Predicting the Association Between Oral Hyperpigmentation and Pregnancy in Women Aged 18-45 Using Machine Learning and Directed Acyclic Graphs: A Cross-Sectional Study

ObjectivesThe study aims to predict the association between oral hyperpigmentation and pregnancy in women between 18 and 45 years using machine learning (ML) algorithms. Directed acyclic graphs (DAGs) were constructed to explore the association between oral hyperpigmentation and demographic factors such as age, occupation, ethnicity, and number of pregnancies. MethodsThis cross-sectional study involved pregnant and non-pregnant participants aged between 18 and 45. The study was conducted at one public and one private sector hospital. DAGs were used to represent relationships between different variables. Three ML models, including Logistic Regression (LR), Random Forest (RF) and Gradient Boosting Machine (GBM) algorithms, were used to predict the association between oral hyperpigmentation and pregnancy based upon relevant theoretically backed predictors. ResultsThe results show that labial mucosa and hard palate hyperpigmentation had a positive correlation with the number of pregnancies, with p-values of 0.003 and 0.018, respectively. All three ML models demonstrated excellent predictive power. However, the LR model with an accuracy rate of 96.67% was particularly effective in predicting the association between oral hyperpigmentation and pregnancy. ConclusionsThis study highlighted the effectiveness of LR, supported by DAGs, in predicting oral hyperpigmentation among women aged 18-45. Key factors such as the number of pregnancies, age, ethnicity, and occupation significantly influenced site-specific hyperpigmentation. The findings support the integration of ML into maternal oral healthcare for early detection and culturally tailored interventions. Simpler ML models proved most accurate in moderate-sized datasets with fewer covariates. SignificanceOral hyperpigmentation is known to be impacted by pregnancy, though few predictive models incorporate demographic factors into consideration. This study uses DAG-supported machine learning models to investigate and predict the association between oral hyperpigmentation and pregnancy in women of age 18 to 45 years. This study contributes to data-driven maternal healthcare by developing predictive models tailored to the ethnically diverse population, aiming to improve patient counselling, early detection, and guide preventive measures in oral healthcare for females.