FDA and M-CERSI Workshop. Complex data analytics - PowerPoint PPT Presentation

Loading...

PPT – FDA and M-CERSI Workshop. Complex data analytics PowerPoint presentation | free to download - id: 80f5e4-YzdjZ



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

FDA and M-CERSI Workshop. Complex data analytics

Description:

FDA and M-CERSI Workshop. Complex data analytics Henry Skip Francis, M.D. Data Mining Research and Evaluation Team, FDA/CDER September 11, 2015 – PowerPoint PPT presentation

Number of Views:80
Avg rating:3.0/5.0
Slides: 31
Provided by: Franci582
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: FDA and M-CERSI Workshop. Complex data analytics


1
FDA and M-CERSI Workshop. Complex data analytics
  • Henry Skip Francis, M.D.
  • Data Mining Research and Evaluation Team,
    FDA/CDER
  • September 11, 2015

2
PharmacovigilanceIT Tools Should Not Replace Our
Own Eyes or Good Clinical Judgment(Slide
courtesy of June Almenoff, M.D., Ph.D.)
  • Human Judgment is irreplacable

3
Workshop QuestionsFrom a Data mining and
informatics viewpoint
  • Mobile health and social media.
  • Tools and technologies for real time analysis.
  • Methodological opportunities and challenges.
  • Understanding trust and influence in social
    media.
  • Standards and Inter-operatibility

4
Complex Data AnalyticsUnifying Goal
  • Convergence is an approach to problem solving
    that integrates expertise from life sciences with
    physical, mathematical and computational
    sciences, medicine and engineering to form
    comprehensive synthetic frameworks that merge
    areas of knowledge from multiple fields to
    address specific challenges.
  • National Research Council. Convergence
    Facilitating Transdisciplinary Integration of
    Life Sciences, Physical Sciences, Engineering,
    and Beyond. Washington, DC The National
    Academies Press, 2014.

5
Converging data sources
6
FDA Complex Data Analytic methods Components of
data mining program (CRISP_DM model)

7
Workshop QuestionsFrom a Data mining and
informatics viewpoint
  • Mobile health and social media.
  • Tools and technologies for real time analysis.
  • Methodological opportunities and challenges.
  • Understanding trust and influence in social.
    Media.
  • Standards and Inter-operatibility

8
RAPID logos 03/13/2014
RAPID logos 02/20/2014
9
Peramivir use during H1N1 influenza A
epidemicUsing a combined system for convergent
data
  • 14 to 34 million cases
  • 63,00 to 153,000 hospitalizations
  • 7,000 to 14,000 deaths
  • 1371 requests for drug
  • 369 AE reports
  • 344 unique patients
  • 900 adverse events

10
RAPID MOBILE APPS
RAPID CLOUD ENVIRONMENT
Visual Look at RAPID Stakeholders
1
3
  • CHIO Cloud Infrastructure
  • Taha Kass-Hout (CHIO)
  • Jim Milto
  • AWS Contractor Team
  • Security
  • Lewis Watson (CISO)
  • Shawn Porter
  • MedDRA / Data SMEs
  • Sonja Brajovic
  • Roger Goetsch
  • Krishna Chary (E2B)
  • Mitra Roca (HL7)
  • Web Services, Database, Dashboards
  • FDA RAPID Contractor Team

2
FDA COMMUNICATIONS
  • FDA RAPID LEADERSHIP
  • Dr Henry Francis
  • Richard Zhang
  • Bruce Weaver
  • Syed Haider
  • 3500A / MEDWATCH
  • Joseph Tonning
  • Dr Robert Ball
  • MOBILE DEVELOPMENT
  • NIH National Library of Medicine George Thoma,
    Sameer Antani, Stacey Arnesen
  • FDA RAPID Contractor Team Booz Allen Hamilton,
    Program Manager Dinesh Kolla

4
  • Office of Communications
  • Paul Buckman
  • Sherunda Lister
  • Kim Rawlings
  • OIM (Streaming)
  • Josh Lehman
  • Division of Drug Information (DDI)
  • Mary Kremzer
  • Catherine Chew
  • DMEP (Emergency)
  • Mary Beth Roberts

EXISTING ANALYTICAL TOOLS
5
  • ArcGIS (Geolocation from Mobile Devices, Heat
    Maps)
  • Martha OConnor
  • Wayne Gorski
  • Nathan Beck
  • Newland Agbenowosi (RAPID Contractor Team)

Data Processed in Cloud Data is processed and
stored in FDA Cloud. Adhering to FDA Security
and Data best practices. Dashboards provide
leadership, comm staff and QC views into MCM AE
Data.
Perform Analytics Utilize existing tools for
location-based analysis (ArcGIS) and signal
detection (Empirica) of captured MCM AE data.
3
5
Mobile Data Collection Clinicians / Reporters
enter MCM AE, attach digital file (ie, photo) and
app auto captures geolocation (lat/long)
1
2
Data Transferred to FDA Data is submitted from
mobile device and sent over Cellular network or
WiFi to FDAs Cloud Environment
  • Empirica (Signal Detection)
  • Ana Szarfman
  • Marilyn Pitts
  • Michael Johnston

Response Sent to Reporter Within 24 hours a
targeted response is sent via email containing
link to digital file (ie, Podcast) with
additional information
4
11
Workshop QuestionsFrom a Data mining and
informatics viewpoint
  • mobile health and social media
  • Tools and technologies for real time analysis
  • Methodological opportunities and challenges
  • Understanding trust and influence in social media
  • Standards and Inter-operatibility

12
App Flow
13
The cloud-based back end design will support the
ingestion, analysis and visualization of
streaming adverse event data for
pharmacovigillance study.
Visualization Layer
Single Drug Analysis Modules
Multi-Drug Analysis Modules
Main Screen
14
Propylthiouracyl (left) vs. Methimazole (right)
15
3. RAPID Phase II
The proposed RAPID Biosurveillance Platform
includes a cloud-based open source big data
analytic tool to facilitate detection of adverse
event signals in near real-time
Alerts Display Drugs with gt 50 increase in the
number of reports in the last 2 weeks
  • My Projects
  • Project 1
  • Project 2
  • Project 3
  • Create Project

Drug-Event Tracker Drug-Event Tracker Drug-Event Tracker
Drug Change in Number of Reports (daily) Algorithm Ranking Statistic (all data)
Paramivir-H1N1 3 2.0
Avandia-Diabetes 0.17 1.05
MPA-Steroid Injections 1.0 1.3
Zanamivir-H1N1 0.17 1
ALL 10 2.5
Time Series Visualization for Number of Reports
for Drug of Interest Adverse Event Data and
Social Media Data
FAERS
10
Pro-MED-mail
14
FAERS Pro-MED-mail
24
FAERS Pro-MED-mail
-4
Mouseover or use the arrow keys to inspect values
  • Drug-AE Filter
  • All drugs in which 1 or more cases are fatal
  • All drugs in which 50 of the adverse event cases
    are reported in pediatrics
  • Disproportionality Metric Filter
  • ROR
  • PRR
  • GPS
  • IC
  • Corresponding Ranking statistic e.g., 5 per cent
    quantile Q0.05of the posterior distribution

Clicking on a drug in the list above activates
the time series and geographic visualization
Mouseover bubbles to view information on adverse
events reported at different geographic locations
16
RAPID GIS Visualizing
17
Growing volumes of rich and complex data sets
have outstripped basic tools and methods,
resulting in a data analysis gap
Small amounts of data or samples(megabytes to
gigabytes)
Large (or all) Amounts of data (gigabytes to
petabytes)
  • Existing Capabilities
  • Basic/Advanced Analytics
  • Analyze the average number of violations at a
    single manufacturer over time
  • Simulate the impact of a foodborne illness on the
    wider population
  • Future Capabilities
  • Big Data/Cloud Analytics
  • Predict which manufacturers are at greatest risk
    for producing contaminated products based on a
    variety of data sources
  • Analyze relationships between contaminated
    produce and multiple environmental variables

ADVANCED ANALYTICS
CLOUD ANALYTICS
Analytic Complexity
Predictive and real-time analytics capabilities
BASIC ANALYTICS
BIG DATA ANALYTICS
Accurate historical observations
Volume, Variety, Velocity of Data
18
Workshop QuestionsFrom a Data mining and
informatics viewpoint
  • Mobile health and social media.
  • Tools and technologies for real time analysis.
  • Methodological opportunities and challenges.
  • Understanding trust and influence in social
    media.
  • Standards and Inter-operability.

19
Social Media World
20
Mining Social Media for Product safety
  • Description of the problem
  • -Like drinking from a firehose
  • -90 of data stored in the world today has been
    created in the past
  • two years. (Kass-Hout, T. and Alhinnawi, H.
    British Medical Bulletin 2013108 5-24)
  • Solutions to the problem
  • -Informatics, data mining, bi directional
    communication.
  • - Real time adaptation to public health events
  • How will the solution impact FDA?
  • - Develop multidimensional vision of medication
    use and drug safety

21
Whats the Value of Social Media?
  • Responsibility to monitor public discussions
  • Expectation that someone is listening
  • Massive volume of discussion and patient reported
    outcomes of unknown depth (and quality)
  • Unique stream of intelligence that is not
    necessarily captured in other data sources
  • Not intended to supplant traditional
    post-marketing surveillance

22
Vernacular-to-Regulatory Translation
23
Workshop QuestionsFrom a Data mining and
informatics viewpoint
  • Mobile health and social media.
  • Tools and technologies for real time analysis.
  • Methodological opportunities and challenges.
  • Understanding trust and influence in social
    media.
  • Standards and Inter-operability.
  • Where do you do the work?

24
The cloud-based RAPID Bio-surveillance System
will support collaboration between FDA and other
Federal agencies to enhance monitoring emerging
health threats
  • RAPID Biosurveillance System
  • Tier 1 Regulatory Action/Guidance
  • FDA and external partners issue guidance to
    ensure patient safety
  • Tier 2 Data Visualization
  • Dashboards summarizing key information provide
    safety alerts
  • Disproportionality metrics and detailed analyses
    allow FDA and collaborators to understand
    emerging issues
  • Tier 3 Data Management Analytics
  • RAPID data is combined with existing MedWatch
    and Medwatcher reports and data from external
    collaborators
  • Advanced analytics support AE signal detection
  • Tier 4 Processing of Adverse Events
  • Healthcare professionals submit AE data via the
    RAPID mobile app
  • AE data is stored in a data lake to support
    real-time access

medwather
Drug Shortage
25
The cloud-based RAPID Biosurveillance System will
support collaboration between FDA and other
Federal agencies to enhance monitoring of
investigational therapies for Ebola and other
emerging health threats
  • RAPID Biosurveillance System
  • Tier 4 Processing of Adverse Events
  • Healthcare professionals submit AE data via the
    RAPID mobile app
  • AE data is stored in a data lake to support
    real-time access

26
The cloud-based RAPID Biosurveillance System will
support collaboration between FDA and other
Federal agencies to enhance monitoring of
investigational therapies for Ebola and other
emerging health threats
  • RAPID Biosurveillance System
  • Tier 1 Regulatory Action/Guidance
  • FDA and external partners issue guidance to
    ensure patient safety

27
Four points worth discussion
  • Convergence of information sources
  • What solutions will solve convergence issues
  • Each information source has its own message
  • Each information source has its own problems

28
Holistic Drug Safety
29
PharmacovigilanceIT Tools Should Not Replace Our
Own Eyes or Good Clinical Judgment(Slide
courtesy of June Almenoff, M.D., Ph.D.)
  • Distinguishing Humans and non humans

30
Data Mining team
  • Syed Haider
  • Alfred Sorbello
  • Ana Szarfman
  • Joseph Tonning
  • Bruce Weaver
  • Lin Yu (collaborator)
  • Richard Zhang
About PowerShow.com