Design: Prognostic model development with bootstrap and temporal validation
Population: 8,909 Swedish primiparous women with second- to fourth-degree tears, 2014–2019
Primary outcome: Anal incontinence (flatus and/or faecal) at 1 year postpartum
Effect: AUC 0.78 (95% CI 0.76 to 0.80); temporal validation AUC 0.78 (0.75 to 0.81)

Swedish investigators linked three national registers to build a model for anal incontinence one year after a second-, third- or fourth-degree tear. They used 8,909 first-time mothers delivering vaginally at 34 weeks or later between 2014 and 2019, and tested the model on a separate 2020–2022 cohort.

Of 17 candidate predictors, seven stayed in the final model, with a high Wexner incontinence score at the eight-week check the strongest. Discrimination was good: AUC 0.78 (95% CI 0.76 to 0.80) for any anal incontinence, 0.80 for faecal and 0.78 for flatus incontinence, and 0.78 again on temporal validation. Calibration was best in women with third-degree tears.

The value is in triage. Symptoms reported at eight weeks already carry most of the information about who will still be affected a year later, which argues for asking structured questions at that visit rather than waiting for women to raise it.

The model has been validated only within Sweden and over time, not in another country or health system.

Use a structured incontinence score such as Wexner at the eight-week postnatal perineal review rather than an open question.
Women with a high score at eight weeks are the group most likely to still have symptoms at one year.
Consider early referral to pelvic floor physiotherapy or a perineal clinic for women with persistent symptoms at that visit.
Second-degree tears are included in the model, so do not limit follow-up questions to obstetric anal sphincter injury.

Why it matters

Symptoms at eight weeks are a strong signal of long-term outcome, so that visit is where most of the risk can be spotted.

The statistics, in plain English

An AUC of 0.78 means that if you pick one woman who had incontinence at a year and one who did not, the model ranks the affected woman higher about 78% of the time. Holding at 0.78 in a later cohort suggests it was not just fitted to the first dataset.