Search

International audience; Suitable causal inference in biostatistics can be best achieved by knowledge representation thanks to causal diagrams or directed acyclic graphs. However, necessary and sufficient causes are not easily represented. Since existing ontologies do not fill this gap, we designed OntoBioStat in order to enable covariate selection support based on causal relation representations. OntoBioStat automatic ontological causal diagram construction and inferences are detailed in this study. OntoBioStat inferences are allowed by Semantic Web Rule Language rules and axioms. First, statements made by the users include outcome, exposure, covariate, and causal relation specification. Then, reasoning enable automatic construction using generic instances of Meta_Variable and Necessary_Variable classes. Finally, inferred classes highlighted potential bias such as confounder-like. Ontological causal diagram built with OntoBioStat was compared to a standard causal diagram (without OntoBioStat) in a theoretical study. It was found that confounding and bias were not completely identified by the standard causal diagram, and erroneous covariate sets were provided. Further research is needed in order to make OntoBioStat more usable.

BACKGROUND Word embedding technologies are now used in a wide range of applications. However, no formal evaluation and comparison have been made on models produced by the three most famous implementations (Word2Vec, GloVe and FastText). OBJECTIVE The goal of this study is to compare embedding implementations on a corpus of documents produced in a working context, by health professionals. METHODS Models have been trained on documents coming from the Rouen university hospital. This data is not structured and cover a wide range of documents produced in a clinic (discharge summary, prescriptions ...). Four evaluation tasks have been defined (cosine similarity, odd one, mathematical operations and human formal evaluation) and applied on each model. RESULTS Word2Vec had the highest score for three of the four tasks (mathematical operations, odd one similarity and human validation), particularly regarding the Skip-Gram architecture. CONCLUSIONS Even if this implementation had the best rate, each model has its own qualities and defects, like the training time which is very short for GloVe or morphosyntaxic similarity conservation observed with FastText. Models and test sets produced by this study will be the first publicly available through a graphical interface to help advance French biomedical research.

Background Adverse drug reactions (ADRs) are an important cause of morbidity and mortality. Classical Pharmacovigilance process is limited by underreporting which justifies the current interest in new knowledge sources such as social media. The Adverse Drug Reactions from Patient Reports in Social Media (ADR-PRISM) project aims to extract ADRs reported by patients in these media. We identified 5 major challenges to overcome to operationalize the analysis of patient posts: (1) variable quality of information on social media, (2) guarantee of data privacy, (3) response to pharmacovigilance expert expectations, (4) identification of relevant information within Web pages, and (5) robust and evolutive architecture. Objective This article aims to describe the current state of advancement of the ADR-PRISM project by focusing on the solutions we have chosen to address these 5 major challenges. Methods In this article, we propose methods and describe the advancement of this project on several aspects: (1) a quality driven approach for selecting relevant social media for the extraction of knowledge on potential ADRs, (2) an assessment of ethical issues and French regulation for the analysis of data on social media, (3) an analysis of pharmacovigilance expert requirements when reviewing patient posts on the Internet, (4) an extraction method based on natural language processing, pattern based matching, and selection of relevant medical concepts in reference terminologies, and (5) specifications of a component-based architecture for the monitoring system. Results Considering the 5 major challenges, we (1) selected a set of 21 validated criteria for selecting social media to support the extraction of potential ADRs, (2) proposed solutions to guarantee data privacy of patients posting on Internet, (3) took into account pharmacovigilance expert requirements with use case diagrams and scenarios, (4) built domain-specific knowledge resources embeding a lexicon, morphological rules, context rules, semantic rules, syntactic rules, and post-analysis processing, and (5) proposed a component-based architecture that allows storage of big data and accessibility to third-party applications through Web services. Conclusions We demonstrated the feasibility of implementing a component-based architecture that allows collection of patient posts on the Internet, near real-time processing of those posts including annotation, and storage in big data structures. In the next steps, we will evaluate the posts identified by the system in social media to clarify the interest and relevance of such approach to improve conventional pharmacovigilance processes based on spontaneous reporting.

SIGAPS score determines to a great extent each university hospital's funding for research. This score is considered to reflect the scientific production. Due to its financial impact, he may modify authors' attitudes leading them to publish their articles in French. Our objective was to evaluate the association between the SIGAPS score and publications in French.Based on Rouen University Hospital's (RUH) SIGAPS data from 2007 to 2010, we used Spearman's correlation to establish an association between SIGAPS score of each author and the fact that they published in French.A positive correlation was found between SIGAPS score and author's number of publications in French (r(s)=0.51; IC(95%)=[0.44; 0.58]). The relationship between the SIGAPS score and the author's publication rate in French was negative but higher (r(s)=-0.87; IC(95%)=[-1.0; -0.68]; author's deciles).The relationship between the SIGAPS score and the number of publications in French is not surprising as the SIGAPS score is based on the number of publications. As regards to the publication rate in French, this was even more interesting as our results showed that the more productive an author was, the less they published their results in French. Publications in French did not appear to be the best way to improve individual SIGAPS score. There is high heterogeneity between authors who are very prolific, with one third having a publication rate in French high above the average and one third who had a publication rate in French well below the average.There was a high negative correlation between SIGAPS score and the publication rate in French for RUH's researchers (author's deciles).

ATC classification is a WHO international classification used to classify drugs. The aim of this paper is to evaluate two lexical methods in English and in French to map ATC to UMLS. Several applications have been impemented to illustrate the use of the ATC mapping in English and French: (a) MeSH translation in Norwegian, (b) Drug Information Portal, and (c) ATC to PubMed tool. Two lexical methods were used to map ATC to UMLS. The first approach used a French natural language processing tool to map French terms of ATC to the French terminologies of UMLS. The second approach used the MetaMap tool to map English terms of ATC to UMLS. The English MetaMap provides slightly more mappings than the French NLP tool (3,170 vs. 2,992). On the other hand, the French NLP tool provides a slightly better precision than MetaMap (88% vs. 86%). Using a manual mapping between ATC and MeSH, the union of the validated mappings between ATC and MeSH provides 2,824 mappings (68.7% of ATC codes of the fifth level). Lexical methods are powerful methods to map health terminologies to the UMLS Metathesaurus. Manual mapping is still necessary to complete the mapping.

CCAM is a French terminology for coding clinical procedures. CCAM is a multi-hierarchical structured classification for procedures used in France for reimbursement in health care, which is external to UMLS.The objective of this work is to describe a French lexical approach allowing mapping CCAM procedures to the UMLS Metathesaurus to achieve interoperability to multiple international terminologies. This approach used a preliminary step intended to take only the significant characters used to code CCAM corresponding to anatomical and actions axes.According to the 7,926 CCAM codes used in this study, 5,212 possible matches (exact matching, single to multiple matching, partial matching) are found using the French CCAM to UMLS based mapping, 65% of the corresponding anatomical terms in the CCAM code are mapped to at least one UMLS Concept and 37% of the corresponding action terms in the CCAM code are mapped to at least one UMLS Concept. For all the exact matches found (n=200), 91% were rated by a human expert as narrower than the mapped UMLS Concepts, while only 3% were irrelevant.

SIGAPS is a bibliometric score that aims at making an inventory, evaluating and promoting scientific publications of hospitals that perform research. It has become a major stake in France since it is one of the most important components of the MERRI (Mission Training, Research, Reference and Innovation) founding of hospitals. This score is based on the points attributed to the authors of articles published in journals indexed in Medline, according to the rank of the authors and the Impact Factor of the journal.to compare the reliability of the score when applying different way of computing it, and different weights for the rank or the Impact Factor.we computed the scores of all the physicians of a University Hospital, using the rules that are actually applied at the national level. We then used 4 different scenarios, with different weight given to the rank of authors or the Impact Factor. We compared the scores obtained by each author according to the different scenarios with the Spearman's rank and Pearson's correlation coefficients.The score is not significantly affected when no points are given to the fourth authors and above, when the last author get more points or to change the points according to the Impact Factor of the journal.The different scenarios do not lead to significant changes for the physicians' scores, and therefore for the cumulated score of the hospital. Despite the well known limits of bibliometric indicators, the SIGAPS score appears reliable to compare the hospitals for founding decisions.

The aim of this study is to provide to indexers MeSH terms to be considered as major ones in a list of terms automatically extracted from a document.We propose a method combining symbolic knowledge - the UMLS Metathesaurus and Semantic Network - and statistical knowledge drawn from co-occurrences of terms in the CISMeF database (a French-language quality-controlled health gateway) using data mining measures. The method was tested on CISMeF corpus of 293 resources.There was a proportion of 0.37+/-0.26 major terms in the processed records. The method produced lists of terms with a proportion of terms initially pointed out as major of 0.54+/-0.31.The method we propose reduces the number of terms, which seem not useful for content description of resources, such as "check tags", but retains the most descriptive ones. Discarding these terms is accounted for by: 1) the removal by using semantic knowledge of associations of concepts bearing no real medical significance, 2) the removal by using statistical knowledge of nonstatistically significant associations of terms.This method can assist effectively indexers in their daily work and will be soon applied in the CISMeF system.

French pharmaceutical theses are rarely quoted. If the main obstacles originate from language or access barriers, proper indexation could also be blamed. Manually extracted key-words don't necessary come from a structured thesaurus. In the following work, this manual indexing method is compared to an automated one, "Nomindex", based on UMLS. The automated method is improved by the addition of a relevance scoring system. The first indexing step consists of downloading, adapting and indexing theses in electronic format. Results will then be analyzed and sorted by relevance, through the comparison of classic statistical indices (noise, silence and relevance). It was assumed that the manually obtained key-words were always relevant. The silence of manual indexing is nevertheless high: seven new key-words are proposed by Nomindex, which results are mixed (10% of silence, but 50% of noise). These results are promising on the first experiment on pharmaceutical document without lexicon improvement. The indexing, if it is currently insufficient for a real life use, could easily be improved by specific updates of the lexicon.