In this month’s article, Mike Farahbakhshian discusses the coronavirus epidemic and what Health IT can do to combat it. Suggested drink pairing: Lustau Brandy de Jerez Solera Gran Reserva, neat.
Passion! Excitement! Betrayal! C-SPAN!
Happy February, Useketeers! Once again, we’re in the month that’s so short, yet packed full of so many things: The Superbowl, Black History Month, and the Trump impeachment trial (which is like the Superbowl for debate nerds like me). Whatever you feel about the result, the trial itself was full of high pomp and drama: Constitutional law arguments that will change the bounds of the presidency; tense call-outs of the Chief Justice from senators; Adam Schiff’s closing statement, powerful rhetoric worthy of Cicero; and a telenovela-like breaking of ranks by Mitt Romney, who cited his conscience in a (now bipartisan) vote to convict.
Though the President was acquitted, partisan tensions run high, with a divisive State of the Union Address – both delivered by POTUS and subsequently ripped apart by the Speaker of the House – and Presidential declarations of anger, vindication and retribution in the wake of the impeachment. Wow! All this, not even a week into the month.
Yet despite living in some of the most divided times in recent history, we are facing a threat that might just bring us back together. This threat does not discern Republican from Democrat. This threat knows no ideology, borders or reason. I’m of course talking about the Wuhan Coronavirus 2019-nCoV. As many of you in the Federal Health IT community gird your loins and harden your livers for the inevitable HIMSS trip, let’s talk about what this disease is and does, and what Health IT can (and can’t) do to combat it.
From Pangolin to Pandemic: The Story So Far
2019-nCoV was first discovered in the Chinese city of Wuhan, Hubei province. 2019-nCoV, much like Ebola, rabies, the stanky leg and pretty much every other terrible disease out there, originated in a bat population. However, since most people tend to avoid contact with bats, the disease required an intermediary animal.
Enter the humble, adorable pangolin, a critically endangered animal which is often illegally trafficked for meat in China, Africa and elsewhere. A pangolin, for those of you who don’t know, is an anteater from Lord of the Rings who wears mithril armor. Here’s a hot take from Mike: I would advise all members of the Federal Health IT community to abstain from eating pangolin meat, no matter how delicious it may be. Just Say No to pangolin meat!
But someone didn’t say no to pangolin meat, so we’re stuck with 2019-nCoV. As a coronavirus, it is a cousin of Sudden Acute Respiratory Syndrome (SARS) and Middle East Respiratory Syndrome (MERS). You may remember the SARS scare, and if you don’t, I definitely recommend the highly, extremely factually correct 2004 Thai movie entitled SARS Wars: Bangkok Zombie Crisis.
Kidding aside, while SARS was highly infectious, it had a mortality rate of 9.6%. MERS, on the other hand, infected far fewer people – less than 3000 — but had a mortality rate of 34.4%. We would prefer a low mortality rate for 2019-nCoV. Right now, the official death rate of 2019-nCoV is around 2%, but that figure is in dispute. For one thing, the number has remained exactly constant for days on end, which is highly improbable. I’ll get to that later. As of February 7th, the World Health Organization’s official counts list 31,481 confirmed cases and 637 deaths.
A Comedy of Errors
The response to date has not been reassuring. I will cover some of the areas where the epidemiological response to date has been handled poorly. I’ll also make observations on how this can be fixed.
Error #1: Epidemiological Countermeasures Without Proper Resources or Incentives:
The entire city of Wuhan is on lockdown and the infected are being rounded up into quarantine camps. This might have something to do with the fact that the disease has “super spreaders,” which lead to extremely fast growth; a single patient infected fourteen health care workers. The population of Wuhan has been ordered to report their temperatures daily to the authorities. In essence, this is compulsory crowdsourcing of symptomatic and syndromic information. The problem here is that the Chinese authorities are ordering compulsory crowdsourcing without any incentives. Anyone who reports an elevated temperature gets moved to a quarantine camp, which lacks the resources to treat effectively and has a mortality rate double the official rate. For all of you Health IT vendors who harp about Human-Centered Design and User Experience, here’s a lesson to take away: it is in the best interest of population health to have people report their illnesses. You need to make it in the patient’s best interest – i.e., an improved outcome – to encourage self-reporting. Being put in a worse situation is a negative incentive. On this side of the ocean, the CDC has instituted the first Federal quarantine in over 50 years (the last being for smallpox). Right now, our public health resources are enough to ensure that quarantines lead to an improved outcome, and that self-reporting leads to improved patient outcomes. As the epidemic spreads, and our public health resources are taxed, will we make the same mistakes as China? Good planning by our Federal Health IT community, starting today, will help forestall that fate.
Error #2: Reliance on Technology with Bad Models
When faced with a global pandemic, the first instinct is to rely on technology to solve the problem. And why not? AI and machine learning are great modeling tools when it comes to epidemiology. In fact, an AI model predicted the 2019-nCoV outbreak. However, AI is only as good as its data and modeling. Bad data led to the failure of Google Flu Trends. Sometimes the data is good but does not account for all variables in a model. For example, an AI model using a Recurrent Neural Network of the 2019-nCoV spread has successfully predicted the following day’s numbers of infected, dead and recovered within a 3% margin.
Yet, left to extrapolate, the AI model is predicting 52 million deaths within the next 45 days. Useketeers, I don’t know about you, but these numbers sound about as realistic as the chance of an actual pangolin invasion.
That said, if you want to avoid bad numbers like this, any AI model needs to continually update variables based on factors like responses to date, tweaks to mortality rate based on asymptomatic carriers, and a whole lot more.
The bottom line regarding AI is simple. We can’t and shouldn’t rely on unmitigated extrapolation – it’s too dangerous and can lead to mass panic and counter-incentive activities like not reporting when you’re showing symptoms.
There are some promising uses of AI in fighting 2019-nCoV. Remember those generative adversarial networks I keep harping about? They’re great for identifying proteins and enzymes that can combat the coronavirus. A Maryland startup has already identified promising antiviral treatments. Another multinational collaboration has identified that 2019-nCoV might be vulnerable to an HIV antiviral, azatanivir. Using AI and Machine Learning in this context will be much more useful, but it takes a long time to go from modeling to solution. The Federal Health IT community should start working now to ensure that potential treatments become actual treatments.
Error #3: Public Health Reporting is Suspect
Governments need to communicate clearly and honestly during an epidemic. The communication needs to engender trust because public health authorities need honest cooperation from a scared populace. Yet as with most things in the authoritarian Chinese regime, facts are often concealed for political motives. A whistleblower tried to reveal information on the spread, but was rebuked; he later died of 2019-nCoV. This is causing serious unrest in China. Adding fuel to the fire, the Chinese social media service, Tencent, allegedly accidentally leaked the true 2019-nCoV infection and death toll. One grab purports to show the death toll at 24,589 – 80 times the official figures.
These numbers have appeared multiple times, each time with a silent “correction,” leading some to speculate that this could be:
- Evidence of a cover-up, with two sets of numbers: real and “official”
- Evidence of a doctored smear campaign, although whom it would benefit is currently unclear
- Evidence of a coding error, though why it is recurrent is puzzling
- Evidence that someone on the inside is trying to leak the real numbers, right before the numbers are updated/overwritten
Without a fair and free press observing fact-based norms, we will never know. For all of our political fractiousness here in the United States, I hope we still have enough faith in our institutions and data-driven decisions to be honest about epidemiology numbers.
We need to be honest because 2019-nCoV is spreading like wildfire. The first death outside of China has occurred in the Philippines. The second was in Hong Kong. The first person-to-person spread in the US has been documented. The disease is projected to become a permanent endemic disease, like many flu strains, periodically flaring up around the world. If we are to have any hope of containment, we need honest, straightforward, reliable reporting.
Frankly, I don’t have a lot of hope. Here at home, the complete failure of the Iowa Democratic Caucus reporting app has directly led to multiple conspiracy theories. Let it be noted that there is no evidence for shady conspiracy theories, no matter what the Right may be pushing. But there is plenty of evidence – from a postmortem code review – that the code was slapdash and poorly tested before being rushed to market within two months’ time.
Now imagine a code error in reporting for a terrifying pandemic. Perhaps, say, a blip in infection or death rates; during an election year in an unparalleled time of global unrest; with an infection that can cripple our allegedly strong economy. Useketeers, I ask you: Do you think cooler heads will prevail if public health reporting is anything other than perfect? Moreover, I ask you: Do you think conspiracy theories will go away if the reporting is perfect?
We are fighting virulence on two fronts: the physical, in the form of 2019-nCoV; and the memetic, in the form of conspiracy theories and mutual distrust. A coronavirus can be treated with medicine. The viruses of the mind, that break down bonds between citizens and their Government, is far worse. I don’t know if anything can be done about this. Yet I know we are doomed if we don’t bother to try.
The only thing I can suggest for the Federal Health IT community is to be a better citizen and demand the same of your partners and customers. Test your damned reporting code. Make sure your product is reliable. Work with your corporate and Government public relations teams to shut down conspiracy theories and misinformation when they start. Public health requires tight coordination between IT and Strategic Communications. Work with your counterparts closely. The alternative is discord, panic, and death.
This Message Is Directed Achoo
Here’s an equally unpleasant fact: at the rate this is spreading, some of you may get this virus. Maybe I will. Maybe it will be like a mild cold. Maybe some of us will die. Maybe it will be like the Spanish Influenza of 1918. Maybe it will peter out like SARS. The short answer is, we don’t know. But for those of us who work in Federal Health IT, we need to stop treating public health products and services as deliverables and start treating them like what they truly are: tools to preserve our lives, the lives of our loved ones, and the lives of our fellow citizens.
Our society is more divided than ever before: acrimonious political rifts between Left and Right; OK-Boomering between young and old; gender mistrust in a post-MeToo world; even Laurel versus Yanny. Yet 2019-nCoV, or the next global pandemic, or the next one, cares nothing for these fractures. And so it is that we must join together, extending a hand and depending on our Government and our fellow citizens.
Ironic as it is, maybe a global disease will spur us to heal our cultural sickness.