Q&A: Closing biomarker testing gaps in early-stage NSCLC with AI-enabled clinical intelligence

Karen Huelsman: One of the biggest hurdles is the sheer pace of change. The NCCN guidelines for non-small cell lung cancer changed nine times in 2026 and 11 times the year before. It’s incredibly difficult for clinicians, especially those in community or rural settings who see all tumor types, to keep up with every update, like when c-MET was added as a key biomarker. This is compounded by operational challenges, such as pulling data from non-discrete fields in the EHR and the manual effort of chart review.

Heather Rudicil: From a health system perspective, we are always trying to create the ‘easy button’ for our clinicians. The complexity of oncology, the constant guideline changes, and the administrative burden are significant. We need to find ways to support our providers and ensure that even with these challenges, every patient gets the right test at the right time without adding to provider burnout.

Dr. Ezra Cohen: The core idea is to create an AI-enabled bridge for the gaps that occur in a patient’s journey. The Tempus Next platform works by integrating and analyzing multimodal data from the EHR, including both structured data like lab results and unstructured data like progress notes and imaging reports. Advanced machine learning algorithms screen patient records in real time to identify individuals who have a potential care gap—for instance, a patient with eNSCLC who has not had guideline-recommended biomarker testing. The system then generates a notification for the designated care team member, allowing them to review the case and close the gap. It’s a safety net that supports the provider, who always makes the final clinical decision.

Dr. George Simon: It is critically important. Identifying 18% more patients with EGFR mutations has profound therapeutic consequences. Without that data, you would not give a patient adjuvant osimertinib, which we know from the ADAURA trial data provides an impressive survival benefit. You might also prevent a patient from receiving a treatment that is ineffective and potentially toxic, like immunotherapy in this context. The same is true for ALK. These numbers have direct consequences for patient outcomes. It surprised me that we saw a lift in PD-L1 testing, as I would have thought it was automatic, but it shows that gaps can exist even for the most established biomarkers.

Heather Rudicil: You have to identify your key stakeholders, create transparency, and define the “why” behind the initiative. For us, a key part of the process was leveraging our specialty testing coordinators. They act as a safety net and manage the in-basket messages, giving clinicians the autonomy to make decisions but ensuring the care gap is addressed. To avoid alert fatigue, you must be disciplined. We have a strong committee that governs when an alert is activated, and if it has low engagement or isn’t actionable, we turn it off.

Karen Huelsman: We built a lot of trust in the early days by being the “human in the loop.” I was manually reviewing the care gaps surfaced by the platform to ensure they were accurate before alerting a clinician. Now that the model is so well-trained, the accuracy is spot-on. The feedback from providers has been overwhelmingly positive. When I reach out about a care gap, the response is usually, “Thank you, I’m so glad you pointed this out.” It’s a low lift for me now, but it has a huge impact.

Karen Huelsman: We’ve seen several downstream benefits. First, we used the data from this quality improvement project for our Commission on Cancer (CoC) accreditation, which was very helpful. Second, it has advanced our health equity goals. By automating the identification of patients, we remove potential biases. We ran internal reports and found no differences in our tumor profiling rates by zip code, race, or social vulnerability index. I call it “health tech-quity.” Finally, by partnering with our specialty pharmacy team, we used the data to demonstrate the downstream return on investment of our precision oncology testing program.

Dr. George Simon: The lessons learned here can be translated to all malignancies. I believe every cancer patient should be sequenced. What was non-actionable yesterday is actionable today, like with KRAS. My message to my team is, “If in doubt, sequence.” For health systems, the key is to involve stakeholders early, over-communicate, and make the process as easy as possible for busy clinicians.

Heather Rudicil: My advice is to get a baseline. Understand where you are today so you can tell your story and show the “why” to get clinician buy-in. Don’t be afraid to ask questions and lean on experts. We are all learning this new space together.

Karen Huelsman: Embrace innovation and just start somewhere. Many groups ask where to begin, but the important thing is to get moving. You learn so much along the way, and that informs what you do next.