The Reopening is a Gamble Based on Terrible Metrics.

Online Craps - Play Real Money Craps Casino

I’ve never been a gambler, so I don’t really know anything about the game of craps. I just see how it’s portrayed on the big screen–it always seems so fun and dramatic. A guy rolling the dice, surrounded by a group of cheering fans. He says something like  “Seven, c’mon seven!” and tosses the dice in the air. For that moment, a drawn out moment, all the faces freeze in gleeful anticipation, watching the dice fly through the air, waiting with hope and anticipation to see where they land. It’s completely left to chance, or maybe luck, or perhaps prayer. For that moment everybody waits, everybody dreams.  They live in that moment, waiting to celebrate. 

The scene is very much like the one we find ourselves in now. The decision to reopen the economy has been made, the dice have been cast into the air, waiting to land. Already we’re seeing signs that the economy is improving, people are venturing out again for summer getaways. Restaurants and shops are reopening, and there are predictions of improvement in several economic metrics like jobs, and consumer spending. While we have metrics to measure progress on the economy, metrics that measure the concurrent COVID-19 risks are, unfortunately, terrible. In other words, while the potential winnings at the craps table are plain for all to see, the participants in this game are simultaneously blind to the downside risks. Like I said, I’m no gambler, but it would seem to me that a good gambler always knows their risks. Let’s have a look then, at the metrics we’re using to warn us of the risks in the gigantic game of chance.

COVID-19 Testing.

 The media has never been very good with numbers, and the statistics heavy reporting during the pandemic has done nothing but reaffirm this. Media outlets continue to highlight “the new cases of coronavirus” without really putting them into the context of the increase in testing. Over this past Memorial Day weekend, media outlets reported that North Carolina experienced a record high number of new cases, with 1107 new cases reported on May 23rd. However they did not report the dramatic jump in testing that had also occurred. Looking at the graph below, you can see that testing has been increasing steadily over the previous weeks. Though if you look closely, you would rightly question my point, because on the day of the record spike, 5/23, there was a relative reduction in testing compared to previous days. This highlights another serious problem with testing, one that CDC director Dr. Robert Redfield recently eluded to when he said “testing sucks, bro”. Ok so he didn’t actually say that. But he did note that there are serious problems with data reporting, he stated that by the time the data reaches the CDC, “regularly the data is delayed and it’s incomplete.

And for that, I hereby nominate him for the award of “Greatest Understatement of the Pandemic by Physician or Scientist”. To put it more simply, test result data is a freaking mess. The data on test results are reported by various different labs, at different times, and in different ways. Some of them might take a week to come back, some might take a few hours. While all  labs report all of their positive results, not all labs report negative tests, or report them at the same time. Taking all of this together, if I were to look at that record number of new positive cases on 5/23 could I say with confidence that the data portray an accurate depiction of what was happening on that day? Not at all. The sad thing is that nobody really knows.  We’ve taken to using moving averages of data not just because there is day-to-day variability in the data, but because there is a significant variability in the reporting of the data. 

Percent positive tests. 

The White House guidelines for “Opening Up America Again” called for a 2 week decline in the number of new cases to be achieved before proceeding to a phased reopening. However, it was apparent that as testing volume increased, case numbers would continue to climb. So as an alternative to this they also proposed that a 2 week decline in percent positive COVID-19 tests could also be used as a gating criteria. As with many things in life, this seemed like a good idea at the time, but later had us scratching our heads and wondering what we were thinking. COVID testing initially focused on high risk patients who presented to health care settings. As the pandemic has progressed and testing numbers have increased, we’re now testing more people who are not at the same high-risk profile. We’re performing more surveillance testing at group settings like nursing homes, prisons, and meat processing facilities. We’re screening people before surgeries and outpatient procedures. Screening these lower risk groups has the effect of reducing the percentage of positive tests. This in turn has the effect of making it seem like the prevalence of the disease is decreasing, when it may not be, and very likely going in the opposite direction. Now then, we find ourselves in a rather bizarre situation with two COVID-19 test-based metrics, total positive tests, and percent positive tests. Both of these metrics are telling us two different things yet both are somehow incorrect. 

CLI: Syndromic Trending. 

Every flu season, people start going to emergency departments and physicians’ offices with flu-like symptoms like fever and cough. Several of these locations are part of a statewide and nationwide network that track the seasonal flu. This system of patient care sites tracking symptoms of flu collectively make up the Influenza Like Illness Network, or ILINet. It’s a great system for tracking flu and gives clinicians a valuable tool as they make preparations for seasonal flu. Sometime in March, someone had the bright idea of using this network to help track COVID-19. It made a lot of sense, after all many of the symptoms were similar, cough and fever, and there was an obvious shortage of testing which perhaps this network could help make up for. At that time, it was the only way for officials to monitor COVID-19 activity. Thus ILINet was rebranded and the COVID-Like Illness Network, CLINet was born. 

But there’s a number of reasons why this system, while great for tracking flu, is an imperfect tool for COVID-19 surveillance. First, the system requires that patients present to their doctors offices and ED’s with illness when they are sick. We know that around the country, patients are shunning visits to the health system, my own office visits are down about 50%. Even people with COVID-19 are delaying going to the hospital, and by the time they get there, they tend to be very sick. People are presenting to the hospital often a week or more after their first symptoms.  This tells us that there is a considerable time lag between when people are infected and presenting to the ED. By the time CLINet tracks an upward trend, there will already have been a several week backlog of people who have been infected but haven’t yet presented to the healthcare system. In other words, CLI is what we call a lagging indicator, it’s not so much of an early warning indicator as it is an alert that screams “brace for impact!”. 

Antibody testing.

Sars CoV2 antibody testing is emerging as an important way to tell if people have been exposed to COVID. It can’t really tell if there’s an uptick in infections, but knowing whether people have been infected is an important part of modeling. Knowing how many people have already been infected reduces the number of susceptible people, and thus the number who could potentially get sick from Covid. The antibody tests have a lot of positive things going for them, they’re  relatively quick, painless, point of care, widely available and have fast turnaround times. As such, you’d think that they were the perfect test. Except there’s some issues with them. A positive test certainly could mean that you have had coronavirus, or it may also mean that you haven’t. Or maybe you’ve had another coronavirus that’s cross reacting with the test. With the current test we don’t truly know.  But at least if you’ve had a negative test you can be sure that probably maybe have not had it. Or maybe you have. Again, it’s a tough call. For now, it appears that the antibody test will need to wait until the CDC determines that they’re no longer nearly useless. 


Hospitalizations is probably the only indicator that we have that is not subject to errors from the various sources above. The obvious problem of course is that hospitalizations data is that they are a lagging indicator. If we wait until people start landing in the hospital, it’s probably too late to prevent more people from landing in the hospital. Aside from deaths, it’s the ultimate lagging indicator. Lagging indicators are not useless, they can help tell us when a surge has peaked and provide valuable data for modeling future outbreaks. But they’re not the indicator we need to help us determine when and where the next outbreak will happen and allow us to act in time to stop it. 

What are the Metrics we Need?

Tests, tests, and more tests. In order for testing to be a more valuable metric, there needs to be many more of them. Testing sites need to be ubiquitous, and we shouldn’t have to wait 4 days for the results. Currently in the U.S we’re testing about 400,000 people per day, that’s much better than we were a few weeks ago. But several groups who look at the numbers of tests required to safely monitor the public are calling for much higher numbers, and there are a number of estimates that cover a very wide range. On the low end you have a group from Harvard  calling for 900k tests per day at a minimum. At the high end there’s another group, also from Harvard, calling for 2 million tests per day to start, ramping up to 5 million tests per day by early June. Harvard apparently has a very large campus and it was too far for the two groups to walk and meet each other.

By increasing the numbers of tests in this way, we could more easily know the true prevalence in any given community, and see when it is increasing. In lieu of widespread COVID patient testing, we could be helped by some other metrics that give us clues about disease activity, and thus help in predicting disease outbreaks before they happen. Testing of wastewater samples, for example can identify the presence of coronavirus in a community even before the first case is identified. Kinsa, a smart thermometer, has shown that it’s aggregated data can predict disease activity 3 weeks in advance. Increased community mobility, as monitored by tech companies like Facebook and Google, can show that a community is at higher risk of a COVID outbreak. In fact IHME is now using mobility data in constructing their models of COVID activity. 

Maybe we’ll finally get testing to where it needs to be. Maybe some of these other metrics will help in predicting future outbreaks before it’s too late to stop them. Maybe coming weeks will see more effective treatments for sick people and meaningful progress towards a vaccine. These are the only things that change the trajectory of the dice as they fly through the air. As healthcare workers all we can do is prepare, and wait for the dice to fall where they may. 

Deep Ramachandran, M.D. is a Pulmonary, Critical Care, Sleep Medicine physician, founding CHEST Journal Social Media Editor, and co-Chair of ACCP Social Media Work Group. He blogs at Caduceusblog. He is on twitter @Caduceusblogger.

Covid Journal 10: My own Covid test shows that our health system is still very sick.

Drive Thru Testing

About a week ago, while getting ready to come in to work, I noticed my left eye was red. This gave me pause for two reasons. First because pink eye has become recognized as one of the many signs of COVID-19, and second because regularly perform bronchoscopy, an aerosolizing procedures on patients. I was referred by employee health to get a COVID-19 test. Except I couldn’t because while our hospital does offer it’s employees the option of drive-thru testing, that test is only available for limited blocks during the day, and that time had already passed. It was a Friday afternoon, now I would have to wait until Saturday to get tested. But on the weekends employee health does not offer the test at my hospital, so instead I would have to drive to a site 50 miles away from my house in order to get the test on Saturday. Instead of waiting,  I opted to go to the outpatient testing center, the same place that I send patients to to get tested for COVID-19.  After a quick drive-thru nose swab (take heart, it’s no longer the terrible brain tickler it used to be) I was on my way to self isolation at home until my results came back from LabCorp. 

That process took 4 days. Fortunately they were negative, but let’s pretend that they had been positive.  In the days since the bronchoscopy I had been interacting with people, co-workers and patients all over the hospital. Had I been infectious, I could have potentially infected many of them, after all we know that asymptomatic people with COVID-19 can still spread the disease. All of those people whom I had infected at home and work had continued to be out in the community for the 4 days since the time I took the test. Many of them may have gone to church (now open), restaurants (now open) and to their own homes. Each of these people would have to be tracked down  and potentially tested, then  those tests would take another 4 days. If the tests in those contacts had been positive, many of their contacts in turn would have to be tested, who would also have to wait another 4 days to get their test results. At this rate, my one infection could easily cause a breakout cluster that could not be contained. It would only be a matter of time before the disease would find a vulnerable person and kill them. It does not matter how many surveillance  and tracking people we hire to track and isolate cases, if the tests take too long to come back, the disease will always be several steps ahead of us. 

The President said that we have “prevailed” on testing. We have not, and still have much work to do. Testing was initially constrained by a number of factors, most recently a lack of reagents. As we have solved that problem we have run into a shortage of swabs, and now we’re again running short of PPE, in particular gowns. Testing is just one part of many interdependent parts that we need to get right in order to control the infection. Until we prevail on all of them we have not prevailed at all.

This is our Dystopian Future: Brought To You By COVID-19 Antibody Testing.

(Omar Marques/Getty Images)

Right now pulmonary critical care physicians are fielding myriad calls and texts from all over the country. These callers come with frantic questions. Questions like “Hello Deep this is Minnie, your aunty. Your uncle’s toe is hurting and I read on the internet that  it’s a sign of coronavirus. What should we do?” Notice that they don’t ask “could I have coronavirus?” because they are absolutely convinced that they have it. What they want to know is “what should I do about my coronavirus”. Of course I give them the usual measured, sober and socially acceptable response; You’re ok, stay at home, wash your hands. 

But here’s what I WANT to tell them.

You beat coronavirus? Celebrate, CELEBRATE, YOU’RE FREE! Throw away your hand sanitizer! Run to the grocery store and grab an unsanitized cart, feel free to touch every cereal box. Use the bathroom and dry your hands with the air dryer. Take in a movie, get some popcorn, and go ahead and lick those buttery fingers. No need to be afraid.

Soon,  we’ll be through the initial hellish phase of this pandemic with it’s spiraling death counts. But the fight won’t be over. Until we have an effective vaccine or treatment, we’ll enter a prolonged stalemate with the disease. Now that we’re looking at opening the country back up again, people have been contemplating what that might look like. 

That future will inevitably split us into two groups based on whether we have immunity from the disease or not. Immune people will have their status bestowed on them from previous infection or documented by results of antibody testing. Eventually someone will come up with a catchy title for Immunes and Uninfecteds. Perhaps there will be a Twilight-esque movie about a forbidden romance between the two. Spoiler alert, one of them dies.

Life for Immunes is going to be good, their lives will look, to the Uninfected, like an unending party. They are going to go out early, and stay out late.They will restart their lives, and live them like it was the last day of their lives, such will be their appreciation for their rewon freedom. 
Uninfecteds will still be staying at home, watching the glamorous lives of the the Immunes as if from behind prison bars.  While people will know their own status there will be no way to tell who is who in public. For that reason, governments will still require some form of social distancing to protect these people. Many Uninfecteds will with violate that, and they will get sick and die. 

Immunes will feel sorry for the Uninfecteds, “I feel really bad for Uni’s” they will say,  “but I have to work and feed my family. Why should I go on with the quarantine when I’m not sick and there’s nothing wrong with me?”
Life for Uninfected’s will be tough. They will eventually need to go back into the workforce. But in order to work they will need to have protections against infection.So they’ll ask for accomodations from their employers based on the American with Disabilities Act. And as they are accomodated, workplaces will become increasingly segregated as Uninfecteds seek protections  like individual spaces with appropriate physical distancing. They’ll have different restrooms, different lines in the cafeteria and different dining tables. In some cases an Uninfected worker will be paired with an Immune for their own protection. The Immunes will naturally resent the social isolation. The Uninfecteds will see the Immunes new found freedoms and resent them right back. 

Meanwhile the employment pictures will look very good for Immunes, particularly health care workers. While employers will insist that they do not discriminate based on immune status, everyone knows that they do. Employers know that Uninfecteds could get sick and then require time off that could range from a few weeks to a few months. Even worse, they might die. For that reason, insurance premiums, including life, disability, and health insurance, rise astronomically for Uninfecteds.  For all of these reasons Uninfecteds have a much harder time finding work especially obese men over 50 who appear to be more affected by the virus. 

After the tragedy of the intial part of the health crisis in the U.S., healthcare workers were seen as heroes and had enjoyed a new found solidarity. That quickly falls apart as health care systems bid up the prices on Immune workers. Across the country every one is tested for antibody status, laws are passed requiring that everyone have their COVID-19 Antibody status determined. 

Health disparities that existed before the crisis become wider, while new disparities emerge.  Covid-19 antibody status becomes a ticket to a better life. Illicit dealers emerge to sell samples of the virus to people who want to infect themselves.  Political differences emerge too, one political party declares itself a champion of Immune freedoms. SImultaneously they accuse the other party of promoting a socialist agenda by promoting accomodations for Unifected.

The Immune party blocks a bill in the House stating that it is an invasion of privacy, but it’s really about blocking funding needed to build factories to make the vaccine. The vaccine is now the only thing that separates the Immune from Uninfected. Facebook posts appear showing the dangers of vaccination, stating that the vaccine come from a WHO plant built by the Chinese. The vaccine, it is rumored, contains another secret virus. This virus, they say, is even deadlier than COVID-19. And the new virus would do things to our country that no one could imagine. 

Deep Ramachandran, M.D. is a Pulmonary, Critical Care, Sleep Medicine physician, founding CHEST Journal Social Media Editor, and co-Chair of ACCP Social Media Work Group. He blogs at Caduceusblog. He is on twitter @Caduceusblogger.