People struggle to assess risk, especially in a pandemic

The writer directs the Leverhulme Centre for Demographic Science at Nuffield College, the University of Oxford

I remember being presented with an illustrated risk card when I was pregnant with my first child. Stick figures marched towards the higher odds of having a baby with genetic alterations as I aged. As someone who works with numbers all day, you would think I easily grasped the true risk. You can understand the odds, but not comprehend the reality. Many lives have been enriched by children stacked on the side I was meant to be concerned about. 

For decades, scientists questioned whether people could assess personal risk, particularly under conditions of uncertainty. The pandemic has generated daily doses of numbers, league tables and quandaries ranging from vaccine passports to varying messages about the Oxford/AstraZeneca vaccine. When given odds and risk-benefits, what is our benchmark? What is an acceptable risk? 

This week the UK and European regulators suggested that the AstraZeneca vaccine’s association with rare blood clots may have a causal relationship, but the risk is extremely low. Until then, we had faced the correlation-causality conundrum. Correlation is an association with an event (vaccine) and outcome (blood clot), while causation is the established pathway or mechanism between the two. Correlations range from 0 (no relationship) to 1 (strong association) but they can be spurious; a relationship might appear causal but is not.

To see just how far off apparently strong associations can be, visit a blog by Tyler Vigen called Spurious Correlations, which shows a 0.95 correlation of per capita cheese consumption in the US with the number of people who died by becoming tangled in their bedsheets. The divorce rate in Maine has a near perfect correlation with the per capita consumption of margarine. Although it is tempting to make causal theories here, there is no pathway. 

Psychologist Paul Slovic’s classic work on perceptions of risk showed that people need an anchor or numerical predictor to gauge an acceptable risk-benefit trade-off. We know that the risk of blood clots after the AstraZeneca jab is extremely low at four in a million, with chances of death at one in a million. This is one in 250,000 or 0.0004 per cent; you have a 99.9996 per cent chance of not developing a blood clot. Women on the birth control pill have a risk of blood clots averaging 300 to 900 per one million. Although we cannot forget those who die or suffer, it is a timely reminder that every drug, vaccine and intervention has risks — but the benefits almost always triumph.

For a mass vaccine rollout, the psychology of risk perceptions needs to be considered. In the 1970s, Amos Tversky and Daniel Kahneman argued that when people need to make judgments under conditions of uncertainty, they rely on mental strategies, or heuristics, derived from previous experiences that are often hindered by biases from family, friends, cultural context or the media.

Kahneman later wrote Thinking, Fast and Slow, differentiating how we think fast, intuitively and emotionally, but also slowly, deliberatively and logically. Experts focus on slow technical estimates while people think fast, often perceiving risks as higher than they are. We are more willing to accept risks from voluntary activities such as skiing than those outside our control like food preservatives or vaccines. 

Beyond vaccine safety, the pandemic has presented persistent uncertainty while creating unprecedented dynamic, complex problems, such as whether to introduce vaccine passports. The ceaseless debate adds to a feeling of chaos and lack of co-ordination, which undermines trust in authorities. Individual US states are building their own DIY passport solutions. Just as New York introduced the Excelsior App, Florida and other states passed legislation to ban them. With President Joe Biden’s administration reticent to adopt a federal standard, the lack of co-ordination may come back to haunt the US as people cross borders.

Could we have done it differently? In 1976, a swine flu scare resulted in the fast-tracked production of a vaccine and inoculation of 45m Americans. David Sencer, then director of the Centers for Disease Control and Prevention, and J Donald Millar, a public health administrator, concluded that when large numbers of people are exposed to a vaccine, unrelated illnesses or deaths will be attributed to vaccines (in their case Legionnaires’ disease). Excess cases of syndromes may appear when there is uncertainty about a causal relationship.

Sencer and Millar wrote that the media and public awareness “can be a major obstacle to implementing a large, innovative program responding to risks that are difficult, if not impossible, to quantitate”. It does not look like we have become much better at managing risks under conditions of uncertainty.