Serosurveys can monitor key health disparities in SARS-CoV-2 infections and vaccinations

Well-designed serological studies can reveal what proportion of the population has been previously infected with SARS-CoV-2. We also show how these data can also be used to monitor health disparities in both infection and vaccination.

Like Comment
Read the paper

It has been over two years since the first COVID-19 lockdowns began around the world, but I want to take the reader back to that time in order to understand the origins of our recent study. As San Francisco went into lockdown, we had just begun collecting data for a new serosurveillance study. At that time there was still so much uncertainty about the SARS-CoV-2 virus, including about how common asymptomatic infection was. At the time tests of any kind were hard to come by and not widely available, meaning COVID cases were hugely underreported. It was difficult to ascertain just how many people were getting infected. As a result, we became interested in designing a study to better assess what proportion of the San Francisco population was previously infected.  

We launched a serosurveillance platform, where we combined Electronic Health Records and clinical laboratory records to algorithmically select blood samples taken for testing in clinical labs - both within the University of California, San Francisco health network, and the San Francisco Department of Public Health Network. We then tested these samples using an in-house assay for antibodies to SARS-CoV-2.  We wanted as representative a population as possible, so we chose samples taken for routine blood tests - primarily complete metabolic panels. We drew on samples from both a large health network and a public safety net network that captures underinsured and uninsured patients. We also excluded samples from individuals whose blood was drawn as part of a visit for COVID symptoms. We randomly selected stratified samples by age and zipcode. And then we sent the final sample list to the clinical labs, who set aside the samples for us after they were done with them. There was then a huge effort from many of the staff research associates and other contributors to this project, to then process these samples. carry out and validate the serological assays. 

 Our first study estimated that between 3 - 10% of all infections were reported between March and May 2020. These initial results also highlighted stark disparities within San Francisco – identifying large disparities in seroprevalence by race/ethnicity, neighborhood and homelessness status.

Box and whisker plots showing posterior estimates of seroprevalence, stratified by race/ethnicity. The shaded triangle shows the raw seroprevalence estimate. The midline of the boxplot shows the median of the posterior, the upper and lower edges of the box show the 25% and 75% quantiles, whiskers show 95% credible interval of the posterior. Points show posterior estimates outside of this interval.  Stars (*) indicate the race/ethnic groups where the 2.5% and 97.5% quantiles of the differences in posterior estimates for seroprevalence between samples from Hispanic patients and that group did not cross zero. Crosses (†) indicate the ethnic groups where the 2.5% and 97.5% quantiles of the differences in posterior estimates for seroprevalence between samples from Black or African American patients and that group did not cross zero.

 Now, I’d like to take the reader to another period of time that feels both close and far away - the very beginning of vaccine rollout in early 2021. Most who were not over 65 years old or healthcare workers were still not eligible for the vaccine at this point. It was then that we decided to carry out a round of further sample collection, to understand how things had changed since June 2020, and to measure disparities in both vaccination coverage and infection by race/ethnicity and neighborhood within San Francisco.

 This time, we tested samples using two different assays. One assay (Ortho Clinical Diagnostics VITROS Anti-SARS-CoV-2 ) tested for antibodies to the spike protein, which can be produced in response to either vaccination or past infection. When people are infected with SARS-CoV-2, they produce antibodies to another part of the coronavirus called the nucleocapsid protein, which aren’t produced in response to vaccination. We carried out a separate assay (Roche Elecsys Anti-SARS-CoV-2 ) which tests for antibodies to the nucleocapsid protein, to identify those who had previously been infected with SARS-CoV-2.   We then used the results of these assays, as well as data  from the San Francisco Department of public health on vaccination coverage by age and race within San Francisco to estimate the probabilities of being previously infected and previously vaccinated by zipcode, race/ethnicity and age. Once again, we found stark disparities, which have echoed those reported across many parts of the United States and around the world. We estimated that infection risk of Hispanic/Latinx residents was five times greater than of White residents aged 18–64 and that White residents over 65 were twice as likely to be vaccinated as Black/African American residents.  We also found that socioeconomically-deprived zipcodes in the South-East of the city had a higher probability of past infection and lower vaccination coverage than less deprived zipcodes.

Maps show a estimated probability of prior infection and b probability of vaccination by zipcode in San Francisco, as of February 2021.

As further waves of infection continue to affect communities and additional booster allocations are planned, serology continues to provide a powerful lens through which we can quantify these disparities directly, and continue to monitor the transmission dynamics of SARS-CoV-2. 

Isobel Routledge

Postdoctoral Scholar, University of California, San Francisco