METHODS FOR CLUSTERING TIME SERIES DATA ACQUIRED FROM MOBILE HEALTH APPS

Area of Science:

• Digital Health
• Data Science
• Respiratory Medicine

Background:

• Mobile health apps collect extensive patient data for chronic disease management.
• Longitudinal data from asthma apps present challenges due to missing values from varying user engagement.
• Identifying patient subgroups and predicting adverse events is crucial for asthma care.

Purpose of the Study:

• To develop and validate a novel statistical method for analyzing complex, longitudinal mobile health data with missing values.
• To identify distinct asthma patient phenotypes using data from the Asthma Mobile Health Study (AMHS).

Main Methods:

• Proposed a probability imputation model to infer missing data in longitudinal survey responses.
• Employed a consensus clustering strategy combined with multiple imputation.
• Validated the method through simulation studies and application to AMHS data.

Main Results:

• The proposed imputation and clustering method demonstrated superior performance compared to low-rank matrix completion.
• Analysis of AMHS data identified several patient groups with distinct asthma phenotype patterns.
• The method effectively handles heterogeneous missing data structures common in mobile health studies.

Conclusions:

• The developed method offers a robust solution for analyzing large, complex mobile health datasets with missing information.
• This approach can facilitate the identification of at-risk asthma populations for targeted interventions.
• Further validation could enhance the clinical utility of mobile health data in asthma management.