Acute Risk Monitoring for Oncology Therapy Regimen (ARMOR)

Acute Risk Monitoring for Oncology Therapy Regimens (ARMOR): A Silent Prospective Validation of a Machine Learning Model

Patients undergoing outpatient infusion systemic therapy for cancer are at risk for potentially preventable, unplanned acute care in the form of emergency department (ED) visits and hospitalizations. These events impact patient outcomes, treatment decisions, and healthcare costs. To address this need, the Centers for Medicare & Medicaid Services developed the chemotherapy measure (OP-35). Recent randomized controlled studies indicate that electronic health record (EHR)-based machine learning (ML) approaches accurately direct supportive care to reduce acute care during radiotherapy. This study aims to develop and prospectively validate ML approaches to predict the risk of OP-35 qualifying, potentially preventable, acute care events within 30 days of infusion systemic therapy.

Study Overview

• Status: Completed
• Conditions: Cancer; Acute Care Service Utilization
• Intervention / Treatment: Other: Medical record review

Detailed Description

OBJECTIVES:

I. Develop and retrospectively validate electronic health record-based machine learning models using routinely collected clinical data from patients receiving systemic therapy to predict risk of potentially preventable OP-35 qualifying acute care events. (Phase 1: Retrospective)

II. Prospectively validate machine learning models across distinct time periods. (Phase 2: Prospective)

III. Understand patterns of care by stratifying and analyzing model performance by treatment type, cancer diagnosis, and race/ethnicity to assess bias and disparities in outcomes.

OUTLINE:

Retrospective and prospective clinical data obtained from medical records will be used to develop and validate predictive machine learning models. Prospective data will be divided into 2 phases: Prospective validation (PV) 1 and PV 2.

• Study Type: Observational
• Enrollment (Actual): 4740

Contacts and Locations

Study Locations

• United States
  • California
    • San Francisco, California, United States, 94143
      • University of California, San Francisco

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

Adult patients receiving care for cancer

Description

Inclusion Criteria:

• Patients 18 years or older diagnosed with cancer who receive care at UCSF and/or one of the UCSF affiliate locations.

Exclusion Criteria:

• Patients under the age of 18.
• Patients receiving care as part of a clinical trial.

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Patients receiving cancer therapy at University of California, San Francisco (UCSF); All adults undergoing systemic cancer-related therapy from July 2017 to March 2024 at any UCSF outpatient, infusion center with available OP-35 data.	Other: Medical record review; Retrospective chart reviews for data collection will be conducted.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Area under the receiver operating characteristic curve (AUROC) for OP-35 prediction model.	UCSF patients receiving infusion systemic therapy had clinical data incorporated into machine learning (ML) models to predict risk of Centers for Medicare & Medicaid Services Chemotherapy Measure (OP-35) qualifying acute care events within 30 days of infusion. Models included variables such as cancer diagnosis, therapeutic agents, and laboratory values. Three ML approaches were employed to train models in predicting OP-35 events. Models were trained and retrospectively validated on data from July 7, 2017, to February 11, 2021, and prospectively validated on 2 cohorts: April 17, 2023, to October 29, 2023 (PV1) and February 19, 2024, to March 31, 2024 (PV2) to generate a validation AUROC. The initial prospective validation occurred over a pre-planned period with the assumption of a 2% event rate, based on the model development data, with an alpha of 0.05 and 84% power to detect an AUROC of 0.75, requiring a sample size of at least 8000 infusions.	Up to 6.75 years

Collaborators and Investigators

• Sponsor: University of California, San Francisco
• Collaborators: National Cancer Institute (NCI); Conquer Cancer Foundation
• Investigators: Principal Investigator: Julian Hong, MD, MS, University of California, San Francisco

Study record dates

Study Major Dates

• Study Start (Actual): July 1, 2017
• Primary Completion (Actual): March 31, 2024
• Study Completion (Actual): March 31, 2024

Study Registration Dates

• First Submitted: May 15, 2026
• First Submitted That Met QC Criteria: May 15, 2026
• First Posted (Actual): May 22, 2026

Study Record Updates

• Last Update Posted (Actual): May 22, 2026
• Last Update Submitted That Met QC Criteria: May 15, 2026
• Last Verified: May 1, 2026

More Information

Terms related to this study

• Keywords: Chemoradiation; Chemotherapy; Artificial Intelligence; Radiation therapy; Machine Learning; Acute Care; Risk Cancer Care Delivery
• Additional Relevant MeSH Terms: Neoplasms
• Other Study ID Numbers: 20721 (DAIDS-ES Registry Number); R01CA277782 (U.S. NIH Grant/Contract)

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: YES
• IPD Plan Description: De-identified data may be shared with study collaborators during the course of the study.

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No