What can an audit of national surveillance data tell us? Findings from an audit of Canadian vancomycin-resistant enterococci surveillance data.

Background The Canadian Nosocomial Infection Surveillance Program (CNISP) has conducted surveillance for vancomycin-resistant enterococci (VRE) in sentinel hospitals since 1999. In 2010, a reliability audit of the 2008 data was conducted. Methods Stratified random sampling was used to obtain a proportional sample of VRE case forms submitted in 2008 from 36 CNISP hospitals. The original VRE data were compared to re-abstracted data for congruence on 16 pre-selected variables. Any discrepancy between the original and re-abstracted data was identified as a discordant response. Results Re-abstracted data were received from 35 out of 36 hospitals, providing 98% (n=428) of the 437 case forms requested. Of these, 37% (n=157) had zero discordant responses, 29% (n=126) had one discordant response, 16% (n=70) had two discordant responses and two forms (0.5%) had eleven discordant responses. Among the 35 hospitals, one hospital (3%) submitted forms with no discordant responses. Overall, the percentage of discordant and missing responses was 5%, ranging from 1% (n=29) for type of infection to 22% (n=93) for previous hospitalization. Conclusions Overall case forms were complete. However, discordant responses were more likely for variables that require interpretation and judgment or using historical data. Clearly defining variables and providing applicable response options may improve data quality. As well, the implementation of electronic health records may also improve the reporting of historical data. [ABSTRACT FROM AUTHOR]