Daily COVID-19 Briefing: Thursday


Top news, reports and insights for today:

  1. Daily deadline summaries for Thursday:
  • U.S. daily coronavirus case count crosses 50,000 setting a single day record as some states reverse course on reopening and hospitals were hit by a surge of patients (Wall Street Journal)
  • Experts suggest a plastic face shield, if properly fitted and warn, has advantages over cloth masks for use by the public and should be used more widely (JAMA)
  • The American Academy of Pediatrics issues guidance in favor of reopening schools in the fall. Guidance stresses that children do not appear to be driving the epidemic, are less likely to be infected than other respiratory illnesses and that in-person education is better than online teaching (American Academy of Pediatrics)
  • Italian researchers studied nearly all residents of one town before and after lock-down. Main results were that 40% of those infected never developed symptoms. None of the children in the study became infected. Those with and without symptoms did not differ on viral load. Half of those infected had cleared the virus 2 weeks later. The lockdown was very effective at halting transmission (Ars Technica)
  1. U.S. enters new dangerous phase, smashing previous daily high record new cases, decline in deaths stops
     I have said for months that a return of widespread community transmission with Rt>1 was inevitable by July 1 given that states reopened recklessly. However, the ferocity of the resurgence in cases has caught me and many experts off guard. I apologize for showing you 4 different graphs for today’s briefing, but the story they tell is chilling. The first graph is our usual epidemic curve of new reported lab-confirmed COVID-19 cases by day. A record high 52,554 new cases were reported yesterday, the first day with more than 45,000 cases and 16,000 more than the peak set in late April. For the first time since the epidemic, the U.S. reported more than 300,000 new cases in one week. In just 7 days, the U.S. reported more infections than were accumulated from the start of the epidemic until April 4. The U.S. now has more than 2.6 million reported cases, 26% of all cases on the planet and twice as many as Brazil, still ranked second.
    Deaths flatten: The second graph shows daily deaths. This remains a mystery. The 7-day moving average has been falling since Memorial day. In the last few days, that decline looks to have stopped. Yesterday, Arizona, a state on the front edge of the case surge, reported a spike in deaths with 88, a rise of 100% from the previous day. Arizona, Iowa, Oklahoma, South Dakota, Kentucky, Texas, Virginia, and Rhode Island, all have seen deaths rise by 20% or more in the last week. I believe we will see deaths rising sharply in the states where cases first started surging.
    Cases increased last week in all but 3 U.S. states: The next graph shows the weekly growth factors by state. Nationally, the U.S. had it’s first week with 300,000 new cases. Astonishingly, the growth factors for last week compared to the week before are >1 in all U.S. states except Rhode Island, New Jersey and the nation’s capital. Weekly cases rose by 50% or more in Alaska, Idaho, Montana, Kansas, Florida, Georgia, Louisiana, Mississippi, and Tennessee.
    13 States reported more than 15 new daily cases per 100,000 population last week. The last graph shows the rate of new daily cases last week. Arizona continues to be the most extreme, now reporting more than 47 new cases per 100,000 every day. That is the highest rate we have seen anywhere at any point. In the South, 9 of 13 states are above this mark, which is 3-times higher than the 5 cases per day per 100,000 benchmark for low transmission. Florida, Georgia, Louisiana, Mississippi, South Carolina and Texas are all over 20. Only 1 Northeastern state is over 5.
    The bottom line: The U.S. epidemic is out of control with no end in sight.
  1. U.S. capturing about 2/3 of the true number of COVID-19 deaths
    We already know that deaths from COVID-19 are undercounted for a variety of reasons. We also know that an accurate tally of the death toll is very important. As I have emphasized here before, our best tool for understanding the true mortality picture is to examine deaths from all causes using the best available data and comparing the numbers from 2020 to previous yearly averages. This allows us to capture excess deaths above what is expected in a “normal” month taking seasonal variation into account. It’s not perfect, but it is closer to the truth than our current count. Previous studies using similar methods have found that the ascertainment rate (% of true COVID-19 deaths that are counted) is about 50%. Over time, we hope that will improve. Yesterday, a new study was published by my former colleague and friend demographer Steven Woolf and in JAMA that adds important new information. They used a state-of-the art model to estimate total excess deaths that are probably COVID-19 related. This includes cases where COVID-19 is mentioned on the death certificate, but also more indirect causes. They specifically examine rises in deaths above temporal trends in past years that may reflect those who died because they didn’t seek care for existing problems, had exacerbations of chronic diseases, or may have succumbed to secondary distress (e.g., drug overdoses). Their main finding is that in the 8-week period from March 1 through April 25, about 65% of the excess deaths likely related to COVID-19 were actually counted as such. That means, the remaining 35% of deaths were not counted. The 5 states with the most COVID-19 deaths experienced large proportional increases in deaths from non-respiratory underlying causes including diabetes (96% increase), heart disease (89%), Alzheimer disease (64%) and cerebrovascular disease (35%). New York City, which reported the highest number of COVID-19 deaths, saw a 398% rise in the number of heart disease deaths. The graph below, taken from the paper, shows that in the 5 hardest hit states, a significant jump in deaths can be seen in March and April of this year for heart disease, diabetes. Smaller but still substantial increases happened for stroke and Alzheimer disease as well. Diabetes deaths rose from 40% to more than doubling depending on the state.
    The bottom line: The true number of COVID-19 deaths is probably about 1/3 higher than our current numbers suggest. If true, the U.S. now has an estimated 177,000 deaths rather than the official count of 131,000. Significant numbers of people are dying of non-respiratory manifestations of SARS-CoV-2 and are not being counted. States vary widely in the accuracy of their counts.
From Woolf et al, JAMA Published online July 1, 2020. doi:10.1001/jama.2020.11787

Daily COVID-19 Briefing: Friday


Top news, reports and insights for today:

  1. Daily deadline summaries for Friday:
  • Florida just reported nearly 9,000 new daily cases, a record high, bringing the state’s total to 122,960. This is almost 4,000 more cases than the 5,004 reported yesterday (CNN)
  • Both Florida and Texas moved to close bars on Friday after cases exploded in those states. Florida’s Governor left it to the Secretary of Department of Business and Professional Regulation to tweet the announcement. The Texas governor followed suit reinstating some restrictions on dine-in restaurants (NBC News)
  • Top doc “Uncle Tony” Anthony Fauci said today that the White House coronavirus task force is considering the need for a new testing strategy, admitting that the current approach is not working (CNN)
  • It’s more clear than ever that medical workers should be using N95 respirators, not just regular masks. While production has ramped up, shortages remain. N95 masks were designed to be single use. Researchers have now shown an effective method of decontaminating N95 masks using steam and a microwave in just 3 minutes (FastCompany)
  1. Zooming in on weekly growth-factors: key trends in deaths and cases by state
    Every detective has to decide which clues to pay attention to and which to ignore. A good detective will often pay attention to details others miss. This is just as true for disease detectives. While most people are watching case and death totals by day, I watch 1-week growth factors. If you have read my blog you may recall that these are important metrics to gauge recent trends. They tell us how fast the epidemic is moving in a state right now. Below are 7-day growth factors for cases and deaths by state, showing the ratio of cases/deaths in the last seven days and the 7 days before that. For example, 1.5 means there were 50% more cases/deaths last week compared to the week before. In contrast, 0.6 means there were 40% fewer cases/deaths last week. We know that deaths keep falling even as cases are surging. That’s a mystery. What do these growth-factors tell us?
    For cases, outbreaks have slowed in all Northeast states except Delaware (1.22) and Pennsylvania (1.21) where cases are rising. Numbers are too small in Vermont. That, by the way was the only state in the U.S. with fewer than 50 new cases last week. In contrast, growth factors are >1 (cases are rising week-over-week) in all western states, all southern states except Alabama and 9 of 13 mid-west states. While everyone is watching headline states like Arizona, Florida and Texas, the bigger picture is that cases rose by 10% or more in an astonishing 32 states. Idaho spiked the most with 1,122 new cases last week, up from 441 the week before. Arizona (+19,587), California (+34,472), Montana (+148), Nevada (+2,783), Michigan (+1,688), Missouri (+2,518), Florida (+28,092), Georgia (+10,183), Mississippi (+3,875), and Texas (+32,066) all saw weekly cases jump 50% or more.
    Deaths were more volatile despite the overall falling trend. Delaware stands out after reporting 43 deaths last week, more than triple the previous 7 days. Arizona (+219), Kansas (+14), Alabama (+79), and South Carolina (+70) all saw deaths rise by 50% or more last week. Falling deaths in states where cases are surging still begs for an explanation.
    Bottom line: The virus is spreading fast in 32 states, not just the few in the headlines, with trends indicating a return to exponential growth in some. Deaths remain a mystery. It may be the calm before the storm. Few expected the U.S. to be back to the April peak as we head into the heart of summer.
  1. Dark times in the sunshine state: a deep dive in Florida
    Florida is a bellwether state, worthy of close attention for many reasons. It has played a key role in the national epidemic as a hub of early transmission and a touch-off point for national spread after spring break. It is a state that has been under scrutiny for it’s slow actions, confusing messages and challenges in data collection and reporting. Perhaps of biggest concern is that Florida is home to a high concentration of the most vulnerable of our citizens: older people. One fifth of it’s population is over 65 and there are about 700 nursing homes with more than 80,000 beds. It is a treasure trove of America’s grandparents. What happens in Florida will disproportionately contribute to the overall burden of death and suffering in America’s outbreak.
     Florida is our patient, and the patient is sick. As numerous news outlets like CNN are now reporting, Florida is a top candidate for the next hot spot or epicenter as numerous sun-belt states battle the front edge of the U.S. outbreak. New record cases are being set almost daily with no apparent end in sight. In desperation, the governor closed bars this evening. Let’s order a battery of tests and have a look at what’s going on. The gallery below contains 8 figures that represent those test results. What do we learn:
    1. Graph 1: This is the overall epidemic curve for Florida (from STATNews COVID-19 Tracker) showing daily cases. It shows what all the fuss is about. Around June 6, the curve switches from linear growth to exponential growth in cases. There were 60,000 state-wide cases then, that has doubled in 20 days. Today, a record high 8,942 new cases were reported, a rise of 79% over the previous day. The first case was reported 109 days ago.
    2. Graph 2: In states like Minnesota and Arizona, big surges can be blamed on a small number of counties that become white hot for a period. Lets see how wide-spread the problem is. First, we see Miami-Dade County, home of the highest number of cases in the state (28,664 cases and 935 deaths). We see the previous spike there around April 9 that eventually lead the state to shut things down. The shut-down worked and cases fell until Memorial day. Then the old peak was surpassed in just 7 days, and now the county is seeing more than 600 a day, with a 50% rise in the last week.
    3. Graph 3: Maybe the problem is just in the urban south. How about the other end of the state, way up in Nassau county, a suburban area north of Jacksonville. Looks like a carbon copy. First peak is clear in early April, general decline till end of May and then an even more severe spike starting June 10. Here the new peak is twice as large as the April peak. It is the same pattern in a less populous county in the north. While there are only 136 total cases, the weekly new case total jumped 167%.
    4. Graph 4: Ok, what about the Northwest of the state? Leon County is home to the state capital. That part of the state hasn’t been on the radar yet. Now at 642 cases, the picture is fairly similar. First peak occurred later in mid-April at 15 daily cases. Here there was a smaller aftershock in the last two weeks of May, and then around June 15, the firestorm started, surging to the previous peak and then 70% higher in a bit over a week. In the state capital, average new cases have risen 280% week-over-week.
    5. Graph 5: What about the sleepy middle of the state? Perhaps lower density communities in smaller towns and big farms offers some insulation. Polk County covers a large area right in the center of Florida, home to just 603,000, it’s largest city is Lakeland; universities, farming, mining and the headquarters of Publix are key features. The story here is different and the same. Polk county was spared the first peak, never rising above 20 cases a day in April. But even in this low-population, largely agricultural county, a firestorm took off around June 9, with cases surging from 20 a day to over 130. Last week averaged 117 a day, compared to 50 the week before, a rise of 137%. There are some counties in Florida where surge hasn’t happened. The main point is that the surge is occurring all over the state, north to south, urban to suburban, with a similar pattern of exponential growth in June.
    6. Graph 6: Maybe this surge is all about more testing. This graph from rt.live shows positive daily tests and total testing volume for the state. Testing doubled from 20K per day to 40K around May 20. Since then, it’s been up and down but the overall trend has been steady. There is nothing in these numbers that would convince me that the June surge comes from testing increases. The curves for testing and cases are not the same shape. What is evident is that the surge in cases started a month after shelter-in-place orders were lifted.
    7. Graph 7: The effective Reproduction rate (or Rt) is an important test result. We now have pretty compelling models that allow us to see how the R-value changes over time. This graph also comes from rt.live, a model I and others have come to trust. A reminder, the Rt is the time-specific estimate of how many additional people each infected person subsequently sickens. Currently, Florida is estimated to be 1.4, which is the 5th highest in the country. The model shows that Rt was below 1.0 during the SAH phase, then started rising above 1, where it has been flat since Memorial day. A sustained Rt of 1.4 means each infected person passes the illness to 1.4 others. That is sufficient to drive nearly exponential growth. Without any epidemic control measures, Florida experienced an initial Rt of about 3.0. The control measures put in place knocked this down to below 1, where the virus might have then failed to thrive. Reopening the state however has brought the state back to a place where the virus is spreading half as fast as when nothing was being done.
    8. Graph 8: Models like this can be used to predict the near and intermediate term future. My current favorite model uses machine learning and combines many models and various data sources from COVID-19 Projections. This graph tells us where Florida is now, taking account not only of those testing positive, but also for asymptomatic infections that are never tested. This tells us that currently about 3% of Florida’s residents have been infected. That is light years away from the 55-70% attack rate needed for herd immunity to halt the epidemic. It estimates Florida is now actually acquiring around 11,000 true infections every day. Based on these parameters, the model predicts cases will peak again in Florida, but not until July 26, one month from now, at which point there will be an estimated 13,716 infections each day. Given the uncertainty in the model, that could be as high as 24,000 or as low as 4,700. Estimated current infections will rise from 150,000 to over 200,000 before this peak occurs.

Daily COVID-19 Briefing: Thursday

Top news, reports and insights for today:

  1. Daily headline summaries for Thursday:
  • Scientists at Yale have found that viral RNA found in sewage sludge predicts COVID-19 cases and hospital burden. This paper, while not yet peer-reviewed, suggests the potential power of poop in biosurveillance (medrxiv)
  • Asked why New York City was hit so hard by the first waves of COVID-19, experts point to 4 key factors: 1) large influx of travelers from Europe, 2) early reluctance to close subways and buses, 3) hesitation to issue stay-at-home orders by the Governor, 4) sending recovering patients back to nursing homes (Bloomberg)
  • U.K. scientists honing in on the identification of blood markers for mysterious Kawasaki-like multisystem inflammatory syndrome in children (MISC-R) (NBC News)
  1. A focus on growth factors by state: Hot spots in Kansas and the south
     As the epidemic has slowed in recent weeks, it makes sense to shift the time scale of how we track it. Daily ups and downs of cases and deaths at the national level have been choppy and unclear. We know that looking at numbers for the whole country can be misleading given that the outbreak has slowed in New York, Detroit and New Orleans, only to disseminate to smaller towns and rural areas. Let’s zoom out a bit and focus instead on longer term trends at the state and regional levels. The graphs below show 7-day week-over-week growth factors, which I have said previously can be a better metric to gauge the current situation compared to looking at the daily numbers. Growth factors are ratios, so looking at them gives a different perspective. In epidemiology, we get lots of practice with ratios. If we want to know the strength of a relationship between a risk factor and a disease, we can look at the absolute amount of disease attributed to the risk factor (also called Attributable risk) or the ratio of disease in those with and without the risk factor (also called Relative risk). There are definite plus-minus trade-offs in choosing which numbers to focus on, but in this case, we can improve the signal-to-noise ratio by looking at the ratio of cases/deaths last week to the cases/deaths the week before.
     The top graph shows the growth factors for new cases. States with values greater than 1 are more active, with positive growth in new cases week over week. What this shows is that all states with stable numbers in the northeast saw negative case growth except Maine. The midwest was a mixed bag with cases rising in 6 states and falling in 6. Cases grew 20% or more in Iowa, South Dakota and Wisconsin. Michigan is slowing fastest with just over half the cases last week as the week before. While little attention has been paid to the West, new cases rose in three states (California by 16%, Idaho by 24%, and Nevada by 23%). In my view, the biggest take-home message here is that cases grew last week substantially in the south, where 10 of 12 states saw cases rise. West Virginia more than doubled new cases adding 322 compared to the 141 the week before. Cases grew sizably also in Alabama (48%), Arkansas (66%) South Carolina (26%), Tennessee (28%) and Virginia (21%).
     The bottom graph covers growth in deaths. In the northeast, deaths rose in Maine (14%) and Rhode Island (54%). All western states saw deaths decrease last week. The fastest rise in new deaths occurred in Kansas where deaths nearly doubled week-over-week. In the south, new deaths increased in half the states, led by Alabama (24%), Arkansas (30%), Georgia (21%), South Carolina (31%) and Tennessee (22%).
  1. The state of testing: which states are ready to reopen?
     Today’s Top Pick of the Day article from Vox is based on interviews and discussions with a number of infectious disease experts who were asked to list the top criteria that should be used to decide when it is safe for states to relax outbreak control measures and reopen. I thought I would take my own deeper look into two of those criteria related to testing. The two benchmarks suggested are:
     1) New daily tests of at least 150 per 100,000 population to insure adequate testing capacity for spotting new infections and tracing their contacts.
     2) A test positivity rate of less than 5% indicating that the testing is no longer selective and that a wide net is being cast.
     You can read the paper to see what they conclude about these numbers. I created two new graphs using numbers from the COVID Tracking Project (the same source they used) but looking slightly differently. The top graph below ranks 50 states and DC from highest to lowest on new daily tests per 100,000 residents, averaged over the last month. As was true before, the testing regimes in the northeast are considerably more mature than in most southern and midwestern states. Rhode Island is still in the lead, testing an average of 176 per 100,000 per day and is the only state clearly above the threshold of 150 per day. New York, Massachussetts, Louisiana and Utah are the other states that are over 100 per day. North Dakota is the only midwestern state testing more than 75 a day, which may explain why cases are growing in that state. Twenty six states are now testing less than a third of the 150 benchmark. Arizona continues to have the lowest testing rate of all states by any metric. States below average are very likely under-estimating their current case prevalences, some quite dramatically.
     The bottom graph shows the test positivity rate (or TPR) for the last month. Since lower TPR is “better” in this case, the ranking of states is reversed. A total of 21 states have a TPR in the last month that is below 5%, while 16 states still have a TPR that is high (greater than 10%). Maryland, Washington DC, Delaware and Virginia constitute a block of Atlantic seaboard states that have ramped up their testing capacity (top graph) but still are seeing a large fraction of tests come back positive.
    The bottom line: No states qualify to reopen just using these two benchmarks as proposed in the VOX analysis. Testing remains widely discrepant across states. This shows what happens when there is a complete absence of national leadership and standards for testing. We remain largely and unsettlingly in the dark about this epidemic in the vast majority of U.S. states.