There has been talk about the flattening of the curve of new reported COVID-19 cases indicating at the very least a slowing growth of the caseload in New York City.   I found the following data in an interesting “working paper” by Jeffrey E. Harris from the National Bureau of Economic Research: The Coronavirus Epidemic Curve is AlreadyFlattening in New York City.  It was posted April 7th.  Others have also opined on the same topic: Joseph Goldstein in the NY Times; Bob Herman on Axios.

Here is a brief summary of the Harris paper.

Note the inflection of the curve of new reported cases/day around March 20, 2020, in NYC but not in Los Angeles. Also note the different case numbers: Y-axis 1-1000 for LA, 1-100,000 for NYC.   Most certainly the situation in LA reflects an earlier time point in the progression of the local Coronavirus outbreak.  NYC may be an indicator of what is yet to come in LA.

The most valuable part of the Harris paper (I think) is the discussion of the possible confounding variables.

(1) Constraints on Testing: on March 20 New York City issued a directive “Healthcare resources must be saved to treat the sickest patients who require inpatient and critical care.” The Department directed providers and hospitals to “immediately stop testing non-hospitalized patients for COVID-19 unless test results will impact the clinical management of the patient. In addition, do not test asymptomatic people, including HCWs [healthcare workers] or first responders.”

In deed there was an immediate dip in the number of tests by about 50% and a simultaneous dip in the number of reported cases by about 30%.  But the dip in hospitalizations was much less prominent (about 10%) and they provide further data that show that the tests performed on hospitalized patients remained constant at about 17-23% of all tests performed:

Note: the Y axis is a logarithmic scale.

(2) Asymptomatic cases not considered: as many as 50% of COVID-19 cases may be asymptomatic and untested, yet infectious and contributing to the rise in the counts of newly diagnosed each day.  But Harris argues: “…there is no clear reason to believe
that the extent of understatement has changed significantly since the one-month takeoff period covered in Figure 1. So long as the ratio of undetected asymptomatic cases to detected symptomatic cases has remained constant”

(3) We are missing the false negatives:  tests based on nasal swabs (RT-PCR tests) have a false negative rate in the range of 26.7–46.4% (Yang et al. 2020). That means routine nasal swab testing could be missing a lot of coronavirus infections. But has the percentage of false negatives changed significantly during the month of March 2020 to explain the flattening of the curve in NY City?

(4) What is the breakdown of the data by borough:  There are reports of hospitals in Queens and the Bronx being overwhelmed and a heat map from the Dep. of Public Health, shows high levels of pos. tests in the Bronx, Queens and Brooklyn.  Harris presents data on new cases since March 21 broken down by borough. All the doubling times remain longer than the initial 1.3 days doubling time in early March, regardless of the borough.  But the data for Manhattan are particularly good suggesting a flattened peak and leveling off.

(5) Diaspora from the city: assuming that those that fled the city (for their Hamptons 2nd homes, for example) have equal rates of infection than those that did not or could not, Harris has calculated that 90% of the New York City’s inhabitants would have needed to flee the city in order to cause the flattening of the curve in the top figure.

(6) Is the initial rapid progression realistic?  The Ro number (‘basic reproductive number’, or number of people infected by one positive person) differs slightly depending on the city:

New York City: Ro=3.4     population density 27,016/square mile

Los Angeles: Ro=1.8        population density 7,544/square mile

Wuhan: Ro =2.2 -2.4       population density 3,200/square mile

This is most consistent with the high population density in NYC compared with LA, or even Wuhan.  Population density in NYC has previously been discussed in the context of the pandemic.  This is the most plausible cause of the rapid progression observed in New York City.

(7) Super-Spreaders:  this theory suggests that the apparent flattening of the incidence curve is a result of extreme heterogeneity in the infectivity of the New York City
population, with a small proportion of the total population – on the order of 5,000 individuals – subject to the super-spread of the virus.  While there are anecdotal reports of such super-spread in the New York City area (Williamson and Hussey 2020), Harris has not found clear evidence of a major source comparable to the 77 COVID-19 cases reportedly emanating from a late February Biogen meeting.

(8) Supporting Ancillary Data?  Harris correctly discusses daily death rates as a poor and lagging indicator.  By one estimate, it takes an average of 16 days from the onset of symptoms until a patient dies of complications and the time from initial
infection to death may average about 3 weeks.  Harris also looks at thermometer data from Kinsa!  I have previously posted on this blog about this exciting new way of assessing pandemic progression.

The downward slope of the red line is interpreted as a resolution of the Coronavirus outbreak.  BUT it is very indirect data and there is no data presented for Los Angeles !!!  Why not? If Harris is correct there should be no downward slope for Los Angeles.

In asking what makes the curve flatten, Harris discusses the obvious: social distancing, aggressive case tracking and effective public policy such as De Blasio’s order limiting  gatherings, closing gyms and other places of congregation and reinforced by subsequent orders that all New Yorkers (except workers in essential businesses) must stay at home.

The discussion of voluntary behaviors is very relevant.  Ridership on New
York City subways was already down 19 percent by March 12 and 60 percent by March 16 (Metropolitan Transportation Authority 2020).

The critical ingredient in the public policy mix may have been the successful communication of consistent, clear, accurate and  timely information to millions of individuals, who responded by taking action without government coercion. Put bluntly, what flattened the curve was no more than the naked truth.

Finally, in NY City about 1/226 residents have so far contracted COVID-19.  In Los Angeles, by contrast, it is about 1/4,100. The corresponding probability of knowing at least one infected person in a comparable size social circle is much less in L.A. than in NYC.  People are therefore more motivated to take action to reduce risk in New York.

Harris, Jeffrey E., The Coronavirus Epidemic Curve Is Already Flattening in New York City (April 1, 2020). Forthcoming, National Bureau of Economic Research, Working Paper Series Electronic . Available at SSRN: https://ssrn.com/abstract=3563985 or http://dx.doi.org/10.2139/ssrn.3563985