How China Should Use Forecasting To Guide Strategic Loosening of Anti-COVID Measures

Recent protests in China against draconian zero-COVID lockdown measures have “roiled investor sentiment.” Some initially might have an ounce of schadenfreude to see China suffer through repeated lockdowns, but in global health, all the delight of others’ downfall is a kind of masochism. It’s a bit like cutting off your nose to spite your face. When China suffers, the whole world suffers.

Leave it to a globalized economy to inspire empathy for the Chinese people, suffering under lockdowns and the lack of basic human rights. For all the reasons for the world to care about global health, it seems the economic reasons are still the most eye-catching. Forget the intrinsic value of health, health as wealth, health as a human right, or health as creating the capacity for freedom and well-being. It’s only when investments are at risk that suddenly we care about other countries’ COVID policies—and their flavor of authoritarian repression.

China’s recent announcement to shift from its zero-COVID path is not surprising. As we all know by now, there are only two pathways towards higher immunity: vaccinating with efficacious vaccines, or community spread (which has variations as well). There is no third option here. That’s it. But by loosening a lockdown strategy, China still has a few options left on the table. China is not as cornered as many are making it out to be.

Strangely, the country in its zero-COVID situation has not much improved since 2020, which was our pre-vaccine year. China’s immunity has not greatly increased, with its vaccines of low and waning efficacy and low vaccination coverage, particularly of the elderly, who are most susceptible to the severe impacts of COVID-19, namely, hospitalization and death. Its lockdown strategy has also meant that most of the population is susceptible to contracting the virus.

In terms of vaccination, it is obvious that China should expand vaccination with more efficacious vaccines. But it seems unlikely it will do so, for reticence against taking US or foreign vaccines. Perhaps there is an ally to China who can coax the country to appreciate the importance for its self-interest. But even if China does import better vaccines (or approves and uses its own homegrown mRNA vaccine), it will take time for the effect of vaccination to be realized. In the best case, we’re talking at least a few months before immunity improves.

So what are China’s options for relaxing measures? In the spectrum of measures for limiting the spread of COVID-19, zero-COVID lockdowns is on one end (epitomized by the Chinese)—and on the other end is a laissez-faire approach (epitomized by the US), which says that the whole thing is over and we don’t have to worry about a thing. It’s safe to say it’s neither extreme. Adding more mitigation measures does not make one more authoritarian, and letting go of measures does not make one more democratic. We have to move away from such a simplistic, black-and-white way of thinking.

While waiting for vaccination coverage and other countermeasures, China’s policymakers could either double-down on its lockdown strategy (like a person negotiating a bad position, digging into their heels, resulting in more protests), or China could choose to strategically relax its mitigation measures, by taking a watch-and-see methodology to see how quickly COVID-19 spreads, and whether the hospital system can bear the exponential growth or not. It might even take a play from some jurisdictions (e.g., Honolulu and many others) which made a tiered or graded system in which the stringency of measures depended on the hospital capacity indicators and the immediate forecast in the next two weeks. The latter is clearly preferable and wise.

China has wiggle room in rolling back its COVID restrictions; and it may not be as dire as many scientists have made it out to be. Some scientists have been sounding an alarm that rolling back will lead to many more deaths, as cited in a Nature paper published in May 2022, which argued that China is not ready to abandon its lockdown strategy. But a key message of this paper—one that is important to note—is that, even if the full lockdown is rolled back to maintain some non-pharmaceutical interventions, the healthcare system may not be overwhelmed. Maintaining masking, distancing, handwashing, oral antivirals, and a variety of other measures are still part of the COVID control toolkit.

Of course, the models and forecasts disagree. That might be expected since we have a limited history of what spread might be like in the absence of a lockdown, and a lot of information in China is not so easy to access. There are a lot of known unknowns and a lot of unknown unknowns when it comes to modeling. A different model from the Economist says removing lockdowns will overwhelm the system. But as usual with multiple models, it’s not so easy for policymakers to understand all the assumptions and the exact scenarios, as well as how valid and predictive any of these models are. Policymakers are not modeling experts, and modeling experts are not policymakers. And herein lies the rub.

How exactly should China’s policymakers wade through competing models and guidance? We wrote a paper about how they should use models using the case of the state of Hawaii (see my brownbag seminar at Harvard here on this paper). Based on what we did in Hawaii, I’d recommend that China’s National Health Commission take a thoughtful approach to forecasting (outlined below)—which will help to counter the shrill, slightly frivolous use of modeling bandied by the media.

1. Focus on forecasting rather than scenario planning

Collaborate with modeling experts in the country to produce an unbiased COVID forecast, ideally with a scientific technical advisory group or taskforce that can be tapped for policymaker queries.

Focus on forecasting (what will happen to the cases if nothing changes, also called the base scenario or business-as-usual scenario) rather than on scenarioing (what will happen to the cases if a policy is introduced compared to the base scenario), with the guiding principle being to build trust in the forecast as an unbiased source of information:

• Emphasize the business-as-usual case, that is, with no changes in policy.
• Produce a forecast no more than two weeks ahead in order to emphasize the potential changes that may occur if a new policy is implemented today.
• Emphasize that the forecast is dynamic, that any change in policy action or individual behavior can greatly change the forecast result after two weeks, and that the forecast will likely be wrong as soon as it is published because behaviors and actions including by the public will change upon publication of the forecast.

2. Communicate the forecast widely and broadly

Communicate the forecast results widely and broadly, and in an unbiased way and without consideration of new scenarios or policies, keeping in mind the building of trust as the overarching principle in all communications about the forecast. By communicating the forecast broadly, preferably on an ongoing basis, the forecast can be used by both the public and policymakers to closely track hospital capacity. Public perception of the COVID forecast can also naturally encourage the public to voluntarily take measures and precautions on its own, which in turn dampens the spread.

Whether policymakers reintroduce measures, or relax measures, will depend not only on hospital capacity but several other considerations and data sources including economic impacts. Policymaker communication on reintroducing policies should be kept separate from communication by the modeling and forecasting and scientific advisory group in order to build trust in the forecast itself.

Develop and train scientific experts to communicate simply to the public and emphasize the importance of an unbiased scientific approach rather than a partisan or politically motivated approach. Emphasize that experts should not have opinions about what policies should or should not be taken. Experts need to be able to communicate the potential impacts of a scenario without being driven by their personal preferences about what interventions should be done.

If the experts behind the forecast are asked about their preferences for what measures should be taken, the experts should default to the following answer: it is the responsibility of policymakers to make decisions about policies drawing on wide sources of information, which can include the forecast but also other sources of information. The job of the forecast is to provide unbiased guidance about what the COVID situation will be in the absence of any changes and maintaining business as usual.

Experts should be humble in recognizing that the model has many limitations and weaknesses, rather than assume their model is the best or most definitive model, and that the model is a static picture into what is by nature dynamic and constantly changing situation.

3. Use scenarioing very carefully

When applying scenarioing—that is, the use of modeling to consider alternative scenarios—be aware of its pitfalls. Scenarioing can be perceived as politically motivated depending on the choice of scenarios, resulting in negatively impacting the trust in the overall COVID forecast to begin with.

Further, the communication of assumptions between the base scenario and the alternative scenario is quite challenging for a general audience unless there is a strong public communicator who is able to understand both the technical dimensions and what the general audience is concerned about.

In Hawaii, we publicly communicated scenarioing results very sparsely, and generally only in request to specific policymaker requests, generally in private rather than public. We believe this helped to aid the perception of neutrality and build trust in the overall COVID forecast.

How the Chinese use forecasts and models to inform their decisions will be something the whole world should be watching. China's repressive authoritarian tactics aside, a thoughtful approach to COVID forecasting by the Chinese government will be welcomed by all.

With thanks to Javier Guzman and Justin Sandefur for helpful comments and the Hawaii Pandemic Applied Modeling Workgroup.