1
Engineering for Connected
Patients in the IoT Era
Anurag Singhal
Director of Engineering
Clinical Ink Inc.
Sep 17, 2015
2
Introductions
Anurag Singhal, MS
Director of Engineering
anurag.singhal@clinicalink.com
@anurag_singhal
www.linkedin.com/in/singhalanurag
3
About Clinical Ink
 Founded in 2007, ~100 Employees
 Winston-Salem, NC, Horsham, PA and Cambridge, MA
 Acquired CentrosHealth, VC Funding 2015
 SureSource & CentrosHealth Platforms
 Various Awards & Recognitions
Society for Clinical
Data Management
2014 Best
Practices
eClinical Company of the
Year
Founders 50 Club
Center for
Entrepreneurs
4
About You!
5
Agenda
• Introduction to CentrosHealth Platform
• MongoDB in Healthcare Environment
• Tackling Data from Wearables & Sensors
• Q&A
6
CentrosHealth Platform
 Clinical Trial Specific Mobile Apps
 For Connected & Engaged Patients
 In Partnership with Novartis
 Intro Video
http://youtu.be/U2fYY0KBtJ4
 Website www.centroshealth.net
7
CentrosHealth Platform
 Surveys, Diaries
 Tasks & Reminders
 Medication Adherence
 Sensor Integration
 Communications with Clinic
 Education Content
 Configured Per Study Needs
8
MongoDB @ Clinical Ink
9
MongoDB within CentrosHealth
 Single Replica Set in AWS
 Single database  Database per Tenant
Self-Service Access
End of Study Exports / Archival
Study Level Software Upgrades
 Separated Database for Personally Identifiable Data
HIPAA Security
10
HIPAA, 21 CFR Part 11, Safe Harbor &
EU Data Privacy
 Does Mongo do HIPAA?
 MongoDB Consulting Helped!
 Recognized need to upgrade
to Enterprise edition
Admin Audit Trails
Stronger Encryption
Point in time recovery
Support is Awesome!
11
Key DB Requirements / MongoDB Fit
Requirement DB Implications
 Store Surveys Document / Name - Value Pairs
 Capture Sensor Data Big Data
 Varied Research Applications Flexible Schema Design
 Global scale clinical trials Data locality
 Startup Low cost & fast to develop
 Critical Data / Enterprise Security, Scalable, Availability
 Audit Trails for Compliance Transactional Documents
Strong Fit Good Fit Worked Out Unmet Requirement
12
Limitations / Workarounds
 Tech Support
Not allowed to directly edit JSON
Offer Data Correction APIs
MongoChef for Queries
 Ad Hoc Reporting for Business Users
Not easy over JSON / NoSQL
Offer ETL APIs
Implement Data Marts
13
Tackling Data from Wearables &
Sensors
CentrosHealth now offers streamlined connectivity for a wide range of
medical devices!
14
Business Case
 Enable clinical trial data capture
at patient’s home
using patient’s phone
 End to End Medical Device Connectivity
 Medical grade health data integrity & security
 Integrate Qualcomm Life’s 2net Platform
An open medical device ecosystem
15
An End to End Business Process Flow
Configure
Study & Kit
Devices
Patients
Wirelessly
Capture
Data
Transport,
Transform &
Store Data
Deliver
Actionable
Insights to
Clinicians
& Pharma
16
Wearables & Biometric Sensors = Big Data
10’s of
Months
100’s of
Clinical
Trials
100’s of
Patients
Millions
of
Measur-
ements
~100s of Billions
of
Measurements
17
Engineering Sensor Data with MongoDB
Device Memory -> Firmware Data
Format
2net SDK Normalized Device Data
CentrosHealth Sensor Logs (Device
Specific)
CentrosHealth Aggregated
Measurement Data (Device Agnostic)
18
Embedded 2net Data as Subdocuments
 Schema within
collection varies by
device!
 Seamlessly embed 3rd
party JSON data as a
subdocument
 Store all data
received!
Note: Data Model shown is similar,
NOT actual
19
Love that Aggregation Framework!
 4GB Test Data
 100M Data Points!
 De-dupe to 33K
Records
 No Indexes
 223 Seconds
 On Macbook Air!
 Two function calls
20
Wrestling with Wearable Activity
Tracker Data
 Activity Tracking device sends Last 15 days of data over
and over every 15 minutes!
 Clinicians needed to monitor all data using our web UI!
 Server side challenges with reporting over sensor data
De-duplication
Data streaming to resolve memory issues
21
Market Waves!
 Successfully Completed a Pilot Clinical Trial
 A Partnership Effort by Clinical Ink, Novartis, PPD, Qualcomm
22
Questions? Comments?
23
Thank You Folks!
BTW, We’re hiring a team of full stack,
server, web and mobile nerds!
Email careers@clinicalink.com

Engineering for Connected Patients in the IoT Era

  • 1.
    1 Engineering for Connected Patientsin the IoT Era Anurag Singhal Director of Engineering Clinical Ink Inc. Sep 17, 2015
  • 2.
    2 Introductions Anurag Singhal, MS Directorof Engineering anurag.singhal@clinicalink.com @anurag_singhal www.linkedin.com/in/singhalanurag
  • 3.
    3 About Clinical Ink Founded in 2007, ~100 Employees  Winston-Salem, NC, Horsham, PA and Cambridge, MA  Acquired CentrosHealth, VC Funding 2015  SureSource & CentrosHealth Platforms  Various Awards & Recognitions Society for Clinical Data Management 2014 Best Practices eClinical Company of the Year Founders 50 Club Center for Entrepreneurs
  • 4.
  • 5.
    5 Agenda • Introduction toCentrosHealth Platform • MongoDB in Healthcare Environment • Tackling Data from Wearables & Sensors • Q&A
  • 6.
    6 CentrosHealth Platform  ClinicalTrial Specific Mobile Apps  For Connected & Engaged Patients  In Partnership with Novartis  Intro Video http://youtu.be/U2fYY0KBtJ4  Website www.centroshealth.net
  • 7.
    7 CentrosHealth Platform  Surveys,Diaries  Tasks & Reminders  Medication Adherence  Sensor Integration  Communications with Clinic  Education Content  Configured Per Study Needs
  • 8.
  • 9.
    9 MongoDB within CentrosHealth Single Replica Set in AWS  Single database  Database per Tenant Self-Service Access End of Study Exports / Archival Study Level Software Upgrades  Separated Database for Personally Identifiable Data HIPAA Security
  • 10.
    10 HIPAA, 21 CFRPart 11, Safe Harbor & EU Data Privacy  Does Mongo do HIPAA?  MongoDB Consulting Helped!  Recognized need to upgrade to Enterprise edition Admin Audit Trails Stronger Encryption Point in time recovery Support is Awesome!
  • 11.
    11 Key DB Requirements/ MongoDB Fit Requirement DB Implications  Store Surveys Document / Name - Value Pairs  Capture Sensor Data Big Data  Varied Research Applications Flexible Schema Design  Global scale clinical trials Data locality  Startup Low cost & fast to develop  Critical Data / Enterprise Security, Scalable, Availability  Audit Trails for Compliance Transactional Documents Strong Fit Good Fit Worked Out Unmet Requirement
  • 12.
    12 Limitations / Workarounds Tech Support Not allowed to directly edit JSON Offer Data Correction APIs MongoChef for Queries  Ad Hoc Reporting for Business Users Not easy over JSON / NoSQL Offer ETL APIs Implement Data Marts
  • 13.
    13 Tackling Data fromWearables & Sensors CentrosHealth now offers streamlined connectivity for a wide range of medical devices!
  • 14.
    14 Business Case  Enableclinical trial data capture at patient’s home using patient’s phone  End to End Medical Device Connectivity  Medical grade health data integrity & security  Integrate Qualcomm Life’s 2net Platform An open medical device ecosystem
  • 15.
    15 An End toEnd Business Process Flow Configure Study & Kit Devices Patients Wirelessly Capture Data Transport, Transform & Store Data Deliver Actionable Insights to Clinicians & Pharma
  • 16.
    16 Wearables & BiometricSensors = Big Data 10’s of Months 100’s of Clinical Trials 100’s of Patients Millions of Measur- ements ~100s of Billions of Measurements
  • 17.
    17 Engineering Sensor Datawith MongoDB Device Memory -> Firmware Data Format 2net SDK Normalized Device Data CentrosHealth Sensor Logs (Device Specific) CentrosHealth Aggregated Measurement Data (Device Agnostic)
  • 18.
    18 Embedded 2net Dataas Subdocuments  Schema within collection varies by device!  Seamlessly embed 3rd party JSON data as a subdocument  Store all data received! Note: Data Model shown is similar, NOT actual
  • 19.
    19 Love that AggregationFramework!  4GB Test Data  100M Data Points!  De-dupe to 33K Records  No Indexes  223 Seconds  On Macbook Air!  Two function calls
  • 20.
    20 Wrestling with WearableActivity Tracker Data  Activity Tracking device sends Last 15 days of data over and over every 15 minutes!  Clinicians needed to monitor all data using our web UI!  Server side challenges with reporting over sensor data De-duplication Data streaming to resolve memory issues
  • 21.
    21 Market Waves!  SuccessfullyCompleted a Pilot Clinical Trial  A Partnership Effort by Clinical Ink, Novartis, PPD, Qualcomm
  • 22.
  • 23.
    23 Thank You Folks! BTW,We’re hiring a team of full stack, server, web and mobile nerds! Email careers@clinicalink.com

Editor's Notes

  • #3 2nd hire at CentrosHealth Acquired this year by ClinicalInk 16 Yrs Oracle / 1 Year MongoDB
  • #5 Raise Hands if you are Using Mongo Community Longer than a year? Enterprise? Healthcare / Life Sciences?
  • #11 How many have worked on HIPAA compliant software?
  • #13 Raise your hands if you will NOT allow your tech support to directly edit your JSON documents to resolve data issues? APIs implement audit trail functionality Higher effort, but better eventual solution than Triggers anyways!
  • #15 How many of you have use a wearable device such as Fitbit, Apple Watch?
  • #17 High volume, mid variety and high velocity big data (Long Term Projection)
  • #18 Intimately Understand Device data SDKs like 2net can greatly simplfy the integration Yet, must do some device specific processing / learning Always store all raw data Aggregated - Data Model and Semantics are very important