Hey Everyone,
Are any of you healthcare workers or know someone who is?
As many of you know I’m very bullish about AI in healthcare and over the decades I believe it will drastically improve our healthcare systems. However, it will take a lot of time. Relieving some of the burden from clinicians and physicians is going to be a huge part of this.
So how are MDs on the field actually viewing this? I asked , Sarah Gebauer, MD, for her take on this.
If you want to get more deep dives like this and support my 10+ Newsletters that cover emerging tech, join as premium reader.
Is AI about to augment and support Physicians and Clinicians?
If this topic interest you please read more from the author in her Newsletter.
Crucial AI concepts every physician should know:
The Advent of Ambient Clinical Documentation
In an ere with key Generative AI startups in Healthcare getting more funding, Copilots for clinicians are also becoming more common. Ambient clinical documentation is a booming business. The technology allows doctors to record conversations with patients to automatically turn them into clinical notes and summaries using AI and is a major topic at Healthcare conferences like HIMSS conference this year, where more than 30,000 health and tech professionals gathered in Orlando, Florida.
Earlier in March, Salesforce announced Einstein Copilot: Health Actions will allow doctors to book appointments, summarize patient information and send referrals by prompting AI with conversational language, and they are not alone.
Companies like Microsoft’s Nuance Communications, Abridge and Suki believe their solutions will help reduce doctor’s administrative workload and prioritize connections with patients.
As I cover startups in AI, there are quite a few making waves in healthcare.
Ambience Healthcare
Hippocratic AI and so many others have enormous potential. But what do doctors really need?
How AI Can Improve Patient Experiences
Ambient clinical documentation tools have a side effect, more time for patients. Better patient experiences with face-to-face conversations with patients. This technology allows doctors to consensually record their visits with patients. The conversations are automatically transformed into clinical notes and summaries using artificial intelligence.
Automating Documentation can help reduce feelings of burnout
Administrative workloads are a major problem for clinicians across the U.S. health-care system. A survey published (via CNBC) by Athenahealth in February found that more than 90% of physicians report feeling burned out on a “regular basis,” largely because of the paperwork they are expected to complete.
More than 60% of doctors said they feel overwhelmed by clerical requirements and work an average of 15 hours per week outside their normal hours to keep up, the survey said. Many in the industry call this at-home work “pajama time.”
What if AI can save time and improve clinician well-being, how does this improve the experience of patients? Our expert MD talks about this below in more detail.
Many doctors said their biggest concern about AI is that it could lead to a loss of human touch in health care. But what if the opposite is true?
Documentation Tools for Clinicians (“Ambient Scribes”)
Nuance’s DAX Copilot (Nuance was acquired by Microsoft)
Other articles by Sarah MD:
Overview of Physician Roles in Health Tech/AI
Clinical AI Risk: Current and Future State
Sarah is one of my favorite medical futurists to read, her posts are grounded in experience and research.
How Physicians are using AI to do less documentation
Guest post by of
Imagine it’s your first day running a multi-billion dollar company. You’re the new CEO, and can decide how your employees spend their time. One of your employees presents you with a productivity report that states that your highest-paid, most specialized employees are spending 35%-50% of their time writing down what other people say or should do. That means that you’re paying these people at least a third of their salary to act as transcriptionists. You immediately ask your assistant to find another way for that information to be recorded, and have just saved hours of work and can use those people to expand your operations.
This is a calculation that many CEOs are now doing as they try to figure out how to use AI to use employees more effectively.
Physician Documentation: Trying to do everything, everywhere, all at once
But would you believe me if I told you that the employees in this scenario are actually physicians? That’s right: your doctor spends at least a third of his or her time documenting information rather than seeing patients or thinking. I’m guessing your first reaction is “that makes no sense” and you’d be right. Most physicians dread this aspect of the job, which has increased with the advent of electronic medical records and increasing regulatory requirements from insurers and the government.
Can AI help physicians get out from under the paperwork?
Clearly we need a major update to the way physicians spend their time. Many hope AI will be that change that frees physicians from mountains of paperwork. AI has already helped simplify some aspects of healthcare administration and paperwork through solutions like ambient scribes, referral letters, and billing optimization. And there’s potential for AI to go much further to make better use of physician time.
Has documentation made doctors into notetakers rather than healers?
I used to be part of an admissions committee for a medical school. When I interviewed idealistic young people applying to medical school, “typing” and “filling out forms” was never once mentioned as a reason for becoming a physician. “Helping people” was by far the most common answer. A recent editorial in the New England Journal of Medicine addressed this issue in a well-titled article, “On Calling – From Privileged Professionals to Cogs of Capitalism?”. The author, a physician, describes the transition from viewing medicine as a calling to viewing it as just a job.
One study showed that physicians spend two-thirds of their time talking to patients, but a quarter of that was talking to the patient while typing. Have you ever had a client meeting in which you had to take notes at the same time? And that the notes had to be excellent, because you could potentially be sued for millions of dollars if you missed a detail in the official record of the meeting? That’s what it’s like to be a physician trying to form a connection with a patient while trying to use an electronic health record.
If we assume the average physician works 50 hours a week, that means physicians are spending 25 hours in which they’re able to actually connect with a patient and think about the patient’s problems.
Another survey showed that some physicians spend almost 20 hours a week on paperwork and documentation:
More time charting means less time with patients
Because physicians generally make their money through volume instead of quality, they’re expected to maximize the number of patients they see. However, each patient then adds to their paperwork burden. A study by the Agency for Healthcare Quality and Research showed that physicians feel they need 50% more time for each scheduled visit to do a good job. Many studies have shown that physician burnout is increasing in part due to the documentation burden.
Physicians’ staff shoulders a huge amount of the paperwork also, and one study estimates that each practice spends $70,000/year on average just on interacting with insurance plans. That doesn’t include physician time spent charting and transcribing. That interaction with health plans translates into $31 billion in opportunity cost. If you were the CEO of this company, I’m guessing you’d find a better way to spend that money.
AI is already making an impact in healthcare
Physicians are most enthusiastic about AI that allows them to spend more time doctoring and less time notetaking. Thankfully, AI is already making an impact in the paperwork burden for many forward-thinking physicians.
Prior Authorizations
The process of physicians requesting that their patients receive a service then respond to denial letters is one of the first areas in which physicians enthusiastically started using generative AI. These prior authorizations often require a letter to be written to justify the use of a drug or therapy with details of the patient and why that patient needs the therapy. Often physicians in a specialty have to write basically the same letter to the same insurance companies with slightly different patient details. These letters are particularly frustrating because almost 75% of these rejections are ultimately approved. A third of medical offices employ full-time staff to take care of the average 30 prior authorizations per physician per week.
More importantly, according to the American Medical Association, 89% of providers said that prior authorizations had a negative impact on patient care.
Just a week or two after GPT 3 appeared, many of my colleagues started to “write” these largely pointless and formulaic letters using an LLM, and the use has only increased in the past year. Multiple software companies now market their ability to enable automatic prior authorization letters for physician offices including Thoughtful, Waystar, and many of the companies that bill on behalf of physicians, known as revenue cycle management companies. Still, less than half of prior authorizations are performed fully electronically (and even fewer with AI); most offices still use fax.
Of course, insurers are also using AI to streamline their workflows, as this McKinsey graphic demonstrates. The concept of taking humans out of the loop for the processes that rely on humans just looking up straightforward information like eligibility makes a lot of sense.
Predictably, the insurers have started to use AI to deny claims more quickly. This practice is the subject of lawsuits against United Healthcare and Humana, which allege that the insurers are denying the claims so quickly that there is no human judgment involved in the process.
Referral Letters
A similar use of AI from many software companies, especially those that interface directly with EHRs, are referral letters from one physician to another. Like prior authorization letters, these are pretty formulaic. They usually say something like: “Dear Dr. Smith, thank you for referring Mr. Jones for gallbladder surgery. This delightful patient underwent a successful cholecystectomy…” They’re a way for physicians to communicate with each other and to give updates, and much of the formal language is hundreds of years old. They’re also usually easier for other physicians to read and get to the point more quickly than scrolling through a patient’s medical record.
Some of the AI referral letter products can even learn the physician’s usual voice, tone, and format, and write referral letters based on those preferences. Because these letters are often more about maintaining relationships (and therefore a steady flow of patients), the fact that they exist is often more important than the specifics of the letter. Not sending a letter like this would imply that you don’t care enough about the other physician to give them an update about their patient.
Scribes
One of the hottest areas for AI technology right now is the AI scribe space. These tools use generative AI to listen in to the conversation between the patient and the physician and create a templated note for the medical record. This technology allows physicians to avoid looking at a screen and typing while they’re trying to connect with a patient.
History of scribes in healthcare
Having someone else type notes during a doctor visit is not new. Older physicians recall that they had human scribes who wrote the notes by hand. This concept gave way to physicians typing or dictating information about the visits. Dictation is still widely used, though it is now electronic rather than sent to a human to transcribe as it was 20 years ago. Dictation differs from ambient scribes in that it’s direct transcription rather than a summary of a visit.
Human scribes
Human scribes have made a comeback in recent years, with about 12-20% of physicians in the US currently using a person to stand in the room and take notes directly into the EHR. These scribes decrease physician documentation time by three minutes per patient on average, which can save up to two hours a day for physicians who see 40 patients. Some hospitals have experimented with virtual human scribes, in which offshore humans listen in to a visit and summarize the encounter, but it’s not a common option.
This graphic from the American Academy of Family Practice shows the advantages and disadvantages of these technologies:
Ambient Scribes
And now we have ambient scribes, which are pretty incredible already. I’ve tried versions from multiple AI scribe companies (including TORTUS AI, which – full disclosure – I consult for) and they do an amazing job of filtering out irrelevant information and putting the information in the right spot.
Think of the technological challenge inherent in this process: patient visits are often interrupted by clinic staff or phone calls, meander off into conversations about kids and dogs, and use abbreviations and technical jargon. They’re often circular, meaning a patient will mention a symptom and the physician won’t ask a follow up question about it until several minutes later. These tools produce a full transcript that uses generative AI to find the important information and put it into a form that’s indistinguishable from what a physician would actually type. Many of my friends have reported that ambient scribes actually do a better job of including important details than they would have included themselves.
This graphic from Elion describes the way most ambient scribe software works:
Results from early ambient scribe studies
A recent study from The Permanent Medical Group (TPMG), which is affiliated with Kaiser, showed favorable results for physician engagement and accuracy of documentation. The metric that physicians are most excited about is the decrease in what’s referred to as “pajama time”, or the time doctors spend charting at night and on weekends. Physicians who used the AI scribe software spent less time writing notes during off hours, which is a huge deal for doctors who want to spend time with their families instead of typing notes and dictating.
So what’s the hold up? Why isn’t there an ambient scribe in the hands (or voice!) of every physician?
Crucial concerns about ambient scribes
The main concerns are outlined by Elion, which reviews ambient scribes:
Note quality: This is the big one. A small error in the ambient scribe can mean a big malpractice lawsuit if the information isn’t captured correctly. This is already a risk with dictation and typing, but the physician is less in control of the process with ambient scribes. The doctor has to sign off on the note, but it seems unlikely they will thoroughly read all the notes, especially if it seems close enough. Aspects of the software that affect the accuracy include:
Background noise
Multiple speakers
Multilingual conversations
Note customization: Doctors are particular and like to have their notes sound like themselves and organized in a certain way. Some vendors allow for customization at the specialty or individual physician level, and others just have a standard template.
Medical coding: Most of the function of an electronic medical record is for billing and coding. If the ambient scribe is not integrated with the coding software, the doctor still has to click through the coding boxes.
Cost: Many vendors charge an additional $2-7k/physician for this technology. Although some health systems are hopeful this technology will allow them to see more patients, many physicians see as many patients as they can already during the day, and save the notes for the “pajama time”, so the ambient scribes won’t necessarily mean more patients (and thus more hospital revenue).
EHR integration: Many patients have complex histories. It’s a major advantage if the AI software can pull information out of the chart to populate the note. And not all the ambient scribes still require the physician to copy and paste the note into the EHR, which is clearly not ideal. Future iterations of ambient scribes will likely notice when a physician tells the patient she will order a medication and enter that into the EHR also, going beyond just the note summarization functionality.
The Ambient Scribe Market
A variety of companies offer AI scribe services, including some big players like Augmedix and Nuance. They’ve been raising money steadily and the race is on to see who will win major market share:
Major players in the market are shown in this graphic from Hospitalogy:
It’s also worth noting that some of these companies have partnered with cloud computing services, which will be an increasing issue in healthcare as the volume of data increases with AI:
Conclusion
As physicians continue to grapple with the dual demands of patient care and administrative burdens, AI offers a promising avenue to reclaim valuable time for what matters most: patient interaction and care. While challenges such as note quality, customization, medical coding, cost, and EHR integration remain, the advancements in AI technology and its integration into healthcare practices hold the promise of significantly reducing the “pajama time” many physicians currently endure. This shift not only has the potential to improve physicians’ work-life balance but also to enhance patient care by allowing doctors to focus more on the human aspect of medicine rather than the clerical. As the healthcare industry evolves, it’s clear that the role of AI will be pivotal in shaping the future of medical documentation, signaling a shift towards more efficient, patient-centered care.
About Sarah Gebauer MD
Dr. Gebauer is a physician, AI expert, and consultant with extensive administrative and quality experience. She is passionate about engaging physicians in healthcare AI and started a Slack group, video series, and website dedicated to the topic. She also writes a weekly Substack newsletter about the intersection of healthcare and AI.
Dr. Gebauer attended Stanford Medical School, completed an anesthesiology residency at the University of California, San Francisco and a fellowship in Hospice and Palliative Medicine at the San Diego Hospice Institute for Palliative Medicine. After fellowship, she joined the faculty at the University of New Mexico and she completed a Graduate Certificate in Clinical Informatics.
Sarah takes on really fascinating topics:
Will AI make bad doctors better and good doctors worse?
Healthcare and Data Brokers
Read More in AI Supremacy
