Untitled

Abstract

The purpose of this retrospective study is to measure machine learning models' ability to predict glaucoma drainage device failure based on demographic information and preoperative measurements. The medical records of 165 patients were used. Potential predictors included the patients' race, age, sex, preoperative intraocular pressure (IOP), preoperative visual acuity, number of IOP-lowering medications, and number and type of previous ophthalmic surgeries. Failure was defined as final IOP greater than 18 mm Hg, reduction in intraocular pressure less than 20% from baseline, or need for reoperation unrelated to normal implant maintenance. Five classifiers were compared: logistic regression, artificial neural network, random forest, decision tree, and support vector machine. Recursive feature elimination was used to shrink the number of predictors and grid search was used to choose hyperparameters. To prevent leakage, nested cross-validation was used throughout. With a small amount of data, the best classfier was logistic regression, but with more data, the best classifier was the random forest.