Project

Back to overview

Automated detection of adverse drug events from older inpatients’ electronic medical records using structured data mining and natural language processing

English title Automated detection of adverse drug events from older inpatients’ electronic medical records using structured data mining and natural language processing
Applicant Csajka Chantal
Number 167381
Funding scheme NRP 74 Smarter Health Care
Research institution Division de Pharmacologie clinique Département de Médecine Université de Lausanne
Institution of higher education University of Lausanne - LA
Main discipline Public Health and Health Services
Start/End 01.09.2017 - 31.08.2022
Approved amount 602'560.00
Show all

All Disciplines (6)

Discipline
Public Health and Health Services
Medical Statistics
Pharmacology, Pharmacy
Clinical Pharmacology
Geriatrics
Information Technology

Keywords (16)

Polypharmacy; Multimorbidity; Medication errors; Patient safety; Inappropriate prescribing; Adverse drug reactions; Clinical decision support system; Adverse drug events; Older inpatients; Multicenter study; Quality of hospital care; Interdisciplinary research; Natural Language Processing; Electronic medical record; Aged 65 and older; Automated adverse drug event reporting system

Lay Summary (French)

Lead
Les personnes âgées sont particulièrement exposées aux effets secondaires médicamenteux. L’élaboration d’outils d’aide à la détection des effets secondaires et de procédures préventives permettra d’optimiser la sécurité médicamenteuse en gériatrie.
Lay summary

Contexte. Les effets secondaires médicamenteux sont observés chez environ un tiers des personnes âgées hospitalisées. Parmi les médicaments les plus à risque figurent les antithrombotiques largement utilisés pour prévenir la thrombose. Des outils d’aide à la détection de problèmes liés aux médicaments existent à l’hôpital, mais peuvent être améliorés afin de mieux identifier les facteurs susceptibles de conduire à des effets secondaires.

Objectifs.  L’étude vise à optimiser des outils d’aide à la détection automatique des effets indésirables médicamenteux en exploitant les informations disponibles dans le dossier patient informatisé. L’objectif est de quantifier le nombre d’hémorragies et de thromboses associées à la prescription d’antithrombotiques, d’identifier les facteurs déclenchants et de proposer des améliorations pour la pratique clinique.

Description du projet. Le projet est conduit par une équipe interdisciplinaire dans cinq hôpitaux de Suisse romande et alémanique. La liste des médicaments antithrombotiques et les facteurs de risque associés sera établie afin d’élaborer des outils de détection. Les informations cliniques et biologiques disponibles sous forme de texte libre ou de données structurées seront ensuite extraites à partir des dossiers médicaux informatisés des patients âgés hospitalisés. Des algorithmes informatisés permettant la reconnaissance et le traitement de ces informations seront élaborés afin d’identifier les événements indésirables et leurs facteurs déclenchants. Finalement, la validité de ces algorithmes ou outils de détection sera vérifiée.

Impact. Ce projet permettra d’introduire des mesures permettant de sécuriser la prescription des antithrombotiques. Les résultats seront implémentés dans la pratique clinque au moyen d’indicateurs d’effets indésirables pour la gestion des risques, de formations aux professionnels de santé ; les outils et des méthodologies développés seront diffusées pour de nouvelles recherches dans ce domaine.

Direct link to Lay Summary Last update: 29.08.2017

Lay Summary (English)

Lead
Elderly people are particularly exposed to adverse drug events. Developing tools to help detecting these adverse effects along with preventive procedures will help optimise medication safety in the elderly hospitalized population
Lay summary

Automatic detection of adverse drug events in the geriatric care

Background. Adverse drug events are observed in approximatively one third of elderly people in hospitals. The medications most at risk include anti-thrombotic drugs, which are widely used in geriatrics to prevent thrombosis. Tools that help detecting medication-related problems already exist in hospitals but are not well developed and could be improved to better identify factors likely to trigger adverse drug events.

Aims. This objective of the study is to optimise tools able to automatically detect undesired effects of medication based on information available in the patient’s electronic files. The objective is to quantify the number of haemorrhages and thromboses associated with prescription of anti-thrombotics, identify the triggering factors and propose improvements for clinical practice.

Description. The project is being conducted by an interdisciplinary team at five hospitals in German and French-speaking Switzerland. First, the list of anti-thrombotic drugs will be established along with their side effects, risk and confounding factors in order to develop detection tools. Second, clinical and biological information available in the form of free text and structured data will be extracted from the electronic medical files of elderly patients. Third, software algorithms will be developed allowing this information to be understood and processed in order to identify adverse events and their triggering factors. Finally, the validity of these algorithms and detection tools will be verified.

Significance. The project will allow the introduction of measures aiming at improving safety when prescribing anti-thrombotic medication. The findings will be implemented in clinical practice by means of indicators of adverse events for risk management, and training for healthcare professionals; the tools and methodologies developed will be disseminated for new research in this field.

Direct link to Lay Summary Last update: 29.08.2017

Responsible applicant and co-applicants

Employees

Project partner

Collaboration

Group / person Country
Types of collaboration
Helsana Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Industry/business/other use-inspired collaboration
Prof Rény HUG Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
- Exchange of personnel
Pr C. Quantin, Department of biostatistics and bioinformatics, Dijon University France (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication

Associated projects

Number Title Start Funding scheme
167381 Automated detection of adverse drug events from older inpatients’ electronic medical records using structured data mining and natural language processing 01.09.2017 NRP 74 Smarter Health Care

Abstract

Research topic Older inpatients are particularly exposed to adverse drug events (ADEs), most of them being preventable. ADEs may induce or worsen frailty, geriatric syndromes, functional and cognitive disability, and lead to loss of autonomy, frequent and longer hospitalization, institutionalization and finally death. They also lower the patients’ satisfaction and quality of life, and increase resource use and costs. Inappropriate prescribing, interactions, and inappropriate medication administration and monitoring are the leading causes of preventable ADEs (pADEs) in older inpatients. Indeed, prescribing is particularly challenging in these patients affected by polypharmacy, multimorbidity, frailty, and age-related alteration in pharmacokinetics and pharmacodynamics that influence drug elimination and response.Many studies have found antithrombotic drugs to be frequently associated with ADEs in older patients. These drugs are widely used in the older population for the prevention and treatment of arterial and venous thromboembolism. They are highly associated with bleeding complications in older patients and with thromboembolic events in case of suboptimal use. The growing use of antithrombotics in the geriatric population and their substantial risk of toxicity and inefficacy have therefore become an important patient safety and public health issue worldwide.Few studies on ADE measurement have been conducted in Switzerland. Nation-wide pharmacovigilance database mainly contains adverse drug reactions which collection is subject to selection and underreporting, and ADE spontaneous reporting in hospitals is still under development and prone to underreporting. More comprehensive automated detection tools of ADEs should be developed to improve ADE reporting and monitoring in Swiss hospitals. Many interventions have been conducted to improve the quality and safety of medication prescribing among which clinical decision support (CDS) tools within computer provider order entry systems. CDS may detect from electronic medical records (EMRs) drug-related problems and provide feedback to health care professionals through alerts and reminders. Compared with voluntary reporting, chart reviews or automated detection of ADEs from hospital discharge data, CDS consume fewer resources and are faster. However, regarding sensitivity and specificity, they may be less competitive than enhanced manual chart reviews such as the “Global Trigger Tool”. New ADE detection and monitoring systems are currently being developed based on multiple sources of data (i.e. structured data and free texts from EMRs) and methods involving structured data mining (SDM) and natural language processing (NLP). Apart from CDS system targeting drug interactions, dosing errors, or prescription orders, no ADE detection and monitoring system based on EMR SDM or NLP is currently available in Switzerland. Nor can “ready-made” systems from other countries be adapted as they were developed for EMRs written in English.Our research hypothesis is that the automated detection of ADEs from EMRs using SDM and NLP could significantly improve risk management and patient safety in hospitalized older inpatients combining multimorbidity, frailty and polypharmacy. It could additionally provide reliable data for health care professionals, patient safety organisations and policy-makers. Objectives The objectives of our project are to:1.Generate a list of relevant antithrombotic drugs (i.e. heparins, vitamin K antagonists, direct thrombin and factor Xa inhibitors, fondaparinux, antiplatelet drugs) with their potential ADEs and pADEs (i.e. thromboembolic and hemorragic adverse events), and confounding factors (i.e. patient characteristics, concomitant health conditions and drugs);2.Develop and validate an electronic application for the automated detection of ADEs related to defined drugs. The application will process data derived from EMRs by means of both SDM and NLP;3.Develop and adapt NLP tools for the specificities of the French and German medical languages;4.Assess the performance in terms of efficacy, reliability, reproducibility and implementability of the ADE automated detection tool; 5.Implement strategies to improve ADE reporting and prevention through knowledge transfer.The overall strategy comprises the development of automated detection tools that will provide quantitative measures and triggers of ADEs associated with antithrombotic drugs and the elaboration of measures for implementation, replication and prevention.Research planThis project is a collaborative, interdisciplinary research between experienced researchers from various disciplines (i.e. geriatric medicine, clinical pharmacology and pharmacy, nursing, medical informatics, data sciences, biostatistics, health services research and linguistics). Five hospitals will participate in this project including two hospitals in the French speaking part of Switzerland (Geneva and Lausanne university hospitals), two in the German speaking part (Zurich university hospital and Baden cantonal hospital), and one in both linguistic regions (Valais hospital). The research project will be divided in five work packages (WP) to answer the four first objectives: WP1 - Drug and ADEs selection; WP2 - Data extraction and management; WP3 - Structured data mining (SDM); WP4 - Free-text and narratives processing (NLP); WP5 - ADE detection tool assessment.Timeframe and milestones The project will last 4 years: January 2017 - December 2020ImplementationThe implementation of the project’s results will include five complementary components:1.Dissemination of the project results to various stakeholders; 2.Diffusion of ADE indicators for hospital risk management optimisation and internal safety monitoring of antithrombotic drug prescriptions; 3.Sensitization and training of healthcare professionals and patients on ADEs caused by antithrombotic drugs; 4.Knowledge transfer of tools and methodologies for automated ADE detection; 5.Development of Federal surveillance and improvement programs for the quality and safety of antithrombotic drug prescription. Several deliverables (DL) targeting different audiences will be produced: DL1 Delphi evidence report, DL2 SDM & NLP algorithms, DL3 ADE indicators, DL4 Delphi questionnaire, DL5 ADE assessment form, DL6 NLP data set, and DL7 NLP pipeline.SignificanceThe present research project should provide a reliable and applicable automated tool to detect ADEs related to antithrombotic drugs in older hospitalized patients using EMRs. This tool could be implemented within EMRs and be completed by e-alerts and reminders notifying providers on probable ADEs and may feature an automated causality assessment facilitating pharmacovigilance reports. Showing that analytical approaches are applicable to free text and narratives will also foster the possibility of using them for numerous other purposes, such as the Swiss Personalized Health Initiative. Finally, our project should contribute to improve patient safety in older inpatients with multimorbidity and polypharmacy and therefore reduce age-related health inequities, as well as hospital use and costs.
-