This study aims to determine if artificial neural networks can classify sleep apnea patients by severity level using 3,281 patient records with 20 input features. The neural network architecture tested varied parameters like learning rate, batch size, optimizer, activation function, and number of hidden layers and nodes. The best performing model achieved an accuracy of 88.13% for 2 output classes and 61.95% for 4 output classes, but complete replacement of manual labor was not deemed feasible due to data imbalances and room for improved clinical data and neural network accuracy. Further development has the potential to replace both manual processing and overnight sleep studies.