Transcript ICPE_25Aug2015_Social Media Medication Discussions

What are People
Talking About?
Medication
Discussions on
Social Media
ICPE August 25 2015
Michele Thomas
Authors
– Michele Thomas, MBA2, Bradley Turnbull, MS3, Jeffery L
Painter, BS, JD1, Rachael L DiSantostefano, PhD1 and Greg
Powell, PharmD, MBA1.
1GlaxoSmithKline,
RTP, NC, United States; 2Blue Zone Industries, Inc.,
Chester Springs, PA, United States and 3Department of Statistics, North
Carolina State University, Raleigh, NC, United States
Abstract submitted to
International
Conference on
Pharmacoepidemiology
(ICPE)– January 2015
2
Conflict of Interest Disclosure
•The
project was funded from the following sources:
–GSK
•The following personal or financial relationships relevant to
this presentation existed during the past 12 months/during
the conduct of the study:
–Employment by commercial entity
–Consultancy with GSK
Overview for Today
–
–
–
–
Background and Research Objective
Methodology
Results
Brief look at clustering used to rapidly identify themes in
1000’s of social media posts
– Conclusion
– Q&A
Background: Approach & Objective
Contextualization of Real-World Drug Use Through Social Listening
– f
Publically available
data is
collected from multiple
sources (e.g. Facebook,
Twitter, Internet forums)
System filters out noise,
highlights important
information, and adds
supplemental data (e.g.
MedDRA coding, scoring etc.)
Users can interactively
explore post information
using an easy to use
interface that helps facilitate
the review process
OBJECTIVE: Describe discussion topics in social media among individuals communicating
about 15 GSK medicines on Twitter and Facebook over 1 year period.
5
Data to Support Research Objective
Figure 1. Consort Diagram
Original posts
(n=212,635)
Post-processed posts
(N=81,300)
Proto-AEs
Mentions
Proto-AE 100% sample
(n=8,202)
Mentions sample from 73,098
(n=8,361)
English only
(n=8,022)
English only
(n=7,852)
Correct medication
(n=7,916)
Correct medication
(n=7,252)
Manual adjudication
Adjudicated as Proto-AE
(n=4,447)
Adjudicated as Mention
(n=10,712)
Results
– What are patients talking about in social media regarding
medications?
Twitter Medication Discussions
Themes (Patient)
Dosing-Related
6%
Product Complaints
12%
Medication Benefits
35%
Lack of Efficacy
47%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
7
Note: 45% of ‘Mentions’ were posted by patients; 43% had indication; 71% of Twitter clusters had notable time-based peaks .
Post Data & Clustering
Cluster Size
Additional
Indication Info
Product
Text
Domain
Cluster
Drug A
Drug B
NA
NA
Cluster 1
Cluster 1
33
33
Y
Y
Drug C
Drug C
have to Someone this stupid Drug A on my nose cause im sick ... t(-_-) f - this [link]
using Drug B cause my teeth are f****** stupid and hurt
Can't believe I have another stupid headache! This is like 3 in just over a week. The
Drug C didn't help either.... [link]
There is a Drug Y pill at home with my name all over it... ~
NA
NA
Cluster 1
Cluster 2
33
237
Y
N
I
B
Drug C
Drug C
Drug C
Drug C
Drug C
Drug D
Drug D
Drug D
Drug D
Drug D
Drug D
Drug A
Drug A
Drug A
Drug E
Drug E
Drug E
Drug F
@user I have, not working :( I need to go back to drs for some more Drug Y
@user I have Drug Y but it sometimes doesn't work
This Drug C doesn't work. I still have migraine
This Drug C isn't doing s*** for my migraine
Drug C doing s*** for this migraine. Please let me not have to get up in 5 hrs...
Drug D always has my back
Someone get my Drug D
I need a mf Drug D
That's given me f****** heartburn. Drug D please
I need some Drug D for this heartburn
I really need a Drug D right now. Heartburn.
Even with a f****** Drug A across my nose I can't breathe
Got this Drug A on my nose cause I can't breathe _ _
@user I got a Drug A_
This Drug E has got me all f***** up
I got a Drug E on and its not doing s***
I need a Drug E
RT @user: :/ RT @user: I need some Drug F
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Cluster 2
Cluster 2
Cluster 3
Cluster 3
Cluster 3
Cluster 4
Cluster 4
Cluster 4
Cluster 5
Cluster 5
Cluster 5
Cluster 6
Cluster 6
Cluster 6
Cluster 7
Cluster 7
Cluster 7
Cluster 8
237
237
634
634
634
530
530
530
155
155
155
456
456
456
419
419
419
207
N
N
Y
Y
Y
N
N
N
Y
Y
Y
N
Y
N
N
N
N
N
B
I
I
I
I
B
0
0
0
0
B
0
B
I
0
0
C
I
0
0
Visualization of Clusters
Conclusion
– We examined themes among products and changes over
time for social media medication posts
– More research is needed to understand the utility of theme
analysis for safety surveillance.
10