Reimagining Clinical Documentation with Artificial Intelligence eloquently imagines leveraging the power of artificial intelligence (AI) to “autoscribe” in the EHR as a means of reducing physician documentation burden, improving productivity and reducing burnout. The authors are forward-thinking, innovative physician leaders from prestigious institutions who have significantly contributed to the improvement of health care.
Using AI as a scribe assistant could very well be a dream for many overburdened physicians. But what if the authors of the paper have missed the bigger picture? What if the game-changing answer isn’t investing in cutting-edge technology to enable physicians to spend less time typing into an EHR system that was designed to benefit billing insurance companies grounded in archaic claims-based systems? What if the smart investment points AI at enabling patients instead?
If I contrast the suggestion of using AI to enable more efficient doctors with other industries, the idea comes up short:
- Airline industry: Instead of standing in shorter lines because airline staff have access to more efficient technology, I can independently book a ticket, choose my seat, change flights and check bags without assistance or a line.
- Financial industry: Instead of blindly taking the advice of my investment banker who is assisted with technology, I can monitor my own portfolio, evaluate financial trends data and make my own trades.
- Real estate industry: Instead of relying on a realtor with access to cutting-edge technology to manage my house sale, I can list, market, advertise, manage and negotiate the sale of my home.
This, of course, is not a new concept in health care. Patient-centered design is commonly advocated for by thought leaders such as the Institute for Healthcare Improvement. When I first heard Harvard Business School professor Regina Herzlinger lecture on consumer-driven health care nearly a decade ago, I thought she was delusional. I grew up, professionally, within the walls of health care and only considered incremental improvements within the current delivery system. I now think of my previous perspective as the “Health Care System Stockholm Syndrome.” As I look at the stack of publications suggesting that incremental EHR evolution is the solution to physician burnout, I wonder if we aren’t all collectively falling into this trap.
This point was made personally clear to me when my 14-year-old daughter experienced her first urinary tract infection (UTI). By the time I arrived home, she had researched her symptoms (burning, frequent urination, mild hematuria) online. She determined that UTIs were common in her age group and, without risk factors, were easily treated with antibiotics.
Given our busy schedules and the late hour, I contacted the on-call nurse expecting a fast resolution. It turns out that no amount of clinical information, discussion of evidence and guidelines or request to submit a urine specimen could prevent a visit. Horrifyingly, I was even offered the option of seeking care at an emergency room. With no alternative, we went to a nearby urgent care. When the physician finally appeared after 10:00 p.m., she immediately declared that she would be prescribing an antibiotic based on the positive urine dip. She then repeated the same questions previously asked and documented twice the same evening.
My apparent frustration could seem unreasonable. Wasn’t it important that a licensed medical professional trained in physical examination and differential diagnoses assess a teenage girl to assure that a less common but more serious condition was not in play? It turns out the answer is no. According to Thomas M. Hooton, MD, and Kalpana Gupta, MD, who recently co-authored a paper on the subject, most women with classic UTI symptoms and no risk factors do not require additional testing to make the diagnosis. In fact, treatment can be safely started without diagnostic testing based on the responses to standard interview questions.
Equipping the urgent care physician with an AI autoscribe might have made her life better, but it would not have improved health, value or efficiency. Compared to so many health care experiences, my self-proclaimed insult (wasted $120 copay and six hours of time) seems slight. Although the insult was marginal, if I consider our experience in the context of a health care system that costs too much and produces too little value for too few people, I suggest that we think more radically about leveraging AI to support more self-service health care, where applicable and safe, such as:
- Personalized options for self-scheduling health care services.
- Home monitoring with self-management action plans.
- Price comparisons for various health care services.
- Shared decision making for high-cost services.
- Insurance prior authorization support, application, determination and communication.
Of course, the decision to equip doctors or patients with AI is not a binary, forced choice. I suggest that we think about AI as a means of enabling every individual on the care team to act to their fullest potential. The plea is to include the patient and family within the care team circle and that the focus be on overall value.
