All posts by petezajonc

Covid-19 Deaths by County Characteristic – Diabetes Incidence

Acknowledgment

This US County health analysis is dedicated to individuals who are the victims of the Covid-19 virus and to all the people working hard to keep our US Counties safe, healthy and nourished.

St. Paul, MN native Marion Goldetsky Klein was devoted to her family, her community and her Jewish faith.  “She didn’t know the meaning of the word ‘no.’ If you asked her to do something, she did it. She was active in every Jewish women’s organization possible, not just as a member, but she was a leader.” Klein died after contracting COVID-19 in the Memory Care Unit of Sholom Home West in St. Louis Park. She was 87. Star Tribune

This analysis relates County diabetes incidence to covid-19 deaths across 3,142 US Counties. Are Counties with populations most affected by diabetes also Counties with high covid-19 death rates?

County health statistics report on public health. The information supports community investment decisions made by local, County, and Federal administrators to improve County health services and public health.

Summary — Data as of May 14, 2020

In summary, public health and health services in some large Counties and many small, rural Counties with high diabetes incidence are also combating relatively high covid-19 death rates.  Mid-size Counties, with between 250,000 and 1 million residents, have a large share of covid-19 deaths but the death rates from covid-19 in those with or without above average diabetes incidence are about the same  —  mid-size Counties with large Hispanic American populations has an effect on these results.

Perhaps unsurprisingly, Counties with high median incomes are associated with slightly lower covid-19 death rates than Counties with low median incomes. This is indicated by the upward tilt of the black linear trend line in the first chart below (highest income Counties on the left but the highest covid-19 death rates are to the right). A “negative” association is also found between County median income and County diabetes incidence, but it is stronger.

Covid-19 Deaths v. Diabetes: US Counties Ranked by Median Income

How strong is the association between County diabetes incidence and County covid-19 death rates? In fact, there is no overall association between County diabetes incidence and County covid-19 death rates. This is illustrated by the “flat” black linear trend line in the chart below. But, groups of Counties do differ, of course, and, as indicated above, this analysis illustrates those where diabetes incidence and covid-19 deaths are related.

Covid-19 Deaths v. Diabetes: US Counties Ranked by Low-to-High Diabetes Incidence

Prosperity is more complex than just consumption. The covid-19 crisis has proven that a healthy and safe workforce is required to sustain consumer societies. It is likely we’ll be better off if US economists begin to measure our markets and shareholders value our corporations in ways that will account for not just the goods and services they create, but also the value they contribute to sustaining community health outcomes.

Analysis of Covid-19 Deaths

Many of these trends relate to County public health and County health resources. The pandemic has fallen heavily on 6 NY state Counties that were ill-prepared for it. It especially affected the 44 largest US metropolitan “A” Counties with 1 million residents or more.  Nationwide “A” Counties account for 28% of the US population and 45% of covid-19 deaths.  The table below shows that covid-19 death rates in “A” Counties are 210% higher than in the rest of the US (the calculation of death rate ratios, such as 2.1 : 1, is explained in the Analysis Approach section at end of this analysis).

US Population and Covid-19 Deaths by County Size Group

Excluding New York state brings “A” County death rates back in balance – a 1 : 1 covid-19 death rate ratio compared to the rest of the US.  Covie-19 death rates in 250,000 to 1 million resident “B” Counties are also very high.  Their vulnerability is more pronounced when New York state is excluded.

Relating covid-19 death rates to US Counties with above an average diabetes incidence, which affects 8% of the US population, helps understand the imbalances.  The chart below highlights Counties with above average diabetes incidence (orange bars).

Taken as a whole above average diabetes incidence Counties represent 35% of the US population.  As indicated above, overall, their covid-19 death rates are the same as the rest of the US though they account for 46% of US diabetes cases.  But the even 1 : 1 death rate ratio for the US overall masks differences behind large and small Counties with high diabetes incidence.

Only a small fraction of urban “A” Counties with populations over 1 million have above average diabetes incidence rates.  However, this 2% slice of the US population (or 6.8 million) reports 15% (or 12,000) of all US covid-19 deaths.  Residents in “A” Counties with above average diabetes incidence have 500% higher covid-19 death rates (6 : 1 ratio) than the remaining “A” Counties.  Death rates are 360% higher (4.6 : 1 ratio) than other “A” Counties when New York state is excluded.  A few “A” Counties have large diabetes and covid-19 public health problems and are listed below.

“B” Counties with 250,000 to 1 million residents have the highest number of covid-19 deaths; however, covid-19 death rates are slightly higher in “B” Counties with below average diabetes incidence (blue bars).  “B” Counties that keep the incidence of diabetes low have succeeded holding off high covid-19 death rates.

71.7 million Americans reside in rural “C” and “D” Counties with populations under 250,000.  In “C” Counties, with between 30,000 and 250,000 residents, covid death rates are 10% higher (1.1 : 1 ratio) in Counties with above average diabetes incidence than the remaining “C” Counties.  In “D” Counties with less than 30,000 residents, they are 80% higher (1.8 : 1 ratio) in Counties with above average diabetes incidence.

In some large “A” Counties and smaller “C” and “D” Counties where diabetes incidence is above average, public health and health services struggling with diabetes are also struggling with relatively high covid-19 death rates.  While most covid-19 deaths are in “B” Counties, the death rates are lower in the B Counties with above average diabetes incidence.

Counties where Diabetes Incidence are 8% or more of the population (orange)

The largest share of covid-19 deaths among “B” Counties falls on those with above 18% Hispanic American populations — they have 90% higher covid-19 death rates than the remaining “B” Counties (see previous article in this series). But among them the share of deaths is largest in those with below average diabetes incidence. Above average Hispanic American Counties with below average diabetes incidence have the highest “B” County covid-19 death rates. Above average Hispanic American “B” Counties with above average diabetes incidence have 20% lower covid-19 death rates. As the table indicates “B” Counties with smaller, below average Hispanic American populations and above average diabetes incidence report 10% higher covid-19 death rates (1.1 : 1 ratio) when compared to the remaining below average Hispanic American “B” Counties.

“B” Counties where Diabetes Incidence are 8% or more of the population — with above and below average Hispanic American populations

Current covid-19 deaths and death rates for the highest diabetes incidence counties in each urban-rural County size group are listed below.  As a reference, the overall US covid-19 death rate is currently 0.03%.  All of these Counties have above 8.0% incidence of diabetes.

Top Counties where Diabetes Incidence are 8% or more of the population by County Size

 

 

Many American Counties have alarming covid-19 death rates, and many have alarming diabetes incidence.  Counties require community investment in parks, nutrition services, hospitals, health clinics, and, very soon, contact tracing, if the US wants all County resident to prosper. Resources are required so those that are under-served can take ownership of and sustain their future.

Analysis Approach

This is one of a series of analyses that relate covid-19 deaths to key County health and demographic characteristics, such as cancer death rate or percent of adults over 60. In each, side-by-side bar charts illustrate how Counties are faring in the covid-19 crises when grouped by size of population and median income and split by those above and below the characteristic’s average.

Each analysis focuses on Counties where the health or demographic characteristic is higher than some “average”. The “average” is defined here as the characteristic’s incidence or percentage — think of average cancer death rates or percentage of adults over 60 — that accounts for roughly 1/3rd of the US population when the Counties that qualify are pooled together. Covid-19 deaths are then totaled for Counties that come in above or below the “average”.

Side-by-side bar charts illustrate how Counties are faring in the covid-19 crises when grouped by size of population and median income and split by those above and below the characteristic’s “average”.

The numbers also tell a story as 1/3 versus 2/3 population comparison is written as an population ratio of 1 : 2, which is the same as 0.5 : 1 since we want 1 to represent the basis of comparison. Of course, these “odds” ratios change as the data and the populations change. Take, for example, the ratio that helps us compare the covid-19 death rates for Counties with above average percent adults over age 60 to those with below average percent adults over 60. All things being equal, the ratio of covid-19 death rates should match the ratio of the County populations. Typically it won’t: “For Counties with 1 million or more residents, the covid-19 death rate ratio of those with above average percent adults over age 60 to that of the remaining Counties with 1 million or more residents is 1.2 : 1. In other words, residents living in Counties with above average percent adults age over 60 have a 20% higher chance of a covid-19 death than residents living in the remaining Counties that have 1 million+ population.”

Here’s an example. First, the covid-19 death rate is computed by the covid-19 deaths divided by population. For the entire US, as of today, the covid-19 death rate is 0.03% = 83,752 / 325,719,178 meaning 3 out of 10,000 residents have died. Next, we create the covid-19 death rate ratio by comparing the death rates of the Counties with and without the characteristic. In this example, the 1.2 : 1 ratio means residents in “A” Counties with above average percent adults age over 60 have a 20% higher chance to die from covid-19 than residents of Counties without the characteristic.  The ratio is calculated here as: 1.2 = (3,615 /7,585,831) / (34,189/ 85,069,756). The 1.2 represents the numerator, or counties with the characteristic, and the 1 represents the denominator, or the counties without the characteristic.  Again: “Residents in Counties with the characteristic are 20% more likely to die from covid-19 than residents in Counties without it.”

Example: “A” Counties where Adult Age 60+ is 23%+ of County population

Update on Past Analyses

Above Average White American, Hispanic American and African American Counties. Counties with above average White American populations (85% or more White American) have covid-19 death rates 70% higher than all remaining US Counties.

Counties where White Americans are 85% or more of the population (orange)

Counties with above average Hispanic American populations (18% or more Hispanic American) have death rates 200% higher than other Counties to have covid-19 deaths.

Counties where Hispanic Americans are 18% or more of the population (orange)

In Counties with above average African American populations (13% or more African American) death rates from covid-19 are 250% higher than other US Counties.

Counties where African Americans are 13% or more of the population (orange)

Age 60+. 35% of the US population live in older Counties defined as above average Age 60+ populations (orange bars on the left graph). Since 8 of 10 covid-19 deaths are adults 65 years and older Counties with above average Age 60+ populations, D Counties, are vulnerable – though covid-19 death rates are higher in D Counties with below average Age 60+ populations (blue bottom bars).

Counties where Adult Age 60+ is (orange) or is not (blue) 23%+ of County population by County Size

Heart Disease. It may seem like Heart Disease is a “city problem”, and it is, but even when New York state is removed higher covid-19 death rates are found in A counties with above average Heart Disease deaths. Populations in rural C and D Counties with above average Heart Disease deaths also have relatively more covid-19 deaths.  This pattern of covid-19 deaths mirrors that of diabetes incidence, though the top ranking Counties differ.

Counties where Heart Disease causes 0.2% or more population deaths (orange) or not (blue)

Cancer. The above average Cancer death Counties (orange bars in the charts) account for 38% of the US population. This population resides frequently in rural C and D Counties and have covid-19 death rates 20-30% higher than other C and D Counties.  Few A Counties have above average cancer deaths, most likely because the populations in these Counties tend to be younger.

Counties where Cancer causes 0.19% or more population deaths (orange) or not (blue) by County Size

More on County Groups, more on the US Excluding NY

3,142 US Counties are geographic and administrative districts for which lots of statistics are gathered, including covid-19 deaths. Each County can be characterized as urban and rural, and they range in population from 10.2 million (LA County, CA) down to 88 (Kalawao County, HI) residents. In these reports, comparisons are made between four groups of counties depending on population size and by median income bands, as indicated in the chart axes.

New York state accounts for 6% of the US population and disproportionately more covid-19 deaths, so it is useful to consider total US results excluding New York state. This is done in the chart below which shows a disproportionate majority of covid-19 deaths occurring suburban B Counties across the US but excluding New York State (second bar on right graph).

US Population and Covid-19 Deaths by County Size Group — Excluding NY State

What Next?

In future reports we’ll continue to examine county health statistics that might help put covid-19 statistics in a helpful light.

Some other measures that may tell the developing covid-19 story:

  • How do results compare if we adjust the data to look at covid-19 deaths for each County one month after the first one was reported?
  • Do Counties’ previous flu or respiratory deaths tell us about County covid-19 death rates?
  • Do County median income or education tell us about the likelihood of covid-19 deaths?
  • Do results change if we use the exact density of County population instead of County Size?
  • Do average temperatures or levels of particulate matters matter?
  • How do 2016 voter preference associate with covid-19 deaths?

If you have any questions about this analysis or the data, or if you have suggestions, please don’t hesitate to contact me.

-Pete

Sources: US CensusUSA Facts Covid-19IHME forecasts, US Cancer Deaths NIH 5 year average, CDC Handbook on Death Reporting, 2003New York TimesMedian IncomeCDC Wonder Detailed MortalityCDC Fine Particulate Matter 2003–2011

 

Covid-19 Deaths by County Characteristic – Above Average White, Hispanic, and African American Counties

Acknowledgement

Takuo Aoyagi, an inventor of the pulse oximeter, an ingenious and indispensable medical device that measures oxygen in the blood and has become a staple of hospitals around the world, emerging in recent months as a key tool in the fight against the novel coronavirus, died April 18. He was 84.  Washington Post

This US County health analysis is dedicated to individuals who are the victims of the Covid-19 virus and to all the people working hard to keep our US Counties safe and nourished.

Summary — Data as of May 7, 2020

This analysis uses County statistics to report on public health outcomes relating to the covid-19 crisis. The hope is that this information will help support community investment decisions by Federal and County administrators to improve resident health services and public health. This analysis examines County ethnicity statistics to illustrate their association with covid-19 deaths across 3,142 US Counties.

In summary, Counties with above average White American populations are faring well in the covid-19 crisis.  In fact, among the lowest median income US Counties, only 0.1 deaths occur in those with above average White American populations for every single death that occurs in all other low-income Counties.  Public health expert guidelines would suggest that better diet and exercise coupled with better health services contribute to these positive outcomes.

On the other hand, Counties with above average Hispanic American and African American populations have had the worst outcomes in the covid-19 crisis.  In mid-size population Counties with above average Hispanic American populations 2.0 deaths occur for every single death in other Counties of similar size.  Even among America’s highest median income Counties, those with above average African American populations have 3.3 covid-19 deaths for every single death occurring in the other high-income Counties.  Most alarmingly, 7.0 deaths occur in above average African American population Counties that have the lowest median incomes for every single death that occurs in all other Counties in the lowest median income band – that is 700% more deaths.  According to expert guidance, above average Hispanic American and African American Counties require community investment in parks, nutrition services, hospitals, health clinics, and, very soon, contact tracing if the US wants all Counties to prosper.  Resources are required so those that are under-served can take ownership of and sustain their future.

Covid-19 Deaths and Ethnic Percentages for Each State

5-7-20 #1 pct eth

 

Prosperity is more complex than just consumption. The covid-19 crisis has proven that a healthy and safe workforce is required to sustain consumer societies. So, can US economists measure our markets and can shareholders value our corporations in ways that account for not just the goods and services they create, but also the value they contribute to sustainable community health outcomes?

 

Analysis of Covid-19 Deaths

The pandemic has fallen heavily on 6 NY state counties. This especially affects the results of the group of 44 large US metropolitan “A” Counties with 1 million residents or more. They account for 28% of the US population and 47% of covid-19 deaths — for every covid-19 death in the rest of the US 2.2 have occurred in “A” Counties. When New York state is removed the rate of covid-19 deaths is the same for “A” Counties as it is for the rest of US Counties.

However, on removing New York state, another imbalance comes to light.  For 228 “B” Counties with between 250,000 to 1 million residents, outside of New York state, 2.1 covid-19 deaths occur for every single death in other US Counties.  Many of these trends relate to County ethnicity, but most importantly also to the public health and health resources available to each.

US Population and Covid-19 Deaths by County Size Group

5-7-20 #1 pct z county size

To better understand these imbalances, this analysis begins with the US Counties with the fewest covid-19 deaths which are above average White American population Counties. In the graph below, US Counties are split into two groups with the results for above average White American population Counties highlighted (orange). Above average white population Counties experience 0.3 deaths for every death that occurs in the remaining US Counties.  This suggests the above average White American population Counties may be making the kinds of public health and health care investments that the rest of US Counties should consider.

Counties where White Americans are 85% or more of the population (orange)

5-7-20 #2 white

County median income bands, with roughly 80 million US residents living in each, rank US Counties from high-to-low median income. The highest income US Counties have median income above $73,000. The bottom 25% of Counties have median income below $38,000.  The lowest median income Counties are smaller, and range in size from 692 to 424,000 residents.  Currently, only 0.1 deaths occur in the lowest median income Counties that have above average White American populations for every single covid-19 death experienced in other low median income Counties.  The lowest median income above average White American Counties have been healthiest during the covid-19 crisis.

Counties where White Americans are 85% or more of the population (orange) — ranked by County Median Income band

5-7-20 #11 white by income

The relatively few covid-19 deaths in above average White American Counties must be made up for elsewhere.  Next we’ll consider Counties with either above average Hispanic American or African American populations.  (Note: some Counties qualify in more than one ethnic segmentation of the US population.)

Above average Hispanic American population Counties account for 39% of the US population but 2.1 deaths for every single covid-19 death occurring in the rest of the US.  The previously mentioned imbalance in “B” County covid-19 deaths appears in above average Hispanic American populations where 2 deaths occur for every single death in other “B” Counties.  When New York state is excluded the results are similar with 1.2 of every covid-19 death occurring in “B” Counties with above average Hispanic American populations.

Counties where Hispanic Americans are 18% or more of the population (orange)

5-7-20 #3 hisp

When Counties are ranked from high-to-low median income the rate of covid-19 deaths is consistently higher than their population share would suggest.  Even among America’s top median income Counties, those with above average Hispanic American populations experience 1.4 for every single covid-19 deaths that occurs in America’s other high-income Counties.

Counties where Hispanic Americans are 18% or more of the population (orange) — ranked by County Median Income band

5-7-20 #11 hispanic by income

Counties with above average African American populations account for 37% of US residents. All things being equal, one would expect the same percent of covid-19 deaths, however, 2.6 deaths occur in above average African American population Counties for every single death in other US Counties.  The chart below is a mirror image of what was illustrated above for above average White American Counties where fewer, not larger, than expected covid-19 deaths occur.  When New York state is excluded from above average African American population Counties 1.8 deaths occur in for every single death that occurs in other US Counties.

Counties where African Americans are 13% or more of the population (orange)

5-7-20 #4 black

One may hope that above average African American population Counties with the highest median incomes might at least be safer but they are not.  They experience 1.8 deaths for every single death that occurs in the rest of US’s highest median income counties — 80% more covid-19 deaths. But, most alarmingly, 7.0 deaths occur in above average African American population Counties that have the lowest median incomes for every single death that occurs in all other Counties in the lowest median income band – that is 700% more deaths.

Counties where African Americans are 13% or more of the population (orange) — ranked by County Median Income band

5-7-20 #11 black by income

It is worth noting that race and ethnicity are defined separately on the US Census so that many Hispanic residents can also report their race. As a result, when population and death figures are combined from this analysis they may end up being larger than the national totals.

Counties with above average White American populations are faring well through the covid-19 crisis. Counties with above average Hispanic American populations, especially “B” Counties, and all above average African American population Counties require more investment if they are to reverse the current trend in covid-19 deaths and improve their public health and health care services in the future.

Analysis Approach

This is one of a series of analyses that relate covid-19 deaths to key County health and demographic characteristics, such as cancer death rate or percent of adults over 60. Each analysis focuses on Counties where the health or demographic characteristic is higher than some “average”. The “average” is defined here as the characteristic’s incidence or percentage — think of average cancer death rates or percentage of adults over 60 — that accounts for roughly 1/3rd of the US population when the Counties that qualify are pooled together. Covid-19 deaths are then totaled for Counties that come in above or below the “average”.

The resulting 1/3 versus 2/3 population comparison then gets written as a ratio, such as, 0.5 : 1. Of course, the ratios change as the data changes. For example, the ratio helps us compare the covid-19 death rates for Counties with above average percent adults over age 60 to those with below average percent adults over 60. All things being equal, the ratio of covid-19 death rates should match the ratio of the County populations. Typically it won’t: “For Counties with 1 million or more residents, the covid-19 death rate ratio of those with above average percent adults over age 60 to that of the remaining Counties with 1 million or more residents is 1.2 : 1. In other words, residents living in Counties with above average percent adults age over 60 have a 20% higher chance of a covid-19 death than residents living in the remaining Counties that have 1 million+ population. They are younger.”

Here’s an example. First, the covid-19 death rate is computed by the covid-19 deaths divided by population. For the entire US, as of today, the covid-19 death rate is 0.02% = 80,140 / 325,719,178, meaning 2 out of 10,000 residents have died. Next, we create the covid-19 death rate ratio by comparing the death rates of the Counties with and without the characteristic. In this example, the 1.2 : 1 ratio reported for “A” Counties with above average percent adults age over 60 is the ratio of covid-19 death rates of the with the characteristic and those without, or, in this case, 1.2 = (3,522 /7,585,831) / (33,043/ 85,069,756). The 1.2 represents the numerator, or counties with the characteristic, and the 1 represents the denominator, or the counties without the characteristic.

“A” Counties where Adult Age 60+ is 23%+ of County population

Side-by-side bar charts illustrate how Counties are faring in the covid-19 crises when grouped by size of population and median income and split by those above and below the characteristic’s average.

Update on Past Analyses

Above Average Minority Counties. US counties where populations of African American or Hispanic are 50% or above the US averages — that is, 19%+ for African American or 27%+ for Hispanic — have 43% of the US population and account for 52% of US covid-19 deaths. We note that D Counties with above average Minority populations also have very large covid-19 death rates (bottom orange bar at right), though this is a small number overall. Note this chart will be replaced with the current analysis charts next week.

Counties with 19%+ African American or 27%+ Hispanic (orange) or not (blue) by County Size

5-7-20 #6 min

Age 60+. 35% of the US population live in older Counties defined as above average Age 60+ populations (orange bars on the left graph). Since 8 of 10 covid-19 deaths are adults 65 years and older Counties with above average Age 60+ populations, especially C and D Counties, are vulnerable in the covid crisis.

Counties where Adult Age 60+ is (orange) or is not (blue) 23%+ of County population by County Size

5-7-20 #7 60

Heart Disease. It may seem like Heart Disease is a “city problem”, and it is, even when New York state is removed most covid-19 deaths are found in counties with above average Heart Disease deaths. Populations in rural C and D Counties with above average Heart Disease deaths also have relatively more covid-19 deaths.  These Counties also have older age populations.

Counties where Heart Disease causes 0.2% or more population deaths (orange) or not (blue)

5-7-20 #8 heart

Cancer. The above average Cancer death Counties (orange bars in the charts) account for 38% of the US population and 27% of covid-19 deaths. The chart shows that above average Cancer deaths trends with covid-19 deaths. When New York state is removed, however, the share of covid-19 deaths exceed the population percentage of above average Cancer death Counties among the 38 remaining “A” counties.  These Counties also have older age populations.

Counties where Cancer causes 0.19% or more population deaths (orange) or not (blue) by County Size

5-7-20 #9 cancer

More on County Groups, more on the US Excluding NY

3,142 US Counties are geographic and administrative districts for which lots of statistics are gathered, including covid-19 deaths. Each County can be characterized as urban and rural, and they range in population from 10.2 million (LA County, CA) down to 88 (Kalawao County, HI) residents. In these reports, comparisons are made between four groups of counties depending on population size and by median income bands, as indicated in the chart axes.

New York state accounts for 6% of the US population and disproportionately more covid-19 deaths, so it is useful to consider total US results excluding New York state. This is done in the chart below which shows a disproportionate majority of covid-19 deaths occurring suburban B Counties across the US but excluding New York State (second bar on right graph).

US Population and Covid-19 Deaths by County Size Group — Excluding NY State

5-7-20 #11 x-ny

What Next?

In future reports we’ll continue to examine county health statistics that might help put covid-19 statistics in helpful light.

Some other measures that may tell the developing covid-19 story:

  • How do Covid-19 deaths change when adjusted for the first date one was reported?
  • Do Counties’ previous flu or respiratory deaths tell us about County covid-19 death rates?
  • Do County median income or education tell us about the likelihood of covid-19 deaths?
  • Do results change if we use the exact density of County population instead of County Size?
  • Do average temperatures or levels of particulate matters matter?

If you have any questions about this analysis or the data, or if you have suggestions, please don’t hesitate to contact me.

-Pete

Sources: US CensusUSA Facts Covid-19IHME forecasts, US Cancer Deaths NIH 5 year average, CDC Handbook on Death Reporting, 2003New York TimesMedian IncomeCDC Wonder Detailed MortalityCDC Fine Particulate Matter 2003–2011

 

Covid and County Ethnicity

Acknowledgement

 

Takuo Aoyagi, an inventor of the pulse oximeter, an ingenious and indispensable medical device that measures oxygen in the blood and has become a staple of hospitals around the world, emerging in recent months as a key tool in the fight against the novel coronavirus, died April 18. He was 84.  Washington Post

This US County health analysis is dedicated to individuals who are the victims of the Covid-19 virus and to all the people working hard to keep our US Counties safe and nourished.   Takuo Aoyagi was 84.

 

Analysis Background — Data as of May 7, 2020

 

This analysis uses County statistics to report on public health outcomes relating to the covid-19 crisis.  The hope is that this information will help support investment decisions by County administrators to improve resident health services and public health.  This analysis uses County ethnicity statistics to learn about the impact of covid-19 across 3,142 US Counties.

 

The upshot on ethnicity is that Counties with above average White American populations have fared very well in the crisis.  While it is early still and this may change, the analysis suggests Counties with above average Hispanic American and African American populations require investment to improve their public health and health care services and reverse the current covid-19 trend.

 

The chart shows covid-19 deaths as a percentage of state population and the ethnic percentages for each state.  States are ranked by number of covid-19 deaths from low to high.

 

 

“Prosperity is more complex than just consumption.”  The covid crisis has proven that a healthy, safe workforce is required for a strong consumer society and is essential to a sustainable economy.  Can US economists begin to measure our markets and corporations by valuing what they create not just in terms of goods and services, but also in how much sustainable value they deliver in the form of improved health outcomes for our workforce and consumers?

 

 

Analysis of Covid-19 Deaths

 

This analysis reports covid-19 deaths against key County health and demographic statistics, such as ethnicity.  These characteristics are used to split Counties into two unequal-sized groups, with Counties having the characteristic accounting for roughly 40% of the US population and the balance of Counties without the characteristic accounting for about 60% of the US population.  County health and demographics, as well as covid-19 deaths, are then examined.  All things being equal, covid-19 death percentages should also split by the same 60/40 ratio, but of course, they do not.  Side-by-side bar charts illustrate what’s going on in each County group and for the County populations when split by those with and without the characteristic in question.

 

The pandemic fell early and heavily on 6 NY state counties.  This especially affects the results of the group of 44 large US metropolitan “A” Counties with 1 million residents or more.  They account for 28% of the US population and 47%, or about 33,000 of the 70,000 covid-19 deaths today.  When New York state is removed, the proportionate balance is restored: 1 million+ “A” Counties are 27% of the US population excluding New York state and report 27%, or about 12,000 of 45,000 covid-19 deaths.

 

However, on removing New York state, another imbalance comes to light for 228 “B” Counties having between 250,000 to 1 million residents.  They account for 36% of the US population and 55% of covid-19 deaths when New York state is removed.

 

To better understand these imbalances, this analysis goes back to all states and works backwards looking at the County characteristics associated with the fewest covid-19 deaths, beginning first with above average White American populations according to the Census.  In the graph below, US Counties are split into two groups with the results for above average White American population Counties highlighted in orange.  The above average white population Counties have 34% of US residents, yet they only account for 13%, or 9,000 of all covid-19 deaths.

 

The population percentage and covid-19 death percentage among above average White American population Counties are an imbalance.  The graphs show for largest to smallest County Sizes that each group of above average White American population Counties (orange bars) has a proportionately larger share of the US population percentage (left chart) than share of US covid-19 deaths (right chart).  If covid-19 public health and health care were equally available to all US Counties the left and right orange bars would be the same size, but there is an imbalance indicating above average White American Counties are safer.

 

Counties where White Americans are 85% or more of the population (orange) or not (blue)

 

The imbalance of covid-19 deaths in above average White American Counties must be made up in Counties with either above average Hispanic American or African American populations, or both.

 

Above average Hispanic American population Counties account for 39% of the US population.  They account for 57% of covid-19 deaths as of today.  The imbalance in “B” Counties is apparent for this split of the data as above average Hispanic American populations in “B” Counties have a disproportionately larger share of covid-19 deaths than their population would indicate.  When New York state is excluded this imbalance in more than expected covid-19 deaths in “B” Counties with above average Hispanic American populations appears.

 

Counties where Hispanic Americans are 18% or more of the population (orange) or not (blue)

 

 

 

Counties with above average African American populations account for 37% of US residents.  All things being equal, one would expect the same percent of covid-19 deaths but their share of covid-19 deaths is 60%.  The imbalance described here is the mirror image opposite of what was illustrated earlier for above average White American Counties.  Where Counties have above average Where American populations the share of covid-19 deaths is smaller than expected.  Where Counties have above average African American populations the share of covid-19 deaths is larger than expected.

 

Counties where African Americans are 13% or more of the population (orange) or not (blue)

 

It is worth noting that race and ethnicity are defined separately on the Census, so many Hispanic residents also report their race.  As a result, some end these numbers end up what we think of as the White American or African American totals.  This is the double counting in the Census data when working with ethnic populations that needs to be managed: Whites are 77%, Hispanic are 18%, African are 13% which comes to 108%.  Adjustments are not commonly made and would not change the outcome of this analysis in any case, so none have been attempted here.

 

Counties with above average White American populations are faring well through the covid crisis. Counties with above average Hispanic American populations, especially “B” Counties, and above average African American populations require more investment if they are to reverse the current trend in covid-19 deaths and improve their public health and health care services.

 

“Prosperity is more complex than just consumption.”  The covid crisis has proven that a healthy, safe workforce is required for a strong consumer society and is essential to a sustainable economy.  Can US economists begin to measure our markets and corporations by valuing what they create not just in terms of goods and services, but also by in how they create sustainable value delivered as improved health outcomes for our workforce and consumers?

 

Update on Forecasted Covid-19 Deaths

 

The University of Washington’s IMHE is the pre-eminent US source for estimates of future covid deaths.  It creates the “covid curves”.  Last week the IMHE raised the estimate to 134,000 deaths by August, 2020.  It had previously been 70,000 but US Counties already reported that many, so the IMHE update was obviously needed.  (Our leaders should learn from TV weather forecasters who call for rain even when there’s a relatively small chance.  We’d at least like to know – and I’m referring to County health officials here — about the worst case scenario.)

 

Estimated infections were added to the IMHE data repository and contribute to improvements in their latest models.  Newly available covid-19 test samples, stratified by US geography, help estimate infection rates across each state.  Currently, IMHE assumes the US has a total of 11 million covid-19 infections.  This is close to 10X more than the 1.2 million confirmed cases we see in today’s news.

 

That said, confirmed cases are difficult to make much use of because of how they are reported.  The same goes for number of tests administered.  Confirmed cases are positive test outcomes made at hospitals but there are a lot of “if’s” behind them: a) if the test is given to someone with symptoms b) if a trip was made to a test center c) if results were reported to county health commissioners, d) if there was no mistaken false negative, and e) if someone doesn’t have symptoms we’ll never know.  The IMHE uses testing per capita perhaps as a public health measure but not the raw number of confirmed cases in their covid-19 models.  They use estimated infections.  And, TV news would do well to report that there have been an estimated 11 million covid-19 infections to date.

 

The IMHE models continue to rely on factors that affect transmission rates including state school and business closures, travel restrictions and mobility, hospital resource availability, and social distancing measures.  (It would surprise me if IMHE does not use County health and census demographics in their models.  I would hope they do but don’t seen any indication so far.)  Some numbers to keep in mind:

 

  • Deaths per Capita 0.02% rising to 0.04% by August — 1 to 2 out of every 5,000 people
  • Deaths per Infection 0.53% (1 out of every 200 covid-19 infections die)

 

But as is evident in this chart, and in the discussion above, Deaths per Capita is a generalization.  The likely outcome from a covid-19 infection depends heavily on your County health and demographic characteristics – as well as your own health.

 

State Deaths per Capita and per Infection — States ranked from low to high number of covid deaths

 

The chart ranks states from low to high by number of covid-19 deaths, similar to the ethnicity graph shown in the beginning of this analysis.  The message is simple: if the number of infections grow then more people die.  That may be seem obvious, but some County administrators are making decisions as if this is not going to happen.  The IMHE model does account for ill-advised decisions when they see them.  And, in case you’re wondering, “immunity percent” does not enter into the IMHE equations.

 

 

Update on Past Analyses

 

Above Average Minority Counties.  US counties where populations of African American or Hispanic are 50% or above the US averages — that is, 19%+ for African American or 27%+ for Hispanic — have 43% of the US population and account for 52% of US covid-19 deaths.   We note that D Counties with above average Minority populations also have very large covid-19 death rates (bottom orange bar at right), though this is a small number overall.  Note this chart will be replaced with the current analysis charts next week.

 

Counties with 19%+ African American or 27%+ Hispanic (orange) or not (blue) by County Size

 

 

Age 60+.  35% of the US population live in older Counties with above average Age 60+ populations (orange bars on the left graph).  Since 8 of 10 covid-19 deaths are adults 65 years and older Counties with above average Age 60+ populations, especially C and D Counties, are vulnerable in the covid crisis.

 

Counties where Adult Age 60+ is (orange) or is not (blue) 23%+ of County population by County Size

 

 

Heart Disease.  It may seem like Heart Disease is a “city problem”, and it is, even when New York state is removed most covid-19 deaths are found in counties with above average Heart Disease deaths.  Populations in rural C and D Counties with above average Heart Disease deaths also have relatively more covid-19 deaths, but this is not related to these Counties have older age populations.

 

Counties where Heart Disease causes 0.2% or more population deaths (orange) or not (blue)

 

 

 

Cancer.  The above average Cancer death Counties (orange bars in the charts) account for 38% of the US population and 27% of covid-19 deaths.  The chart shows that above average Cancer deaths trends with covid-19 deaths.  When New York state is removed, however, the share of covid-19 deaths exceed the population percentage of above average Cancer death Counties among the 38 remaining “A” counties.  We’ve seen C and D Counties have older populations and this makes their populations vulnerable to both Cancer and covid-19.

 

Counties where Cancer causes 0.19% or more population deaths (orange) or not (blue) by County Size

 

 

More on County Groups, more on the US Excluding NY

 

3,142 US Counties are geographic and administrative districts for which lots of statistics are gathered, including covid-19 deaths.  Each County can be characterized as urban and rural, and they range in population from 2,600,000 down to 88 residents.  In these reports, comparisons are made between four groups of counties depending on population size as indicated in the chart axes.

 

 

New York state accounts for 6% of the US population and disproportionately more covid-19 deaths, so it is useful to look at total US results excluding New York state.  This is done in the chart below which shows the disproportionately large number and majority of covid-19 deaths occuring suburban B Counties (second bar on right graph).

 

US Population and Covid-19 Deaths by County Size Group – Excluding NY State

 

 

 

What Next?

 

In future reports we’ll continue to examine county health statistics that might help put covid-19 statistics in helpful light.

 

Some other measures that may tell the developing covid-19 story:

  • How do Covid-19 deaths change when adjusted for the first date one was reported?
  • Do Counties’ previous flu or respiratory deaths tell us about County covid-19 death rates?
  • Do County median income or education tell us about the likelihood of covid-19 deaths?
  • Do results change if we use the exact density of County population instead of County Size?
  • Do average temperatures or levels of particulate matters matter?

 

If you have any questions about this analysis or the data, or if you have suggestions, please don’t hesitate to contact me.

 

-Pete

 

Sources: US Census, USA Facts Covid-19, IHME forecasts, US Cancer Deaths NIH 5 year average, CDC Handbook on Death Reporting, 2003; New York Times; Median Income; CDC Wonder Detailed Mortality; CDC Fine Particulate Matter 2003-2011

Geotargeting in Programmatic

A study of digital campaigns show traffic targeting 10 DMAs mapped well to those geographies.  Exceptions are some datacenter traffic, and micro-targeting.  Zooming in, local “freckles” of hundreds of IP addresses were shown to have the same geo-location, and not the exact user GPS location — desktop, mobile, and tablet shared the same clusters of geo-locations.

Co-Author: Dr. Augustine Fou – Cybersecurity and Ad Fraud Researcher

Bots vs Humans on a Map

Geographic distribution of human and bot traffic in internet advertising is generally comparable when viewed on a map.  Mass fraud commited out of large datacenters is the one exception.  Close examination of the bot traffic does show some unlikely geographies but most is likely sourced from malware sitting on a system attached to a legitimate IP address location.

Co-Author: Dr. Augustine Fou – Cybersecurity and Ad Fraud Researcher

The New WMD’s

The New WMD’s

WMD? Is Cathy O’Neil creating a hysterical media frenzy so she can sell more copies of “Weapons of Math Destruction“, or is she sincerely onto something?

Well, if the word “math” in your book title, it’s best to do anything you can to get some other relevant attention! Anything! Foment hysteria? Well, worth a try… But we’re into something more than just fomenting hysteria here. It’s a radical book, but fair. And well written….

Predictive Models I’ve Built

While at Epsilon, from 2000-2016, and at Time Inc. prior to that, my primary use of data was to create analyses and predictive models for client applications.  I also created predictive models that became data products in their own right, many of which are still sold to Epsilon clients today.

Generally working with customer file data using SAS and R, I created the following types of modeling solutions:

  1. Likelihood to Pay on a Marketing Offer – Predict the likelihood of payment and non-payment using ensemble modeling techniques
  2. Response Optimization – Enable optimal business results in client marketing campaigns through response models
  3. Segmentation – Used nearest neighbor and hierarchical segmentation techniques to segment the population into categories, or personas
  4. Prospect Models – Predict prospects that look most likely to exhibit a specific behavior, such as, response, pay, or product purchase
  5. Cross-Sell Models – Response models using customer transactional data to predict likelihood to respond to a new product offer
  6. Product Optimization – Given the likelihood of response, predict which product will be most profitable to offer to an existing customer.
  7. Area-Level Data Imputation – Impute missing data values using geographic roll-ups to enable complete data set coverage
  8. Propensity Models – Use census-balanced weighting to model look-alikes to self-reported consumer behaviors
  9. Econometric Models – Predict estimates for household characteristics such as income, net worth, and home value, applied as census-balanced scores across the US market
  10. Area-level Models – Predict geographies with populations most likeyl to exhibit an interest or purchase behavior

 

Data I’ve Worked With

Most recently I worked on ad fraud detection using browser data.  Prior to that, I was responsible for Epsilon’s Analytical Data Assets team.  In this role I was tasked with analytic product development based on both Epsilon and 3rd party data.  While at Epsilon, from 2000-2016, my primary use of data has been to create analyses and predictive models for client applications.

I’ve worked with the following types of data:

  1. JSON Browser Data – Developed python code base to read and analyze browser session data containing indicators of potential human or ad fraud.
  2. Compiled Consumer Data – Household characteristics such as age, income, home ownership, and purchase transactions primarily on Epsilon’s TotalSource Plus file containing thousands of data points on 160MM US households.
  3. Consumer Self-Reported Data – Self-reported warranty and survey data relating to interests, purchases, and ailments (for example, interest in gardening, shops at Walmart, and suffers from diabetes), primarily from Epsilon’s Target Source database of over 40MM US households.
  4. Consumer Transactional Data – Bank data relating to credit and debit card line items summarized by merchant and merchant category by month, primarily Epsilon’s Market View data.
  5. Credit Data – Consumer credit line and credit inquiry data from Credit Bureaus
  6. Web Data – Email and digital data streams for web analytic analysis
  7. Syndicated Market Research Data – Census-balanced self-reported consumer research data, primarily the MRI syndicated survey research file.
  8. Business Firmagraphic Data – Characteristics and contacts associated with business entities
  9. Aggregated Credit Data – Credit bureau information aggregated at a zip+4 level
  10. US Census Data – Typically compiled and resold by companies such as, Epsilon and others
  11. Customer Relationship Management Data – Client specific transactional data in various industries (publishing, travel, financial).

About our logo

cropped-Wheel.pngIllustration of 10 congruent equilateral triangles that have the same center. Each triangle has been rotated 12° in relation to the one next to it. The outer vertices are connected with a smoother curve to form a circle. Hence, the circle is circumscribed about the triangles.  ref. http://etc.usf.edu/clipart/42900/42974/tri2_42974.htm  Copyright © 2004–2016 Florida Center for Instructional Technology. ClipArt ETC is a part of the Educational Technology Clearinghouse and is produced by the Florida Center for Instructional Technology, College of Education, University of South Florida.