Back to the Future- Can Artificial Intelligence (AI) Get Doctors & Patients Talking Again?


Disclosure- I am an advisor to Cloudmedx

Thirty years ago, when I started my medical career, we would document clinical care notes on reams of yellow paper—the more experienced the doctor, the shorter the note. Medical students would go on for 15 pages, the resident might write 3 pages, and the senior physician would write 3 lines. Over the years, I’ve witnessed 3 waves of change in healthcare.

The First Wave: Physicians, Take Note

During the first wave, physicians used handwritten notes to chronicle patient care. Although this workflow was convenient for the physician, this method had some serious disadvantages. Practice patterns evolved based on what the individual physician had learned over time, not necessarily best practice. For a particularly challenging case, one would go to the library to look for an article. Furthermore, the often-illegible paper notations were prone to misinterpretation by other members of the care team. Checks and balances—for example, making certain that the patient wasn’t allergic to a prescribed drug—was a slow, manual process. Finally, and perhaps most importantly from the hospital’s perspective, care records were incomplete, which meant that the hospitals could not bill completely for the treatment patients received during their stay.

The Second Wave: Enter the Electronic Medical Record

The electronic medical record (EMR) forces physicians to enter data into predefined boxes (a process referred to as structured data entry), as opposed to jotting down handwritten notes. With the EMR, additional information can be entered in order to capture all information necessary for billing, the lifeblood of any hospital. Ideally, the EMR would make recommendations based on best practice (sometimes referred to as clinical decision support), but this is rarely the case.

Partly incentivized by the federal government, the EMR has become ubiquitous. But billing, as well as quality metrics, make doctors feel like slaves to the EMR as they try to figure out which button to click rather than using their time and energy to focus on the patient. Many doctors spend hours after work just trying to catch up entering information into the EMR so they can see as many patients as they did before the advent of this technology. This additional demand on physicians is one factor contributing to their burnout.

Physicians’ engagement with the EMR makes the patient feel like a third wheel during their own clinic visit rather than an active participant in the doctor–patient relationship.

The Third Wave: Are the Robots Coming to Take Your Job, or Are They Coming to Serve You?

In almost every industry, workers fear competition from automation. Medicine is not immune to this potential threat. Several years ago, a well-known venture capitalist suggested that most physicians would ultimately be replaced by cheaper, faster, more accurate computers. However, I still believe that most people want their care from people. It’s really up to us, both providers and patients, to figure out how to get the right mix of man and machine to improve care, lower costs, and maybe let people get back to taking care of each other.

The Wave of the Future: Enter CloudMedx Artificial Intelligence

Tashfeen Suleman, the founder of CloudMedx, is a computer scientist who spent many years working for Microsoft. When his father, a former fighter pilot, became ill he went to see a doctor. Unfortunately, the symptoms did not produce the right diagnosis from the treating physicians, and Tashfeen’s father almost died. It was then that Tashfeen realized that he could use his skills to put information into the hands of physicians in time. 

Tashfeen set about building CloudMedx, with the vision of helping physicians and patients use their own data to make data-driven decisions around risk, care management, and revenue cycles. Over the years, the company has grown and provides its services to an impressive set of health systems, including Mount Sinai Health System in New York, Crescent Medical Center in Texas, The University of California–San Francisco, and others.

From Chess to Healthcare

In 1997, the computer “Big Blue,” competed against the sitting world chess champion, Gary Kasparov, and won (though Kasparov beat Big Blue the year before). Somewhat frustrated, Kasparov used the experience to create a new chess tournament—one that had 3 types of competitors: humans, computers, and human with computers. Time and again, the winners of that tournament were the human–computer teams: The computers performed complicated calculations, but could not replace human intuition and creativity.

Rather than look at artificial intelligence (AI) as a competitor or a replacement for humans, Tashfeen and his team replicated Kasparov’s early work. Presuming man and machine would be more successful than man or machine, Tashfeen’s team went about proving their hypothesis by having 3 groups take a modified version of the medical boards.

The Outcomes

The mean score of humans who took the boards was 75 (range, 68-81), the mean score of the computers was 85, and the human–computer teams achieved a mean score of 91. Notably, it took the AI several weeks of training to understand context and take the exam. The same level of understanding takes doctors 12 years. The human group took 70 minutes to take the exam, whereas the software completed it in under 5 minutes. During the post-exam interviews, one of the residents taking the exam with software assistance noted that the software gave him more confidence in his answers, particularly as he tired during the exercise. Exhaustion is an all too common feeling for today’s physicians.

What Does This Mean for Healthcare?

The results of this test demonstrate that computers can learn at an incredibly fast pace and they may be able to assist physicians in an augmentative manner. As it integrates into the EMR, the software becomes contextual. It knows enough about the patient to offer advice to the physician, accessing a medical library and incorporating clinical outcomes into the patient’s care plan. It also weighs likely risks, suggests tests to order, and saves time by offering to complete the documentation with the physician’s oversight.

I firmly believe that AI in healthcare is not a replacement but supplement providers, a way to change our relationship to data. Doctors will stop being documentarians and go back to being humanitarians, with better answers at their fingertips. Patients will be offered a chance to learn about their problems without being burdened by worry of the unlikely. It is the combination of human and machine that offers a path forward.

Reader Interactions

Leave a Reply

Your email address will not be published. Required fields are marked *