Estimating the sensitivity and specificity of matching name-based with non-name-based case registries

Abstract

Because non-name-based case registries have recently been used for reporting human immunodeficiency virus infection, this study attempted to define the sensitivity, specificity and accuracy of case registry matches using non-name-based registries. The AIDS, sexually transmitted disease (STD), and tuberculosis (TB) case registries were matched using all available information to establish the standard. The registries were then matched again using five increasingly less specific criteria to compare sensitivity, specificity and accuracy. The registries were then also transformed into non-name-based codes as if they were the HIV registry and matched again. With name-based registries, sensitivities increased as the matching criteria became less exacting, while the accuracy declined slightly. Specificities remained close to 100% due to the relatively small number of matched cases. Results from matches of non-name-based registry matches were similar to those of the name-based registry matches. Non-name reporting can be used for data matching with acceptable accuracy.

Practical Considerations for Matching STD and HIV Surveillance Data with Data from other Sources

“…Considerations regarding selection of match criteria and how such selection can affect the sensitivity, specificity, and positive predictive value of the match have been published. 1,3,6,9 Aside from the obvious requirement that potential matching elements must be present in both datasets, a number of considerations may affect the discriminating power of specific data elements. If either dataset has a significant proportion of records with missing data or unknown values for a given element, this will reduce the discriminating power of that item in any type of match conducted.…”

CDC and Progress toward Integration of HIV, STD, and Viral Hepatitis Prevention

“…CDC support for program integration began in earnest with the recognition of STDs as co-factors for HIV transmission 6 and the subsequent formation in 1995 of a National Center for HIV, STD, and TB Prevention. That organizational change promoted collaboration across the realigned programs at CDC, resulting in the release of STD treatment guidelines with a new emphasis on HIV infection and hepatitis B vaccination, 7 the creation of STD/HIV prevention training centers (http://depts.washington.edu/nnptc/) augmented by a national prevention information network (http://www.cdcnpin.org), and support of special projects to match surveillance registries 8 and promote hepatitis B vaccination and hepatitis virus (HCV) counseling and testing. 9,10 HOW CDC CAN REMOVE BARRIERS IDENTIFIED BY STATE AND LOCAL PROGRAMS State and local health departments seeking to improve preventive care for their clients welcomed the federal support 11 and encouraged CDC to "demonstrate concerted and focused leadership in the development of integrated service delivery" 12 and "remove the lack of integration as one of the barriers to more effective programs.…”

The Effect of Case Rate and Coinfection Rate on the Positive Predictive Value of a Registry Data-Matching Algorithm

“…[3][4][5][6][7] All data-matching projects use either a deterministic or probabilistic method, or a combination of the two, to identify matches. Some published analyses have reported the sensitivity, specificity, and positive predictive value (PPV) of their method; 5,8 however, as of this writing, none has addressed the effect of the case rate and coinfection rate in the population on the PPV of a registry data-matching algorithm. One statistical modeling study reported that the prevalence of false matches declined as the events became rarer (i.e., lower case rate) or the number of matches increased (i.e., higher coinfection rate).…”