COVID-19 Data Dispatch

Tag: Contact tracing

  • We need better contact tracing data

    We need better contact tracing data

    Last week, New York Times reporter Apoorva Mandavilli questioned the scientific basis for recent public health guidance against small gatherings. Politicians and public health officials are telling us to cancel Thanksgiving dinners, she writes, but it’s difficult to find data that actually demonstrate a link between small gatherings and COVID-19 transmission.

    Mandavilli acknowledges that the majority of states do not collect or report detailed information on how their residents became infected with COVID-19. This type of information would come from contact tracing, in which public health workers call up COVID-19 patients to ask about their activities and close contacts. Contact tracing has been notoriously lacking in the U.S. due to limited resources and cultural pushback.

    I came to a similar conclusion about the contact tracing data deficiency in October, when I investigated the practice in this newsletter. Still, the data that are publicly available suggest that larger gatherings and congregate facilities are still the major sources of virus spread, as Mandavilli writes:

    But in states where a breakdown is available, long-term care facilities, food processing plants, prisons, health care settings, and restaurants and bars are still the leading sources of spread, the data suggest.

    The piece faced criticism for potentially undermining important guidances about the holidays. Even CDC Director Robert Redfield pushed back against it. When asked about this story on Fox News, he said, “From the data that we have, that the real driver now of this epidemic is not the public square… It’s really being driven by household gatherings.”

    For me, this distinction between Mandavilli’s story and Redfield’s statement underscores that either a.) the CDC has access to some contact tracing data that the rest of us don’t, or b.) nobody has access to complete contact tracing data, and public health officials are communicating the conclusions that seem more politically salient. I don’t love either outcome!

    The volunteer project Test and Trace compiles information on each state’s contact tracing efforts. Check out how your state is faring, and if you’re unsatisfied, contact your local politicians and ask them to do better.

  • Your Thanksgiving could be a superspreading event

    Your Thanksgiving could be a superspreading event

    Between 10% and 20% of people infected with COVID-19 are responsible for 80% of the virus’ spread.

    You might have seen this statistic before, but take a second to think about what it means. Imagine that one unlucky person at a crowded restaurant, infected with the novel coronavirus but not yet symptomatic enough that she has noticed, spreads the virus to ten others. Meanwhile, her ten coworkers, who all contracted the virus at the same time as she did, do not spread the virus to anyone else at all. This type of dispersal—what epidemiologists call a large k value—introduces a level of randomness to COVID-19 spread which makes it difficult to track and protect against.

    But scientists are learning to better understand COVID-19 spread by keeping tabs on those instances where one person infects many, which they call superspreading events. While research continues about the underlying biology driving who is infectious and who isn’t, investigating the events in which people get infected can help us better understand how to protect ourselves and our communities.

    For more thorough explanations into the science of superspreading, I’d recommend you read Christie Aschwanden in Scientific American, Zeynep Tufekci in The Atlantic, or Martin Enserink, Kai Kupferschmidt, and Nirja Desai with an incredible series of scrolly visualizations in Science Mag.

    Here, I’m focusing on the data around these events: how we identify them, what the data tell us so far, and why we should keep them in mind as temperatures drop and cases rise.

    How do we find superspreading events?

    The CDC defines a superspreading event as one in which “a few persons infect a larger number of secondary persons with whom they have contact.” This leaves a lot of room for interpretation, as different researchers have different thresholds for determining how many people must be involved. Depending on who you talk to, anything from a 500-person rally to your extended family’s Fourth of July gathering might fit the definition.

    So, when you see a sensationalist article claiming that some event caused superspreading, it is important to consider what definition was being used and how the given event was identified as one that fits.

    There are three ways superspreading events can be identified:

    • Continuous tracking of an outbreak site: This is the easiest way to find superspreading. You have a place with a lot of people—say, a nursing home, a prison, a college campus—and you watch how many cases erupt over time. This may be an easier means of finding events because local administrations or public authorities are conducting regular testing and making data public; meanwhile, the sites themselves may have large groups of people living and working in close quarters, which is a prime environment for virus spread. Scientists count these sites as superspreading events even though they are not “events” in the way we usually think of the word because this type of long-term superspreading can have the biggest impact. California’s San Quentin State Prison, for example, was ordered to reduce its prison population after over 2,000 prisoners tested positive.
    • Contact tracing: This strategy, in which public health officials contact individuals who test positive and ask them about their contacts to find other infected individuals, has not taken off in the U.S. as it has in other countries, which makes it harder for us to identify superspreading events. It works like this: if contact tracers find that one new case is a teacher at an elementary school, for example, they can call other teachers and school administrators to find out which other cases are connected to that location. Japan has famously avoided widespread lockdowns by employing a “cluster-busting” strategy in which officials contact-trace backwards from new cases in order to find how those people got infected, then tell other people at the spreading events to isolate. Scientists in Europe and the U.S. are now promoting this approach as our cases surge.
    • Scientific studies: This strategy of superspreading identification is perhaps the least consistent, but it gets the most press. Epidemiologists may use publicly available case data, cell phone tracking data, or other information to look for patterns in new cases after major events. Such studies may draw attention, as a working paper on the Sturgis, South Dakota motorcycle rally did in September, but it can be difficult for scientists to investigate events when they don’t have access to data on precisely which cases are connected and how. The Sturgis paper was criticized for making estimates based on unreliable data. A similar new paper on the COVID-19 impact of Trump rallies is currently undergoing peer review.

    Where do superspreading events happen?

    Full-screen dashboard link.

    Independent researcher Koen Swinkels started a database to answer this question. The database is compiled from media reports, scientific papers, and public health dashboards, as well as volunteer reports. (You can submit an event through a form on the database’s site.)

    As of November 7, the database includes about 1,600 superspreading events, ranging from churches to dinner parties to meat processing plants. About 1,100 of these events took place in the U.S. For those American events, the most common superspreading settings by far are prisons (50,000 cases), rehabilitation/medical centers (27,000 cases), nursing homes (26,700 cases), meat processing plants (13,900 cases), and other medical centers (12,200 cases). Parts of the Northeast, West Coast, and South are heavily represented in the database, while other areas of the country have yet to see significant superspreading events logged.

    You can explore the map pictured above, as well as a bar chart which organizes superspreading settings by their COVID-19 case numbers, in a pair of interactive Tableau visualizations which I built based on this database.

    Swinkels emphasized in an email to me that the database is not at all representative of all COVID-19 superspreading events which have taken place, in America or around the world. “Hundreds of millions of people have been infected with SARS-CoV-2, while we have only about 200,000 cases linked to the 1,600 superspreading events in our database,” he said.

    He and other members of the team, including professors at the London School of Hygiene and Tropical Medicine, are currently compiling events from the most easily available public sources, which he admits is not a comprehensive strategy. Swinkels also noted that the events identified by public sources may be biased by where public health officials direct their focus, which can lead to settings that were closed in the spring or are now operating under restrictions being left out of this database and of superspreading research more broadly. The database is also biased by the team’s English-language familiarity; they are looking to find more events described in non-English language publications.

    What does this mean for the holidays?

    This newsletter topic was inspired by a reader question I got last week: Ross asked me how post-election gatherings and holiday celebrations might contribute to COVID-19 spikes.

    The evidence so far suggests that protests have not yet been a major cause of COVID-19 spikes. But “so far” is doing a lot of work in that sentence. While protests are generally outside and see high mask compliance, Swinkels explained, they tend to involve talking and singing in close contact, and instances of transportation and socialization around a protest might pose more risk. (Imagine, for example, shouting “FUCK TRUMP!” in a crowd of 500 with two friends, going to an outdoor bar together afterward, then each taking the bus home to three different parts of the city. That’s a lot of risk for one evening.)

    More research on protests is necessary to truly determine how much risk they might pose to the communities around them. And, as contact tracing apparatuses in different parts of the country scale up—slowly but surely—such research will get easier.

    Holiday celebrations, on the other hand, are a definitive cause for concern. These celebrations almost always occur indoors, involve talking and eating, and bring people together from disparate locations. Superspreader events also almost always occur indoors, may involve loud talking, and expand COVID-19 risk from one area to another. There’s a reason that Dr. Anthony Fauci’s daughters are not traveling home for Thanksgiving.

    I asked Koen what he’d learned from compiling and comparing hundreds of superspreader events. “Knowing more about where and when superspreading events occur can help you to avoid high-risk situations and live more freely in low-risk situations,” he said. He listed several key risk factors: indoors, poor ventilation, many people, close together, prolonged periods, loud vocalization (such as singing or shouting), and cold, dry air.

    He also highlighted the importance of understanding aerosol transmission. The six feet rule we’ve all come to know and flaunt is based on the dispersal of larger air particles, which don’t travel far from an infected person. But aerosols, which are smaller particles, are able to travel further and stay in the air longer—especially in indoor, poorly ventilated spaces. You can sit all the way across the room from Grandma while you eat, but if masks are off and all the windows are closed, it won’t make much difference. This FAQ document by aerosol scientists provides much more detail about how this type of COVID-19 spread works.

    I’m not going to tell you to avoid traveling for the holidays; I’m not a public health expert, I don’t have that authority. But I can give you this fact: your Thanksgiving could be a superspreading event. So could the train you take to get to your relatives’ house. So could the bar where you go for outdoor drinks a few days before traveling. In order to make it through this winter, we must all be aware of our risks and adjust our behavior accordingly.

  • Contact tracing: Too little, too late, no public data

    Contact tracing: Too little, too late, no public data

    Most states are not ready to find and trace all of their new COVID-19 cases as the country heads into a new wave of outbreaks. Screenshot via Test and Trace, taken on October 18.

    On October 1, a little over two weeks ago, I received an email from New York Governor Andrew Cuomo’s office.

    The email invited me to download a new COVID-19 phone application, developed by the New York State Department of Health along with Google and Apple. The app, called COVID Alert NY, is intended to help New Yorkers contact trace themselves. (Side note: I am not entirely sure how Cuomo’s office got my email, but I suspect it has something to do with the complaints I left about his budget back in June.)

    Here’s how Cuomo’s office describes the app:

    COVID Alert NY is New York State’s official Exposure Notification App. This is a free smartphone app available to anyone 18+ who lives and/or works in New York. The app uses Bluetooth technology—not location data—to quickly alert users if they have been in close contact with someone who has tested positive for COVID-19. Once alerted, users can quickly protect themselves and others by self-quarantining, contacting their physician and getting tested.

    The app is intended to fit into New York’s contact tracing efforts by automatically informing app users that they have been exposed to COVID-19 and prompting them to take the necessary precautions. It also features a sypmtom checker, which asks users to note if they have exhibited a fever, cough, or other common COVID-19 symptoms, and a page with the latest case and testing data for every county in New York.

    Contact tracing, or the practice of limiting disease spread by personally informing people that they have been exposed, has been a major method for controlling COVID-19 spread in other countries, such as South Korea. But in the U.S. the strategy is—like every other part of our nation’s COVID-19 response—incredibly patchwork. We have no national contact tracing app, much less a national contact tracing workforce, leaving states to set up these systems on their own.

    Back in May, I researched and wrote an article for Stacker about this problem. I compared contact tracing targets, calculated by researchers at George Washington University, with the actual numbers of contact tracers employed in every state, compiled by the public health data project Test and Trace. GWU’s estimates started at a baseline 15 contact tracers per 100,000 people, then were adjusted based on COVID-19 trends in every state. Now, this story should be seen as a historical snapshot (the summer’s Sun Belt outbreaks hadn’t yet started when I wrote it), but it is telling to scroll through and see that, even several months into America’s COVID-19 outbreak, the majority of states had tiny fractions of the contact tracing workforces they needed to effectively trace new cases. New York, for example, had a reported 575 contact tracers employed, compared to an estimated need of over 15,000 contact tracers.

    Today, many states are doing better. New York is up to 9,600 contact tracers, according to Test and Trace’s latset counts, and has planned to hire thousands more. This state, along with Massachusetts, New Hampshire, New Jersey, Connecticut, Vermont, and Washington D.C., has received high marks from Test and Trace’s scoring system, with 5 to 15 tracers employed for every new positive COVID-19 case. But all of these high-scoring states are in the Northeast, where COVID-19 outbreaks peaked in the spring. The Midwestern states currently seeing spikes, such as Wisconsin and Missouri, all rank far lower on their preparedness to trace new cases. (See the screenshot above.)

    Meanwhile, actual data on the efficacy of these contact tracers are difficult to come by. To continue using New York as an example: since the application’s release on October 1, New York’s Department of Health has not released any data on how many people have downloaded the application, much less how many positive cases have been logged or how many contacts have been traced. Data have neither been mentioned in Cuomo’s press releases nor have they appeared on the state’s COVID-19 dashboard.

    According to tech website 9to5Mac, as of October 1, 11 states had registered contact tracing apps with Google and Apple’s exposure notification technology. These states include Alabama, Arizona, Delaware, Nevada, New Jersey, North Carolina, North Dakota, Pennsylvania, Virginia, and Wyoming, as well as New York. Six more states have apps in development.

    A brief analysis by yours truly found that, of those 11 states with contact tracing apps, only four post contact tracing data: Delaware, New Jersey, North Dakota, and Wyoming. Delaware and New Jersey both have dedicated data pages detailing the share of COVID-19 cases which have successfully participated in the state’s contact tracing efforts (57% and 71%, respectively). North Dakota and Wyoming both post statistics on their cases’ source of COVID-19 exposure, including such categories as “contact with a known case,” “community spread,” and “travel”; these data must be sourced from contact tracing investigations. 11.1% of North Dakota’s cases and 27.1% of Wyoming’s cases have an exposure source listed as “unknown” or “under investigation,” as of October 18. Meanwhile, Pennsylvania and North Carolina have both posted statistics on their contact tracing workforces, but no data on the results of these workforces’ efforts.

    Other states without registered apps may also be posting contact tracing data. But it is still a notable discrepancy that, among the states that have systematic contact tracing technology, tracing results are lacking. Compare these states to South Korea, which at the height of its outbreak publicly posted demographic information and travel histories for individual COVID-19 cases in alerts to surrounding communities. South Korea’s approach has faced criticism, however, for reporting private information about people who tested positive.

    And that brings me to the biggest weakness for American contact tracing: lack of public trust. Americans, more than residents of other nations, tend to be concerned about personal privacy and, as a result, are more wary of speaking to strangers on the phone or using an application that sends their data to the government, even if all those data are anonymized. Olga Khazan explained this issue in an article for The Atlantic, published in late August:

    Still, contract tracing depends on trust, and many Americans don’t trust the government enough to give up their contacts or follow quarantine orders. Of the 121 agencies Reuters surveyed, more than three dozen said they had been hindered by peoples’ failure to answer their phone or provide their contacts. About half of the people whom contact tracers call don’t answer the phone, because they don’t want to talk with government representatives, Anthony Fauci, the director of the National Institute of Allergy and Infectious Diseases, said during a June news conference.

    Black and Hispanic or Latino communities are particularly likely to distrust the government and avoid contact tracers’ calls. This attitude makes sense, given how both America’s government and medical systems are inexorably tied to racist histories. But for the public tracers hoping to help these communities—which have been disproportionately impacted COVID-19—it’s another barrier to stopping the virus’ spread.

    Even I, as someone who understands more about the need for contact tracing than the average American, am wary about using New York’s COVID Alert app. The app asks me to turn on both Bluetooth and location data, and even though COVID Alert purports to be anonymous, Twitter, Instagram, and other applications have made no such promises. So far, I have been using the application when I go to the park, grocery shop, or ride the subway, but for the vast majority of my days it sits dormant on my phone.

    And of course, I have to wonder: where was this app in March, when the city shut down and my neighborhood filled with ambulance sirens? Like most other parts of America’s COVID-19 response, contact tracing has been scattered and difficult to evaluate, but the data we do have indicate that most states are doing too little, too late.

    more contact tracing content
  • I am once again asking: why are journalists doing this?

    I am once again asking: why are journalists doing this?

    President Trump and the First Lady tested positive for COVID-19 in the early morning on Friday, October 2. As I draft this newsletter on Sunday morning, at least 15 other people connected to the President have tested positive, ranging from Bill Stepien, Trump’s campaign manager, to New York Times Washington correspondent Michael Shear.

    You might expect me to source this number and these names from a federal public health agency, which is conducting all of these tests and making their results public. Not in this pandemic! My source is, of course, a dashboard compiled by volunteer journalists and science communicators.

    This dashboard, called the COVID-19 At The White House Contact Tracker, is attempting to trace over 200 contacts in connection with the President and his staff. The team behind it includes Benjy Renton, independent reporter on COVID-19 in higher education, Peter Walker, data visualization lead at the COVID Tracking Project, and Jesse O’Shea, MD, infectious disease expert at Emory University.

    The Contact Tracker is an incredible public service. In its current form, the dashboard lists 235 White House contacts who should get tested for COVID-19, along with their positions, test results (if known), symptoms (if they test positive), and the date of their most recent test. You can also view the data as a timeline, based on each person’s last contact with the President, and as a map based on the Rose Garden ceremony, the debate, and two other potential spreading events.

    It is not surprising, after months of poor data reporting from the federal government that, instead of the CDC or the HHS, the best source of data on this high-profile outbreak is—as Dr. O’Shea puts it— “three awesome dudes [contact tracing] from our homes.” But it is worth emphasizing.

    https://twitter.com/JesseOSheaMD/status/1312644102027177985

    What are federal public health agencies prioritizing right now, you might ask? The HHS is planning a $300 million-plus ad campaign with the goal of “defeating despair” about the coronavirus. And this money came out of the CDC’s budget. I was planning to devote a bigger section to this campaign before COVID-19 hit the White House, but instead, I will direct you to an excellent (and terrifying) POLITICO feature on the subject. Dan Diamond also discusses his investigation of the campaign on his podcast, POLITICO’s Pulse Check.