Using Machine Learning to Optimize the Quality of Survey Data: Protocol for a Use Case in India

Area of Science:

• Public Health
• Data Science
• Survey Methodology

Background:

• Data quality is crucial for reliable survey findings, traditionally managed through quality assurance.
• Data analytics, especially with electronic tools, is increasingly vital for real-time quality monitoring.
• Machine learning offers potential for advanced real-time data quality assessment.

Purpose of the Study:

• To describe methods for assessing household survey data quality.
• To integrate traditional and machine learning analytics for enhanced quality assurance.
• To improve data quality in a maternal and child health survey in India.

Main Methods:

• Utilized traditional and machine learning (ML) based quality assurance procedures.
• Implemented a seven-component strategy including tool development, training, monitoring, and feedback loops.
• Applied ML algorithms for outlier analysis, creating features from timestamps, "don't know" rates, and skip rates; employed supervised and unsupervised learning.

Main Results:

• Data collection occurred from October 2019 to March 2020.
• Quality assurance results were anticipated by August 2020.
• The study focused on improving data quality for the Kilkari impact evaluation.

Conclusions:

• Machine learning is underutilized for improving survey data quality in low-resource environments.
• The proposed approach is expected to enhance the Kilkari survey's data quality and impact evaluation robustness.
• The quality assurance approach has broader implications for health data collection applications.