Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Jan 14, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Predicting Postpartum Hemorrhage Using Clinical Features Extracted With Large Language Models.

Elizabeth G Woo¹, Israel Zighelboim¹, Tyler Gifford¹

¹Center for Computational Medicine and Clinical AI, Department of Medicine, and the Section of Ultrasound, Genetics, and the Fetal Neonatal Care Center, Department of Obstetrics and Gynecology, University of Chicago, Chicago, Illinois; the Department of Obstetrics and Gynecology and the Division of Gynecologic Oncology, St. Luke's Cancer Center, St. Luke's University Health Network, Bethlehem, Pennsylvania; the Department of Biomedical Data Science, Stanford University, Stanford, California; and Maternal Fetal Medicine, Oregon Health & Science University, Portland, Oregon.

|October 20, 2025

Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

BRIDGE: benchmarking large language models for understanding real-world clinical practice texts.

Nature biomedical engineering·2026

Same author

Amniocentesis prior to cerclage: practice patterns and obstetric outcomes.

American journal of obstetrics & gynecology MFM·2026

Same author

Physician-Reported Safety Outcomes of AI-Generated Hospital Course Summaries.

JAMA network open·2026

Same author

The role of pre-operative sit-to-stand assessment in robotic hysterectomy surgical outcomes: An observational cohort study.

Gynecologic oncology reports·2026

Same author

Interventions and Perinatal Outcomes Associated With Recipient Twin Cardiomyopathy in the Setting of Twin-To-Twin Transfusion Syndrome: A Systematic Review.

Prenatal diagnosis·2026

Same author

Acceptability and barriers to rapid HIV/syphilis testing among pregnant Wayuu women in Colombia: A cross-sectional pragmatic study.

International journal of gynaecology and obstetrics: the official organ of the International Federation of Gynaecology and Obstetrics·2026

Same journal

Addressing Obstetric and Gynecologic Needs in Thalassemia.

O&G open·2026

Same journal

Parental Leave Policies in Obstetrics and Gynecology Training Programs in the United States and Canada.

O&G open·2026

Same journal

National Trends in CA 125 Testing and Transvaginal Ultrasound Use in Older Women Receiving Routine Gynecologic Care.

O&G open·2026

Same journal

Association Between Resident Involvement in Cesarean Delivery, Complications, and Operative Times.

O&G open·2026

Same journal

Geospatial Analysis of Abortion Access for U.S. Servicewomen After <i>Dobbs</i>.

O&G open·2026

Same journal

Ovarian Reserve and Planned Cryopreservation Outcomes in Physicians and Nonphysicians.

O&G open·2026

See all related articles

This summary is machine-generated.

Large language models (LLMs) can predict postpartum hemorrhage (PPH) using clinical notes, outperforming traditional methods. LLM-based approaches offer improved risk stratification for better PPH prevention.

Area of Science:

Medical Informatics
Artificial Intelligence in Healthcare
Obstetrics and Gynecology

Background:

Postpartum hemorrhage (PPH) remains a leading cause of maternal morbidity and mortality.
Accurate prediction of PPH risk before labor onset is crucial for timely intervention.
Current prediction models often rely on structured data, potentially missing critical information in clinical notes.

Purpose of the Study:

To evaluate the efficacy of large language models (LLMs) in predicting postpartum hemorrhage (PPH) from prenatal clinical notes.
To compare the performance of LLM-based prediction models against traditional methods using structured data.
To assess LLM performance across different PPH outcome definitions, including a novel intervention-based definition.

Main Methods:

More Related Videos

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

Related Experiment Videos

Last Updated: Jan 14, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

A retrospective cohort study was conducted using electronic medical record data from a large health network.

Two PPH definitions were used: estimated blood loss (EBL-QBL) and a clinical intervention-based definition (cPPH).

Three prediction pipelines were evaluated: structured data-only machine learning, LLM-direct prediction from notes, and LLM-extracted features combined with structured data.

Main Results:

LLM-based direct prediction achieved the highest predictive performance (AUROC 0.79-0.80) for PPH.
Interpretable models combining LLM features and structured data showed strong performance (AUROC 0.76-0.78).
Structured data-only models had significantly lower performance (AUROC 0.65-0.71), and LLM feature extraction identified 47 significant predictors.

Conclusions:

LLM-based approaches demonstrate significant potential to enhance PPH risk stratification beyond structured data.
The LLM feature extraction method provides a balance of predictive accuracy and clinical interpretability.
Integrating LLMs into clinical workflows could enable earlier PPH detection and targeted preventive strategies.