the government respond once the lockdown was declared.

For example, in Mumbai, we were able to work directly with the municipal corporation to get approvals and support. We worked with the public Kasturba Gandhi Hospital and Tata Institute of Fundamental Research to get together the scientists needed to implement the sero-survey, and worked on the survey design with them, because we knew that slums were going to be particularly at-risk. In Pune, [the local administration] did the sero-survey independently, and very quickly. We've also been working with the Andhra Pradesh government, but even before that, they did a rural and urban sero-survey of four districts. Around that time, Bihar was also doing random testing of returning migrants; not antibody tests but RT-PCR. Then in Karnataka, [former health commissioner] Pankaj Pandey and others in the health ministry were open to doing a statewide COVID-19 survey that would really give us a sense of what was going on in rural areas. IDFC also collaborated with the Centre for Monitoring Indian Economy (CMIE), which provided the Karnataka sample, and a number of local laboratories. Once we had those two successes [in Karnataka and Mumbai], we got more requests from state governments, as in Tamil Nadu and the repeat survey in Mumbai.

IDFC plugged all these pieces of information together, correlating official data with seroprevalence, to be able to advise states on how to go about suppressing the spread of COVID-19. Now, we're advising on vaccine allocations, again using the seroprevalence route. We're not just gathering data, but trying to guide policy based on that data.

The Bihar example is interesting because it shows the level of creativity we've had to get into to fill in the blanks for data that we simply don't have.

I agree. When we talk at times about how slowly parts of India's governments respond, at other times we have to acknowledge that there are pockets that act very quickly, and that there are big returns to facilitating that independence and letting people act without the bureaucratic process that would otherwise slow things down.

We saw that with COVID-19. The fact that there was willingness to set aside a lot of bureaucratic processes to act quickly is what allowed us to do these [sero-surveys]. That's when you will see examples like Bihar, where talented officials like Chanchal Kumar [principal secretary to the Chief Minister] and his team, who immediately realised the need to do population-level surveys, and not just look at confirmed cases, and that they should target suppression efforts at the areas where cases were popping up. Being willing to act on that rather than just defer to standard guidelines is, I think, what allowed Bihar to give us a lot of information on COVID-19. In fact, I would argue that it would have been great if early on, we'd paid a little bit more attention to Bihar. Because, by separately testing migrants coming in from different states, randomly, they were able to get a sense through Bihar data of what was going on throughout India. If we'd highlighted Bihar's success, that may have led to even faster ramping-up of random population testing.

Were some of the lessons from sero-surveys taken too seriously? Where high seroprevalence was taken as an undisputed fact, things like antibody decay became caveats which we didn't pay enough attention to, and the path to herd immunity seemed more clear and direct than it really was.

I agree, for a number of reasons. One, as you pointed out, is antibody decay. We know that the fraction of the population that is infected with a disease can only increase, and as long as immunity is durable, the fraction of the population that's immune can only increase. But seroprevalence measures what happened, say in the last three months. As the rate of infection declines, and because of antibody decay, COVID-19 seroprevalence will actually stay constant and maybe even decline. What that means is over time, seroprevalence is under-estimating more and more the level of natural COVID-19 immunity present in the population. There are statistical ways to adjust for that, and also ways to directly check for that.

What's also important is that, when we walked into this pandemic, we learned a lot from epidemiologists about compartmental models and this concept called 'herd immunity'. But we took it a bit far. What we said was there's some immunity threshold that, when you hit it, the epidemic dies out. Everybody was talking about what is the COVID-19 herd immunity threshold, what is the attack rate, and so on. The difficulty was that when we started seeing these really high rates of seroprevalence, say 55% in Mumbai slums, immediately people said we're close to herd immunity, we can relax. But we don't actually know what the level of herd immunity is. It's yet to be estimated for COVID-19.

Also, herd immunity is not an absolute level but depends on human behaviour. If people, say in Mumbai, interact the same way as in 2019, then the herd immunity threshold is going to be higher than in a lockdown where everyone's controlling their behaviour. Even today, when people know there's still a risk, they're not back to interacting in exactly the same way as in 2019. So the herd immunity threshold will rise as activity rises. Do I know that, assuming a seroprevalence of 55% was correct, that we were close to herd immunity? No! I had no idea what the threshold was. I knew that as activity rose, that threshold would keep increasing. Because of that, I think we took seroprevalence too seriously.

Does that mean that knowing seroprevalence is not useful? No. It's very useful. First, if you have 55% seroprevalence in slums and 15% in non-slum areas in July, it says you're doing a much better job of controlling the epidemic in non-slums than in slums, and that there is rapid spread in slums even during the lockdown, and we need to understand why. Either you ease the lockdown, or understand what the lockdown did, which was that by reducing mobility, people got stuck in crowded housing and that could have led to rapid spread of the COVID-19 epidemic. I don't know the answers. We have to learn about this so we can prepare for the next epidemic. So, we can learn a lot from seroprevalence because it gives you a sense, just like measuring your temperature gives you a sense of what might be wrong, but you should be careful in interpretation.

It seems that right now we're going to see the epidemic move at multiple speeds. There is the population that hasn't yet been infected, also reinfection, and the new COVID variants. One seroprevalence number is never going to capture it all.

You bring up a very important point, which is that the national epidemic is a series of local epidemics and they may have different timings. One way to think about it is through birthday parties. Every week there's a birthday party in different places in different households in India. If you look at the entire country, the number of birthday parties seems relatively constant over time, but that doesn't mean in my family I have birthday parties all year long! In an epidemic, there is some contagion even across areas. So you'll find some correlation, but it's not a perfect correlation. Take Tamil Nadu, for example. Tamil Nadu has, based on a recent preprint IDFC put up, 32% seroprevalence, but if you disaggregate that, going from the Nilgiris all the way to Perambulur [over 300 km away], you're seeing seroprevalence that ranges from 12% to 52%. So there's still a lot of heterogeneity within India. We tend to focus on the state-level aggregates, but even state-level aggregates are a lot. In a state like Bihar with 100 million people, which is the size of many countries, I don't know if state-level is the right way to aggregate. It's very important we understand that these are a series of local epidemics which are differently timed, but that follow common dynamics. Once you see that, it will alter what you expect to see in India. Yes, it's possible we will continue seeing outbreaks of COVID-19, but we need to expect those, and also not over-interpret them.

We welcome feedback. Please write to respond@indiaspend.org. We reserve the right to edit responses for language and grammar.