October 9 COVID Update: Midwest and Great Plains See Continued Growth

The following summary and its attached report are the twenty-second in the ongoing series by the W. Capra Data Science team on the impact of the Covid-19 outbreak on the industries we service over time.  The previous reports can be found in the following links: April 14, April 23, April 30, May 6, May 13, May 20, May 28, June 4, June 11, June 18, June 26, July 2, July 9, July 16, July 23, July 31, August 7, August 14, August 21, September 3, and September 17.

Introduction

The continuing COVID-19 pandemic has changed the American way of life and its business landscape.  Cases, deaths, and business closures have had an immense impact on every single community across the nation.  To keep W. Capra, its clients, and its external partners updated on the latest trends with COVID-19, this report is published with state-by-state data trends.  The trends try to predict how states handle the pandemic over time, which states are nearing containment of the outbreak, and what factors – regulatory, cultural, etc. – are important to determining future outcomes.  Some of the trends examined to determine those factors include the testing, case, and death velocity and acceleration, along with other fatality rates and the reproduction rate of the virus itself across states.  The data can give much information but is incomplete for many reasons.  The largest reason is the fact that patients can be either symptomatic or asymptomatic with a wide range of estimates of the number of patients who are asymptomatic.  Because of the existence of both symptomatic and asymptomatic patients, testing is skewed towards the most symptomatic patients, which inflates the case rate and fatality rates.  Expanding testing gives a clearer picture of how widespread and deadly the disease is.  Secondly, the differing guidelines across communities and the variance in social adherence make determining effectiveness difficult, and only approximate trends are listed.  Finally, the data reported does not fully reflect state and national trends due to changes in reporting from government agencies, causing delays due to new requirements and quality control issues – so those metrics are left out of the report due to their propensity to be incorrect.

A note before we begin: the outbreak and the data surrounding it changes daily. This report was created when looking at the outbreak as a data problem that might benefit from data-driven solutions and insights. It is not intended to be a substitute for medical or safety advice, nor is it a recommendation on outbreak response currently in place in various locations around the country. Individual assessment of local laws and current official government and health guidance should be reviewed before making any decisions.

Analysis

Testing plays a crucial part in achieving containment and maintaining safe levels of the pandemic.  Currently, enough tests have been administered for 32.47% of the US population, with the actual percentage of Americans tested being lower than the stated number as some individuals have been tested multiple times.  Distribution of the tests is also uneven, with some states testing a higher percentage of their population than others.  Inconsistent testing reinforces the idea that testing is not performed at the necessary pace and can skew towards symptomatic patients and inflate the bias of the death rate. 

The resulting case rate – or the number of tests that are positive – of the outbreak varies by state, with some states reaching over 45% at some point during the pandemic.  Case rate fluctuation has occurred as asymptomatic and less symptomatic people were told to self-quarantine and would only receive a test later when directed by a doctor.  The true case rate is much lower, but since every individual cannot be tested every day, the true rate remains unknown.  Right now, the case rate metric being referenced is a two-week positivity rate to reduce a state’s bias throughout the course of the pandemic and focus on much more current trends.  This, coupled with the fact that testing has increased significantly since the beginning of the pandemic, has resulted in an overall case rate of 7.1%.  Some states are much higher than the national rate, with North and South Dakota, Idaho, and Wisconsin all reaching over 20%.  These numbers suggest that testing in becoming less accessible, not more accessible as needed.

Case fatality rates track directly behind the case rate.  This report utilizes a lag-adjusted case fatality rate, which is currently 1.7% for the US.  There exists a lag because there is an average two-week lag between a case being reported and that case resulting in death.  This rate is for all discovered cases, but the true infection fatality rate is lower as some cases are never discovered.  Right now, many states are near the national average, but there are several states that are above the national average, led by FL and MA with both states being over 3%.

While there is much that remains unknown due to missing or incomplete information, state and national level performance can be tracked to monitor the spread of COVID-19.  Various states are in various stages of the pandemic and their numbers, between testing, cases, and deaths, all fluctuate.  Positive outcomes are when new daily cases and deaths drop – signifying a peak in those metrics.  These velocity metrics, and their associated accelerations, help indicate statewide trends.  While a state experiences increasing growth in the outbreak, it is considered in the worst of four stages – the Exponential stage.  After peaking, the number of cases will still increase but at a shrinking rate – the Linear growth stage.  Once the acceleration turns negative, the number of new cases and deaths will generally shrink day-to-day – the Improvement stage.  Once the change in daily new cases and deaths reaches a steady-state near zero, the state is considered in the Containment stage.  To make a rough estimate of the acceleration to place states in one of these four stages, their velocities and accelerations utilize a weighted moving average for both cases and deaths, which removes the skew from data prone to large swings.

Nationally, the US has seen many fluctuations in the case and death rates around the country.  The first peak in cases experienced happened in late April and early May, with a long, flat decrease followed by another rapid increase in cases that peaked sometime in July.  The country has experienced case declines for the months following the July peak, but recently the number of cases has been increasing again.  The death rate has followed a similar trend, peaking in May followed by a prolonged period of decline, then peaking again.  The difference, while the July peak in cases dwarfed the previous peak, the death rate peak was nowhere near the previous peak in deaths.  The hypothesis for the difference in first and second death rate peaks is the growing familiarity with the disease and treatment. 

The states themselves are seeing many different macro trends.  Originally, the Northeast was the epicenter of the outbreak in the US with NY and NJ being hit particularly hard.  With extended lockdowns and other measures, the Northeast has been in a much-improved position for a while.  Over the summer, the South and West became the outbreak epicenters as many communities and states did not implement similar measures as the Northeast or for long enough periods.  Eventually, the outbreak was in the Exponential stage for the entire country outside of the Northeast.  Late summer saw the outbreak reach better control, as many states moved out of the Exponential stage with many reaching the Improvement stage.  As states have moved to reopen more fully, many have reverted to previous circumstances.  Currently, 17 states are in the Exponential stage, mostly those in the Rockies and Midwest.  The second worst stage, the Linear Growth stage, currently contains 19 states across the country, mostly in the Midwest and South.  The second-best stage, the Improvement stage, has 11 states, including some of the largest with CA, TX, and FL all in this stage.  Finally, the Contained stage only has 3 states: NY, NJ, and CT.  To read an unabridged version of the results, please see the attached report for a complete view of specific states.

Conclusions

The United States’ new death curve continues to decrease, but the decreases in the new case rate have recently completely reversed and cases have been increasing.  The decline in the death rate continues slowly. Overall, the trend in death rate continues to decrease with improved treatments and more resources coupled with the declining age of the infected population.  Recently, North Dakota has become a hot spot for the virus and provides a clear example of how deflated testing affects latter measures.  North Dakota has witnessed a continued decrease in testing, resulting in a case rate of 28% with deaths increasing as well.  Expanding testing will help to identify infected individuals faster and help bring down the case and death rate with more outbreak visibility.  However, the early indications of the new case velocity and acceleration showed that the recovery period will most likely be much longer than the rapid acceleration of the virus and new cases.  Recovery for North Dakota and other states like it will take much longer than the few weeks it took to revert to exponential spread.  Overall, many Midwest and Great Plains states, not just North Dakota, are experiencing exponential growth.  Southern and Western states are seeing a plateau in improvement, with some seeing case growth after weeks or months of improvement.  All of this contrasts with NY, NJ, and CT which have had virus containment since June.

For further discussion of data modeling or anticipated COVID-19 business impacts, contact the W. Capra Data Science team:

Nate at nrao@wcapra.com

Stu at sgreenlee@wcapra.com

Sources

The COVID Tracking Project, https://covidtracking.com/

New York Times COVID Data, https://github.com/nytimes/covid-19-data

COVID-19 Community Mobility Reports, Google, https://www.google.com/covid19/mobility/

Normalized Maximum Heatmap by Sergy Bryl, https://analyzecore.com/2020/05/04/the-spread-of-covid-19-across-countries-visualization-with-r/

R~t COVID-19 Estimates, https://rt.live/