Hop­kins virus track­ers: Look be­yond the red dots

The Washington Post - - FRONT PAGE - BY KYLE SWEN­SON

She found her­self lis­ten­ing for the sirens all the time. Be­fore, they had been just back­ground noise, am­bu­lances reg­u­larly blast­ing through Beth Blauer’s neigh­bor­hood in Bal­ti­more to or from a nearby crit­i­cal-care fa­cil­ity for the elderly.

Now she knew ev­ery emer­gency trip from the nurs­ing home could turn up later as a statis­tic on her com­puter screen.

The novel coro­n­avirus was work­ing its way through the United States, and Blauer — along with dozens of col­leagues at Johns Hop­kins Univer­sity — was ac­tively track­ing its path.

“The sirens now feel dif­fer­ent,” she said re­cently. “They come with a dif­fer­ent flood of emo­tions.”

The noise out­side her win­dow was a tan­gi­ble re­minder of the lives hid­den in the maze of anony­mous data that had come to dom­i­nate her days.

Since launch­ing in Jan­uary, the univer­sity’s Coro­n­avirus Re­source Cen­ter has ex­ploded in scope and pop­u­lar­ity, gar­ner­ing mil­lions of page views and pop­ping up in news cov­er­age and daily con­ver­sa­tion. Through numbers, the tracker has told the story of what the virus is do­ing while the story is still un­fold­ing, of­fer­ing a nearly real-time pic­ture of its silent march across the globe.

But even as data has jumped to the fore­front of in­ter­na­tional

dis­cus­sions about the virus, the Johns Hop­kins team wres­tles with doubts about whether the numbers can truly cap­ture the scope of the pan­demic, and whether the pub­lic and pol­i­cy­mak­ers are fail­ing to ab­sorb the big pic­ture. They know what they are pro­duc­ing is not a high-res­o­lu­tion snap­shot of the pan­demic but a con­stantly shift­ing Etch a Sketch of the trail of covid-19, the dis­ease caused by the virus.

Case counts are con­sis­tently in­con­sis­tent. Re­port­ing prac­tices dif­fer from coun­try to coun­try, state to state, even county to county. If au­thor­i­ties fail to con­tex­tu­al­ize the virus with other fac­tors — such as Zip codes, race or Med­i­caid use — the hard­est-hit com­mu­ni­ties can go un­seen.

“Numbers in some ways in­still this sense of com­fort. But then on the other hand, they can be wrong,” said Lau­ren Gard­ner, the as­so­ciate pro­fes­sor at Johns Hop­kins’s Whit­ing School of En­gi­neer­ing who has spear­headed the global tracker since Day 1. “And they can be wrong for lots of dif­fer­ent rea­sons.”

For those look­ing closely, the Hop­kins project does lay out a clearly leg­i­ble story about Amer­i­can life to­day, one in­volv­ing eco­nomic in­equal­ity, racial dis­par­i­ties and poor ac­cess to health care. Many of the same is­sues flared up in street protests after the May 25 killing of Ge­orge Floyd in Min­neapo­lis po­lice cus­tody, an in­ci­dent that sparked civil un­rest and tem­po­rar­ily pushed the pan­demic off the coun­try’s front pages.

The tracker data can offer a bridge be­tween the two news cy­cles, those work­ing all hours to main­tain it say. And as cases again be­gin to spike in the South and West, the Johns Hop­kins project re­mains a key re­source for un­der­stand­ing the virus’s im­pact.

“This is the first time data has been such a cen­tral part of the nar­ra­tive,” said Blauer, the ex­ec­u­tive di­rec­tor of Johns Hop­kins Univer­sity’s Cen­ters for Civic Im­pact. “The hu­man con­nec­tion — I think we need more of that in the larger na­tional nar­ra­tive. It just feels like the com­pas­sion is get­ting lost.”

Launch­ing the project

It started over cof­fee.

The virus had clouded En­sheng Dong’s thoughts all Jan­uary. Ev­ery time the first-year PHD stu­dent called home to China’s Shanxi province, he heard about the sick­ness spread­ing from Wuhan.

Dong mod­eled out­breaks as part of his stud­ies. He had lived through China’s 2003 out­break of se­vere acute res­pi­ra­tory syn­drome, or SARS. Comb­ing the avail­able Chi­nese sta­tis­tics, he re­al­ized each data point could be a for­mer class­mate or neigh­bor or fam­ily mem­ber.

So when Gard­ner, his ad­viser at Hop­kins, sug­gested that Dong cre­ate a map to track the global reach of the in­fec­tion, he read­ily agreed.

“I wanted to use my ex­pe­ri­ence to col­lect data to show the pub­lic,” Dong said. “And the first mem­ber of the pub­lic was me.”

Gard­ner, an ex­pert in mod­el­ing in­fec­tious-dis­ease spread, ini­tially had a fairly mod­est vi­sion for the project. She knew dis­ease re­port­ing — or how au­thor­i­ties track and pub­li­cize the numbers — is in­con­sis­tent. She fig­ured that by not­ing the data in real time, she and Dong could pro­vide aca­demic col­leagues with sta­tis­tics for later anal­y­sis.

“When we started this there was not a sin­gle ded­i­cated covid19 track­ing web­site by a pub­lic health author­ity any­where,” she said.

Dong went straight to work. For 12 hours, he col­lected data, trans­lated in­for­ma­tion from Chi­nese, de­signed ta­bles and bull­dozed the sta­tis­tics into a pro­gram that would cre­ate a map. His stark aes­thetic choices — black for the back­ground, red for dots in­di­cat­ing in­fec­tions — were de­lib­er­ate: “I wanted to alarm peo­ple that the sit­u­a­tion was get­ting worse.”

The next morn­ing, Jan. 22, Dong showed Gard­ner his re­sults. After a few tweaks, the project went live, with red dots bal­loon­ing across coun­tries — and states and prov­inces, when pos­si­ble — to show numbers of known cases, and charts list­ing con­firmed cases and deaths for each ju­ris­dic­tion.

Gard­ner and Dong fished through me­dia re­ports and Twit­ter ac­counts for up­dates, man­u­ally punch­ing in the new fig­ures. Early on, as the virus spread to Ja­pan and South Korea, the project was a crowd­sourced ef­fort, with peo­ple from around the world email­ing about new cases. The data feed­ing into the dash­board was open to the pub­lic in a Google Sheets file, so any­one could click through the numbers and pin­point mis­takes or offer sug­ges­tions.

As the po­ten­tial di­men­sions of the disas­ter took shape — mil­lions sick, wide­spread lock­downs, a scorched-earth global econ­omy — the pub­lic looked to the tracker to make sense of a fright­ful or­deal.

Lainie Rutkow, a pro­fes­sor of health pol­icy and man­age­ment at Hop­kins’s Bloomberg School of Pub­lic Health, was re­minded of the pub­lic anx­i­ety she wit­nessed in the days after the at­tacks of Sept. 11, 2001, a cat­a­clysmic event that had rerouted her from a ca­reer in law to one in pub­lic health.

“That same type of un­cer­tainty, I see now,” she said, “those feel­ings of un­cer­tainty about what hap­pens next.”

She joined the team with the idea that what Gard­ner and Dong had started could grow into a more am­bi­tious tool, one that could not only track the virus but also ex­plain and con­tex­tu­al­ize the spread for a global au­di­ence des­per­ate for in­for­ma­tion.

At the same time, the me­chan­ics be­hind the project were rapidly chang­ing. In Fe­bru­ary, gov­ern­ments be­gan re­leas­ing virus sta­tis­tics. Rather than ease the work­load, the shift re­vealed a ma­jor short­com­ing: There was no in­ter­na­tional body lay­ing out cri­te­ria for how to tally coro­n­avirus in­fec­tions or deaths. Each coun­try put out its own data, some­times re­vis­ing it weeks or months later in ways that dra­mat­i­cally changed the trends the tracker was try­ing to iden­tify.

The pace of in­fec­tion — faster ev­ery day — cre­ated more is­sues for the small team.

“Ini­tially, I was try­ing to up­date it two to three times a day, at 12 p.m. and 12 a.m.,” Dong said. “But peo­ple were so anx­ious to see the dash­board, so I had to up­date ev­ery three to four hours. But some­times by the time I had col­lected it all, the data had up­dated in the orig­i­nal data source, so I would have to tear down ev­ery­thing I’d done and col­lect it again.”

Chart­ing in­equity

Blauer first pulled up the tracker out of per­sonal cau­tion. It was Fe­bru­ary, and she was sup­posed to travel to Is­rael and In­dia soon.

As she scanned the dash­board, Blauer rec­og­nized that what her col­leagues were build­ing hit on con­cerns cen­tral to her work at the Cen­ters for Civic Im­pact, which helps lo­cal and state gov­ern­ments use data in de­ci­sion­mak­ing.

She con­tacted Rutkow and asked to join the team.

Her own plunge into data started after col­lege, when Blauer worked as a ju­ve­nile pro­ba­tion of­fi­cer in Bal­ti­more.

She would get called to po­lice sta­tions in the mid­dle of the night and asked to de­ter­mine whether a child who had just been ar­rested could go home with fam­ily or needed to spend the night in jail.

“I was asked to make a de­ci­sion in the lives of th­ese kids based on no in­for­ma­tion,” she said. “I didn’t know if the child had been in school that day, if they had ac­cess to food at home, if there was a so­cial worker in­volved with the fam­ily.”

There had to be a bet­ter way, she de­cided.

In 2004, she be­came a key player in then-mayor Martin O’mal­ley’s Ci­ti­s­tat pro­gram, a data-driven ef­fort to track and mon­i­tor mu­nic­i­pal work.

When O’mal­ley (D) be­came gov­er­nor, Blauer ran a sim­i­lar statewide ef­fort from Annapolis, track­ing ev­ery­thing from in­fant mor­tal­ity rates to bud­get spend­ing. The large-scale ef­fort — one of the first of its kind in gov­ern­ment — “be­came a kind of re­li­gion,” she said.

The ex­pe­ri­ence schooled Blauer on the nitty-gritty com­plex­i­ties of Amer­i­can in­equity. Any given Zip code was lay­ered with the his­tor­i­cal bag­gage of past poli­cies. Dis­crim­i­na­tory hous­ing, health­care ac­cess, school sta­bil­ity — they were all baked in.

So when Blauer and oth­ers be­gan plot­ting a U.S. dash­board to com­ple­ment the global tracker in March, they de­cided a sim­ple tally of in­fec­tions and deaths would not be enough to fully ex­plain what the virus was primed to do to black, Latino and Na­tive Amer­i­can com­mu­ni­ties.

The team set­tled on three ar­eas of ad­di­tional in­for­ma­tion for each county in the United States. The first was health-care ca­pac­ity — not only the num­ber of in­ten­sive care unit beds and staffing sta­tis­tics but also how peo­ple ac­cessed the lo­cal health-care sys­tem, whether through pri­vate in­sur­ance or Med­i­caid. Next, they de­cided to add in­for­ma­tion on the de­mo­graph­ics of each county, in­clud­ing a racial break­down, un­em­ploy­ment fig­ures and age dis­tri­bu­tion.

The third fo­cus was com­par­ing county dis­ease data against the state’s as a whole. The goal was to mea­sure whether the virus posed an equal-op­por­tu­nity risk or whether all that his­tor­i­cal bag­gage would de­ter­mine who lives and dies.

The U.S. map went live in mid-April and was quickly com­pli­cated by in­con­sis­ten­cies in how dif­fer­ent ju­ris­dic­tions pre­sented data.

The Cen­ters for Dis­ease Con­trol and Preven­tion had just changed its guid­ance, sug­gest­ing that lo­cal health of­fi­cials in­clude prob­a­ble deaths and prob­a­ble in­fec­tions in their counts. Some states did. Some didn’t. Some­times some coun­ties within a state would fol­low it and oth­ers would chart a dif­fer­ent course.

“It’s one thing that this is not con­sis­tent glob­ally, that Spain presents data dif­fer­ently from In­done­sia, and In­done­sia re­ports dif­fer­ently from the U.S.,” Gard­ner said. “The thing that’s crazy to me is how dif­fer­ent the re­port­ing is within a state in the United States, let alone state to state.”

By then, Johns Hop­kins, like the rest of the county, was shut down. The tracker team — which now com­prises dozens of pro­fes­sors, ex­perts and stu­dents from mul­ti­ple de­part­ments — coordinate­d via long Zoom meet­ings, emails and phone calls.

“The first 61/ weeks, we were

2 build­ing the plane as it flew, and we were fly­ing at su­per­sonic speeds,” said Sheri Lewis, a mem­ber of Hop­kins’s Ap­plied Physics Lab­o­ra­tory work­ing on the tracker.

The U.S. map soon il­lus­trated what anec­do­tal re­ports were sug­gest­ing: Mi­nor­ity com­mu­ni­ties were be­ing hit hard­est by the dis­ease.

In Washington, D.C., African Amer­i­cans ac­counted for 46 per­cent of the city’s pop­u­la­tion but 74 per­cent of deaths. In Ari­zona, Na­tive Amer­i­cans — who make up about 5 per­cent of the pop­u­la­tion — ac­counted for 18 per­cent of deaths. And in South Dakota, where African Amer­i­cans are less than 3 per­cent of the pop­u­la­tion, they rep­re­sented 17 per­cent of coro­n­avirus cases; other mi­nor­ity pop­u­la­tions to­tal­ing around 16 per­cent of the pop­u­la­tion, in­clud­ing Lati­nos and Na­tive Amer­i­cans, made up 70 per­cent of cases.

For Gard­ner, the U.S. map put a spot­light on in­equal­i­ties that made the pan­demic more heart­break­ing.

“When you ac­tu­ally start look­ing at the af­fected pop­u­la­tions, the break­down of race and age and eth­nic­ity and so­cioe­co­nomic de­mo­graph­ics, it be­comes so much more hu­man,” she said.

Even Blauer, who has spent her ca­reer teas­ing out the so­cial ills hid­den in­side strings of numbers, felt frus­tra­tion and res­ig­na­tion as she watched the virus’s path.

“We’ve known for gen­er­a­tions that pop­u­la­tions that are poor and liv­ing in highly dense ar­eas have th­ese co-mor­bidi­ties that are pre­sent­ing for risk for covid-19,” she said. “But the re­al­ity of the sit­u­a­tion is we don’t do any­thing about it. If you are born black in this coun­try, it’s harder for you to get a job, harder for you to keep a job and also harder for you to stay alive.”

“The thing that’s crazy to me is how dif­fer­ent the re­port­ing is within a state in the United States, let alone state to state.” Lau­ren Gard­ner, as­so­ciate pro­fes­sor at Johns Hop­kins’s Whit­ing School of En­gi­neer­ing


The team be­hind the John Hop­kins Univer­sity covid-19 dash­board tool mon­i­tors mil­lions of pieces of coro­n­avirus data daily from around the world.


LEFT: A school is ster­il­ized in Taiyuan, in China’s Shanxi province, in March. RIGHT: A pa­tient is trans­ferred in April from a re­hab cen­ter in Bal­ti­more. Hop­kins team mem­bers such as En­sheng Dong, from Shanxi, and Beth Blauer of Bal­ti­more say it’s not lost on them that the data points rep­re­sent peo­ple, not just numbers.


Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.