In this post, we will examine how social distancing measures impact COVID-19 case growth rates. This analysis shows that school closures may not be receiving as much credit as they should for slowing the pandemic. This post focuses on the aggregate US. In later posts, we will address the issues involved in lifting the statewide stay-at-home orders.
When asked why there was not a national stay-at-home order, President Trump replied, "We have a thing called the Constitution, which I cherish, number one. Number two, those governors, I know every one of them, they're doing a great job. They're being very, very successful with what they're doing," Trump replied. "And, as you know, I want governors to be running things. In some cases, we'll supersede."
In the absence of a cohesive national plan, the governors of the individual states have made their own determinations as to the best way to proceed in the face of the COVID-19 pandemic and have issued orders to reduce person-to-person contact and to keep people safe. Neither the content nor the names of the orders are uniform from state to state. Some orders are called Stay-at-Home while others are called Shelter-in-Place; Healthy at Home; Home or Work; Stay Home, Stay Safe; or Safer at Home. Although the orders vary from state to state, they include some common elements:
An appeal to all residents to stay home except for essential trips for supplies or outdoor exercise
The closure of educational facilities
The closure of all non-essential businesses in the state and the call for workers in critical industries to continue working
The implementation of interstate travel restrictions
Regardless of how the order is named, the intent is still the same. It’s to limit the spread of COVID-19.
So, have social distancing measures impacted COVID-19 case growth rates?
In this post, we attempt to answer this question for the United States as a whole (we will address it state-by-state in a later post). Since there wasn't a single national order, we looked at when the various states instituted social distancing and converted the orders into a timeline of the percent of population affected. For simplicity, we only present the stay-at home and educational facility closure orders. Specifically, we have not attempted to sort out the travel quarantine restrictions, and we found that the non-essential business orders generally mirrored the stay-at-home orders and were redundant for this analysis.
To assess the impact of social distancing, we looked at whether the exponential rate of growth of cases decreased following the orders. We expected a lag between when the orders were put in place and when the impact was felt due to the incubation period of the virus, (which is about 5 days on average).
The graph below shows the growth rate in cases (left axis), expressed as the ratio of the given day's total number of cases divided by the number of cases 5 days earlier. The graph also shows the daily percentage of the US population under stay-at-home and educational facility closure orders (right axis).
The above graph suggests the following conclusions:
While broader social distancing measures took longer to roll out across the country (with some states still holding out as of the date of this post - 4/17), suspension of in-person schooling took place early in the spread of the virus and was much more pervasive. Between March 14 and March 19, orders to close educational facilities were applied to more than 90% of the US population. The growth rate peaked above 5x between March 20 and March 22, after which the growth rate began a steady decline, which is continuing.
Because of the partisan politics surrounding stay-at-home decisions, the closures of educational facilities have not been given proper recognition for having a positive role in slowing the epidemic. Beyond suppressing the spread of the virus through classrooms, it seems likely that the suspension of schools also brought awareness to the broader public, triggering social distancing behaviors prior to full stay-at-home orders.
While further imposition of formal stay-at-home orders has no doubt contributed to the further slowing of the epidemic, the long rollout of these orders makes it difficult to pinpoint an effect in the aggregate US graph presented here. Instead we will need to separate the data by state to get a clearer picture, which we will do in a later post. (The Atlantic also has an intriguing alternative explanation for the current decline.)
The variability of the growth rate early in the epidemic, including a decline between March 10 and March 16, is likely due to a lull while the virus was being contained in Washington state and before it began to take off in the rest of the country. Again, breaking out the analysis by state will clarify what was happening during this period. Also, this was a period when there were relatively few cases, as can be see in the graph below, which shows both the case growth rate and total cases. This graph also illustrates why looking at case growth rates is a better way to visualize "bending the curve".
In the absence of a national response, independent researchers and statisticians have stepped into the void in order to keep the public informed. The importance of these independent contributions cannot be understated. The information is effective in changing societal behavior, guiding governor plans, acting as a planning tool for hospitals, and influencing administration policy.
In the immortal words of Sting, “Don’t stand so close to me.” As there is no vaccine to combat COVID-19, social distancing is the best and only measure available to slow the progress of this pandemic.
Bình luận