JMIR Publications

ABSTRACT

Background:

Overactive bladder (OAB) is a prevalent condition, particularly among women, characterized by urinary urgency, often accompanied by frequency and nocturia. Traditional risk prediction methods for OAB are limited, failing to fully integrate multidimensional risk factors, including female reproductive history. Machine learning (ML) offers potential for enhanced predictive accuracy using large-scale datasets like the National Health and Nutrition Examination Survey (NHANES).

Objective:

This study aimed to develop and validate a machine learning-based model to predict OAB risk in women, incorporating reproductive and sociodemographic factors, and to identify key predictors using interpretable methods.

Methods:

We conducted a cross-sectional study using NHANES data (2011–2018) from 7,884 female participants. OAB was assessed via the Overactive Bladder Symptom Score (OABSS). Lasso regression, univariate, and multivariate logistic regression were used to select key variables in the training set (75%, n=5,519). Five ML models (logistic regression, support vector machines, XGBoost, naive Bayes, neural networks) were developed and evaluated using the area under the receiver operating characteristic curve (AUROC). The SHAP (SHapley Additive exPlanations) method interpreted the optimal model, and restricted cubic spline (RCS) curves analyzed dose-response relationships.

Results:

Five variables—age, body mass index (BMI), poverty-to-income ratio (PIR), age at menarche, and number of vaginal deliveries—were identified as significant predictors. The XGBoost model outperformed others, achieving an AUROC of 0.730 (95% CI 0.716–0.744) in the training set and 0.695 (95% CI 0.673–0.718) in the test set (25%, n=2,365). SHAP analysis ranked age, BMI, and PIR as the top contributors to OAB risk. RCS revealed nonlinear associations of BMI and vaginal deliveries with OAB risk, with thresholds at 33.659 kg/m² and >0 deliveries, respectively, and linear associations with age (>58 years), earlier menarche (<13 years), and lower PIR (<3.299).

Conclusions:

The XGBoost model, enhanced by SHAP interpretability, provides a robust tool for predicting OAB risk in women, integrating reproductive and sociodemographic factors. This approach facilitates early identification of high-risk individuals, supporting personalized prevention and clinical management strategies.