Toward Improved Data Quality in Public Health: Analysis of Anomaly Detection Tools applied to HIV/AIDS Data in Africa

Abstract

The study examined the data quality efficiency of the WHO Data Quality Review (DQR) toolkit and PyCaret anomaly detection algorithms. The tools were applied to the African HIV/AIDS data (2015-2021) extracted from a public data repository (data.pepfar.gov). The research outcome suggests that unsupervised anomaly detection algorithms could complement the efficiency of the WHO DQR toolkit and improve Data Quality Assessment (DQA). In particular, the study showed that anomaly detection algorithms through python programming provide a more straightforward and more reliable process for detecting data inconsistencies, incompleteness, and timeliness appears more accurate than the WHO tool. Consequently, the study contributed to ongoing debates on improving health data quality in low-income African countries.