A Multi-Site Dashboard for Evaluating HiGHmed Cardiovascular Sensor Data from Apple Watch

The HiGHmed use case cardiology aims to enable the early detection of clinical worsening of patients with chronic heart failure. Two complementary studies are introduced, comprising: 1) a routine clinical study capturing in-hospital patient data, and 2) a sensor sub-study monitoring patients remotely through Apple Watches. Wearable devices continuously collect cardiovascular metrics, including heart rate and step count, and transmit data to participating medical data integration centers. This study presents a dashboard tool designed to support cardiologists in evaluating multi-site sensor data collaboratively.

We developed a centralized R Shiny dashboard that aggregates Apple Watch sensor data with hospitalization records from 12 of 93 participants across four study sites. The dashboard integrates device-recorded metrics (heart rate and steps), manually recorded clinical parameters (resting heart rate, blood pressure, and weight), Kansas City Cardiomyopathy Questionnaire (KCCQ-12) scores, and hospitalization events. Three analytical modules were constructed: 1) an observation overview displaying stacked multi-metric trends, 2) a heart rhythm analyzer presenting hourly heart rate patterns with three-day rolling windows, and 3) a correlation module visualizing relationships between metrics using 14-day sliding windows.

The observation overview displays pre-hospitalization deterioration, including low KCCQ-12 scores (e.g., ∼50) and reduced activity levels (daily steps dropping below 2,000). Hourly heart rhythm analysis demonstrates a decline observed on hospitalization day compared to baseline. Correlation analysis confirmed expected physiological relationships, including a strong positive correlation between steps and device heart rate (r ≈ 0.8). There are visible gaps in sensor recordings and inconsistent adherence to manual measurement, suggesting data quality challenges. The aggregation of cardiovascular metrics from multiple sources is essential for monitoring patient trajectories during and outside hospital stays. This work introduces a single integrated overview of metrics that could signal clinical worsening in cardiac patients.

The sensor data analysis dashboard provides a practical tool for multi-site cardiovascular data evaluation, enabling clinicians to transition from case-by-case parameter review to systematic multi-metric assessment of cardiac patients.