How to Manage Covid19 Recovery

In the near future (a few days or weeks), local, regional, and national policymakers will grapple with how to configure the reopening of economic and social life, post the covid19 lockdown. There are two reasons to be concerned about how the reopening unfolds in the US: 1) the problems in public policy process and decision making that were exposed during the arrival of covid19 and the subsequent shutdown phase, and 2) the “one step left, two steps right” confusion exhibited by other countries and regions that have attempted reopening. To do this well requires a sensible methodology that can yield specific insights for controlled and systematic reopening.

Lockdown Measures Were Necessary

There is no doubt about the lethality and scale of covid19, and about the seriousness of the health crisis it caused. Likewise, social distancing and other cautionary measures were necessary (indeed, implemented too late, in my view). The lockdown was quite indiscriminate ("shut down everything ..."), but that was fine because or urgency and asymmetry of risk. The measures seem to have been successful at controlling the spread (picture below, and look here for a detailed analysis). This reduction in daily growth rate from over 30% to 10% expanded the doubling period from 3 days to 7 days within two weeks (and now about 15 days) and gave our health system time to catch up.

No alt text provided for this image

Harmful Effects Create Pressure to Reopen

But these measures also have adverse side effects. By crippling economic activity, they cause job losses, disruptions in food supply, individual economic ruin, hoarding and shortage of medical equipment, etc. There will be long-term harm due to loss of education and skills, increased poverty, starvation, homelessness, unemployment, and suicides, all of which will mean even more deaths. For instance, there has already been an incredible surge in demand at food banks, a sharp fall in timely rental payments, and of course a huge widespread collapse in economic activity.

No alt text provided for this image

These harmful consequences will create pressure for loosening the restrictions and reopening economic and social life. This is already happening. Warnings about the shutdown appeared in influential outlets as early as a week into the shutdown! People and businesses are getting impatient and concerned about economic survival. Some state governors and regional leaders are eager to reduce lockdown measures and return to economic activity. This trend will intensify as the numbers of covid cases (and hospitalizations and deaths) slow down, and as other harms increase.

The overarching question then is, when and how best to reopen?

Time to be Smart and Methodical about Reopening

The lockdown was justifiably indiscriminate, but reopening should not be. There is some risk that impatience will lead to a “big bang” reopening — closed till one day, all open the next. That could have awful consequences. Here are a few bad examples to learn from and avoid: 1) Georgia opened up hair salons - seriously!? people in really close proximity for half hour, one of whom is a super spreader? 2) some counties opened beaches, but "only a few" - why? to cause overcrowding in those that do open?

Saner minds have suggested three tactics for handling the reopening, namely

  • Methodical and phased selection of sectors to reopen (e.g., optometry offices before movie theaters or golf before basketball or recreational tennis before stadium sports).
  • Contact tracing and intensive testing.
  • Antibody testing to detect past infections and potential immunity.

Even these elements need to be improved. For instance, contact tracing has been absent in the US, and testing is still lagging. In terms of selection, the much-used criteria of “essential vs non-essential” ignores a vital element of risk. For instance, hair salons and recreational tennis are both not essential, but vastly different in the quantum of infection risk, therefore it makes no sense to club them together. Instead, build a matrix to incorporate value or benefits, infection risk, and the ability to conduct activity in a non-contact mode.

No alt text provided for this image

Then reopen most left-column activities (low risk) but only right-column activities that lack a remote mode and are essential (like the ones in bold face). The analysis should be done at a fairly granular level. Activities in the bottom row should remain shut down, in order to reduce avoidable spread of infection. One could add other dimensions to the matrix including population density and susceptibility density, but my goal here is not to build the best decision matrix but talk about how it is executed.

Reopening: Simultaneous, Rapid, Multiple Science Experiments

The above tactics are necessary but not sufficient. If decided at a high enough level (e.g., by State Governors) and implemented at large scale (e.g., all gyms in a city or all optometry offices in a state), they may still lead to a “reopening done wrong” scenario where there’s a surge in new cases in a few days, and with much confusion about its cause, causing a backpedaling of the reopening. This kind of broad reopening, confusion, and rollback has occurred in other countries. So, what’s the alternative?

The core idea in this note is that reopening should be viewed as a collection of simultaneous, rapid, data-driven science experiments, done in a careful and controlled manner, with treatment and control groups and intensive data analysis. Whatever tactics are employed in identifying what sectors to reopen, they should be embedded into a thorough, precise, detailed and very large-scale scientific experiment. Start at a small scale (e.g., a few selected optometry offices and a few car repair shops in a few selected cities, rather than all), and then monitor-analyze-adapt based on intensive data collection and comparison against carefully chosen control groups.

Data collection is paramount, and monitor-analyze-adapt

This would mean, for instance, that if it is decided in a certain county that banks should be opened first, then i) only a few banks are selected to open, ii) a control group of banks is developed, iii) intensive data collection (covid19 cases) is done in both regions before and after the reopening (for a period of say 7 days). Simultaneously, a similar experimental process occurs for other units chosen for reopening and in other counties, say agricultural operations in one, tool shops in another, and so on.

How to do this? Well, we already have an experimental recipe for medical product development, a sequence of animal trials and phase 1–2–3 trials. That is a useful template for reopening, but it has to be speeded up. We can also learn from Silicon Valley tech companies where simultaneous multiple experiments are part of routine business. Companies like Facebook systematically create overlapping partitions of both users and new product features, efficiently executing hundreds of experiments each day and successively improving products by building on experimental results.

Politics and Public Communication

The reopening strategy needs to account for politics and optics, making it critical to have transparency and public confidence. For example, publicize all the decision rules, especially those related to revocation of a loosening/opening action rerun, emphasize that opening decisions are provisional — understood to be reversible if data so indicates — and announce results on a weekly basis, and incorporate local autonomy by asking local jurisdictions to provide input into the cost and benefit “weights” that go into the decision rules.

Start Slow and Start Early

One of the advantages of this “start slow, monitor and analyze, then expand” approach is that it can be started early. A “big bang” reopening would be safe only if there is very high confidence that exponential viral spread will not recur. But the economic harm that is occurring every day makes it harder to wait that long. In contrast, the experimental approach could start much sooner because there will be greater clarity regarding its effects on spread and because any negative consequences (if they occur) will be small and easier to roll back.

Indeed, a good part of our economy already remains open today — grocery stores, delivery workers, health care staff, and many others. By corollary, this approach demands that the same scientific experimentation approach should now be applied to these crucial groups so that we develop an accurate sense of the risk of non-isolation.

Additional work is needed to identify the exact design of this experimentation approach, and its implications regarding what units should be reopened, in what sequence, and under what conditions. Doing this needs to be an urgent and top priority.

Noah Gift

Founder Pragmatic AI Labs, Pioneering MLOps Leader & Author, Veteran Startup CTO, Duke Data Science & AI EIR, Ex-Caltech/Disney/Sony

3y

One wildcard is an income inequality time bomb in California. Hearing from many people who were already paycheck to paycheck and now they are in big trouble. California, in particular, could have the most unknown variables: high taxes, high pension burden, the leading nation in homelessness, leading in income inequality, and leading in housing affordability problems. The secondary and tertiary effects could be a big deal. A positive hope is the even the most business-friendly people realize that if you don't take care of the most vulnerable it affects the entire network and leads to populism and lack of faith in our institutions and science. States with a "high burn rate" for personal expenses could be the most volatile when the true economic conditions play out over the coming year.

Sanjay Saigal

Energized by learning (in humans) and value creation (in markets). Teaching: Stanford, UC Davis, Minerva, Ashoka, St. Stephen's Instt. • Consulting: Ilog, Decision Focus, Intechné • Startups: Compass Modeling, Skuchain

3y

A key - possibly even fatal - reality working against your modest proposal is that science and (the peculiar American variant of) democracy aren't necessarily compatible. The dissemination of experimental outcomes can be made more transparent. But in our polarized environment, science-based policy-making reliably fails. Even the most transparent and sensible decision is laid waste by conspiracy theorists, power-motivated partisans, and garden-variety trolls. Look at the recent anti-public health protests in Texas, Michigan and elsewhere. Science-phobes only need a homeopathic "vital energy" of a suspected motive to oppose public interest policies. In our oppositional environment, it's difficult to expect even the most transparent community-minded experimental approach lasting beyond the drawing board. That said, I don't want to come off as the pessimist "seeing difficulty in every opportunity." If there ever was a time to try to overcome societal FUD, this is it. Governor Newsom has certainly been out front on things, so there's hope someone in authority will take up your challenge.

Jake Hosier

Director of Access Programs at UC Davis Continuing and Professional Education

3y

Hemant, this is excellent! I hope you are consulted on the big question of how to re-open the University in the Fall.

Karen Mesrobian

Sr. Manager, Operations at Sigma Computing - We're hiring!

3y

Great analysis Hemant! One attribute I would add to be considered is population density. I.e. if some banks, but not all banks, re-open there should be a statistical risk-assessment based on likelihood of spread, and rate of infection, that determines which banks get the green-light. Areas, such as at the county level, with less dense populations should theoretically have a higher risk tolerance for spread than a highly populated metropolitan area.

To view or add a comment, sign in

Insights from the community

Explore topics