Abstract
OBJECTIVE: To assess the accuracy of the South African National Health Laboratory Services (NHLS) corporate data warehouse (CDW) using a novel data cross-matching method. METHODS: Adults (≥18 years) on antiretroviral therapy (ART) who visited a hospital-based HIV clinic in Durban from March to June 2012 were included. We matched patient identifiers, CD4 and viral load (VL) records from the HIV clinic's electronic record with the NHLS CDW according to a set of matching criteria for patient identifiers, test values and test dates. We calculated the matching rates for patient identifiers, CD4 and VL records, and an overall matching rate. RESULTS: NHLS returned records for 3498 (89.6%) of the 3906 individuals requested. Using our computer algorithm, we confidently matched 3278 patients (83.9% of the total request). Considering less than confident matches as well, and then manually reviewing questionable matches using only patient identifiers, only nine (0.3% of records returned by NHLS) of the suggested matches were judged incorrect. CONCLUSIONS: We developed a data cross-matching method to evaluate national laboratory data and were able to match almost 9 of 10 patients with data we expected to find in the NHLS CDW. We found few questionable matches, suggesting that manual review of records returned was not essential. As the number of patients initiating ART in South Africa grows, maintaining a comprehensive and accurate national data repository is of critical importance, since it may serve as a valuable tool to evaluate the effectiveness of the country's HIV care system. This study helps validate the use of NHLS CDW data in future research on South Africa's HIV care system and may inform analyses in similar settings with national laboratory systems.