Welcome to the February/March issue of WonderWell, a newsletter intended to gather the most groundbreaking research and insightful commentaries in evidence-based medicine, wellness, healthcare leadership, writing, and innovation to help you live and work in alignment with your purpose and well-being.
Before we get into the topic for this month, I wanted to share that I had a great response to an article I wrote for the Globe and Mail (Saturday edition) about the pitfalls of the ‘self-improvement’ industry, and had a chance to discuss it on the All Sorts Podcast with the very gracious Desiree Nielsen, RD (whose books you should read, and she has a brand new one as well). It was a wide-ranging interview/discussion, and Desiree’s questions were incredible. Hope you enjoy it. I also realized how much I use the filler words “you know” — will work on that! Consider it a fair warning!
But how do we define it? Atul Gawande penned an excellent article in the New Yorker several years ago which hits the main points in a very compelling way. Further, several years ago, the American Board of Internal Medicine (ABIM) Foundation launched the “Choosing Wisely” initiative in part to support value-based care and reduce wasteful procedures/treatments (including that which has little to no evidence of effect).
Value as it relates to ‘value based care’ (VBC) is defined as “the measured improvement in a person’s health outcomes for the cost of achieving that improvement.” It’s crucial to note that while reducing costs/waste is related to VBC, it can’t be equated to VBC — they aren’t the same thing.
This brings us to a more philosophical argument: what does “value” mean generally, and how does this concept apply to both our health, and the systems that support it (the obvious ‘healthcare system,’ but also the places we work and play and live)?
First we can ponder what value means to ‘health.’ We can probably agree that our ‘health’ is inherently valuable, as it’s derivative: without it we’re limited in actualizing our other needs. As such, we place high value on our health, and are willing to invest in it, though oftentimes it falls by the wayside. Our health is valuable as it links to surviving but also thriving — without our health, our quality of life suffers (the ‘how’ we live), and at the most extreme, we cease to live (ie we die).
But what about value in healthCARE, i.e. the delivery of services for the purposes of optimizing health/well-being and offsetting/treating morbidities? How might we define that? I think we can conceptualize it in a few different ways.
For one: we can use the Costco example (for my international readers: Costco is described here). Most people would agree that Costco is a place where people seek value for household goods (food, appliances, etc). Why? Because per unit its on average cheaper: the consumer pays less per unit, so Costco remains in business primarily due to this perception of delivering ‘value’ by selling products that themselves deliver ‘value.’ Bulk stores are similar: what we save on packaging ends up in our pocket. A good example is shampoo: 20% more for the same price. That’s value.
Effectively, getting MORE for less money, even if we shell out a bit more money at the outset is ‘value.’ It feels like we got a deal. But it’s more than just monetary. Part of assessing value involves something intrinsic and somewhat intangible. It involves a ‘feeling’ of receiving more than we bargained for (in a good way, to be sure). Think about the last time you went to a cash register and realized a purchase was on sale. Or if your local barista threw in an item (I see you: extra shot of espresso!) for free. Or when searching for a hotel or airline ticket you realize you’ve stumbled upon a deal that’s too good to pass up. That feeling is akin to value. We feel ‘good.’
Can this idea apply to healthcare?
Currently, according to the Centers for Medicare & Medicaid Services (CMS), the US spends just over $12,000 on healthcare per person, per year. COVID has brought this into full focus: spending increased by close to 10% in 2020. The US healthcare system is also one of the least cost-effective systems in the world (meaning more money spent with worse outcomes), especially compared with countries like Canada (where I’ve spent most of my life) and where I was born (United Kingdom). But why? Well a large fraction of spending goes to hospitals (31%) and doctors/primary care clinics (20%), and, as Gawande writes in his New Yorker piece, it’s very likely secondary to how physicians (and insurers, and hospital administrators) are incentivized. Wasteful procedures (and spending) become a byproduct of a mismatch between incentives and value.
Allow me to me share a few stories to illustrate this point.
First, up: as a patient. In June 2020, a few months into the pandemic, I broke my left wrist — a classic FOOSH in my NYC apartment. I was lucky in that it involved my non-dominant hand/wrist. I was also lucky to have a friend who is an ER Doc in the city (hi Dan!) and whom I could call right away as I laid there watching the soft tissues of my wrist swell up (ouch!). Dan kindly organized for me to see his colleague at the emergency department (ED) at a big teaching hospital a few blocks away for an Xray and splint. That was done quickly, and Dan’s colleague kindly allowed me to take a photo of my scans. The orthopedic surgeon resident advised that I’d likely just need a cast, based on the scans alone. Then I was referred, as per protocol in the ED, for a one-week followup with a staff orthopedic surgeon. So, a week later, I dutifully went. The surgeon knew I had a medical background and am a physician myself. What he didn’t know was that I has sent the X-ray images to Dan, my brother (An ER doc in Canada) and a few other friends in ED and Ortho (in Canada and the US) — every single one said, based on the X-ray and my verbal history of what happened, that the bone alignment was good and a cast alone would suffice.
But what did the surgeon in clinic advise? Surgery. He explained that I would run the risk of arthritis if I didn’t take this option, as an an active woman in her thirties, this may not be ideal.
What’s the cost difference between a cast for a wrist FOOSH and surgery. ENORMOUS. The ED visit and split, as well as the followup clinic visit amounted to $3,600 USD. Surgery on top would have been at least $10,000 (and that’s being conservative).
And surgery includes risks (of anesthesia, post-surgical infection of the soft tissue/bone, etc).
I was able to explain to the surgeon that I was returning to Canada within days and would prefer to have surgery there if needed. He resigned himself to an ‘ok.’
Back in Canada I was seen quickly and casted, had a great physiotherapist and was about 75% in terms of range of motion and weigh-bearing within 8 weeks, and 100% within 7 months (that last 25% was tricky).
What’s the lesson here? This was a glaring example of the differences in approach between Canada — which has a universal healthcare system and a different setup of ‘incentives,’ and the US. Not assuming ill intent, but was this surgeon aware that surgery would not in fact be better, especially considering the balance of risks and benefits, not to mention costs? Did he know that the risk of arthritis, according to the evidence, is equivocal between a cast and surgery? It’s unclear. But this was an experienced surgeon, so perhaps the incentive to recommend a relatively easy surgical procedure with a high payoff (to him and the hospital system) played a large role.
Now imagine doctors like him making similar recommendations — that is, ones that could be influenced by financial gain, not clinical evidence — to millions of Americans each year, Americans who do not have the privilege of having a medical background or someone they could turn to for an informal opinion before deciding.
This is how the US spends more on healthcare than most other industrialized nations.
Second story: as a doctor. A few years ago as a resident physician in the children’s hospital, I realized that the team was ordering daily ‘lytes’ (short for ‘electrolytes’) for every patient admitted on IV fluids. Allow me a brief digression: our bodies very tightly regulate sodium (Na) and potassium (K), among other things (bicarbonate/HCO3 and chloride/Cl for instance). Intravenous (IV) fluids are a mainstay of supportive care for many patients admitted to the hospital, as hydration is often an issue in we’re sick, due to fluid shifts, insensible losses, etc. IV fluids themselves, as they contain electrolytes, can also lead to ‘too much’ or ‘too little’ electrolytes (specifically K or Na) which can lead to all kinds of issues for the brain and the heart. For an illness like diabetic ketoacidosis (DKA), assessing electrolytes closely and comprehensively is very important for many reasons (more here, note that DKA is managed differently in children vs adults, but the general principle of close electrolyte monitoring remains the same).
But for other illnesses, namely ones that don’t involve massive shifts in electrolytes, and where the main concern is whether the patient could become hypernatremic (high Na) on an NaCl IV fluid, ordering a whole set of ‘lytes’ isn’t usually necessary: checking Na alone should suffice, and potentially, with good kidney function, checking every second day may also be reasonable. So why did the team order daily lytes on everyone on IV fluids? Again: it seemed to be an action that was incentivized, but not due to financial gain per se — it was likely an action incentivized secondary to ‘habit’ (i.e. ‘this is just what we do.’).
Why does this matter? Habits or shortcuts or heuristics save us from additional cognitive load. When we drive a car, we automatically know we need to stop at a red light, and go at the greenlight. We don’t stop to ask ourselves consciously whether we should stop or go (yellow lights on the other hand…). Stopping to think ‘does this patient need this action? Will ordering X change the course of management?’ are important questions, but they take time and effort. It’s often easier, especially in a busy hospital ward, to go into automatic mode and simply order more than what’s actually needed. The idea being: it’s better to be more comprehensive than not.
Except: that assumption is often wrong for patients who are clinically improving.
With this particular example of ‘daily lytes’, out of curiosity, I wondered about two things:
1. What was difference is in terms of volume of blood taken for a patient if we ordered ‘all lytes’ compared with just the one we needed (Na) in this case?
2. What was the cost difference between the two?
So, after a hasty lunch one day, I headed down to the lab second floor and asked a technician, who provided me with a list of costs. I also spoke with a phlebotomist to understand the blood volume issue.
Here’s what I found: blood volume wise, the difference was small, but there was still a difference — a few milliliters per tube of blood taken. Iatrogenic (hospital caused) anemia remains an issue in acute care medicine, secondary to taking too much blood from a patient. For a patient in hospital for a week, a few milliliters a day can be significant enough to cause symptoms (fatigue for instance, on top of what may be normal secondary to an illness).
Cost-wise, the difference was about $5. I can’t recall the exact figures, but each electrolyte was roughly $1 dollar (for ease of explanation I’ll assume equivalence between the Canadian and US dollar). A full lytes panel of 6 electrolytes costs $6. What does this mean? Let’s assume that each patient admitted to hospital requires a full lytes panel on admission, but then those that receive IV fluids only need their Na checked each day after. If the average stay is about 4 additional days on the inpatient unit, we’re dealing with a difference per patient of $5 a day — so $20 in total (with the total cost being $26 if we add that day 1 panel cost). Now imagine there are 500 patients admitted per month into the unit who require IV fluids. Thats 500 x $20 (=$10,000) compared with 500 x $4 (=$2000). Now multiply that by 12 to get the yearly figures, and then by the number of acute care wards in the country and….well you get the point. The financial difference that results from being more prudent with test-ordering is immense. The consequences of not stopping to think if there could be a better, less wasteful (and potentially less harmful, if we consider the volume of blood lost) way of delivering care results in the opposite of value-based care. Mind you: this is one example, of which there are thousands — as such the it’s not surprising that the waste we’re considering is into the millions if not billions. It’s the price we pay for physician/institutional inertia.
Last example — the everyday patient. In June 2021 I was visiting upstate New York, and had a really interesting chat with a taxi driver (as an aside: some of my most interesting chats about healthcare happen to be with ride-share and taxi drivers). I’ll call him Dale, and the moment I entered his car he launched into a discussion about the American healthcare system. In his 70s, Dale is a blue-collar worker with a grade 10 education. He also served in the military for several years, including in Vietnam. Dale has several chronic health issues, including Type 2 diabetes that’s poorly managed and requires insulin, high blood pressure, and high cholesterol. Recently his Medicare coverage stopped covering his insulin. This led to a very frank discussion with his primary care doctor, and Dale explained to me that he had all but “given up” and accepted that ‘death was around the corner’ (without insulin a patient with uncontrolled diabetes can go into organ failure and die). It was a very tragic example and a story that has stuck with me. Here we have an example of where, despite the high cost of care per patient in the US, we still have millions of people like Dale, who actually need the spending, who fall through the cracks, and adopt an almost resigned/pessimistic view of their longterm healthcare.
All three of these stories help us understand the puzzle pieces behind what we know as “value-based care.” But there is one more story that’s crucial, and particularly topical now: that of physician ‘burnout’ (aka anxiety, depression, etc) secondary to buckling under the pressures of performance, including the expectation to provide ‘value based care.’ When I mentioned cognitive load and institutional inertia/physician inertia earlier, it was because we must also understand that both of these concepts are in term impacted by the well-being of the physician. A stressed- out, unsupported, demoralized healthcare professional is unlikely to have either the time or the energy to stop to re-evaluate if they are ordering the appropriate test, or more broadly if they are providing the best possible care — one that maximizes outcomes for the patient and minimizes costs (not to mention makes the physician ‘feel’ like they’re making a difference).
Value-based care, in other words, must also, as it’s core assumption, place ‘value’ on physician health and well-being. It should be easier to provide value-based patient care, not more difficult, and there is value inherent in ensuring the physician feels good, remains healthy and thriving before/during/after delivering patient care.
Now that I’ve [hopefully] painted the problem clearly, we can agree that the *system* has to change. But for it to change in a sustainable way, the tweaks can’t just be topdown — from government or insurers for instance. We need to be thinking more creatively, like an entrepreneur. We need to ask how, to paraphrase Buckminster Fuller, of ways in which we can make the old system of how physicians are incentivized to deliver care, obsolete. The only way to do that is to create a better system, one that provides incentives that align with what doctors value.
Is there a way to provide value-based care while optimizing physician health/well-being and cutting down on unnecessary time wasted on administrative tasks? Yes.
Is there a way to improve patient outcomes while lowering costs, saving the most money on complex procedures that are actually needed. Yes.
Could this also involve taking a more whole-person/patient approach to healthcare? Yes.
So what might this potentially ‘better system’ look like? That’s the topic of the next newsletter, and an exciting announcement I have about a pivot in my own professional focus.
Have a healthy, joyful, and safe month,
Amitha Kalaichandran, M.D., M.H.S.
Mild and dire forecasting models serve different purposes, and can be tricky to interpret. But when they appear similar, it may signal the end of the pandemic.
CONSIDER THIS THOUGHT experiment: J is a 55 year-old patient who has smoked two packs of cigarettes a day since he was 22. He has just been diagnosed with stage III non-small-cell lung cancer. His doctor uses a series of methods, including a model, to decide his prognosis.
In Situation 1, his doctor uses the “precautionary principle” and presents the worst-case scenario based on a model of the worst case: J has about six months to live.
In Situation 2, the doctor bases her prognosis on future-projecting J’s present situation, by definition not the worst-case scenario and more “optimistic”: J has another two years to live.
Which scenario is better?
The answer isn’t so straightforward. In medicine, prognostication is fraught with its own challenges and depends largely on the data and model used, which may not perfectly apply to an individual patient. More importantly: The patient is part of the model. If the information used then shifts the patient’s behavior, the model itself changes–more precisely, the weights given to certain variables in the model change either toward a more negative or positive outcome. In the first scenario, J may decide to shift his behavior to make the most of his next six months, perhaps extending it to nine months or longer. Does that mean the model was inaccurate? No. It does mean that knowledge of the model helped nudge J toward a more optimistic outcome. In the second scenario the opposite may happen: J may continue his two-pack-a-day smoking habit, or only cut down to a pack a day, which may hasten a more negative outcome. It’s entirely possible that J in Situation 1 lives for two years, and in Situation 2 lives for six months.
This pattern exists everywhere, from prognosticating climate change to even polling (knowing poll results can affect voting behavior, potentially changing the outcome). We’ve seen a similar dilemma with Covid-19 pandemic modeling, which may help explain the divisiveness over everything from when the pandemic may end to whether lockdowns are appropriate. Last year, just as the World Health Organization declared Covid-19 a global pandemic, I wrote about uncertainty and risk perception. When faced with uncertainty we defer to experts, but a month later the National Institute of Health’s Anthony Fauci correctly noted that experts are fraught with predicting what was (and still is) a “moving target.”
Over the past few weeks we’ve seen more opinion pieces focused on optimism: that herd immunity will be reached by April, and summer will be more like 2019, wide open and carefree. We’ve also seen how this optimism, based on a “present-day accurate model” can sway behavior: from schools opening (but then locking back down) to Texas’ recent removal of its mask mandate potentially contributing to an uptick in cases. Others have taken a more pessimistic approach, saying it may be another two years until things “return to normal,” and the virus variants are a “whole other ballgame.” Today, in Michigan and in Canada, a potential variant-fueled third wave suggests a less optimistic outlook (for now). We’re all deeply familiar with how this pattern has repeated itself several times over the past year, and even experts disagree (and some have changed tack). It’s more than just bad news bias. But how do we reconcile this dichotomy between the “optimists” and the “pessimists”? It may come down to how we understand the purpose of epidemiological models in general, and the two types of pandemic forecasting models.
Justin Lessler is an associate professor of epidemiology at Johns Hopkins University and is part of a team that regularly contributes to the Covid-19 Forecast Hub. He specifies that there are four main types of models: theoretical, which help us understand how disease systems work; strategic, which help public officials make decisions, including to “do nothing”; inferential, which help estimate things like levels of herd immunity; and forecasting, which project what will happen in the future based on our best guess how the response and epidemic will actually unfold.
When it comes to forecasting models, there are those whose forecasts are not worst-case scenario by definition (thus more optimistic), which aim to describe present-day patterns in transmission and susceptibility and project out, assuming the current patterns stay the same. In these “dynamic causal models” a variety of different variables are added to also include, as University College London based biomathematician Karl Friston described, unknown factors that affect how the virus spreads, dubbed “dark matter.”
Then there are forecasting models guided by the “precautionary principle,” aka “scenario models,” where the assumptions are often the most conservative. These account for the worst-case scenario, to allow governments to best prepare with supplies, hospital beds, vaccines, and so forth. In the UK, the government’s Scientific Advisory Group for Emergencies focuses on these models and thus guides policy around lockdowns. In the US, President Biden’s Covid-19 task force is the closest equivalent, while the epidemiologists and actuaries that appear nonconformist may be the closest we get to a group like the Independent SAGE (which Friston works with).
“The type of modeling we do for the Independent SAGE is concerned with getting the granularity right, ensuring the greatest fit–with minimal complexity–to help us look under the hood, as it were, at what is really going on,” Friston told me. “So, the fundamental issue is namely, do we comply with the precautionary principle using worst-case scenario modeling of unmitigated responses, or do we commit to the most accurate models of mitigated response?”
This gets to the heart of the tension between various “experts.” For instance, epidemiologists like Stanford’s John Ioannidis have tended to be more concerned with modeling the pandemic to accurately explain current patterns (and extending this pattern into the future), which can come off as more optimistic and isn’t typically used to guide policy.
**Originally published in Wired, March 2021**
How to make sense of recent concerns about the AstraZeneca vaccine
Last week, several European countries paused their use of the AstraZeneca vaccine due to concerns about clotting and bleeding risks. Though the World Health Organization (WHO) and European Medicines Agency (EMA) have both said that it is safe to use, most countries have resumed using the vaccine, and the company released data on Monday showing it is 79% effective in preventing symptomatic disease in the United States, many people may still be wondering about the risks. There are five major things to clear up when understanding the concerns about blood clots.
When most people think of blood clots, they think of a scab on the skin or clots in menstruation: congealed, thickened blood. In medicine, we’re talking about something more serious, involving the blood that circulates in our veins and travels from the tissues to the lungs to get reoxygenated. Blood clots are a general term for what’s known as deep vein thrombosis (DVT) and pulmonary embolism (PE).
Think of DVTs as blood clots that are often found in the calves or in the arms. Sometimes they resolve on their own, but they become dangerous when they break off and travel through the circulation and into the lungs, causing a PE, which in turn causes chest pain, decreases oxygen, and can lead to death. Sometimes DVTs can break off and travel backward to the heart and through the body again, making their way into the brain and causing a stroke. This is called a paradoxical embolism. A more rare clot in the brain is called a cerebral venous thrombosis (CVST). CVSTs may be the main clot of concern associated with the AstraZeneca vaccine. DVTs, PEs, and CVSTs are medical emergencies.
Most of the time, blood clots form in order to help us heal from wounds — injured tissue, internally or externally. Their formation involves the “coagulation (fancy word for clotting) cascade,” which comprises the extrinsic pathway, intrinsic pathway, and common pathway. The extrinsic pathway refers to factors in the coagulation cascade that are external or extrinsic from blood when studied in a test tube. The intrinsic pathway refers to factors in the cascade that are found in the blood when studied in a test tube.
These pathways require many components to work together effectively, including various clotting factors, most of which are named using Roman numerals and some that aren’t, like protein tissue factor (TF) and Von Willebrand factor (VWF). Other proteins block abnormal clots from forming, so they are said to have “anticoagulant” effects. These include Protein C, Protein S (both work with Vitamin K), and antithrombin III.
Some individuals bleed more easily than others. This can be due to deficiencies in coagulation factors — Factor VIII and Factor IX deficiencies, for instance, cause hemophilia, as does a deficiency in VWF. Other people have a lower platelet count. Since platelets are important to forming a “clotting plug,” which helps prevent blood loss by temporarily sealing an injured blood vessel, a dip in platelets often means bleeding risk may increase.
Glad you asked. First, anyone with a deficiency in an anticoagulant is at risk. Put another way, anyone who doesn’t have clotting blockers or who clots easily is at risk. An individual with antithrombin III deficiency, for example, would typically clot more easily.
But someone can have perfectly normal coagulation factors and a perfectly well-oiled coagulation cascade and still be at risk. Many athletes (as I’ve written about previously) fall into this category. This brings us to Virchow’s triad. Over a century ago, the German scientist and physician, Rudolf Virchow, described three components that increase the risk of a blood clot.
The first is “venous stasis,” which refers to moments when the blood sitting in our veins is stagnant. Imagine honey or ketchup in a squeezy bottle that’s stuck because it’s been sitting around. The way ketchup or honey congeals is similar to how stagnant venous blood forms. Except in the body, this can lead to a clot. In humans, this happens when we are stagnant. Long flights where we aren’t moving around is a common situation, but so is lying in a hospital bed for days on end, which is why many patients receive a blood thinner and are encouraged to move around.
The second component is vessel injury. If a blood vessel gets injured, the body responds by forming a clot, much as it would if you injure your skin through a scrape or a dog bite. Except when this happens in the body, there’s a chance the clot can become large and break off, blocking vessels and preventing blood (and therefore oxygen) from reaching the tissues, which can be deadly when it comes to the lungs or brain. These blood vessel injuries often happen during surgery.
The third factor involves other factors that increase hypercoagulability, which can refer to everything from cancer to inflammatory disease to being on estrogen hormone therapy (like the birth control pill). The mechanisms vary, but they are generally due to the impact on components of the coagulation system that drive it toward more clotting and away from anti-clotting.
Everything! We’re almost there. Let’s get some facts straight first. First, the incidence of DVT and PE, due to the issues described above, is about one per 1,000 people per year. For CVSTs, it’s even more rare: five per 1 million. This is the normal pre-pandemic and pre-vaccine incidence and reflects individuals at risk due to Virchow’s triad and issues with their coagulation system.
Back to the vaccines. Robust vaccine monitoring systems in many countries specifically look for potential adverse events after the vaccine, as part of what is called “active surveillance.” In general, however, we don’t have active surveillance for blood clots. No one calls families randomly to ask if anyone has had a blood clot. So, the fact that about 37 people who got the AstraZeneca vaccine have reported blood clots, out of 5 million who received the vaccine, doesn’t necessarily mean it’s caused by the vaccine. In all likelihood, these same 37 people would have had the same blood clot even if they weren’t vaccinated. And this is likely, given that the rate isn’t particularly high, compared with the baseline risk of blood clots. While the year still has nine months left, the current rate is about 0.006 per 1000 people per year for clots in general, which is lower than baseline.
It’s possible, given that the AstraZeneca vaccine is generally easier to store and manufacture in larger volumes (e.g. by India), that more people in total have received it. If that is the case, it may seem like the AstraZeneca vaccine is associated with more clots compared to the other vaccines, but the reality could be that more people have received it, period.
The Pfizer/BioNTech vaccine has been given out in 72 countries, and AstraZeneca to 71 as of March 18, but the number of people who have received it in those countries is not known. If each vaccine were distributed with the same frequency, it would be much more straightforward to compare the rate of adverse events, and it’s possible we would see the same pattern with them (which isn’t much of a pattern at all if it’s less than or equal to the baseline risk).
This is where the Bradford Hill criteria of causation comes in. They essentially say that temporality — the fact that an outcome comes after an exposure (in this case, an adverse event comes after a vaccine) — isn’t sufficient to prove causality, for the same reason that wearing a yellow T-shirt a few hours before the sun comes out doesn’t mean your T-shirt caused sunshine. We need more. Specifically, a biological gradient and plausibility: A biological explanation for the cause, much like we know that smoking causes lung cancer because the elements in cigarette smoke are known to be carcinogenic (even in a lab, they can cause mutations in lung cells that result in cancer).
Now that you’re an expert in clotting and causality, we can ask three crucial questions.
The first is whether the incidence of blood clots is statistically significantly higher among those that received the AstraZeneca vaccine compared to those that received no vaccine or another vaccine. (Statistically significant means that it’s unlikely to be due to chance.) Here’s the easiest way to think of it: In a random sample of 1,000 individuals, half of whom received the AstraZeneca vaccine and half of whom received another vaccine or no vaccine, does the AstraZeneca group show a statistically significant increased incidence of DVT, PE, or CVST? When testing a large number of rare events, the Bonferroni correction must also be applied to avoid the erroneous finding of statistical significance when testing several things, which apparently was missing from the EMA’s initial work.
The second is whether the dip in platelets observed in people who got the AstraZeneca vaccine is different from what is seen with other vaccines and viruses. Viruses, in general, can sometimes cause temporary dips in platelets (known as thrombocytopenia), and vaccines that are made from inert viruses may also do this. Though they usually cause a mild decrease in platelets, a severe decrease can be concerning and can cause a paradoxical overactivation of platelets, which can cause clots.
The third is whether there is a component in the AstraZeneca vaccine that would impact the coagulation cascade, specifically the hypercoagulability element of Virchow’s triad. This seems unlikely as most vaccine adjuvants (which boost the “immunogenicity”) and stabilizers are inert, meaning they don’t have medicinal or biological impacts. Alternatively, finding other biological mechanisms to explain the body’s abnormal response to the vaccine is also possible.
In summary, it’s unlikely that the clotting issues discovered by active surveillance are caused by the vaccine. However, it’s understandable why some countries are pausing vaccine administration until the above three questions, and possibly others, are answered.
The WHO continues to back the vaccine, while the EMA simply wants to add a warning, and countries like Canada are considering updating its guidance. The crucial thing to understand is that in a battle of risks, the harm from halting a vaccine campaign aimed at putting a stop to a deadly pandemic, which has a risk of mortality and long-term complications, appears to be much higher than the risk of blood clots.
A physician’s suicide reminds us that the plague of COVID-19 creates deep emotional wounds in health care workers
One of the oldest tales in the history of medicine is the story of the archetypal “wounded healer,” Chiron. As legend goes, Chiron, an immortal centaur, who both taught medicine and served as a physician, attended a gathering hosted by another centaur named Pholus. After a series of events involving other centaurs fighting over wine, Heracles (aka Hercules), in his attempt to intervene, accidentally unleashed a poisoned arrow that hit Chiron’s knee. Chiron, being immortal, was forced to endure unbearable pain.
Despite his ability to heal others, Chiron was unable to heal himself. Filled with shame, he retreated back to his cave, still committed to teaching his disciples. Eventually, after nine days, his pain became unbearable and Chiron requested that Zeus remove his immortality so he could die. Though a myth, it serves as the first documented story of a physician suicide, albeit assisted, and suggests that the challenge of healing our healers stretches back centuries.
The recent suicide of Lorna Breen, an accomplished and compassionate physician, researcher, colleague, friend, sister and daughter, after she served on the front lines of a busy New York City emergency department, reminds us that the plague of COVID-19 also creates deep emotional wounds in health care workers. As her father Philip Breen described her, she“was like the fireman who runs into the burning building to save another life and doesn’t regard anything about herself.” Her death was not due to COVID-19; it was due to a system and culture of hospital medicine that failed to value her as a human beyond her profession.
Right now, COVID-19 is a stress test, exposing the vulnerabilities in our financial, social welfare and health care systems. But it’s also a catalyst, giving rise to novel solutions such as providing a guaranteed basic income, expanding blood donation eligibility, reducing bureaucracy in hospitals and encouraging partnerships between tech companies. As such, it must also be a catalyst for improving medical culture so that one day no physician is forced to choose suicide as a result of an inability to cope or seek healing for themselves.
Awareness of the suicide epidemic plaguing the profession has gained ground over the last five years. Doctors have the highest suicide rate of any profession: about 300 doctors die each year in the United States (the size of a typical medical school student body). Effectively, suicide has now become an occupational hazard of the profession. But it’s also the canary in a coal mine serving as a warning for an overwhelmed and unhealthy system, one that doesn’t care for its doctors.
One thing is painfully clear: physician suicide isn’t about resilience. Doctors by definition are resilient; we must be to jump through many hoops to gain admission, serve on long overnight calls often without food, water or sleep, and work unreasonable work hours, often with an inadequate support system. Sadly the overemphasis on individual resilience at the expense of ensuring the work environment is healthy has placed the onus on doctors themselves—which is nothing more than victim-blaming.
While substance use and mental illness may be factors, many doctors do not have a diagnosed mental health disorder like depression and anxiety. This may, in part, be due to stigma around seeking a formal diagnosis, but we also know that symptoms of depression are wildly dependent on the environment; the influence of our situation on our reactions has been understood by sociologists for decades.
While things like mindfulness help to a degree, it’s a lot like expecting a soldier to meditate while bombs are being dropped all around her. The priority must instead be to get that soldier into a safe space with a battalion she can rely on, with the appropriate protective gear. Putting an otherwise healthy person, someone who is driven, intelligent, empathetic, in an environment that is not conducive to her well-being will place additional pressures on her with little room to thrive, or possibly even survive. The consequences can be disastrous, but are not surprising.
The problem of physician suicide is so deep, and the role of culture so paramount, that pontificating on solutions often feels futile, especially as the issue isn’t so much what the solutions are, but how to actualize them.
Culture must change from the top down, and this takes sound policies and commitment. Policies must include limits on work hours, time for self-care, and zero tolerance for bullying and harassment. We must also increase psychological safety (defined by Harvard scholar Amy Edmondson as “a climate in which people are comfortable expressing and being themselves”), a matter that is a pressing issue during the pandemic, as with the firing of doctors in Mississippi who have voiced concerns.
We should also ensure that all physician health programs are free of conflict of interest, completely divorced from licensing bodies, and accessible both geographically and financially. During a crisis especially, as we know from humanitarian aid workers, reentry trauma is common, and so access to these programs now is paramount in order to offset the risk of suffering alone. Isolation is an unsafe breeding ground for trauma, anxiety, and unprocessed grief.
Beyond telling the story of Chiron’s death, the ancient Greeks came to see suicide as primarily due to malfunctional “humors”—the end result of the build-up of black bile (melancholia) or yellow bile (mania). The beauty of medical knowledge is that it evolves; so too must our understanding. We must take lessons from as far back as Chiron, and as recently as Lorna Breen, to understand that environmental factors matter much more than the individual. Breen’s passing during this pandemic offers us a moment to reflect on how best to use our outrage and mourning, as patients and physicians, to finally move out of the clouds of ignorance, willful blindness and institutional inertia to prevent the same tragedy for repeating itself.
Once Chiron died, he left two legacies. The first was in those he taught: like the father of medicine, Asclepius, who in turn was said to have taught Hippocrates. Thousands of medical students take the Hippocratic oath each year. The second legacy, according to the poet Ovid, was through a gift from Zeus, who wanted to ensure Chiron’s spirit lived on in the night sky, so he created the constellation Centaurus—what may now be viewed a literal interpretation of the saying per aspera ad astra (“through hardship, to the stars”).
It shines brightest during the month of May. This year it might remind us of the thousands of physicians who took their own lives while healing others—some during this pandemic—doctors who might inspire us to finally change direction. And for Breen, as one of those bright stars, may we also vow to honor you as the hero you were, illuminating our path forward.
If you are having thoughts of suicide, call the National Suicide Prevention Lifeline at 1-800-273-8255 (TALK) or go to SpeakingOfSuicide.com/resources for a list of additional resources. Here’s what you can do when a loved one is severely depressed. For physicians on the front lines expressing mental distress or suicidal thoughts, or who just wish to talk, call the Physician Support Line 1-888-409-0141, which is open 8am to 3am ET, seven days a week, and provides free and confidential support with a volunteer psychiatrist.
**Originally published in Scientific American**
Let’s get clear on what the problems really are, then divide and conquer.
Recently I recalled one of the most crucial things I learned in medical school: the power of the “problem list.” Each patient came to us with a diagnosis, which was the reason for hospital admission. But as our attending made clear: the easiest and most efficient way to address the condition was to separate it into its component parts. It no longer becomes “let’s manage this patient with dementia,” it becomes “let’s sort out what the smaller problems are that make up the bigger challenge of treating this dementia.” We then understand how each smaller problem feeds into the larger one, which ultimately leads to appropriately managing the patient’s disease.
Tackling COVID-19 in America is an overwhelming and gargantuan task with no clear pathway, with everything so far pointing to failure. Much like how psychologists recommend “chunking” for learning, parsing out a big problem into a smaller set of problems helps us organize our thoughts, delegate tasks appropriately, all while making sure we’re not overlooking anything. It also helps us create contingencies and monitor progress.
If America was a patient, this would be her problem list and items to delegate:
1. Unclear case definition and endpoints
Infectious disease experts must help us more clearly define what a COVID-19 case looks like: the virus attacks the respiratory system, but other systems, like the gastrointestinal system, may also be affected. It also appears that the inflammatory response (how the body responds to the virus) as opposed to the virus itself, may be the primary cause of mortality, and this dictates treatment. Given that the tests available are imperfect, a negative test, in the presence of symptoms, should be treated as a presumptive case. Once a case definition is established as universal, it should further be stratified as mild, moderate and severe, with objective criteria defining each. Our metrics of response success must also be determined: COVID-19 is an unprecedented pandemic that is positioned to barrel through the U.S. and kill anywhere from 100,000 to 1 million people. Is a successful response one that cuts the most conservative projections by half? And are we more concerned about minimizing infections (of which most will be mild) or is the bigger priority to minimize the number of deaths?
2. Confusing public health messaging
Clear public health messaging is a challenge especially during times of uncertainty. Currently the messaging on whether transmission can occur through the air remains inconsistent between the World Health Organization, Centers for Disease Control, the White House, and state governments. This contributes to the spread of the “uncertainty virus” and mistrust, not unlike in a hospital when multiple teams are involved where medical errors are often secondary to communication issues, and was especially so with the case of masks. Groups like Choosing Wisely have disseminated some evidence-informed, best practices but clear public health messaging needs to be centralized. The White House should delegate one expert, ideally Dr. Anthony Fauci, to disseminate up-to-date public health information clearly and succinctly while also communicating uncertainties. Editors of major media in print and online challenged with this crisis will also play a key role in presenting consistent and reliable public health messaging. For months experts as well as the media underestimated the threat of COVID-19, and while contrarian views can help dissuade groupthink and tunnel vision, they risk undermining public health best practices and expert consensus. It is not a black swan. Rather, it was a dark horse: an underdog, one we were too blinded to see coming. We’ve seen dark horses before.
3. Insufficient testing
We need to clarify what we mean by “ramp up testing.” Tests should be two-fold: of secretions for the presence of the virus (presence/absence, and quantitative viral load data if possible) through a swab of the oral cavity and serologic testing for protective antibodies (which dictates prior infection, and likely protection) ideally with a fingerprick test. At this stage home-based testing might make the most sense, and it’s crucial to test a number of candidates against the gold-standard hospital-based test. An ideal test kit might have: a link to an online symptom checker, the swab and fingerprick test, and a self-addressed return envelope to mail back the test to a state lab. Once a kit (which would be priced at $0 to the public) is created, a partnership with Amazon (similar to what was struck in Canada) might make the most sense, given their warehousing and shipping capabilities, but we must ensure their delivery workers are provided with protective gear. Additional tests should be disseminated to homeless shelters. The tests won’t reach everyone, but capturing at least 75% of the population should be enough. As a metric we must set a benchmark for the number of Americans we want tested by April 30th.
4. No clear clinical pathway after a positive test
In China, positives were quarantined away from home. Had we organized early enough we could have used empty hotels for this purpose. Instead we should model symptom monitoring recommendations after asthma action plans, which are based on the traffic-light method. An expert committee — possibly from the American Academy of Emergency Medicine — could create a similar system (for instance, including symptoms like fever for a specific number of days, shortness of breath, and so on) so that those with a positive test know when to go to the hospital. We have enough data now, based on thousands of cases, to create this system.
5. Challenges with logistics, manufacturing, and procurement
While exciting, searching for a vaccine is not the biggest issue right now. Instead, it’s logistics, manufacturing and procurement, and this requires organized and thoughtful public-private partnerships. To be clear, the Defence Protection Act must be formally implemented with clear directions for the manufacturing of ventilators (ideally portable bedside ventilators as these would work better in make-shift hospitals without ready access to outlets), n95 masks, face shields, and gowns for healthcare workers. But currently this is highly decentralized which contributes to chaos: so formally involving the Defence Logistics Agency will also be key. Delegating these tasks to a few major companies who have the ability to manufacturer and ship their products quickly is crucial. We must also set clear pricing: a Forbes investigation recently found the inability to effectively negotiate contributed to the undersupply. Companies like Apple have people skilled at negotiation and procurement, and could offer their most skilled specialists to assist in ensuring we get supplies we need for the next 2–3 months, which appears now to be a focus. Outside the box solutions such as mask sterilizing systems should also be scaled up as well.
6. Lack of a universal policy, treatment, and end-of-life algorithms
We don’t have expert consensus on institutional infection control policy, nor treatment, discharge, or end-of-life best practices. As such, we should consider rapidly adopting a universal infection control policy modelled on Partners Healthcare and have an expert team, perhaps from the Society of Critical Care Medicine use the currently available evidence to create an algorithm for care, stratified by mild, moderate and severe. While imperfect, it will provided a road map that can be refined as we learn more, and would replace the informal crowdsourcing of best practices on social media. An ideal algorithm should dictate the parameters for oxygen, what starter therapies (medications and fluids) might help, criteria for mechanical ventilation (and settings), when to provide experimental treatments (e.g. chloroquine and remdesivir) for compassionate or trial us, and when to discuss comfort care. While abiding by infection control practices, everything possible must be done to allow family members to be present with their dying loved ones — walkie talkies goodbyes aren’t enough. Eack patient that enters the hospital with COVID should have an advanced directive regardless of how severe they are on admission. Given that some deaths have occurred after discharge, every COVID patient released from hospital must have a clear set of criteria on what to do at home, and when to return.
7. Unprotected healthcare workers & whistleblowers
Though doctors may be enlisted, many are struggling with their duty to serve, preparing their wills, and protesting seemingly to deaf ears for personal protective equipment (PPE). Thousands of healthcare workers around the world have died, including at least two resident doctors. The death of New York City-based Dr Frank Gabrin, himself a proponent for physician wellness, need not be in vain. We must have PPE for each healthcare provider, replaced at least once a shift, while also allowing for sufficient recovery time between shifts (in New York, having doctors and nurses serve from around the country helps with this). Punishing whistleblowers was seen first in China and but is creeping up in the U.S among healthcare workers and the military — this reprisal demonstrates a lack of psychological safety which will only worsen outcomes. Everything must be done to protect those that speak up.
8. Scattered research and no centralized database.
While it’s promising to see so much research on therapeutics happening all over the world — snippets shared over social media are mostly of case reports and small trials. We must create a central research database of existing studies — Stanford has a good starting model. Many research questions still remain. We could also leverage electronic medical record systems to help central database of diagnoses, clinical course, and outcomes.
9. Exacerbations of existing inequities
As with any patient, the social history cannot be forgotten. We need to get clear on what Americans with chronic health conditions should do if they can’t get care as they are at risk for dying due to lack of care during this crisis. We must also make every effort to protect and serve the most vulnerable who are at higher risk of poor outcomes — African Americans, those in the South, as well as the homeless and the undocumented (who may often be ‘essential’). Indeed as Alexandra Ocasio-Cortez tweeted last week, “inequality is a comorbidity.” It will be a stain on this nation if this crisis further perpetuates existing inequities. Ensuring healthcare during this time is accessible and universal, as recently underscored by the WHO is key, and could be inspired by other promising social experiments.
10. No clear plan for the “echo pandemic” of mental illness and social unrest
We are beginning to see an echo pandemic of mental illness and we may also see a rise in social unrest the longer we stay in lockdown. We must plan for both of these. To start, mental health experts should, where possible, offer services virtually. City planners must prepare for a possible surge in domestic violence, looting, and rioting. Notably, given the policy around face coverings, many perpetrators of public crimes may be difficult to identify.
This is America’s problem list; it is by no means comprehensive but it might be a starting point to help a strong leader delegate tasks. We can benefit from post-mortems from SARS and study the pandemic response now. As economist Daniel Kahneman popularized, we should also consider creating a premortem — anticipating how our response will fail helps us prioritize an action plan. The first step in any situation and assessment is realizing that one big problem is really a set of smaller problems and progress involves working diligently to address each component part. The intent is not to oversimplify but to make the task of battling COVID-19 more manageable while minimizing decision fatigue and maximizing public trust.
The time is now to divide and conquer. COVID-19 is not a drill. It’s a bitter pill.
**Originally published on Medium [visit for hyperlinks/citations]**
Some of America’s biggest companies should consider leveraging their logistical capabilities—from using drive-thru windows for screening to turning megastores into diagnostic and treatment centers—as part of their corporate social responsibility, during these dire times.
Dear CEOs of McDonalds, Apple, Nike, and Marriott:
As you probably know, the success of both China and South Korea in decreasing the number of new cases of COVID-19 required both social distancing but also widespread testing and isolation of confirmed cases away from their homes. In other instances, testing even more aggressively made a big difference, and the World Health Organization now strongly recommends expanding COVID19 screening as well as isolation. Italy may have waited too long to implement crucial measures and North America has lagged behind for some time: estimates show that the US is now less than two weeks behind Italy and extremely behind in COVID-19 testing.
Testing is not widely available in the US and Canada, with the spread of misinformation leading symptomatic people to head to their local hospital or family doctor to try to get tested (with limited success while overburdening the system). It’s even more dire knowing that, in New York City for instance, an estimated 80% of ICU beds may already be occupied.
As powerful corporations, I hope you consider leveraging your own logistical capabilities, as part of your corporate social responsibility, during these very dire times—particularly in hotspots like Seattle, San Francisco, Toronto, Vancouver, and New York City. Here are some suggestions for what you can do during these perilous times.
Over the past week, McDonald’s announced they are closing seating. There are over 14,000 McDonald’s in the US alone, most of which have drive-thru windows.
So, my first idea involves pausing fast-food manufacturing for a few weeks in some of these outlets and using the existing drive-thru infrastructure for in-person fever screening (window 1) and COVID-9 throat swabs (window 2, if fever is present). These could be staffed with local nurses (wearing personal protective equipment, or PPE) who might typically work in community clinics that are currently closed. The brand recognition of McDonald’s means that most North Americans would easily be able to locate their nearest franchise. These would effectively serve as “Level 1” screening and diagnostic facilities for the next several weeks, with repeat testing weeks later to assess when an infection has cleared.
Second, over the past week, Apple (which has 272 stores in the US) and Nike (which has 350 stores) have closed their stores. Both of these stores, which maximize negative space and average several thousand square feet (so up to 4.5 million square feet of unused space), have design elements that may help reduce transmission during a pandemic. Some of these stores could be refashioned to serve as “Level 2” diagnostic and treatment centers, for more in-depth diagnoses and assessment of confirmed COVID-19–effectively “cohorting” positive cases together. Also, since both Nike and Apple have longstanding manufacturing relationships with China, with independent shipping and warehouse capabilities, they could help store any donated medical supplies from China and the country’s business leaders. Doctors who are not currently skilled to work in an emergency department or intensive care unit (for instance, most general practitioners) could administer the tests and basic treatment at these sites while wearing appropriate PPE, which offloads the burden on hospitals (which in turn serve as “Level 3” treatment sites for more advanced care). This could work better than military tents.
Third, China’s success in reducing transmission was in large part due to effectively quarantining cases away from their family (so as not to infect other family members). Yet building large quarantine centers, as China did, is not logistically feasible in North America. As such, now that there are fewer travelers, Marriott, which has wide reach across North America, could offer designated hotels in which to isolate the confirmed positives for 14 days to help induce “suppression.”
To be sure, North America should still follow the lead of both Britain and France by harnessing local manufacturing capabilities (which requires a Defense Protection Act), specifically for personal protective equipment like N95 masks, gloves, and gowns for first responders–this is even more crucial given the shortage. However, the bigger challenge will remain logistical. We may even end up having enough expensive equipment like ventilators (which may be used to serve multiple patients) if the milder cases are effectively identified and treated early.
I agree that “brands can’t save us” — but companies can leverage their strengths in collaboration with government. In fact, there have been countless examples from history of corporations pivoting to assist in public health challenges. The most prominent one that comes to mind is Coca-Cola. For decades, Coca-Cola offered its cold chain and other logistical capabilities to assist public health programs to deliver vaccines and antiretroviral medications, because donating money, simply put, just isn’t enough.
Through innovation, you’ve been able to place a thousand songs in our pockets, boast the largest market share of footwear, become the biggest hotel chain in the world, and serve as the most popular fast food company. Facilitating widespread screening, diagnostic testing, and facilitating the safe isolation and treatment of mild-moderate cases is not an impossible feat, especially if you work together with the healthcare system. Instead of allowing your brick-and-mortar businesses to sit idle please consider pivoting towards a solution in collaboration with government, as part of a coordinated and effective pandemic response.
Time is running out.
**Originally published in Fast Company on March 19 2020**
Canadian and international initiatives aim to apply AI to help solve global health conundrums
As we grapple with the coronavirus (COVID-19) pandemic, the pattern of viral spread may have been identified as early as Dec. 31, 2019, by Toronto-based BlueDot.
The group identified an association between a new form of pneumonia in China and a market in Wuhan, China, where animals were being sold and reported the pattern a full week ahead of the World Health Organization (which reported on Jan. 9) and the U.S. Centers for Disease Control and Prevention (which reported it on Jan. 6).
Dr. Kamran Khan, a professor of medicine and public health at the University of Toronto, founded the company in 2014, in large part after his experience as an infectious disease physician during the 2003 SARS epidemic.
The BlueDot team, which consists largely of doctors and programmers, numbering 40 employees, published their work in the Journal of Travel Medicine.
“Our message is that dangerous outbreaks are increasing in frequency, scale, and impact, and infectious diseases spread fast in our highly interconnected world,” Khan wrote via email. “If we want to get in front of these outbreaks, we are going to have to use the resources available to us — data, analytics, and digital technologies — to literally spread knowledge faster than the diseases spread themselves.”
In the past, BlueDot has been able to predict other patterns of disease spread, such as Zika outbreak in south Florida. Now its list of clients includes the Canadian government and health and security departments around the world. They combine AI with human expertise to monitor risk of disease spread for over 150 different diseases and syndromes globally.
BlueDot, as a company, speaks to the emerging trend of using AI for global health.
In India, for instance, Aindra Systems uses AI to assist in screening for cervical cancer. Globally, one woman dies every two minutes due to cervical cancer, and half a million women are newly diagnosed globally each year: 120,000 of these cases occur in India, where rates are increasing in rural areas.
Founded in 2012 by Adarsh Natarajan, the Aindra team recognized that, in India, mortality rates were high in part due to the six-week delay between collecting samples and reading pathology during cervical cancer screening programs. It was also a human resources issue: in India, one pathologist is expected to serve well over 134,000 Indians.
With the aim of reducing the workload burden and fatigue risk (misdiagnosis rates can increase if the reader is tired and overworked), Aindra built CervAstra. The automated program can stain up to 30 slides at a time and then identify, through an AI program called Clustr, the cells that most appear to be cancerous.
The pathologist then spends time on the flagged samples. Much like traditional global health programs, Aindra works closely with several hospitals and local NGOs in India, and hopes their technology may later be adopted by other developing countries.
“Point of care solutions like CervAstra are relevant to a lot of countries who suffer from forms of cancer but don’t have infrastructure or faculties to deal with it in population based screening programs,” Natarajan says.
Natarajan also points to other areas where AI is relevant in global health, such as drug discovery or assisting specific medical specialists in areas like radiology and pathology. Accenture was able to use AI to identify molecules of interest within 10 months as opposed to the typical timeline of up to 10 years.
The Vector Institute, based in Toronto, is also plugging into the potential of AI and global health. It works as an umbrella for several AI startups, some with a health focus and all aiming to have a global impact.
Melissa Judd, director of academic partnerships at Vector Institute, points to the United Nations’ sustainable development goals as a framework upon which to help orient AI towards improving global health. Lyme disease, for instance, is a global health issue that also comes up against the topic of climate change, and recently a Vector-supported AI initiative was able to identify ticks that spread of Lyme disease in Ontario.
Last December, the Vector Institute launched the Global Health and AI Challenge (GHAI) — a collaboration with the Dalla Lana School of Public Health to engage students from across the University of Toronto (from business to epidemiology to engineering) in critical dialogue and problem solving around a global health challenge.
The potential of AI for global health is immense. Major academic journals are also taking note. Last April the Lancet launched the Artificial Intelligence in Global Health report. By looking at 27 cases of how AI has been used in healthcare, editors proposed a framework to help accelerate the cost-effective use of AI in global health, primarily through collaboration between various stakeholders.
As well, a recent commentary in Science identified several key areas of potential for AI and global health, such as low-cost tools powered by AI (for instance an ultrasound powered through a smartphone) and improving data collection during epidemics.
Yet, the authors caution against seeing AI as a panacea and emphasize that empowering local, country-specific, technology talent will be key, as inequitable redistribution of access to AI technology could worsen the rich-poor divide in global health.
This warning aside, Khan with BlueDot is optimistic.
“We are just beginning to scratch the surface as there are many ways that AI can play a key role in global health. As access to data increases in volume, variety and velocity, we will need analytical tools to make sense of these data. AI can play a really important role in augmenting human intelligence,” Khan says.
**Originally published in CBC News**
Two recent US initiatives: the New York Times’ rare disease column and a TBS series called Chasing the Cure are pointing to an emerging trend in the media: the idea that medicine can crowdsource ideas to diagnose difficult cases. But, can it be used to help diagnose patients, and what are the potential pitfalls?
Reaching a correct diagnosis is the crucial aspect of any consultation, but misdiagnosis is common, with some studies suggesting that medical diagnoses can be wrong, up to 43% according to some studies. This concern was the focus of a recent report by the World Health Organization. Individual doctors may overlook something, draw the wrong conclusion, or have their own cognitive biases which means they make the wrong diagnosis. And while hospital rounds, team meetings, and sharing cases with colleagues are ways in which clinicians try to guard against this, medicine could learn from the tech world by applying the principles of “network analysis” to help solve diagnostic dilemmas.
A recent study in JAMA Network Open applied the principle of collective intelligence to see whether combining physician and medical students’ diagnoses improved accuracy. The research, led by Michael Barnett, of the Harvard Chan School of Public Health, in collaboration with the Human Diagnosis Project, used a large data set from the Human Diagnosis Project to determine the accuracy of diagnosis according to level of training: staff physicians, trainees (residents and fellows), and medical students. First, participants were provided with a structured clinical case and were required to submit their differential diagnosis independently. Then the researchers gathered participants into groups of between two and nine to solve cases collectively.
The researchers found that at an individual level, trainees and staff physicians were similar in their diagnostic accuracy. But even though individual accuracy averaged only about 62.5%, it leaped to as high as 85.6% when doctors solved a diagnostic dilemma as a group. The larger the group, which was capped at nine, the more accurate the diagnosis.
The Human Diagnosis Project now incorporates elements of artificial intelligence, which aims to strengthen the impact of crowdsourcing. Several studies have found that when used appropriately, AI has the potential to improve diagnostic accuracy, particularly in fields like radiology and pathology, and there is emerging evidence when it comes to opthamology.
However, an issue with crowdsourcing and sharing patient data is that it’s unclear how securely patient data are stored and whether patient privacy is protected. This is an issue that comes up time and time again, along with how commercial companies may profit from third parties selling these data, even if presented in aggregate.
As such, while crowdsourcing may help reduce medical diagnostic error, sharing patient information widely, even with a medical group, raises important questions around patient consent and confidentiality.
The second issue involves the patient-physician relationship. So far it doesn’t appear that crowdsourcing has a negative impact in this regard. For instance, in one study over half of patients reported benefit from crowdsourcing difficult conditions, however very few studies have explored this particular issue. It’s entirely possible that patients may want to crowdsource management options for instance, and obtain advice that runs counter to their physicians’ and theoretically this could be a source of tension.
The last issue involves consent. A survey, presented at the Society of General Internal Medicine Annual Meeting in 2015, reported that 80% of patients surveyed consented to crowdsourcing, with 43% preferring verbal consent, and 26% preferring written consent (31% said no consent was needed). Some medico-legal recommendations, however, do outline the potential impact on physicians who crowdsource without the appropriate consent, in addition to the possible liabilities around participating in a crowdsourcing platform when their opinion ends up being incorrect. Clearly these are issues that have no clear answer: and we may end up in a position where patients are eager to crowdsource difficult-to-diagnose (and treat) sets of symptoms, but physicians exercise sensible caution.
It’s often said that medical information doubles every few months, and that time is only shortening. Collectively, there’s an enormous amount of medical knowledge and experience both locally and globally that barely gets tapped into when a new patient reaches our doors in any given hospital or clinic. Applying network intelligence to solving the most challenging, as well as the illusory “easy,” diagnosis, may give patients the best of both worlds: the benefit of their doctor’s empathetic care with the experience and intelligence of a collective many, but the potential downsides deserve attention as well.
**Originally published in the British Medical Journal**
Here’s why communicating public health risk during an epidemic is so challenging
Ann, a friend and mentor in her 50s, exclaimed over coffee at the end of January: “You know, Amazon is sold out of medical masks. You just can’t get any now. But I’m going upstate this weekend, so I should have better luck there.” I looked at her quizzically. At the time, the World Health Organization (WHO) had not yet announced that the newly named disease COVID-19 (formerly known as 2019-nCoV), caused by the virus SARS-CoV-2 (or simply “coronavirus”) was a Public Health Emergency of International Concern (PHEIC), but this announcement was delayed for several days. Besides, masks should only be reserved for people with symptoms.
Ann is an intellectual, someone who doesn’t easily head into panic mode (this helped her in her law career immensely, and later as a CEO and business leader). But in that moment, she had made up her mind: the masks would be a prudent thing to purchase, despite the lack of indication that they were needed. Effectively, Ann was hedging on the idea that, with the messages she received through the media and her friends, it would be better to be more conservative and overly prepared for the worst, given the potential consequences of being underprepared.
It immediately struck me that, despite being trained in both epidemiology and medicine, I wasn’t entirely sure what to advise Ann at the time: the messages I had received, and articles I had read, were no more consistent. There was still much uncertainty around the coronavirus in terms of how serious it was projected to be and what ordinary citizens could do to minimize risk. We all make decisions every day despite uncertainty, and when emotions come into play it can make things trickier.
But when it comes to public health, where the risks of sending the “wrong” message can have devasting consequences—unnecessary anxiety on the one hand (which can take an immense psychological toll) and thousands of unnecessary deaths on the other. To me, one thing is clear: the messaging around coronavirus thus far has been far from ideal, which suggests that uncertainty in a public health emergency is a wrench that can have devastating consequences if it isn’t harnessed appropriately.
Coronavirus is a moving target, as most epidemics are. As a Canadian, I watched with curiosity when Canadian airports decided on January 17 not to screen travelers for coronavirus (the effectiveness of screening is debatable, but the U.S. had already mandated it). But this then changed a mere one day later. The messaging was all over the place: “We thought it wasn’t necessary, but oops, now it might be.” Initially, the WHO wasn’t as concerned: the information and data about coronavirus wasn’t enough to call it an “emergency,” perhaps in part because the institution was reliant on a whole host of assumptions, such as the accuracy of data from China, a country not exactly known for transparency (with some noting the government may have purposely misled the public).
Gradually, the WHO became more concerned, finally on January 30 labelling coronavirus as a PHEIC, which implies a seriousness and a whole other set of other measures should be taken. Now countries as far and wide as Italy, Iran, Korea, and Spain are reporting a high concentration of cases. As of Wednesday, February 26, over 2,700 people had died worldwide from coronavirus since December and over 81,000 were infected globally. To put that in perspective, the SARS epidemic of 2003, which began in November 2002, infected over 8,000 people and led to 774 deaths in a period of six months.
Today the core messages remain unclear. For instance, the WHO has refused to officially advise no travel to China, but the U.S. State Department made this advisory earlier this month. For weeks we also received mixed messaging about human-human transmission, which is now clear, and more disturbingly that it can occur even when someone isn’t symptomatic (though it is rare). Even epidemiologists had trouble deciding how bad it really is. One reason is that a traditional data point in epidemiology, the R0 value, which is the average number of people an infected person is expected to transmit a disease to, is limited in its predictability.
Still, several doctors and public health professionals have taken to social media to remind the world that the flu kills more, as an attempt to dissuade fears, but COVID-19 is more severe, not just in its the ability to send more affected persons into intensive care (like SARS), and that it can kill even young and healthy hosts (as opposed to the more vulnerable who are more affected by the flu), and by most accounts has a higher case fatality rate (the proportion of those with the virus who die), somewhere around 2 percent (though this rate may be lower—0.7 percent—outside of China’s Hubei province) compared to the flu (which has a case fatality rate of around 0.1 percent).
All of this whiplash points to one perhaps uncomfortable thing: no one really knows how bad COVID-19 is, and how much damage it could eventually lead to. We know from postmortems of how SARS and Ebola were approached—both epidemics that provided an opportunity for bodies like the WHO and the Centers for Disease Control to learn from (the CDC provided a report on their Ebola response, and the WHO released a report on outbreak communication immediately after SARS)—that waiting too long to sound the alarm can be disastrous. We also know that the early predictions were based on assuming that China was being transparent and honest about their situational assessment, something we now understand was not the case.
I recently spoke with Kathryn Bertram, of the Johns Hopkins Center for Communication Programs (JHU CCP), who pointed me to the extended parallel process model as a helpful starting point to examine public health messaging during an epidemic. It considers both our rational reactions and emotional reactions (primarily fear) to help determine the best course of action for behavior. On the rational end, we must ask ourselves about “efficacy”—this refers to the effectiveness of a solution (for instance wearing a face mask or avoiding travel to China) and well as our perceptions on how as individuals we can institute this solution effectively. On the emotional end, we ask ourselves about the severity—how severe might it be if we, as individuals were infected, as well as susceptibility (how likely we might contract it).
Herein lies the issue: the perceived threat rests largely on the information we receive from experts. If the threat is high, we make decisions to take protective action. If we are told that the threat is low or even trivial, we are less motivated to protect ourselves even if we have the resources to do so. When an epidemic is underway, uncertainty can create fertile ground for mixed messages and inconsistency, which in itself can breed mistrust and fear.
Reflecting back to my conversation with Ann, I’m reminded of Annie Duke’s book Thinking in Bets, in which she makes a persuasive argument that, as individuals, we’re often required to make decisions based on having incomplete information. Duke uses the analogy of poker, where decisions are made based on an uncertain future. A good decision, despite this uncertainty, rests on whether we use the right process to come to that decision.
As individuals, we also benefit from thinking back to situations where we may have chosen one way but felt if we had a similar choice again we would choose differently, so our memories play a role as well (and arguably for public health we can rely on our collective memory from other coronavirus epidemics, like SARS). She likens our decisions to bets: given the information available to us, along with our memories of how past decisions panned out, and acknowledging that some of the outcome is due to chance, what might be the best choice to make that would most likely provide the most benefit for our future selves?
Bertram underscores the core risk communication principles, which can also be applied to media covering the epidemic: communicate often, communicate what is and isn’t known clearly, and provide simple action items for individuals to take (so things like handwashing).
Similarly, public health stakeholders should communicate what is and isn’t known, coordinate messages to help ensure consistency, and perhaps most importantly, acknowledge that their views (and thus their messaging) may change quickly; thankfully more recently media organizations are choosing to express this uncertainty and a recent op-ed in the New York Times underscores many of these principles, as “people react more rationally and show greater resilience to a full-blown crisis if they are prepared intellectually and emotionally for it.” The authors also urge that we consider using the term “pandemic” (though the WHO is not yet comfortable with this).
Effectively, while the WHO still presents a hopeful view, it and other organizations played poker on a global scale—and the chips they were playing belonged to entire communities. Their decisions and messages matter, and on balance, it might be best to bet that the consequences of underestimating the severity of the pandemic may be worse than overestimating it. The alternative, which brings to mind the dog meme “this is fine,” could lead to both distrust and potentially thousands of unnecessary deaths. It seems that, despite the WHO finally conceding that COVID-19 continues to poses a “grave threat” to the world and may qualify as the long-dreaded “disease X,” the briefing yesterday remained vague and hesitant, and even domestic messaging about whether the virus is contained or spreading continues to be inconsistent. Some have even suggested we finally accept that COVID-19 may be “unstoppable.” Clearly, we’re still down a few chips.
**Originally published in Scientified American, on February 26 2020**
There’s more than meets the eye — here are some tips to help avoid confusion.
In August 2019, JAMA Pediatrics, a widely respected journal, published a study with a contentious result: Pregnant women in Canada who were exposed to increasing levels of fluoride (such as from drinking water) were more likely to have children with lower I.Q. Some media outlets ran overblown headlines, claiming that fluoride exposure actually lowers I.Q. And while academics and journalists quickly pointed out the study’s many flaws — that it didn’t prove cause and effect; and showed a drop in I.Q. only in boys, not girls — the damage was done. People took to social media, voicing their concerns about the potential harms of fluoride exposure.
We place immense trust in scientific studies, as well as in the journalists who report on them. But deciding whether a study warrants changing the way we live our lives is challenging. Is that extra hour of screen time really devastating? Does feeding processed meat to children increase their risk of cancer?
As a physician and a medical journalist with training in biostatistics and epidemiology, I sought advice from several experts about how parents can gauge the quality of research studies they read about. Here are eight tips to remember the next time you see a story about a scientific study.
1. Wet pavement doesn’t cause rain.
Put another way, correlation does not equal causation. This is one of the most common traps that health journalists fall into with studies that have found associations between two things — like that people who drink coffee live longer lives — but which haven’t definitively shown that one thing (coffee drinking) causes another (a longer life). These types of studies are typically referred to as observational studies.
When designing and analyzing studies, experts must have satisfactory answers to several questions before determining cause and effect, said Elizabeth Platz, Sc.D., a professor of epidemiology and deputy chair of the department of epidemiology at the Johns Hopkins Bloomberg School of Public Health. In smoking and lung cancer studies, for example, researchers needed to show that the chemicals in cigarettes affected lung tissue in ways that resulted in lung cancer, and that those changes came after the exposure. They also needed to show that those results were reproducible. In many studies, cause and effect isn’t proven after many years, or even decades, of study.
2. Mice aren’t men.
Large human clinical studies are expensive, cumbersome and potentially dangerous to humans. This is why researchers often turn to mice or other animals with human-like physiologies (like flies, worms, rats, dogs and monkeys) first.
If you spot a headline that seems way overblown, like that aspirin thwarts bowel cancer in mice, it’s potentially notable, but could take years or even decades (if ever) to test and see the same findings in humans.
3. Study quality matters.
When it comes to study design, not all are created equal. In medicine, randomized clinical trials and systematic reviews are kings. In a randomized clinical trial, researchers typically split people into at least two groups: one that receives or does the thing the study researchers are testing, like a new drug or daily exercise; and another that receives either the current standard of care (like a statin for high cholesterol) or a placebo. To decrease bias, the participant and researcher ideally won’t know which group each participant is in.
Systematic reviews are similarly useful, in that researchers gather anywhere from five to more than 100 randomized controlled trials on a given subject and comb through them, looking for patterns and consistency among their conclusions. These types of studies are important because they help to show potential consensus in a given body of evidence.
Other types of studies, which aren’t as rigorous as the above, include: cohort studies (which follow large groups of people over time to look for the development of disease), case-control studies (which first identify the disease, like cancer, and then trace back in time to figure out what might have caused it) and cross-sectional studies (which are usually surveys that try to identify how a disease and exposure might have been correlated with each other, but not which caused the other).
Next on the quality spectrum come case reports (which describe what happened to a single patient) and case series (a group of case reports), which are both lowest in quality, but which often inspire higher quality studies.
4. Statistics can be misinterpreted.
Statistical significance is one of the most common things that confuses the lay reader. When a study or a journalistic publication says that a study’s finding was “statistically significant,” it means that the results were unlikely to have happened by chance.
But a result that is statistically significant may not be clinically significant, meaning it likely won’t change your day-to-day. Imagine a randomized controlled trial that split 200 women with migraines into two groups of 100. One was given a pill to prevent migraines and another was given a placebo. After six months, 11 women from the pill group and 12 from the placebo group had at least one migraine per week, but the 11 women in the pill group experienced arm tingling as a potential side effect. If women in the pill group were found to be statistically less likely to have migraines than those in the placebo group, the difference may still be too small to recommend the pill for migraines, since just one woman out of 100 had fewer migraines. Also, researchers would have to take potential side effects into account.
The opposite is also true. If a study reports that regular exercise helped relieve chronic pain symptoms in 30 percent of its participants, that might sound like a lot. But if the study included just 10 people, that’s only three people helped. This finding may not be statistically significant, but could be clinically important, since there are limited treatment options for people with chronic pain, and might warrant a larger trial.
5. Bigger is often better.
Scientists arguably can never fully know the truth about a given topic, but they can get close. And one way of doing that is to design a study that has high power.
“Power is telling us what the chances are that a study will detect a signal, if that signal does exist,” John Ioannidis, M.D., a professor of medicine and health research and policy at Stanford Medical School said via email.
The easiest way for researchers to increase a study’s power is to increase its size. A trial of 1,000 people typically has higher power than a trial of 500, and so on. Simply put, larger studies are more likely to help us get closer to the truth than smaller ones.
6. Not all findings apply to you.
If a news article reports that a high-quality study had statistical and clinical significance, the next step might be to determine whether the findings apply to you.
If researchers are testing a hypothetical new drug to relieve arthritis symptoms, they may only include participants who have arthritis and no other conditions. They may eliminate those who take medications that might interfere with the drug they’re studying. Researchers may recruit participants by age, gender or ethnicity. Early studies on heart disease, for instance, were performed primarily on white men.
Each of us is unique, genetically and environmentally, and our lives aren’t highly controlled like a study. So take each study for what it is: information. Over time, it will become clearer whether one conclusion was important enough to change clinical recommendations. Which gets to a related idea …
7. One study is just one study.
If findings from one study were enough to change medical practices and public policies, doctors would be practicing yo-yo medicine, where recommendations would change from day to day. That doesn’t typically happen, so when you see a headline that begins or ends with, “a study found,” it’s best to remember that one study isn’t likely to shift an entire course of medical practice. If a study is done well and has been replicated, it’s certainly possible that it may change medical guidelines down the line. If the topic is relevant to you or your family, it’s worth asking your doctor whether the findings are strong enough to suggest that you make different health choices.
8. Not all journals are created equal.
Legitimate scientific journals tend to publish studies that have been rigorously and objectively peer reviewed, which is the gold standard for scientific research and publishing. A good way to spot a high quality journal is to look for one with a high impact factor — a number that primarily reflects how often the average article from a given journal has been cited by other articles in a given year. (Keep in mind, however, that lower impact journals can still publish quality findings.) Most studies published on PubMed, a database of published scientific research articles and book chapters, are peer-reviewed.
Then there are so-called ‘predatory’ journals, which aren’t produced by legitimate publishers and which will publish almost any study — whether it’s been peer-reviewed or not — in exchange for a fee. (Legitimate journals may also request fees, primarily to cover their costs or to publish a study in front of a paywall, but only if the paper is accepted.) Predatory journals are attractive to some researchers who may feel pressure to ‘publish or perish.’ It’s challenging, however, to distinguish them from legitimate ones, because they often sound or look similar. If an article has grammatical errors and distorted images, or if its journal lacks a clear editorial board and physical address, it might be a predatory journal. But it’s not always obvious and even experienced researchers are occasionally fooled.
Reading about a study can be enlightening and engaging, but very few studies are profound enough to base changes to your daily life. When you see the next dramatic headline, read the story — and if you can find it, read the study, too (PubMed or Google Scholar are good places to start). If you have time, discuss the study with your doctor and see if any reputable organizations like the Centers for Disease Control and Prevention, World Health Organization, American Academy of Pediatrics, American College of Cardiology or National Cancer Institute have commented on the matter.
Medicine is not an exact science, and things change every day. In a field of gray, where headlines sometimes try to force us to see things in black-and-white, start with these tips to guide your curiosity. And hopefully, they’ll help you decide when — and when not to — make certain health and lifestyle choices for yourself and for your family.
**Originally published in the New York Times**