(174) Establishment and Validation of Automatic Monitoring Module for Drug-Induced Arrhythmia Based on Hospital Information System Data

Publication Number: 505

Background: Drug-induced arrhythmias are rare but potentially life-threatening. This study aims to reduce the harm of drug-induced arrhythmias to patients through the use of Adverse Drug Event Active Surveillance and Assessment System -â…¡ (ADE-ASAS-â…¡) that we developed for post marketing surveillance.

Objectives: To establish a drug-induced arrhythmias module in ADE-ASAS-â…¡ that can quickly screen arrhythmia patient. To explore the relationship between drugs and arrhythmias in the real world data based on unstructured electronic medical records (hospital information system).

Methods: In this study natural language processing (NLP) was uesd for improving the keywords of drug-induced arrhythmias module and optimizing alarm condition.To validate the module, we expanded the sample size of monitored inpatients. Through manual evaluation, we analyzed the gender, age, underlying diseases, clinical manifestations, and related drugs of true positive patients.

Results: 29 keywords are identified for drug-induced arrhythmias module. 2,918 electronic medical records were reviewed in preliminary experiment，the positive predictive value (PPV) was 10.82% and the recall rate (R) was 99.18%. The expanded sample was 33007, according to the determination procedure of positive cases, we finally identified 622 positive patients with a PPV of 7.88%. Among them, 50.32% were over 65 years old; the basic diseases were mainly tumors, circulatory system diseases and infectious diseases. 64 categories of 133 drugs were involved, and the top 3 drugs were quinolones antibacterials (83 cases, 13.34%), calcium channel blockers (65 cases, 10.45%), antifungal drugs (61 cases, 9.81%). Extrasystoles, tachycardia and electrocardiogram QT prolonged were the most important types of drug-induced arrhythmias.

Conclusions: This study showed that the age distribution, drug category and type classification of patients with drug-induced arrhythmia were basically consistent with other relevant research results. The establishment of drug-induced arrhythmias module based on ADE-ASAS-â…¡, gives us a chance to obtain the target electronic medical records efficiently, accurately and quickly, to dig the relationship between drugs and arrhythmia in the real world data and to explore the clinical characteristics, occurrence regularity, risk factors and so on. It is advantageous to the post marketing surveillance and provides the accurate basis for the identification and control of clinical medication risk.