February 7, 2023Open Access

Prediction of Preeclampsia from Clinical and Genetic Risk Factors in Early and Late Pregnancy Using Machine Learning and Polygenic Risk Scores

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Preeclampsia (n=1125). Intervention: Clinical and genetic risk factors (SBP PRS) predictive models. Primary outcome: Prediction of preeclampsia (Area Under the Curve) (AUC 0.91).

View Full Paper

Ask AI

Bookmark

Key result

Xgboost machine learning model using clinical variables predicts late-pregnancy preeclampsia with 0.91 AUC.

AUC 0.91
n=1,125

Why the study?

Preeclampsia is a leading cause of maternal and neonatal morbidity and mortality, and predictive tools are needed to identify individuals most at risk.

Does integrating clinical and genetic factors (SBP PRS) using machine learning improve the prediction of preeclampsia in pregnant individuals?

Population

N=1,125 pregnant individuals who delivered between 05/2015-05/2022 at Mass General Brigham hospitals

Comparison

Machine learning and linear regression models using clinical EHR data with or without SBP PRS

Design

Retrospective cohort study

Authors

VKVesela KovachevaBrigham and Women's Hospital BEBraden W. EberhardCalifornia University of Pennsylvania RCRaphael Y. CohenBrigham and Women's Hospital

Discussion

Member takes

Implication

Supports validation of ML models for late-pregnancy preeclampsia prediction; leaves open added value of genetic factors before clinical adoption.

Study Design

Type

Cohort (n=1,125)

Multicenter

Yes

Structured PICO

Does integrating clinical and genetic factors (SBP PRS) using machine learning improve the prediction of preeclampsia in pregnant individuals?

Population

1,125 pregnant individuals with available electronic health record and linked genetic data who delivered at Mass General Brigham hospitals.

Exposure

Machine learning (xgboost) and linear regression models integrating clinical EHR data and systolic blood pressure polygenic risk scores (SBP PRS)

Comparator

Models using only clinical variables or only genetic variables

Outcome

Prediction of preeclampsia risk (measured by Area Under the Curve [AUC])

Main Result

Effect estimate: AUC 0.91

Machine learning models using clinical variables can highly accurately predict preeclampsia in late pregnancy, while polygenic risk scores offer modest predictive improvements in early pregnancy.

Limitations

SBP PRS was developed using a European population
Limited sample size of preeclampsia cases (N=87)
Small cohort of patients
Limited types of analyses (unable to investigate predictions of early-onset preeclampsia)
Some variables are based on billing codes which may be inaccurate and do not reflect disease severity

Cite This Study

Kovacheva et al. (2023) conducted a cohort in Preeclampsia (n=1,125). Clinical and genetic risk factors (SBP PRS) predictive models was evaluated on Prediction of preeclampsia (Area Under the Curve) (AUC 0.91). An xgboost machine learning model using clinical variables achieved an area under the curve of 0.91 for predicting preeclampsia in late pregnancy.

synapsesocial.com/papers/6aa3e7a4fd19ae9969b15363 https://doi.org/10.1101/2023.02.03.23285385

View Full Paper

Ask AI

Bookmark

February 7, 2023Open Access

Prediction of Preeclampsia from Clinical and Genetic Risk Factors in Early and Late Pregnancy Using Machine Learning and Polygenic Risk Scores

View Full Paper

Ask AI

Bookmark

Key result

Xgboost machine learning model using clinical variables predicts late-pregnancy preeclampsia with 0.91 AUC.

AUC 0.91
n=1,125

Why the study?

Preeclampsia is a leading cause of maternal and neonatal morbidity and mortality, and predictive tools are needed to identify individuals most at risk.

Does integrating clinical and genetic factors (SBP PRS) using machine learning improve the prediction of preeclampsia in pregnant individuals?

Population

N=1,125 pregnant individuals who delivered between 05/2015-05/2022 at Mass General Brigham hospitals

Comparison

Machine learning and linear regression models using clinical EHR data with or without SBP PRS

Design

Retrospective cohort study

Authors

VKVesela KovachevaBrigham and Women's Hospital BEBraden W. EberhardCalifornia University of Pennsylvania RCRaphael Y. CohenBrigham and Women's Hospital

Discussion

Member takes

Implication

Supports validation of ML models for late-pregnancy preeclampsia prediction; leaves open added value of genetic factors before clinical adoption.

Study Design

Type

Cohort (n=1,125)

Multicenter

Yes

Structured PICO

Does integrating clinical and genetic factors (SBP PRS) using machine learning improve the prediction of preeclampsia in pregnant individuals?

Population

1,125 pregnant individuals with available electronic health record and linked genetic data who delivered at Mass General Brigham hospitals.

Exposure

Machine learning (xgboost) and linear regression models integrating clinical EHR data and systolic blood pressure polygenic risk scores (SBP PRS)

Comparator

Models using only clinical variables or only genetic variables

Outcome

Prediction of preeclampsia risk (measured by Area Under the Curve [AUC])

Main Result

Effect estimate: AUC 0.91

Machine learning models using clinical variables can highly accurately predict preeclampsia in late pregnancy, while polygenic risk scores offer modest predictive improvements in early pregnancy.

Limitations

SBP PRS was developed using a European population
Limited sample size of preeclampsia cases (N=87)
Small cohort of patients
Limited types of analyses (unable to investigate predictions of early-onset preeclampsia)
Some variables are based on billing codes which may be inaccurate and do not reflect disease severity

Cite This Study

synapsesocial.com/papers/6aa3e7a4fd19ae9969b15363 https://doi.org/10.1101/2023.02.03.23285385

Also Consider

Synapse has enriched 4 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark