key: cord-0078836-er8mbx8x authors: Mehrotra, Megha L.; Lim, Esther; Lamba, Katherine; Kamali, Amanda; Lai, Kristina W.; Meza, Erika; Szeto, Irvin; Robinson, Peter; Tsai, Cheng-ting; Gebhart, David; Fonseca, Noemi; Martin, Andrew B.; Ley, Catherine; Scherf, Steve; Watt, James; Seftel, David; Parsonnet, Julie; Jain, Seema title: CalScope: Monitoring SARS-CoV-2 Seroprevalence from Vaccination and Prior Infection in Adults and Children in California May 2021– July 2021 date: 2022-05-13 journal: Open Forum Infect Dis DOI: 10.1093/ofid/ofac246 sha: 4caec69fbcd64bfc92e2765a69f0396621437ccf doc_id: 78836 cord_uid: er8mbx8x BACKGROUND: Understanding the distribution of SARS-CoV-2 antibodies from vaccination and/or prior infection is critical to the public health response to the pandemic. CalScope is a population-based serosurvey in 7 counties in California. METHODS: We invited 200,000 randomly sampled households to enroll up to 1 adult and 1 child between April 20, 2021 and June 16, 2021. We tested all specimen for antibodies against SARS-CoV-2 nucleocapsid and spike proteins, and each participant completed an online survey. We classified participants into categories: seronegative, antibodies from infection only, antibodies from infection and vaccination, and antibodies from vaccination only. RESULTS: 11,161 households enrolled (5.6%), with 7,483 adults and 1,375 children completing antibody testing. As of June 2021, 33% (95%CI [28%, 37%]) of adults and 57% (95%CI[48%, 66%]) of children were seronegative; 18% (95%CI[14%, 22%]) of adults and 26% (95%CI[19%, 32%]) of children had antibodies from infection alone; 9% (95%CI[6%,11%]) of adults and 5% (95%CI[1%, 8%]) of children had antibodies from infection and vaccination; and 41% (95%CI[37%, 45%]) of adults and 13% (95%CI [7%, 18%]) of children had antibodies from vaccination alone. CONCLUSIONS: As of June 2021, a third of adults and most children in California were seronegative. Serostatus varied regionally and by demographic group. By July 2021, the United States had recorded more than 34 million COVID-19 cases 2 and 600,000 deaths, with over 3.7 million cases and 60,000 deaths in California. 1 Though all 3 adults and children over 12 have been eligible for COVID-19 vaccination since May 2021 in 4 California, vaccine uptake has been uneven; as of July 31, 2021, the percent of persons fully 5 vaccinated ranged from 24% to 79% across California counties. 6 The California Department of Public Health (CDPH) monitors COVID-19 burden and 7 forecasts hospitalizations to determine when additional mitigation measures are required 8 to avoid overwhelming the healthcare system. 2 Both prior SARS-CoV-2 infection and 9 vaccination reduce the risk of symptomatic COVID-19 and hospitalization, although 10 questions remain regarding the relative level and duration of risk reduction. 3-7 Accurately 11 forecasting future COVID-19 surges requires estimating population immunity from prior 12 infection or vaccination in order to determine how many people remain susceptible to 13 infection. Estimating population immunity using routine surveillance data is challenging. 14 Since COVID-19 may be asymptomatic and persons with mild illness may not seek testing, 15 many infections are not recognized or reported. Recent studies estimated that 70% of 16 SARS-CoV-2 infections in California were unaccounted for in the CDPH COVID-19 17 surveillance system by December 2020. 8 Population-based serosurveys can estimate immunity from prior infection or 19 vaccination without the limitations inherent in routine surveillance, and several 20 seroprevalence studies have been completed or are currently underway throughout the 21 United States. 9-15 However, the studies conducted thus far in California have been limited to CalScope is a repeated cross-sectional study using random address-based sampling 10 of households in seven counties in California. The study re-samples households with 11 replacement over three timepoints. Sampling Strategy 13 We used a multistage sampling strategy to allow for region-specific seroprevalence 14 estimates. The sampling approach was guided by principles of causal transportability 18 to 15 ensure that the final study results could be appropriately and efficiently generalized to the 16 general population (Appendix). We sampled households in seven counties: Alameda, El 17 Dorado, Kern, Los Angeles, Monterey, San Diego, and Shasta. 18 We used an address-based sampling frame created by Marketing Systems Group to 19 select a probability sample of households within each county. The frame uses the United 20 States Postal Service Computerized Delivery Sequence File, which covers all residential 21 delivery locations in the United States, with each address geocoded and linked to the 2015 American Community Survey. 19 We oversampled households from census tracts with 1 higher proportions of Black households to ensure adequate representation. To enroll a 2 total of 10,000 households, we sampled 200,000 households per wave distributed across 3 the seven counties proportional to each county's population with a minimum of 15,000 4 households sampled per county. Sampled households could enroll one adult and one child (6 months to 17 years 6 old). To randomize which eligible household members participated, we instructed 7 households to enroll the adult and child with the next upcoming birthday. 19 Antibody Testing 20 Participants were mailed at-home antibody test kits with instructions on how to 21 collect a dried blood spot (DBS) specimen and were asked to return their sample to Enable Biosciences within 30 days. Specimens with inadequate volume or collected >30 days before receipt by the laboratory were rejected. All valid specimens received by the 1 laboratory by August 1, 2021 were included in this analysis. 2 Specimens were tested for both anti-spike and anti-nucleocapsid antibodies using 3 Enable's ADAP SARS-CoV-2 total antibody assay. The assay procedures have been 4 described previously (Appendix). 20 The assay cutoffs were established by testing 100 5 healthy controls and set at 99.7% percentile. The cutoffs for spike and nucleocapsid 6 antibodies were 3.00 ΔCt and 1.50 ΔCt respectively. In a validation study including 31 PCR-7 positive COVID-19 cases and 80 healthy blood donors, the assays were shown to be 100% We anticipated that households that enrolled in the study and completed antibody 12 tests would differ from those that did not respond. Thus, we constructed sampling weights never registered for the study, candidate variables for the Step 2 weights were limited to 17 the neighborhood-level characteristics listed above. 18 We used a cross-validated ensemble machine learning algorithm, SuperLearner, 27 to 19 estimate inverse probability of selection weights for both Step 1 and Step 2. We included a 20 mixture of parametric and machine learning algorithms in the SuperLearner. Weights for Step 1 were estimated separately for adults and children within each county. Step 2 weights were estimated at the household level within each county. Finally, we used the 1 known sampling probabilities for each invited household to construct the Step 3 weights. 2 We multiplied all three weights and used iterative proportional fitting (raking) to 3 calibrate the combined weights to ensure that the weighted distribution of age, sex, Using the sampling weights, we estimated the proportion of the population in each 19 serostatus category and with evidence of prior infection for the whole sample and stratified 20 by county, age, race/ethnicity, and HPI quartile. We used a non-parametric bootstrap with 21 1000 replicates to obtain 95% confidence intervals. We estimated the ratio of the number of SARS-CoV-2 infections to confirmed cases CalScope had higher levels of education, higher household income, and were less likely to 19 be Latinx compared to households that did not participate (Table 1 and Supplementary 20 Table 1 ). Table 1 shows the demographics of the study sample before and after weighting. The median specimen collection date was May 22, 2021, with 60% of specimens collected Finally, adults living in HPI quartiles 1 or 4 were less likely to be seronegative than 10 adults living in HPI quartiles 2 or 3. (Table 3) In contrast, children living in HPI quartile 1 11 were less likely to be seronegative compared to those in the higher HPI quartiles. (Table 4) 12 Our results suggest that these targeted vaccination campaigns have been effective -8 seropositivity due to vaccination is similar for adults in the lowest and highest HPI 9 quartiles (56% in both). Though the proportion of children and adults who had been previously infected was 11 similar, most children remained seronegative as of June 2021 because they were not yet 12 eligible for vaccination. Even among those eligible (age 12-17), vaccination coverage has 13 been low, and many were still seronegative. With in-person schooling resuming in much of 14 the state, it will be important to encourage vaccination of all age eligible children. 15 Other studies using remnant commercial laboratory specimens have found that 16 seroprevalence from prior infection may be higher among children than adults. 32,33 These 17 studies may overestimate seroprevalence from prior infection in children because children 18 receiving bloodwork may be more likely to be sick compared to the general population. 19 However, our estimates are uncertain and do not preclude the possibility that 20 seroprevalence from prior infection is higher in children than adults. We anticipated that those who enrolled in our study might not be representative of 18 our target populations, so we used causal transportability to design our study and survey 19 instrument to generalize our results to our target populations. Our weighted results can be 20 considered unbiased estimates of SARS-CoV-2 serostatus in our target population only if 21 we assume that we were able to measure and adjust for all the differences between the 22 A C C E P T E D M A N U S C R I P T study sample and target population that were associated with SARS-CoV-2 serostatus. 1 Given the low response rate in CalScope, it is likely that our weights may have excluded 2 some relevant characteristics that differed between the sample and target populations, and 3 our estimates may still suffer from residual non-response bias. Additionally, we relied on 4 self-reported vaccination to classify serostatus, which may be subject to social desirability 5 or recall biases. However, our results are in line with those from other studies and known 14 It is critical that public health agencies monitor who does not have SARS-CoV-2 antibodies 15 to accurately forecast future COVID-19 surges. CalScope will begin collecting data for Wave Step 1 Weights, T=1 indicates that an individual has a valid antibody test result, Z is a 9 vector of individual-level measurements from the survey instrument. Step 3 and 2 weights 10 were estimated at the household level. Step 1 weights were estimated at the individual 11 level with weights estimated separately for adults and children. The combined weight is the 12 product of all 3 weights. United States COVID-19 Cases and Deaths by State over Time | Data | Centers for Disease 2 Control and Prevention Efficacy and Safety of the mRNA-1273 SARS-CoV-2 Safety and Efficacy of the BNT162b2 mRNA Covid-8 The Incidence of SARS-CoV-2 Reinfection in Persons 10 With Naturally Acquired Immunity With and Without Subsequent Receipt of a Single 11 Dose of BNT162b2 Vaccine COVID-19 Cases and Hospitalizations by COVID-19 Vaccination Status and 14 Previous COVID-19 Diagnosis -California and Protection Duration of COVID-19 Vaccines: Waning 17 Effectiveness and Future Perspective SARS-CoV-2 Cumulative Incidence and Period 19 Seroprevalence: Results From a Statewide Population-Based Serosurvey in California Cohort profile: a national Sites in the United States SARS-CoV-2 seroprevalence and neutralizing 4 activity in donor and patient blood Prevalence of SARS-CoV-2 antibodies in a large 6 nationwide sample of patients on dialysis in the USA: a cross-sectional study COVID-19 antibody seroprevalence in Santa 9 Seroprevalence of SARS-CoV-2-Specific Antibodies 11 Among Adults in Transportability of Causal and Statistical Relations: A Formal 14 Approach A serological assay to 18 detect SARS-CoV-2 antibodies in at-home collected finger-prick dried blood spots External Validity: From Do-Calculus to Transportability 21 Coronavirus 2 Seroprevalence and Reported Coronavirus Disease SARS-CoV-2 antibody magnitude and 19 detectability are driven by disease severity, timing, and assay Immune correlates analysis of the 22 mRNA-1273 COVID-19 vaccine efficacy clinical trial mRNA-based COVID-19 vaccine 24 boosters induce neutralizing immunity against SARS-CoV-2 Omicron variant Evidence for antibody as a 27 protective correlate for COVID-19 vaccines Neutralizing antibody levels are highly 29 predictive of immune protection from symptomatic SARS-CoV-2 infection