Predicting obstetric anal sphincter injury in the first and second vaginal delivery and after a cesarean delivery: development and validation of an intrapartal model

Objectives

To develop and validate prediction models for obstetric anal sphincter injury (OASI) in three birth scenarios (1st vaginal delivery, vaginal birth after cesarean section and 2nd vaginal delivery). Antenatal and intrapartal predictors were included in the models to construct a web-based, interactive, easy-to-use calculator.

Study design and setting

All 45 maternity units in Sweden participated in the study, with 609,916 first and second deliveries in gestational week ≥37+0 with singleton pregnancies and cephalic presentations between 2009 and 2017. The outcome was OASI, defined as a third- or fourth-degree perineal injury involving the external or internal anal sphincter muscles, or both. We analyzed the relative contribution of predictors with Nagelkerke´s R² (R²_N) after minimization of the Bayesian Information Criterion for the predictor selection in a logistic regression with OASI as the binary outcome. Model performance was evaluated according to overall measures, discriminative ability, and calibration, with optimism-correction by a bootstrap procedure.

Results

OASI occurred in 25,245 women (4.1%). There were 54 relevant, possible predictors, and 47 predictors were kept as candidates for the final models. We included 28, 40, and 46 predictors for the 3 scenarios, respectively. Infant birth weight was identified as the primary predictor, contributing 31% to 45% of the R²_N in the full prediction models, which had R²_N values of 9.3%, 7.4%, and 12.9% for each scenario. In 2-para women, obstetric information from the first birth was important at the 2nd birth, accounting for 50% of the total predictive information. A sphincter injury in the 1st vaginal delivery strongly predicted a repeat injury (R²_N, 40%). Vacuum delivery in the first and second vaginal delivery contributed 33 and 29% of R²_N, respectively. By incorporating information on fetal biometry and labor events, the performance of the models increased substantially (e.g. R²_N increased from 1.7 to 9.3% in the first scenario). An online calculator was developed (www.sphinctercalc.com).

Conclusions

Fetal birth weight is the crucial predictor of sphincter injury, supporting efforts to assess fetal biometrics through imaging techniques. This also applies to obstetric interventions, particularly vacuum delivery, provided that there are alternatives to act upon. The proposed online calculator requires further international validation and refinement before it can be widely used clinically and for women's shared decision-making.