199: Validation of a prediction model of successful trial of vaginal birth after cesarean (VBAC)


      Clinical prediction models are only useful if their ability to predict outcomes is consistent across populations. Thus, we sought to validate the predictive ability of a published prediction model for VBAC success in a different population

      Study Design

      Within a multicenter, retrospective cohort of 25,005 women with at least one prior cesarean, we sought to validate the model for prediction of VBAC success constructed from data from the MFMU Network study (Grobman et al). Using the coefficients of the logistic regression solution from the published model, a unique predictor for VBAC success was created for all women in our cohort who attempted VBAC based on maternal age, race, prior vaginal delivery and timing, and recurrent cesarean indication. Logistic regression and receiver-operator characteristic analysis was used to estimate the accuracy of prediction. The error estimates for the coefficients were used to calculate the minimum and maximum estimates of predictive accuracy for VBAC success


      In the cohort of 13,706 patients attempting VBAC, the model-derived predictor was significantly associated with VBAC success (AUC=0.84, 0.83-0.85 [min-max], p<0.01). The model was most accurate with a sensitivity of 93.6% and a specificity of 58.0%, correctly classifying 85% of patients in the cohort with respect to VBAC success (N=9,763/10,334)
      Figure thumbnail gr1


      The validity of the recently published predictive model for VBAC success was supported in this large, multicenter cohort