OC15.07: Application of machine learning methods to predict the success of expectant or medical management of miscarriage
Year: 2024
Type: conference-abstract
Abstract: The aim of this study is to use machine learning to develop a clinical decision support tool to offer women personalised outcome prediction for expectant and medical miscarriage management. A retrospective, multi-site study of patients selected for expectant or medical management of miscarriage was undertaken. A total of 1075 patients across two hospital early pregnancy units were eligible for inclusion. The clinical variables collected included patient demographics, history and ultrasound features. These 14 variables were selected following a systematic review of studies evaluating predictors of miscarriage management outcomes and discussion with clinical experts. These variables were routinely recorded as part of the routine clinical care of miscarriage patients. A combination of 11 linear, Bayesian, neural-net and tree-based machine learning algorithms were applied to 8 different feature sets. Receiver-Operator Characteristic (ROC) curves and Area Under the Curve (AUC) were created for the training, validation and test set results for our best performing model and feature selection combination. Parameters were in the majority well matched across training-validation and external test sets. The respective optimum training, validation and external test set AUC scores were as follows in the expectant management cohort: 0.72 (95% CI 0.67,0.77), 0.63 (95% CI 0.53,0.73) and 0.70 (95% CI 0.60,0.79) (Logistic Regression (LR) in combination with Least Absolute Shrinkage and Selection Operator (LASSO)). The AUC scores in the medical cohort were 0.64 (95% CI 0.56,0.72), 0.62 (95% CI 0.45,0.77) and 0.71 (95% CI 0.58,0.83) (Logistic Regression (LR) in combination with Recursive Feature Elimination (RFE) feature reduction). This robust machine learning method has been validated and externally tested. This study sets the stage for future clinical trials prospectively offering quantitative personalised outcome prediction for patients making a decision on their miscarriage management.
DOI: 10.1002/uog.27828
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4402359781