Automatic Detection of Adverse Drug Events in Geriatric Care: Study Proposal

Area of Science:

• Geriatric Medicine
• Clinical Pharmacology
• Health Informatics

Background:

• Adverse drug events (ADEs) affect one-third of older inpatients, increasing mortality and healthcare costs.
• Antithrombotic drugs are frequently implicated in ADEs among the elderly population.
• Existing ADE reporting systems have limitations; automated detection using electronic medical records (EMRs) is crucial for proactive monitoring.

Purpose of the Study:

• To develop and validate an automated tool for detecting antithrombotic-related ADEs in older inpatients.
• To assess the incidence of hemorrhages and thromboses linked to antithrombotic drug prescriptions.
• To identify risk factors and propose clinical practice improvements for antithrombotic drug safety.

Main Methods:

• A multicenter, cross-sectional study using EMR data from 2015-2016 from four Swiss hospitals.
• Inclusion of inpatients aged ≥65 years prescribed antithrombotic drugs.
• Development of rule-based and machine learning algorithms to identify hemorrhagic and thromboembolic events from EMR data, followed by manual validation.

Main Results:

• The study will analyze data from 34,522 eligible residents aged ≥65 years.
• Data analysis is scheduled for 2022, with project completion by mid-2023.
• Performance metrics including AUC, F1-score, sensitivity, and specificity will be used for validation.

Conclusions:

• The developed tool aims to enhance safety in antithrombotic drug prescribing, reducing ADEs.
• Findings will inform clinical practice through risk management indicators and professional training.
• This automated detection system, leveraging natural language processing, will be a novel resource for Switzerland.