Month: April 2020 Page 1 of 2

One of the requirements that the President set on April 16^th for reopening the country is that the state or region must  show 14 consecutive days of declining daily case counts  No state has met this criterion in the strict sense.  For some of the states that have reopened or announced reopening soon such as Minnesota, Tennessee, Arkansas, Arizona, Indiana, Iowa, and Colorado the latest 14-day trend in their case counts is actually upward and they should definitely not ease any of the restrictions they have in place.  For some of the reopening states such as Texas, Ohio, Kentucky, Alabama, and Mississippi the trend is mostly flat and not downward.  Only for Florida, Georgia, Louisiana, Oklahoma, Alaska, Idaho, and Montana, are the most recent two-week trend edging downward.  The alternate gating criterion of a downward trajectory of positive tests as a percentage of total tests has also not been met by most states.  This criterion makes more sense in those situations where a state broadens testing protocol significantly to cover both asymptomatic and mildly symptomatic cases such that case counts may even rise temporarily.  

Florida cities and counties responded early and forcefully to tackle the COVID-19 pandemic as soon as the first confirmed case appeared on March 1^st.  So even though the governor only declared a state-wide stay-at-home policy on April 4^th, Florida is one of the few states that passed both versions of the President’s gating criteria on case count for reopening the state.  The figure below shows that a week after the governor’s state-wide order, the confirmed case count (brown squares against left axis) peaked around 1200 cases per day near April 9^th.  Fatality count (blue diamonds against right axis) began to flatten out a week later near 45 per day but still has yet to peak. 

The figure below shows the number of COVID-19 tests performed in Florida each day over the last 5 weeks.  Florida has tested about 1.78% of its population to date — comparable to the US total of 1.88%.  While the number of daily tests (brown squares against left axis) has increased by roughly 50%, the more significant metric is the percentage of tests that returned positive results (blue diamonds against right axis) dropped steadily from near 12% to near 6% — roughly half of that of the US as a whole recently at 12%.  This means that FL has begun to test more of their uninfected and less obviously infected cases.  It is important to continue to broaden the test administration so that Florida can gain a better understanding of where the disease remains so that effective isolation and contact tracing can be conducted to cut the transmission further.  Antibody serology tests which are easier to do could provide complementary information on what portion of the currently uninfected population has already been infected and may have gained immunity.  

Florida is reopening May 4^th with a conservative phased approach to minimize reinfection. Most of the state will reopen certain businesses like restaurants and retailers at 25% capacity and with strict adherence to the social distancing guidelines of the CDC.  Movie theatres, bars, fitness centers, and places that offer personal services will open later.  Miami-Dade, Broward and Palm Beach Counties that are the most infected will not reopen until additional data can be collected to show that they are sufficiently safe.

When we look at the cumulative confirmed cases and fatality data for close to two hundred countries, it seems quite remarkable that some countries seem to do 10 times better than other countries in terms of case fatality rate.  We might conclude that the novel coronavirus discriminates by country or that certain countries manage their pandemics much better than others.  The truth is more complicated. 

When examining mortality rates two issues are important: how we count deaths and how we test patients.  The case fatality rate is a straightforward statistic calculated by dividing number of confirmed deaths by the number of confirmed cases for any group or subgroup of patients.  For an ongoing pandemic, this statistic is almost always understated.  Firstly, because it generally takes several days for a confirmed case to resolve itself into death or a cure, so the denominator using the current case count is overstated.  Secondly, many deaths occur outside the healthcare system and although the patient was never tested positive for coronavirus they often show many signs that they probably died of COVID-19.  The cause of death assignment is often subjective, especially when there are comorbidities involved.  The most probable ones should be counted in both the numerator and denominator to determine the true COVID-19 fatality rate.

The other major problem is that the total number of infectees is rarely known while the pandemic is raging since we are most likely testing only a subpopulation of all infectees. Often the testing protocol stipulates that we test only the most seriously ill and most at-risk.  This protocol is actually not the most efficient way to test the population given limited resources because you wind up testing patients that are most likely to test positive anyway and miss a large number of asymptomatic and mildly symptomatic cases, vastly underestimating the denominator of the fatality rate.  This is one reason it is so important to test early and widely to get an accurate picture of how the infection is spreading within the general population.  The other major impetus for wide testing is, of course, to ascertain that every infectee is identified and isolated as soon as possible and all their potential contacts tracked and isolated as well to contain the spread of the disease.  Once we are able to do this we can more confidently calculate the COVID-19 fatality rate as the total confirmed and probable deaths due to COVID-19 divided by the total number of deaths (confirmed and probable) plus cures (including all those infectees that never showed any symptoms).  While the pandemic is raging we can only estimate this number. 

We need to do this so we can identify those segments of the population that are most at risk early.  The table below lists the top countries in the world in terms of confirmed cases and fatalities plus a few countries that have been identified as doing well in terms of managing their pandemics.  We had already examined the age and gender effect in Spain and Italy and found them to be very similar in form with mortality doubling every 7yrs in the range 35-75yrs old.  Both countries tested nearly 2.9% of their population (as of April 27) which seems like a reasonable amount but since both countries were highly infected 11% to 17% of the tests returned positive.  When a high percentage of tests return positive results it means that we are only testing those that are most suspect and not really testing the full population adequately.

South Korea is now held up as the best practice example in early and widespread testing and contact tracing.  While they have only tested 1.2% of their population, they tested to the point where only 2% of their tests returned positive.  Interestingly enough when we look at their age-dependent data, we see that they follow a steeper exponential form (mortality doubling every 5.9 yrs)  that is lower overall due to more complete sampling of the full population.  The steepening we believe is due to the fact that most countries test their at-risk (older male) population better than they do the younger population which tends to have milder and sometimes asymptomatic cases.  When we look at Germany which is often considered the best practice country in Europe, we see a very similar graph to that of South Korea.  Germany has tested more than South Korea (2.5% vs 1.2% of their population, respectively) but not as thoroughly as South Korea with 8% of their tests returning positive results. 

USA testing completeness and efficiency is very poor and needs to improve by a factor of 2.5 to match Germany’s efficiency.  Testing kits should not be distributed randomly to achieve the same percentage of the population tested, but rather more to those regions that are hotter than others and returning more than 10% positive results.  States like NJ that continue to return near 50% positive test results must increase their testing by more than 5 fold to achieve a more accurate measure of their infection.  If the President wants a lower fatality rate for the USA, he simply needs to improve testing.

Because age is such a strong effect, we see that the apparent difference in fatality rates between Germany (3.8%) and South Korea (2.3%) may be entirely explained by the older population of Germany (47.1 vs 41.8 yrs).  The difference between South Korea and New Zealand could also be explained by age distribution differences in their population.  As testing becomes more complete and efficient in the seven countries with double-digit percentage fatalities, we predict that their fatality rates will decline by more than a factor of 2.  The fatality rate differential now seen among all countries will narrow significantly and we may then search for other comorbidity risk and structural factors that could better explain any residual differences in how each country is impacted by the COVID-19 pandemic. 

Sweden’s response to the COVID-19 Pandemic has been very unusual and it’s justification for continuing the policy is magical thinking.  They have relied on voluntary social distancing and “herd immunity” to solve their pandemic problem.  There are few words to adequately describe the tragedy unfolding in Sweden so I’ll just let these tables and graphs speak for themselves.

The figure above shows the progress of the infection in Sweden since the first case was confirmed on Feb 25^th.  Since then the number of confirmed cases per day (brown squares) has increased steadily with no signs of stabilization.  The first death was reported on March 13^th and with about a 10-day lag, it has tracked steadily upward every day since then.  The death count (shown as blue diamonds) is plotted on the right axis scaled to 16% of the case count on the left axis.  That means the fatality rate is roughly 16%.  This mortality rate is much higher than that for its Scandinavian neighbors, Denmark and Norway, and Finland (shown in the table below).  Sweden’s mortality rate may be so high because they are testing very little and missing many of the asymptomatic and mildly symptomatic cases (last column in the table below), but it’s hard to know for sure.  What is more certain is that their death per million citizens is also the highest, at 213 per million. 

The table also shows that Sweden’s infection rate as measured per million population is the highest and its death per million is also the highest.  Given that it has the poorest testing of the three Scandinavian countries, it is possible that its infection rate is even higher.  Of course, the Swedes don’t care about this because their aim is to get to “herd immunity.”  In fact, the country claims that their infection rate is much higher than that measured and will soon get them to “herd immunity” — generally acknowledged to require 60% of the population to be infected — in Stockholm in a few weeks and the rest of the country soon thereafter.  Having measured so poorly, I am not sure how they know this except with their own proprietary models.  In our estimate, they are picking up about 20% of all real cases so their real infection rate is 0.86% (far from the 60% required for herd immunity) and their real mortality rate is 3.3%.

The tragedy is that it didn’t have to turn out this way.  Sweden now has about 2000 more deaths than its neighbor Norway.  They both started out with mild infections in late February and Norway kept its infection and death counts controlled.  On the graph below, Sweden’s death count (brown squares) is plotted against the left axis and Norway’s (blued diamonds) is plotted against the right axis adjusted for Norway’s smaller total population.  The differences in the graphs are striking and widening. 

Sweden should serve as a cautionary tale for the eight states in the US that have not imposed statewide lockdowns.  They could wind up looking like Sweden rather than Norway. 

New York City (NYC) metro area has been one of the hardest-hit areas in the world during the COVID-19 pandemic.  Whether defined as the Metropolitan Statistical Area (MSA) with 20.3 million, or the broader Combined Statistical Area (CSA) with 23.7 million it has the largest population in the United States and the activities of residents are closely intertwined.  All 31 counties in the CSA have been heavily infected by the coronavirus.  By itself, this CSA would be the most heavily infected country in the world with the heaviest death toll. 

We look at 17 of these counties with the closest commuting distance to Penn Station in Manhattan, summarized in the table above.  These counties have a total population of 18.1 million and comprise 77% of the CSA.  Every day, millions of commuters enter NYC (including The Bronx, Brooklyn, Manhattan, Queens, and Staten Island) for 4.2 million jobs, and then many return to their homes in the NY-NJ-CT-PA area.  Any disease they pick up in NYC quickly and easily spread into the neighboring counties. This has made the entire region the hottest in the world with 1.79% of the population in the17 counties confirmed infected and 0.13% of their population killed by COVID-19.  These numbers are 3 to 4 times worse than for next hardest-hit countries. 

How did metro NYC get this bad?  One reason is the lack of a coordinated regional or national policy and program to deal with the pandemic in the complex multi-state CSA.  The other is the late start to quick, accurate, and comprehensive testing.  Some epidemiological models suggest that on March 1 when the first confirmed COVID-19 case was found in NYC, as many as 10,000 cases had already spread in the region since late January.  Of course, testing has improved several hundred-fold since March and the number of hidden cases is no longer 10,000 times the number of confirmed cases.  However, several antibody test programs suggest that the hidden population may still be 30 to 50 times larger than the confirmed population in areas like Santa Clara County and Los Angeles County.  Were this true in NYC it would imply that NYC has already achieved herd immunity which it obviously has not as evidenced by the several thousand new cases reported very day.  A more recent antibody detection test conducted in NYC found that 21% of the population may be infected already.  We believe that this, too, is an over-estimate due to very high false positives from the test. 

Our estimate, based on current COVID-19 testing efficiency (43% of tests return positive results) and testing completeness (2.9% of the population tested) in the CSA is that testing is uncovering about 85% of all the serious cases and 30% of all the infected cases in the region.  The vast majority of the hidden population comprises asymptomatic or mildly symptomatic cases that currently do not qualify for testing due to limited availability.  This means that 6% (=1.79%*3.3) of the 17-county metro-NYC area could already be infected and that herd immunity (generally believed to require 60% of the population) could be achieved before August or sooner if regional lockdowns were relaxed or lifted.  The good news is that as of April 22, testing has doubled from last week’s average to over 300,000 per day, although many pockets of shortages remain and it remains to be seen whether this pace is sustainable.  In addition, the mortality rate is probably not as bad the 10.5% measured for NYC (including all probable cases as currently recommended by the CDC), but possibly just 5% when all COVID-19 cases are counted.  This would nevertheless translate to 25,000 deaths in NYC and 45,000 deaths for the NYC CSA by July 1.

One lesson to be learned is that early, strong, and coordinated action by the regional and national government is required to minimize the pandemic toll.  Early and comprehensive testing and tracking, and treatment is an essential part of that government response.  Another lesson from metro-NYC is that without a lockdown, the testing and health care systems would have been overwhelmed.  As it turned out, peak hospitalizations surpassed capacity for many days in NY, and on some days testing returned near 70% positive results in NJ.  That meant that they were testing near-certain cases (actually the wrong thing to do with limited testing capacity) and missing many serious cases.  The eight states who refused to impose state-wide lockdowns would do well to learn from NYC’s and other states’ examples.

The novel coronavirus (SARS-CoV-2) attacks the human body primarily through the airways and the lungs.  Classic symptoms of COVID-19, the disease caused by the coronavirus, are fever, cough, and shortness of breath.  But for many — perhaps as many as half of those infected — there are few if any symptoms.  People who suspect they have been infected are asked to shelter in place and to check their temperatures every day.  But for those who do not develop a fever or obvious shortness of breath or cannot get a COVID-19 test due to shortages, they may ignore their problems for many days while the disease silently ravages their lungs.  A better diagnostic tool than a thermometer could be an oximeter that is widely available and not much more expensive than a thermometer and much more accessible than a COVID-19 test.  Oxygen saturation levels below 88% were found to be the most useful marker of critical illness in NY hospital admissions. 

We now know that COVID-19 mortality is strongly associated with age, gender, and comorbidities such as obesity.  Obesity (BMI>40) and heart failure were the next most important indicators after age.  What is it about obesity that makes COVID-19 so much more deadly while other more suspicious conditions such as asthma do not seem to?  We suggest that another medical condition that is associated with obesity but sometimes ignored may be more predictive of hospitalization and mortality: obstructive sleep apnea (OSA).  OSA is known to be associated with age, gender, and obesity.  Older men who are obese are more likely to have sleep apnea than any other combination of demographic factors.  While OSA has been clinically diagnosed in only 4% of American males and 2% of American females, it is suspected to afflict nearly 10% of the US population in milder/ undiagnosed forms.  One additional demographics factor that has emerged recently is race: African Americans (AA) are more likely to die from COVID-19 than Whites.  Many socioeconomic factors such as poorer access to health care in general and testing, in particular, have been suggested, but one factor that has not is OSA.  It turns out that OSA prevalence among AA may be as high as 24% — much higher than the 5% formerly diagnosed.

Sleep apnea is a breathing disorder that may be greatly exacerbated by COVID-19 leading to respiratory distress and possibly failure.  Moreover, we think that persons with sleep apnea may have learned to adapt to mild oxygen deprivation and so that they may not notice a change in their breathing pattern until the coronavirus has caused severe damage to the lungs — leading to hospitalization and eventual death.  Sleep apnea patients on CPAP (Continuous Positive Airway Pressure) and BiPAP (Bilevel Positive Airway Pressure) machines may actually feel no change because small initial damage can be easily compensated by the machines at night, and more rapid breathing during the day.  However, the use of these machines can cause the disease to spread more rapidly because CPAPs, unlike true ventilators, blow contaminated air into the room that may circulate into the building causing widespread infections in apartment buildings, nursing homes, veteran homes, prisons, etc.

It is important to test this hypothesis immediately, but equally important to have the CDC recommend the use of oximeters to detect and monitor potential victims of COVID-19.  As soon as they are identified, they should be tested, isolated, and treated to reduce the risk of infection to others.

With over 2.4 million COVID-19 cases and over 160,000 deaths globally, there is a lot of data that can be analyzed to better understand this virus as a function of age, co-morbidities, medical history (such as prior exposure to coronaviruses like SARS, and other vaccines like BCG), etc.  The two important factors that have been noted since the early days of this pandemic are age and gender and they continue to be the most important factors.  We look at the age and gender dependencies with the best single data set provided by the Spanish Health Department containing over 200,000 cases (or 0.4% of the total population) and over 20,000 deaths.  Spain has tested over 2.0% of its population which while short of a comprehensive sample does begin to give us some ideas about how the disease has affected the entire country. 

We look at the probability of a random person acquiring the disease as a function of age.  We know that this sample is not truly random nor comprehensive but it does give an intriguing result that should be examined further.  The figure below shows that an 80+ years old person has about 100 times the likelihood of getting infected compared to a person under 10 yrs old (the left axis measures percent of the population a certain age group has been infected with 0.4% of the population as a whole).  Some of this large difference may result from the fact that an 80+ years old person has a higher probability of getting approved for a test than a young person.  But if there is a real effect, it is all the more important that we perform random testing of a large sample of people to find the true infection rate by age.  The infection rate varies very little by gender.  Slightly more women than men get the disease, 53.6%, but Spain is 51.0% female.  

Another interesting question to ask is how and why the mortality for this disease varies so significantly by age and gender.  For regular seasonal flu, there is a modest mortality differential by age with older people more likely to die from it than younger people.  However, with COVID-19 someone older than 80+ has nearly 100 times the chance of dying than someone younger than 30 yrs old (see figure below where the mortality rate as a percentage is shown on the left axis).  One important note is that this age variation is continuous and not a step function as some people commonly assume.  Between 35 and 75 the mortality risk doubles for every 7 years increase in age.  There is no cliff at 65+ where your risk increases all of a sudden — the risk increases continuously for both men and women above the age of 35.  The graph does show that men (dashed grey line) are at a higher risk of dying from this disease than women (dotted red line), by about a factor of 2 for all ages from 35–75, with the gap widest at 2.6x in the 50–59 age group.   

The same mortality versus age trend was observed in Italy (shown as the solid blue line in the figure above).  Italian mortality was generally 20% higher than for Spanish most likely due to the fact that the sample taken as of March 30 was biased toward 1.25 times more males than females.  (However, the underlying population split was the same 51% female to total as Spain so it is unclear why Italian men are more likely to get infected than Spanish men vs women — probably a large sampling bias.)  The same mortality versus age and gender relationships were also observed in China.  So this age and gender effect seems to apply to this pandemic globally.

If both infection rate and mortality rate are strongly biased against the elderly it means that an 80+ years old man has 10,000 times higher chance of dying from COVID-19 than a kid — an astonishing disparity.

Governor Mike DeWine announced on April 16th that he wants to re-open Ohio (OH) for business on May 1st, but it looks dangerous.  May 1^st would make OH the first state to re-open, while other states such as New York (NY) and Pennsylvania (PA) just extended their lockdowns to May 15^th.  He says he will do so cautiously in phases which is the right approach but the latest upsurge in newly confirmed cases should cause him to re-think his strategy.  The OH governor has received praise for his early recognition and response to the pandemic, but his initial response was tepid and he only imposed a strong lockdown and stay-at-home policy four days after CA on Mar 23^rd when his state was about 1.5 (±0.1) times more infected than California (CA).  By then OH had 38 cases per million population whereas CA had 25 cases per million four days earlier (of course both states are much better than NY). 

Until recently OH had followed an infection record as controlled as CA (see figure below where the left axis for CA and right axis for OH have been adjusted for their differing populations; the horizontal axis tracks real-time for CA and has been shifted back 4 days for OH’s delayed start).  However, over the last two days, OH has experienced a renewed surge.  OH did update its definition for COVID-19 on April 8^th to include probable cases, but that should not have produced an upsurge on April 15^th unless there was a delay in implementation.  Another reason may be OH has lots of infected neighbors.   

If any state can re-open early it is CA because it was the first state to lockdown and its major cities and metropolitan statistical areas (MSAs) are self-contained — with the Rockies and the Pacific serving as very effective shields for the state.  However, Governor Newsom has said that he may extend the CA lockdown into June or at least until he can get more thorough testing.  OH is in a much tougher position to control its pandemic or re-open by itself.  It is part of two large MSAs: one in Cincinnati with Kentucky and Indiana and the other with Pittsburgh in PA.  OH also borders West Virginia, and heavily infected Michigan.  This means that Governor DeWine should not re-open without close coordination with all these 5 states.  If he re-opens earlier than his neighbors he can infect his neighboring states.  Complex MSAs like NYC and Cincinnati should have been managed as a unit with full cooperation between federal, state and local governments.  This is why the lack of Federal government coordination in testing and lockdown has had such as a disastrous impact on the US war against COVID-19.

While 95% of the US is locked down now, and the President is pushing the governors to make plans to re-open their states for business, eight states have yet to institute a statewide lockdown or closure of non-essential businesses and stay-at-home order.  These states are all infected and may experience uncontrollable outbreaks in the near future.  All of them have made some efforts to mitigate the pandemic and three of these states (highlighted in pink in the table below) have partial lockdowns for certain cities and counties within their state imposed by local governments that help to mitigate their risk.  But unless they take quick and decisive action to lockdown their state in full they will continue to expose their residents to unnecessary deaths and expose the entire country to re-infection even as the US daily confirmed case count appears to be topping out. 

Six out of the eight states (top 6 out of 8 rows in the table above) are seriously infected already and will have to institute state-wide lockdowns to control the spread.  Two of these, South Dakota (SD) and Iowa (IA) may suffer very badly before they gain control of their outbreaks.  No other state has ever let their infections get as bad as SD before imposing state-wide lockdown and only one state, New York (NY), locked down after where IA is today, and no one wants to follow in the footsteps of the tragedy in NY.  Our forecast for SD and IA is that before it all ends this year more than 1% of their population will be infected and more than 400 per million of their citizens will die.  Relying on good behavior and herd immunity does not work as evidenced by Sweden, and we urge the legislators and citizens in these states to warn their governors to act before it is too late for them and for the rest of the country.

Two weeks ago we posted a commentary about the importance of quick, accurate and thorough testing in the war against COVID-19 and suggested that after an abysmal start the US was finally on the right track.  We are less optimistic now — particularly after reading about the President’s recent appeal to the South Korean government for 600,000 test kits.  What is the problem with the much-touted US COVID-19 testing program? 

The figure above shows that after a slow start in February and through early March, cumulative testing (blue circles) accelerated exponentially from early to late March when its growth slowed again.  Since April 9^th when daily tests peaked (brown squares against left axis in the figure below) at 163,172 completed tests, testing has steadily declined to 129,854, yesterday.  This is highly concerning since the President and some governors have promoted the accessibility, speed, and accuracy of US COVID-19 testing for many weeks.  Moreover, if US testing has hit another bottleneck due to lack of testing personnel, equipment and/or supplies, some of the recent declines in daily confirmed case count may be artificially hiding an underlying increase in real daily COVID-19 cases. 

To check this possibility we looked at the percentage of tests that are returned positive (blue diamonds in the figure above plotted against the right axis) over the last six weeks. Back in early March testing was definitely constrained and on several days the percentage of positive results approached 20%.  Then as the US ramped up testing by permitting commercial labs to run these tests and removing the CDC as the bottleneck, the percentage of positive test results fell below 10% for many days near mid-March indicating that we were beginning to test many weakly symptomatic or asymptomatic cases.  However since then the percent positive has climbed back to 20% in early April and stayed above 1 out of every 6 tests.  Compared to best in class countries such as Taiwan (1in 118), South Korea (1 in 49), and Singapore (1 in 25) this is still abysmal performance. Compared with good European nations such as Norway (1 in 19), Germany (1 in 10) and Austria (1 in 10) our record is still bad.  As we mentioned yesterday in regards to the UK, countries with a high percentage of positive results usually have very high mortality rates either because they have not measured all those that have been infected by a coronavirus (leading to an artificially low denominator), or they have truly let this asymptomatic portion of the population silently infect and kill many older and vulnerable citizens.  We need to at least double our testing to 300,000 per day to match the effectiveness of the best Europeans at uncovering most if not all COVID-19 cases in the US.

There are some states that are much worse than average on this metric.  New Jersey (NJ) and New York are the two worst states with 50% and 41% cumulative positive test results, respectively.  Moreover, on a daily basis, NJ has gotten worse.  On April 13^th, 69% of all completed tests on that day came back positive.  NJ has one of the more restrictive COVID-19 testing protocols that could be missing half of all cases including many that are asymptomatic or weakly symptomatic.  This means that the true infection rate in NJ is probably at least double that of the 0.73% currently reported.  The NJ governor has recognized this problem and has ordered 15 Abbott ID NOW machines from the Federal government on April 10^th to augment NJ’s stressed testing program.   It is unclear when relieve will come and when they will be able to get an accurate picture of their infected population.

Many analysts have asked why is California (CA) doing so much better than New York (NY) in this pandemic.  NY experienced its first case on March 1^st, while CA experienced its first case earlier on January 26^th.  In both states, the infection did not become obviously serious until March 11^th when the infection count crossed 200 cases in both states.  On March 19^th when CA confirmed case count reached 25 per million population Governor Newsom declared a lockdown in CA.  On March 22^nd, NY had already reached an infection rate of 812 per million before Governor Cuomo declared a lockdown.  Thus, even though NY was only 3 days behind in enacting strong mitigation measures, it was 33 times more infected by then and that much further along the exponential growth curve.  In the early days of an infection’s exponential growth phase quick and decisive action is crucial.  NY’s slower response then has led it to have a measured infection rate that is 16.2 times worse than CA now. 

In addition the NY situation was complicated by its strong infection epicenter in metro-New York City (NYC) which involved a strong-willed mayor and two other governors who needed to cooperate to shut down the entire region to prevent cross-infection.  This did not actually happen until March 23^rd when CT locked down and NY state infection had already reached 1073 per million.  These delays have cost NY thousands of unnecessary lives.  We had urged the government to act forcefully on Mar 9^th.  Had any of them acted then, much of the tragic loss of lives could have been avoided and we would have been 2 weeks further along on our recovery. 

The magnitude of the difference between the two states is actually more than a factor of 16.2. NY’s death count per million is 28.1 times worse than CA.  Some of this difference may be due to the younger population in CA (36.8 vs 39.0 yrs) especially in Santa Clara County where many of the early CA cases were located.  But the delayed reaction in NY that led to overwhelmed hospitals and testing facilities in NY might have also played a part.  When you look at the number of tests performed in NY vs CA it might look like NY is doing better with 2.6% of its population now tested against just 0.5% of the CA population tested to date.  But this not the best metric to compare.  When the infection rate is higher you need to test more to prove that you have tested all the mild and asymptomatic cases.  The percentage of all tests that yielded a positive result is a measure of how thoroughly a state has tested its population.  On this measure, CA is doing better than NY with 13% of completed tests yielding a positive result compared to NY’s 42%.  NY’s number shows that it is probably only testing the most seriously ill patients and not probing the true spread of the disease in NY.  Thus the true infection rate in NY might be 2% of its population and the ratio between the 2 states may well be not just 16.2 times but 25 to 35 times worse in NY.  This would mean that NY with an infection rate that was 33 times worse at the time of lockdown has continued to maintain that same ratio of infection disadvantage 3 weeks later.  Other considerations such as socioeconomic factors like race, income, and behavior, as well as population density, probably also matter but we argue that all the observed differences could be just due to the math of exponential infection.  

NY suffers from the terrible legacy of delayed and botched testing and delayed and weak lockdown decision-making during a pandemic which has carried through to this day.  Of course, things could have been much worse for both states if the local and state governments had waited for the President to act.  Both states would have been still experiencing exponential growth instead of showing clear signs of a top last week (see figure above).