Data Quality in Norwegian Surgical Site Infection Surveillance. Implementation and validation of a national system for surveillance of surgical site infections in Norway

Abstract

Surgical site infections (SSIs) constitute about one quarter of all healthcareassociated infections in Norway and are associated with a substantial cost for hospitals, patients and the community. Surveillance with feedback has been proven to be a useful tool in prevention of adverse events. In Norway, SSI surveillance was initiated through the establishment of the Norwegian Surveillance System for Antibiotic Consumption and Healthcare-Associated Infections (NOIS) by regulation in 2005. The overall aim of this thesis was to describe and validate NOIS-SSI in order to ensure good data quality through a uniform, robust and feasible system for monitoring and controlling SSIs. This was achieved through three sub-studies, each reported in a separate paper. The first study gives a description of the methodology of NOIS-SSI, and the value of a mandatory system with automated data collection and post-discharge surveillance. The second study is an investigation of the completeness, representativeness and accuracy of the denominator data by comparing NOIS-SSI to administrative data. The third study is an investigation of the necessity of post-discharge surveillance and the consequences of differing duration and intensity of follow-up of patients after hospital discharge. Paper I reports a steadily increasing participation in NOIS-SSI during 2005-2009, with many hospitals implementing computerized systems and submitting extra data voluntarily. We found reasonable quality of the risk-adjustment variables and a very good post-discharge follow-up, with 81% of SSIs being detected after hospital discharge. Paper II shows that the completeness of NOIS-SSI's denominator data improved greatly between 2005-2010 and that hospital participation became more representative over time. However, NOIS-SSI did not achieve representativeness for all surgical procedures. The hospitals which participated, submitted accurate denominator data and all the computerized systems delivered data of high quality. In paper III, we found that 82% of the deep SSIs after hip arthroplasty in 2005-2011 were detected after discharge from hospital. All of the patients with deep SSIs that were detected between 30 days and one year after hospital discharge were readmitted, which means that they could have been detected through the hospitals' computer systems instead of by patient questionnaire. A mandatory surveillance system should give more complete and unbiased data than a voluntary system. NOIS-SSI was not complete and representative and this was mostly due to a flexible and incremental implementation strategy. The upside of this flexibility was the establishment of electronically based surveillance systems in almost all hospitals. Electronically based systems have led to good quality of riskadjustment variables and excellent accuracy. Because most SSIs are detected after hospital discharge, active follow-up of the patient after discharge is important for correct case ascertainment. Active follow-up is resource demanding, but without it most SSIs would go undetected and the infection rates would be incorrect. Validation of correct classification of SSIs according to surveillance definitions has not been performed in Norway. Some studies have indicated good-to-excellent sensitivity and specificity in SSI validation by retrospective chart review, whereas others have found poor inter-rater agreement. Information technology in surveillance is in rapid development, and much has been published about computer algorithm-assisted case detection in existing clinical and administrative databases. Although there are many potential pitfalls in utilizing more automated surveillance, it is time and cost efficient and circumvents many of the issues associated with manual systems. All-year, all-procedure mandatory surveillance was implemented in Norway from 2012/2013. The findings of this thesis have led to changes in the Norwegian surveillance system, and one-year follow-of hip arthroplasty patients by letter will be replaced by surveillance through readmissions in 2015. By continually improving and upgrading the system we hope to achieve a surveillance system that is robust, efficient and useful. Acting on the basis of surveillance data in a clinical setting is the ultimate goal of a surveillance system. High quality data are essential in this process, and more use of automated case-finding may be an important asset in achieving this.