Skip site navigation

Three weeks ago, we wrote that three critical measures of the COVID-19 pandemic in the United States—cases, hospitalizations, and deaths—were on the rise. This week, the story is more complicated, but includes some unqualified good news: nationwide, new cases dropped from 468,000 cases last week to 448,198 cases this week, the first week-over-week decrease since early June. Hospitalization numbers appear to be nearly flat this week, but this crucial and previously stable data point has suffered significant losses in quality and completeness, leaving the reality of COVID-19 hospitalizations unclear. More about that in a moment.

COVID-19 deaths are still rising, up 16% this week after several days of more than 1,000 death reports. This week, states reported 6,779 COVID-19 deaths.

Within individual outbreaks, the relationship between falling new cases and rising deaths is easy to explain: According to the CDC, it takes an average of about two weeks for a patient to progress from the onset of symptoms to death. The average lag between death and the reporting of a death is just over seven days, though our research suggests that different methods of reporting require different amounts of time.

In the national data, though, it took 27 days after cases began to rise in early June for deaths to start rising as well.

Why so long? The demographics of early-summer surges in new cases may have had something to do with it, as younger people are much less likely to die of COVID-19, though they are not exempt from long-term health effects. But the simplest explanation is that the United States does not have one outbreak of COVID-19: it has dozens, often including multiple discrete outbreaks within a single state. At the regional level, deaths continued to fall in the Northeast long after case growth had declined, which in the national data obscured the initial weeks of rising deaths in hotspot states in the South and West.

Before this week’s decrease, we saw five weeks of sharp increases in new cases. This is our third week of sharp rises in COVID-19 deaths. We can expect that the spike in cases from the last several weeks will continue to translate into rising deaths in the first half of August, though the absolute number of COVID-19 patients who will die of the disease in the next few weeks remains unknowable.

Testing growth has also slowed in the last two weeks, which makes it difficult to understand how much of the decline in new cases reflects declining outbreaks or backlogged reporting. One data point we would usually look to as a measure of outbreak severity, current COVID-19 hospitalizations, has become much less useful.

Hospitalizations in (and out of) focus

Over the last two weeks, the number of people hospitalized with COVID-19 in the United States has approached previous peak numbers from the spring outbreaks concentrated in the Northeast. At the same time, changes in federal reporting methods have placed a significant strain on hospitals throughout the country and, at least temporarily, made the data hospitals report to states incomplete and unreliable.

Earlier this month, the Trump administration issued a directive requiring all US hospitals to switch from reporting COVID-19 data to the Centers for Disease Control and Prevention to reporting substantially more data directly to Health and Human Services. Hospitals—many in regions with major outbreaks—were given two days’ notice to complete these changes, and many have not yet been able to comply fully with the new directive.

Although our data comes directly from states and territories, not from a federal source, the hospitalization data we rely on has degraded significantly in quality since the new reporting directive went into effect. Many hospitals have struggled to report their data to anyone under the strain imposed by the changeover. As a result, some states have disclosed incomplete data reporting from hospitals—or taken their hospital data offline entirely—while many others posted declining numbers without explanation.

Our June 26 analysis of the COVID-19 hospitalization data reported from the new HHS system reveals that the federally maintained hospitalization figures are, on average, about 24% higher than those reported by the states themselves. It appears many states are reporting hospitalization data using more restrictive criteria than the HHS data, which includes people hospitalized with suspected cases of COVID-19, but we are also looking into other possible causes for these discrepancies. If state reporting stabilizes again and we can pinpoint the reasons for the differences between state and federal hospitalization numbers, both datasets will be useful in understanding outbreaks of COVID-19—but we’re not there yet.

Taken together, the gaps and uncertainties in the previously stable hospitalization data mean that this crucial indicator has become much less useful for understanding the true severity of COVID-19 outbreaks. This is particularly damaging at a moment when more than 20 states are in what the White House coronavirus task force considers the “red zone” of COVID-19 outbreaks—and when many states and communities are making difficult decisions about whether to send children back to classrooms this fall. 

A better outlook in hotspot states

A look at the four states hit earliest and most intensely in the surge of COVID-19 cases that began in late June reveals more good news this week. Arizona, California, Florida, and Texas have all continued to report declines in new cases, indicating that they are past the peaks they experienced in early July. Arizona, which peaked first, is now seeing COVID-19 deaths declining. We hope to see the same in California, Florida, and Texas in the next two to three weeks.

As cases decline in the states with the largest current outbreaks, many smaller surges are appearing in other states across the South and Midwest. In Georgia, the sharp rise in cases is reminiscent of Arizona’s in June, and Georgia’s seven-day rolling average for new cases—a metric less subject to reporting artifacts than the daily number—has already come within 100 cases of Arizona’s peak seven-day average of 3,844.

Tennessee is also posting a sharp increase in new cases, and although Missouri, Ohio, and Oklahoma are reporting less than 2,000 new cases per day, their case and hospital numbers—incomplete as the latter may be—are concerning. We will be watching these states closely over the coming weeks.

To keep up to date on our work, follow us on Twitter and join our low-frequency email list.