510. Good enough? Surgical Site Infection Surveillance compared to Administrative Data in Norway 2005-2010
Session: Poster Abstract Session: Surgical Site Infections
Thursday, October 3, 2013
Room: The Moscone Center: Poster Hall C
  • NOIS- Good enough.pdf (289.8 kB)
  • Background:

    This study compares SSI data in the Norwegian Surveillance System for Healthcare-Associated Infections (NOIS) to administrative data in the Norwegian Patient Register (NPR) on an aggregated level. This is achieved through (i) describing data quality by evaluating coverage, completeness and representativeness of denominator data and (ii) investigating the reason for any discrepancies between NOIS and NPR; two largely electronic systems for data collection.


    We compared de-identified data in NOIS with NPR in 2005-2010 on an aggregated level by breaking the data down by several variables, and comparing the distribution of the events within each grouping. We calculated coverage and completeness for each procedure and measured representativeness by comparing frequencies for key variables in NOIS with three groups of data in NPR. We investigated the discrepancies in the two registers by evaluating differences between regions, hospital size, hospital type, and IT-systems.


    NOIS had a total coverage of 59% of the procedures reported to NPR. Procedure specific coverage varied from 12% to 92% and completeness varied from 62% to 99% depending on priority of the procedure in the surveillance system. NOIS was representative by age group and sex. There was a shift in the representativeness by type of hospital from 2005-2007 to 2008-2010. There were no differences between the two registers that could explain the discrepancies in the denominator data.


    This study shows improvement in the coverage and completeness of NOIS from 2005 to 2010 compared to the "gold standard" NPR. Hospital type influences the representativeness more pronouncedly during the first three years of the surveillance system operation. Discrepancies between the two registers must be assumed to be due to human influence. Health care personnel evaluate each patient in the surveillance system. It is possible that they choose to include or exclude certain patients. Although NOIS is not person identifiable, thus limiting the data analysis to a more superficial comparison on an aggregated level, we see indications NOIS and NPR data are very comparable and of high quality. In order to validate NOIS in-depth, linked data are necessary.

    Hege Line Loewer, BS, BA, MoH, Infectious Disease Epidemiology, Norwegian Institute of Public Health, Oslo, Norway


    H. L. Loewer, None

    Findings in the abstracts are embargoed until 12:01 a.m. PST, Oct. 2nd with the exception of research findings presented at the IDWeek press conferences.