“It can bring you in the right direction”: Episode-Driven Data Narratives to Help Patients Navigate Multidimensional Diabetes Data to Make Care Decisions

Title of the Paper

“It can bring you in the right direction”: Episode-Driven Data Narratives to Help Patients Navigate Multidimensional Diabetes Data to Make Care Decisions

Paper Information

• Subject Area: Health informatics, patient data analysis, digital health management
• Keywords: diabetes, personalized health data, human-data interaction, data-driven decision support, health data visualization, personal health informatics, patient-generated data, multidimensional data interfaces

Research Background and Problem Statement

• Problems and Challenges:

  • Diabetes patients need to process multiple streams of health data to support self-care, but existing tools have limited capability to transform data into actionable information.
  • The complexity of multidimensional data, such as blood glucose, diet, insulin, and physical activity, makes data analysis and decision-making difficult.
  • Patient-generated health data reports often focus on functionality rather than intuitiveness, leading to high cognitive load and difficulty in formulating action plans.

• Significance:

  • Diabetes is a chronic disease, and data-driven self-management is crucial for improving blood glucose control and reducing medical dependency.
  • Given the comprehensive impact of multidimensional factors on health, tools need to help patients understand and connect data to facilitate effective decision-making.

• Motivation and Related Work:

  • Previous research has primarily focused on collecting and visualizing patient data, with limited exploration of how multidimensional data can support decision-making.
  • Commercial diabetes data visualization tools are mainly designed for clinicians and are not suitable for independent patient use, limiting data engagement and personal decision-making capabilities.

Solution

• Methods and Solution:

  • Proposed a new set of data interfaces called “Episode-Driven Data Narratives.”
  • The design aims to help patients independently engage with multidimensional data to support critical decision-making tasks in diabetes management.

• Innovations:

  • Emphasizes structuring data presentation around “episodes,” enabling users to focus on data performance and related factors within specific time periods.
  • Integrates domain-inspired heuristic rules to interpret relationships between data streams, using representative data to demonstrate connections.
  • Provides a panoramic view of different episodes and potential factors, supporting patients in selecting and implementing personalized interventions.

• Implementation Steps and Techniques:

  1. Define design goals for the data interface, including supporting multidimensional data interaction, episode-driven data presentation, and contextualization of domain-inspired questions.
  2. Create prototype data narratives using diabetes patient data (CGM and insulin pump), incorporating domain knowledge to establish interpretable causal inferences and association questions.
  3. Test the usability and comprehensibility of the data-driven tool through interactive interface design (implemented in Figma) and compare it with existing commercial tools.

Research Findings

• Specific Findings:

  • Compared to commercial diabetes data reports, episode-driven data narratives improved data comprehension and reduced cognitive load during data analysis.
  • Facilitated patients in identifying and prioritizing complex multidimensional data issues, such as selecting episodes requiring intervention and constructing action plans tailored to influencing factors.

• Comparative Advantages:

  • Significantly improved performance in data comprehension tests (91 points vs. 69 points for commercial tools).
  • NASA Task Load Index assessments revealed a significant reduction in task complexity burden.
  • Encouraged patients to proactively analyze potential interventions, establishing connections between influencing factors and validating evidence.

• Experimental and Evaluation Results:

  • Collected participant behavior data during analysis using two tools, comparing task efficiency and outcome quality.
  • Using episode-driven data narratives, patients were able to more clearly identify potential intervention factors and prioritize high-impact factors.
  • Experiments validated the value of intuitive episode-driven tools when working with multidimensional data.

• Limitations and Future Directions:

  • Small sample size; broader applicability of findings requires larger-scale validation.
  • Research subjects were primarily limited to diabetes patients; applications for other diseases need further exploration.
  • For chronic conditions like depression, causal factors may not be directly measurable, requiring adjustments to episode-driven design for such complexities.

Conclusion and Design Insights

• Tools designed to assist patients in using multidimensional data should simultaneously reduce unnecessary data exploration while allowing a degree of exploration to promote decision-making autonomy.
• By structuring data presentation around episodes and contextualizing domain knowledge, users can interpret and act on health data more intuitively, enhancing the potential for data-driven health management.