Objectives: Valid deduplication of human immunodeficiency virus (HIV) case reports is critical to the utility of these data to inform HIV programs. The Haitian Ministry of Health (MSPP) and partners operate a case-based, national HIV/AIDS surveillance system (HASS), using deterministic and probabilistic procedures to identify duplicate records. These procedures are described and validated based on expert classifications.
Methods: Two samples of HASS records identified as duplicates were selected: 100 pairs from deterministic and 100 pairs from probabilistic matching procedures (total: 200 pairs, 400 case reports). Clinical data from the national electronic medical record (iSanté) were reviewed and consensus gold-standard determinations on the status of duplications were made. False positive rates (FPR) were estimated by reviewing these records, while false negative rates were calculated (FNR) by using LinkPlus™ probabilistic linkage software. The effect of deduplication on total HIV case counts was demonstrated.
Results: Review of deterministic matches yielded 99 true positives and 1 false positive (FPR, 1 per 100; 95% CI, 0.71-5.4). Review of probabilistic matches yielded a FPR of 6 per 100 (95% CI, 2.7-12.4). LinkPlus identified 1,491 probable matches among 68,393 records, representing a FNR of 2 per 100 (95% CI, 0.55-7.0). After adjustment, the estimated unique count of reported HIV patients in HASS was 211,885 (95% CI, 207, 293-213, 232) as of December 2013.
Conclusions: Based on application of the established procedures, HASS conforms to the duplication performance standard recommended by the Centers for Disease Control and Prevention for HIV surveillance.