Automated Analysis of Public Health Laboratory Test Results

This study investigates the use of machine learning methods for classifying and extracting structured information from laboratory reports stored as semi-structured point-form English text. This is a novel data format that has not been evaluated in conjunction with machine learning classifiers in previous literature. Our classifiers achieve human-level predictive accuracy on the binary Test Performed and 4-class Test Outcome labels. We consider symbolic approaches for predicting the highly multi-class Organism Genus and Organism Species labels. Results are discussed from the viewpoint of interpretability and generalizability to new incoming laboratory reports. Code has been made public at https://github.com/enchainingrealm/UbcDssgBccdc-Research/tree/master/src.