Estimates of Surgical Site Infections through a Population-Based Linked Administrative Data System in Alberta, Canada

Background: Surgical site infection (SSI) surveillance has been collected through in-patient chart review and includes data only on infections developing before discharge or during readmission to acute care facilities in Alberta, Canada. Shortage of professionals and short post-operative stays represent an extra challenge to determine the ‘actual’ rate of SSIs according to the CDC definition (30 days after surgery - one year with prostheses).
Objective: 1-To obtain SSI rate estimates including infections developing post-discharge.  2- To build a population based data system to identify SSI by linking surgery hospital records, physician billings, outpatient and emergency department visits that can be used as a tool to develop concurrent province wide electronic SSI surveillance system.

Methods: 1.Study period: April 1^st, 2002 to September 30^th, 2007; 2.Selection of six major elective surgeries to be studied. Re-interventions were not considered;  
3.Identification of procedure codes (ICD10-CCI) related to the selected surgeries; 4.Identification of disease codes (ICD10 and ICD9) related to surgical site infections; 5.Extraction of all inpatient surgical stay records from the provincial health records databank; 6.Extraction of all health care encounters related to SSI for each inpatient surgical record: all in hospital and outpatient records and all physician claims were scanned to a preset amount of days (30, 60, 90, 120, 180 and 360) post surgery. Results: Searching for SSI developing up to 30 days post-surgery in all health care encounters by type of surgery resulted in infection rate estimates 1.7 to 5.2 times higher than those calculated using only information from hospital admissions and readmissions.
 Table 1 - SSI rate estimates up to 30 days post-surgery using different search opportunities

Surgery	SSI Rates (no of infections/no of surgeries X 100)
	During Admission or readmission	Identified at all encounters
CABG	6.4	11.5
C-Section	1.7	8.9
Hip replacement	2.6	4.4
Hysterectomy	2.5	8.8
Knee replacement	3.7	7.1
Spine procedures	1.7	3.5
Cardiac Valve	4.2	7.1

Up to 85% of all infections/year were detected within 30 days of surgery, following the distribution expected when using active surveillance with CDC definition criteria. Furthermore, by using administrative data estimates, a 60 day cut off raised the detection of SSI to up to 95% without incurring extra costs or effort.

Conclusions: Electronic medical records data may be used to estimate SSI closer to reality. It will help surveillance and to raise awareness to a problem that has been underestimated by short post-operative follow ups. Although personal communication from surgeons indicates similar SSI rates between active surveillance and administrative data, a limitation to this study is the absence of published SSI data from active surveillance to compare with our findings.