EHR Model Flags High-Risk Periods in Patients with Metastatic Breast Cancer

Determining when patients with metastatic breast cancer are nearing the end of life remains difficult, often leading to intensive interventions with limited benefit. Prognostic uncertainty can delay goals-of-care discussions, supportive services, and planning for symptom management, while clinicians may also overestimate survival. To support earlier, patient-centered decisions, researchers have developed a data-driven tool that identifies periods of elevated mortality risk across hospital and outpatient settings.

The metastatic breast cancer–specific prognostic model was developed by a team led at UNC Lineberger Comprehensive Cancer Center and the University of North Carolina at Chapel Hill School of Medicine (Chapel Hill, NC, USA). It estimates the probability of death within 30 or 90 days. The regression-based approach uses routinely collected electronic health record data, including laboratory results, vital signs, breast cancer subtype, and medications. The goal is to help oncologists recognize high‑risk periods and initiate timely discussions about care needs and patient preferences.

New findings published in JCO Oncology Practice on September 4, 2026 describe the development and validation of the model using CancerLinQ Discovery, a national oncology database with real‑world data. Unlike tools that combine multiple cancer types and underrepresent breast cancer, this model is specific to metastatic breast cancer and can identify high‑risk periods even for patients who have lived a long time with the disease. The variables most associated with increased risk align with familiar clinical signals, including worsening liver function, increased heart rate, escalating needs for pain medication, and declining ability to perform self‑care.

Investigators indicate the tool could be implemented in routine practice to support better end‑of‑life care. It may encourage reassessment of treatment options, earlier involvement of palliative care specialists when appropriate, and closer alignment of care plans with patient values and goals. Future studies are planned to determine how best to integrate the model into clinical workflows and to evaluate its impact on decision‑making by oncologists, patients, and caregivers.

"In our prognostic model, the variables most closely correlated with increased risk align with signals clinicians already recognize, like worsening liver function, increased heart rate, rising needs for pain medication and declining ability to care for oneself at home. Any clinician could tell you those are signs that a patient is getting sicker, but sometimes we still miss them and don't do enough to prepare patients and families,” said Emily Ray, M.D., M.P.H., medical oncologist at UNC Lineberger Comprehensive Cancer Center and associate professor in the UNC School of Medicine's Division of Oncology.

"Our model reinforces that clinical intuition and could help oncologists better recognize these patterns and act on them to arrange more support for patients and caregivers," said Ray.

Related Links
UNC Lineberger Comprehensive Cancer Center 
UNC School of Medicine