Towards iconic language for patient records, drug monographs, guidelines and medical search engines.

Practicing physicians have limited time for consulting medical knowledge and records. We have previously shown that using icons instead of text to present drug monographs may allow contraindications and adverse effects to be identified more rapidly and more accurately. These findings were based on the use of an iconic language designed for drug knowledge, providing icons for many medical concepts, including diseases, antecedents, drug classes and tests. In this paper, we describe a new project aimed at extending this iconic language, and exploring the possible applications of these icons in medicine. Based on evaluators' comments, focus groups of physicians and opinions of academic, industrial and associative partners, we propose iconic applications related to patient records, for example summarizing patient conditions, searching for specific clinical documents and helping to code structured data. Other applications involve the presentation of clinical practice guidelines and improving the interface of medical search engines. These new applications could use the same iconic language that was designed for drug knowledge, with a few additional items that respect the logic of the language.

Can F-MTI semantic-mined drug codes be used for adverse drug events detection when no CPOE is available?

BACKGROUND: Adverse Drug Events (ADEs) endanger the patients. Their detection and prevention is essential to improve the patients' safety. In the absence of computerized physician order entry (CPOE), discharge summaries are the only source of information about the drugs prescribed during a hospitalization. The French Multierminology Indexer (F-MTI) can help to extract drug-related information from those records. METHODS: In first and second validation steps, the performance of the F-MTI tool is evaluated to extract ICD10 and ATC codes from free-text documents. In third step, potential ADE detection rules are used and the confidences of those rules are compared in several hospitals: using a CPOE vs. using semantic mining of free-text documents, diagnoses and lab results being available in both cases. RESULTS: The F-MTI tool is able to extract ATC codes from documents. Moreover, the evaluation shows coherent and comparable results between the hospitals with CPOEs and the hospital with drugs information extracted from the reports for potential ADE detection. CONCLUSION: semantic mining using F-MTI can help to identify previous cases of potential ADEs in absence of CPOE.

Adverse drug events prevention rules: multi-site evaluation of rules from various sources.

Adverse drug events are a public health issue (98,000 deaths in the USA every year). Some computerized physician order entry (CPOEs) coupled with clinical decision support systems (CDSS) allow to prevent ADEs thanks to decision rules. Those rules can come from many sources: academic knowledge, record reviews, and data mining. Whatever their origin, the rules may induce too numerous alerts of poor accuracy when identically applied in different places. In this work we formalized rules from various sources in XML and enforced their execution on several medical departments to evaluate their local confidence. The article details the process and shows examples of evaluated rules from various sources. Several needs are enlightened to improve confidences: segmentation, contextualization, and evaluation of the rules over time.

Automatic construction of dictionaries, application to product characteristics indexing.

Summary of Product Characteristics (SPC) indexing enables to extract all the information needed to analyze a prescription and find some inappropriate medications. We evaluate a French Multi-Terminology Indexer tool (F-MTI) for SPC automatic indexing. This tool uses a dictionary containing the textual forms that are likely to appear in natural language text for the drug clinical particular terms contained in the Vidal thesaurus (TUV). We developed a method to automatically generate this dictionary. The evaluation showed a precision of 52.9% and a recall of 46.2%. F-MTI will be integrated in a semi-automatic indexing tool.

Multi-terminology indexing for the assignment of MeSH descriptors to medical abstracts in French.

BACKGROUND: To facilitate information retrieval in the biomedical domain, a system for the automatic assignment of Medical Subject Headings to documents curated by an online quality-controlled health gateway was implemented. The French Multi-Terminology Indexer (F-MTI) implements a multiterminology approach using nine main medical terminologies in French and the mappings between them. OBJECTIVE: This paper presents recent efforts to assess the added value of (a) integrating four new terminologies (Orphanet, ATC, drug names, MeSH supplementary concepts) into F-MTI's knowledge sources and (b) performing the automatic indexing on the titles and abstracts (vs. title only) of the online health resources. METHODS: F-MTI was evaluated on a CISMeF corpus comprising 18,161 manually indexed resources. RESULTS: The performance of F-MTI including nine health terminologies on CISMeF resources with Title only was 27.9% precision and 19.7% recall, while the performance on CISMeF resources with Title and Abstract is 14.9 % precision (-13.0%) and 25.9% recall (+6.2%). CONCLUSION: In a few weeks, CISMeF will launch the indexing of resources based on title and abstract, using nine terminologies.

Using multi-terminology indexing for the assignment of MeSH descriptors to health resources in a French online catalogue.

BACKGROUND: To assist with the development of a French online quality-controlled health gateway(CISMeF), an automatic indexing tool assigning MeSH descriptors to medical text in French was created. The French Multi-Terminology Indexer (FMTI) relies on a multi-terminology approach involving four prominent medical terminologies and the mappings between them. OBJECTIVE: In this paper,we compare lemmatization and stemming as methods to process French medical text for indexing. We also evaluate the multi-terminology approach implemented in F-MTI. METHODS: The indexing strategies were assessed on a corpus of 18,814 resources indexed manually. RESULTS: There is little difference in the indexing performance when lemmatization or stemming is used. However, the multi-terminology approach outperforms indexing relying on a single terminology in terms of recall. CONCLUSION: F-MTI will soon be used in the CISMeF production environment and in a Health MultiTerminology Server in French.