Why have you or have you not done this? Is it important to understand from direct observation how clinicians make decisions, often jointly with patients, often in a negotiation involving feelings, fears, misunderstandings, things patients have heard from family members or read online? Thoughts on this topic?
Biostatisticians, have you ever rounded in the hospital with a clinical team or observed some days of work in a clinic?
I work with ED docs and even just seeing how the chart data gets entered into the ehr would be extremely helpful. I think it’s a good idea but I have not done it.
My first jobs were working for physicians, where I helped patients and spent hundreds of hours doing chart abstractions and coding gastroenterological procedures into key-sort cards for an early research database. I went on to work in cardiology, spending time in the cardiac care unit and cardiac cath lab, which gave me perspective while programming signal detection algorithms for automatic cardiac catheterization hemodynamic analysis. For 19 years of my career my office was located in a cardiology group. I went on rounds only once, but these other experiences were highly valuable in building my abilities to collaborate with physicians. Regarding medical decision making, hundreds of hours of conversations with cardiologists and being the lead statistician on SUPPORT (Study to Understand Prognoses Preferences and Outcomes of Treatment) for 8 years (and end-of-life medical decision making study funded by the Robert Wood Johnson Foundation) were most valuable in teaching me about medical decision making. I also attended national conferences of the Society for Medical Decision Making and read many papers on the subject.
I believe it is extremely important for biostatisticians to spend time on the “front lines.” This is the way to learn how physicians use data, make decisions, correctly use probability, and misunderstand probability.
Very neat answers. I think it would be a good program to have biostatisticians interested in how physicians think accompany them on rounds or in clinics on many occasions.
These are the only settings in which physicians make many rapid decisions. There are some other settings, like in the office, when they get a notification that a patient has a question. They’ll make a decision or two at that time, as well, but this is more difficult to observe. And the process of decision-making is likely different.
For a medical student to learn how physicians think when it counts – under pressure, in direct discussion with patients – requires rounding hundreds of times. Does this type of stats rounding program / clinic program exist anywhere?
medical students attend, and we want to add funny-looking, curious biostatisticians to the onlookers? i’ve visited eg neonatology units to see equipment, data collection. i’d try to stay out of sight …
Only for those interested in / commenting on how physicians think: yes, I think they should have extensive experience observing it in the wild. If that’s not an area of interest, then no value, presumably, yep.
For example, you probably wouldn’t critique how firefighters think about putting out a fire without watching exactly how they do that work over a period of time. The baseline assumption would be: they have an approach they’ve developed over time that is effective for them. To offer a better approach would require deep understanding of their current approach, as practiced, I believe? The risk of skipping that step and offering advice anyway would be that the advice might be incorrect or impossible to implement?
Looking over at what Einstein wrote about the schism between epistemology and practical application of understanding (in science in his case), here’s what he said – partly motivating my question:
“The reciprocal relationship of epistemology and science is of noteworthy kind. They are dependent upon each other. Epistemology without contact with science becomes an empty scheme. Science without epistemology is—insofar as it is thinkable at all—primitive and muddled. However, no sooner has the epistemologist, who is seeking a clear system, fought his way through to such a system, than he is inclined to interpret the thought-content of science in the sense of his system and to reject whatever does not fit into his system. The scientist, however, cannot afford to carry his striving for epistemological systematic that far. He accepts gratefully the epistemological conceptual analysis; but the external conditions, which are set for him by the facts of experience, do not permit him to let himself be too much restricted in the construction of his conceptual world by the adherence to an epistemological system. He therefore must appear to the systematic epistemologist as a type of unscrupulous opportunist: he appears as realist insofar as he seeks to describe a world independent of the acts of perception; as idealist insofar as he looks upon the concepts and theories as free inventions of the human spirit (not logically derivable from what is empirically given); as positivist insofar as he considers his concepts and theories justified only to the extent to which they furnish a logical representation of relations among sensory experiences. He may even appear as Platonist or Pythagorean insofar as he considers the viewpoint of logical simplicity as an indispensable and effective tool of his research. (Einstein 1949, 683–684)”
If you replace “scientist” with “physician,” you’ll get a feeling for what I’m saying.
As a Cardiologist and a Biostatistician, I had the opportunity to observe and live the deficiencies of both professions. There is lack of communication. Neither a cardiologist/MD nor a biostatistician alone can conduct a good research. For example, since most statisticians are far from clinical terminology and do not experience clinical decision-making stages and thinking that they can do alone all stats without basic clinical experience. On the other hand Most MDs dont familiar to basic concepts of probability/stat and think that they can do a good research with basic statistical knowledge using user-friendly softwares. Cooperation between the two professions is compulsory and education is essential. Finally, let me state; MDs are actually thinking probabilistic or even Bayesian in their mind when deciding (they are unaware to do this), but in practice unfortunately they are classifier and frequentist. As a Cardiologist, thanks to the biostatisticians I follow here ( Frank, Andrew, Christos, Darren, Maarten, Senn…) I’ve started to improve my behavior in clinical practice.
I see some aspects of what you’re saying.
For example, I think most physicians decide whether or not do a stress test–and nearly all other tests–based upon consideration of pretest probability & the extent to which the result is likely to alter the perceived probability of a disease or other actionable state of health. I’m not sure this is easily or widely understood unless one is with physicians who are making decisions in real-time, since the literature tends to suggests that everything is viewed in a frequentist mode.
Most of all, I would second your point about neither group being able to operate successfully without the other. Also, and this is important in my opinion, sweeping criticisms of how people think are antithetical to improving the current situation IMO. Even if those critiques are informed by extensive direct bedside observation of clinicians–is this usually so?–sweeping critique of the other party’s thinking style precludes serious collaboration in my experience. Food for thought in the ongoing schism between clinicians & statisticians, fueled in my opinion by the “clinicians think X,” “clinicians do X” line of argumentation that I see often.
As a side observation, I have repeatedly had the experience of watching physicians make great decisions and using a great process for doing so, but hearing physicians describe the process in a way that statisticians would say violate principles of decision theory. In other words, I’ve often seen physicians poorly describe the process they actually use with patients, e.g., saying they use a hard threshold for treatment according to some medical guideline when in fact they make the needed compromises when the patient is above a threshold but has many other things in their favor.
Interesting conversation. I am an ex-GP and worked in the Neonatal Intensive Care Unit for a while. So, I had a lot of hours making a snap decision on daily basis. Given that experience, it does allow me to empathize better with the work of clinicians.
A lot of clinical decisions carry tons of assumptions and potentially many out-of-date ‘evidence’-based practice. More often than not, they are well-meaning and that is the only information known to them.
So, if we can have a better collab between biostatisticians and clinicians it would be great. And we should find a way to communicate both sides of the worlds so that everyone can understand I guess.
I also learn that there is no shortcut to this.
I think some practical advice would be:
For biostatisticians, be kind and patient with clinicians. Do take your time in explaining statistical findings. Observing clinician works can help with building your ‘patience’.
For clinicians, it’d be great if they are allowed to have the opportunity of getting good probabilistic lectures/advice/courses, hahaha. It was horrendous when I was doing my training, the lecturer literally just read the slides… that experience definitely deterred 99.99% of clinicians from statistics.
Very good ideas Andrew. At the heart of good collaboration is mutual respect, and good education. Biostatisticians receive minimal education in medicine, learning from their physician colleagues on the job instead. And as you said, the biostat education of physicians is generally terrible. If I were designing a curriculum for physicians for quantitative methods it would bear almost no resemblance to what is used today.
Another issue is finding the optimum division of labor, which could affect how we teach if we were able to assume that the physicians would operate in an optimum-division-of-labor environment going forward. My very best collaborations have been with cardiologists who not only respect what I do but also do not want to spend time learning how I do it. Instead, they want to know why I do it, what is the general basis for a method, and especially how to interpret the results. On the other hand, the clinical researchers who have wanted to be able to duplicate my analyses or argue against my choices have in the majority of cases created an inefficient and dissatisfying collaboration for me. There are unfortunately many clinical researchers who want to make statistical choices of methods, not just inform my choice of methods. This is the statistical equivalent of me telling a clinician how to treat a patient, which I would never do.
This a great description of the paradox underlying clinical decision making.
Really enjoy all the comments on this thread. It’s analagous to my experience in informatics. Good EHR improvement requires IT analysts to understand physician workflows, and physicians to appreciate different IT solutions to a clinical needs. Not everything is solved by adding a pop-up alert; every pop-up alert need not show for every physician in all settings. Lack of perspective and communication leads to poor or inefficient EHR builds. Good collaboration and shared perspective saves time and achieves desired goals.
I was an RN for about 7 years before retraining with an MA and PhD in psychology (with heavy emphasis on quantitative work). I believe that experience has made me a much better consultant in all sorts of ways. For example, having experienced the often frantic day-to-day of a clinician, you understand immediately why the quality EHR is so challenging, and to proceed with extreme caution using.
I’ve attended tumor boards, surgeries, rounding a few times, observed nursing care for outcomes and management effects and even performed a time-motion study of ICU nurses. Listening to clinicians at all levels (RN-MD/PhD) generated my best ideas and pointed to holes in how we think or why practice varies from evidence, thus generating a pragmatic trial or new guideline review with summarization of evidence that measured the observed bias. As an epidemiologist, I’m not a fan of most typical epi cohort studies that jin a ton of pages from one design that, unfortunately rarely replicate. As a clinical epidemiologist, I think it’s pretty much a requirement.