PSIP - Patient Safety through Intelligent Procedures in medication

 

WORK in PROGRESS (October 2009)

 

STAGE 1: IMPROVE KNOWLEDGE ON ADVERSE EVENTS

The second task of WP1, data extraction generalization, was developed and extraction was performed on 7 700 stays from Lille s hospital and on 15 000 discharge letters.

The authorization for extension of the Danish data to be used for further data mining was obtained from Region H.

Standardisation and automation of data extraction in HIS was performed by Oracle (the WP1 leader).

The second task of WP2, results of data and semantic mining, was continued.

In this WP, Vidal was in charge of defining and choosing techniques for the medical documents semantic mining. It was decided to use F-MTI, the French multi-terminology indexer to extract semantic data from the medical documents. F-MTI is an automatic indexing tool based on a multi-terminology context and developed in another language than English.

To validate the method Rouen s team (CHU Rouen) have submitted to this F-MTI tool a set of discharge summaries from the Denain hospital (CH Denain) as well as 4 000 discharge summaries from internal medicine and geriatrics of the Rouen university hospital (CHU Rouen).

Ideea Advertising (WP2 leader) has designed and improved a software called Expert Explorer used by the experts to review the rules and the hospital stays. A set of rules was elaborated and qualitatively validated by Lille s team (CHRU Lille) in collaboration with Vidal, Rouen university hospital and Denain hospital. 250 validated rules from automatic calculation were placed on the common repository. This repository was completed by about 100 another rules built with the Vidal s database.

To elaborate and validate the rules, at least 20 steps should be checked in details and modulated to avoid abnormalities. Some effects or some class of drug should be redefined. A new classification of the drugs could be generated from this work.

The third task of WP3, final structured description of candidate adverse events, is progressing.

A sample of about 40 normal stays combined with 40 abnormal stays (i.e. stays targeted by a PSIP rule and presenting an effect ) has been collected from the CH Denain and from the Region H hospital.

Two French speaking experts, a pharmacist and a clinician, have reviewed the 85 stays from the CH Denain and two Danish speaking physicians from the Patient Safety Unit in Region H have reviewed the 80 stays from Region H. The review has been supported by a computerised review form. The review process has also been monitored with portable usability labs allowing to record the experts reasoning (think aloud procedure). The analysis of the corresponding verbal protocols is completed, along with the quantitative and qualitative analysis of the reviewers forms.

A second round of onsite observations of the situations at risk corresponding to the ADEs validated by the reviewers have been performed in the Denain hospital. An ergonomist from Kite has participated to all the onsite observation rounds. The description of the work situations under the SHELL formalism is in progress.

Rouen, Lille and Vidal teams validated the interest of the semantic mining extraction in discharge summaries. To that end, several validation phases were realized.

STAGE 2: DEVELOPMENT OF CLINICAL DECISION SUPPORT SYSTEM (CDSS)

Work Package 4 (WP4) Knowledge Engineering constitutes the link between WP3 (Knowledge Elicitation) and WP5 (Design & Development of CDSS Modules), as WP3 delivers elicited and validated knowledge in natural language (e.g. rules, qualitative/quantitative features, statistics, etc) to WP4, which in turn has to select the appropriate knowledge representation scheme and construct a knowledge-based system (KBS) that will constitute the basis for the CDSS for ADEs prevention developed in WP5.

The first deliverable of WP4 (D4.1) entitled 1^st (Intermediary) Knowledge Framework came up this year. It s major outcome is the initial definition of the PSIP Knowledge Base (KB) components, as well as the KBS platform selection for the project (jointly performed with WP5).

The first activity performed in WP4, was the identification and subsequent analysis towards specifications of the knowledge sources that are considered in the project for ADEs prevention. In particular, the sources considered are ADE rules generated from data mining techniques applied in WP2 (Data and Semantic Mining), drug interaction rules provided by the partner VIDAL that are consolidated in a commercial database, semantic mining results originated also from WP2, tacit knowledge of domain experts and human factors analysis applied in the entire PDAC (Prescribing - Dispensing - Administration - Compliance) drug chain, as well as literature sources. For each of the above knowledge sources, an analysis has been performed in terms of the requirements posed from a knowledge engineering viewpoint, to the extent that this was possible subject to the progress made in relevant WPs, i.e. WP2, WP3 (Knowledge Elicitation) and WP11 (Usability Engineering), and the availability of sufficient information.

An initial experimentation and development has been performed, as soon as the KBS was selected and became available. The purpose of this work was to verify the applicability of the PSIP knowledge model in representing in practice knowledge derived from the input sources. To this end, standard terminologies have been imported in the system and a small number of representative example rules have been implemented.

Work Package 5 (WP5) Design and development of CDSS modules aims at developing clinical decision support systems (CDSS) for ADEs prevention. WP3 delivers elicited and validated knowledge in natural language (e.g. rules, qualitative/quantitative features, statistics, etc) to WP4, which in turn has to select the appropriate knowledge representation scheme and constructs a knowledge-based system (KBS) that will constitute the basis for the CDSS for ADEs prevention developed in WP5.

According to the specifications of WP4, the Knowledge Based System (KBS) has been chosen to implement the knowledge rules and terminology designed in WP2 and 3. The suitability of the selected system (GASTON) to PSIP has been verified and its use has been decided.

The second task Implementation of rules in a first set of CDSS modules has been achieved this year and has been leading to the writing of the Deliverable D5.1 First CDSS Module , which aims at developing a first Module in a medical niche to demonstrate the feasibility of the project s approach.

Vidal provided drug information for anticoagulant and non-steroidal anti-inflammatory drugs. These two pharmacological classes represent the identified niches for the first CDSS.

Region H was able to verify some rules in the first CDSS

The feasibility of developing a prototype CDSS based on the framework described by WP4, utilising the knowledge extracted from the data and semantic mining phases (WP2), and delivered in the form of a taxonomy (WP3) was demonstrated.

Work Package 6 (WP6) Validation of the CDSS modules started on September last year with the task Methodology and specifications for the validation of CDSS module . The aim of this workpackage is to validate the accuracy of CDSS Modules developed in WP5.

Testcases have been composed in cooperation with authors at Region H, CHU Rouen, UMIT, and University Hospitals of Innsbruck. Each test case contains a semi-structured presentation of a patient case, describing demographic data, diagnosis, lab values, and prescriptions. It also describes any alert that should come out, and the priority of the alerts. The test cases comprise both simple and complex cases, and also a few normal cases.

STAGE 3: INTEGRATION OF CONTEXTUALISED-CDSS (Cx-CDSS) IN EXISTING IT SOLUTIONS AND USAGE

The objectives are to develop prototypes, to develop an integration platform of the CDSS modules and to create User-Centred-Design interfaces for Contextualised-CDSSs.

The specification and realisation of the first releases of the Connectivity Platform (CP) were done by Oracle.

IBM and Medasys carried out the technical specifications to allow the communication between their Computerised Physician Order Entry system (CPOE) and the CP. Propositions have been made for the PSIP platform integration in the IBM s graphical interface.

A Web prototype is currently under specification to support Healthcare professionals by allowing an access to the PSIP Platform in hospitals which do not have a CPOE.

WP10 was launched in April 2009. The state of the art about patient support was performed. Work is currently running to check the usability of Vidal rules for patient prototype.

HUMAN FACTORS ENGINEERING (WP11) AND EVALUATION (WP12)

WP11 has modelled the PSIP machine as a game. Scenarios have been designed and tested in a simulated clinical environment with clinicians and actors as patients.

WP12 wrote the deliverable D12.1 which reviews the available evidence on the impact of CPOE systems and summarizes basic and advanced evaluation methods. Given the ongoing developments in the PSIP projects, and the situation in the specific participating hospitals, an adaptive evaluation strategy was proposed, comprising of modules defined according to the following axis: WHAT: Type of intervention; WHERE: Location of evaluation (hospital); WHICH: Evaluation focus and evaluation questions; HOW: Evaluation methodology

4) MEETINGS:

12-14 January 2009: PSIP meetings: Plenary meeting with report on state-of-affairs, Advisory Board meeting, Project Board meeting, Kickoff meeting for WP7-8-9 in Hall, Austria.

2-3 March 2009: WP7-8-9 meeting in Paris (France).

21-22 April 2009: WP7-8-9 meeting Copenhagen (Denmark).

2-6 June 2009: Workshop PSIP in Bucharest (Romania) organized by Ideea Advertising.

30 August-2 September 2009: Participation to the international congress of the European Federation for Medical Informatics (MIE), Sarajevo (Bosnia-Herzegovina).

A workshop PSIP, organized by R gis Beuscart and the PSIP Consortium, was held on 31^st August 2009 during this meeting.

23 and 26 September 2009: PSIP plenary meeting in Belgirate (Italy).

24 and 25 September 2009: First international PSIP workshop held in Belgirate (Italy) organized by KiteSolutions.

30 September 2009: First international conference on medication safety, from medication error to medication safety held in Copenhagen, Denmark.

4) UPCOMING EVENTS:

12-13 November, 2009: IMIA Working Group on Human Factors Engineering in Health Informatics (HFE 2009) will be holding a symposium in the Sonoma Valley, California.

14-18 November 2009: American Medical Informatics Association (AMIA) 2009 Annual Symposium in San Francisco.

30 Nov 2 Dec 2009: Int. Conf. on Intelligent Systems Design and Applications (ISDA), Special Session on Intelligent Systems Design and Applications in the Health Domain, will be held in Pisa, Italy.

11-13 January 2010, PSIP meetings: Plenary meeting with report on state-of-affairs, Advisory Board meeting, Project Board meeting in Issy les Moulineaux (France).

1-4 June 2010: EFMI STC 2010 Conference in Reykjav k ICELAND, Seamless care - safe care: The challenges of interoperability and patient safety in health care

23-24 June 2010, IT in Health Care: Socio-technical Approaches, Fourth International Conference in Aalborg, Denmark.

12 15 September 2010, IMIA: The 13th International Congress on Medical Informatics will be held in Cape Town, South Africa.

 

5) PUBLICATIONS:

Ammenwerth E., Schnell-Inderst P., Macahn C., Siebert U., The effect of electronic prescribing on medication errors and adverse drug events: a systematic review, JAMIA, 15, 5, 2008.

Stefan J Darmoni, Saoussen Sakji, Suzanne Pereira, Taieb Merabti, Elise Prieur, Michel Joubert, Michel Thirion. Multiple terminologies in an health portal: automatic indexing and information retrieval. AIME"09, 12th Conference on Artificial Intelligence in Medicine, Verona, Italy, July, 2009.

Saoussen SAKJI, Catherine LETORD, Suzanne PEREIRA, Badisse DAHAMNA, Michel JOUBERT, St fan J. DARMONI, Drug Information Portal in Europe: Information Retrieval with Multiple Health Terminologies, In Stud Health Technol Inform. 2009;150:497-501.

Dirieh Dibad, AD; Sakji, S; Prieur, E; Pereira, S; Joubert, M & Darmoni, SJ. Recherche d"information multi-terminologique en contexte : Etude pr liminaire. Risques, technologies de l"information pour les pratiques m dicales : comptes rendus des treizi mes journ es francophones d"informatique m dicale (JFIM2009), Nice, France, Avril, Informatique et sant , Volume 17, Pages 101-112, 2009.

Sakji S, Letord C, Dahamna B, Kergourlay I, Pereira S, Joubert M, Darmoni S.

Automatic indexing in a drug information portal.

In Stud Health Technol Inform. 2009;148:112-22.

Suzanne PEREIRA, Blandine PLAISANTIN, Michelle KORCHIA, Nicolas ROZANES, Elisabeth SERROT, Michel JOUBERT, St fan J. DARMONI, Automatic Construction of Dictionaries, Application to Product Characteristics Indexing, in Stud Health Technol Inform. 2009;150:512-6.

R gis BEUSCART, Peter McNAIR, St fan J. DARMONI, Vassilis KOUTKIAS, Nicos MAGLAVERAS, Marie-Catherine BEUSCART-ZEPHIR, Christian NOHR, and the PSIP PROJECT CONSORTIUM, Patient Safety: Detection and Prevention of Adverse Drug Events, in Stud Health Technol Inform. 2009;150:968-71.

Sylvia PELAYO, Fran oise ANCEAUX, Janine ROGALSKI, Marie-Catherine BEUSCART-ZEPHIR, Organizational vs. Technical Variables: Impact on the Collective Aspects of Healthcare Work Situations, in Stud Health Technol Inform. 2009;150:307-11.

Jean-Fran ois GEHANNO, Ga tan KERDELHU , Saoussen SAKJI, Philippe MASSARI, Michel JOUBERT, St fan J. DARMONI, Relevance of Google-Customized Search Engine vs. CISMeF Quality-Controlled Health Gateway, in Stud Health Technol Inform. 2009;150:312-6.

Catherine Letord, Saoussen Sakji, Suzanne Pereira, Badisse Dahamna, Ivan Kergourlay, Stefan J Darmoni. Recherche d"information multi-terminologique : application un portail d"information sur le m dicament en Europe. Ing nierie et Recherche Biom dicale / BioMedical Engineering and Research, Number 29, Pages 350-356, 2008.

Suzanne Pereira, Aur lie N v ol, Ga tan Kerdelhu , Elisabeth Serrot, Michel Joubert, St fan J. Darmoni. Using multi-terminology indexing for the assignment of MeSH descriptors to health resources in a French online catalogue. AMIA2008.

Nicolas LEROY, Brian BJ RN, Adrian BACEANU, Marie-Catherine BEUSCART-ZEPHIR, Human Factors Methods to Support the Experts Review of Automatically Detected Adverse Drug Events, in Stud Health Technol Inform. 2009;150:542-6.

Emmanuel CHAZARD, Cristian PREDA, B atrice MERLIN, Gr goire FICHEUR, the PSIP consortium, R gis BEUSCART, Data-Mining-Based Detection of Adverse Drug Events, in Stud Health Technol Inform. 2009;150:552-6.

Chazard E, Ficheur G, Merlin B, Genin M, Preda C, Beuscart R., Detection of adverse drug events detection: data agregation and data mining., In Stud Health Technol Inform. 2009;148:75-84.

Chazard E, Ficheur G, Merlin B, Serrot E, Beuscart R. Adverse Drug Events Prevention Rules: Multi-site Evaluation of Rules from Various Sources. Stud Health Technol Inform. 2009;148:102-11.

Băceanu A, Atasiei I, Chazard E, Leroy N., The expert explorer: a tool for hospital data visualization and adverse drug event rules validation. Stud Health Technol Inform. 2009;148:85-94.

Chazard E, Merlin B, Ficheur G, Sarfati JC, Beuscart R. Detection of adverse drug events: proposal of a data model. Stud Health Technol Inform. 2009;148:63-74.

Leroy N, Chazard E, Beuscart R, Beuscart-Zephir MC, The Psip Consortium., Toward automatic detection and prevention of adverse drug events. Stud Health Technol Inform. 2009;143:30-5.

V. Koutkias, G. Stalidis, I. Chouvarda, K. Lazou, V. Kilintzis and N. Maglaveras, A Knowledge Engineering Framework towards Clinical Support for Adverse Drug Event Prevention: The PSIP Approach , in: Detection and Prevention of Adverse Drug Events, R. Beuscart et al. (Eds.), Studies in Health Technology and Informatics, Vol. 148, pp. 134-144, IOS Press, 2009.

V. Koutkias, K. Lazou, V. Kilintzis, R. Beuscart and Nicos Maglaveras, On Intelligent Procedures in Medication for Patient Safety: The PSIP Approach , in Proc. of the Int. Conf. on Intelligent Systems Design and Applications (ISDA), Special Session on Intelligent Systems Design and Applications in the Health Domain, Nov. 30 Dec. 02, Pisa, Italy, 2009 (accepted for presentation).