Case Counts, Infections, and Deaths in the US and Western Europe

Are diagnostic-positive case counts an accurate method of tracking the infection rate? First, the UK:

After the initial surge in the spring, covid contagion dropped consistently and dramatically. From June through August the dx-positive case counts¬† and deaths remained at a relatively low and steady rate. Then, starting around the beginning of September, case counts started climbing rapidly, tripling over the last three weeks. The death count remained low and steady until September 14; now deaths too are on the rise. So there’s been a two-week lag between new cases and deaths, which makes sense given the typical progression of the disease.

From Sept. 14-21, the UK averaged 22 covid-related deaths per day. Deaths are a reliable lagging indicator of infections. Using a rough fatality rate of 1 percent, those 22 daily deaths in the third week of September would have been produced by 22/.01 = 2,200 new daily infections during the first week of September. And when you look at the daily test-positive case counts, they averaged around 2,000 new cases per day during that one-week interval. In other words, covid testing in the UK is identifying nearly every new infection.

Over the past week the UK averaged 3,900 new cases daily; if the pattern persists, then around the second week of October these new cases will likely be producing around 40 covid-related deaths per day.

Next, France:

After the spring surge, cases and deaths dropped dramatically, remaining low and steady from May through July. In early August case counts began to climb rapidly. Death counts too began to increase around the middle of August. Over the past week France has averaged 55 covid deaths per day. Divide by .01 = an estimated 5,500 new infections during the first week of September. The average dx test-positive case count in France during that week was 6,800 — more than the number estimated from death rate.

A plausible explanation is that the recent surge has infected a disproportionate share of younger people, who have a lower than average fatality rate. So, e.g., if the people who’ve been infected in France during the current surge are on average 2 years younger than the national median age, then the recent French fatality rate would drop from 1% to 0.8%. Using this age adjustment, the lagging estimate of infections for early September would equal the case count.

During the past week France has averaged 10,100 new cases per day, which projects to a daily death rate of 82 for the second week of October.


It’s generally the same pattern as UK and France: spring surge, rapid decrease, long low plateau of new cases and deaths. In Spain the increase in case counts began in early July, followed by increasing death counts. Over the past week Spain has averaged 116 daily deaths; using the 1% fatality rate, there would have beeen around 11,600 new daily infections during the first week of September. The daily dx-positive case count in Spain for that week was 8,500 — like UK and France, pretty close.

Over the past week, Spain’s daily case rate has dropped to 4,600 per day. Forecasting deaths from cases, Spain is projected to have a death count of around 63 per day during the second week of October.

Now the USA:

In contrast to these three Western European nations, the relationship between dx test-positives and deaths is more tenuous in the US. Death rates peaked in April, dropped in May, plateaued in June, increased in July, and have plateaued again since August. Case rates followed a different pattern: peak and plateau from April through the middle of June, consistent rapid rise through the end of July, slow decline through the August, plateau during September.

During the past week an average of 770 people died daily of covid in the US. Using the national age-adjusted mortality rate of 0.85, and using the two-week lag between infections and death characteristic of the European data, the current death tally would have been produced by an average of around 90,000 new cases daily during the first week of September. During that 7-day interval the daily US case count was 46,000. I.e., using death as lagging indicator of infection, the US’s diagnostic testing is identifying only half of the new infections. And since there is no consistent 2-week time lag between case rate fluctuations and death rate fluctuations, case counts aren’t a reliable indicator of current infections or future deaths.

Finally, a comparison of projected death rates. The UK, France, and Spain have all experienced abrupt increases in infections, with rising death tolls lagging two weeks behind. The US covid toll, meanwhile, is either plateaued or in slow decline. Let’s forecast that, for the second week of October, the US death count will drop slightly, to 730 per day. How does that rate compare with the projected death counts of the three European countries, adjusting for national population?

  • US = 730 deaths/328 million population = 2.2 deaths per million
  • UK = 40 deaths/67 million population = 0.6 deaths per million
  • France = 82 deaths/67 million population = 1.2 deaths per million
  • Spain = 63 deaths/47 million population = 1.3 deaths per million

While the US is still experiencing more covid deaths than the UK, France, and Spain, the gap has narrowed considerably over the past couple of months. It remains to be seen whether they’ll be able to stem the tide and reverse the flow, as they did so effectively in the spring.

Racial/Ethnic Disparity in US Covid Infection Rates

Death rates. Here are the updated, officially counted covid deaths per 100K population by race/ethnicity in the United States:

  • Overall = 61 deaths per 100K
  • White = 39 deaths per 100K
  • Black = 93 deaths per 100K
  • Hispanic = 60 deaths per 100K

Fatality as lagging indicator of infection. Official death counts probably underestimate the number of fatalities attributable to covid infection. However, the cumulative death rate is the most accurate indicator of the population’s infection rate. Using the IHME’s estimated US covid fatality rate of 0.85% for the US , here are the imputed cumulative infection rates by race/ethnicity:

  • Overall = 61/(.0085 x 1000) = 7.2 percent infected
  • White = 39/(.0085 x 1000) = 4.5 percent infected
  • Black = 93/(.0085 x 1000) = 10.9 percent infected
  • Hispanic = 60/(.0085 x 1000) = 7.1 percent infected

Age adjusted fatality rates. The 0.85% fatality rate estimate is an overall average for the US population. Age, however, is a significant predictor of covid mortality; e.g., per the CDC, 65-74 year olds infected with covid are 90 times more likely to die of the disease than are infected 18-29 year olds. All else equal, populations with a higher median age will have a higher mortality rate. Here is median age by race/ethnicity of the US subpopulations:

  • Overall = 38 years
  • White = 44 years
  • Black = 34 years
  • Hispanic = 30 years

Based on the CDC’s data, as well as national seroprevalence studies conducted in various countries, each additional year of population median age results in about a 10% higher mortality rate. Conversely, younger populations have a 10% per year lower mortality rate compared to the overall average. The year-by-year age adjustments are geometric rather than linear, compounded annually. So, calculating the age-adjusted covid fatality rates by race/ethnicity:

  • Overall = 0.85% fatality rate
  • White = 0.85% x 1.16) = 1.51% age-adjusted fatality rate = 1.78 times the overall rate
  • Black =¬†.0085/1.14 = 0.58% age-adjusted fatality rate = 0.68 times the overall rate
  • Hispanic = .0085/1.18 = 0.40% age-adjusted fatality rate = 0.47 times the overall rate

Age-adjusted infection rates. Combine these intermediate calculations to arrive at cumulative US age-adjusted covid infection rates by race and ethnicity:

    • Overall = 7.2 percent infected
    • White = 4.5/1.78 = 2.5 percent infected
    • Black = 10.9/0.68 = 16.0 percent infected
    • Hispanic = 7.1/0.47 = 15.1 percent infected