Prediction models for perineal lacerations during childbirth: A systematic review and critical appraisal

Perineal lacerations could lead to substantial morbidities for women. A reliable prediction model for perineal lacerations has the potential to guide the prevention. Although several prediction models have been developed to estimate the risk of perineal lacerations, especially third- and fourth-degree perineal lacerations, the evidence about the model quality and clinical applicability is scarce. To systematically review and critically appraise the existing prediction models for perineal lacerations. Seven databases (PubMed, Embase, The Cochrane Library, Cumulative Index to Nursing and Allied Health Literature, SinoMed, China National Knowledge Infrastructure, and Wanfang Data) were systematically searched from inception to July 2022. Studies that developed prediction models for perineal lacerations or performed external validation of existing models were considered eligible to include in the systematic review. Two reviewers independently conducted data extraction according to the Checklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies. The risk of bias and the applicability of the included models were assessed with the Prediction Model Risk of Bias Assessment Tool. A narrative synthesis was performed to summarize the characteristics, risk of bias, and performance of existing models. Of 4345 retrieved studies, 14 studies with 22 prediction models for perineal lacerations were included. The included models mainly aimed to estimate the risk of third- and fourth-degree perineal lacerations. The top five predictors used were operative vaginal birth (72.7 %), parity/previous vaginal birth (63.6 %), race/ethnicity (59.1 %), maternal age (50.0 %), and episiotomy (40.1 %). Internal and external validation was performed in 12 (54.5 %) and seven (31.8 %) models, respectively. 13 studies (92.9 %) assessed model discrimination, with the c-index ranging from 0.636 to 0.830. Seven studies (50.0 %) evaluated the model calibration using the Hosmer–Lemeshow test, Brier score, or calibration curve. The results indicated that most of the models had fairly good calibration. All the included models were at higher risk of bias mainly due to unclear or inappropriate methods for handling missing data and continuous predictors, external validation, and model performance evaluation. Six models (27.3 %) showed low concerns about applicability. The existing models for perineal lacerations were poorly validated and evaluated, among which only two have the potential for clinical use: one for women undergoing vaginal birth after cesarean delivery, and the other one for all women undergoing vaginal birth. Future studies should focus on robust external validation of existing models and the development of novel models for second-degree perineal laceration. CRD42022349786. The existing models for perineal lacerations during childbirth need external validation and updating. Tools are needed for second-degree perineal laceration.