"Less words, more pictures": creating and sharing data visualizations from a remote health monitoring system with clinicians to improve cancer pain management.

Abstract

Background: The Behavioral and Environmental Sensing and Intervention for Cancer (BESI-C) is a remote health monitoring system (RHMS) developed by our interdisciplinary team that collects holistic physiological, behavioral, psychosocial, and contextual data related to pain from dyads of patients with cancer and their family caregivers via environmental and wearable (smartwatch) sensors.

Methods: R, Python, and Canva software were used to create a series of static and interactive data visualizations (e.g., visual representations of data in the form of graphs, figures, or pictures) from de-identified BESI-C data to share with palliative care clinicians during virtual and in-person 1-hour feedback sessions. Participants were shown a sequence of 5-6 different data visualizations related to patient and caregiver self-reported pain events, environmental factors, and quality of life indicators, completed an electronic survey that assessed clarity, usefulness, and comprehension, and then engaged in a structured discussion. Quantitative survey results were descriptively analyzed and "think aloud" qualitative comments thematically summarized and used to iterate data visualizations between feedback sessions.

Results: Six to 12 interdisciplinary palliative care clinicians from an academic medical center, a local hospice, and a community hospital within Central Virginia participated in five data visualization feedback sessions. Both survey results and group discussion feedback revealed a preference for more familiar, simpler data visualizations that focused on the physical aspects of pain assessment, such as number of high intensity pain events and response to pharmacological interventions. Preferences for degree of data granularity and content varied by discipline and care delivery model, and there was mixed interest in seeing caregiver reported data. Overall, non-physician participants expressed greater interest in visualizations that included environmental variables impacting pain and non-pharmacological interventions.

Conclusion: Clinicians desired higher-level (i.e., less granular/detailed) views of complex sensing data with a "take home" message that can be quickly processed. Orienting clinicians to unfamiliar, contextual data sources from remote health monitoring systems (such as environmental data and quality of life data from caregivers) and integrating these data into clinical workflows is critical to ensure these types of data can optimally inform the patient's plan of care. Future work should focus on customizing data visualization formats and viewing options, as well as explore ethical issues related to sharing data visualizations with key stakeholders.