Is DOH manipulating the data?
In the last few days, some statisticians and data scientists I follow on Twitter have been raising a warning flag about Department of Health (DOH) statistics. Yesterday, the total number of COVID-19 cases recorded since the start of the pandemic some 14 months ago topped 1 million. That’s a grim milestone, and worrying enough. But the real source of discomfort has to do with the remarkable drop in the number of active cases.
From 203,710 active cases on April 17, the total fell to 77,075 on April 25. (It dropped further yesterday, to 74,623, with 11,333 recoveries and 8,929 new cases recorded.) This should call for, if not a celebration, then at least a round of congratulations. This is good news, right?
“But DOH reported more than 93,000 [new] cases in just the past 10 days, higher than today’s [number of] active cases,” Edson Guido, a PhD candidate in economics at the University of the Philippines and data analytics lead of ABS-CBN, wrote on April 25.
If I understand him correctly, that means that — even assuming that all 93,000 new cases are mild and asymptomatic — all those 93,000 new cases by the DOH’s own standards must still be considered active on the 10th day. (The standard guidance today is if you test positive but do not show any symptoms, you must isolate for 10 days.) Why was the total number of active cases pegged at 77,075? Puzzling, to say the least.
A couple of days before, Guido explained on the ABS-CBN News Channel what the DOH was doing. “The scale of recoveries is something we haven’t really seen before. Because what DOH used to do before was report thousands of recoveries every Sunday, where they do what they call the time-based tagging of recoveries. They usually tag mild and asymptomatic cases as recovered 14 days after the onset of symptoms, as long as their status did not progress to critical or severe. But what they’re doing right now is they’re doing this tagging every day, and we’re seeing the result. It’s really a tweak in the protocol more than anything else.”
A tweak in the protocol is polite language for a potentially disruptive change, yet again, in how the DOH collects and presents COVID-19 data.
Peter Julian Cayton, a PhD who teaches statistics at the University of the Philippines, questioned the unusual decrease, too, but in less polite terms. “From 203k to 77k in 8 days? Really? Are we just being lazy with monitoring COVID-19 cases, just tagging recoveries without due diligence?”
“These recovery rates are pretty ridiculous,” he wrote in another tweet.
Others have also questioned the sudden acceleration in recoveries, which started soon after the Philippines surpassed the Indonesian record for highest number of active cases in Southeast Asia.
On the macro level, the change in reporting protocol is concerning, especially when the resulting rapid decrease in number of active cases is not reflected, however imperfectly, in hospital bed utilization and other such data-based measures or in anecdotal measures like easier access to medical services.
But the micro level needs a closer look, too. For instance, Jules Guiang is a TV personality who tested positive on April 17. The contact tracers in touch with his family (three other members had also tested positive) told him he would be considered recovered on April 25 “dahil sinabay kami sa day of symptoms ng Dad ko”—that is, because they counted the rest of the family from the first day his father, the first to fall ill, exhibited symptoms. Does this even make sense?
To draw out the implications of the contact tracers’ advice: On April 25, all four COVID-19-positive members of the Guiang family would have been classified as recovered, despite the fact that not all of them tested positive on April 15.
To be sure, the suspiciously rapid rise in recoveries is not the only problem with government data collection or presentation. Jason Haw, a PhD epidemiology student who teaches at Ateneo de Manila, was among those who criticized the National Task Force Against COVID-19 for deliberately fudging the numbers on the country’s vaccination rollout. “The Philippines’ vaccination rollout is so bad that the government had to (1) use total number of doses administered instead of number of people vaccinated and (2) omit the denominator which is total population.” Yes, this is spin. In absolute terms, 1.6 million doses sounds high, but not when you divide that by two and then divide that further by a population of 110 million.
But it’s the DOH data on active cases that potentially give rise to the possibility of manipulation. And if the DOH data are manipulated, or even just collected and presented without any method or rigor, that means the national government is only fooling itself about the true state of the public health emergency. “There’s no due diligence with poor and incomplete data,” Cayton wrote. Also, no real chance against the coronavirus.
On Twitter: @jnery_newsstand, email: firstname.lastname@example.org