Why state reopenings have sent projected coronavirus death rates back up

Several top Trump administration officials are continuing to predict Depression-era unemployment numbers for the month of May.

On Friday, the Bureau of Labor Statistics reported that the US economy lost a record 20.5 million jobs in April, with the unemployment rate reaching 14.7%. But that’s based on claims filed by mid-month.

Treasury Secretary Steven Mnuchin gave an even steeper figure on Sunday in an interview with Fox News when asked where unemployment stands. Mnuchin, asked by Fox’s Chris Wallace whether “we (are) talking close to 25% at this point, which is Great Depression neighborhood,” responded, “We could be,” adding that the administration expects the numbers to improve in the coming months.

Larry Kudlow, the chairman of the President’s National Economic Council, said Sunday that he expects May to see “very difficult numbers,” but also sought to call attention to what he called “a glimmer of hope” in the data for Americans to return to work.

Yet Kevin Hassett, a top White House economic adviser and former CNN contributor, continued to paint a dark picture for future unemployment, telling CNN’s Jake Tapper on “State of the Union” that the rate will “probably” be “close to 20%” in the May jobs report, which will be released in June.

How reopening affects projected death rates

We’re now well into May and public health officials say there still isn’t enough testing or contact tracing to reopen the country safely. But as unemployment rates soar, governors are under increasing pressure to reopen more businesses in a gamble that no one will know the health effects of for weeks.

The early indications aren’t good.

See where your state stands on reopening here.

An influential coronavirus model from the Institute for Health Metrics and Evaluation at the University of Washington recently doubled its projections for US deaths. It’s now forecasting that more than 134,000 people will die from Covid-19 by August.

Here’s how the model shifted in April, when the rapid spread of state-by-state shutdowns and widespread compliance with social distancing sent death estimates downward. CNN spoke Sunday to Dr. Theo Vos, an IHME professor of health metrics sciences who is working on the model, about what’s changed since then.

The conversation, conducted over email and lightly edited for flow, is below:

CNN: Your model has shifted upwards to now forecast more than 134,000 US deaths from the coronavirus. I’m curious what the main driving forces behind that shift were?

TV: With more data showing a longer lingering peak and a slower downward trajectory of deaths in quite a number of locations compared to what has happened in earlier epidemics (Wuhan, Italy, Spain) our model follows these new data points and extends the first wave curve of death in time and that makes the cumulative death count by early August go up.

By creating a hybrid model between the statistical death model and a transmission model, we can better pick up changes in predictors of transmission. Change in mobility and testing per capita are the most important drivers in the transmission model. There also is a modest effect of temperature (warmer temp, a bit less transmission) and an effect of population density (which obviously does not change over time). Thus the uptick in mobility we see in some states and the predicted change in mobility from lifting of some of the social distancing measures tells the transmission model there will be an uptick in new infections and in time those will lead to more symptomatic cases, hospitalized patients and eventually an increase in deaths.

CNN: The model added a tool to visualize mobility. Can you explain what exactly that refers to and what role it plays in your forecast as more states relax social distancing measures?

TV: It is based on combining the information from four sources into a single indicator (Google, Facebook, Safegraph and Descartes). This gives us day-to-day measures of mobility and change in this variable affects the probability of transmission as explained above. For the effect of relaxation of some social distancing measures we look at the effects their implementation had in an opposite direction and data on change in mobility in places that have started relaxing measures to predict the effect into future weeks.

CNN: The IHME model had previously projected that US coronavirus deaths would come to a halt this summer. Now with more available data, how has that timeline changed?

TV: The previous model assumed all social distancing measures would remain in place. At the time our focus was on informing policy makers about the surge in cases in hospitals/ICUs and the timing of the peak of this first wave. With the policy focus shifting towards relaxation of some of the social distancing measures, we incorporated the transmission model into the death model, so that we can better reflect the effect of such changes on mobility and consequently on the risk of transmission in the community.

CNN: As we head into the summer months, what could cause the model to shift upward again?

TV: As mobility and testing are the main predictors, it will be a balancing act between increased mobility and increasing testing intensity that will determine what will happen next. It will be important to get the balance right between these to avoid another upsurge in cases and deaths if transmission increases above the level where a new case of infection transmits to more than one other individual.

Of course, the testing per capita variable we use in our model as an important indicator (for which we have data) it is important that there are aspects of testing that determine how effective the strategy is beyond just the total numbers of tests: directing testing at everyone with mild symptoms or at higher risk of becoming infected; quick results and immediate contact tracing + testing of contacts and isolation of infected contacts (including those who are asymptomatic).

CNN: Is there anything specific you want people to know about your model moving forward?

TV: We will continue to regularly update the model, extend to other countries and incorporate new information that provides feedback on how good our predictions were.