Data cleaning process for HIV-indicator data extracted from DHIS2 national reporting system: a case study of Kenya

Area of Science:

• Public Health
• Health Informatics
• Data Science

Background:

• The District Health Information Software-2 (DHIS2) is crucial for national health reporting.
• Ensuring high-quality DHIS2 data is essential for effective decision-making.
• Systematic data cleaning is vital for preparing DHIS2 data for analysis.

Purpose of the Study:

• To describe a systematic and replicable data cleaning method for HIV data in DHIS2.
• To report on the methods and outcomes of cleaning Kenyan HIV data from 2011-2018.
• To enhance DHIS2 data quality for secondary analyses.

Main Methods:

• Extracted six HIV indicator reports from DHIS2 (2011-2018) for Kenyan facilities.
• Employed a semi-automatic, multi-phase data cleaning framework (screening, diagnosis, treatment).
• Applied a generic five-step cleaning sequence to 93,179 facility-records.

Main Results:

• Removed 50.23% of records due to missing data.
• Identified and addressed various data quality issues, with significant variations across programmatic areas (p < 0.001).
• A clean dataset suitable for secondary analysis was successfully generated.

Conclusions:

• Transparent reporting of data cleaning is crucial for research validity and data use.
• Semi-automatic procedures significantly improved DHIS2 data quality beyond automated methods alone.
• The cleaned dataset enables reliable secondary analyses of HIV trends.