estimating missing deaths in delhi apos s covid 19 data CORD-Papers-2021-10-25 (Version 1)

Title: Estimating missing deaths in Delhi's COVID-19 data
Abstract: A sero-prevalence survey in Delhi measured an infection rate of 23.48% and an implied infection fatality rate (IFR) of 0.06%. Modeling using age group based IFRs from France, Spain and Lombardia project an average IFR that is significantly higher than currently estimated. We show that at least 1500-2500 COVID-19 deaths in the 60+ age group are missing.
Published: 7/30/2020
DOI: 10.1101/2020.07.29.20164392
DOI_URL: http://doi.org/10.1101/2020.07.29.20164392
Author Name: Chakravarty, S
Author link: https://covid19-data.nist.gov/pid/rest/local/author/chakravarty_s
sha: a2f996dc91ce3a4f2474e4cb929c492a2e38e5c7
license: medrxiv
source_x: MedRxiv; WHO
source_x_url: https://www.who.int/
url: https://doi.org/10.1101/2020.07.29.20164392 http://medrxiv.org/cgi/content/short/2020.07.29.20164392v1?rss=1
has_full_text: TRUE
Keywords Extracted from Text Content: IFRs IFR India's COVID-19 France's in Table 1 . people Pulla 2020b medRxiv France's COVID-19 France Spinnicci May. A large Spain ( Delhi's Dengue antibodies IDSP COVID-19 France's Mumbai (BMC 2020 b medRxiv preprint Figure 2 metros Figure 1 I Delhi IFR curve Pulla 2020a Delhi's IFR Delhi's death Delhi's IFR curve Delhi COVID-19 Figure 2 SARS-CoV-2 Delhi's IFRs IFRs 5 medRxiv preprint Delhi's lower Delhi's death distribution. https://doi.org/10 IFRs Surat (SMC 2020 IFR sero-surveys Delhi's COVID-19 Delhi Delhi's average
Extracted Text Content in Record: First 5000 Characters:A sero-prevalence survey in Delhi measured an infection rate of 23.48% and an implied infection fatality rate (IFR) of 0.06%. Modeling using age group based IFRs from France, Spain and Lombardia project an average IFR that is significantly higher than currently estimated. We show that at least 1500-2500 COVID-19 deaths in the 60+ age group are missing. India is a relatively young country and the IFR is expected to be lower than countries with older populations like Spain, France or Italy. A country wide infection fatality rate comparison between countries with very different population structures like India or France wouldn't make sense but comparing the fatality rates in similar age groups would be appropriate. For example, a comprehensive study (Salje et al. 2020 ) of France's COVID-19 outbreak estimated these age group based IFRs (with error ranges). The IFR goes up rapidly with age, especially past the age of 50. India's significantly younger population would result in a lower average IFR compared to France. In Figure 1 I use a recent Census population projection (MoHFW 2019) to estimate the 2020 Delhi population distribution in different age groups. I will use both 2011 and the estimated 2020 distributions as the Delhi of 2020 is older (share of people in the 80+ group nearly doubles), and this makes a big difference to COVID-19 fatalities. I use age based infection rates of China (Verity et al. 2020) , France (Salje et al. 2020) , Lombardia (Italy) (Modi et al. 2020) , New York (Yang et al. 2020) and Spain (Bevand [2020] 2020) as models to estimate Delhi COVID-19 deaths for a sero-positivity of 23.48%. I assume that all age groups are uniformly infected as age based infection rates were not released. As Figure 2 shows, Delhi's IFR turn out to be significantly lower than what we would expect if the age group based IFRs of any of the five models regions considered above. For example, Delhi's average IFR using France's numbers are less than half of France's in Table 1 . Delhi's COVID-19 deaths are anywhere between one-half to one-tenth of what the various model regions suggest, and lower than any lower error bound as well! We also see that using the estimated 2020 population distribution instead of 2011 would lead to significantly higher number of projected COVID-19 deaths. Delhi's lower than expected IFR is surprising as there is no good reason to expect that an Indian in a particular age group is healthier, have better immunity, or access to better health care than a person 2 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 30, 2020. . 3 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 30, 2020. . https://doi.org/10.1101/2020.07.29.20164392 doi: medRxiv preprint Figure 2 : Delhi had 3004 officially confirmed COVID-19 deaths on 4th July, 2020, leading to an infection fatality rate (IFR) of 0.06% based on a seroprevalence of 23.48%. This is exceptionally low compared to the expected number of deaths (and IFR) if we apply the age-based infection fatality rates of countries and regions that have had a large COVID-19 outbreak. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 30, 2020. . https://doi.org/10. 1101 in Europe in the same age group. An age group wise infection fatality rate of Delhi and the other models regions can shed some light on this puzzle. Unfortunately, it seems that the last time Delhi release any data of deaths by age group was on the 20th of May (DHGS Delhi 2020) when Delhi had 176 COVID-19 deaths only. Many cities and states have been more forthcoming with such data. In Figure 3 , I plot the cumulative share of deaths vs. age groups for nation-wide COVID-19 data from the Integrated Disease Surveillance Programme (Ghosh 2020) , Mumbai (BMC 2020), Bengaluru (BBMP 2020) and Surat (SMC 2020) on the same graph. The distribution of deaths by age groups in large urban areas is remarkably similar, perhaps reflecting their similar age and comorbidity profiles. Cities like Kolkata and states like Kerala are likely to have older populations but the differences are likely to be small. Thus far, about 60% of COVID-19 deaths have been recorded in the 10 largest metros so the IDSP data also has a similar shape. Given the absence of a better alternative, I will use the Mumbai age group wise death distribution data to model Delhi's death distribution. I distribute Delhi's COVID-19 deaths in different age groups using Mumbai as a model, and compare with the expected share of
PDF JSON Files: document_parses/pdf_json/a2f996dc91ce3a4f2474e4cb929c492a2e38e5c7.json
G_ID: estimating_missing_deaths_in_delhi_apos_s_covid_19_data
S2 ID: 220846971