The objective of this report is to both summarize and critique the COVID-19 prevalence survey released by the Ohio Department of Health and The Ohio State University [1].  The motivation for this report is that Governor DeWine promoted this survey as very important, and it used tax dollars so these results should be accessible to the general public.  As noted below, there are some serious concerns about how this survey was performed and interpreted.  

In this study, 30 census tracts were sampled throughout the state during 20 days in July.  Both PCR and antibody tests were conducted on study participants, and demographic data and behavioral data regarding mask wearing and social distancing behaviors were also collected.  The study used Bayesian latent class models to estimate prevalence, arriving at an estimate of COVID-19 prevalence of 0.9% (95% credible interval, 0.1% to 2.0%), and “past COVID-19” (presumably derived from the antibody survey) at 1.5% (95% credible interval, 0.3% to 2.9%).   In a press conference, one of the clinical scientists involved in the study commented on this somewhat low antibody seroprevlence by concluding that antibody levels must only persist for 3 months.  

However, there are a great deal of methodological concerns with this study.  First, the original objective was to sample 1200 individuals, but the final sample was 727.  What was the reason for this 60.5% recruitment success rate, which is somewhat low for clinical study standards?  Second, Northwest Ohio was oversampled quite a bit (N=162 for this region, which is not 1/8 of 727).  This does not reflect population density, since Northeast Ohio, Central Ohio near Columbus, and Southwest Ohio are actually the most populous regions of the state.  Only 81% had valid antibody tests, which is not explained, and draws some concern about the validity of this test in general.  

In addition, these results do not correspond with what is known about COVID-19 prevalence both in Ohio and across the country.  Projections from this prevalence survey leads to an estimate of ~133,000 individuals have "evidence of past COVID-19".   As of October, the case count was over 153,000.  The USC seroprevalence study showed that seroprevalence was 28-55 times greater than case count [2,3].  That study showed a seroprevalence of 4%, which in Ohio would yield a number ~468,000.   Moreover, the sensitivity and specificity of the test used was not stated, which has been reported by other seroprevalence studies, because it is well-established that there is wide variability in the accuracy of these tests.  The conclusion that antibody responses do not last long is inconsistent with recent published studies showing that antibody production persists for 5-7 months [4,5].  An alternative hypothesis is that subjects who were exposed but did develop symptoms as early as January may no longer show an antibody response; this hypothesis is consistent both with aforementioned literature suggesting antibody responses wane after 5 months and COVID-19 cases appearing in the Ohio coronavirus .csv file as early as January.  Looking at the data another way, it should raise concern that the estimated prevalence of both COVID-19 positivity based on PCR and antibody are so similar, when they are testing two very different things.  

In addition to the limitations mentioned above regarding subject recruitment and COVID-19 data, there are several other limitations that raise concerns about the study’s validity.  While the demographic characteristics of the sample are described, they are not compared to the demographic characteristics of the state of Ohio as a whole.  This raises the question, is this sample representative? The California seroprevalence study cited above adjusted for demographic differences between their sample and the overall population, and that is lacking here.  While average age is a valid statistic, it is not as useful as distribution, to better understand if older or younger individuals were oversampled.  Third, there was a lost opportunity to examine the association between behavioral factors with antibody positivity and PCR positivity.  Do people that practice more restrictive behaviors have a lower frequency of antibody response and/or COVID-19 positivity?  A recent study published by the CDC demonstrated that COVID-19 positive or negative status was not associated with mask-wearing behavior [6].  If the Ohio Department of Health wants to promote mask-wearing and social distancing, they should use actual data to support those activities.  When data analyses such as these are neglected, it suggests they were left out of the report because of “negative” (non-significant) findings, which is of serious concern.  Lastly, and perhaps most importantly, there is no mention of how many COVID-19 positive individuals actually demonstrated symptoms.  There is a wealth of literature (cited separately, below) suggesting that the rate of transmission from truly asymptomatic individuals is quite low, most studies estimating 2-4%, with some large studies showing zero transmission. 

Legislators should take serious notice of this critique. Tax dollars may not have been used appropriately in this endeavor, and the conclusions touted by the governor and his clinical research team are potentially invalid given the weaknesses of the study.  

Cited References

1.  https://coronavirus.ohio.gov/static/dashboards/prevalence-covid19-ohioadults.pdf?fbclid=IwAR2Z_RhhfqYKAp9udYfR7COs536XV-0mU1LCesmcFAIjNl7B-oYhVM3LX_U

2. https://pressroom.usc.edu/preliminary-results-of-usc-la-county-covid-19-study-released/

3. https://www.youtube.com/watch?v=C_jXKcp4Zyg&feature=youtu.be

4. https://www.cell.com/immunity/fulltext/S1074-7613(20)30445-3

5. https://science.sciencemag.org/content/early/2020/10/27/science.abd7728

6. https://www.cdc.gov/mmwr/volumes/69/wr/mm6936a5.htm 

Literature showing asymptomatic spread is rare (<5%)

https://www.nature.com/articles/s41591-020-0962-9?fbclid=IwAR1FJ5ys6pibMA7fP4J9YjZix6YlQk2suBr1-UxEF5RfzCMmb00FtVovCAs

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7219423/pdf/main.pdf

https://www.acpjournals.org/doi/10.7326/M20-2671?fbclid=IwAR1Uw760JaCjG4g6GtWs1TUDVjdiwV6cluqSrcNUyUqmL6i9dLOWHNCbd4w&

https://pubmed.ncbi.nlm.nih.gov/33061473/

https://pubmed.ncbi.nlm.nih.gov/33035448/

https://pubmed.ncbi.nlm.nih.gov/33015099/ 

https://pubmed.ncbi.nlm.nih.gov/33015099/ 

https://pubmed.ncbi.nlm.nih.gov/32632012/

https://pubmed.ncbi.nlm.nih.gov/32574282/

https://pubmed.ncbi.nlm.nih.gov/32562791/

https://pubmed.ncbi.nlm.nih.gov/32544669/

https://pubmed.ncbi.nlm.nih.gov/32525469/ 

https://pubmed.ncbi.nlm.nih.gov/32497212/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7166025/