Close-up of CT scan results reflected in glasses of physician.
Universal Images Group via Getty Images
Four months before Senator Lindsey Graham passed away, I addressed the United States Senate, tackling a seemingly straightforward question: Where can artificial intelligence have the most impact in healthcare?
To provide clarity, I highlighted aortic dissection.
“The most dangerous failure is not a machine failure. It’s a missed or delayed diagnosis.”
This weekend, early reports suggested that Senator Graham succumbed to this very condition.
While the timing is unfortunate, I am unaware of the specifics of Senator Graham’s medical care, and nothing in this article should imply that alternative technology could have altered the outcome. However, this incident underscores why conditions such as aortic dissection are central to discussions about healthcare AI.
While I could have chosen diabetes, heart failure, or numerous other ailments, aortic dissection exemplifies one of healthcare’s toughest challenges. The condition is rare, and its symptoms often mimic more prevalent illnesses like heart attack, stroke, or severe back pain. Accurate recognition requires clinical expertise, yet diagnosis is just the start. Once suspected, swift imaging interpretation, mobilization of specialists, and often emergency surgery are essential. Every moment counts.
Reflecting on it, my example was less about aortic dissection itself and more about the crucial role of time.
Discussions around healthcare AI often focus on whether computers can diagnose diseases better than doctors. This is indeed a key area of innovation, but with over two decades of experience in deploying clinical technology, I believe recognition is only the initial step.
Throughout my career in health systems, I’ve encountered clinicians at various stages, and none ever wanted to miss a diagnosis. I believe that failure is sometimes sought in the wrong places. Healthcare is an incredibly human endeavor. Radiologists end their shifts, emergency physicians move to the next trauma, specialists concentrate on the task at hand, and patients return home overwhelmed, contemplating how to manage daily life. Information is not lost intentionally but often gets buried under daily pressures.
Before my Senate testimony, Dr. Andrew Ibrahim, Chief Clinical Officer at Viz.ai, spoke to the House Energy and Commerce Committee about healthcare AI. His testimony resonated with what I hoped lawmakers would grasp. “What matters,” he stated, “is not the algorithm in isolation, but how these tools are integrated into real clinical workflows to solve real, high-stakes problems.”
The part of Ibrahim’s testimony that struck me most was his personal story about his father.
When his father showed signs of a stroke, Ibrahim, drawing on his clinical expertise, informed the emergency department in advance and coordinated the care that led to a positive outcome.
When we discussed this recently, Ibrahim shared that the experience highlighted something he hadn’t fully grasped during his medical training. “I didn’t appreciate how hard it is for information to get to the right person,” he remarked. “There’s a person at the end of that telephone game who has to make a time-sensitive decision.”
He explained that AI’s potential lies not only in identifying life-threatening conditions but in ensuring that the necessary images and information reach the specialist in time.
His testimony also emphasized a broader point beyond stroke:
“Not everyone has a medically trained son ready to coordinate their emergency care. We need systems that ensure patients get the right treatment without relying on luck or personal connections.”
That statement has stayed with me. For years, the focus has been on whether AI can diagnose diseases more effectively than physicians. I now believe AI excels when it makes exceptional outcomes less reliant on sheer luck.
What Better Looks Like
Consider a patient undergoing imaging for kidney stones. A radiologist spots an unrelated small spot on the lung and advises follow-up. The urologist, however, concentrates on the kidney stone. Months later, after a car accident, the patient has another CT scan. The spot remains and has grown. Again, follow-up is advised, but the emergency team prioritizes the injuries.
The patient leaves the emergency department overwhelmed, not focusing on the pulmonary nodules.
Life continues.
The patient resumes everyday activities, driving their child to school and caring for an aging parent, while the recommendation fades into the background.
Years later, unexplained pain leads to a new scan. The diagnosis is grim: advanced cancer. The once-active parent is now too weak from chemotherapy to drive, relying on their aging parent for transportation.
This outcome wasn’t intended or ignored. The system functioned as systems often do when built around busy humans.
Imagine the same scenario in a differently structured healthcare system:
The radiologist still detects the lung spot. The physician still uses clinical judgment. The patient leaves the emergency department focused on immediate concerns rather than an incidental finding.
Life continues.
The patient maintains their routine, but the healthcare system doesn’t forget.
The spot measurement is automatically included in the report. Evidence-based follow-up recommendations are provided. The patient receives not just a recommendation but a clear plan for next steps. If follow-up imaging doesn’t occur, the system identifies the oversight and contacts the patient via the portal. If that fails, another message goes through the primary care office. An appointment is arranged.
Upon follow-up, the spot has progressed but is still treatable. The treatment is straightforward.
No chemotherapy.
No radiation.
No major surgery.
The patient continues driving their child to school.
They keep attending to their parent’s needs.
Life continues.
Recently, Andrew Menard, a healthcare executive and cancer survivor, shared insights that resonated with me. He embraces automation and automates much of his day. Yet, after his cancer journey, he believes technology should handle tasks humans can’t reliably manage, allowing clinicians to focus on uniquely human tasks.
He also believes patients should not be passive in their care.
“Listen to your own body,” he advised. “Trust your instincts. Push.”
He makes a valid point.
Patients should not be passive in their healthcare. They play a role in the safety system, noticing when something is amiss, asking questions, and seeking clarity when answers are lacking.
AI won’t eliminate uncertainty. Medicine has never been absolute, and likely never will be. Clinical judgment amidst uncertainty remains a human responsibility. Technology’s role is not to dehumanize but to bolster healthcare’s resilience to human realities.
Clinicians change shifts. Patients feel overwhelmed. Families have responsibilities. Life moves on. Rather than ignoring these realities, technology can help health systems by remembering what we forget, communicating lost information, and ensuring critical findings don’t vanish amidst daily demands.
Four months ago, this was the message I hoped to convey to the Senate.
Senator Graham’s passing is a somber reminder that the future of healthcare AI isn’t about replacing physicians. It’s about developing healthcare systems that better support clinical teams, empower patients, implement preventive health principles, and safeguard the ordinary moments that healthcare aims to protect. The objective isn’t merely accurate diagnosis. It’s about giving people more opportunities to drive their children to school, care for their aging parents, and live the lives they aspire to lead.
