The Doctor's Advocate

The Doctor’s Advocate

First Quarter 2025 | Archives
Author photo

Daniel Kent Cassavar, MD, MBA, FACC, Medical Director, The Doctors Company and TDC Group

ARTICLE AT A GLANCE
Healthcare practitioners are ready to let AI lighten their loads—but first, we need to achieve EHR interoperability. AI can help us with that, too.

My checkout desk is run by an individual who has been with my practice for 20 years. I see her being more useful doing more important jobs, because we should be using automation to discharge patients and book their next appointments.

I am optimistic that tools powered by artificial intelligence (AI) can make our lives easier and more successful by relieving administrative burdens, streamlining workflows, improving patient safety, and making risk management easier to manage.

But first, we need to achieve interoperability for our EHRs. Fortunately, AI can help us with that, too.

Our AI Wish List

Most practitioners spend so much time swatting through clouds of administrative nuisances that we miss the experience of focusing on and advocating for patients.

The wish list of administrative and business operations functions that AI can help us with includes prior authorizations, billing, scheduling, and a host of front-desk and checkout tasks. Further, I’d welcome some automated assistance with certain essential items completed by medical assistants, like medication reconciliation, that take up precious, limited appointment time. Witnessing improvements in AI for healthcare, I’m looking forward to my slice of the assistance pie.

Here are the top three items on my personal AI wish list:

  1. Booking: Many healthcare systems are already experiencing success with referral automation and other scheduling tools. These generate cost savings for practices and organizations while increasing healthcare access for patients. For example, a pilot program at Montage Health in California has shrunk their time from referral received to patient scheduled from 23 days to 1.6 days.
  2. Emails: Generative AI can draft email responses to common questions like “What does this test result mean?” Sharing the inbox load—and preserving clinician attention for more complex patient situations—combats one of the primary sources of burnout.
  3. Triage: After heart catheterization, a chatbot sifting a patient’s answers to questions can refer a patient with signs of infection to a human clinician. Meanwhile, that chatbot can answer questions for patients with routine procedure aftereffects, reserving the attention of human clinicians for patients with pressing concerns.

Triage chatbots are already live and patient facing, with mixed results so far. It’s early days—as developers make improvements, chatbots will help us more quickly distinguish between minor questions and major concerns. Already, chatbots providing patient support are one of the three generative AI use cases chosen for review by an expert panel at the IHI Lucian Leape Institute “as being broadly representative of anticipated clinical uses of AI in the next several years.”

AI Can Help With Teamwork, Too

Back to that heart catheterization: Let’s say I perform the procedure, but the next day, another practitioner is rounding. They don’t need to know procedure details—they just need to know that the result was normal. AI’s capability to slice and dice summaries, with varying degrees of complexity, can pull a lot of weight through transitions in care. Patient handoffs are charged moments for patient safety and practitioner liability, so AI’s capacity for summation—if carefully integrated into workflows—can elevate patient safety and bring relief to overburdened practitioners.

Other AI benefits to teamwork are harder to specify, but imagine how lifting administrative burdens can lift up teamwork: We all have more capacity to reflect and to interact in our best collegial manner when we reduce unnecessary hassles and stressors and recover some cognitive bandwidth.

Any of these wish list solutions must seamlessly integrate into the EHR, and they have to achieve better results than we’re getting now. Not perfect results, but better than our present state.

Let’s say my wish list of AI-powered tools becomes perfect and available for free tomorrow: My practice, like many, will still be hampered by interoperability challenges.

The Risks of Our Siloed Medical Records

When I complete an ultrasound here at ProMedica, the test result comes to my inbox. I can ask my nurse to call up the patient and say, “Your result was normal.” But if my patient gets that test completed at the other local healthcare system, no notification reaches me. If I get a call months later from the patient, asking about their test results, this is less than ideal from the perspective of risk management, relationship management, patient satisfaction, or clinician satisfaction.

The other big hospital in Toledo, Ohio, where I practice, is an Epic shop, just like we are—but they use a different version, so we’re still not interoperable. If my patient gets studies done at their hospital (perhaps their insurance demands it), I don't get those results until they are transcribed, printed out, faxed over, scanned in, and put into our records here. It’s silly. And potentially dangerous.

The benefits of interoperability to test tracking alone will be, if we ever achieve it, a boon to patient safety and practitioners’ protection from liability risks. For many practices and systems, test tracking is a chronic headache at best—and a safety and liability risk at worst.

In a mobile, high-tech society, patients should not have to leap over so many hurdles to see a practitioner outside their usual EHR, and healthcare should not still be doing this much printing, faxing, and scanning.

EHR Interoperability Is Table Stakes for AI

AI excels at digging through mountains of data, and it can help us wrangle our EHRs. Some healthcare systems have turned AI loose to find information lost in nondiscrete fields in their own EHRs. Elsewhere, researchers are using large language models to translate clinical data into standardized forms that are more easily transmissible across platforms.

Imagine the recovery of cognitive bandwidth and the benefits to professional satisfaction if we and our teams could spend less time chasing down records and more time practicing medicine.

Still, with all that we’ve learned about biases and other dangers inherent to AI for healthcare, developers have some trust building to do with clinicians. Part of that trust can be built by recognizing practitioners’ priorities: As we consider our investments in AI wish list items, we can let developers know that AI has some basic table-stakes promises to fulfill, starting with EHR interoperability.


The Doctor’s Advocate is published by The Doctors Company to advise and inform its members about loss prevention and insurance issues.

The guidelines suggested in this newsletter are not rules, do not constitute legal advice, and do not ensure a successful outcome. They attempt to define principles of practice for providing appropriate care. The principles are not inclusive of all proper methods of care nor exclusive of other methods reasonably directed at obtaining the same results.

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.

The Doctor’s Advocate is published quarterly by Corporate Communications, The Doctors Company. Letters and articles, to be edited and published at the editor’s discretion, are welcome. The views expressed are those of the letter writer and do not necessarily reflect the opinion or official policy of The Doctors Company. Please sign your letters, and address them to the editor.