Matching AIDS and tuberculosis registry data to identify AIDS/tuberculosis comorbidity cases in California

Abstract

The purpose of this study was to evaluate the sensitivity and positive predictive value (PPV) of a registry data linkage procedure used in the California AIDS and Tuberculosis (TB) Registry Data Linkage Study to identify AIDS/TB comorbidity cases in California. The California AIDS registry data from 1981 to 2006 were linked to the California TB registry data from 1996 to 2006 using LinkPlus, a probabilistic record linkage program developed by the Centers for Disease Control and Prevention, and matched results were manually reviewed to determine true or false matches. We estimated the sensitivity of this procedure to range from 98.0 per cent (95% confidence interval, CI: 97.3%, 98.7%) to 98.8 per cent (95% CI: 98.1%, 99.2%), and the PPV to be 100 per cent (95% CI: 96.8%, 100.0%). Our study demonstrated the feasibility of using this linkage procedure to match AIDS and TB registry data with a very high degree of accuracy.

A New Method for Assessing How Sensitivity and Specificity of Linkage Studies Affects Estimation

“…In addition, the size and representativeness of the study sample may be increased [11]. However, record linkage studies are constrained by the quality of the datasets being linked and by the methods of linkage used [11]–[13]. Particularly, linkage can be more complex if the amount or quality of identifying data for individuals are limited.…”

Rifampin monoresistant tuberculosis and HIV comorbidity in California, 1993–2008

“…Manual review of all matched cases was performed, with only those matches above a pre-determined priority threshold considered to represent a co-infected case. [14]…”

Transnational Record Linkage for Tuberculosis Surveillance and Program Evaluation

“…We used established linkage nomenclature to define key terms. [17][18][19][20][21] A comparison pair was considered to be any possible comparison of a pre-immigration record with a record from the California TB Case Registry. A link was defined as a comparison pair accepted as having a high likelihood of belonging to the same person after manual review.…”

“…16 Because no common identifi er was present in both data sources, records were linked on the likelihood of having the same or similar personal identifi ers. Consistent with established techniques, [17][18][19] we used New York State Identifi cation and Intelligence System (known as NYSIIS) phonetically recoded fi rst and last names and dates of birth for blocking or data partitioning and matched on fi rst and last names, date of birth, overseas examination year, and sex. We calculated M-probabilities using the default direct method.…”