In the last 10 days of July, Delhi, Mumbai and Ahmedabad, three of India’s major cities, released the results of COVID-19 seroprevalence surveys that their respective municipal corporations had conducted in association with research institutes. In the absence of any published countrywide sero-survey results, these studies, notwithstanding their limitations, are the only ones that give us information about the prevalence of COVID-19 infection in the general population of these cities, and possibly also give us some insight into the infection prevalence in typical urban Indian populations.
It should be borne in mind that sero-surveys, by testing for IgG antibodies to the infection that form around two weeks after infection, identify people who had been infected in the past and would have since recovered. So, such surveys identify only infected people who have circulating antibodies and only give an estimate of the prevalence of infection about two weeks before. Also, migrant workers, who constitute a good chunk of the population in Indian cities, possibly get missed out in these surveys as they are likely to have left for their hometowns during the lockdowns.
As reported earlier (“COVID cover-up” and “Chinks in the armour”, Frontline , July 3 and 17 respectively), the Indian Council of Medical Research (ICMR) conducted a nationwide cross-sectional sero-survey in mid May that covered over 70 districts, including only the containment zones of the 10 cities reporting the highest number of cases as of April 25, and tested over 26,000 individuals. Although the results of this national survey are yet to be published, the ICMR stated in a press briefing that the seroprevalence rate found among the general population, that is, excluding the hotspot zones of the 10 cities, was only 0.73 per cent. Since this survey was conducted in mid May, it gives the seroprevalence rate of April-end.
On the basis of leaked information, some news reports said in June that the ICMR study had found over 30 per cent prevalence in the cities. The actual findings, however, remain unknown as the ICMR has not made the results public. In the absence of the ICMR data, these city-specific surveys conducted in June-July, give us a more recent picture of seroprevalence in these three cities. This article discusses the data released by these surveys and also looks at, for what it is worth, the nationwide seroprevalence data gathered by the private diagnostics company Thyrocare Technologies Ltd. Neither the scientific community nor the Ministry of Health and Family Welfare has taken note of this private effort as it falls well short of being a properly conducted survey.
Delhi
In a press release dated July 21, the Health Ministry said it had commissioned the seroprevalence study for Delhi as a follow-up to the sero-survey the ICMR carried out in the containment zone of Delhi’s South East district. The National Centre for Disease Control (NCDC), an institution under the Ministry, in collaboration with the Delhi government carried out the community-based cross-sectional sero-survey in all the 11 districts of Delhi between June 27 and July10. The IgG antibody tests were done using the ICMR-approved indigenous COVID KAVACH IgG ELISA (enzyme-linked immunosorbent assay) kit.
Following a “multistage sampling study design”, 21,387 sera samples were collected. The results of this seroprevalence study (Figure 1) show that the (population-weights adjusted) average infection prevalence across Delhi (from mid June to the third week of the month) was 22.86 per cent. According to the Ministry, a large number of infected people were asymptomatic. Given the current population of Delhi (over 20 million), this means that about 4.6 million people were infected in mid to end June. However, the official data for the number of confirmed cases in Delhi for that period was 43,000 to 82,000.
While this clearly indicates a gross failure of the testing strategy, the Ministry, however, in its usual self-congratulatory style, said: “Nearly six months into the epidemic, only 23.48 per cent [unweighted average] of the people are affected in Delhi, which has several pockets of dense population. This can be attributed to the proactive efforts taken by the government to prevent the spread of infection including prompt lockdown, effective containment and surveillance measures, including contact tracing and tracking,…”
In an article in The Financial Express (August 1), Padam Singh, a former head of the ICMR’s medical statistics division and former Additional Director General at the ICMR, and Davendra Verma, former Director General of the Central Statistics Office of the Ministry of Statistics and Programme Implementation, wrote a severe critique of the sampling design and survey methodology even as they gave some additional data about the Delhi survey that had not been made public. The survey had found that the seropositivity rate (SPR) among females (24.2 per cent) was higher than among males (21.63 per cent) and that the SPR among the younger age group (<18) was 23.13 per cent, while that among the higher age group (>18) was 22.86 per cent. They also expressed surprise at the large variation in the SPRs between districts, ranging from 12.95 per cent to about 28 per cent.
While remarking that the above data were somewhat strange and unlikely, the authors pointed out that the SPRs found would actually be underestimates because the the detection kit used had a low sensitivity of 92.1 per cent, and the NCDC data analysis had not accounted for this. In the main, Padam Singh and Davendra Verma criticised the survey for its faulty four-stage sampling design. They pointed out that the inclusion of dispensaries as the primary sampling units in the third stage appeared to be basically because of administrative convenience. The selection of individuals at the fourth stage, which was left to the dispensaries, was, therefore, not random. “Thus, there is no sampling of wards and obviously there was no multistage sampling. The question is how dispensaries became part of the primary sampling units if they were not part of the sampling frame,” they wrote.
Meanwhile, despite this erroneous sampling methodology, the second round of sero-surveillance in Delhi has already begun.
A more sound survey in Mumbai
The design and conduct of the Mumbai survey was perhaps technically more sound as it involved many front-ranking research institutions and a larger group. Its limitation was its relatively (as compared with Delhi and Ahmedabad) smaller sample size as it covered only three of the city’s 24 wards, though this choice was made with some scientific rationale. It is also the only one among the three surveys whose summary report with all the relevant data has been made public.
This sero-survey, which used the random sampling methodology, is the first stage of a bigger project undertaken by a joint venture between the Tata Institute of Fundamental Research (TIFR), Mumbai; the Translational Health Science and Technology Institute (THSTI), Faridabad, Haryana; the University of Chicago, United States; Duke University, North Carolina, U.S.; A.T.E. Chandra Foundation, a Mumbai-based philanthropic organisation; Kasturba Hospital, Mumbai; and the IDFC Institute, a Mumbai-based public policy think tank. The project, which was launched on June 29, aims to conduct the survey at two time points to infer the trajectory of the epidemic in the city. The NITI Aayog, the Brihanmumbai Municipal Corporation (BMC) and the TIFR jointly conducted the initial stage of the survey from June end up to mid July.
The strengths of the study are the use of chemiluminescence immunoassay (CLIA) IgG antibody detection kits with high specificity (100 per cent) and high sensitivity (93 per cent). To gain insights into seroprevalence, the data captured was stratified in terms of slum/non-slum areas, age (in four groups: 12-24 years, 25-39 years, 40-60 years and over 60 years) and gender. In particular, the sampling methodology took into account the number of reported cases and population sizes in each of the wards and, significantly, for slums and non-slums separately. Sampling included people who had been symptomatic but recovered at the time of the survey or were asymptomatic without distinction. The sampling did not include active containment zones. The larger study also includes a survey of health care workers, assessment of the impact of risk factors on prevalence and determination specifically of the presence of neutralising antibodies (as against binding antibodies). These components of the study are still ongoing.
The selection of the wards was based on the following criteria: coverage of city and suburban areas, east, west and north areas and representation of localities with low to high caseload as on June 2. The selected wards were R North (Dahisar region, low), M West (Chembur region, average) and F North (Matunga region, high). Significantly, the chosen wards did not include the Dharavi slum, which lies close to F North and is supposed to be the largest slum in Asia.
“We chose 3 wards out of 24 for many reasons,” said Sandeep Juneja of the TIFR. “Our sampling budget was somewhat limited. It made sense to focus on geographically limited diversity so that our estimates given the sample sizes are more meaningful. Further, this way we make better conclusions about the differential in population density (slums and non-slums) leading to different prevalence rates. There is a large set-up cost involved in going to each new ward. So, from the logistics point of view, this was more manageable.”
Out of an estimated sample size of 8,670 individuals, the first round could gather only 6,936 samples because, strangely, while the participation from slums was 100 per cent, the turnout from non-slum areas was only 70 per cent (Table 1). “Nonetheless,” says the survey report, “the sample size relative to the prevalence is adequate to draw statistically meaningful conclusions.” Sample analyses were done at Kasturba Hospital and the THSTI.
The study has estimated an average prevalence of 56.5 per cent in slums and 15.5 per cent in non-slums in the three wards selected (Tables 2 (a-c) and 3 (a-c)). Assuming that 42 per cent of Mumbai’s population lives in slums, this works out to an overall seroprevalence rate of about 33 per cent for the entire city, which is significantly higher than the value of about 23 per cent found for Delhi. Also, according to the report, the seropositivity for females was marginally higher than for males, but the age-wise prevalence in both males and females was comparable. However, since the data have not been corrected for the lower sensitivity of 93 per cent of the CLIA kits, these figures represent a conservative lower estimate, notes the report.
On the basis of the results, as in the case of Delhi, the authors of this study too have inferred that asymptomatic infections are likely to be a high fraction of all infections. While the higher prevalence in slums could be due to the higher population density and shared common facilities such as water points and toilets, the lower prevalence in non-slums points to better adherence to physical distancing, wearing protective masks and access to better hygiene and sanitation coupled with the other non-pharmaceutical interventions by the BMC. On the basis of the survey-estimated prevalence in the three wards and the BMC’s records of reported deaths, the infection fatality rate (IFR) has been estimated to be a low 0.05-0.1 per cent as against the case fatality rate (CFR) of 5-6 per cent. The low IFR is perhaps a reflection of active measures the BMC has taken to isolate symptomatic cases immediately.“These results,” says the report, “will be valuable to learn more about herd immunity. Although it is still unclear what level of prevalence leads to herd immunity, our findings indicate that at least in slums this could be attained sooner [rather] than later, if the immunity exists and persists in a significant proportion of the population.” It is even likely that slum pockets may have already attained herd immunity. If, as mentioned in COVID-19 literature, we assume that R-nought (the average number of people that an infected person can pass on the infection to) to be 2 to 3, a near 60 per cent SPR would be close to the required value to achieve herd immunity. It is interesting to note that at the time of the Mumbai survey, Dharavi (which was not included in the survey) was already seeing a rapid downward trend in its caseload, and as Juneja concurs, Dharavi may indeed have a higher SPR than the other slum areas surveyed.
Second round of survey
According to Juneja, the planned repeat survey (the second round) is expected to begin in one or two weeks’ time. “Since the intent is to measure change in prevalence, we will be going to the same wards,” he said.
Analysing the Delhi and Mumbai sero-survey data for theWire.in, Murad Banaji, a mathematician from Middlesex University London, wrote: “Limitations aside, the data suggest two things: that the virus has spread wide and the fatalities have been relatively low.” According to Banaji, when the epidemic in London and New York City (NYC) was in its downward trend, the seroprevalence rates in the two cities had been estimated to be 17.5 per cent and 23 per cent respectively. An SPR of about 23 per cent for both Delhi and NYC works out to about 4.4 million and 1.9 million infections respectively shortly after the respective surveys. However, the fatalities in the two cities were drastically different: 3,200 in Delhi and 15,000 in NYC. If you compare the IFRs calculated from this data, it means that COVID-19 was 10 times deadlier in NYC than in Delhi, says Banaji.
Similarly, following Banaji, if you compare Mumbai and London, which had recorded a similar number of deaths (of 5,500) at the time of the surveys, their respective SPRs of 33 per cent and 17.5 per cent work out to 4.5 million and 1.6 million of infected people. On the basis of this data, the rough values of IFRs for the two cities work out to 0.12 per cent and 0.34 per cent respectively. So, the virus was three times deadlier in London than in Mumbai. Of course, it must be pointed out here that Banaji has not accounted for the demographic difference in the cities being compared, with Indian populations being skewed in favour of younger age groups. One of the important factors in the COVID-19 pandemic is that younger people are not affected as severely by the virus as the elderly are. If you take that into account using age-stratified data, the differences should reduce significantly.
A politically motivated survey?
The Ahmedabad Municipal Corporation (AMC) claimed that the sero-survey conducted in the city was the world’s largest, but it released only just as much information about the results as the Delhi survey. In fact, the AMC did not even share the identity of the research institution/laboratory with which it had associated to carry out the survey. According to the AMC press release, the survey was conducted from June 16 to July 11 and 30,054 samples that were “fully distributed [ sic ]” among the seven zones of the city were collected. For a city with a population of 6.3 million, this works out to a sample-to-population ratio of 4,770 per million. Comparing this with the Spanish sero-survey whose ratio was 1,302 per million and a U.S. study in six of its States whose ratio was 255 per million, the release claimed that this was the most extensive sero-survey study in the world so far.
Alongside this claim, the release also seemed to take a swipe at the ICMR, whose as-yet-unpublished sero-survey of May, which was conducted with just 496 samples from containment zones only, had a “minuscule” ratio of “just 79 per million”. It was rumoured in early July that the ICMR had not been allowed to release its sero-survey results for political reasons because its survey had found a very high SPR in Ahmedabad, something that would not have been palatable to the powers that be. Thus, the Ahmedabad survey and the tenor of its press release smack of a politically motivated study, especially when even the institution involved in the study has not been named.
According to the information put out by the AMC, the SPR in Ahmedabad (during June) was 17.61 per cent. The release further said: “Some sections of media have reported about a study by ICMR showing around 49 per cent seropositivity in Amdavad.... [The ICMR study] is not at all representative of [the] actual existing situation and cannot be relied upon for any conclusion (the AMC’s latest study is 60 times larger).”
Like in Delhi, the AMC survey too seems to have found variations in the SPR between different zones of Ahmedabad, with the Central Zone having the highest SPR (28.43 per cent), followed by the North Zone (27.42 per cent) and then the North-West Zone ( 6.43 per cent). Significantly, like in Delhi and Mumbai, this study too found that the SPR among females (17.98 per cent) was a little higher than among males (17.29 per cent), but the release noted that “the difference is not statistically significant”.
Private ‘sero-survey’
Thyrocare Technologies Ltd has been collecting seroprevalence data since the beginning of July and has made this data public through tweets of A. Velumani, its chairperson and founder, who is a scientist-turned-entrepreneur. Although this cannot be termed a proper study because it is not a randomised survey, nevertheless, one has to accept that the data collection strategy was novel and innovative, the data collation and its classification too was pretty quick and the dissemination of information immediate.
Velumani offered antibody tests at a very low cost and, using the vast network of Thyrocare diagnostic laboratories across the country, gathered seroprevalence data from corporate houses, apartment buildings and individuals who came to have themselves tested in more than 600 pin codes across the country. “We have not chosen whom to test; we have only tested those who wanted it. Eighty per cent was the requirement of the corporates, 15 per cent was the requirement of residential societies and 5 per cent was the demand of individuals. We covered pin codes from Nariman Point to Jamshedpur, so big and small cities are all covered,” Velumani told the online news website thequint.com in an interview.
So, in that sense, the sampling is not random at all; it is highly biased and leaves out the large number of people from the lower strata of society to whom the survey did not offer the test. In fact, there is no statistical sampling at all because there is no defined sampling frame to begin with. Therefore, statisticians and medical experts, perhaps, would not even give this survey a second look. Nevertheless, the data from this private survey (which seems to be ongoing), given in age- and gender-stratified form, does perhaps offer some insight into infection prevalence across the country that is most recent.
As of August 4, Thyrocare had tested 1,51,588 individuals (across all age groups), of which 28,347 were seropositive for IgG antibodies. This works out to a crude SPR of 18.7 per cent, albeit in a significantly biased population, across the country. Of course, this should not be immediately taken to mean that 240 million people of the country were infected, but it probably gives a very rough idea of prevalence. The data of July 29 shows that more males (about 2.5 times more) got themselves tested than females. Interestingly, even this biased survey found a higher SPR among females than males, 20.02 per cent compared with 17.62 per cent. Also, if you look at the SPRs for the three cities discussed above, Thyrocare’s figures are roughly in the same ballpark, notwithstanding the much smaller numbers it tested in these cities compared with the population sizes that were sampled in the sero-surveys. If there is something significant in these numbers, it is for the experts to glean from them.
What the city sero-surveys, at least, tell us is that about one out of four to five individuals in major cities has been infected, but this is still far removed from what is required for herd immunity to set in. However, if any part of a city is to become “herd immune”, it will be the slums first, as the Mumbai survey shows. And, it is quite likely that Dharavi may have already achieved herd immunity given its recent caseload trend. Asymptomatic infections seem to constitute a significant proportion of all infections. All surveys seem to indicate that seropositivity in women appears to be higher than in men. This may be an interesting aspect for virologists, medical experts and epidemiologists to look into.
