A female overactive bladder risk model developed by machine learning: based on 2007-2018 NHANES data.

Area of Science:

• Urology
• Medical Informatics
• Machine Learning

Background:

• Overactive bladder (OAB) significantly impacts daily life, with current risk estimation relying heavily on subjective patient symptoms.
• There is a critical need for objective diagnostic tools and risk prediction models for OAB.
• Developing novel risk assessment strategies is essential for timely and accurate OAB diagnosis.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for assessing overactive bladder (OAB) risk in the female population.
• To identify key contributing factors to OAB risk using data-driven approaches.
• To enhance the diagnostic capabilities for OAB in women.

Main Methods:

• Utilized data from 10,807 female participants in the National Health and Nutrition Examination Survey (NHANES) from 2007-2018.
• Trained and compared seven ML algorithms, including Support Vector Machine (SVM), Logistic Regression, K-Nearest Neighbor (KNN), Random Forest (RF), Gradient Boosting, Decision Tree (DT), and Extreme Gradient Boosting (XGBoost).
• Developed models using ten characteristic factors to predict OAB risk.

Main Results:

• The Random Forest (RF) model achieved the highest performance, with an Area Under the Curve (AUC) of 0.879.
• Hypertension was identified as the most significant factor influencing OAB risk.
• Diabetes and sleep disorders were also found to be important contributing factors to OAB risk.

Conclusions:

• Machine learning models demonstrate strong diagnostic performance and interpretability for predicting female overactive bladder (OAB) risk.
• The developed ML-based risk model offers a valuable tool for improving OAB diagnosis in women.
• This approach provides a more objective method for OAB risk assessment compared to symptom-based evaluations.