A Real Good Feeling Something Bad's About to Happen

Thoughts on intuition in the hyper-rational halls of medicine

You’re in the woods at dawn, steam rising into the air as you breathe out. You’ve said prayers, made offerings, and now have spotted the tracks of a deer that has passed this way recently. Your clan is relying upon your prowess to make it through the winter. What are the skills, knowledge, and practices that guide you in a successful hunt? From what I’ve been told (having never hunted myself), hunting combines the keen observation of rational thought with intuition and spiritual knowledge. Success depends on science, knowing the creature, knowing the terrain, knowing the weather. It depends on technique with a gun, bow, or trap. However, for some who hunt, connecting to the spirit world, to the spirit of the hunted animal, to intuition about when, how, and where you will encounter the animal are all key parts of hunting success. To them, knowing and performing the right rituals and prayers is as important as the right equipment. The Emerald podcast captures this wonderfully and inspired me to reflect more on this topic.

This weaving of the rational and the intuitive was “just the way it is'' for much of history, woven into life to the point that the gifts of those who possessed it were cultivated and integrated, neither ostracized nor commercialized. In fact, it is “just the way it is'' now, if we let it be. We all have intuitive capacities to varying degrees, but have been convinced that the information gathered from those capacities cannot be trusted. Intuition within the realms of healing and medicine are a particularly interesting example of this. Traditional healers have always included intuition as one of their “ways of knowing” how to tend to their communities. They often recognize emotional, social, and spiritual ailments as key parts of individual wellbeing and include them in their treatment plans. 

In its current form, many view medicine as hyper-rational. Many believe that a rational, objective means of diagnosing and treating people is the goal. However, the medical literature demonstrates that without “clinician gestalt” (that’s medical jargon for intuition), our diagnostic strategies are less accurate than when intuition is included in the picture. In fact, Wells’ criteria, used to determine the most appropriate modes of testing for diagnosing someone with a blood clot in the lungs, clinician gestalt is worth the most points in the score, the same amount of points as having symptoms of a blood clot in the leg, worth twice as many points as having a blood clot previously. Clinician gestalt isn’t perfect, but in experienced physicians, it becomes an important resource. In trauma patients, we are more likely to overestimate the risk of significant injury in those we think are likely to have severe injury and underestimate the risk in those we think are unlikely to have a severe injury. However, even in that same study, if an experienced physician estimated there was <2% chance of serious injury, they were 100% accurate that the patient did not have a clinically significant injury. This emphasizes the role of experience in using intuition. As students and residents, we tend to operate in a very rational, data-driven realm. We are swimming in a fish-bowl of evidence-based medicine and rational medical education culture, not knowing anything else is out there, and we don’t have enough experience to have real pattern recognition. Several experts, including Daniel Kahneman, who wrote Thinking Fast and Slow, suggest that part of intuition is a subconscious pattern recognition. I suspect intuition is more than simply pattern recognition that you cannot articulate, but I think that is certainly a component of it. 

As I have become more experienced in medicine, the application of medical evidence has become a source of tension for me. I can understand the research very well. That will tell me on average what will be most helpful for people who are like those who met the inclusion criteria (e.g. age range of participants, how ill they are, when and where participants are recruited) in that trial. How much the person sitting in front of me has in common with the people in those trials varies greatly. Should I apply this study, this algorithm, this treatment in the specific context I am in? Sometimes, like when I practiced in Uganda, living conditions and a patient’s ability to return for follow-up are sufficiently different from the urban US or Europe where most research is conducted, that the research needs to be generalized with caution. However, even within the US, studies often apply to relatively small segments of the population. As a very obvious example, the vast majority of medical research has been performed on men, because of this women are 50-75% more likely to suffer from adverse reactions from medications. In fact, I heard at a medical conference that women had more side effects for all ten of the most commonly prescribed medications. It’s pretty hard to take the average response across a population and use it to know with confidence what is best for the specific person sitting in front of you. Over time, you combine what you have learned from the data with what you know about a person’s living situation, goals, risk tolerance, support structure, and more to customize a plan for them. When you look at how much variation there is in each of these categories and how little research has been done on any of those things I just listed, trying to make a medical decision based on a randomized controlled trial alone without taking these factors into account makes about as much sense as trying to make medical decisions based on an astrological chart alone. It’s hard to compile these types of criteria rationally, make a pros and cons list, and make a decision. Sometimes the decision has to come down to shared decision making with your patient and a consultation with your gut. 

Head CT watercolor 3 (work in progress)

At work, we often talk about our “spidey sense” tingling. Sometimes, you “get a real good feeling something bad’s about to happen”, and for all of the hand-wringing about over-testing in the medical system, that feeling is right substantially more than it is wrong. I had a young adult COVID patient who was on the edge of needing to be hospitalized. I knew the moment I saw him that something bad was going to happen, so we admitted him. Several days later, I was putting in a breathing tube and he was on a ventilator for a couple months after that. Or there are the patients that you have seen dozens of times, that because of your experience with them you know something is “just off”. This is literally a daily occurrence, but it’s rarely discussed because it can’t be proven, often can’t even be articulated. 

This is one of my concerns about the increasing systemization of medicine. Every year there are more algorithms, more time goals, and now more AI. I think these tools do raise the lowest common denominator, by which I mean, if you are inexperienced, incompetent, or overwhelmed, it will help you provide adequate care. However, if you are not those things, the five extra minutes it takes you to determine if someone is really having a stroke and needs a CT scan and clot busting medicine or is really septic and needs IV antibiotics probably does not make that big of a difference to the outcome of the patients who are having the condition, but extra five minutes to avoid doing those things unnecessarily matters a lot to the patient who is not. I don’t know if you have tried it, but it is nearly impossible to tune into that still, small voice of intuition when you are being rushed. Just a couple weeks ago, a patient came in for arm weakness. He was old enough that it was reasonably likely he was having a stroke. After seeing the weakness was present, I called a Code Stroke and they whisked him away to the CT scanner. When he returned, and we talked more in depth, I realized it was related to a shoulder injury.

I wonder about this as more and more AI is rolled out in healthcare too. Sure it would be nice to be done with my documentation faster, but that could also be solved by having less onerous documentation requirements. I admit that I am not an expert in this topic, and I am an AI skeptic by nature. If we can only use the data and the rational in decisions guided by AI, what do we lose? And what do we lose in our faith in our intuition and its usefulness? I have enough self-doubt to know that if AI tells me I am wrong, I am almost certainly going to believe it/them. I am sure that there are ways that AI can save lives by recognizing patterns human brains don’t. I’m also sure there are ways that it can harm people, both patients and health care providers, immeasurably by magnifying the biases of the designers and only utilizing the rational part of human intelligence. With time, perhaps we could learn the strengths and weaknesses of both the human and AI systems and how to integrate them to get the best of both worlds. However, we’re not taking that time, and my intuition tells me this is going to be a big problem.