Researchers Created Breast Cancer Mortality Model

A new tool helps oncologists predict 90-day mortality risk for patients with metastatic breast cancer.

Updated on Oct. 6, 2026 in Cancer

Researchers Created Breast Cancer Mortality Model

Live Poll

Should medical AI tools be used to prompt end-of-life conversations for terminal patients?

Researchers have developed a prognostic model that predicts 90-day mortality risk for metastatic breast cancer patients using clinical data from electronic health records. The tool achieved internal accuracy rates of up to 83%, offering a way to identify patients who may benefit from early palliative care.

Why it matters

Current prognostic tools often fail to serve metastatic breast cancer patients effectively, leaving oncologists without reliable support to determine when to initiate end-of-life discussions. This model aims to fill that gap by automating risk identification within routine clinical practice.

The study cohort included 9,270 adults, with the model demonstrating internal validation area under the curve scores of 0.81 and 0.68 in external testing. Accuracy for predicting 90-day mortality ranges from 72% to 83%.

The players

UNC Lineberger Comprehensive Cancer Center

This facility is a cancer research and treatment organization that served as the home institution for the study's lead researchers.

CancerLinQ Discovery

This is a platform designed to aggregate and deidentify patient data from electronic health records for use in clinical oncology research.

The details

The model calculates risk by analyzing variables such as lab results, BMI, vital signs, and recent treatment history. Key indicators include declining sodium or albumin levels and recent chemotherapy discontinuation, all pulled automatically from electronic health records.

The study cohort included patients diagnosed between 2000 and 2020.
External validation data was collected between 2015 and 2017.

The Big Picture

This development represents a shift toward data-driven oncology by utilizing the CancerLinQ Discovery platform to modernize prognostic care. This study follows the precedent of leveraging large-scale real-world data sets to improve clinical decision-making tools.

This model could lead to earlier discussions about palliative care options for patients with advanced breast cancer. It potentially allows clinicians to provide more personalized support based on a patient's specific laboratory values and health indicators.

The takeaway

Predictive models like this help oncologists align medical treatment with patient priorities during the final stages of illness. Integrating such tools into daily practice allows for more timely transitions to palliative care services.