It’s an open secret in the physician community, but perhaps not widely known: Physicians spend more time documenting and updating patients’ electronic health records than we do examining and talking to actual patients. According to a 2020 study in the Annals of Internal Medicine, physicians spend an average of 16 minutes and 14 seconds per visit, with chart review (33%), documentation (24%) and ordering (17%) taking up the majority of the time That time — less than 5 minutes of direct interaction with our patients.
This is a problem because while creating documentation is time-consuming, it is critical to the health of our patients. And our existing solutions are still not up to the task. To make matters worse, many of the proposed solutions may misinterpret the problem.
Consider a woman in her early 40s who was recently admitted to the emergency room where I practice, complaining of fatigue and trouble sleeping:
When I checked on her, the symptoms she reported became more complicated: cough, fever, and shortness of breath. It expanded into her family history to include health and diet, her life insurance policy, the funeral of an aunt who recently died of lung cancer, and her past experience with treating her with SARS when her primary doctor diagnosed her.
After further questioning, the real reason she went to the emergency room finally came to light: She was worried that she, too, had lung cancer, and that her primary care doctor hadn’t thoroughly examined her.
In her case, I had plenty of time to delve into her core concerns. An exhaustive record of my patient’s entire visit and all the symptoms she reported gives an accurate description of what happened. However, this will also include a few red herrings, and may not clearly convey the crux of her visit: She wants reassurance from her doctor that she doesn’t have lung cancer.
Technology has increased recording time, with doctors paying more attention to computers or tablets than patients. To alleviate the demands created by technology, physicians hire scribes, develop macros, build templates and leverage speech recognition to create patient records. There are a number of solutions that try to solve this problem with sophisticated environmental scribing techniques that can listen to the entire encounter and then automatically create clinical documentation for the physician.
I believe these ambient scribing solutions are technical marvels, but in the end, don’t fix the root problem:
The basic goal is to provide an easier way for doctors to put mental models in their heads into a patient’s digital file.
In contrast, environmental transcription provides a record of what happened, rather than a representation of the doctor’s mental model of the patient. The distinction I make is similar to a court reporter versus an opinion judge. The recorder creates the original record; the judge uses that transcript to create a mental model—in effect, an expert narrative that throws out facts that are not relevant to the case.
Environmental dashing is often proposed as a way to improve documentation, but it also addresses bugs. For example, cameras and automatic dictation could theoretically do this, but the burden of explaining it all would still fall on the doctor. (Assuming, that is, they have enough time to review the full transcript.)
Therefore, the real challenge for more effective documentation is to obtain relevant subjective information from patients, obtain objective data from medical records, examinations and tests, and use expert clinical opinion to develop a cohesive medical narrative of the entire encounter. And no matter how amazing AI becomes, ambient transcription is only better at creating more accurate transcriptions—it doesn’t include the unspoken ideas and expertise of clinicians.
As examples of startups tackling documentation challenges from different directions, here are a few promising examples:
-
Omedix By combining artificial intelligence, natural language processing, and virtual scribes to provide a hybrid solution, virtual scribes can ask clinicians questions to clarify their mental models of documents.
-
Suki Use a combination of machine learning, natural language processing, and human-in-the-loop to convert a doctor’s speech into a complete clinical document. Essentially, their system learns from each doctor how to translate the doctor’s brief verbal expressions into a fully documented mental model.
-
abridged An ambient listening technology is deployed that effectively acts as a doctor’s forensic reporter. It listens to conversations, akin to environmental scribing — and then creates a structured summary that physicians can refer to when writing clinical documents at the end of the day.
-
Decoding Health Have sophisticated clinical AI that can think like a doctor. Based on the patient’s symptoms, medical history, and objective findings from the consultation, it generates a graphical interface of the patient’s mental model, which physicians modify to accurately reflect their thinking. The model then allows the software to develop a doctor’s document that sounds like a doctor and is a coherent narrative.
In full disclosure, the latter two are nurturing affect health, an innovation center where I work as a consultant. As I said, it’s hard to find players in this space solving problems from mental model frameworks, but we hope to see promising results from startups like this soon.
Photo: megaflopp, Getty Images
