Labour monitoring and decision support: a machine-learning-based paradigm

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Current labor monitoring tools like the partograph have limitations.
• Purely supervised machine learning models may lack flexibility and interpretability.
• Real-time decision support during labor is crucial for improving outcomes.

Purpose of the Study:

• To propose a novel machine-learning paradigm for real-time labor monitoring and decision support.
• To address the limitations of existing methods in labor management.
• To develop a flexible and interpretable tool for healthcare providers.

Main Methods:

• A hybrid unsupervised and supervised machine learning approach was developed.
• Unsupervised dimensionality reduction was used to visualize labor data.
• Personalized healthy labor trajectories were estimated using historical cohort data.
• The World Health Organization's Better Outcomes in Labour Difficulty (BOLD) study data (9,995 women) was utilized.

Main Results:

• The proposed model achieved a sensitivity and specificity of approximately 0.70 across all women.
• Predictive performance varied across different subgroups.
• The model demonstrated effectiveness in predicting caesarean sections.

Conclusions:

• The developed machine learning approach offers superior flexibility and interpretability compared to the partograph.
• It matches the performance of state-of-the-art supervised models.
• This paradigm provides a valuable real-time monitoring and decision-support solution for labor management.