-
Calculating a Severity Score of an Adverse Drug Event Using Machine Learning on the FAERS Database
- Author(s):
- Robert P. Schumaker, Michael A. Veronin, Rohit R. Dixit (see profile) , Pooja Dhake, Danya Manson
- Date:
- 2017
- Subject(s):
- Machine learning
- Item Type:
- Abstract
- Tag(s):
- Adverse Drug Events, BioInformatics, machine learning, Pharmacovigilance
- Permanent URL:
- https://doi.org/10.17613/d6c2-jz26
- Abstract:
- An Adverse Drug Event (ADE) is a medical injury that can result from a prescription or over the counter drug that causes an allergic reaction, overdose, reaction with other drugs or is the result of a medication error. Vulnerable populations such as children and the elderly are most susceptible to ADEs. This lack of standardized data has kept FAERS from fulfilling its full potential as a pharmacovigilance tool and its limitations have been the subject of numerous studies. Our motivation is to improve drug safety by creating a new type of pharmacovigilence system that 1. Performs data cleaning and standardization of FAERS data, 2. Computes a drug reaction severity score for each ADE based on the reported indications and coded using a modified Hartwig Severity scale, 3. Models the data to A) empirically identify drug-interaction events and their relative strength of event in specific symptom-related incidents and to B) identify drug-disease event severity for specific indications such as hypertension, stroke and cardiac failure, 4. Computes a predicted severity score for the models using machine learning algorithms 5. Evaluates the accuracy of the predicted severity score versus actual severity on a holdout dataset, and 6. Builds a predictive clinical tool for physicians that can interact with a patient’s EHR and identify adverse reaction potential at the point of prescription. We propose a global data-driven approach with the TylerADE System. This system uses advanced machine-learning techniques to sift through data and uncover potentially unknown drug events. This research has the potential to 1) improve the efficiency of pharmacological research by identifying potentially unknown n-drug events that merit further study; 2) create a risk score of potential medication events that physicians can use in a clinical setting; and 3) improve patient safety.
- Notes:
- Author's list: First author: Rob Schumaker; Second author: Michael Veronin; Third author: Rohit Dixit; Fourth author: Pooja Dhake; Fifth author: Danya Manson http://iima.org/wp/wp-content/uploads/2017/04/Proceedings-IIMA-ICITED-2017.pdf
- Metadata:
- xml
- Published as:
- Conference proceeding
- Status:
- Published
- Last Updated:
- 7 months ago
- License:
- Attribution-NonCommercial
- Share this:
An online community for ASEEES members
pages-from-page-3-1-dixit.pdf