Upload
stamos
View
54
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Web-scale pharmacovigilance. Maggie Mahan 16 April 2013. Motivation. Adverse drug events cause morbidity & mortality Typically discovered after drug marketed Increased internet searches of health information (~60% of American adults) - PowerPoint PPT Presentation
Citation preview
Web-scale pharmacovigilanceMaggie Mahan16 April 2013
Motivation Adverse drug events cause morbidity &
mortality Typically discovered after drug marketed
Increased internet searches of health information (~60% of American adults)
Mining web search history to identify unreported side effects of drugs or drug combinations
Logs are inexpensive to collect & mine
Drug safety surveillance
Background (1/2) Drug side effects reported but incomplete and biased
Leads to delayed reporting of adverse events Compounded with multiple drugs
Previous research on tracking seasonal influenza Search logs can be used for health monitoring Health-seeking activity captured in queries to web
search services mirrors trends gathered by traditional surveillance
Background (2/2) Present study used online health-seeking search activity to
identify adverse drug events associated with drug interactions Paroxetine: anti-depressant Pravastatin: cholesterol-lowering drug Interaction reported to create hyperglycemia
Hypothesis: patients taking these two drugs might experience symptoms of hyperglycemia and may have conducted internet searches on these symptoms and concerns related to hyperglycemia before the association was reported
Methods 12 months of search logs
Word used in user queries Pravastatin & brand names Paroxetine & brand names Hyperglycemia-associated words
Disproportionality analysis Assess increased chance of
search for hyperglycemia-related terms given search for both drugs
Reporting ratios based on observed versus expected
Results – user groups & prevalence
Searching both drugs = more likely to search hyperglycemia-associated terms
Difference between groups is consistent
Results – disproportionality analysis
Results - disproportionality analysis for known drug–drug
interactions
Conclusions Log analysis valuable for identifying drug pairs linked to hyperglycemia
Method generalizable, similar to a prediction task
Majority of TP identified provides validation for the set of terms used
Valuable signal even though search logs are unstructured, not necessarily related to health, and include any words entered by users
More in-depth analysis is needed
Patient search behavior directly can complement traditional sources of data for pharmacovigilance
References White RW, Tatonetti NP, Shah NH, Altman RB, Horvitz E (2013)
Web-scale pharmacovigilance: listening to signals from the crowd. J Am Med Infom Assoc. 20(3): 404-408.
http://scopeblog.stanford.edu/2013/03/06/researchers-mine-internet-search-data-to-identify-unreported-side-effects-of-drugs/