A new study from Lawrence Berkeley National Laboratory (LBNL) makes a bold claim that increasing ventilation rates can cut the number of student absences related to illness. The lead author on the paper, Mark Mendell, is quoted by LBNL’s News Center as saying, “Our overall findings suggest that, if you increased ventilation rates of classrooms up to the state standard, or even above it, you would get net benefits to schools, to families, to everybody, at very low cost.”
But what did they have to assume to find that suggestion?
Looking at ventilation rates and illness absence
The study was published in the refereed scientific journal, Indoor Air, and is titled, Association of classroom ventilation with reduced illness absence: a prospective study in California elementary schools. What they did was pretty simple, really. They put dataloggers with carbon dioxide sensors in 162 classrooms of 3rd, 4th, and 5th graders. The classrooms were in 28 different schools in different parts of California. The dataloggers measured CO2 levels and turned them into averages for each 5 minute period. The researchers monitored the CO2 data for two years.
The next step was turning the CO2 levels into ventilation rates. Since people breathing indoors raise the CO2 level, the researchers developed a mathematical model to convert the amount of CO2 in the air into a ventilation rate. The higher the CO2 level in the classrooms, the lower the ventilation rate, and vice versa.
The way that the calculation was made was by dividing the CO2 generation rate of the students by the difference between the indoor and outdoor CO2 levels. The CO2 generation rate they used (0.0043 l/s) came from another study (Haverinen-Shaughnessy et al., 2011). For the outdoor CO2 level, they used an estimate of 400 parts per million. The indoor CO2 level they used was the maximum from each 15 minute moving average of the readings from their CO2 sensors.
When the researchers looked at their calculated numbers, which they did in a number of ways using statistical analyses, they found that ventilation rates in California schools are mostly below the code-mandated rate of 15 cubic feet per minute (cfm) per person. The table below shows what they found. It’s in liters per second rather than cfm. The reference point for your conversions is 15 cfm = 7.1 l/s, so you can see that only one of the lines below shows a rate above the required level.
After converting CO2 levels to ventilation rates, the team compared these rates to the classroom-by-classroom illness absence data. What they found was that as the calculated ventilation rates went up, illness absences went down. Combining the data from all the schools, they found that illness absences decreased by 1.6% for each additional 1 l/s of calculated ventilation. According to the Center News article, the illness absences keep decreasing all the way up to a ventilation rate of 15 l/s (30 cfm) per person.
That’s an interesting result!
Correlation and causality
One of the biggest traps in science is to confuse or conflate a correlation, two things that change together, with causality, a relationship in which one change causes another. It makes up a significant portion of what is called pseudoscience, and it’s also great for humor. Perhaps you’ve heard that global warming is caused by the decreasing number of pirates?
The abstract of the article states that the authors assumed causality from the correlation they discovered between calculated ventilation rates and illness absences: “Assuming associations were causal and generalizable…” The News Center article quotes Mendell on the topic: “We saw a correlation, but we don’t know if it’s directly causal.” That didn’t stop him from saying, as quoted at the beginning of this article, “…you would get net benefits to schools, to families, to everybody, at very low cost” by increasing ventilation rates.
The journal article itself shows a bit more restraint in some areas: “Caution should be exercised in extrapolating these findings,” the authors wrote in the Discussion section. “The relationships found here are consistent with, but do not prove, a causal relationship between increased VRs in elementary school classrooms and decreased illness absence.”
But then they throw that caution to the wind: “These findings suggest a potentially large opportunity to improve the attendance and health of elementary school students in California through provision of increased outdoor air ventilation in classrooms.”
They’ve found is a correlation, and a tenuous one, but they go on to suggest that increasing ventilation rates will result in huge financial benefits: “Total estimated benefits from VRs [ventilation rates] increased to 7.1 l/s-person are $113 million, over 25 times the estimated costs of 4.0 million. Total benefits from an increase to 9.4 l/s-person, $226 million, are over 30 times the estimated costs of 7.3 million.” Although they don’t project the cost, they even recommend going up to 15 l/s (30 cfm) at the end of the article.
The costs they estimated, however, do not include installing ventilation equipment in schools that don’t have any or changing out heating and cooling systems that cannot handle the extra load from the increased ventilation air. As with much of what they say, even the operational costs may be off significantly: “These estimates have a high expected level of uncertainty.”
Too many uncertainties; too many questions
Let’s start with the uncertainties. The researchers measured CO2 levels in the classrooms. They used those numbers to calculate ventilation rates, as I described above. The equation they used has three variables: the indoor CO2 levels they measured, the outdoor CO2 levels, and the rate of CO2 generation of the students in the classrooms.
The number that is most accurate is the one they measured, the indoor CO2 levels, but as with all experimental data, there were uncertainties. One that they mention in the section on strengths and limitations is the challenge of finding the equilibrium CO2 level. If, for example, someone breathes too close to a sensor, that jacks up the level for a while in that classroom. They did get a lot of data, though, so if there’s something to be learned from the CO2 data, they’ve probably learned it.
Going from classroom CO2 levels, though, to ventilation rates means that the uncertainties grow. They used a constant value for the CO2 generation rate of the children in the classrooms, and they used a constant 400 ppm for the outdoor CO2 level. Both of those numbers could be off by significant amounts. They tried to measure the outdoor CO2 at each school but the “data proved unusably erratic.”
They described other potential pitfalls with their methods as well, including problems with the student data. For example, some schools indicated they had “implausibly extended periods with no reported illness absences in any classroom” and some had a lot of “unverified and unexcused absences.”
In fact, the section on strengths and limitations is twelve paragraphs long. They described the strengths in the first half of the first paragraph. The remaining eleven and a half paragraphs were about the limitations.
Correlation does not imply causality
Maybe I missed an important part of their justification, but the case they made for causality in the News Center article and their research paper was… well, they didn’t make a case at all for causality. They just assumed it.
I think everyone in the field of building science understands that ventilation rates are related to indoor air quality (although they’re not the only factor) and certainly can have an impact on health. I tell people people whenever I talk about airtightness that they need to have mechanical ventilation, too.
My problem with what they’ve done here is that, rather than making recommendations to double ventilation rates, they should have concluded their paper by raising the many questions that still need to be answered:
How good is the model for calculating ventilation rates from CO2 levels? If they measured actual ventilation rates, how close would they be to their calculated numbers?
What other variables do we need to account for? Poor kids get sick more than kids from families with more money, but they didn’t do a whole lot with that in the article.
What does an increasing ventilation rate affect student absences more over longer time periods? The authors hypothesized that increasing the ventilation rate would have its maximum effect on absences in a 7-day period but found that the 21-day period was most affected.
There are just too many questions left open to be rushing to judgment about doubling the required ventilation rates in schools. A scientific paper in a refereed journal should not make recommendations based on assumptions. What can happen in such cases is that, because they appeared in a scientific journal, those recommendations get picked up by media and policymakers as having a validity they just don’t have.
Maybe in a few years we’ll have enough data to back up their claims, but for now, it’s way too early to turn their data into policy.
Correlation does not imply causality.