Quantifying emergent drug-resistant TB using laboratory data and record-linkage methods

Clinicians, public health workers, and researchers must link multiple laboratory results to individual patient records to identify emerging drug-resistant TB.

Determine the proportion of individuals with rifampicin-resistant TB (RR-TB) who developed fluoroquinolone resistance using different data matching methods.

Data from the National Health Laboratory Service (NHLS) from the Western Cape Province, South Africa (January 2008 to June 2015), were used to identify patients with RR-TB who developed fluoroquinolone resistance after initial susceptibility. We used four methods to link patient results: A) exact match on NHLS patient number; B) exact match on surname, first name, and date of birth; C) custom algorithm (Stellenbosch University) using approximate matches for patient-identifying variables; and D) custom iterative approximate linkage developed for this study. We manually reviewed matches identified by each method.

We identified 30,381 laboratory results showing rifampicin and fluoroquinolone susceptibility. Estimates of resistance development were 4% using methods A and B, and 8% using methods C and D. Across methods, 688 unique patients who developed fluoroquinolone resistance were identified, 598 of which were correct by manual review. Method D identified 587 (98%) of these correctly matched patients, the largest proportion across these methods.

Customised approximate linkage approaches maximised correct patient matches.