Icon Legend

This session is not in your schedule.

This session is in your schedule. Click again to remove it.

Presentation Icons

Additional registration fee required

Faculty have requested this content not be shared outside of the session

CME Credit Offered

Abstract Award

Recording available 2/16-5/2

AIUM Credit

Foundation Awardee

Poster Icons

Abstract Award

Foundation Awardee

Poster Session 4

Category: Digital Health Technologies (DHT)

Poster Session 4

(1281) Prepregnancy risk factors for placenta accreta spectrum: understanding data structure using principal component analysis

Thursday, February 12, 2026

3:30 PM - 5:00 PM

Submitting Author and Presenting Author(s)

Tess EK Cersonsky, MD (she/her/hers)

Resident
Icahn School of Medicine at Mount Sinai, New York, United States

Coauthor(s)

EA

Eugenia Alleva, MD

Icahn School of Medicine at Mount Sinai
New York, New York, United States
IB

Isabelle Band, MD

Resident Physician
Mount Sinai Hospital
New York, New York, United States
AB

Angela T. Bianco, MD

Professor and Division Director, Maternal-Fetal Medicine
Icahn School of Medicine at Mount Sinai
New York, New York, United States
Henri M. Rosenberg, MD (he/him/his)

Maternal Fetal Medicine Attending
Icahn School of Medicine at Mount Sinai
Icahn School of Medicine at Mount Sinai, New York, United States

Objective:
Although certain risk factors for placenta accreta spectrum (PAS), such as prior cesarean, uterine surgery, or placenta previa, are well established, many causal factors remain poorly understood. This limits our ability to predict and prepare for PAS deliveries, which carry substantial risk for severe maternal morbidity and mortality. Therefore, we sought to investigate the underlying data in a PAS cohort using principal component analysis (PCA). This information will help build clinical models that focus on important risk factors for PAS.

Study Design:
This analysis was performed on a cohort built for machine learning analysis comprised of parturients with confirmed PAS and controls. Only variables available prior to the index pregnancy were included. A case-matched subset was selected for dimensionality reduction using Kernel PCA. The proportion of variance explained by each principal component was calculated to assess the structure of the data. Sensitivity analysis was performed to estimate the relative contribution of each variable to the components, providing an approximation of variable importance.

Results:
Records from 118890 patients were considered, and a subset of 11889 patients and 260 variables were included. Variables fit into 982 components, of which the first 20 explained 50% of the total variance. Component 1 explained 20% of the variance and included several pre-existing clinical variables, notably, history of cesarean delivery, myomectomy, hysteroscopy, or dilation & curettage.

Conclusion:
Kernel PCA can help identify underlying data structures in a sample of patients with and without PAS. Most significantly, prior surgical history and gynecologic conditions make up a large portion of the first principal component, suggesting that these factors cluster in a way that is critical for PAS risk stratification. This information can help reduce dimensions entered into machine learning algorithms, improving ease of use and applicability in clinical settings.