Surgical site infection surveillance in knee and hip arthroplasty: optimizing an algorithm to detect high-risk patients based on electronic health records.

Abstract

Background: Surgical site infection (SSI) is an important cause of disease burden and healthcare costs. Fully manual surveillance is time-consuming and prone to subjectivity and inter-individual variability, which can be partly overcome by semi-automated surveillance. Algorithms used in orthopaedic SSI semi-automated surveillance have reported high sensitivity and important workload reduction. This study aimed to design and validate different algorithms to identify patients at high risk of SSI after hip or knee arthroplasty.

Methods: Retrospective data from manual SSI surveillance between May 2015 and December 2017 were used as gold standard for validation. Knee and hip arthroplasty were included, patients were followed up for 90 days and European Centre for Disease Prevention and Control SSI classification was applied. Electronic health records data was used to generate different algorithms, considering combinations of the following variables: ≥1 positive culture, ≥ 3 microbiological requests, antimicrobial therapy ≥ 7 days, length of hospital stay ≥ 14 days, orthopaedics readmission, orthopaedics surgery and emergency department attendance. Sensitivity, specificity, negative and predictive value, and workload reduction were calculated.

Results: In total 1631 surgical procedures were included, of which 67.5% (n = 1101) in women; patients' median age was 69 years (IQR 62 to 77) and median Charlson index 2 (IQR 1 to 3). Most surgeries were elective (92.5%; n = 1508) and half were hip arthroplasty (52.8%; n = 861). SSI incidence was 3.8% (n = 62), of which 64.5% were deep or organ/space infections. Positive culture was the single variable with highest sensitivity (64.5%), followed by orthopaedic reintervention (59.7%). Twenty-four algorithms presented 90.3% sensitivity for all SSI types and 100% for deep and organ/space SSI. Workload reduction ranged from 59.7 to 67.7%. The algorithm including ≥ 3 microbiological requests, length of hospital stay ≥ 14 days and emergency department attendance, was one of the best options in terms of sensitivity, workload reduction and feasibility for implementation.

Conclusions: Different algorithms with high sensitivity to detect all types of SSI can be used in real life, tailored to clinical practice and data availability. Emergency department attendance can be an important variable to identify superficial SSI in semi-automated surveillance.