Paul Morrison1, Maxwell Dixon2, Arsham Sheybani2, Bahareh Rahmani1, 3, 1Fontbonne University, USA, 2Washington University, USA, 3Maryville University, USA
The purpose of this retrospective study is to measure machine learning models' ability to predict glaucoma drainage device (GDD) failure based on demographic information and preoperative measurements. The medical records of sixty-two patients were used. Potential predictors included the patient's 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 IOP 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. Overall, the best classifier was logistic regression.With a small amount of data, the best classifier was logistic regression, but with more data, the best classifier was the random forest. All five classification methods discussed at this research confirm that race effects on failure glaucoma drainage. Use of topical beta-blockers preoperatively is related to device failure. In treating glaucoma medically, prostaglandin equivalents are often first-line with beta-blockers used second-line or as a reasonable alternative first-line agent.