Example data entry and display for the BMETS Decision Support Platform, to be published on Oncospace (Johns Hopkins Department of Radiation Oncology), pending peer review and publication.

This tool can be used at the time of consultation for palliative radiotherapy to symptomatic bone metastases in order to estimate patient survival time following consultation. Providers can enter values for the patient, disease, and treatment variables listed below, and a predicted survival plot will be displayed, based on the Bone Metastases Ensemble Trees for Survival (BMETS) machine learning model (link to publication pending).

Briefly, the BMETS model was built using a database of 397 patients seen in Radiation Oncology consultation for symptomatic bone metastasis at the Johns Hopkins Hospital between 2007 and 2013. Each patient’s chart was retrospectively reviewed to collect demographic, disease, and treatment information felt to be potentially useful in predicting survival. His or her actual survival time following consultation was documented. The BMETS model was then built using a random survival forests algorithm to predict survival time following consultation based on 27 patient-specific characteristics.

Enter your patient’s information below.

If a value is unknown, leave the entry blank or unselected.

Chest wall
Select all that apply
Please select
IV therapy
Select all that apply
Select all that apply

Most recent lab values, within the prior 6 weeks:

(cells/cubic mm)
(cells/cubic mm)

Predicted Survival Curve

The interactive orange plot above demonstrates the predicted survival curve within the 12 months following radiation oncology consultation for the specific patient based on the characteristics selected above. The blue curves demonstrate the predicted survival for all other patients with symptomatic bone metastases in the BMETS database, arranged from lowest (dark blue) to highest (light blue) predicted survival. These blue curves are displayed for comparison purposes only.

NOTE: The plot displaying the patient’s predicted survival reflects a predicted value from the BMETS model. While the model is calibrated to be as accurate as possible across all patients, the predicted survival time may underestimate or overestimate an individual patient’s actual survival time.

Click here for a tutorial on using and interpreting output for the predicted survival plot.