External Validation of Second-Stage Vaginal Delivery Prediction Model [ID: 1377356]

INTRODUCTION: We sought to externally validate Gimovsky et al.’s prediction model for spontaneous vaginal delivery (SVD) in the second stage (https://www.pushprescriber.com [PUSH MODEL]) prior to using it for a quality improvement (QI) intervention at our hospital. METHODS: This was a retrospective cohort of 2,296 birthing persons delivering at a single hospital over 1 year. Inclusion criteria were similar to the PUSH MODEL: nulliparous, term, without planned cesarean, singleton, vertex, nonanomalous, epidural anesthesia. We calculated the predicted probability of SVD at every hour from the PUSH MODEL to generate receiver–operator curves and analyzed factors associated with second stage duration in our cohort. RESULTS: Overall, 1,735 (77%) delivered by SVD, similar to 74% in the PUSH MODEL. Head position had the strongest association with SVD (83% OA versus 39% OP, P <.01). The area under the curve from the PUSH MODEL at the start of the second stage was 0.64 (95% CI 0.62–0.77), as compared to 0.73 in the original population. The model had similar performance at hours 1 (n=1,513), 2 (n=917), 3 (n=522), and 4 or more (n=255). Using a cutoff of 77% (baseline rate of SVD in our population), those who “Screen Positive” at the start of the second stage had an 78.5% rate of SVD as compared to 39.3% for those who “Screen Negative” ( P <.01). CONCLUSION: The PUSH MODEL performed well on external validation; we similarly found that head position and second-stage duration were strongly associated with the probability of SVD in our cohort, supporting incorporation of this tool in QI work at our hospital.