More Good News: This Week in COVID-19 Data, Feb 18

All major indicators of COVID-19 transmission in the United States continue to fall rapidly. Weekly new cases have fallen from 1.7 million at the national peak in early January to fewer than 600,000 this week, and cases have declined in every state. As we’ve seen at many points in the pandemic, case numbers are changing most quickly, with hospitalizations and deaths declining after a delay: cases have been falling sharply for five weeks, hospitalizations for four, and deaths for two. In this week’s numbers from nursing homes and other long-term-care facilities, we are now seeing solid declines in deaths correlated with COVID-19 vaccinations of this most vulnerable population.

If we look at cartograms of the United States showing weekly new cases for the worst week in January and for the week ending yesterday, the drop in the number of cases in each state is startling. Although states are still reporting large numbers of cases, many more parts of the country show absolute levels much closer to what we saw before the most recent surge accelerated nationally in October.

COVID-19 hospitalizations, too, have fallen very sharply. Every region and sub-region of the US now shows substantial declines in the number of hospitalized COVID-19 patients. 

Despite the declines, the total number of people hospitalized is still very high—hospitalizations have yet to fall far enough to reach even the peaks of the two previous surges, which both rose to roughly 60 thousand hospitalizations.

Here’s what we can tell about cases and deaths among Latinx people  

Throughout the pandemic, the age-adjusted risk of contracting and dying from the coronavirus has been higher for people of color. With the easing of the huge outbreaks in Arizona and California, we reviewed the data states have reported for Latinx people, a group that saw disproportionate numbers of cases and deaths in those states. According to CDC data, Latinx people have been 1.3 times more likely to be infected and 2.3 times more likely to die throughout the pandemic in the US compare to white people. But the regional story is quite complex: The effects of the pandemic on Latin American populations have shown wide variation from urban Los Angeles to the rural South. In the West this winter, the huge outbreaks in Southern California and Arizona drove a massive spike in deaths among Latinx people, but in the Midwest and Northeast, even as deaths rose, Latinx deaths did not precisely follow. 

The differences by region and demographic group are obscured in the topline numbers and nationwide comparisons. Better state-level race and ethnicity data continues to be necessary to understand these differences in how the pandemic is playing out. 

Vaccines are finally showing up in the data, right where we most want to see them

According to the CDC, more than 16 million Americans have now received two doses of either the Pfizer-BioNTech or Moderna vaccine since late December. At least 1.8 million of these fully immunized people are residents or staff in nursing homes or other long-term-care facilities, and at least 900,000 of that 1.8 million are residents. More than 4 million residents and staff in long-term-care facilities have received at least one dose of vaccine. 

Like many people, we expected to see the effects of vaccine distribution show up first in long-term care facilities. But for weeks, we didn’t see much happening in the data. In retrospect, this makes sense: Many of the people dying of COVID-19 who were reported dead in early February probably contracted COVID-19 in December, given the time it takes for the disease to overcome a body’s defenses and the lags in data reporting pipelines. As each week passed, we watched and waited to see a clear signal that all those first and second doses were having an effect. 

During the clinical trials, all of the front-runner COVID-19 vaccines were 100 percent effective at preventing hospitalizations and deaths, even among the oldest participants. In the past two weeks, preliminary (not yet peer-reviewed) real-word data has started to come in, most recently from Israel, showing similar effectiveness in preventing severe illness and even reducing viral load—a finding that could indicate reduced viral transmission. During Wednesday’s White House press briefing, Dr. Fauci noted that this research has “very important implications from a public health standpoint for interfering and diminishing the dynamics of the outbreak.” 

Now, with the data through February 11, the shift is clear: Deaths are falling in long-term care facilities. Cases, too, have plummeted. As importantly, deaths in these facilities are declining not just in absolute numbers, but as a share of COVID-19 deaths in the whole country. In early January, the percentage of COVID-19 deaths associated with outbreaks in long-term care facilities decreased below 30 percent for the first time since we began collecting LTC data in May, and it has continued to decrease since. This change correlates strongly with mass vaccinations in these facilities. 

First, the absolute numbers: Excluding data from Missouri and New York—both of which dumped large numbers of un-dated deaths into their LTC data in the past month, confounding analysis of both their individual states and the national numbers—the number of deaths in long-term care facilities have fallen 50 percent from the peak of reported deaths in mid-January.

But deaths are clearly falling among all age groups in the US, so one could imagine that the declines in long-term-care facilities merely reflected this broader pattern. To test this theory, we looked at the share of deaths that were linked to long-term care facilities. Remember that these facilities are home to less than 1 percent of the US population. For months, the share of deaths linked to long-term care facilities had bounced between 30 and 40 percent. In mid-January, it was still 29 percent. Then, over the last three weeks, the share of deaths associated with LTC outbreaks began to decline. 

For the week ending February 11, long-term-care linked deaths represented just 18 percent of total reported deaths in the states where we have this data. One way to think about this is that the share of deaths associated with long-term-care facilities has been cut in half since early January.

Cause and effect has been very difficult to establish in the United States throughout the pandemic. Our national patchwork of datasets and policies has confounded many simple analyses that try to explain why cases or deaths are rising or falling. But this week, we have clear evidence that the vaccines are saving lives in exactly the places where we would expect to see their effects show up first. We are—at long last and after so many failures—beginning to protect the most vulnerable.

Data disruptions in Texas

This week, in Texas, a major crisis has unfolded. Record-cold temperatures locked up the state’s energy infrastructure, causing huge power outages. Local news reports suggest that this slowed or stopped vaccination distribution and administration in many parts of the state. Testing sites were also closed in some areas like Hidalgo County.

It’s not just operations, but also reporting that can be disrupted. Over the last year, we’ve seen that major storms can cause significant problems with the COVID-19 data that states report. Sometimes, these problems are easy to spot. A state will simply outright say: we cannot post data today. Other times, the wobbles a storm introduces into the data are not as visible. 

If people are having trouble moving around the city because of a storm, they are also less likely to seek out testing or even seek medical attention. Because authorities are still able to report something, the interruption is not immediately obvious. As we noted in last week’s update, the early February snowstorm in New York created data disruptions that artificially depressed COVID statistics for several days. Then, a week later, with normal data pipelines, the 7-day average bounced upward, making it seem as if the decline in tests and cases had reversed. But that was not the case in reality.

There are more important issues in the state of Texas right now, but it would not be surprising to see a similar dip and then rebound over the next couple of weeks. With alarm about the variants of concern running high, we caution that it might be difficult to distinguish a data artifact from a fast-moving outbreak with just a week of data.

Federal data resources

As the COVID Tracking Project winds up state data compilation, we are conducting a series of training sessions about how to use the federal datasets that are now available. People interested in learning how to access, preprocess, and use these datasets can sign up here. 

The CDC has also replaced its previous COVIDView page with a new COVID Data Tracker Weekly Review that provides “key visuals from the week and narrative interpretations” of pandemic data. It is simple and useful. The weeklies also provide a good starting point for understanding what is available in federal government data repositories.