Abstract
The purpose of automated health surveillance systems is to predict the emergence of a disease. In most cases, these systems use a text categorization model to classify any clinical text into a category corresponding to an illness. The problem arises when the target classes refer to diseases sharing multiple information such as symptoms. Thus, the classifier will have difficulty discriminating the disease under surveillance from other conditions of the same family, causing an increase in misclassification rate. Clinical texts contain keywords carrying relevant information to distinguish diseases with similar symptoms. However, these specific words are rare and sparse. Therefore, they have a minor impact on machine learning models' performance. Assuming that emphasizing specific terms contributes to improving classification performance, we propose an algorithm that enriches training samples with terms semantically similar to specific terms using the deep contextualized word embeddings ELMo. Next, we devise a weighting scheme combining chi-square and semantic scores to reflect the relatedness between features and the disease under surveillance. We evaluate our model using the SVM algorithm trained on i2b2 dataset supplemented by documents collected from Ibn Sina hospital in Rabat. Experimental results show a clear improvement in classification performance than baseline methods with an F-measure reaching 86.54%.
Highlights
Public health surveillance is a significant focus of National health policies
Evaluation A health surveillance system must be efficient enough to accurately detect the onset and progression over time of a disease, so our proposed model is designed to meet the following two requirements: 1) Reduce the proportion of mild flu-related documents classified as severe flu, this has the effect of avoiding false outbreak alerts
A system for detecting the occurrence of severe forms of flu by using only clinical texts recorded in electronic health record (EHR) is devised through a text classification model with the challenge of discriminating between severe and mild flu-related documents containing many common features
Summary
Public health surveillance is a significant focus of National health policies. It is ensured by collecting epidemiological data from various healthcare facilities to detect disease outbreaks and subsequently plan appropriate response strategies early. The risk of misclassification increases, especially for documents related to severe influenza cases, since the frequency of specific features that characterize severe cases is low compared to common features frequency In this respect, many research efforts attempt to improve feature selection algorithms by highlighting the discriminative power of infrequent specific terms. The idea behind this algorithm is to mitigate the deficiency caused by the scarcity of specific features by adding new features to training samples in order to counterbalance the preponderance of common features This algorithm is based on a deep contextualized word representation method named: Embeddings from language models (ELMo), renowned for its power in detecting the finest syntactic and semantic characteristics of words. Experimental results show significant improvement compared to ontology-based feature methods and static word embeddings models with a notable decrease in misclassification rate of test clinical notes related to severe flu by reaching an F-measure of 86.54%.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
