In this post, a summary of several studies on the spread of COVID-19 in child care and school environments. During the pandemic I spent considerable time collecting data on COVID in schools. These studies, however, come from other researchers.
Child Care Staff
As regular readers of this newsletter will know, this summer when I was collecting data on child care centers with my Google Forms, I talked up a study coming out of Yale on child care providers. It’s here! Or, at least the first paper is. It was published in Pediatrics. I contributed to a commentary on it, which was fun.
This paper focuses on risks to staff in child care settings; I assume there will be a follow-up on children, but given the much higher risk to staff, it’s really good to have started here. The Yale team collected data on about 57,000 child care providers early in the pandemic. They have data on whether the individual tested positive or were hospitalized with COVID (427 people) and whether they continued to see children in person. This last piece is really important because it allows them to have a control group: their main analysis compares people who continued to see kids in person to those who did not.
Their conclusion is that, in this sample, providers who see children in person are at no greater risk than those who did not (see below). There are COVID risks in both groups, but they are not higher in those who see kids.
Also informative in this study is the precautions taken. Not surprisingly, this was a fairly careful group. They report high rates of frequent hand washing and disinfecting toys, and about three-quarters are doing some kind of symptom screening or temperature checking. They did curbside pickup/drop off in nearly all cases to limit parent interaction. ( Masks were less common: only about 11% of kids and a third of staff).
As we titled our commentary: “Under the Right Conditions, Center-Based Child Care is an Unlikely COVID-19 Threat to Staff”.
Household Spread in India
I have been remiss in not discussing this paper from Science from late September on contact tracing in India. This is a really impressive effort to do comprehensive contact tracing in a low resource environment (two states in India) as well as look at infection and death rates.
The authors use evidence on 575,000 contacts of 84,000 cases, and they look at “secondary attack rate” — that is, the chance that each contact is infected by a case. In their data, this infection happens about 11% of the time for high risk contacts, and about 5% of the time for low risk. Being in a household together: about 9% transmission; health care setting, 1.2%; being in a car together for 6 hours, 80%.
The authors have some helpful graphs of transmission by age groups below, which basically finds that there seems to be a lot of transmission from very young kids to other very young kids. This got a lot of attention and you can see it in the data here. It’s definitely more concerning around child-to-child transmission, something we haven’t seen much of.
Caveats to this study include the fact that it is difficult for them to identify who is the index case; we saw results like this in South Korea for older children, and in that case it turned out that a lot of the transmission was actually simultaneous infection. These authors note that it is difficult to know whose infection was first. And from the standpoint of the US, the household and general environment in India is somewhat different.
Bottom Line: In this setting, we’re seeing evidence of more kid-to-kid transmission. Overall risk of infection from household contacts is about 10%.
School Reopening in Germany
Moving to older kids… One of the key school questions is whether kids will become infected at school and spread to the community. Will we see opening of schools drive community rates? I will say this has become a big question in the last week as many places have seen rates tick up and their schools are open. Of course, the general move together isn’t enough and this is begging for an analysis of community rates based on school reopen dates, but we do not yet have that in the US.
Where we do have that analysis, as of this week, is Germany.
The paper is here, entitled: School Re-Openings after Summer Breaks in Germany Did Not Increase SARS-CoV-2 Cases. The authors exploit the fact that German states differ in when they return to school after summer break. They are able to look at whether cases spike after school return. Basically, this is like asking: Georgia goes back to school in August, and Rhode Island not until September. Do we see spikes in Georgia in August and Rhode Island in September? This type of analysis is commonly called an “event study”.
Results are typically shown in graphs like the ones below (from their paper) which show the evolution of infection rates around school reopens. Basically, you can think of these as lining up all the German states so “time zero” is the date of school reopen, which differs by state, and then looking at how infections evolve in different age groups as time goes on. What is useful about having places with different start dates is the authors can control for general time effects.
Their bottom line, as suggested by the title, is that school reopening did not drive case rates. If anything, for kids aged 0 to 14 the rates actually seem to decline a bit after school reopens. They attribute this to the fact that there is more mixing and travel on summer break.
Notable here is that they see flat rates among adults. At least in this setting, the concern that kids are getting the virus, being asymptomatic and spreading it around at home does not seem to be valid.
Secondary Schools in Sweden
Finally, we have data from Sweden on infection around secondary school kids, parents and teachers. In case you haven’t been following: Sweden took a somewhat different approach from the rest of Europe and didn’t lock down very much. Many schools stayed open, and they generally kept restaurants and other businesses open. This has been criticized by many, an certainly their death rates early in the pandemic were far worse than comparable places in Europe.
The idea in this paper is to compare risks for teachers and parents of lower versus upper secondary school students. Lower secondary schools (think of this as the first part of high school) remained open. Upper secondary schools did not. (Primary schools and child care centers were open, and earlier data suggested limited risk to staff in those settings).
The authors find evidence that lower secondary school teachers are at higher risk than upper secondary school. That is: it would appear that being in person at work increased the risk of infection, both overall and serious illness. Parents of children in lower secondary school did not seem to be affected, but there is some evidence of effects for partners of teachers.
In terms of magnitudes, the authors estimate that 150 cases of COVID-19 would have been prevented if lower secondary schools had closed.
The authors put in a number of cautions — they do not know if teachers got COVID-19 at school or from kids, and it is notable that there were very few precautions taken. Schools were open at normal density, and masks are rare. But at a minimum this points to the need for precautions among staff in settings with older children.
(You might ask about student infections — they do not study them. Due in at least large part to very limited testing, there were only 87 confirmed cases of COVID-19 in students 7 to 16 in Sweden in this period, in a population of 1.2 million. This is too small a number to analyze).
Any Closing Thoughts?
This is a lot of new data. Most of it is reassuring, I think, especially on younger kids and especially the data from the US. But it also underscores the need for vigilance. The child care centers were sanitizing. Schools in Germany have some significant precautions (limited masks, though). Household transmission in India, where precautions are likely to be low and density high, is high among kids. Unmasked, fully dense secondary school does seem to increase infections among teachers.
What I am most encouraged by, I guess, is that we are starting to get these data in. I hope the next weeks will see more evidence of this type, and we can start to build a more coherent picture of what is safe and what is not.
