Thesis

Cancellation and No Show Predictions Using Machine Learning in Dental Practices

In healthcare, medical practices/clinics specialize in care for the entire body whereas dental practices/clinics specialize in care for teeth and mouth. Medical and dental practices alike face considerable cancellations and no shows rates. Most of the research on these cancellations/no shows (C/NS) are for medical practices but they can also be applied to dental practices. The resulting methods have not, to the author's knowledge, yet been applied to a dental practice. For a dental practice, these cancellations/no shows are costly because of the employee and doctor idle times associated with them. In this study, we perform exploratory data analysis (EDA) on appointments and basic demographics of a dental practice with the goal of finding patterns that allow the practice to predict C/NS using machine learning. The data is used to train five statistical models on a year's worth of real-world patient data taken from a California dental practice. We then use these models to guide the prediction of patient cancellation/no shows and measure these predictions using 75% of the appointments to train and 25% of the appointments to test (aka holdout data). The five models used were Logistic Regression, K Nearest Neighbors, Support Vector Machine, Feed Forward Neural Networks and Decision Tree. The models are then adjusted by adjusting the variables/features used. In this research, the findings and suggestions for future work are presented.

Relationships

Items