Should this school's data be merged with a school that's fully open?
My thoughts about Covid in schools-
Our studies are crap. We haven't put appropriate resources into studying this. Merging "in-person school" v "remote school" across the country makes little sense since there's so much heterogeneity in what "in-person school" looks like. 🧵
Should this school's data be merged with a school that's fully open?
Until then, we must rely on what we know on where the virus spreads best, and admit that schools have many of those qualities.
More from Education
Normally I enjoy the high standards of journalism in @guardian . Not today as disappointed with misleading headline that suggest infections are spreading fastest in children. It'll worry parents/teachers & I doubt most readers will unpick the
The latest REACT1 report shows prevalence of infection in ALL age groups has fallen, including children aged 5-12 from 1.59% in Round 8 to 0.86% in Round 9a. The authors of REACT1 report also (wisely) didn't try to interpret the prevalence figures.
If this were a research trial you wouldn't place much weight on the age differences in % prevalence because of the wide confidence intervals, i.e. differences weren't statistically significant.
3/
I've previously tweeted on the challenges (& dangers) of interpreting surveillance data. One would need lots more contextual info to make sense of it & arrive at sound
Undoubtedly some will extrapolate from the prevalence of infection figures in children to other settings i.e. schools based on the headline. I'd advise caution as there is a real risk of over-interpretation through extrapolation of limited data. Association is not causation.
5/
The latest REACT1 report shows prevalence of infection in ALL age groups has fallen, including children aged 5-12 from 1.59% in Round 8 to 0.86% in Round 9a. The authors of REACT1 report also (wisely) didn't try to interpret the prevalence figures.
If this were a research trial you wouldn't place much weight on the age differences in % prevalence because of the wide confidence intervals, i.e. differences weren't statistically significant.
3/
I've previously tweeted on the challenges (& dangers) of interpreting surveillance data. One would need lots more contextual info to make sense of it & arrive at sound
Misinterpretation of surveillance data is a serious issue. Surveillance data needs to come with a warning label - Open to biases - interpret with caution! Some may not realize that surveillance often does not measure all infection, it's a proxy for actual disease incidence.
— Andrew Lee (@andrewleedr) February 14, 2021
1/
Undoubtedly some will extrapolate from the prevalence of infection figures in children to other settings i.e. schools based on the headline. I'd advise caution as there is a real risk of over-interpretation through extrapolation of limited data. Association is not causation.
5/