Predicting Post-Cesarean Surgical Site Infections: A Secondary Analysis of the Consortium on Safe Labor [27P]. (May 2020)

• Record Type: Journal Article
• Authors: Liao, Caiyun; Aziz, Michael; Chiu, Chi-Yang; Mari, Giancarlo; Tate, Danielle L.
• Is Part Of: Obstetrics and gynecology. Volume 135(2020)Supplement 1
• Publication Date: 2020-05
• Subjects: Obstetrics -- Periodicals; Gynecology -- Periodicals; 618
• DOI: 10.1097/01.AOG.0000663868.86684.a9
• Languages: English
• ISSNs: 0029-7844

Abstract

Abstract : INTRODUCTION: Surgical site infections (SSI) have been estimated to complicate 2–18% of cesarean deliveries (CD). We sought to establish prediction models for post-CD SSIs, using the Consortium on Safe Labor (CSL), a study involving 12 institutes in all the nine districts in the American College of Obstetricians and Gynecologists (ACOG). METHODS: All the CDs in CSL were included. The primary outcomes were endometritis and wound infections. Covariates were selected from clinical and demographic variables that might alter post-CD SSI risks, using backward stepwise selection. Multi-level modeling was used to account for data arising from the same patients. Model discrimination was assessed by area under a receiver-operator characteristic curve (C-statistics), while model calibration was assessed by comparing the predicted vs empirical probabilities. The study was approved by the Institutional Review Board. RESULTS: In CSL, 208, 695 women delivered 228, 438 pregnancies between July 2002 and July 2008. There were 65, 979 CDs (28.9%), all of which were included in this study. The overall incidences of post-CD endometritis (n=490) and wound infections (n=547) were 0.74% and 0.83%, respectively. The C-statistics for the endometritis and wound infection models were 0.89 and 0.80, respectively, demonstrating good discrimination capacity. The endometritis model demonstrated good calibration, mainly when the estimated risks were low (<2%). The wound infection model was less accurate. CONCLUSION: Methodological challenges remain in predicting relatively rare yet morbid events, such as post-CD SSIs. However, statistical modeling based on national data, such as CSL, and the Birth Registry led by ACOG, may provide opportunities to better predict and prevent post-CD SSIs.