This study was a single-, case-control study conducted at Sunnybrook Holland Orthopaedic & Arthritic centre, the largest joint arthroplasty centre in Canada. The majority of the procedures performed at this facility were elective surgeries of hip or knee arthroplasty. This study was approved by the institutional review board (REB- CR6108), and patient consent was waived due to the retrospective nature of the study. Given the infrequent occurrence of surgeries after infected cases, a cases-control approach was utilized within the cohort of every patient in the hospital database between 5th January 2015 and 21st December 2018.

This study involved a detailed examination of OR scheduling logs during the four-year period from 2015 to 2018 to identify cases of scheduled elective joint arthroplasty, osteotomy, infected cases, and arthroscopic surgery. The initial step involved identifying surgical cases with interventions indicative of the presence of an infection, which included procedures such as debridement of infected surgical wound, incisional and drainage, arthrotomy with debridement, debridement with implant retention, first-stage debridement, and amputation due to unmanageable infection.

Subsequently, we identified and included into the study cohort patients undergoing elective arthroplasty or surgery of lower extremities classified as clean-wound procedures, namely corrective osteotomy, arthroscopic surgery, and patellofemoral reconstructive surgery, which were performed before and after the index infection-related procedure. Cases associated with history of infection including second-stage revision following initial stage debridement of PJI were excluded. Included patients were then divided into two distinct cohorts, the cases and the controls, based on their sequence in relation to the infected cases. Cases were any elective clean-wound surgeries done after an infected case in the same OR on the same day, to which we referred as the “post-infection sequence cohort”. Conversely, the controls were clean-wound procedures performed before the first case of infection in the same OR on the same day, a more typical scenario that we categorized as the “pre-infection sequence cohort”.

The OR scheduling logs and follow-up results were reviewed by three surgeons [PR, ST, BR]. Each case was independently assessed, and cross-verified among the team members to ensure the reliability and validity of the data. Patients’ baseline characteristics, particularly those influencing postoperative infections [3, 4] including cancer, diabetes mellitus, genitourinary disease, hepatobiliary disease, neurological diseases including Alzheimer’s disease and Parkinson’s disease, obesity, psychological disease including depression, active smoking, and peripheral vascular disease, were assessed via the hospital database.

The primary outcome was any superficial or deep infection occurring within one year after the index surgery. Superficial wound infection, as defined by the Surgical Infection Study Group (SISG), encompasses the presence of pain, edema, erythema, warmth, and impaired function around the surgical site [8]. Deep infection was defined as an infection that persists after surgery and necessitates any debridement or surgical intervention into the previous surgical area, in addition to antibiotic treatment. This study primarily focused on short-term one-year infection rates as the primary outcome because existing literature demonstrates that most infections related to microorganism from intraoperative contamination occur within this timeframe [9]. The secondary outcome was an infection occurring over the longest available follow-up time. Participants’ medical records were reviewed and documented in March-April 2024 by one of the three study surgeons for notes indicative of the presence of superficial or deep infection post-surgery. All instances of infection were then documented to assess the one-year postoperative infection rates. To ensure comprehensive follow-up, all participants were monitored for the most recent infections recorded in both our hospital database and the interconnected community hospital database. This approach includes verifying any infections treated elsewhere or managed with antibiotics, as reported in patients’ histories or noted from patients’ presentations at any hospital. Additionally, all email communications from patients were included in the virtual care notes within the hospital database system to ensure no details were missed.

Categorical data are described using frequency and percentage and compared using the Chi-squared test. Continuous variables are summarized and presented as mean with standard deviation or median with interquartile range (IQR) as appropriate. They were compared using the student t-test when normally distributed. We employed propensity score matching to mitigate confounding variables and create two balanced cohorts between the cases and the controls. Utilizing R version 4.3.3, a full propensity score matching was conducted via MatchIt package [10]. We calculated a propensity score for each participant by considering significant baseline characteristics based on an absolute standardized difference of 0.2 [11]. The characteristics under consideration were age, gender, and pre-existing conditions that can increase the risk of post-operative infection, including cancer, diabetes mellitus, genitourinary disease, hepatobiliary disease, neurological disease, body mass index (BMI) over 30 kg/m2, psychiatric disease, smoking, and vascular disease [4]. The technique included a full matching method on propensity scores with a caliper of 0.2, based on the logit of the propensity score, to balance the aforementioned characteristics between the two cohorts. We evaluated the balance improvement by determining standardized differences for each covariate pre- versus post-matching, ensuring that any significant disparities were rectified before proceeding to the main analysis. Following this, the cohorts underwent a survivorship analysis to determine the infection rates within one year of the procedure and at the most recent follow-up. The time-to-event data are visually represented as the Kaplan-Meier survival curves through the utilization of the Survival [12] and Survminer packages [13] in R version 4.3.3. Results are reported as the hazard ratio (HR) with 95%Confidence Interval (95%CI) and the p-value from the log-rank test. A p-value of < 0.05 was considered statistically significant. There was no missing data in the present study.