The District of Columbia is sharing more information than ever about the spread of the novel coronavirus. With a click of a button, you can pore over reams of data about what sort of people are testing positive, in which neighborhoods, after participating in what activities, and more.
With those details, however, come new questions about how the city is calculating the numbers and whether the data is guiding sound public policy.
For example, contact-tracing data published Monday indicated that dining out remains a common activity among those who go on to test positive for the virus, while fewer than 1 percent of all coronavirus patients had been to a gym or played a sport before they got sick. On the same day, Mayor Muriel E. Bowser (D) banned high school sports and affirmed her recent decision to keep indoor dining open but shut down indoor fitness classes.
Bowser’s actions have been criticized from many corners, from gym owners and recreational basketball players to members of the D.C. Council. In a meeting this past week, council member Elissa Silverman (I-At Large) said the difference between the data and the restrictions has confused residents.
“They see .6 percent on the exposure data for gyms and fitness,” she said. “And they’re asking, ‘Well, why are we closing gyms to group fitness activities when, for example, dining out is at 19 percent?’ ”
Experts say contact-tracing interviews — in which a person who has tested positive answers questions about their activities while contagious — are meant to track infection from one individual to the next, not to draw broad conclusions about the safety of certain activities.
But contact tracers are struggling because the virus moves faster than test results, and many people won’t reveal their contacts. Meanwhile, elected officials trying to balance health risks with economic damage have seized on the data for another purpose: to argue for and against shutdown decisions.
“What are we trying to do? We’re trying to find contacts who may have been infected so that you can ask them to quarantine,” said Emily Gurley, a Johns Hopkins epidemiologist whose course on contact tracing is used to train workers across the country. “In terms of interpreting those data as ‘here’s where these people were infected,’ it’s problematic. And it’s problematic because of very basic concepts in epidemiology.”
Health Director LaQuandra Nesbitt said contact-tracing data, “while it can be a rich data set, does have limitations. . . . We can’t be put in a position as a health authority to only make recommendations for things that show up in contact-tracing data.”
Much of the newly released D.C. data was prompted by questions from reporters and a recent report from D.C. Auditor Kathy Patterson that called for increased transparency.
The District shares more than many states: The Atlantic’s Covid Tracking Project gave the District an A+ for providing so much information. Virginia also scored an A+, while Maryland got an A grade, losing points for not detailing ventilator availability or the preexisting health conditions of patients. Other jurisdictions fared far worse, with three given Ds for offering no information about the racial or ethnic identity of patients.
Still, experts have identified systematic flaws in the District’s reporting, which since Monday has included the types of settings where “outbreaks” have been identified, and graphs detailing the answers that coronavirus patients gave to contact tracers.
The outbreak list was quickly criticized by education advocates and members of the council because it identifies an “outbreak” as two cases within a two-week period linked to one establishment — even if the cases weren’t linked to each other. Advocates say the criteria unfairly tilted the data toward identifying schools as sources of outbreaks, and Silverman tussled with Nesbitt in a meeting Wednesday about why the health department won’t share more information about those outbreaks.
The second data set, a page of graphs and tables from the contact-tracing task force, was revealing — up to a point.
It showed a downward trend in the size of social gatherings from August through late fall but growth in how many of those gatherings were indoors. The data also showed notable consistency in the types of activities that coronavirus patients say they partook in before they got sick.
Restaurant dining, social events and employment outside the house were the most common activities, followed by travel. Far less common: salon visits, worship services, sports and gym workouts.
But experts cautioned against interpreting this data to mean that restaurants are unsafe or that gyms are low risk because the report does not show how common each activity is for the general population or how many people engaged in a particular activity without contracting the virus.
“You’d like to be able to say, ‘What are the high-risk activities that you can close?’ But does this say that?” asked Jaime Ashander, a computational biologist in the District.
Such broad conclusions about categories of activities aren’t what contact tracing is meant for, experts say. The goal is “isolating people while they’re still infectious and preventing further infection,” said Alex Perkins, a professor of epidemiology at Notre Dame university, who has identified problems with the District’s data. “But that’s most effective when transmission is much lower than it is now.”
The D.C. auditor’s report says contact tracing is most effective within three days of a test, but it often takes longer than that for test results to come back. Once they have results, contact tracers in the District report that they call 99 percent of people who test positive within 24 hours.
About 75 percent of those patients agree to an interview. But while they might answer questions about whether they went to work or a social gathering, about 60 percent refuse to provide names of their close contacts, which is key to halting spread.
The combination of widespread infection and lack of cooperation means the District’s contact tracers have fallen far short of their most important goal. Among all people in the city who test positive, no more than 10 percent had already been identified as a close contact of someone else with the virus. To reopen the city, the health department has said, that number will need to rise above 60 percent.
Resolve to Save Lives, a public health program that makes recommendations on coronavirus data, has praised the District for disclosing that metric. Among the 50 states, only Oklahoma does so, the D.C. auditor’s report said.
Experts say D.C. has enough contact tracers to call every patient but needs other interventions to curb the virus so that contact tracing can work, perhaps including further restrictions on indoor gatherings.
“If simultaneously you’re able to push with other measures to push down the base case rate, then [contact tracing] can be more and more effective,” said Noam Ross, a computational disease modeler who has advised New York state and several cities on their coronavirus responses.
Gurley said the practice can still work even in widespread outbreaks. “Those programs still reduce transmission, and they reduce cases counts,” she said.
Other metrics in the city’s abundant data also have raised eyebrows among experts.
Ashander questioned why the city’s rate of transmission, commonly labeled Rt, has not risen in parallel with the steep rise in actual cases. “There’s a disconnect there,” he said.
When Perkins examined the data, he criticized the District’s choice of a model that converts facts about the virus into a daily number. “I can’t recall seeing a [Rt] number that fluctuates quite this much” in other jurisdictions, he said.
D.C. Health spokeswoman Alison Reeves said the city excludes cases that seem likely to have originated outside the District from its Rt calculation to give a clearer picture of the rate of transmission within the city. That means any positive case from someone who recently traveled outside the region doesn’t count in this model.
She also defended how the department calculates the seven-day average test positivity rate, a metric that drew questions from data scientist Ryan Stahlin.
In most cities, a seven-day average would mean that the city adds up all the tests conducted for seven days, then adds up all the positive results, then reports what percentage are positive, said Stahlin, who with a friend created DCcovid.com to showcase the city’s daily data in a readable way.
But the District figures out what percentage of tests are positive on one day — and then takes the average of seven days’ percentages in a row. Stahlin called this method misleading and “incredibly dumb.”
“The fact that the entire city’s reopening plan is based on these metrics, I think it’s worth defining in precise terms,” he said.
Reeves said other jurisdictions, including the state of Indiana, use the same method.
“This measure is more sensitive than the approach of averaging over the total tests in a 7-day period and can help identify potentially important fluctuations that may happen during a specific week,” she wrote in an email. “If one day of the week has a high positivity rate, that should be incorporated into our measure so that we can understand why a higher percentage of positive tests came from a specific day.”
All the scientists said they were grateful that the District was publishing the data at all.
The D.C. auditor recommended that the city publish additional data, such as the percentage of cases not linked to any other known case (an indicator of rampant disease spread that is less stringent than the quarantined contacts measure), the trend in hospital patients who present with covid-19 symptoms, the time it takes between a person’s coronavirus test and that same person entering isolation, the number of cases among health-care workers and more.
On Thursday, the city listened once again, publishing more charts that the auditor had asked for. They showed the results of contact tracers’ studies of mask-wearing, reporting that more women than men wear masks and that mask-wearing is least common in Wards 7 and 8.
Bowser said the information might lead to more public education campaigns, targeting those who need them most.