Skip to main content

COVID-19 Antibody Seroprevalence in Santa Clara County, California

Our take —

Serosurveys allow us to estimate the number of people infected, which is important for understanding the spread and severity of the disease, as reported data are biased towards the most severe cases (and are a function of availability of testing). This is informative for predicting future spread and impact of the virus. Results of the first seroprevalence study in the United States from Santa Clara, California, suggest that seroprevalence is <5%, but the conclusion of dramatic under-reporting (1/50 to 1/85 infections are being detected) is not supported by the data. Given even a modest change in the assumed specificity of the serologic assay, the conclusions regarding prevalence and the infection fatality ratio change substantially. While under-reporting of COVID-19 cases is expected, we do not have sufficient evidence to conclude such dramatic under-reporting given the level of uncertainty in the estimated seroprevalence and the potential for bias.

Study design

Cross-sectional

Study population and setting

Serologic (IgG/IgM) testing by lateral flow immunoassay (LIFA) of 3,330 adults and children in Santa Clara County, California on April 3rd and 4th, 2020 (2718 adults, 612 children). Volunteer participants were recruited from the county’s general population by targeted Facebook ads who registered online and then were invited to visit a drive-thru test site. Each adult was allowed to bring one child from the same household with them.

Summary of Main Findings

Overall, 50/3300 individuals tested positive for antibodies to SARS-CoV-2 in Santa Clara County, resulting in a crude/unadjusted prevalence of 1.5% (95% CI: 1.11, 1.97%). Sampling weights were applied to correct for imbalances in zip code, race, and sex relative to the underlying population, and the variance was adjusted to account for clustering of adults and children from the same households. The adjusted prevalence was 2.81% (95% CI: 1.45, 4.16). The investigators further went on to make adjustments for sensitivity and specificity of the new diagnostic test, using the manufacturer’s data and their own internal validation procedure, resulting in updated prevalence estimates of 2.49% (95% CI: 1.80, 3.17%) and as high as 4.16% (95% CI: 2.58, 5.70%), respectively. Based on these estimates, the investigators infer that 48,000 to 81,000 people in the county have been infected. When compared to surveillance data showing that there were 956 confirmed cases in the region on April 1st, they infer that only 1/50 to 1/85 infections were being detected. Based on surveillance data of deaths (considering that some deaths may not have occurred yet), for 100 deaths in the county, they infer that the infection fatality ratio is 0.12 to 0.2%.

Study Strengths

This is the first reported serosurvey in a US population.

Limitations

Several limitations are noted. This was a volunteer-based convenience sample, raising concerns that people may have selectively signed up and received testing because they had Covid-19 symptoms and wanted an antibody test as confirmation. Secondly, given the limited data and evolving nature of current antibody diagnostics, the sensitivity and specificity of serological tests, including this one, need to be better characterized using diverse and well-characterized samples. The authors’ internal validation on their locally tested data estimated low sensitivity (as low as 67.6%) and high specificity (as high as 100%), which resulted in the highest upward adjustment of seroprevalence (4.16%). This is because it assumes that there are false negatives, but no false positives, while other studies using other LIFA methods have shown specificity for IgG ranging from (93% – 100%). Thirdly, weighting results had a large impact on prevalence, indicating non-representativeness of the sample and potential for highly influential observations; further, certain ages were under-represented and age was not included in the adjustments.

Value added

Findings indicate that seroprevalence is low in this county (<5%), informing the state of the currently limited potential for herd immunity.