Semi-automated detection of surgical-site infections using a simple and effective hospital data-based algorithm in the national surveillance system in France

BACKGROUND: A semi-automated surveillance system for surgical site infections (SSIs), SPICMI (Surveillance and Prevention Program for Infectious Risk in Surgery and Interventional Medicine), has been implemented in French hospitals, leveraging data from electronic health records (EHRs). OBJECTIVE: To evaluate the performance of the SPICMI algorithm in detecting SSIs in orthopedic and digestive surgery. SETTING: Surveillance data were collected annually from the EHRs. The algorithm identified suspected SSIs based on two criteria: (1) surgical revision during the index stay or readmission, (2) positive microbiological samples from the wound. Suspected SSIs identified were subsequently validated by surgeons. METHODS: A stochastic modeling approach was used to estimate probability intervals for performance indicators. Various detection scenarios were constructed based on SPICMI criteria. Logistic regression analysis was performed using surveillance data. Data unavailable in the database were estimated through a literature review and expert opinions. RESULTS: The probability of surgical revision following an SSI varied significantly between surgical specialties, ranging from 92% in orthopedic surgery to 45.2% in gynecology. In orthopedic and digestive surgery, the SPICMI algorithm demonstrated good reliability for detecting SSIs in minimizing false-negative and false-positive cases (Youden index: 0.96 and 0.79, respectively). Sensitivity (Se) was lower in digestive surgery (0.7-0.9) compared to orthopedic surgery (0.9-1), while specificity (Sp) remained high (0.9-1) in both specialties. CONCLUSION: The SPICMI algorithm shows potential to support efficient use of time and resources in SSIs surveillance management. Further evaluation is needed with a broader panel of surgery procedures.