key: cord-0833746-f0xzuzjr authors: Barchuk, A.; Skougarevskiy, D.; Kouprianov, A.; Shirokov, D.; Dudkina, O.; Tursun-zade, R.; Sergeeva, M.; Tychkova, V.; Komissarov, A.; Zheltukhina, A.; Lioznov, D.; Isaev, A.; Pomerantseva, E.; Zhikrivetskaya, S.; Sofronova, Y.; Blagodatskikh, K.; Titaev, K.; Barabanova, L.; Danilenko, D. title: COVID-19 pandemic in Saint Petersburg, Russia: combining surveillance and population-based serological study data in May, 2020 - April, 2021 date: 2021-08-02 journal: nan DOI: 10.1101/2021.07.31.21261428 sha: b378e23114b5d1d74054628ec0b9689b77e876ba doc_id: 833746 cord_uid: f0xzuzjr Background: The COVID-19 pandemic in Russia has already resulted in 500,000 excess deaths, with more than 5.6 million cases registered officially by July 2021. Surveillance based on case reporting has become the core pandemic monitoring method in the country and globally. However, population-based seroprevalence studies may provide an unbiased estimate of the actual disease spread and, in combination with multiple surveillance tools, help to define the pandemic course. This study summarises results from four consecutive serological surveys conducted between May 2020 and April 2021 at St.Petersburg, Russia and combines them with other SARS-CoV-2 surveillance data. Methods: We conducted four serological surveys of two random samples (May-June, July-August, October--December 2020, and February-April 2021) from adults residing in St.Petersburg recruited with the random digit dialing (RDD), accompanied by a telephone interview to collect information on both individuals who accepted and declined the invitation for testing and account for non-response. We have used enzyme-linked immunosorbent assay CoronaPass total antibodies test (Genetico, Moscow, Russia) to report seroprevalence. We corrected the estimates for non-response using the bivariate probit model and also accounted the test performance characteristics, obtained from independent assay evaluation. In addition, we have summarised the official registered cases statistics, the number of hospitalised patients, the number of COVID-19 deaths, excess deaths, tests performed, data from the ongoing SARS-CoV-2 variants of concern (VOC) surveillance, the vaccination uptake, and St. Petersburg search and mobility trends. The infection fatality ratios (IFR) have been calculated using the Bayesian evidence synthesis model. Findings: After calling 113,017 random mobile phones we have reached 14,118 individuals who responded to computer-assisted telephone interviewing (CATI) and 2,413 provided blood samples at least once through the seroprevalence study. The adjusted seroprevalence in May-June, 2020 was 9.7% (95%: 7.7-11.7), 13.3% (95% 9.9-16.6) in July-August, 2020, 22.9% (95%: 20.3-25.5) in October-December, 2021 and 43.9% (95%: 39.7-48.0) in February-April, 2021. History of any symptoms, history of COVID-19 tests, and non-smoking status were significant predictors for higher seroprevalence. Most individuals remained seropositive with a maximum 10 months follow-up. 92.7% (95% CI 87.9-95.7) of participants who have reported at least one vaccine dose were seropositive. Hospitalisation and COVID-19 death statistics and search terms trends reflected the pandemic course better than the official case count, especially during the spring 2020. SARS-CoV-2 circulation showed rather low genetic SARS-CoV-2 lineages diversity that increased in the spring 2021. Local VOC (AT.1) was spreading till April 2021, but B.1.617.2 substituted all other lineages by June 2021. The IFR based on the excess deaths was equal to 1.04 (95% CI 0.80-1.31) for the adult population and 0.86% (95% CI 0.66-1.08) for the entire population. Conclusion: Approximately one year after the COVID-19 pandemic about 45% of St. Petersburg, Russia residents contracted the SARS-CoV-2 infection in, or 2.2 mln people. Combined with vaccination uptake of about 10% it was enough to slow the pandemic until the Delta VOC started to spread. Combination of several surveillance tools provides a comprehensive pandemic picture. Funding: Polymetal International plc. This study summarises the four consecutive rounds of populationbased serological study based on two representative samples of adults residing in St. Petersburg, Russia, between May 2020 and April 2021. In addition, we combine the seroprevalence estimates with all other available surveillance data: official case count, hospitalisation data, SARSCoV2 VOCs monitoring data, COVID19 specific mortality, excess mortality, vaccination uptake, mobility trends, and search term trends. Thus, we aim to assess the different surveillance tools validity and present a comprehensive pandemic course in the fourth largest European city with a more than 5 million population. St. Petersburg serological study settings and design are described in detail in our previous report [8] . In brief, St. Petersburg COVID19 study is populationbased epidemiological survey of a random sample from the adult population to assess the sero prevalence of antiSARSCoV2 antibodies. The study was based on a phonebased survey and an individual invitation to the clinic for blood sample collection. Eligible individuals were adults residing in St. Petersburg older than 18 years and recruited using the random digit dialling (RDD) method. RDD was accompanied by the computerassisted telephone interviewing (CATI) to collect the information on both individuals who accepted and declined the invitation for testing. Blood samples from the same population group were collected between May 25, 2020, and June 28, 2020, in the first cross section "May-June 2020 survey" henceforth) and between July 20, 2020, and August 8, 2020, in the second "July-August 2020 survey" crosssection). Considering the risks of low response in the next planned crosssection, we created a new popula tion sample applying the similar strategy of RDD followed by CATI ("October-December 2020 survey"). The initial response to the RDD was higher in autumn and winter 2020-2021 compared to the first crosssection in summer 2020. The fourth crosssection ("February-April 2021 survey") involved individuals from both population samples invited between February 15, 2021, and April 4, 2021. Repeated blood sampling allowed seroconversion assessment for individuals who tested pos itive in previous surveys. Also, in this crosssection, some participants reported at least one vaccine shot. They were in cluded in the study as nonresponders as the initial survey does not fully address the characteristics associated with vaccina tion status. However, the vaccinated individuals were still tested. The participant flow for all four crosssections is reported in Figure 1 . The full study protocol is available online ( ). 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 August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint 53,179 were not reachable, did not re spond or refused to participate in phone survey 66,250 mobile phone num bers generated (RDD) in May-June, 2020 46,767 mobile phone numbers generated (RDD) in October-December, 2020 34,787 were not reachable, did not re spond or refused to participate in phone survey 6,671 not eligible or excluded : 3,048 were not residents of selected city districts 81 were younger than 18 years old 2,924 interrupted the interview 618 were surveyed after the recruit ment was over 13 ,071 agreed to participate in phone survey in May-June, 2020 11, 980 During the four surveys, we assessed antiSARSCoV2 antibodies using three different assays. Even though our report was selected among studies of higher quality in the recent systematic review, a significant limitation was related to the absence of own test performance validation [9] . We conducted a validation that revealed the decrease in sensitivity for one of the assays [17] . Finally, to report seroprevalence, we selected enzymelinked immunosorbent assay (ELISA Coronapass) CoronaPass total antibodies test (Genetico, Moscow, Russia) that detects total antibodies (the cutoff for positivity 1.0) and is based on the 4 . 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. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2021. ; recombinant SARSCoV2 spike protein receptor binding domain (Department of Microbiology, Icahn School of Medicine at Mount Sinai, New York, NY, USA). We used ELISA Coronapass through all four surveys. We also used the results of our validation study to correct the seroprevalence estimate for test performance. Sensitivity is equal to 92% and specificity to 100% for ELISA Coronapass (for full validation see [17] ). We summarised the data that included the official registered cases statistics, the number of patients hospitalised, the number of COVID19 deaths, excess deaths, and tests performed for COVID19 detection. Although this information was not available from one source, we used a combination of different sources to restore the pandemic course in St. Petersburg. We have also used the leading Russian search engine Yandex search history in St. Petersburg region to obtain search trends for three terms: "loss of smell", "smell", and "saturation". In addition, Yandex provided mobility trends for St. We used the information on the official COVID19 mortality and derived excess mortality to estimate the IFR. IFR was calculated for the four periods covered by our seroprevalence surveys. We treated the true number of deaths as an interval censored random variable bound downwards/upwards by the number of deaths 14 days after the crosssection start/end date (see Supplementary Materials ). The sample size calculations and statistical analysis plan for the serological survey were described in detail in our previous report [8] . The initial sample size of 1550 participants was calculated assuming the hypothetical prevalence of 20% to obtain the resulting sampling error of 2% using a 95% confidence interval. The actual sample size was lower, which resulted in a maximum error of about 4% when hypothetical seroprevalence reached 40%. The study's primary aim was to assess the seroprevalence based on antibody tests accounting for nonresponse bias and test sensitivity and specificity. Nonresponse was evaluated by comparing answers provided during the CATI by those who visited the test site and all surveyed individuals, estimated using a binomial probit regression of individual agreement to participate in the study and offer their blood sample on their observable characteristics. In the first report, we described the variables that we had chosen to estimate the correction. The observable characteristics associated with response and positivity were reported any disease symptoms before the test and the COVID19 testing history. We used similar variables to correct the seroprevalence estimates for nonresponse during all four crosssections. To account possible sample nonrepresentativeness in a sensitivity analysis, we computed raking weights to match the survey age group and educational attainment proportions in the 2016 representative survey of the adult city population with R package used to compute the weights. The original report also explored individual risk factors for test positivity in the sample participants who completed clinic paperbased surveys. This report assessed individual risk factors using a binomial probit 5 . 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. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2021. The Research Planning Board approved the study of the European University at St. Petersburg (on May 20, 2020) and the Ethics Committee of the Clinic "Scandinavia" (on May 26, 2020). All research was performed following the relevant guidelines and regulations. Informed consent was obtained from all participants of the study. The study was registered with the following identifiers: Clinicaltrials.gov (NCT04406038, submitted on May 26, 2020, date of registration -May 28, 2020) and ISRCTN registry (ISRCTN11060415, submitted on May 26, 2020, date of registration -May 28, 2020). Official statistics, VOCs monitoring data, search terms trends, and mobility trends were obtained from open sources as aggregated data. Analysis based on opensource aggregated data does not require additional ethical permission in Russia. All analyses were conducted in with the aid of GJRM package [20] , study data and code is available online ( ). The study's funder had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The resulting 14,118 individuals responded to CATI questionnaire -6,400 in the first population sampling and 7,718 in the second (see Supplementary Appendix Table S5 for details regarding missing records on variables of interest). The respondents represent city population in terms of their gender, employment status, and household size, but were younger than the adult city population as of 2016 and had higher levels of educational attainment (see Supplementary Appendix Table S6 ). Overall, 2,413 individuals provided blood samples through the seroprevalence study course that were analysed using ELISA Coronapass: 1,035 in the first May-June 2020 survey and 503 of them in the second July-August survey, and 1,378 newly recruited participants in the third October-December survey. Finally, samples from 1,182 participants from previous surveys were collected and analysed in February-April 2021. The adjusted seroprevalence in May-June 2020 was 9.7% (95%: 7.7-11.7) and increased to 13.3% (95% 9.9-16.6) in July-August 2020. We noticed a major increase through the third (22.9% 95%: 20.3-25.5) and between the third and fourth cross sections of the seroprevalence study (see Figure 3 and Supplementary Figure S1 for the weekly data), resulting in seroprevalence equal to 43.9% (95%: 39.7-48.0) in February-April 2021. Naïve antibodies seroprevalence to SARSCoV2 and seroprevalence corrected for nonresponse only and corrected for nonresponse and test performance are presented in Table 1 . Seroprevalence estimates adjusted through raking weights were similar and are available in Supplementary Table S7 . Sero prevalence by different subgroups are reported in Supplementary Table S8 . History of any symptoms, history of COVID19 6 . 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 August 2, 2021. ; tests, and nonsmoking status were significant predictors for higher seroprevalence. The SARSCoV2 antibodies test results trajectories showed that most individuals remained seropositive with a maximum followup of 10 months ( Figure 2 ). Among 177 participants who have reported at least one vaccine dose by the end of April, 2021, 92.7% (95% CI 87.9-95.7) were seropositive. ticipants who tested positive at least once, excluding the 20200720 -20200808 crosssection. Solid blue line is the loess smoother, blue areas report its 95% CI. 7 . 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 August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint The number of cases officially registered in the spring 2020 was much lower than in the autumn and the winter 2020-2021. Number of SARSCoV2 tests reached its maximum in the winter 2020-2021 in contrast to a relatively low number of tests reported in the spring 2020. Official case statistics contrast the number of hospitalisations, official deaths, and excess deaths reported in the spring 2020. The official number of cases, the number of hospitalisation and deaths from COVID19 never reached zero between and after the pandemic waves. The number of COVID19 deaths and excess deaths from all causes peaked in both periods and was in line with hospitalisation dynamics ( Figure 3A ). Internetbased search terms trends were in line with pandemic dynamics. They reflected the changes in hospitalisation and death count better than the official case count, especially during the spring wave ( Figure 3A ). In addition, urban activity trends showed an apparent response to the first spring wave, somewhat less evident response during the second winter wave, and return to prepandemic activity levels in the late spring of 2021. The SARSCoV2 circulating lineages diversity in 2020 was low. All samples from this period were attributed to the B. Figure 3B ). Using excess deaths data, the IFR was equal to 1.04 (95% CI 0.80-1.31) for the adult population for the whole pandemic period. IFR based on the official COVID19 deaths counts was lower and amounted to 0.43% (95% CI 0.11-0.82). When we considered the entire population of the city rather than the adult population for IFR, we obtained the estimate of 0.86% (95% CI 0.66-1.08) based on the excess deaths data. Full results for IFR are reported in Supplementary Table S2 . There was a clear upward trend in IFR by age. IFR was higher in men in all age groups. Our study is the first comprehensive attempt to characterise the pandemic dynamics in the fourth largest European metropolitan area. We used all available sources for surveillance, including populationbased seroprevalence study, the monitoring of SARS CoV2 VOCs, data on registered cases and deaths, relevant search term trends and city activity. Combining this data provides an overall global picture how the pandemic evolved through 2020 and 2021 in St. Petersburg. In April 2021, approximately one year after COVID19, we estimated that about 45% contracted the SARSCoV2 infection in St. Petersburg, roughly 2.2 mln residents. Together with more than 10% vaccination uptake to that moment, less than 45% susceptibles were there in the population. Nevertheless, it was enough to avoid a new pandemic wave in the absence of mitigation measures till the spread of the Delta VOC (B.1.617.2) at the end of May 2021. The first year of the COVID19 pandemic in St. Petersburg can be characterised by two waves of similar intensity but different 8 . 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. . 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 August 2, 2021. ; lengths. In the spring 2020, the first pandemic wave resulted in unprecedented mitigation measures and population reaction that helped flatten the pandemic curve and preserve the healthcare system functionality. As a result, the daily registered number of cases have plateaued in summer. It helped reorganise the hospital capacities in St. Petersburg and prepare for the subsequent increase in the case count. The hospitals had to experience the entire load in autumn and winter, although many additional beds were allocated to COVID19 patients. The number of daily cases plateaued during the winter holidays went down to 700-900 officially registered cases per day in the spring 2020. The halt of most mitigation measures has not resulted in the subsequent pandemic wave until the Delta VOC started to spread rapidly in May 2021. The new summer 2021 pandemic wave is yet to be analysed. Possibly the number of individuals who already have antibodies to SARSCoV2, which was reported to be a strong protection marker from reinfection [21] combined with mitigation measures still in place in winter, has played its role in the pandemic dynamics in 2021 in St. Petersburg. In our study, we did not see any seroreversion events with a maximum followup of ten months, which is in line with some other studies [22] . Populationbased vaccination was introduced in St. Petersburg in early 2021 and progressed slowly but involved primarily individuals who have not contracted the disease. Therefore, the sum of individuals seropositive after infection and the vaccinated individuals can approximate the number of protected individuals, yielding around 50-55% individuals with antibodies to SARSCoV2 by the end of April 2021. However, this approximation may not be valid in the future as more and more individuals who contracted the disease proceed to vaccination. One of the surprising findings, which other studies reproduce [23, 24] , is an association between seropositivity and smoking status. Seroprevalence was lower for smokers. That association was evident for both population samples in our study. Our study, however, does not answer the question, whether smokers are less likely to be infected or to develop less durable protection against infection [25] , which is more likely given higher IFR in men who smoke more often in Russia. More than 20,000 excess deaths have already been reported in St. Petersburg during the pandemic year [11] . The results of our seroprevalence study combined with the data on excess mortality give the IFR equal to 0.86% for the entire population, which is in line with other estimates across Europe [7, 24] . The IFR based on serological study results and excess mortality was stable for all four surveys. However, the official COVID19 death count provided lower IFRs, which were not stable and was even lower during the pandemic waves. Thus, it seems that the number of deaths during the both waves was unprecedented for St. Petersburg to timely provide official data collection and cause of death specifications in mortality records. We continue to monitor the pandemic in St. Petersburg using all available sources and plan to run the following survey to estimate the number of individuals with antibodies to SARSCoV2 after the summer wave. In addition, we aim to detect the 10 . 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 August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint herd immunity threshold in St. Petersburg if any exists given the Delta VOC basic reproductive number and diminished vaccine effectiveness [12] . Several possible limitations of our serological survey may require further explanation. Small sample size and high nonresponse rate compared to the number of phone numbers generated pose a challenge in two cases. First, when the obtained sample is small enough to make the study underpowered. Our sample size calculations show that under the 50% hypothetical prevalence scenario, our sampling error does not exceed 3% [8] . Second, a high nonresponse rate is a problem when there is an unac counted selection on observables or unobservables into the tested subsample. Under our study design, we observe a rich set of characteristics of individuals to account for nonresponse. In conclusion, our study provided an overall description of SARSCoV2 pandemic progression in the fourth largest European 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 August 2, 2021. . 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 August 2, 2021. Statistical appendix: estimating the IR and IFR with the Bayesian evidence synthesis model Observables We conduct K crosssections of serosurvey of adult population of St. Petersburg, Russia. In each crosssection k = 1,...,K we randomly select T k individuals to get tested out of P k individuals at risk of infection in the city. Out of those tested we identify CC k seropositive individuals with confirmed cases of SARSCoV2. In each crosssection k we also observe the cumulative number of deaths D k attributed to COVID19 since the pandemic onset in the city. We need to make inference on following variables: • C k -the cumulative total number of infected individuals by wave k, • IR k -the true infection rate (proportion of population which has been infected by crosssection k), which is the expected • IF R k -the true underlying infection fatality rate, which is the expected value App. 1 . 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 August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint To estimate the IFR across the study crosssections we closely follow [19] who proposed a simple framework for Bayesian evidence synthesis. We make the following assumptions on the distribution of the latent variables: Following [19] , to improve the MCMC mixing we replace the assumption for CC k with CC k ∼ Binomial (T k ,IF R k ). Then we can replace the conditional assumption D k |C k with the unconditional Next we assume that percrosssection IF R k and IR k are distributed according to a random effects model: Data To fit the model we need to acknowledge multiple data constraints. For P k we assume that the entire adult (≥ 18 years old) population of the city is at risk of infection (in a sensitivity analysis we consider the entire city population instead). For crosssections one and two we take the adult city population count as of January 1, 2020 from the Federal State Statistics Service of Russia * , 4 451 025 individuals. The data on the adult population as of January 1, 2021 is not available at the time of writing of this paper. However, the official data on the total city population is available † and amounts to 5 384 342 people as of January 1, * † App. 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 August 2, 2021. ; 2021 and 5 398 064 as of January 1, 2020. We assume that the adult population followed the same trend as the total population (a 0.25% decline) in 2020 and assume P k = 4 451 025 × (1 − 0 · 0025) = 4 439 897 individuals for crosssections three and four. We do not take the values T k and CC k directly from the percrosssection test data. To arrive at the seroprevalence estimate in our study we adjusted those naïve figures for test performance and nonresponse bias. Instead of using the raw counts we invert the reported 95% CI for the seroprevalence estimate for crosssection k. Using a beta prior on the probability of success for a binomial distribution, we can determine a twosided confidence interval from a beta posterior for any given T k and CC k . We define the values of T k and CC k that correspond to the reported seroprevalence 95% CI for ELISA Coronapass from Table 1 adjusted for nonresponse and test characteristics as the effective Another concern is reliability of the reported deaths data. We use two sources for D lower/upper k . The first is the official national government website ( ) that provides daily data on COVIDrelated deaths in St. Petersburg. The second is excess deaths estimation based on monthly data from the Federal State Statistics Service of Russia [11] . We find it valuable to compute the IFR and IR using the data from both sources given the voiced concerns about underreporting of COVIDrelated deaths in the country. For monthly excess deaths data we consider the cumulative excess deaths from January 1, 2020 to the month of the crosssection start to define D lower k and the cumulative excess deaths from January 1, 2020 to the month of the crosssection end to define D upper k . All the variables used in the estimation are reported in Table S1 . Table S2 . Perage and sex IR and IFR Our approach can be easily applied to another problem. Suppose now that k indexes sex and age groups within one serosurvey crosssection. Then we can use the same logic to estimate IR and IFR for each age groupsex combinations. We predict ELISA Coronapassbased seroprevalence within each sex and age group combination from our baseline univariate . 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 August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint and CC ef f ective k (see Table S3 ). For pergroup population P k we use data as of the beginning of 2020 since no data for 2021 is available yet. When it comes to D lower k and D upper k we need to acknowledge that, to the best of our knowledge, no official data on deaths from COVID19 disaggregated by age and sex exists. For this reason, we rely on excess deaths data estimation. We gather official yearly data on deaths in 2016-19 by age group and sex and quarterly data on deaths in 2020-21 to compute our D The estimated IR and IFR for each age and sexspecific group combination are in Table S4 . App. 4 . 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 August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint . 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 August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint * -current smoking status variable is gathered from the paperbased survey of tested individuals in the clinic during the first crosssection and is extrapolated for the same individuals for the second crosssection. For the third crosssection all individuals were asked about their smoking status during the phone interview. ** -for the purposes of the analysis we excluded vaccinated individuals from the tested subsample of individuals in the fourth crosssection, assumed that they failed to agree to get tested, and used their predicted seropositivity status from our univariate imputation model rather than the actual test results. App. 6 . 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. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint 95% confidence intervals in parentheses. "Interviewed" means individuals who agreed to participate in the respective crosssection of the phone survey. KOUZh2018 is the 2016 round of the Comprehensive Monitoring of Living Conditions household survey carried out by the Federal State Statistics Service of Russia. We subset this survey to include only adults in St. Petersburg. We report only completecase observations in terms of all variables, therefore the number of observations is slightly lower due to listwise deletion. * -current smoking status variable is gathered from the paperbased survey of tested individuals in the clinic during crosssection 1 (N = 949). App. 7 . 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. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint Figure S1 . Naïve and adjusted seroprevalence by study crosssection and week (ELISA Coronapass) Week of blood sample draw Seroprevalence, % (dots: naïve, solid line: adj. for non-response + 95% CI) App. 8 . 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. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint App. 9 . 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. (which was not certified by peer review) The copyright holder for this preprint this version posted August 2, 2021. ; https://doi.org/10.1101/2021.07.31.21261428 doi: medRxiv preprint Tracking excess mortality across countries during the COVID19 pandemic with the World Mortality Dataset World Health Organization. WHO Coronavirus (COVID19) Dashboard; 2020 Effect of changing case definitions for COVID19 on the epidemic curve and transmission parameters in mainland China: a modelling study The paramount importance of serological surveys of SARSCoV2 infection and immunity Prevalence of SARSCoV2 in Spain (ENECOVID): a nationwide, populationbased seroepi demiological study. The Lancet Seroprevalence of antiSARSCoV2 IgG antibodies in Geneva, Switzerland (SEROCoVPOP): a population based study. The Lancet Infection fatality risk for SARSCoV2 in community dwelling population of Spain: nationwide seroepidemiological study Seroprevalence of SARSCoV2 antibodies in Saint Petersburg, Russia: a populationbased study Serological evidence of human infection with SARSCoV2: a systematic review and metaanalysis. The Lancet Global Health Excess deaths associated with COVID19 pandemic in 2020: age and sex disaggregated time series analysis in 29 high income countries Excess mortality reveals Covid's true toll in Russia SARSCoV2 Delta VOC in Scotland: demographics, risk of hospital admission, and vaccine effectiveness. The Lancet Pathogen genomics in public health Association of the COVID19 pandemic with internet search volumes: a Google Trends analysis Human mobility trends during the early stage of the COVID19 pandemic in the United States We acknowledge personal support from Vitaly Nesis (Chief Executive Officer, Polymetal International, plc). We thank Alla Samoletova (European University at St. Petersburg) for administrative support and management of the study, Yulia Stepantsova (Chursina) for coordinating phonebased interviews, Lizaveta Dubovik, and Irina Shubina for the science communication. We thank the interviewers, nurses, general practitioners, and the Clinic "Scandinavia" personnel. We also thank all study partici pants.