BACKGROUND: The Canadian Nosocomial Infection Surveillance Program (CNISP) has conducted surveillance for incident cases of methicillin-resistant Staphylococcus aureus (MRSA) in sentinel hospitals since 1995. In 2007, a reliability audit of the 2005 data was conducted. METHODS: In 2005, 5,652 cases were submitted to the CNISP from 43 hospitals. A proportional sample of submitted forms (up to 25) from each site were randomly selected. Stratified random sampling was used to obtain the comparison data. The original data were compared with the reabstracted data for congruence on 7 preselected variables. RESULTS: Reabstracted data were received from 30 out of 43 hospitals (70%), providing 443 of the 598 case forms requested (74%). Of these, 397 (90%) had matching case identification numbers. Overall, the percentage of discordant responses was 7.0%, ranging from 3.5% for sex and up to 23.7% for less well-defined variables (eg, where MRSA was acquired). CONCLUSION: Our findings suggest that, in general, the 2005 MRSA data are reliable. However to improve reliability a data quality framework with quality assurance practices, including ongoing auditing should be integrated into the CNISP's surveillance programs. Providing training to data collectors and standard definitions with practical examples may help to improve data quality, especially for those variables that require clinical judgment. Crown Copyright Â
BACKGROUND: The Canadian Nosocomial Infection Surveillance Program (CNISP) has conducted surveillance for incident cases of methicillin-resistant Staphylococcus aureus (MRSA) in sentinel hospitals since 1995. In 2007, a reliability audit of the 2005 data was conducted. METHODS: In 2005, 5,652 cases were submitted to the CNISP from 43 hospitals. A proportional sample of submitted forms (up to 25) from each site were randomly selected. Stratified random sampling was used to obtain the comparison data. The original data were compared with the reabstracted data for congruence on 7 preselected variables. RESULTS: Reabstracted data were received from 30 out of 43 hospitals (70%), providing 443 of the 598 case forms requested (74%). Of these, 397 (90%) had matching case identification numbers. Overall, the percentage of discordant responses was 7.0%, ranging from 3.5% for sex and up to 23.7% for less well-defined variables (eg, where MRSA was acquired). CONCLUSION: Our findings suggest that, in general, the 2005 MRSA data are reliable. However to improve reliability a data quality framework with quality assurance practices, including ongoing auditing should be integrated into the CNISP's surveillance programs. Providing training to data collectors and standard definitions with practical examples may help to improve data quality, especially for those variables that require clinical judgment. Crown Copyright Â