Abstract
OBJECTIVE: Primary cesarean delivery rates vary by race and ethnicity. We determined whether the inclusion of race and ethnicity substantially improved predictive ability in predictive models for primary cesarean delivery. STUDY DESIGN: Data from 371,468 women who were at risk for primary cesarean delivery were obtained from 2003 California birth certificates. A logistic regression model for primary cesarean delivery was built with maternal age, race and ethnicity, medical conditions, gestational age, multiple births, insurance, nulliparity, complications of pregnancy, and the trimester in which prenatal care began. The model's predictive validity was then compared with a model that excluded race and ethnicity. RESULTS: The C statistics (also called the area under the receiver operating characteristic curve for models) with (0.766) and without (0.764) race and ethnicity were similar and demonstrated that the addition of race and ethnicity did not substantially increase the predictive discrimination of the model. Additionally, there was no substantial difference in model calibration. CONCLUSION: Risk-adjustment models with and without race and ethnicity do not differ substantially in discrimination or calibration.