Why Changing COVID-19 Demographics in the US Make Death Trends Harder to Understand

On June 18, new daily COVID-19 cases in the US exceeded 27,000 for the first time since May 8, and stayed over 27,000 for the next four days. This was the first five-day run of new cases above that number since April 22-26. The 32,325 new cases reported on June 20 were the most since May 1, when just over 33,000 cases were reported. After plateauing in May, the trend is heading in the wrong direction.

Deaths, on the other hand, continue to decline: from June 18-22, fewer than 700 people in the US died each day, a first for a five-day period since March. The disconnect between case and death trends is striking; in Brazil and India, the two other large countries reporting a rapid increase in infections, deaths have been rising in recent weeks as well.

Should we expect that trend to plateau or move upwards? The lag times from infection to death, and death to reporting, mean it’s a possibility. But there’s also a lot we don’t know about this new phase of the pandemic, meaning trends in new cases and deaths may not correlate.

What we know: the death lag

According to the CDC, the average period from symptom onset to death is about two weeks. The average lag between death and the reporting of a death is just over seven days.

New daily positive cases only began to exceed the plateau of the previous two weeks around June 18-19, which means that an increase in deaths as a result of the rise in new cases would not be expected to show up until July.

This may be one reason why the US deaths trend currently differs from Brazil and India. In those, cases have risen consistently; the current surge in the US comes after a decline and a plateau. 

But there’s also evidence new cases are increasingly concentrated in age groups at lower risk of death from COVID-19.

What we only sort of know: the changing demographics of the disease

Some anecdotal and statistical evidence suggests that the average age of people with COVID-19 is declining, which complicates expectations that deaths will increase in step with new cases. On June 16, for instance, Texas Gov. Greg Abbott said that a majority of people testing positive in three counties were under 30, which he said “typically results from people going to bars”; Dr. David Persse, public health authority of the Houston Health Department, told The Texas Tribune that “it is my current theory that elder persons have become more vigilant in taking precaution.” Florida Gov. Ron DeSantis has also said that infections are starting to skew younger.

In the Dallas-Fort Worth region of Abbott’s state, the University of Texas Southwestern reports that the age distribution of positive COVID-19 tests has shifted dramatically from March to June, with a peak under 30 years old. Hospitalizations and ICU admissions in the DFW region have also shifted younger if less dramatically; 50% of hospitalized patients are under 50, as are 30% of ICU patients.

In California in mid-May, three-quarters of all COVID-19 cases were split evenly between residents 18-34, 35-49, and 50-64, at almost exactly 25% for each group. By June 13, the 18-34 group represented a third of cases, while those 50-64 fell to 19%. Over a similar timeframe in Florida, the median age of people testing positive for COVID-19 fell from 54 to 35.

Untangling this shift in age groups from increased testing is a challenge. With additional testing available, more people in lower-risk populations are likely being tested now than when tests were being strictly rationed to severe cases. As businesses reopen across the country, workplace testing may also increase case numbers for the working-age population, and perhaps particularly younger-skewing service workers. Testing, however, is not universally adequate; local and regional spikes in cases are putting pressure on testing infrastructure. The Upshot reports that the testing capacity situation is “acute” in Arizona. In Florida, The COVID Tracking Project’s data indicates testing has actually slowed by 10% in the last two weeks.

In a Twitter thread, University of Florida biostatistics professor Dr. Natalie Dean offers three possible explanations for why the median age of cases might be falling, and what data signals we should look for. 

1. If it’s simply a matter of more testing, hospitalizations should not increase, and test positivity should decline or hold steady. In the South and West, positivity rates appear to be rising, but regional numbers can mask very different state trends. In Texas, Florida, and Arizona, test positivity and cases are both rising; in California, by contrast, new cases are way up but the positivity rate has remained at five percent in June, and in Georgia the positivity rate is up just two percentage points while testing is up.

2. If “elderly people are more cautious,” then cases, test positivity, and hospitalizations should decline. In the Northeast, tests are way up, positivity is way down, and new cases are flat. In New Jersey, new daily cases are down to 10% of April peaks, and hospitalizations have dropped precipitously. New daily cases in Connecticut have been in the double digits for the past couple weeks, and hospitalizations are down to 124 as of June 23 from over a thousand in mid-May.

3. If younger people are less cautious—or if they’re more exposed as young service workers return to their jobs—cases, test positivity, and hospitalizations should rise. This is happening in Texas and Arizona. Cases and test positivity are up in Florida; statewide hospital data was only available for a few days in May before the state removed it from public view, but in Miami-Dade County, hospitalizations increased from 601 to 776 from June 9 to June 22.

Dean ultimately concludes that it’s slightly too early to tell, but the real answer is likely a combination of all three, and that better age-stratified data is needed. High-quality samples exist for the country as a whole, but as we’ve seen, trends vary greatly from state to state and even city to city. The United States is a big country that needs a lot of detailed data.

The best source for nationwide data on testing and positivity by age is the CDC’s COVID-NET, a hospital surveillance network that serves as a sample—the data is very detailed, but it’s from only 250 hospitals spread across 14 states. COVID-NET data is best summarized in the CDC’s weekly COVIDView reports. Looking at that data, Dr. Trevor Bedford of the Fred Hutchinson Cancer Research Center finds hints of Dean’s explanations: a substantial decline in positivity among tests in the 50-plus age group from early April to early June, and a slower decline among younger age groups, with signs of a plateau. He also found a very small increase in cases in the below-50 age group in that same time. (Disclosure: Both Dean and Bedford are members of The COVID Tracking Project’s advisory board.)

What changing age demographics have to do with deaths

In areas where younger adults are driving new infections, we might not see deaths spike until infections overflow into more vulnerable populations. “If what is happening are outbreaks in young people, it seems likely that these young people will go on to transmit to others in their communities,” Dean writes in an email. “This spillover would cause a subsequent rise in cases among older people, followed by a lagged rise in deaths.” She points to a pattern in Florida in which new cases in the 0-44 age group began climbing gradually in early-mid May, echoed by a smaller climb in the 45+ age group in late May-early June. Cases in the first age group began increasing rapidly around the beginning of June, a worrisome portent for the more vulnerable 45+ age group.

[Ed. note: Dr. Dean was also interviewed on the Brian Lehrer show today about the changing demographics of COVID-19 infections and other aspects of the pandemic and the US public health response.]

Race and ethnicity may also affect death rates among younger Americans

If case counts are truly rising among younger Americans, another variable in the death toll will be which younger Americans contract COVID-19. A recent working paper from the Harvard Center for Population and Development Studies, looking at deaths from February 1 through May 20, found that Black people with COVID-19 in the 25-34 age group had a mortality rate 7.3 times that of non-Hispanic white people. The disparity was nine times higher in the 35-44 age group… and then actually falls back to 6.9 times higher for the 49-54 group. The mortality rate ratios for Hispanic people compared to non-Hispanic white people was also elevated: 5.5, 7.9, and 5.8 for those same age groups.

But as we wrote here on June 4, race and ethnicity data is a good example of why better, more consistent state-level data is needed—48 states are reporting it for cases, and just 43 for deaths, and only data for white and Black people is being reported consistently across the country. The authors of the working paper on disparities in mortality by age note, for example, that “Poor quality of AIAN [American Indians/Alaska Natives] mortality and population data likely means the estimated excesses are underestimates.” Until that data improves, its absence introduces yet another lag in our understanding of COVID-19 and mortality.