MedTech Quality in the Age of Big Data - Are you ready?

Are you Ready?
ASQ Tampa Section Meeting
February 12th, 2018
MedTech Quality in the
Age of Big Data
Naveen Agarwal, Ph.D.
Email: creativeanalytics1@gmail.com
Website: https://www.ExeedQM.com
© Creative Analytics Solutions, LLCInnovative Quality Solutions

© Creative Analytics Solutions, LLC
12 Feb 2018 MedTech Quality in the Age of Big Data Slide 2
“By 2018, 30% Of Interactions With Technology Will Happen Through Conversation.” Gartner
“I think of her as a person…but she is really just a machine”
Meet 81 Year Old Yvonne Meyer ….and Alexa
 Medication Reminders
 Connect with Healthcare Aides
 Connect with Family
 Integrate, Report and Analyze
Home Care Data
Source: CNBC, Orbita

Topics for Today
 Quick Introduction – My Story
 Changing Landscape of MedTech Industry
 MedTech Scope, Industry Performance and Drivers for Change
 Growing Momentum for Innovation
 So What?
 Are you Looking at Your Data? FDA is…
 Increasing Expectations for QA/RA Professionals
 Challenges and Opportunities of Big Data
 Case of Customer Complaints – The Big Data Problem
 Why Data is ”Big” now and Why it Hasn’t Delivered
 Post-market Surveillance Process – Evaluating and Reacting to Complaints
 Looking Ahead and How You Can Prepare
 Traditional Analysis and Tools
 (Brief) Introduction to Artificial Intelligence, Machine Learning and Deep Learning
 Common Machine Learning Algorithms
 Emerging Roles in the Big Data Industry

M.S.. Engineering
Product Development
New Ventures
Business Analytics
Product Quality
Technology Development
My Story……….. Ph.D. Engineering
Product Development
DFSS Black Belt
Innovative Quality
Solutions

The Changing Landscape
of the MedTech Industry
2

MedTech Industry Spans a Lifetime of Care

$25.1 B
MedTech Industry Performed Well in 2016….
Sources: EY 2017 Pulse of the Industry Report
Company annual reports
Revenue
$364.4 B
5%
Net Income
$16 B
17%
Market Cap
$750B
3%
Pure Play Leaders Conglomerates
$29.7 B
$18.3 B
$8.4B
$11.3B
$18 B
$7.7 B
$3.7 B $2.7 B
$2.4 B
$6 B $13.1 B
$15.2 B
$7.7 B

…However Industry Remains in Significant Transition
Portfolio
Optimization
Changing
Business
Models
Technology
Disruption
Regulatory
Changes
 From product-based to
comprehensive care
 Evidence based care, Payer
Partnerships
 EU MDR and IVDR,
 MDSAP
 Cyber-security
 Personalized Solutions
 Wearable Devices
 Robotics, 3D Printing
 M&A, Divestitures
 Digital Deals
 East -> West Capital Flow
Prellis Biolgoics
Orthopedic
implants
Source:EY 2017 Pulse of the Industry Report

Growing Momentum for Innovation at FDA
Source:EY 2017 Pulse of the Industry Report

Source: Abbott Medical

Source: FDA, Edwards Lifesciences
Edwards SAPIEN 3 Trans-catheter Heart Valve

So What?
3

FDA Focus on Data and Surveillance
 223 MM members (2000-2016)
 17 partner institutions
 425 million person-years observation time
 5.9 billion pharmacy dispensing
 7.2 billion unique medical encounters
 42 million people with at least one lab test result
Source: Sentinelinitiative.org, fda.gov/safety/FDASentinalinitiative
Distributed Database and Common Data Model
The active surveillance capability of the Sentinel System does not replace the FDA’s existing surveillance tools, but
complements other FDA safety surveillance capabilities by allowing the FDA to proactively assess the safety of regulated
medical products.

Increasing Expectations for QA/RA Professionals
External
Forces
Internal
Forces
Rapid
Innovation
Regulatory
Changes
Competitive
Intensity
M&A
Integration
Cost
Pressures
Audit
Frequency
Skills
Gap
Digital
Transformation
Design Quality
• Shrinking development cycle, complex technology
• Role changing to enabler of innovation
Product Quality
• Many “customers” – not just the end user
• Role changing to brand/product quality steward
Operations Quality
• Pressure on operating costs and productivity
• Role changing to enabler of product fulfillment
Regulatory Compliance
• Gap to changing standards and requirements
• Role changing to enabler of efficient compliance
Regulatory Affairs
• Rapid regulatory changes across the world
• Role changing to enabler of business strategy
New
Standards

Challenges and Opportunities of
Big Data
4

World as we know it! Big Data World!
Complaint means any written, electronic or oral communication that alleges deficiencies related to the identity, quality, durability, reliability,
safety, effectiveness or performance of a device after it is released for distribution.
§ 820.198(b): Each manufacturer shall review and evaluate all complaints to determine whether an investigation is necessary
§ 820.198(d): Any complaint that represents an event which must be reported to FDA shall be promptly reviewed, evaluated and investigated.
The Case of Customer Complaints – Big Data Problem?

Big Data = Volume, Variety and Velocity
Structured Data Unstructured Data
Employee Data
Sales Data
Survey Data
Lifestyle
Data
Geo Data
Vision Test
Data
Complaints
Data
Search
Data
Social Media
Chatter
Video Data
Voice Data
Image Data
Calls Data

Big Data has Big Potential, but Mixed Record of Success
Weak
Economy
Talent
Org Culture
Technology
Org Culture
Slow to
Change
Source: McKinsey Global Institute Report – The Age of Analytics (2016)

Data Analytics Maturity Model
Operations
Efficiency
Reporting &
Data
Warehousing
Data based
Decision Making
Self-Service
Analytics
Democratization
of Data
New
Business
Models
New Sources of
Revenue
Uses of Data
BusinessValueofData
Limited
Automation of
Data and
Processes
Structured Data,
Reporting and
Visualization
Reporting and
Analytics
Throughout
Organization
Analytics Driving
New Revenue
Growth

25th Jan 2018 Excellence Through Customer Experience Slide 20
Effect of
CAPA
Trigger
Time Series Analysis for Action and Follow Up
Confirm

Detecting Low Frequency Issues
Proportional Reporting Ratio (PRR) assesses frequency of a
product-specific event relative to other similar products
Event (R) All Other
Events
Total
Product (P) A B A+B
Other Similar
Products
C D C+D
Total A+C B+D N=A+B+C+D
Standard Deviation =>
95% Confidence Interval =>
As an example, we can trigger a signal if the lower
bound on PRR exceeds 1
Say, we are tracking the frequency of serious medical events
related to a device with respect to all other medical events
and we find the following data in a given month
MDR All Other
Medical
Total
Product (P) 2 46 48
Other Similar
Products
6 822 828
Total 8 868 876
According to our rule, we will trigger this signal as a
“potential” signal – Should we act?
PRR = 5.75
Lower Bound = 1.54

Other Methods – Text Analytics
Data-mining of verbal feedback from customers about their actual experience, or their conversations on social media can provide
insights into patterns and possibly early warning of an issue
As an example, experience of general discomfort with soft contact lens wear is very hard to quantify and understand. By monitoring
frequency of “key words” associated with this experience, we can better understand shifts in customer experience over time
We can study:
1. Time series of indicators
2. Correlations between indicators
3. Correlation to demographic or
geographic factors
4. Association with specific product lots
or manufacturing timeframe to
indicate potential impact of variation

Case of Complaints – What’s Next?
Cluster Analysis
Anomaly Detection
Classification
Sentiment Analysis
Data Mining
NLP

Looking Ahead and
How You Can Prepare
5

Traditional Statistical Analytics Tools
Descriptive
What happened?
Inferential
Why?
Predictive
What could
happen?
Prescriptive
What should
we do?
Basic reporting
• Summarize past data
• Mean, median, mode,
min, max, variance
• Growth rates
• Compare against
benchmarks or goals
• Data distributions,
process capability
• Charts, graphs, tables to
visualize simple trends
Basic prediction
• Relating sample data to
general population
• Finding statistically
significant factors
• Regression Analysis
• Correlations
• Hypothesis testing
• DOE, ANOVA, GLM
• Multivariate analysis
Forecasting
• Delphi methods
• Trend analysis
extrapolation
• Moving averages, data
smoothing
• Time series, ARIMA
• Regression analysis
Future outcomes
• Data modeling and
simulations
• Sensitivity analysis
• What-ifs and
probabilities
• Decision tree analysis
• DOE/Robust Design and
optimization

Next Frontier - AI > ML > DL
Artificial Intelligence (AI) – enables machines to have human-like intelligence
Machine Learning (ML) – enables machines to “learn” with experience
Deep Learning (DL) – enables machines to build a “brain”

Common ML Algorithms
Logistic
Regression
Decision
Tree
Random
Forest
Neural
Networks
Polynomial
Regression
Linear
Regression
Linear equation
best fit
Adapted for yes/no
Uses sigmoid function
Polynomial equation
best fit
Mapping all outcomes
and probabilities
Multiple decision trees
Random sampling
Interconnected layers of
nodes like a human brain
 Simple calculations
 Easy to train/validate
 Numerical prediction
 Main factors
x No higher order terms
x Model optimization issues
x Feature scaling
x Overfitting issues
 Easy to visualize
relationships
 Easy to train/validate
x Over simplification
x Limited predictability
 “Average” of many
 Fast to train
 Higher quality
output
x Slow to output
predictions
x Difficult to interpret
 Model complex tasks
 High accuracy
 Suitable for deep
learning
x Very slow to train
x Challenging to
optimize
x Nearly impossible
to interpret

Emerging Role in Big Data Industry
Statisticians
Engineers
Analysts
Data Scientists
IT professionals
Chief Executive Officers
Presidents/VPs
Senior Directors
Both Technical and
Business
Management Skills
Source: * McKinsey Global Institute Report – The Age of Analytics, 2016
Functional Experts Senior Business Leadership
2M - 4M
Projected US
demand over
the next 10
years*

All models are wrong, some are useful
George E. P. Box

…He is my friend!

About ExeedTM
Portfolio of Innovative Quality Solutions in 4 Broad Areas
Email: Info@ExeedQM.com
Web: www.ExeedQM.com
Phone: 1-833-MY-EXEED
Customer Experience Regulatory Compliance
Risk Management Quality Culture

MedTech Quality in the Age of Big Data - Are you ready?

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to MedTech Quality in the Age of Big Data - Are you ready?

Similar to MedTech Quality in the Age of Big Data - Are you ready? (20)

Recently uploaded

Recently uploaded (20)

MedTech Quality in the Age of Big Data - Are you ready?