What fraction of populations are exposed in huge #COVID19 epidemics & what drives transmission & exposure?
Serosurvey from Spain finally published!
Tons of major findings here & I haven't seen a detailed summary so here's one.
Thread
doi.org/10.1016/S0140-…
Serosurvey from Spain finally published!
Tons of major findings here & I haven't seen a detailed summary so here's one.
Thread
doi.org/10.1016/S0140-…
Goal of study: assess the fraction of people across Spain that have been exposed to SARS-CoV-2 by screening for antibodies.
Spatial scale: country wide!
Sample size: huge! 61K
Why does this matter? Can look at spatial variation, within population comparisons (age, gender, prof.)
Spatial scale: country wide!
Sample size: huge! 61K
Why does this matter? Can look at spatial variation, within population comparisons (age, gender, prof.)
Methods details (that matter!)
1) Study recruited participants using random sample of national database of households. So much better than convenience samples (e.g. using Facebook ads, blood donors, grocery stores). Refusal rate low (18%) so results unlikely to be heavily skewed.
1) Study recruited participants using random sample of national database of households. So much better than convenience samples (e.g. using Facebook ads, blood donors, grocery stores). Refusal rate low (18%) so results unlikely to be heavily skewed.
2) Used 2 tests - fast 10 min point of care test (POC) & an Abbott chemiluminescent microparticle immunoassay test (IA)
Internal validation:
POC test: sensitivity 82.1%, specificity 100%
IA: sens: 89.7, spec: 100%
Tests didn't always agree so results presented multiple ways
Internal validation:
POC test: sensitivity 82.1%, specificity 100%
IA: sens: 89.7, spec: 100%
Tests didn't always agree so results presented multiple ways
2 cont) In results below I mostly present serop from POC test; Serop from IA test was often, but not always, ~0.5% lower-diff. individs sampled
3) Unlike other studies they re-weight results carefully to match pop they want to describe within each province by age, income, sex.
3) Unlike other studies they re-weight results carefully to match pop they want to describe within each province by age, income, sex.
Key Results (many!)
As elsewhere, HUGE variation in seroprevalence 1.2%-14.4% (Madrid 11.3), with higher prev in more populous provinces but variation w/in those as well. "Spain" didn't have huge COVID epidemic - some provinces had large epidemic while others much less affected.
As elsewhere, HUGE variation in seroprevalence 1.2%-14.4% (Madrid 11.3), with higher prev in more populous provinces but variation w/in those as well. "Spain" didn't have huge COVID epidemic - some provinces had large epidemic while others much less affected.
Note that this survey was done Apr 27-May 11. About ~80-85% of today's cumulative cases had occurred by 4/20 & would have been detectable, so July seroprevalence is likely a bit higher in many locations (10-40% higher: 6% instead of 5% for Spain overall)
ourworldindata.org/coronavirus/co…
ourworldindata.org/coronavirus/co…
AGE
Interesting & puzzling patterns & different b/w 2 tests (?); please read CAVEATS below
-Both tests show 16%-33% lower seroprev in 5-19 yr olds
-POC test shows increase w/ age peaking at 70yr
-IA test: less variation; small N for kids
-Both show lower serop for old but N low
Interesting & puzzling patterns & different b/w 2 tests (?); please read CAVEATS below
-Both tests show 16%-33% lower seroprev in 5-19 yr olds
-POC test shows increase w/ age peaking at 70yr
-IA test: less variation; small N for kids
-Both show lower serop for old but N low
CAVEATS for AGE analysis
Despite large sample sizes, error bars are big, and large overlapping confidence intervals. This means variation among ages might be much smaller than crude estimates - i.e. 95% CIs of kids 10-19 overlap w/ adults 20-39.
SEX: No diff b/w male/female
Despite large sample sizes, error bars are big, and large overlapping confidence intervals. This means variation among ages might be much smaller than crude estimates - i.e. 95% CIs of kids 10-19 overlap w/ adults 20-39.
SEX: No diff b/w male/female
SYMPTOMS
-Fraction asymptotic 32.7% (based on recall so likely slight overestimate); sadly not broken out by age
-Only 16.9% of people w/ COVID symptoms seropositive (most - 83%- that thought they had COVID, didn't)
-Fraction asymptotic 32.7% (based on recall so likely slight overestimate); sadly not broken out by age
-Only 16.9% of people w/ COVID symptoms seropositive (most - 83%- that thought they had COVID, didn't)
ANTIBODY DEVELOPMENT
89% of people that self-reported a PCR+ test >14 days ago tested positive for antibodies suggesting that the fraction of infected people that develop T-cell immunity but no antibodies is <10%. Phew.
89% of people that self-reported a PCR+ test >14 days ago tested positive for antibodies suggesting that the fraction of infected people that develop T-cell immunity but no antibodies is <10%. Phew.
SOCIO-ECONOMIC STATUS
Surprisingly little variation; if anything seroprevalence a little higher in highest/richest percentile.
Surprisingly little variation; if anything seroprevalence a little higher in highest/richest percentile.
JOB STATUS
Highest for active workers (expected) and retired (surprising). Lowest for unemployed & unpaid social work.
Highest for active workers (expected) and retired (surprising). Lowest for unemployed & unpaid social work.
PROFESSION
Highest in health care & nursing home (yikes!). Lowest in retail, cleaning, "other"; moderate in 1st responders, transport
Highest in health care & nursing home (yikes!). Lowest in retail, cleaning, "other"; moderate in 1st responders, transport
HOUSEHOLD SIZE
Surprisingly to me, highest in smaller households; lowest in big households. Hard to explain this given high serop if household had confirmed infection. Do larger families mix less w/ other households?
Surprisingly to me, highest in smaller households; lowest in big households. Hard to explain this given high serop if household had confirmed infection. Do larger families mix less w/ other households?
Interpretation
-Fantastic study (great when science is done right!)
-Seroprev still quite low despite large number of deaths, especially in Madrid
-Children infected but slighly lower serop. Study speculates ACE-2 receptors but behavioral "shielding" also possible explanation
-Fantastic study (great when science is done right!)
-Seroprev still quite low despite large number of deaths, especially in Madrid
-Children infected but slighly lower serop. Study speculates ACE-2 receptors but behavioral "shielding" also possible explanation
-Serop in older individs still low - large at-risk pop remaining
-Expected higher exposure of health care; worrisome high serop of nursing staff (infection from SNF residents?)
-Unexpected: little variation by socio-economic status
-Surprising decreasing serop w/ household size.
-Expected higher exposure of health care; worrisome high serop of nursing staff (infection from SNF residents?)
-Unexpected: little variation by socio-economic status
-Surprising decreasing serop w/ household size.
-Highly spatially variable serop - higher in central provinces w/ many people, lower in coastal regions; result of more introductions or differential behavior?
Here's a thread & a paper that also indicate fraction with just T-cell response & no antibodies is LOW.
medrxiv.org/content/10.110…
medrxiv.org/content/10.110…
