harnessing the web to streamline statistical programming ...harnessing the web to streamline...
TRANSCRIPT
1
Daniel BoisvertTT03
Harnessing the Web to Streamline Statistical Programming Processes
2
• Share high level overview and key features of our applications
• Encourage web development for other Biometrics applications
Purpose
3
12 Applications2 Validated
200+ Users
2 Developers
Biogen Biometrics Web Applications
4
Web Application Architecture
INTRANET
5
• Open source, no licensing:o Written in PHP, JavaScript, Python, and MySQL, which are
available for use with no licensing fees and are backed by large user communities.
• Rapid development: o Many existing modules/procedures which have undergone
rigorous QAo Packages to automate common tasks are available and can be
applied with little configuration (e.g.. PDF generation).• Accessibility, Easy Deployment:
o No separate software installation besides a web browsero Global updates are applied on the server side and no
reinstallation is required.
Why Web?
6
Mock Shells SDTM ADaM TLF Define.xml
Challenges in the Statistical Programming Process
Decentralized Information
High Project Management Time
Tedious Tasks
7
Workflow• Tracks requests through their work cycle
(Programmer to QC to Stats to Med. Writing)
Phoenix• Interface to enter SDTM/ADaM Dataset specifications• Define.xml creation
Venus • Interface to enter Mock TLF Shells
Biogen Web Applications
8
0
50000
100000
150000
200000
250000
300000
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Requests Transactions
Project Management in Statistical Programming
2016 Data Requests 18,269Transactions 250,12613.7 Transactions per Request
9
Workflow
10
Workflow - Project Management
Workflow data is analyzable and reportable
11
Workflow - Quantity, Quality, Duration Metrics
12
0
50000
100000
150000
200000
250000
2012 2013 2014 2015 2016 2017Variables Revisions
Decentralized Information –Dataset Specifications
2016 Data118,818 Variables155,997 Revisions
4,098 Datasets
13
Phoenix –Electronic Dataset Specifications
14
Phoenix – Write Data Once Use Multiple Times
15
Phoenix – Enforce Standards
Granular editing control
16
Tedious Tasks - Mock Table Shell Creation
17
Venus – Paper to Data,Targeted Functionality
18
Venus – Enforce Table Standards
19
20
• Predictive Analyticso Predict Request Assignmento Predict Total Work
• Automationo Tie systems together to reduce manual
intervention• Innovation
o Venus to create tables/listings using only PHP, Python
Future
21
• High level overview of key functionality• Web development presents a opportunity
to improve statistical programming processes with low overhead
Conclusion
22
Questions