Explainable machine learning-based stratification of severe obstetric complications using small-scale EMR data

I Gusti Njoman Teguh Budhi Bimantara, I Gusti Putu Asto Buditjahjanto, Lilik Anifah, Gusti Ayu Az Zahra Shabrina Putri Bukian

Abstract


Severe obstetric complications related to miscarriage remain a major contributor to maternal morbidity. This study aimed to develop and evaluate explainable machine learning (ML) models to classify patterns of severe obstetric complications using electronic medical record (EMR) data. A cross-sectional dataset of 57 patients with miscarriage-related conditions was analyzed. Severe-complication patterns were defined as incomplete, missed, or incipient miscarriage, whereas complete miscarriage and threatened/imminent miscarriage were categorized as lower-risk outcomes. Predictors included maternal age, parity, hemoglobin level, infection status, and miscarriage history. Multiple ML algorithms were evaluated using repeated stratified k-fold cross-validation and out-of-fold (OOF) analysis, with synthetic minority oversampling technique (SMOTE) applied to address class imbalance. Model interpretability was assessed using Shapley additive explanations (SHAP). Extreme gradient boosting (XGBoost) with SMOTE achieved the best performance (balanced accuracy=0.792, area under the curve (AUC)=0.806), with an accuracy of 0.877 and an F1-score of 0.926. SHAP analysis identified infection status and hemoglobin level as the leading contributors to model predictions. These exploratory findings support the feasibility of explainable ML for complication-pattern stratification using routinely available, small-scale EMR data, with potential relevance to resource-constrained maternal-health settings; however, larger multicenter datasets, calibration assessment, and external validation are required before clinical application.

Keywords


Electronic medical records; Explainable artificial intelligence; Machine learning; Severe obstetric complications; Shapley additive explanations

Full Text:

PDF


DOI: https://doi.org/10.11591/eei.v15i6.13309

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Bulletin of EEI Statistics

Bulletin of Electrical Engineering and Informatics (BEEI)
ISSN: 2089-3191 , e-ISSN: 2302-9285
This journal is published by the Institute of Advanced Engineering and Science (IAES) .