d CISMeF & TIBS, LITIS, Rouen Summary of Product Characteristics - PowerPoint PPT Presentation

Suzanne PEREIRA a,c,d , Blandine PLAISANTIN a , Michelle KORCHIA a , Nicolas ROZANES a,b , Elisabeth SERROT a , Michel JOUBERT c and Stefan J DARMONI d a Vidal, Issy les Moulineaux, France b LINALCO, Paris c LERTIM, Faculté de Médecine, Université de la Méditerranée, Marseille d CISMeF & TIBS, LITIS, Rouen

• Summary of Product Characteristics (SPC) – basis of information for health professionals on how to use the medicinal product safely and effectively – contains a description of a certain medicinal product’s Drug composition Pharmaceutical form and strength properties and the conditions Authorized applications (indications) Adverse reactions attached to its use Cautions 2 Safety regulations

• SPC indexing : extraction of all the information needed to analyze a prescription and find some inappropriate medications Prescription Drugs Therapeutic indications Contraindications TUV* terminology Undesirable effects Precautions for use Drug Data Base of Vidal 3 *Unified Vidal Terminology

• SPC manual indexing : a tedious task provided by pharmacists that need to be helped • Evaluation of the performances of F-MTI (French Multi-terminology indexer) for SPC automatic indexing 4

A Document A corpus of documents (SPC, Discharge summarie, Web site, and others in txt format) Identification of items, paragraphs and sentences + localisations Sentences Corpus of sentences NOOJ + NOOJ + Bag of Words Dictionnary Dictionnary Algo of terms of constituant MeSH, CCAM, SNOMED, TUV et ICD10 terms Agreggation Restriction to one or several terminologies Post-treatment Indexing proposition 5

• Contains pre-validated variants – All the different textual forms of the actual terms : inflected forms, synonyms, inflected forms of synonyms ex : asthma : asthma – asthmatic – asthmatics – DELA Format ex : asthma,asthma,TUV11 asthmatic,asthma,TUV11 asthmatics,asthma,TUV11 7 6 26

• Application to a SPC indexing Identification of the Sentence: It is contraindicated in case of gastric ulcer. different items (*Contraindications*) SPC F-MTI Identification of sentences *Therapeutic indications* Dictionary of terms *Contraindications* Gastric ulcer,Gastric ulcer,TUV11 Indexing ulcères gastriques, Gastric ulcer,TUV11 *Undesirable effects* ulcère de l’estomac, Gastric ulcer,TUV11 *Special warnings and DELA Format precautions for use* NOOJ tool *Overdose* *Effects on ability to drive and use machines* Contraindications indexing for the drug: TUV11 Gastric ulcer *Contraindications* 7 7 28

• Method for automatic construction of dictionaries 1) Integrate the existing variants • Variants contained in the terminology • Variants coming from previous projects (UMLF, MeSH dictionnary of MAIF, lexicons of the VUMeF project) 2) Automatic gathering of new variants • Variants extraction from medical and pharmaceutical corpus (~50,000 documents) • Using syntactic graphs and a word dictionnary Example : « diminution des facteurs de coagulation » 3) Creation of new forms (plurals and singulars) « decrease of coagulation factors » • Using a simple algorithm (terms < 3 words) 7 8 29

• TUV application – 23,800 forms found in other lexical resources – 550 new inflections generated (After precision of 55%) • Creation of 11,975 transducers linked to a dictionnary of 38,219 word forms • Corpus of 50,000 medical and pharmaceutical documents – 4,279 plurals/singulars created (After precision of 92,3%) TUV dictionary contains 40,266 inflection forms and synonyms for 11,965 terms 7 9 30

Automatically Identification of the different items and sentences Code Term Item *Contraindications* TUV11 Gastric ulcer SPC *Contraindications* TUV63 Colitis, ulcerative *Contraindications* TUV1200 Hypersensitivity *Therapeutic indications* TUV200 Child *Precautions for use* TUV101 Diarrhea *Precautions for use* *Contraindications* Indexing *Undesirable effects* *Special warnings and Manually precautions for use* Code Term Type *Overdose* TUV11 Gastric ulcer Contraindication *Effects on ability to drive TUV63 Colitis, ulcerative Contraindication and use machines* TUV1200 Lactose intolerance Precautions for use TUV200 Child less than 15y. Precautions for use TUV101 Diarrhea Precautions for use 5,191 SPCs 10

Categories Precision (%) Recall (%) Indication 48.1 21.7 Contraindication 46.1 23.5 Adverse effect 77.0 59.4 Precaution for use 28.4 49.3 All types of terms 52.9 46.2 The entire document 57.6 43.4 indexing 11

• Advantages and disadvantages of the dictionary of terms method – very quick method of indexing documents (less than one minute for our entire corpus) – inflection forms and synonyms are validated before indexing ( generate a minimum of errors) – the quality of the indexing depends on the coverage of the different forms for the terms of a terminology (our dictionary is actually insufficient) 12

• Enhance the quality of SPC indexing – Integration of TUV indexing rules – A better assignation of term types – XMLization of SPCs • Integration in an helping tool for manual SPC indexing (BIBLIS) 13

• Development of an automatic dictionary construction method • Evaluation of a French Multi-Terminology Indexer to index SPCs with TUV terms • To help drug properties information extraction for drug prescription safety 14

• The authors would like to thank Vidal, CISMeF and LERTIM teams for their help in the study design and result analysis. • J.F. Bergmann and al. Good medical practice for drugs. Definition, guidelines, references, fields of action and applications, Therapie 63 (2008), 4:267-80. • S. Pereira, P. Massari, M. Joubert, E. Serrot, S.J. Darmoni. Exploring Multi- terminology Indexing of Discharge Summaries, Stud Health Technol Inform (2008). • S. Pereira, A. Névéol, G. Kerdelhué, E. Serrot, M. Joubert, S.J. Darmoni. Using multi-terminology indexing for the assignment of MeSH descriptors to health resources in a French online catalogue. AMIA Annu Symp Proc (2008), 586-90. 15

• Contact • suzanne.pereira@vidal.fr 16