About 80 percent of a diagnosis comes from the patient’s story. The electronic health record has no place to keep it.
I learned the first half of that in my second year of medical school, in a physical diagnosis lecture by the late Dr. Arthur Aufses Jr., chairman of surgery at Mount Sinai for 22 years and a man of extraordinary skill and intellect. In that lecture, he said that about 80 percent of the information needed to make a diagnosis comes from talking to the patient. The rest comes from the physical examination and tests. Surgeons are not known for being talkative, which is why it stuck. At the time, it sounded almost incredible. Studies of new outpatients put the figure at 76 to 82.5 percent. After more than 20 years in practice, I’d call his advice 100 percent accurate.
When I started my endocrine practice, I applied his principle from day one. Whether it’s diabetes, osteoporosis, or thyroid disease, I start at the very beginning and walk through every pertinent detail in order. I ask the patient to tell me about the disease as if writing a book for a complete stranger. I spend a lot of time establishing the story and writing it down. Truth be told, most physicians do the same on the first visit. It is what happens on subsequent visits that differs between my charting and most other charts I have seen.
In my practice, a very detailed history of a condition, with all important landmarks throughout the years, migrates from the history of present illness in the first visit to past medical history in the subsequent visit. Unnecessary details are stripped, but core events remain. With every subsequent visit, if something changes, it is added to the past medical history. So for each condition, the story grows. In reality, when you keep only pertinent important facts, the story is not too long. Half a page to one page of past medical history can contain all the clinically relevant facts in one place, in each note, and is readily seen during each visit. Here is an example. For brevity’s sake, to keep the patient anonymous, and comply with HIPAA, I changed the dates and some of the language.
Real physician voices, twice a week
Free, and one click to unsubscribe.
“PMH: Osteoporosis diagnosed in 2000 or earlier. Initially treated with Actonel between 2000 and 2003, Forteo for 2 years between 2003 and 2005, Actonel between 2005 and 2006, Boniva between 2006 and 2008, Actonel between 2009 and 2011, Atelvia between 2011 and 2012. Bone mineral density decreased despite the treatments, partially because of sedentary lifestyle after ankle tendon rupture and ankle infection in 2012 necessitating 6 weeks of IV antibiotic therapy. Received Reclast in 2013, 2014 and 2015. 2014 DEXA (bone density scan) showed stable bone mineral density compared to 2013. DEXA in 2016 showed lumbar spine T score -2.8, left femoral neck -2.6, left total hip -1.7, left forearm -2.3, not statistically significantly changed when compared to bone mineral density in 2014. First injection of Prolia in 2016, subsequent injections every 6 months through early 2024. Switched to risedronate in 2024 in preparation for a drug holiday. Did not tolerate risedronate because of GI (stomach) side effects, Prolia resumed in 2024 and continued in 2025. Nontoxic multinodular goiter: thyroid ultrasound in 2012 with 6 nodules, the largest 0.8 cm. Ultrasound in 2015 with stable thyroid nodules bilaterally and slightly increased right midsection nodule (1.2 cm). Thyroid ultrasound in 2016 showed enlarged right midsection nodule which measured 1.5 cm. Status post FNA (needle biopsy) of the dominant right midsection nodule in 2016: benign. Thyroid ultrasound in 2019 showed an increase in the size of the right midsection nodule, which measured 1.7 cm. Status post FNA of the right midsection nodule in 2019: benign. Thyroid ultrasound in 2021, 2022 and 2023 showed stable thyroid nodules bilaterally. Thyroid ultrasound in 2025 showed slight increase in the size of the right midsection nodule, otherwise stable nodules. Thyroid ultrasound in 2026 showed stable thyroid nodules bilaterally, at which point the decision was made, given the patient’s age, preferences and other comorbidities, to not continue ultrasound surveillance.” In reality, there is a bit more detail, but you get the gist.
The same patient as a list:
- Osteoporosis
- Multinodular goiter
- Prolia every 6 months
- Intolerance: risedronate (stomach upset)
Every item on the list is accurate. None of it tells the next doctor what to do or what not to repeat. Imagine a doctor who meets her through the list. She orders another ultrasound, sees a nodule that has grown to 1.7 cm, and sends her for a third biopsy. Or she decides the patient needs a bone-building drug and starts Forteo, not knowing the patient already had two years of it. The label limits Forteo to two years in a lifetime unless the patient is still at high risk of fracture. The story prevents both missteps. Without the story, no one would ever look as far back as 2003.
When electronic records arrived, I assumed, incorrectly, that the story principle would be applied there also. Nothing could be further from the truth. When I worked in Epic, the past medical history was a laundry list of diagnoses, with no story attached. I found a workaround: a free-text box for each active problem, called the Overview, visible to every doctor on the chart and editable, much like my dictated history. Most doctors were never taught to use it. One university health system found its primary care doctors used Epic’s built-in problem-oriented tools in fewer than 10 percent of visits. After it built its own easier version, clinicians across the system used it in more than 92,000 visits a month. And according to one Epic health system’s training manual, the Overview is lost when the problem is moved to Medical History. The one field built for the story is erased the moment the problem becomes history.
I still have access to Epic, and I can attest that learning the history of any significant chronic condition there means becoming a medical archaeologist. You dig through years of notes, results, and orders and reassemble the pieces into a story yourself. It takes so long that most physicians don’t do it. In a 15-minute visit, nobody has the time or the desire to click through dozens of notes to fish out what matters. The record has everything and is so disorganized that it’s of little use for reconstructing a patient’s history. So we treat what’s in front of us.
This is a design choice, not a property of computers. Builders of Epic have regulators, administrators, and billers in mind because that is who their customers are, not physicians. Those same builders were never taught the value of the story. So it is not there. In my own office, I created a HIPAA-compliant, AI-based dictation system which still follows my rules and simply converts the detailed history of present illness of the first visit into past medical history at a subsequent visit automatically, so my story never disappears. The difference is that I make the rules and I know the importance of the story.
Is there a solution for the current EMRs with AI? There definitely is, if the importance of the narrative is finally recognized. AI could draft a story from the entire record: notes, labs, imaging, orders, and the reason every drug was started or stopped. Every sentence links to its source. The story is pinned at the top of the problem list and of every note, so nobody opens a chart without seeing it. The physician who manages the problem reads and corrects it, then adds it to the past medical history with one click. From then on it grows the way mine does, a line for each change.
The payoff? Think of how many unnecessary procedures can be avoided, how many repetitive uses of medications that failed in the past can be avoided, even how many consults can be avoided once the information about the patient’s past medical history is not a laundry list but in fact a clinically relevant, relatively compact story. The efficiency and money saved are difficult to overestimate. Anyone who wants to measure it can do it on their own time.
In my practice, I decided on day one that the story matters. Anyone who controls a record can decide the same. The story is never lost and never diminished in my office. It always grows, improves efficiency, and results in better patient outcomes.
Michael Duben is a board-certified endocrinologist in solo private practice in Fairfield, Connecticut, and is affiliated with Bridgeport Hospital. He runs both Endocrinology of Fairfield County, a traditional insurance-based practice, and Restore Health, a cash-based functional endocrinology practice. He completed the Institute for Functional Medicine’s training program.
He did his internal medicine residency at Mount Sinai and his endocrinology fellowship at Montefiore/Einstein, and was a clinical instructor in the Department of Endocrinology at Albert Einstein College of Medicine. He has practiced since 2003, with clinical interests in thyroid disease and thyroid cancer, osteoporosis and bone metabolism, parathyroid disease, pituitary disorders, and obesity and metabolic disease. He has prescribed glucagon-like peptide-1 (GLP-1) receptor agonists since the first one reached the market in 2005, and has lectured to practicing physicians on diabetes and thyroid management.
He designed and built the HIPAA-compliant AI-assisted clinical documentation system he uses in daily practice, without formal training in computer science, along with a range of other tools for his patients and his office. He shares updates on LinkedIn.
