AI Can Lift Administrative Burdens and Restore Joy in Practice

Oct 09, 2023 at 04:17 pm by Staff


 

By Aaron Neinstein, MD, Chief Medical Officer, Notable

 

I think Eric Topol, MD, got it right. In his book Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again, Dr. Topol invites us to reframe our expectations of artificial intelligence (AI). Instead of focusing on trying to use AI to perform miracles like curing cancer, he encourages us to first eliminate the burdensome administrative tasks that are interfering with our role as healer and taking away from our ability to connect with patients.

Relieving unnecessary administrative burdens restores our opportunities to focus on patient relationships and provide care, and this can provide a route to joy in practice.1 Fortunately, we have many opportunities to leverage technology to drive operational efficiency with improvements that allow us to cultivate time for our most precious resource: the clinician-patient relationship.

A Smoother Workflow for Referrals

At UCSF Health, we are using AI to ease the administrative burden of the 1.5 million faxed documents we receive per year. Roughly one quarter of a million of those are referrals. We were using an army of people to manually enter that information, but we deployed AI to ingest that information more accurately, quickly, and cheaply.2 Now referred patients are seeing their providers sooner, and staff members are transitioning to higher-value opportunities to interact with patients. This example shows the rewards of harnessing AI as an administrative workhorse.

A Series of Nudges

We need not fall into the trap of thinking that AI must help by making the diagnosis or telling us the treatment plan. The question is not practitioner or machine, but practitioner and machine: How can AI augment care teams to help them perform at their best? As addressed by my UCSF colleague Julia Adler-Milstein, PhD: In clinical care, we can think of how AI can provide a series of nudges. This line of thought frames AI less as the source of one “Eureka!” diagnostic breakthrough and more as a clinician’s assistant through what she calls “wayfinding.”3

Diagnostic decisions and treatment plans may not happen in a single sitting but, rather, they often unfold through a series of events over time. We need little nudges of information all along the way. For an endocrinologist like me, it might be a question like: “Hey, did you forget to order the PTH in that person with osteoporosis?” Or “Did you notice that person’s creatinine level has been steadily rising over time?”

A Virtual Assistant

Similarly, I am hopeful regarding the prospects for an unlimited array of virtual assistants or copilots. These would help surface pertinent information about your patient that you might not have been aware of—details allowing you to be prepared and have all of the information at your fingertips when you walk into the patient’s room. An AI virtual assistant would allow you to focus on the patient relationship rather than spending your time digging for information.

While these capabilities may sound far into the future, they may be here far sooner than many think. The latest advances in AI technology, namely large language models (LLMs) and generative pretrained transformers (GPTs), are already being built into the fabric of some extremely promising tools and workflows. Imagine a virtual assistant that can comprehensively review millions of data points from across an individual’s medical record to surface personalized recommendations for each patient throughout the care continuum.

The key to successful implementation of this sort of virtual assistant will be integrating it into the patient’s care and clinician’s workflow. In fact, equally as important as the quality of the AI algorithm is understanding how a person is making a decision and how the moment of interacting with AI sits within their workflow to influence behavior and decision making. You could have the best algorithm, but if dropped into the wrong workflow, it may be irrelevant or, worse, harmful. For instance, the AI processing our faxed referrals at UCSF Health is more successful because it is tightly integrated into our staff members’ workflow.

Ambient Listening for Clinical Documentation

Ambient listening is another tool that offers an opportunity to reduce the burden of clinical documentation. The voice-recognition technology can automatically document patient encounters in the exam room—saving clinicians time by improving note-taking efficiency.

Tools for Top-of-License Practice

The next generation of physicians will vote with its feet when selecting a medical practice. Physicians will choose to work in an environment that reduces the burden of administrative paperwork, facilitates easier documentation, and provides high-quality decision support. They will be looking for a toolset that makes it easier to practice and helps enable the physician-patient relationship.

Opportunities to lift administrative burdens present themselves throughout healthcare organizations. For example, prior authorization paperwork and staffing predictions can be streamlined using existing AI solutions.4,5 More equitable and personalized development of patient education materials and more efficient, individualized clinical trials recruitment and enrollment are other possible near-term opportunities.6 As new AI tools emerge, they can be thoughtfully implemented to complete mundane tasks that enable clinicians to practice at the tops of their licenses.

Rather than being one more technology burden, AI offers tools that can enable us to regain time to focus on our patients and restore joy to the practice of medicine.

Our thanks to Aaron Neinstein, MD. Dr. Neinstein, who is board certified in endocrinology, clinical informatics, and internal medicine, maintains an active clinical practice focused on the care of people with diabetes. A nationally recognized expert in diabetes technology, he has led the development and implementation of a wide range of health technologies at UCSF, with a focus on digital transformation and improving patient access and empowerment through virtual care models.

 

Reprinted with permission. ©2023 The Doctors Company (thedoctors.com).

The guidelines suggested here are not rules, do not constitute legal advice, and do not ensure a successful outcome. The ultimate decision regarding the appropriateness of any treatment must be made by each healthcare provider considering the circumstances of the individual situation and in accordance with the laws of the jurisdiction in which the care is rendered.

 


References

  1. Sinsky C. The patient safety risks of burnout—and the path to professional fulfillment. The Doctor’s Advocate. Published December 2022. https://www.thedoctors.com/the-doctors-advocate/fourth-quarter-2022/the-patient-safety-risks-of-burnout-and-the-path-to-professional-fulfillment/
  2. Hagland M. UCSF Health’s AI breakthrough: what’s been learned so far. Published January 24, 2022. Healthcare Innovationhttps://www.hcinnovationgroup.com/analytics-ai/artifical-intelligence-machine-learning/article/21254285/ucsf-healths-ai-breakthrough-whats-been-learned-so-far
  3. Adler-Milstein J, Chen JH, Dhaliwal G. Next-generation artificial intelligence for diagnosis: from predicting diagnostic labels to “wayfinding.” 2021;326(24):2467-2468. doi:10.1001/jama.2021.22396
  4. Falcone S. Google launches 3 AI tools for faster health preauthorizations. Nurse.org. Published April 25, 2023. https://nurse.org/articles/google-ai-for-health-preauthorizations/
  5. Burky A. From finding the right candidates to keeping them, how hospitals are using AI to address workforce needs. Fierce Healthcare. Published November 23, 2022. https://www.fiercehealthcare.com/ai-and-machine-learning/finding-right-candidates-keeping-them-ai-aiding-healthcare-industry-meets
  6. Lee P, Goldberg C, Kohane I. The AI Revolution in Medicine: GPT-4 and Beyond. 1st ed. Pearson; 2023.
Sections: Business/Tech