Master's Student McGill University Health Centre Mercier, Quebec, Canada
Background: Despite the growing number of clinical trials investigating digital health technologies (DHTs) for cardiovascular disease (CVD) prevention and monitoring – particularly in the post-COVID-19 era – data missingness in contemporary DHT studies involving heart failure (HF) patients, specifically in sensor data, remains largely unexplored. This study examined the rate of sensor data missingness in a DHT trial focused on HF patients.
METHODS AND RESULTS: To assess data missingness, we implemented a data quality pipeline to analyze baseline data. This pipeline integrated statistical frameworks to evaluate two key factors: data missing (not-a-number [NaN] score) and background noise (signal-to-noise ratio [SNR]). This approach helped identify sessions that required further analysis, including a collaborative review of session notes.
We implemented our framework in the WEAR-HF clinical trial (NCT06335264), a cross-sectional case-control study designed to establish a digital biomarker for detecting prevalent HF in 50 participants (24 cases, 26 controls; 45% female; median age: 62.5 years [interquartile range: 55.25-70.0]). Our analysis revealed that 16 participants (32%) had poor signal quality or data incompleteness, primarily due to electrodermal activity (EDA) and photoplethysmography (PPG) sensors. Specifically, 12 cases were linked to EDA, 2 to PPG, and 2 to both. Participants with low-quality data were recontacted for repeat sessions, which lead to significant improvement in the clarity and stability of the PPG signal (fig. 1). Additionally, the following device enhancements were implemented: real-time signal quality feedback and improved PPG sensor alignment. Following these modifications, the proportion of high-quality recording sessions increased from 47.83% to 62.35%. Overall, the number of flagged participants decreased from 16 (32%) of the total, to 6 (12%) subsequent to the implementation of these two strategies.
Conclusion: Our findings underscored the importance of standardized pipelines for identifying potential missing data in DHT-related clinical studies involving HF patients. Implementing such standardization can enhance trial design and improve result reliability.
