{"title":"Evaluation of Pregnancy Risks in Women with Subchorionic Hematoma Using Machine Learning Models.","authors":"Lan Wang, Aiping Qin, Yihua Yang, Yufu Jin, Qiuyan Huang, Xinyue Huang, Yu Feng, Ting Liang","doi":"10.12659/MSM.945472","DOIUrl":null,"url":null,"abstract":"<p><p>BACKGROUND Subchorionic hematoma (SCH) can lead to blood accumulation and potentially affect pregnancy outcomes. Despite being a relatively common finding in early pregnancy, the effects of SCH on pregnancy outcomes such as miscarriage, stillbirth, and preterm birth remain debated. This study aims to address these gaps by systematically evaluating the influence of SCH-related clinical factors on pregnancy outcomes using robust analytical techniques. MATERIAL AND METHODS Data from SCH and non-SCH pregnant women were collected and split into training and test datasets. Machine learning classifiers and regression models were used to assess the impact of clinical indices on outcomes such as delivery type, NICU transfer, gestational age, and birth weight. Results were evaluated using ROC and calibration plots. RESULTS (1) SCH women had a significantly higher risk of stillbirth or miscarriage than non-SCH women (P<0.001). Logistic regression and XGB models showed AUCs of 0.858 and 0.916, respectively. Key factors affecting delivery outcomes included the first positive HCG level, hematoma duration, CA125 level, gestational sac diameter, fibrinogen level, and spouse age. (2) 12.7% of successfully delivered SCH newborns required NICU transfer, but clinical indices did not predict NICU need (AUC 0.589 and 0.629). (3) Successfully delivered SCH women had longer gestational ages than those with miscarriage/stillbirth (38.8 vs 10.1 weeks), but indices did not predict preterm/full-term birth (AUCs 0.449 and 0.503). (4) Birth weight was significantly affected by live birth times and gestational age (P<0.05), though the adjusted R-square was 0.226. CONCLUSIONS (1) SCH increases miscarriage or stillbirth risk. (2) the first positive HCG level, the hematoma duration, serum CA125 level, the gestational sac maximum diameter, fibrinogen, and the spouse age highly impacted the delivery outcome. (3) SCH indices do not affect NICU transfer or birth weight. (4) Miscarriage/stillbirth mainly occurs in the first trimester; passing this stage often leads to successful delivery. (5) The birth weight of full-term newborns is significantly higher than that of preterm infants. The clinical indices of SCH pregnant women have no impact on the birth weight of the newborn.</p>","PeriodicalId":48888,"journal":{"name":"Medical Science Monitor","volume":"30 ","pages":"e945472"},"PeriodicalIF":3.1000,"publicationDate":"2024-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11669258/pdf/","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Medical Science Monitor","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.12659/MSM.945472","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"Medicine","Score":null,"Total":0}
引用次数: 0
Abstract
BACKGROUND Subchorionic hematoma (SCH) can lead to blood accumulation and potentially affect pregnancy outcomes. Despite being a relatively common finding in early pregnancy, the effects of SCH on pregnancy outcomes such as miscarriage, stillbirth, and preterm birth remain debated. This study aims to address these gaps by systematically evaluating the influence of SCH-related clinical factors on pregnancy outcomes using robust analytical techniques. MATERIAL AND METHODS Data from SCH and non-SCH pregnant women were collected and split into training and test datasets. Machine learning classifiers and regression models were used to assess the impact of clinical indices on outcomes such as delivery type, NICU transfer, gestational age, and birth weight. Results were evaluated using ROC and calibration plots. RESULTS (1) SCH women had a significantly higher risk of stillbirth or miscarriage than non-SCH women (P<0.001). Logistic regression and XGB models showed AUCs of 0.858 and 0.916, respectively. Key factors affecting delivery outcomes included the first positive HCG level, hematoma duration, CA125 level, gestational sac diameter, fibrinogen level, and spouse age. (2) 12.7% of successfully delivered SCH newborns required NICU transfer, but clinical indices did not predict NICU need (AUC 0.589 and 0.629). (3) Successfully delivered SCH women had longer gestational ages than those with miscarriage/stillbirth (38.8 vs 10.1 weeks), but indices did not predict preterm/full-term birth (AUCs 0.449 and 0.503). (4) Birth weight was significantly affected by live birth times and gestational age (P<0.05), though the adjusted R-square was 0.226. CONCLUSIONS (1) SCH increases miscarriage or stillbirth risk. (2) the first positive HCG level, the hematoma duration, serum CA125 level, the gestational sac maximum diameter, fibrinogen, and the spouse age highly impacted the delivery outcome. (3) SCH indices do not affect NICU transfer or birth weight. (4) Miscarriage/stillbirth mainly occurs in the first trimester; passing this stage often leads to successful delivery. (5) The birth weight of full-term newborns is significantly higher than that of preterm infants. The clinical indices of SCH pregnant women have no impact on the birth weight of the newborn.
期刊介绍:
Medical Science Monitor (MSM) established in 1995 is an international, peer-reviewed scientific journal which publishes original articles in Clinical Medicine and related disciplines such as Epidemiology and Population Studies, Product Investigations, Development of Laboratory Techniques :: Diagnostics and Medical Technology which enable presentation of research or review works in overlapping areas of medicine and technology such us (but not limited to): medical diagnostics, medical imaging systems, computer simulation of health and disease processes, new medical devices, etc. Reviews and Special Reports - papers may be accepted on the basis that they provide a systematic, critical and up-to-date overview of literature pertaining to research or clinical topics. Meta-analyses are considered as reviews. A special attention will be paid to a teaching value of a review paper.
Medical Science Monitor is internationally indexed in Thomson-Reuters Web of Science, Journals Citation Report (JCR), Science Citation Index Expanded (SCI), Index Medicus MEDLINE, PubMed, PMC, EMBASE/Excerpta Medica, Chemical Abstracts CAS and Index Copernicus.