Identifying patients at risk for post-cesarean infection allows for implementing prevention strategies. Our study aim was to build and validate a statistical model to predict the risk of infection after cesarean section.METHODS:
Risk factors and outcomes within 30 days of surgery were collected on 2419 women who underwent cesarean delivery between 1/1/2013–12/31/2013. Two thirds of data (N=1611) containing 43 candidate predictors were randomly placed into a training set for model building and internal validation. Logistic regression models were fit to the data. Model discrimination was measured using the concordance index and accuracy using calibration curves. Internal validation was performed using bootstrapping correct for bias. The final model was externally validated on the remaining 1/3 of data (N=808).RESULTS:
Post-operative infection occurred in 8% (95% CI 7.3–9.5) of patients, including 5% meeting CDC criteria for SSI (95% CI 4.6–6.5). Cesarean section for failure to progress in labor (P<.0001), higher parity (P=.1611), earlier gestational age (P=.0004), longer operative time (P=.0598) and increasing BMI (P<.0001) were the strongest predictors for infection. The model was able to discriminate between women with and without infection on internal validation (concordance index=0.683, 95% CI 0.637–0.732) and external validation (concordance index=0.680) and calibration curves demonstrated predicted versus actual probabilities were accurate through a useful range (3–45%) of risk for infection.CONCLUSION:
Our model accurately predicts risk of infection after cesarean delivery. Identification of patients at high risk for postoperative infection allows for implementation of preventative measures.