Data Cleaning Process for mHealth Log Data to Inform Health Worker Performance

Area of Science:

• Health Informatics
• Data Science
• Mobile Health

Background:

• Log data from mobile health (mHealth) applications offers insights into user engagement and health provider work patterns.
• Raw log data requires significant cleaning and curation for accurate analysis.

Purpose of the Study:

• To describe a systematic data cleaning process for mHealth-derived logs.
• To evaluate the impact of data cleaning on the usability of log data for analyzing health provider work patterns.

Main Methods:

• A systematic data cleaning process based on Broeck's framework (iterative screening, diagnosis, treatment) was applied.
• Log data from the mUzima mHealth application was used as a case study.
• Data quality issues such as inconsistencies, duplicates, and missing values were addressed.

Main Results:

• The data cleaning process identified numerous data quality issues in the raw logs.
• Only 39,229 out of 91,432 initial log records (42.9%) were retained after cleaning.
• Significant data reduction was necessary to ensure data suitability for analysis.

Conclusions:

• A systematic data cleaning approach is essential for deriving meaningful insights from mHealth log data.
• Effective data curation is vital for accurate analysis of health provider engagement and performance.
• The study underscores the challenges and importance of data quality in mHealth research.