Automatic Segregation and Classification of Inclusion and Exclusion Criteria of Clinical Trials to Improve Patient Eligibility Matching

Abstract

Clinical trials are aimed to observe the effectiveness of a new intervention. For every clinical trial the eligibility requirements of a patient are specified in the form of inclusion or exclusion criteria. However, the process of eligibility determination is extremely challenging and time-consuming. Such a process typically involves repeated manual reading followed by matching of the trial descriptions mentioning the eligibility criteria and patient’s electronic health record (EHR) for multiple trials across every visit. Thus, the number of patients to be evaluated gets reduced. In this work, we have focused on a small but important step towards automatic segregation and classification of inclusion-exclusion criteria of clinical trials to improve patient eligibility matching. Accordingly, we have proposed an attention aware CNN-Bi-LSTM model. We evaluate our model with different word and character level embeddings as input over two different openly available datasets. Experimental results demonstrate our proposed model surpasses the performance of the existing baseline models. Furthermore, we have observed that character level information along with word embeddings boost up the predictive performance of criteria classification (in terms of F1-Scores), which is a promising direction for further research.