Quality assessment of data for decentralised antiretroviral therapy referrals and laboratory results in the South African national electronic HIV management register TIER.Net

Three Interlinked Electronic Register (TIER.Net) is South Africas national electronic HIV patient database, used to monitor antiretroviral therapy (ART) delivery and laboratory results. However, few published evaluations have quantified TIER.Net data quality relative to national sources. We aimed to evaluate how well decentralised ART referral and laboratory result data are captured in TIER.Net. We conducted a retrospective analysis comparing TIER.Net to national electronic health systems. For decentralised ART, we used de-identified data from 56 clinics in eThekwini (2020-2023) and compared the annual number of TIER.Net decentralised ART referrals to ART prescriptions in the Synchronised National Communication in Health (SyNCH) database. For laboratory data, we used de-identified records from 103 clinics in KwaZulu-Natal (2015-2022) and compared the annual number of TIER.Net viral load (VL) and CD4 tests with the number in National Health Laboratory Service (NHLS). The proportion of SyNCH decentralised ART prescriptions and NHLS VL and CD4 counts captured in TIER.Net were calculated by clinic, and trends were assessed using linear mixed-effects models (LMMs). The median proportion of SyNCH decentralised ART prescriptions captured in TIER.Net was 104.4% (IQR: 99.85-115.1%) in 2020 and 102.44% (IQR: 100.5-104.5%) in 2023. The LMM estimated an annual decrease of 2.8% (95% CI: -4.8; -0.9%). The median proportion of NHLS VLs captured in TIER.Net was 85.7% (IQR: 70.0-97.9%) in 2015 and 99.1% (IQR: 94.5-102.5%) in 2022. The LMM estimated an annual increase of 1.8% (95% CI: 1.4-2.1%). The median proportion of NHLS CD4s captured in TIER.Net was 74.3% (IQR: 63.9-85.4%) in 2015 and 80.1% (IQR: 68.4-89.1%) in 2022. The LMM estimated no statistically significant trend over time (-0.08%, 95% CI: -0.6; 0.4%). Reassuringly, capture of TIER.Net for decentralised ART and VL data has improved to near 100%, but CD4 count capture remains sub-optimal, highlighting strengths and limitations of conducting analyses with this critical HIV programme database.