Adverse Drug Reaction Information for Asthma patients



  • Our aim is to democratize knowledge from medical research articles with help of data science. ADRIA accesses a collection of biomedical research articles in MEDLINE, the National Library of Medicine database.
  • ADRIA takes information from complex but highly comprehensive and authentic medical literature and makes it easy to access for asthma patients.
  • Using a chat format, ADRIA answers basic but important questions about adverse drug reactions, more commonly referred to as drug side effects, related to asthma medication, while bypassing the medical jargon in research articles.

The Bot

Value Proposition

What is the value?

  • Adverse drug reactions are estimated to be the fourth leading cause of death ahead of pulmonary disease, diabetes, AIDS, pneumonia, accidents and automobile deaths.
  • While drug label inserts do contain adverse drug reaction information, most drugs are approved with an average of only 1,500 patients. This relatively small trial group is often not reflective of all demographics.
  • Research articles are the only authentic and comprehensive knowledge source for adverse drug reactions. While doctors can interpret these research articles, patients do not have the training to understand the information.
  • In summary, ADRIA provides easy access to adverse drug reaction information embedded in the complex research articles of MEDLINE.

Why did we choose Asthma?

  • Asthma is a chronic disease that impacts 235 million people worldwide. Asthma rates have risen 15% in the last decade, and it is the most common chronic disease among children.
  • Asthma costs the United States $56 billion each year, and requires 8.9 million doctor visits, 1.9 million emergency room visits, and almost 500,000 hospitalizations each year.
  • Asthma has no cure. This means patients rely on medication for symptom management, frequently for decades.
  • Medications can be expensive, and many patients use them on a daily basis. This creates significant exposure for patients to experience adverse reactions to these drugs, many of which are serious.
  • The high prevalence of asthma, the widespread and long term use of its medication, and high cost of asthma treatment make asthma a meaningful candidate for ADRIA to tackle.

Technical Overview

ADRIA uses machine learning techniques powered by multi-step data processing that applies a medical translation dictionary to answer user questions. Here for more technical discussions

Machine Learning

  • ADRIA uses machine learning techniques, multi-step data processing, medical translation dictionaries and natural language processing algorithms to read over 10,000 scientific abstracts about chemicals and drug ingredients reported to cause adverse reactions in asthma patients. From these abstracts, with the help of medical dictionaries (MeSH and RxNorm), ADRIA learns to identify which chemicals are associated with which medical conditions.
  • The general assumption for the chatbot design is that if a chemical occurs often in an article that also contains a chemically induced adverse drug reaction, and rarely in other articles, then this chemical will be associated with that adverse reaction.
  • In order to provide more accurate answers, ADRIA aggregates the side effects into 53 more general medical categories ranging from gastrointestinal diseases to worsening of asthma.

Data Pipeline

  • The US National Library of Medicine and the National Institutes of Health provides programmatic access to the biomedical literature via its search engine service, PubMed API.
  • Taking advantage of the Medical Subject Headings, we could select the relevant articles for adverse drug reactions related to asthma medications or ingredients that asthma medications contain.
  • From this text, we tagged the chemicals by using the tmChem API, checked which ones were drug ingredients using the RxNorm API and found the corresponding drug brand again via the RxNorm API.


ADRIA is not intended as a substitute for the expertise and judgement of your physician, pharmacist or other healthcare professional. It should not be construed to indicate that the use of any medication in any country is safe, appropriate or effective for you. Consult with your healthcare professional before taking any medication.

Our Vision

Our product has been designed with functional scalability in mind. ADRIA can be extended to support other diseases by incorporating relevant data. The current data processing steps and model can accommodate pubmed data for other diseases. We intend to open source the code for data processing pipeline.

Our Team

Adam Lenart
Alison Walker
Anamika Sinha
Jerry Song

Get in Touch

Email us at with feedback!