Life Science Analytics

Baker Tilly refers to Baker Tilly Virchow Krause, LLP,
an independently owned and managed member of Baker Tilly International. © 2012 Baker Tilly Virchow Krause, LLP
Baker Tilly
Management Consulting
Realizing Business Value from Unstructured Data

THE VALUE OF UNSTRUCTURED DATA
ANALYTICS
2

There is a tremendous opportunity to gain a competitive advantage by
analyzing unstructured data
Industries continue to struggle with integrating unstructured analytics into their business models. It is time
consuming to identify all of the relevant data source and technically challenging to consume the data into an
analytics environment, where additional processing needs to occur before the data can be analyzed.
Leveraging a Broad Variety of Data:
Companies must be able to transform and parse data from multiple sources and in multiple formats:
databases, text files, scientific devices, transactions, and even social media postings. End users also need easy,
consistent access to all of this data to create a 360-degree view- of their customers, their products, or their
brand.
The Value of Unstructured Data Analytics
Unlocking the True Potential of Big Data
3

Mapping data sources to use-cases
Business use-cases that can benefit from an analysis of unstructured data include:
• Clinical Trial Development-Analysis: Expedite the analysis of patient diary and Patient Reported Outcome data to reduce
time-to-market (and potentially uncover unanticipated benefits in early stage trials)
• Clinical Trial-PRO Development- Analyzing publicly available discussion forum data can accelerate the development of
Patient Reported Outcome measures and streamline the FDA’s protocol review process
• Active Market Surveillance (Pharmacovigilance): Are patients using and experiencing your product in a manner that is
consistent with your Clinical Trial data?
• Market Intelligence: Understanding how your customers are describing their experiences with specific medications can
inform market positioning and facilitate targeted messaging
• Labelling Claim Expansion: Are there unanticipated applications and benefits that are being articulated by customers
that can be used to inform programmatic expansion of an existing compound?
Data Sources Include: Clinical Research
 PubMed
 www.clinicaltrials.gov
 FDA.gov
Patient Support Sites
 patientslikeme.com
 dailystrength.org
 askapatient.com
Social Media Platforms
 Reddit
 Twitter
 Facebook
 Clinical Trial Data (e.g., Patient Diaries)
 Call Center Notes
 Documents
Internal

The abundance of data provides tremendous opportunity…

And an overwhelming amount of data points:

7
We can help separate meaningful from meaningless

MAPPING DATA SOURCES TO USE CASES
8

Availability of Data
Relevant data is readily available:
9

CASE STUDY: PHARMACOVIGILANCE
10

Case-Study: Pharmacovigilance
Data Source:Reddit
11
234M Unique Users 853,824 Subreddits 11,464 Active Communities
217 Countries 8 Billion Page Views Monthly 13+ minutes spent on
Average

Sample Use-Case: Pharmacovigilance
Data Source: www.reddit.com
Analyzing all of the
post titles can yield
value…
But analyzing the
conversations people are
having, and associated
metadata like post date
and # of comments can
be infinitely more
powerful

The Challenges with Analyzing Externally Sourced Unstructured Data:
• There are thousands of posts, and tens of thousands (and more) comments
• Without technology and a methodical text mining processes, gaining insight would require manual review
and data collection.
The amount of time to mine insights from the data would take on the order of months making
it difficult to impact business decisions
13

• To source the data we wrote a Python script to crawl the site and scrape
the data
• We ran a query on Reddit, using ‘Lipitor’ as the search term and analyzed
the results using Python and Oracle Big Data Discovery
• The following are some visualizations and insights we were able to glean
from the data.
14

15
We are able to view a quick top-line summary of the data set and KPIs:
And a distribution of where posts have been submitted

16
Symptoms that people discuss, buried in the comments section of the posts have been tagged,
aggregated and visualized in a Tag Cloud:
And we can see how the volume of comments about the symptoms has changed over time:

17
We are able to see distribution of comments by location…
And limit our analysis to a geographic location of interest. Our summary data updates automatically based on
this refinement:

18
We can set up alerts that tell us when Pfizer products are mentioned:
And configure the alerts to show us the terms that were used to flag them:

19
Users have complete visibility into the source data and finding key words and phrases is facilitated by
powerful search technology:

Summary
• Robust publicly available unstructured data provides opportunities to inform multiple use-cases, including:
- Pharmacovigilance
- Competitor Analysis
- Market Research
- Expedited Clinical Trial End-Point Development
• For most companies these data points represent a difficult ‘aspirational’ data source for inclusion in Business Processes
• Barriers include:
- Identifying the relevant publicly available data sources
- Technical challenges associated with sourcing the data
- Methodology/Technical approach to generating insights (Text Analytics)
- Integrating insights into Business Processes
• Baker Tilly can help!
Proposed next steps
• Custom Demo Development
 Conduct 1/2 day onsite Discovery working-session
 Define high-value use-case for demo
 Identify 2-3 high value unstructured sources for inclusions in demo
 Develop 4-5 visualizations to demonstrate value and surface insights
20
Summary & Proposed Next Steps
Interested in learning more? Contact Andrew Malinow, PhD

Life Science Analytics

More Related Content

What's hot (20)

Viewers also liked (6)

Similar to Life Science Analytics (20)

Recently uploaded (20)

Life Science Analytics

Editor's Notes