Human-Centered Design and Iterative Refinement of Tools and Methods to Implement a Surveillance and Risk Prediction...

Area of Science:

• Oncology
• Health Informatics
• Clinical Surveillance Systems

Background:

• Clinical deterioration in cancer outpatients is a significant cause of preventable harm.
• Timely intervention is critical for medically complex patients to mitigate adverse events.
• Effective clinical surveillance, early recognition, and prompt notification are essential for managing patient decline.

Purpose of the Study:

• To assess the feasibility of developing a surveillance-and-risk prediction system for cancer outpatients.
• To create tools and processes for detecting clinical deterioration in this population.
• To evaluate the system's potential for early risk identification and intervention.

Main Methods:

• Employed systems engineering and human-centered design to develop a prototype system.
• Integrated passive (wearable sensors) and active (patient reporting, EHR data) surveillance methods.
• Utilized deep learning models for risk prediction and evaluated usability via patient/clinician interviews and the System Usability Scale (SUS).

Main Results:

• The system successfully predicted 7-day risk of unplanned treatment events (UTEs) using patient-reported outcomes and EHR data.
• Deep learning models achieved high performance (AUC-ROC: 0.983) in predicting UTEs.
• The risk communication prototype received favorable clinician ratings (SUS score: 76).

Conclusions:

• Demonstrated the feasibility of a surveillance-and-risk prediction system for cancer outpatients.
• The system shows promise for detecting and reporting clinical deterioration.
• Further research is required for full implementation and evaluation of system adoption and effectiveness.