arXiv subjects

Gunnar Treff

Publications and source records attributed to Gunnar Treff.

4 recordsLinked to original sources

arXiv2026

Exploring Self-Tracking Practices of Older Adults with CVD to Inform the Design of LLM-Enabled Health Data Sensemaking

Wearables and mobile health applications are increasingly adopted for self-management of chronic illnesses; yet the data feels overwhelming for older adults with cardiovascular disease (CVD). This study explores how they make sense of self-tracked data and identifies design opportunities for Large Language Model (LLM)-enabled support. We conducted a seven-day diary study and follow-up interviews with eight CVD patients aged 64-82. We identified six themes: navigating emotional complexity, owning health narratives, prioritizing bodily sensations, selective engagement with health metrics, negotiating socio-technical dynamics of sharing, and cautious optimism toward AI. Findings highlight that self-tracking is affective, interpretive, and socially situated. We outline design directions for LLM-enabled data sensemaking systems: supporting emotional engagement, reinforcing patient agency, acknowledging embodied experiences, and prompting dialogue in clinical and social contexts. To support safety, expert-in-the-loop mechanisms are essential. These directions articulate how LLMs can help translate data into narratives and carry implications for human-data interaction and behavior-change support.

cs.HC↗

arXiv2025

Towards Data-Enabled Physical Activity Planning: An Exploratory Study of HCP Perspectives On The Integration Of Patient-Generated Health Data

Physical activity planning is an essential part of cardiovascular rehabilitation. Through a two-part formative design exploration, we investigated integrating patient-generated health data (PGHD) into clinical workflows supporting shared decision-making (SDM) in physical activity planning. In part one, during a two-week situated study, to reduce risk of working with cardiovascular disease patients, we recruited healthy participants who self-tracked health and physical activity data and attended a physical activity planning session with a healthcare professional (HCP). Subsequently both HCPs and participants were interviewed. In part two, findings from part one were presented to HCPs in a card-sorting workshop to corroborate findings and identify information needs of HCPs alongside patient journeys and clinical workflows. Our outcomes highlight HCP information needs around patient risk factors, vital signs, and adherence to physical activity. Enablers for PGHD integration include adaptive data sense-making, standardization and organizational support for integration. Barriers include lack of time, data quality, trust and liability concerns. Our research highlights implications for designing digital health technologies that support PGHD in physical activity planning during cardiac rehabilitation.

cs.HC↗

arXiv2025

Towards a Real-Time Warning System for Detecting Inaccuracies in Photoplethysmography-Based Heart Rate Measurements in Wearable Devices

Wearable devices with photoplethysmography (PPG) sensors are widely used to monitor heart rate (HR), yet often suffer from accuracy issues. However, users typically do not receive an indication of potential measurement errors. We present a real-time warning system that detects and communicates inaccuracies in PPG-derived HR, aiming to enhance transparency and trust. Using data from Polar and Garmin devices, we trained a deep learning model to classify HR accuracy using only the derived HR signal. The system detected over 80% of inaccurate readings. By providing interpretable, real-time feedback directly to users, our work contributes to HCI by promoting user awareness, informed decision-making, and trust in wearable health technology.

cs.HC↗

arXiv2024

The Last JITAI? Exploring Large Language Models for Issuing Just-in-Time Adaptive Interventions: Fostering Physical Activity in a Conceptual Cardiac Rehabilitation Setting

We evaluated the viability of using Large Language Models (LLMs) to trigger and personalize content in Just-in-Time Adaptive Interventions (JITAIs) in digital health. As an interaction pattern representative of context-aware computing, JITAIs are being explored for their potential to support sustainable behavior change, adapting interventions to an individual's current context and needs. Challenging traditional JITAI implementation models, which face severe scalability and flexibility limitations, we tested GPT-4 for suggesting JITAIs in the use case of heart-healthy activity in cardiac rehabilitation. Using three personas representing patients affected by CVD with varying severeness and five context sets per persona, we generated 450 JITAI decisions and messages. These were systematically evaluated against those created by 10 laypersons (LayPs) and 10 healthcare professionals (HCPs). GPT-4-generated JITAIs surpassed human-generated intervention suggestions, outperforming both LayPs and HCPs across all metrics (i.e., appropriateness, engagement, effectiveness, and professionalism). These results highlight the potential of LLMs to enhance JITAI implementations in personalized health interventions, demonstrating how generative AI could revolutionize context-aware computing.

cs.HC↗