Most physicians and nurses understand the value of national practice guidelines, but they also see the importance of bedside experience and real-world patient outcomes. CODE-M addresses all three issues.
By John Halamka, M.D., M.S., Dwight and Dian Diercks President, Mayo Clinic Platform and Paul Cerrato, MA, senior research analyst and communications specialist, Mayo Clinic Platform
In many respects, practice guidelines take the guess work out of diagnosis and treatment by relying on large randomized controlled trials and observational studies to make reliable recommendations. This evidence-based approach has taken medicine out of the dark ages so to speak when practitioners based their actions on a handful of case reports or the pronouncements of experts. Unfortunately, there are many situations in which these guidelines don’t provide clear answers to enable clinicians struggling with a patient who isn’t responding the officially sanctioned therapy. Clinical guidelines are based on ideal patient populations; they typically exclude patients with co-existing diseases, for example. Patients in everyday practice, on the other hand, often present with co-morbidities, prior treatments, changes in their lab values over time, incomplete medical records, and much more.
To address this disparity, Svetomir Markovic, MD, PhD, an oncologist with the Mayo Clinic Comprehensive Cancer Center, and his colleagues have developed an AI-enhanced system called CODE-M. It “extends evidence-based oncology care when standard therapy no longer provides clear answers, helping clinicians explore patients like mine, understand real-world outcomes, and prepare for decisions with greater context and confidence.” In addition to combining the best that national guidelines and clinical experience can offer, it captures all relevant data on all available patients who are similar to the individual patient in front of them. (Figure 1)
The “Patients like mine” approach is relatively new and provides a third perspective that complements national guidelines from clinical trials and real-world clinical experience. Nigam Shah, MBBS, PhD, Chief Data Scientist at Stanford Health and his team have developed a specialty consultation service consisting of medical and informatics experts that provide clinicians with EHR data to help fill in the gaps for patients who don’t fit the idealized results of clinical trials.
Dr. Markovic has taken a similar approach. However, CODE-M’s approach doesn’t just look for a diagnosis that is similar to that of other patients; it aligns each patient based on clinical context, treatment sequence, molecular features, and outcomes, taking a longitudinal perspective that enables clinicians to shift their therapeutic plan over time. And clinicians can see why patients matched, review attribute importance, and adjust cohort definitions. One of the most unique aspects of CODE-M is ordinal patient alignment. The system is not simply matching diagnoses or demographics; it is aligning patients at comparable points in their disease and treatment journey.
And if that treatment plan fails and no other options are available, CODE-M will use its AI capabilities to search for an ongoing clinical trial that is accepting new candidates. The system evaluates eligibility, identifies near matches, and ranks opportunities.
Figure 1
The first step in the CODE-M process is the pre-visit summary, which provides the attending physician with a detailed AI-enabled roadmap that prepares them for the patient about to walk in the door by generating intel most practitioners rarely have at their fingertips. As Figure 2 illustrates, it provides a clinical snapshot, the latest clinical guidelines, the experiences of several patients who had traveled a similar road as the current patient, and a list of clinical trials that may fit their circumstances.
Figure 2
The patients like mine part of the process, illustrated in Figure 3, offers clinicians a detailed comparison of the cohort of similar patients. The ability to show subsequent therapies and outcomes among similar patients is one of the most valuable features for clinicians facing difficult decisions.
Figure 3
Subsequent steps in the CODE-M system include an in-depth look at matching clinical trials, longitudinal tissue and molecular markers, and what is referred to as “real-time patient reported events aligned to the treatment journey.” That includes any new symptoms they report and new wearable data. This unique feature (Figure 4) enables the physician to make adjustments to the patient’s regimen. Its Vigilant capability enables earlier detection of symptoms and providing clinicians with additional context for intervention. These signals are often missed in most doctor/patient encounters but provide a indispensable checkpoint in the CODE-M system.
Figure 4
CODE-M will provide the Mayo Model of Care in a quick summary, in near real time. The provider is presented with a tool that alleviates some of the challenges of today’s doctor/patient relationship, including the 15-minute window most clinicians are given to collect and assess the patient status. It also enables them to investigate treatment options, and evaluate trial opportunities, while still providing relational care to the patient. Finally, the technology design allows CODE-M to go beyond cancer and into other specialties.
Share this:
