Application of Machine Learning Techniques for Enuresis Prediction in Children

Area of Science:

• Pediatric Medicine
• Artificial Intelligence
• Data Science

Background:

• Machine learning techniques (MLTs) offer powerful tools for analyzing large datasets in healthcare.
• Childhood enuresis remains a common condition requiring effective predictive models.

Purpose of the Study:

• To develop and validate a machine learning model for predicting enuresis in children.
• Identify key predictive factors for childhood enuresis.

Main Methods:

• Utilized a dataset of 8,071 elementary school students, with 704 diagnosed with enuresis.
• Selected 14 significant independent variables from an initial 34 using stratified sampling and feature importance analysis.
• Trained a predictive model using logistic regression, identified as the best-performing MLT algorithm.

Main Results:

• The logistic regression model achieved an 81.3% prediction accuracy for enuresis.
• Key predictors identified include: toilet training age, urinary urgency, voiding postponement, defecation frequency, family history of enuresis, child's own room, parental education, sibling history, consanguinity, incomplete bladder emptying, frequent voiding, gender, UTI history, and past surgeries.
• Variable importance ranking highlighted specific factors significantly influencing enuresis prediction.

Conclusions:

• MLTs provide an efficient and objective method for enuresis prediction, especially with large datasets.
• The developed model and identified variables can guide screening and proactive interventions for enuresis.
• MLT application can minimize cognitive biases in clinical diagnosis and treatment planning for enuresis.