Evaluation of Doc’EDS: A French Semantic Search Tool to Query Health Documents from A Clinical Data Warehouse

Abstract

Background: Unstructured data from electronic health record is a gold mine. Doc’EDS is a pre-screening tool based on textual and semantic analysis. The system provides an easy-to-use interface to search documents in French. The aim of this study is to present the tools and to provide a formal evaluation of its semantic features. Material & Methods: Doc’EDS is a search tool built on the top of the clinical data warehouse developed in the Rouen University Hospital. This tool is a multilevel search engine combining structured and unstructured data. It also provides basic analytics features and semantic utilities. A formal evaluation has been conducted to measure the implemented Natural Language Processing algorithms. Results: About 17,3 million of narrative documents are contained in this CDW. The formal evaluation has been conducted over 5,000 clinical concepts that were manually collected. Negation concepts detection F-measure was 0.89, hypothesis concept detection F-measure was 0.57. Conclusion: We hereby present Doc’EDS, a semantic search tool which deals with language subtleties to enhance an advanced full text search engine dedicated to French health documents. This tool is currently used on a daily basis to help researchers identifying patients thanks to unstructured data.

Using Clinical Data Warehouse to Optimize the Vaccination Strategy Against COVID-19: A Use Case in France

“…In these cases, the automatic processing of clinical narrative is used. It combines several NLP algorithms to ensure that searched keywords are relevant (i.e., not in a negative sentence for example) or present in text specific segments (such as conclusion for example) even if those documents are natively unstructured [15].…”

“…Processing CN with NLP is never perfect: some false positives or false negatives can occur (e.g., in detection of negations) [15]. Thus, an evaluation should be conducted to measure the precision of each query.…”

When Context Matters for Credible Measurement of Drug-Drug Interactions Based on Real-World Data

“…In this paper, we use a unique prescription dataset from the Rouen University Hospital data warehouse (7) to test the impact of different contextual items on the DDI frequency measurement.…”

Immunomodulatory and/or immunosuppressive drugs should not be stopped prior to skin tests for the assessment of drug allergy

“…In Rouen, patients were included from the Clinical Dataware House (EDSaN) with a cross-search using the keywords in narrative documents: 'allergic tests' and 'immunosuppressive therapy' (list of drug commercial names or substances). 1 ST were performed in the same way in the two centres according to the international guidelines. 2,3 All included patients (n = 61) gave written consent for allergological investigations and were informed of the potential use of the data for research purposes.…”

“…6 We assume that the heterogeneity of our real-life population (type of IT and culprit drugs, Accuracy of commercially available smartphone applications for the detection of melanoma DOI: 10.1111/bjd.20903 DEAR EDITOR, Artificial intelligence (AI) has shown promise in the analysis of images for detection of melanoma. 1 The number of available dermatology smartphone applications ('apps') is rapidly growing and there is increasing interest in apps that provide diagnosis or triage of skin lesions. 2,3 A 2020 systematic review found that nine studies evaluating six apps had poor study design and high risk of bias.…”