Record linkage without patient identifiers: Proof of concept using data from South Africa's national HIV program

Abstract

Linkage between health databases typically requires patient identifiers such as names and personal identification numbers. We developed and validated a record linkage strategy to combine administrative health databases without identifiers for South Africa's public sector HIV program. We linked CD4 counts and HIV viral loads from South Africa's TIER.Net with the National Health Laboratory Service (NHLS) database for patients receiving care between 2015-2019 in Ekurhuleni District (Gauteng Province). Linkage variables were result value, specimen collection date, facility of collection, year and month of birth, and sex. We used three matching strategies: exact matching on exact values of all variables, caliper matching allowing a ± 5 day window on result date, and specimen barcode matching using unique specimen identifiers. A sequential linkage approach applied specimen barcode, followed by exact, and then caliper matching. Exact and caliper matching were validated using barcodes (available for 34% of records in TIER.Net) as a "gold standard". Performance measures were sensitivity, positive predictive value (PPV), share of patients linked, and percent increase in data points. We attempted to link 2,017,290 laboratory test results from TIER.Net (523,558 unique patients) with 2,414,059 NHLS test results. Exact matching achieved 69.0% sensitivity and 95.1% PPV. Caliper matching achieved 75% sensitivity and 94.5% PPV. Sequential linkage matched 41.9% using specimen barcodes, 51.3% through exact matching, and 6.8% through caliper matching, for 71.9% (95% CI: 71.9, 72.0) of test results matched overall, with 96.8% (95% CI: 96.7, 97.1) PPV and 85.9% (95% CI: 85.7, 85.9) sensitivity. This linked 86.0% (95% CI: 85.9, 86.1) of TIER.Net patients to the NHLS (N = 1,450,087), increasing laboratory results in TIER.Net by 62.6%. Linkage of TIER.Net and NHLS without patient identifiers attained high accuracy and yield without compromising privacy. The integrated cohort provides a more complete laboratory test history and supports more accurate HIV program indicator estimates.

Fig 1. Flow diagram: laboratory results from TIER.Net and NHLS submitted for linkage.

Fig 2. Record linkage performance for each linkage strategy in the full dataset. Note: The figure compares the estimated performance of each linkage strategy with respect to sensitivity, positive predictive value (PPV), lab-level linkage yield, and patient-level linkage yield. Estimates were based on extrapolation from the barcode subsample, under the assumption that barcodes were missing completely at random (S1 Text). Overall, the sequential linkage approach outperformed the other approaches with the highest linkage yield at the test (71.9%) and patient level (86.0%), high sensitivity (85.9%), and PPV (96.8%).

Fig 3. Enrichment of test results in TIER.Net through linkage with NHLS National HIV Cohort (test data generated between January 2015-March 2018).