Saturday, July 25, 2020

Assessing State COVID-19 Performance

I perform my own analysis of data from multiple reputable sources to better inform myself about COVID-19.  I share some of that analysis here. I am not an epidemiologist or government official.  My analysis is my own personal pursuit of better understanding in an often unclear and conflicting reporting environment. Some of my blog readers may find it interesting and maybe even helpful.  It is a layman's work, so keep that in mind.  In this specific post I use the following sources:  The Covid Tracking Project : Rtlive ; and the Bureau of Labor Statistics.

First, I created the Venn diagram at right to provide a context for my writing.  As is the normal course in present day America there are those that try and divide by making every issue an either/or binary choice for political alignment.  I reject that.  The COVID-19 pandemic is complex and at least three areas - public health, rights and responsibilities, and economic impact - should be a part of any pandemic policy.  A skeptical public should take responsibility for demanding balance and evidence of efficacy in policy.

The lack of depth in media reporting is what instigates my own analysis.  There is an overwhelming inclination in government and the media to report raw numbers of COVID-19 cases and deaths. Raw numbers tell us little to nothing of value. What does a statistic mean if it is not compared or contrasted with something else to measure its relative impact, value, threat, or risk? It is in comparisons where we find value and knowledge. For example, one can compare states to measure relative performance and gain insights into policy effectiveness or compare a state's performance over time to reveal trends within a state.

The tables that follow rank states in order of performance largely by comparing them based on the percentage of their population that has died, is currently hospitalized, has been tested, etc.  Using the percentage of the population comparison normalizes the measurement, unlike a raw number.  The color coding breaks the states down into 20% increments displayed red (worse) to green (better).   

The State Name column assumes the same color code as the percentage of the population that has died with COVID-19 (% POP DEATHS) and retains that color code in all tables.


In my judgement the most important reflection of a state’s performance throughout the pandemic is the percentage of its population that has died with COVID-19.  This table is sorted by that statistic (% POP DEATHS), but also contains related variables.

There are statistical definition, collection and reporting issues that cause uncertainty in reporting COVID-19 deaths.   For example, early cases may not have been detected, and a death is reported for anyone who dies "with" not necessarily "from" COVID-19.  The % POSITIVE DEATHS column reflects the percentage of people tested positive for COVID-19 that died in the state.  This can reflect disproportionate institutional elderly deaths in a state or reflect on how well a state's health care system responded to the virus.  The % Tot US Deaths column reflects the percentage of all United States COVID-19 deaths that occurred in a state.  Note that the worst performing eleven states represent 23% of the U.S. population but 56% of the deaths from COVID-19.

The morbidity of COVID-19 still cannot be calculated.  Morbidity is calculated by dividing the number of deaths by the number of cases.  We still do not have an accurate number of cases. There are many more cases of COVID-19 that go undetected than are detected.  The head of the Centers for Disease Control said that there may be ten times as many cases as are detected and a recent report indicates it could be in a range of 6 to 24 times the reported cases depending on the state.   See previous blog posts I have written to understand better the uncertainty of collecting and reporting data about COVID-19.


To better understand how a state is performing currently one can look to testing, lagging 7 day averages of cases, percentage of the population that is hospitalized, and other variables.   In the table below several of these variables are indicated in columns.  It is sorted by the percentage of the state population that is currently hospitalized (% POP in Hospital).   This variable is used because others, such as cases (% Tested Positive), are influenced by many other variables, such as testing volume, and are not necessarily predictive of the intensity of either hospitalizations or deaths.  

Hospitalizations is a critical data element because one major goal in regulating the rate of virus spread is to prevent overwhelming hospitals.   There is no standardization among states or at the federal level of data reporting on hospitalizations.  Kansas still reports nothing.   The timeliness and extent of hospital data variables (e.g. percent bed capacity used, ICU, demographics of patients, etc.) reporting is different in every state.   It is essential that this change.  The threat of overwhelming a state or regions hospital capacity cannot be assessed without detailed data on a number of variables.

I will use Florida as an example as it has been the subject of much media attention because it has seen a dramatic increase in cases.

Florida increased its testing significantly in recent weeks to average nearly 100,000 tests daily.  The percentage of those tested that are positive is about 13-14% over the past two weeks.  That was a significant increase.  It is true, as some have stated, that if you increase testing you will increase the number of detected cases because we know we may be missing many infected people who have no symptoms and do not get tested.  However, that does not entirely explain the increased percentage of positive tests - there is likely increased spread. Florida's increasing number of cases lowered it from number 18 to number 47 in state ranking over the past few weeks.

Many states are increasing testing.   The purpose of the testing is to underpin Contact Tracing.   A person is tested and state authorities are supposed to use the report of that positive test to reach out to anyone who had contact with the infected person, test them and isolate them, in order to slow virus spread.  This only works if there is a quick turn around of tests.  If the test result is not reported within 48 hours it's value is greatly diminished in the contact tracing effort.  Labs are overwhelmed with the increased volume of testing in many states and there are reports of up to two weeks in the turn around of test results.

Contact tracing is critical to a strategy to throttle spread of the virus, but it's effectiveness is questionable and its reporting of detail to the public is rare.  This issue is deserving of serious governmental and journalistic inquiry.   When a governor says he/she is shutting down beaches the first journalist with the opportunity should ask, "Governor, can you provide us with the number and circumstance of infection and spread on a beach based on contact tracing evidence?" 

In another indicator of the increased spread in Florida there is an increase in hospitalizations.  Florida is in the unenviable position of last place in hospitalizations today as indicated in the table below.   The spread and impact is largely, once again, in the Miami-Dade, Broward, and Palm Beach county areas. The state is challenged to ensure that spread does not keep increasing and overwhelm the hospitals. Will lower average age of cases and better treatments result in a much shorter stay in the hospital with less serious interventions?   If the length of stay is shorter the number of new daily cases that can be tolerated is greater and the issue becomes one of churn.  The number of hospitalizations has been decreasing in recent days.

Despite increased cases and hospitalizations Florida is ranked in the middle of the pack among states for deaths.  It has gone in recent weeks from number 25 to number 29 in rank for the percentage of its population that has died with COVID-19.   Deaths have and will increase with increased spread, but how much?  Thus far, it is not the same as it was in the early days of COVID-19 because the population that is getting infected is younger, healthier and does not suffer such severe consequences as older patients, particularly those in institutional settings.  In April the median age of those infected in Florida was 65.  It dropped to about 35 in June and rose to and steadied at about 41 right now.


Any state that is pounding its chest crowing about its recent performance should be very cautious.   We still do not know a lot about the virus.   Social distancing and other actions to control spread can work, but to what extent is not known or measured. Many variables are at play, so there is a great deal of uncertainty about why or whether cases will rise or fall in any state.  

One data element that may help to predict spread intensity is the Effective Reproduction Number or Rt.   Rt is the average number of people who become infected by an infectious person. If Rt is above 1.0, the virus is spreading and the higher the Rt above 1 the faster it is spreading.  When Rt is below 1.0, the virus spread is slowing.   Twelve states had an average Rt below 1 on July 24, an increase from seven in the past week or so.  There is a trending of improvement for many states while others are declining.   This could make the case increases of the previous weeks in many states a blip rather than a trend.   It may also portend blips in states that currently are thinking they are doing well.  It is the nature of what we are doing - throttling, assessing, and reacting in order to maintain balance between COVID-19 suppression, economic expansion, and protection of constitutional rights.

Governor Charlie Baker of Massachusetts this week issued further restricting orders on the Massachusetts population and threatened more severe actions. He also placed restrictions on those traveling to Massachusetts from most other states.  Hospitalizations and deaths have been relatively low in Massachusetts in recent weeks.  Why is the Governor increasing restrictions?  One reason may be that the Massachusetts Rt has been over 1 for more than two weeks and may portend an increase in cases.  At the same time, Florida Governor Ron DeSantis may be optimistic as the Florida Rt has been below 1 for about two weeks and improving, and hospitalizations may have begun a downward trend.    


Locking down the nation for public health reasons has unintended consequences that include the decline of the economy, personal financial ruin for many, and psycho/social suffering.    The economy cannot be stifled forever or with each new virus.  Political construct narratives try to frame the debate in an either/or argument of LOCK DOWN or FULL OPEN.   What we must actually achieve is a balance of continuing economic expansion while preventing hospitals from being overwhelmed and moderating the number of deaths.

I use the table below to see how well a state is doing in moving its economy forward.   The ranking is based on the change in unemployment within the state from May to June, 2020.   

One can see the priorities of states in their performance.  For example, at state rank 51, Massachusetts had the worst unemployment rate in the U.S. in June at 17.4% and ranked 47th in percent change from May to June.  Florida ranked 32nd in state unemployment in June and 18th in change from May to June from 13.7% to 10%.   Florida is more aggressive in trying to open its economy.  Massachusetts is more aggressive in public health behavior restrictions.  Their economic numbers reflect their priorities.  Again, it is about balancing these concerns to find an acceptable outcome in all areas.


Federal and state governments have fairly broad authority constitutionally, in law, and in court precedent to take action in response to a public health emergency.  However, it is not unlimited.   There has been little challenge to the decrees of governors in the courts.   One could expect cases to begin to materialize from both civil liberty advocacy groups and individuals who question the capriciousness and lack of evidence-based-reasoning for some restrictions.

Americans do not give up their due process guarantees under the Constitution because there is a public health threat.  Restrictions have to be reasoned with evidence, targeted, non-discriminatory, and offer avenues of challenge to the individual.  A public health emergency is not a blank check for presidents and governors that ONLY considers the risk of infection and death from a virus.

The restrictions on movement between states may be ripe for challenge.  Several states have implemented restrictions on travel into their states from other states.  The latest restrictions placed by Governor Charlie Baker of Massachusetts is an example.   It is a broad stroke restriction against other states that may be illegal.  California could not ban people migrating from Oklahoma in 1941 under the Commerce Clause.   The Privileges and Immunities Clause has been used in cases to stop states from discriminating against the citizens of other states.

My point is that there is always the opportunity for abuse in emergencies.   If the restriction of rights is not tested, especially in the worst of times, those precedents will potentially stand in future emergencies as the foundation for further governmental power expansion to the detriment of individual liberties.  As Justice Jackson said about government emergency powers to detain Japanese Americans during WWII, “[it] lies about like a loaded weapon, ready for the hand of any authority that can bring forward a plausible claim of an urgent need.”    Fear cannot be allowed to turn heads from rigorous protection of constitutional rights.

No comments:

Post a Comment

Comments to blog postings are encouraged, but all comments will be reviewed by the moderator before posting to ensure that they are relevant and respectful. Hence, there will be a delay in the appearance of your comment. Thank you