Big Data A Starring Role in Healthcare Information ......Big Data –A Starring Role in Healthcare...

Preview:

Citation preview

Big Data – A Starring Role in

Healthcare Information

Architectures

October 26, 2016

2:00 – 3:00 pm ET

**Audio for this webinar streams through the web. Please make

sure the sound on your computer is turned on and you have

speakers. If you need technical assistance, please contact

ReadyTalk Customer Care: 800.843.9166.

2

Housekeeping Issues

All participants are muted• To ask a question or make a comment, please submit via the chat feature and

we will address as many as possible after the presentations.

Audio and Visual is through www.readytalk.com• If you are experiencing technical difficulties accessing audio through the web,

there will be a dial-in phone number displayed for you to call. In addition, if you

have any challenges joining the conference or need technical assistance, please

contact ReadyTalk Customer Care: 800.843.9166.

Today’s slides will be available for download on our

homepage at www.ehidc.org

3

About eHealth Initiative

Since 2001, eHealth Initiative has been advocating the value of technology and innovation in healthcare through research and education.

eHI convenes its multi-stakeholder members, from across the healthcare ecosystem, to discuss how to transform healthcare through information and technology.

eHI members released The 2020 Roadmap. The primary objective is enable coordinated efforts by the public and private sector to transform healthcare by the year 2020.

3

4

Multi-Stakeholder Leaders in Every Sector of Healthcare

5

The 2020 RoadmapKey Focus Areas in 2016

Interoperability

Privacy & Security

Business & Clinical Motivators

Health IT Policy

Data & Analytics

Innovation

5

6

This webinar was made possible through the

generosity and support of PHEMI!

7

AgendaWelcome and Housekeeping

Introductions:

• David Foster, VP Data Insights and Operations,

Healthwise

Speakers:

• Trent Rosenbloom, MD, Vice Chair for Faculty Affairs,

Associate Professor of Biomedical Informatics with

secondary appointments in Medicine, Pediatrics and

the School of Nursing at Vanderbilt

• Kirk Kirksey, VP & CIO, UT Southwestern Medical

Center

• Dr. Paul Terry, CEO and CTO, PHEMI Systems

Big Data - A Starring Role in Healthcare Information Architectures

Using a Large Clinical Data Network to Support Research

S. Trent Rosenbloom, MD MPHAssociate Professor and Vice Chair of Biomedical Informatics

Associate Professor of Internal Medicine, Pediatrics & Nursing

eHealth Initiative Foundation Panel – October 26, 2016

@TrentRosenbloom

PCORNet

• The National Patient-Centered Clinical Research Network (PCORNet), is an innovative initiative of the Patient-Centered Outcomes Research Institute (PCORI).

• PCORNet is designed to make it faster, easier, and less costly to conduct clinical research than is now possible by harnessing the power of large amounts of health data and patient partnerships.

http://pcornet.org

PCORNet

• PCORnet aims to create a “network of networks” that harnesses the power of large amounts of health information and unique partnerships among patients, clinicians, health systems and others.

• In the process, it seeks to transform the culture of research from one directed by researchers to one driven by the needs of patients and other healthcare stakeholders.

Includes: Coordinating Center, CDRNs, PPRNs

http://pcornet.org

PCORNet CDRN Coverage

PCORNet CDRNs

• “Clinical Data Research Networks are networks that originate in healthcare systems and securely collect health information during the routine course of patient care”

- Phase I: 01/2014 – 09/2015, 11 sites

- Phase II: 10/2015 – 09/2018, 13 sites

http://pcornet.org/clinical-data-research-networks/

PCORNet’s CDRN Goals

• Each CDRN should:– engage at least 1 million patients across 2 or more

health systems

– build infrastructure to share data, build novel

informatics tools, and engage key stakeholders

– Perform comparative effectiveness research and

pragmatic clinical trials

Vanderbilt Medical Center: hospitals, >100 clinics engaging 2 million

patients. Meharry/Metro General Hospital: 100,000 patients

VHAN: 7 health systems, 34+ hospitals, 350+ clinics

engaging >3 million patients

Greenway: 1600 clinics engaging 14 million patients

Carolinas Collaborative with > 6 million patients

The Mid-South CDRN

Patient and stakeholder engagement

Large scale enrollment

Standardized data

De-identified

data sharing & regulatory processes

Capability to implement

clinical trials

Three specific cohorts

populated

Efficient biospecimen

banking

Phase I – Key Milestones

Three Initial MS-CDRN Cohorts

• Healthy Weight Cohort– Initial recruitment goal 10,000 individuals for a study

of weight-related health issues

• High Prevalence Cohort– Initial recruitment goal 10,000 individuals with

coronary heart disease

• Rare Disease Cohort– Initial recruitment goal 400 families affected by sickle

cell disease

PCORNet Common Data Model

http://www.pcornet.org/pcornet-common-data-model/

PCORNet Common Data Model

http://www.pcornet.org/pcornet-common-data-model/

PopMedNet

VURDW

VHANRDW

GreenwayRDW

CDM

Mid-South CDRN PCORNet

1. Queries

and Analytic

Software

Packages

from PCORI

2. CDRN

returns

Counts and

Aggregate

resulting data

CDM CDM

UNCRDW

DukeRDW

HSSC

RDW

CDM

Meharry

RDW

CDM CDM CDM

Data Aggregation CDRN-wide

Vanderbilt’s Data Flow Model

Linkage to TN State Health Data (hospitalizations, birth/death data)

Linkage to Tenncare Data

Linkage to CMS Data ( Virtual Research Data Center, RESDAC, CMMI data)

Linkage to Vanderbilt Health Plan (Aetna) health data (claims and PBM data)

Surescripts

Linkage to VU Home Health Data

Linkage to Nursing Home data

Linkages for “Complete” Data

Site Sites in CDM Production CDM Refresh Rate Patients Encounters Dynamic Rate

VanderbiltVanderbilt University Health System

1/09 - 11/15 Quarterly update 1,458,542 13,631,309 Monthly**

VHANWilliamson Medical Center, Maury Regional Medical Center, West TN Health

12/13 - 4/16 Quarterly update 346,241 968,254 Monthly**

Greenway Health

952 sites 1/10 - 12/15 Bi-Annual Update 10,000,000 110,823,801 NA

UNC at Chapel Hill

UNC Health Care System 1/04 – 1/16 Quarterly update 4,078,704 15,270,648 NA

Duke UniversityDuke University 1/05 – 1/16 Quarterly update 2,062,439 34,172,582 NA

HSSC

Greenville Health System (GHS), MUSC Health (MUSC), Palmetto Health (PH), and Spartanburg Regional Healthcare System (SRHS)

GHS: 1/07 – 12/15PH, SRHS: 1/11 – 3/16

MUSC: 1/07 – 9/15Quarterly update 2,879,835 39,068,523 NA

Meharry Medical Center

Meharry Medical College and Nashville General Hospital

1/04 – 6/16 Quarterly update 189,874 315,518 NA

All data nodes have gone through Cycle 1 data characterization

Data Characterization

Prep-to-Research –

Data characterization of data node

has been completed, but did not

complete an interview with the

Coordinating Center prior to the

end of Cycle 1. Sites will complete

this process in Cycle 2.

•HSSC

•Meharry

Research Ready –

Data characterization of data node

has been completed, site has had

interview with Coordinating Center

to review data, and any updates

have been made and reported

back to coordinating center.

•Duke

•UNC

•VHAN

•VUMC

•Greenway

Step 1

DataMart refreshes planned

Step 2

DataMart team responds to

query package and locks SAS

DataMartStep 3

DRN OC approval

(verification of pre-specified requirements)

Step 4

DRN OC analyzes DC

responses and solicits more

information as needed

Step 5

DC Discussion Forums

• All Mid-South sites will run cycle 2 data characterizations which will include

– Labs results

– Prescribing and Dispensing

– Death

• Cycle 2 data characterization begins November 7th

Data Characterization

Data Source Description Dates

Vanderbilt/VHAN Health Plan (Atena)

Universal Medical/ Dental Plan: Service, ICD, CPT, Admission, DRG Codes and Provider information

January 2014-September 2015

Pharmacy Plan: dispensing data, dose, strength duration, refill info, pharmacy location

TennCare Demographics, cause of death, death, diagnosis, dispensing, encounter, enrollment, and procedure

2000-2014

TDOH Statewide hospital/emergency discharge claims, birth/death certificates, demographics, diagnosis, dispensing, encounter, enrollment, and procedure

2011-2013

North Carolina Blue Cross/Blue Shield

All inpatient and outpatient claims, including pharmacy2008-2015

North Carolina Medicaid DUA in process, not available yet. Will be all inpatient and outpatient claims 2008-2015(requested)

South Carolina Claims Discharge database including inpatient, outpatient surgery, emergency, ambulatory free-standing clinic, home health encounters. Demographic, encounter, diagnosis, procedure, and charges.

2000-2015

ADAPTABLE Pilot Inpatient, Outpatient, and Part D Demographics, encounters, diagnosis, procedures, vital, death, labs, prescribing and dispensing

2011-2013

CMS Data Inpatient and outpatient claims In progress

CMMI* All RIFs, Part D, Master Beneficiary summary files, and crosswalks from BENE ID to HIC and SSN

In progress

Data Linkage Projects

Novel Informatics Tools

• Tools for quickly running queries and analyzing electronic health data

• Tools for identifying and contacting patients

– Email, Text, Phone (> 300K emails at VUMC)

– My Research at Vanderbilt (20K)

• New electronic consent process

• Expanded survey tools for collection of patient reported outcomes (via web/mobile platforms, automated phone, embedded video/audio, etc.)

• Integration of PROMIS measures into REDCAP

• Electronic payment processes for study participation

• Potential integration of patient survey data into the EHR for clinical use

• Expansion of clinical decision support tools

ADAPTABLE Trial

• The Aspirin Dosing: A Patient-centric Trial Assessing Benefits and Long-Term Effectiveness Trial will compare the effectiveness of two aspirin doses widely used to prevent heart attacks and strokes in individuals living with heart disease

• Pragmatic Comparative Effectiveness Trial

• PCORI-funded

• First PCORNet demonstration project

http://theaspirinstudy.org

ADAPTABLE Trial

http://theaspirinstudy.org

Questions?

trent.rosenbloom@vanderbilt.edu

An Abridged Tour of Select Data Use atUT Southwestern Medical Center. Dallas

Kirk KirkseyVP Information Resources and CIO

Adjunct Professor, University of Texas at DallasChancellor’s Fellow for Health Information Technology, UT System

What I Will Cover TodayA Bit about UT Southwestern

Why We Use Clinical Data the Way We Do

How We Use the Information

How We Want to Use the Information

Our Biggest Barriers

UT Southwestern – An Overview

Not a Hospital. An Academic Medical Center

Three strategic missions:Patient careResearchEducation

Two hospitals. 40+ clinics.14K employees. $2.8B operating budget.

Affiliations with Parkland, Dallas Children’s, and Dallas VA.

Clements University Hospital Opened 12/2014

SW Health Resources – partnership with Texas Health Resources.

UTSW Campus

Clements University Hospital

580 employees, $85M operat ing budget. $18m - $23M capital projects per year.

Major customer groups: Cl inical professionals, Cl inical Scientists, Basic Science Faculty and Professionals , Students, Physician Partners, University Staff.

Oversee al l UTSW cl inical systems (ambulatory, inpatient, revenue cycle. Approximately 600 t ier 0,1 systems with 200+ interfaces).

Most Wired. Level 7 HIMSS designation. Winner 2016 Healthcare Informatics Innovator Award. Coordinate/col laborate with cl in ical aff i l iates. All voice, date, wired, wireless and cel lular. Research admins support (cl in ical tr ials mgt system, eIRB). Central ized server/OS support. Data center operat ions. CoLo/DR operat ions Help Desk. Desktop Support. About 350 basic products and services. University/Health System ERP. Enterprise data warehouse.

IR SCOPEUTSW – Information Resources Overview

1951 - Late 1970s – MonolithicOne System Must Do it All1970 - 2000(?)– Best of Breed

Maximum Business Unit Functionality“Make Systems Talk”. The Rise of HL7

2000 - Today(?) – Best of ClusterData Integration Trumps (sometimes) BU functionality

The Pendulum of Healthcare Information Architecture

2001(?) - The Rise of the Commercial Electronic Medical RecordIt’s Here to Stay.

The Commercial Electronic Medical Recordas a Data Aggregator

Collects traditional ancillary data (LIS,RIS, Rx, etc) in real time.

Collects non-structured data (clinical documentation, reports)

Allows patient engagement(more data for analytics)

Controls some data quality through workflow

Patient centric organization

• Can’t collect lots of non-clinical data (e.g. building systems).

• Minimal capture of meta-data

• Research and teaching challenged

• Patient centric organization

• Not engineered (yet) for Big Data (diagnostic images, genetic sequencing)

The good The Not-So-Good

How We Use the Data

Member – Enterprise Data Warehouse and i2b2 Clinical Data Repository

Multi-system (e.g. ERP, LMS) integration

MS based. Dimensional

Interoperability with other CTSA sites

Teaching

Certification program using non-production EMR environment

Teaching patient integration with Learning Mgt System (in development)

Research

IRB Protocols

Clinical Trials – EMR integration with Clinical Trials Mgt System (Velos)

Operations research

Big Data UTSWFirst and Foremost – What is it?

Recognize the Big Data Holy Trinity

Move Big DataStore Big DataCompute Big Data

Dedicated 10G network connects from selected laboratories toUTSW Data Center and Texas Advanced Computing Center (TACC) at UT Austin

Support of HPC clusters (trained IR personnel)

Dedicated local Co-Lo and DMZ area open to specialized departmental computing personnel

Partner with Bio-informatics. Leading formation of Clinical Informatics Program

CTSA Grant Holder with i2b2 research data warehouse (not the EDW or CDR)

Active remote disaster recovery/Co-Lo site

Some Examples

Duplicative Diagnostic Imaging in the Dallas Ft Worth Region

CHF Patient Readmission and Migratory Patterns

Two Factor Prediction Model for CHF Patient Readmissions

Patient Portal and CHF Patient Outcomes

DashboardsKPIsMgtReports

PredictInterveneEvaluateAdjust

PredictPublishGet CitedUpdate Resume

RetroactiveAnalytics

AcademicPredictive

OperationsPredictive

ProcessAnalytics

PredictAuto-Intervene (e.g. self driving car)EvaluateAdjust

Our (the Royal ‘Our’) Analytics Challenge

Interventions:A Thought Experiment

Register and train CHF patients at point of care

At home get-connected support

Train and support family and caregivers

More education for Home Health professionals

Multiple regression – which type of portal interaction contributes most to better outcomes

Purchase equipment???????

Major Barriers (IMHO)Implement Data Governance

data definitionsdata transformation rulescross-hierarchy categories (tags)political willIT operations implications

Understand Analytics Architecture Components First (not products)loads/interfaceswarehousedata modelsrequired outputdata scrubbingknow your data

Training the Analytics Professional – The New Frontierhypothesis based investigationsound statistical testingwritingprocess reengineering

Dr. John Snow

One Last Thought:What Does Big Data Buy Us?

New kinds of data. High res diagnostic images. Genetic sequence. Building systems output.

Much more of the same data.

New presentation/simulation techniques.

Compute intensive modeling.

Reduced sampling error (?).

But…

Google Uses Big Datato Predict the Incidence ofH1N1 Flu.

Wal-Mart Uses Big Datato Predict Pre-HurricaneBuying Patterns (2004)

Thank You.

#eHIWebinarPHEMI Systems Copyright 2016

Big Data’s Starring Role in Next-Generation Healthcare Information Architectures

Dr. Paul Terry

PHEMI Systems

#eHIWebinarPHEMI Systems Copyright 2016

The Data Dilemma

Healthcare Data is Locked in Silos

Share and Protect?

Data Available

analysis, insights,

and new services

Analytics

Applications

#eHIWebinarPHEMI Systems Copyright 2016

How do I share data

for secondary use

without compromising privacy,

security, and governance?

#eHIWebinarPHEMI Systems Copyright 2016

Digital Library

#eHIWebinarPHEMI Systems Copyright 2016

Managing Digital Assets

Index of a Book Library Card

#eHIWebinarPHEMI CONFIDENTIAL

Turning Your Data into Strategic Assets

Collect

Index and catalog for findability

Protect and Share

Privacy, Security & Governance

Transform

Analytics-ready

METADATA

#eHIWebinarPHEMI Systems Copyright 2016

Collect Everything

Gene SequenceX-Ray

EMR Form

Code

Fragment

Database

Web Pages

Documents

Virtual MachineSocial Media

Internet

of Things

Video

Customer Service Call Audio

Emails

#eHIWebinar

Index and Catalog with MetadataKnow what complex data you have and be able to find it

Rate = 72

Derived

Rhythm =

Sinus

Derived

LastName =

Smith

Derived

#eHIWebinar

TransformMake complex or multi-structured data analytics-ready

Semi-structured StructuredEchocardiogram Reader

Simplified Distributed Computing Analytics-Ready

Aorta 39 mm

LA 31 mm

LVd 48 mm

LVs 25 mm

FS 48 %

LVEF 65 %

VS 9 mm

PW 9 mm

Ht 170 cm

Wt 77 Kg

#eHIWebinar

Protect and ShareControl access to protect private and sensitive information

Automatically de-identify, mask, and

process data, then publish to 3rd party

applications or Analytics Tools

Enforce consent and data sharing

agreements

#eHIWebinar

Automatically Support Different Data Views for Different Users

#eHIWebinarPHEMI Systems Copyright 2016

Use Cases

#eHIWebinarPHEMI Systems Copyright 2016

• Early warning system

• Prevent, delay, mitigate

• Quarterly molecular screening

• Grow to 25,000 patients

• 15+ varied data sources

• Integrate “omics” with

• clinical data

• Longitudinal study

• Rich research

“How can we integrate genomics information into regular care

so we can advise clients on better behavior for overall

wellness?” Molecular You Co.

#eHIWebinarPHEMI Systems Copyright 2016

Analytics-Ready Data

How do we offer better service while reducing our call center

costs? Global Insurance Company Uses PHEMI Big Data

• Use Data Science to make data

actionable

• Sentiment analysis

• Natural Language Processing

(NLP)

• Replace ODS

• Publish Datasets

• Active Archive

Claims Voice

Recording

Customer

Survey

Social Media Clickstream Call Center

Records

Email

Self Serve Data

Catalog

Custom

ApplicationsAnalytics

Tools

“How do we offer better customer service while reducing call

center costs?” Global Healthcare Insurer

#eHIWebinarPHEMI Systems Copyright 2016

“How do we find targets in all other cancer

patients?” Province of British Columbia

Picture: cbc.ca

#eHIWebinarPHEMI Systems Copyright 2016

“What are the best intervention strategies that drive better

outcomes and have less strain on the system?”

Diabetes Research

#eHIWebinarPHEMI Systems Copyright 2016

“Is there a connection between autism and the microbiome in

the gut?” Autism Research

#eHIWebinarPHEMI Systems Copyright 2016

PHEMI Central Big Data Warehouse

How do I dynamically manufacture datasets for different users?

Filter, Transform, and

Release

Users

Secure

Research

Environment

Structure

Data Marts

Data Cubes

Tables

Applications

Collect

#eHIWebinarPHEMI Systems Copyright 2016

Big data works with your existing information architectures

#eHIWebinarPHEMI Systems Copyright 2016

Big data is here now. It is here to stay.

pterry@phemi.com

#eHIWebinar

64

Questions and Answers!

Please use the chat feature to ask questions

Today’s slides will be available for download on our

homepage at www.ehidc.org

If you have any questions, please contact Claudia

Ellison, Claudia.Ellison@ehidc.org

65

This webinar was made possible through the

generosity and support of PHEMI!

Slides are available at www.ehidc.org/resources

Recommended