Cancer treatments can cause a variety of poorly-managed symptoms and toxicities that can impair quality of life and functioning and lead to early discontinuation of life-prolonging medical treatment. The goals of the proposed study are (1) to develop a novel mobile sensing system to passively monitor symptom burden during chemotherapy and (2) to evaluate the feasibility and acceptability of using this system to recommend patient-provider communication. To accomplish Aim 1, we will enroll 200 oncology patients starting a new chemotherapy regimen and will collect smartphone and wearable sensor data continuously as well as daily patient-reported symptoms via smartphones for 90 days. Using machine learning methods, we will retrospectively analyze data from the first 100 participants and develop a generalizable and parsimonious population-level model to identify severe symptom days. For the second 100 participants, we will run these computational models in real-time and prospectively evaluate the accuracy of our classifications relative to patient-reported symptoms.
For Aim 2, we will enroll 50 patients in a prospective single-arm trial of a system that uses inferences based on sensing and machine learning to recommend contact with providers and will evaluate patient accrual, attrition, and compliance as well as information from both patients and their providers about acceptability and perceived usefulness of the system. The scientific premise of this proposal is that mobile sensing of subtle fluctuations in behavior coupled with real-time computational modeling could enable earlier detection and ultimately better management of severe symptoms during cancer treatment. The proposed project builds on our prior research and completes the necessary development and feasibility work to support a large multisite trial comparing the mobile sensing system to standard of care to determine effects on symptom burden, quality of life, health care utilization, and survival.
Symptoms during cancer treatment are common, compromise patient quality of life and ability to tolerate treatment, and often go undetected at clinic visits. This project aims to combine smartphone and wearable sensor data with machine learning to remotely monitor symptoms during chemotherapy and to develop a system that automatically recommends patient-provider contact when severe symptoms are detected.
