Risk vs. Uncertainty in the Question of Opening Schools

We've had lots of conversation in our household lately on the question of whether to open schools for in-person learning this fall. Even if you don't have children or grandchildren affected, this is a broad question of public health: Not only should we personally go back? but also should anyone be allowed to go back given the current state of the pandemic? or even should anyone be allowed to keep their kids at home? (e.g., based on hypothesized risks to mental health or learning). These are difficult questions because they involve competing goals such as education versus health, parents' work demands versus children's support needs, and also because they must be made with limited information and under time constraints. These two features make real-world decisions particularly difficult, as I described in a recent post about naturalistic decision-making.

Another framework that may be helpful as we think about whether to re-open school buildings is that of risk versus uncertainty, articulated in 1921 by economist Frank Knight. The difference between the two is that risks can be calculated with some level of precision -- they are "known unknowns" like the jar on the left in the image above, where a certain color of ball comes up with a certain probability. Uncertainties, on the other hand, are "unknown unknowns" that we are aware of but unable to calculate with precision. They are like the jar on the right, where you have to pick a ball and will get a certain color but you can't see the balls in advance. When authorities like the American Academy of Pediatrics or Colorado's Governor argue that school buildings will be "reasonably safe," they are suggesting that the situation is one of calculated risk rather than one of uncertainty.

To treat uncertainty as risk means dealing with it through the Narrative mind; if it's a calculated risk that means we can see the calculation. The most honest of the recent proposals to re-open school buildings do offer these kinds of statistical nuances: "a very small percentage of children will die," or "relatively few cases of COVID-19 have been linked to children as the source of transmission." (Both of these are true, so far as the literature goes, although as we will see below that literature is still very thin). There are still questions to be asked based on one's risk tolerance: How many dead elementary schoolers are too many? Or how many dead grandmothers infected by their grandchildren? But in principle it seems like one could make a cold-blooded calculation that balances health risks against competing objectives like economic turmoil, poor mental health, and sub-par learning outcomes. The relative values of these social goods are of course open to debate, but let's not get bogged down in details yet -- my point is just that when risks are knowable, a rational decision can be made.

Let's not be fooled, though: The situation is still one of uncertainty rather than risk. On many points, we may have a feeling that the answer is obvious but in reality the data are still coming in (I actually re-wrote this section three times because of emerging findings, and by the time you read it the information may be out of date again!) Here are some huge unknowns that are each highly relevant to the case of re-opening school buildings, all taken from the National Academies Report, which did recommend re-opening "when safe" but declined to say what situation would be safe enough:

1. We don't actually know the rate of deaths or severe illnesses among children infected with COVID-19, because most of them have been protected so far by stay-at-home measures. Although the death rate for children under age 10 was 0.7% in published research, in Colorado there have been no deaths in this group and the percentage in children ages 10-19 is only 0.1%. Yet the small number of children previously exposed may not be representative of the large number who would potentially be exposed in school buildings; recent reports from Israeli schools and Texas daycares show that outbreaks are possible whenever children are concentrated in closed settings. Additionally, Colorado's hospitalization rate is between 3-5% for children and adolescents, resulting in significant medical costs to families (e.g., $6,000 per hospital day) as well as fear and distress. And even among people with "mild" coronavirus cases, emerging evidence suggests that long-term neurological, pulmonary, or cardiovascular complications can occur. The exact risks to children's health are therefore uncertain.
2. We don't actually know the percentage of COVID-19 cases that can be tracked back to children. Again, most of the data showing that children "don't spread the virus" were collected after schools were already closed; Georgia data show that children were equally involved in transmission before that happened. There's some argument about immune functioning in kids making them less vulnerable to infection even when exposed to coronavirus, but this is still pretty speculative. Other results suggest that asymptomatic children's viral load is just as high as adults' which may make them equally infectious even if they don't show any symptoms. Data showing limited spread of COVID-19 in European schools may not be relevant because community rates were much lower in those countries at the time of re-opening, and even under those conditions schools have seen significant outbreaks. The norm around the world is still for schools around the world to remain closed, so we have very limited data about school-based transmission risks.
3. We don't really know how much transmission of COVID-19 is due to surface transmission (touching contaminated objects), aerosols (small particles that stay airborne, travel long distances, and are hard to block with masks), or droplets (large particles that travel only about 6 feet, quickly drop to the ground, and are easily blocked by a cloth mask). The best current guess is that it's some of each, and a brand-new simulation study of the Diamond Princess cruise ship cohort suggests that each method accounts for about a third of new cases. At least some transmission is clearly via aerosols, and the more it's aerosol-based, the less effective our standard precautions like masks and 6 feet of physical distance will be in preventing transmission. One thing that we do seem to know, incidentally, is that symptom screening misses up to 75% of active coronoavirus cases, so elaborate entry procedures for schools might make our Intuitive minds feel reassured but probably won't change the true risk equation very much.
4. We still don't know enough about the efficacy of various mitigation strategies, such as mask-wearing, physical distancing while indoors, or changing building airflow to use more filters, air movement, or outdoor air. Even the term "social distancing" or "physical distancing" can mean different things: The strongest benefits clearly occur when "social distancing" means "not interacting with people outside your own household." When it means "keeping 6 feet away from others" a number of other factors likely come into play, such as whether the interaction happens outdoors, whether you are both wearing masks, and even the previous question about whether long-range aerosols versus short-range droplets are the primary method of transmission. Because of the long duration of aerosols in indoor environments, and their ability to move through built environments even when (like the Diamond Princess) there is 100% outside air circulating through the HVAC system, aerosol transmission also decreases the effectiveness of cohort-based strategies for containing transmission to just one group of students in a shared building. Complicating all of this is the fact that viral load is associated with COVID-19 disease severity, which might suggest that combined infection via multiple methods or a longer stay in a virus-containing building adds up to illness when any one method would not have done so alone.

These are very large unknowns, and the health risks are significant. On the other side of the risk-benefit equation, there are also important open questions about the benefits of in-person school:

1. To what extent will in-person schooling during a pandemic improve children's mental health? A recent review article summarized risks associated with social isolation in adolescents, but the studies reviewed were all conducted outside the context of a pandemic. People’s reactions to isolation may be different when the experience is shared with others and when there is meaning or purpose to the restrictions. Data on adults’ mental health do show greater anxiety and depression in April 2020 than in April 2018, and recent data on children's mental health show increased mental health concerns as well as food insecurity, loss of health insurance, and loss of childcare. But these increases are by no means overwhelmingly large, and they might be attributable to the pandemic itself or its broad economic effects rather than to the closure of school buildings in particular. Finally, some experts have suggested that returning to school will itself be a traumatic experience because the environment will be so restrictive and so different from what children are used to. Overall, we can realistically say that there are unknown mental health risks from continued social distancing, and that re-opening schools will be of unknown benefit in addressing them.
2. What are the outcomes of online versus in-person learning? Many anecdotal stories have been shared about online learning challenges, and many people argue that large-scale online teaching this spring meant no learning occurred. But many studies show that online learning is just as effective as in-person learning. In some studies, online learning actually produces greater educational gains than traditional teaching, including for students in grades K-12. It has also been widely noted that the experience of most students and teachers in March 2020 was not a well-planned online curriculum, but rather an emergency shift to remote teaching, during which parents also were rapidly transitioning to distance work, joblessness, pandemic-related life stressors, shortages in stores, travel restrictions, and a broad range of emotional reactions in both their children and themselves. Spring 2020 was therefore not representative of online learning in general, and even at that we have no solid outcome data yet to know whether or not it worked. One valid concern is that not all children have equal access to resources such as technology and parental assistance, but re-opening school buildings for all students is clearly not the only way to address this social problem.
3. Are there good alternate ways to deliver social service programs, outside of schools? It’s true that schools are a gateway to social service programs including meals, shelter, vision and hearing screening, mental health or drug use screening, support for parents, and others. Many of the mental health impacts noted in the recent study of children were actually about lack of access to social programs delivered through schools, like food assistance. Yet there might be other ways to accommodate in-person learning for the most vulnerable students, perhaps in outdoor tents or in very small groups that would be feasible only if most children stay remote. This plan would require online learning as the default option for most children; perhaps parents could request in-person learning if they met pre-established criteria such as poverty or learning disabilities. Other social remedies with less health risk also could be pursued. We aren't strongly pursuing other solutions as a society, so their efficacy is unknown.

As demonstrated above, there are significant uncertainties in the available evidence both for and against re-opening buildings. Leading scientist Dr. Anthony Fauci at the National Institute for Allergies and Infectious Diseases therefore appropriately characterized the national push to re-open school buildings as "an experiment." The uncertainties above apply to children only, but at the societal level there are also potential collateral risks of in-person school for teachers or other adults who are present in school buildings.

How can public health professionals, school districts, or individual families make decisions under these conditions of uncertainty? Mishel's nursing theory of illness uncertainty may be instructive: It suggests that people cope with health decisions by either (a) instrumental coping strategies like learning more, taking concrete action to reduce risks, or identifying contingency plans; or else (b) emotion-focused coping strategies like seeking support in the form of agreement by other people, taking refuge in the opinions of authority figures, or trying not to think about the problem.

High-stakes health decisions generate anxiety, so one answer to the dilemma is to avoid the decision. we might ask what others have decided to do or simply accept the judgment of authorities based on the argument that "they wouldn't re-open the schools if it wasn't safe." Unfortunately, the conflicting public health guidance we are receiving from well-regarded authorities like the CDC, WHO, Surgeon General, and political figures makes it hard to know which authorities to trust. Just as there is no agreement yet among scientists, there is no agreement among higher-level decision-makers. We might then throw up our hands and just go with our gut. That approach leaves us vulnerable to any number of emotion-driven biases that affect Intuitive-level decision-making, including the tendency to over-value known present outcomes (I can't work at home with everyone else here!) and under-value unknown future risks (maybe returning to school will lead to extended illness, a hospital stay, or worse, and then really no work will get done). Alternately, we might be inclined to think "I'll just roll the dice" and throw our lot with random chance, but economic philosopher Nassim Nicholas Taleb argues that "in practice, randomness is [just] incomplete information." The only reason randomness seems OK is because we don't yet know the details. It's also true that humans are bad at estimating risks, tending to view small but severe consequences as impossible, and events with probabilities of 90% or more as certain. We're particularly poor at estimating cumulative risks that only add up gradually over time or based on the total number of people present in a school building.

More thoughtful decision-making, therefore, is likely to produce better outcomes under conditions of uncertainty. In Mischel's model, active coping is also more psychologically healthy, e.g. in coping with a chronic illness, and over the long term people who don't go through a thoughtful process are more vulnerable to later regret about their health decisions. Active engagement to reduce health risks also conveys a sense of control that can be helpful regardless of whether the ultimate risks are reduced. Recommendations therefore might be for decision-makers (parents, teachers, or public health officials) to learn as much as they can about COVID-19 risks as well as the risks of continued online learning, to advocate publicly for needed changes (e.g., by writing to school boards or elected officials), and to develop multiple levels of contingency plans depending on whether school buildings are open or closed, and on the level of community transmission of COVID-19. 

If we don't know the underlying risks with any accuracy, how can we know which options to advocate or plan for? Of course, we hope that the actual probabilities of each outcome will become clearer over time, which will move us closer to the situation of risk management, but we aren't there yet. In the meantime, Naturalistic decision making theory suggests that case-based thinking is probably more helpful than rational analysis. We don't have great mental models for a pandemic, but we do have other experiences that might help us to make decisions. For example, think about what it's like to have a sick child at home, versus a child at home doing remote learning. Or consider what an average school day might look like with pandemic precautions in place, and whether that new reality seems better or worse than home-based schoolwork. (Remember that the scenario we all actually want, a return to regular school life with no precautions, isn't currently on the table in most areas).

Decisions about whether or not to re-open school buildings are extremely difficult because of the level of uncertainty involved, and as many people have noted there are "no good choices." Yet policy-makers, communities, and individual families must soon decide what to do. An understanding of uncertainty's effects on our thinking, as well as careful consideration and visualization of alternatives based on our most relevant past experiences, can help us to make more careful and hopefully better decisions about this important public health question.