Clinical Report: Machine Learning Prediction of Postpartum Urinary Incontinence

Overview

This prospective study describes development and external validation of a postpartum urinary incontinence (UI) prediction model combining large language model (LLM)-assisted extraction of electronic medical record information with machine learning algorithms. The supplied article excerpt reports recruitment of 1,243 women for model development and 452 women for external validation, but does not provide the outcome definition, model performance, or specific predictors.

Background

UI is common during pregnancy and after childbirth, and postpartum UI is associated with UI later in life. The article reports pooled prevalence estimates of 41.0% during pregnancy and 33.3% postpartum, and describes effects including discomfort, social embarrassment, reduced work ability, and lower social participation. It notes that postpartum UI may remain unrecognized in routine care and that uptake of pelvic floor muscle training, described as a first-line preventive and treatment measure, is inadequate. The authors identify limited use of unstructured electronic medical record information and difficulty representing nonlinear relationships and interactions as limitations of earlier prediction approaches.

Data Highlights

Key Findings

• The study aimed to develop and validate a postpartum UI prediction model integrating LLM-assisted information extraction from electronic medical records with multiple machine learning algorithms.
• The development cohort comprised 1,243 consecutively recruited eligible women at one campus of a tertiary maternity hospital in Hangzhou in 2020.
• External validation used 452 consecutively recruited eligible women at a second campus serving a relatively independent patient population in another administrative district in 2025.
• Eligibility was restricted to adults with singleton term pregnancies; specified exclusions included active urinary tract infection, stillbirth, fetal congenital malformation, and severe comorbidities.
• The article introduction cites pooled UI prevalence of 41.0% during pregnancy and 33.3% postpartum, and reports that postpartum UI is associated with UI later in life.
• The supplied excerpt does not report the model’s predictors, discrimination, calibration, or other validation results; related articles on postpartum hemorrhage prediction, postoperative nausea and vomiting prediction, and simulated obstetric LLM use concern different outcomes and do not establish findings for this UI model.

Clinical Implications

The authors frame risk prediction as a potential means of supporting early risk stratification and targeted prevention, particularly given that postpartum UI may be underrecognized and pelvic floor muscle training uptake is inadequate. The supplied excerpt does not provide model performance or implementation evidence, so it does not establish the model’s suitability for clinical decision-making.

Conclusion

The study describes a two-cohort approach to postpartum UI prediction using LLM-assisted extraction and machine learning. Assessment of its predictive value is not possible from the supplied excerpt because outcome and validation results are absent.

Related Resources & Content

1. Study authors not provided, Article excerpt, Year not provided -- Development and Validation of a Machine Learning Model Using Large Language Model Assistance to Predict Postpartum Urinary Incontinence
2. European Association of Urology, EAU Guidelines on Non-neurogenic Female LUTS, 2025 -- Disease Management
3. National Institute for Health and Care Excellence, NICE Guideline NG123 -- Urinary incontinence and pelvic organ prolapse in women: management
4. International Continence Society and International Urogynecological Association, 2022 -- Joint report on terminology for the assessment and management of obstetric pelvic floor disorders
5. Frontiers in Medicine — Creation and assessment of a risk prediction model for postpartum hemorrhage in older mothers undergoing vaginal delivery: an evaluation of various machine learning techniques
6. JMIR Medical Informatics — Forecasting Postoperative Nausea and Vomiting Within 24 Hours Through Machine Learning with Enhanced Interpretability from Large Language Models: A Development and Validation Study
7. Archives of Gynecology and Obstetrics — The role of large language models in the diagnosis and management of obstetric patients: a pilot feasibility study using simulated obstetric cases
8. MDSpire News — Postpartum Depression Predicted Before Discharge
9. Disease Management - EAU Guidelines on Non-neurogenic Female LUTS
10. Urinary incontinence and pelvic organ prolapse in women: management - NICE Guideline NG123
11. ICS/IUGA joint report on terminology for obstetric pelvic floor disorders
12. Optimizing Postpartum Care | ACOG
13. Pressure-Mediated Biofeedback With Pelvic Floor Muscle Training for Urinary Incontinence: A Randomized Clinical Trial.
14. Conservative Interventions for Urinary Incontinence on Postpartum Women: A Systematic Review and Meta‐Analysis - Chen - 2024 - Journal of Midwifery & Women's Health - Wiley Online Library
15. Impact of postpartum exercise on pelvic floor disorders and diastasis recti abdominis: a systematic review and meta-analysis - PMC
16. Prevalence and factors of urinary incontinence among postpartum: systematic review and meta-analysis - PMC
17. Longitudinal analysis of the association between parity, mode of delivery and urinary incontinence in midlife using the SWAN cohort data | Scientific Reports