Evaluation of a new prediction model for the estimation of risk of obstetric anal sphincter injuries.

Abstract

BACKGROUND Obstetric anal sphincter injuries are complications to vaginal birth with potential to cause substantial maternal morbidity. Prediction of these injuries might help to improve maternal care but also antenatal counselling and patient information. Previous attempts to create prediction models have in many cases involved variables only known post-partum which limit their use in an antenatal setting. Other models include parameters that are not applicable to a Northern European population. OBJECTIVE To develop and validate a clinically useful model for prediction of risk of obstetric anal sphincter injuries. STUDY DESIGN The model was developed using a retrospective nationwide cohort from the Swedish Medical Birth Registry consisting of 1,209,421 deliveries between 2005-2016. After exclusion criteria (caesarean section, forceps delivery, missing data) were applied, the data set was randomly divided into a development data (n=422,011) set and a validation data set (n=422,010). Using the development data set all variables were assessed with univariable analysis with modified Poisson regression analysis. A prediction model was then built by multivariate analysis using modified Poisson regression where vaiables with p= <0.2 were included. Both forwards and backwards selection were used and variables with p= >0.05 were excluded. Validation was performed by evaluating the agreement of the predicted and true observed rate of obstetric anal sphincter injuries after the prediction model was applied to the validation data set. RESULTS Primiparity, previous cesarean delivery, previous sphincter injury, increasing age, increasing birth weight and maternal origin from Sub Saharan Africa or South/ Southeast Asia were all antenatal variables associated with increased risk of obstetric anal sphincter injury. Smoking, increasing maternal height and BMI appeared to lower the risk. Vacuum extraction also increased the risk of sphincter injury. We developed one model including previously mentioned antenatal parametres and one also including vacuum extraction. The final prediction model including instrumental delivery can be used for predicting the risk of sphincter injury for delivery with or without vacuum extraction with higher accuracy. This model had good discrimination with AUC 0.79 (95% CI 0.78-0.79) and was able to predict risks up to 24% with good to moderate accuracy. CONCLUSION Using antenatally available data, obstetric anal sphincter injuries can be predicted with moderate certainty. This prediction model has been externally validated and can be used for individualised antenatal counselling as well as identifying persons at high risk where preventative strategies might improve outcomes. Further validation in other populations outside of Scandinavia is recommended before clinical implementation.