Abstract:
|
Advances in digital health technologies (DHTs) enable us to derive new clinical endpoints from domains impractical or impossible to measure before, leading to a better understanding of more meaningful health outcome measures and disease characterization. While most research has been conducted to support device selection, algorithm development, and the reliability of digital measures, there is an unmet need to address missing data observed in the complex time series signals collected by DHTs. In this work, we will discuss the causes and mechanisms of DHT missing data using examples from accelerometry and continuous glucose monitoring. We will then review emerging methods to address missingness in different DHT data types such as functional data analysis and robust feature engineering for both epoch level and daily summary data. Lastly, we will discuss strategies for study design and conduct to prevent missing data in clinical trials, including optimizing DHT deployment and capturing the participants’ perspectives. We hope this session serves as a starting point for further discussion among clinical trial stakeholders and enables the development of robust guidance.
|