Methods for outlier detection in clinical registries

Clinical registries collect data about specific patient groups (including people with a certain disease or undergoing a certain surgical procedure) to monitor their outcomes and improve their quality of care. Registries will often compare patient care and outcomes (such as mortality) between hospitals (benchmarking) and then use statistical methods to find any hospitals that have significantly worse outcomes (outliers) that can be targeted for improvement. This thesis evaluates the accuracy of different benchmarking and outlier detection methods, and makes recommendations for the best clinical registry settings for improving outlier detection accuracy.