utilizing ‘big data’ in health care research · 2020-03-10 · utilizing ‘big data’ in...

Utilizing ‘Big Data’ in

health care research

Eric D. Peterson, MD,MPH

Executive Director, Duke Clinical Research InstituteMarch 2016

All Rights Reserved, Duke Medicine 2007 sb/Strategy & Innovation Group | 2

How Big is Big Data

• Just in 2010, 4 exabytes (4 x 10 18) of unique data was generated, which was more than in the preceding 5,000 years altogether

• "Every 2 days we now create as much information as we did from the dawn of civilization up to 2003” (Eric Schmidt, Google CEO)

• The amount of new data is now doubling every 13 months-and will soon double every 12 hours according to IBM

• For college students in technical degree, half of what they learn in their first year of study will be outdated by their third year

Brett King, Huff Post Tech; June 4, 2014 [http://www.huffingtonpost.com/brett-king/too-much-content-a-world-_b_809677.html]; Ray Kurzweil, The Law of AcceleratingReturns; March 7, 2001 [http://www.kurzweilai.net/the-law-of-accelerating-returns]; David Russell Schilling, Knowledge Doubling Every 12 Months, Soon to be Every 12 Hours; industry tap into news, April 19th, 2013 [http://www.industrytap.com/knowledge-doubling-every-12-months-soon-to-be-every-12-hours/3950]; http://www.youtube.com/watch?v=9XXaZRHhmxY


Value of Aggregated Medical Data

Database Size Examples

Clinician’s Experience 10-100’s Osler, Harvey

Single Center Database 1,000-10,000’s Duke, Emory

Epidemiological Cohort 10,000’s Framingham, MESA

National Registries 1,000,000’s AHA, ACC, STS

EHR / patient-powered

Registries

10,000,000’s PCORnet, Health

eHeart registry


Advances that are Changing Medicine

mHealth/Apps

Apple Health App InterfaceA. B. C.

D.Figure 1. Data Flow. A) Patient is seen by health provider and consents to study. B) Patient provided with JawboneUp24, Withingswireless scale, iHealth BP cuff, and iPod with Health App. C) The

patient selects each health monitor (s) that they would like to allow for integration into HealthKit where through the Health App, they can organize data under one dashboard. D) The provider places a request in EPIC for patient to share their data; the patient, through Health App selects any Health data they would like to share with

their physician and sends data to MyChart. This data will then be automatically uploaded to the Duke Epic EMR (Maestro Care). Provider is alerted that data is shared upon opening patient’s chart.

Apple Health App InterfaceA. B. C.

D.Figure 1. Data Flow. A) Patient is seen by health provider and consents to study. B) Patient provided with JawboneUp24, Withingswireless scale, iHealth BP cuff, and iPod with Health App. C) The

patient selects each health monitor (s) that they would like to allow for integration into HealthKit where through the Health App, they can organize data under one dashboard. D) The provider places a request in EPIC for patient to share their data; the patient, through Health App selects any Health data they would like to share with

their physician and sends data to MyChart. This data will then be automatically uploaded to the Duke Epic EMR (Maestro Care). Provider is alerted that data is shared upon opening patient’s chart.

Precision

Medicine


The Big Data Age of Wonder!


“Big data is like teenage sex: everyone talks

about it, nobody really knows how to do it,

everyone thinks everyone else is doing it,

so everyone claims they are doing it.”

– Daniel Ariely, Ph.D.

Duke University


THE BIG CHALLENGES

Research remains too complex,

slow, inefficient, and expensive.

Data, data everywhere.

Yet limited insights and

knowledge

And even when we know what

to do…

Adoption is delayed and

incomplete. Care and

outcomes are highly variable.


IT and Healthcare…Promise and Challenges

90% of world’s

data has been

created in the last

2 years.

DATA INTEGRATION ANALYTICS ACTION

But 80% of that data

is unstructured…

and stored in

separate systems.

Big data analytics

is coming to

medicine: Google,

IBM Watson…

Will need to

integrate these

data into practice.

FRAGMENTATIONSCALABILITY DECISION SUPPORT LEARNING SYSTEMS


The Complexity of Big Data in Heath Care


Cross Sectional Studies

Longitudinal Evaluations

CER – Safety Surveillance

Practical Clinical Trials

Translational Discovery

10

Clinical Registry

Clinical Registry

Clinical Registry

Clinical Registry

Clinical Registry

Claims Data

BiomarkerGenetics Samples

LongitudinalOutcomes

Detailed Pharm + Device Info

Device/DrugRCT



Value of Various Forms of Big Data


Prediction Tools for Risk of Death after Stroke (circa 2012)

Based on 900 US hospitals and 1 Million patients

Smith EE. Circulation. 2010;122:1496-1504


The New Era of Precision Science

January 2015


Baseline Study

A longitudinal cohort study to extensively characterize participants at

baseline and serially using a battery of clinical, imaging, psychosocial,

behavioral, socioeconomic, geospatial, physiometric, and molecular tools.

A comprehensive study of human health and the transition to disease


Baseline: Human Health and Transition to Disease


Post Market Safety Surveillance

27 Institutions, 200 experts, 1 CC

> 60 million people

FDA, The Sentinel Initiative

July 2010


SENTINEL BY THE NUMBERS

180M

4B

137

4

Individuals in the Sentinel Distributed Database

Prescription drug dispensings

Assessments of products, conditions, and

product-outcome pairs

FDA drug safety communications


Big Data and the Pragmatic Clinical Trial


Informatics Solutions: Developing EHR-based

Clinical Research Networks

Internal Data

WarehouseResearch

Datamart

Research

Datamart

Data

WarehouseEHRResearch

Datamart

Study specific

Clinical

Research

Network

Research

Site A

Internal Data

Warehouse

Centralized

disease

registry

Research

Site B

Research

Site C

Clinical Study

Database

EHRs can contribute some basic data to CT

database

EHR

EHR


PCORI-Supporting Research

Selby JV et al. Sci Transl Med 2013;5:182fs13

Clinical Data Research

Networks

• $56 million

• 8 networks

• 1 million pts/network

All Rights Reserved, Duke Medicine 2007

Pre-study• Utilize EHR to identify

local

subjects/population

• Feasibility dashboard

• Embed encounter

instructions and site

content into EHR

• Pre-consent & study

specific consent

• Model outcomes

• Assess sites’ use of

EHR to facilitate

research

• Usability of inclusion

and exclusion criteria

• Define & refine cohort

• Cohort’s interaction

profiles with health

system

• Feasibility analysis

• Recruitment plan

Study SetupRecruitment

Study Conduct

• Incorporate screening

criteria into EHR for

• Scheduling subjects

• Contacting subjects

• Recruiting subjects

• Alert provider of patient

eligibility

• EHR Health Portals

• Patient opt in/out for types

of studies

• Trials specific data

capture at care delivery

• Auto-populated CRFs

fields from EHR

• Extract data to facilitate

work of study

coordinator

• Rules, Alerts & Checks

• Data completeness

• Quality compliance

• Hospitalization/AEs

• Event rates

• Patient retention and

education

Opportunities to Leverage the EHR for RCTs

20


IOM, November 27, 2012

Swedish Registry-Trial Hybrids

TASTE Trial: Thrombus-Aspiration in MI


Randomization• Demographics

• Medical Hx

• Procedural data

Auto-populate

Unique pages for trial

U.S. Registry-Trial Hybrid: Safe PCI in Women

Analytic

Database

*Enrichment factors

• age > 65 years

• creatinine > 1.5

• diabetes

• known 3-vessel

coronary artery

disease

• current cerebro-

vascular disease

and/or peripheral

artery disease,

• known ejection

fraction <50%

• current smoker

Study

designPatients with known coronary artery disease

(MI, or CAD or Revasc) + ≥1 “enrichment factor”*

Identified through EHR/direct pt. consenting in clinics and hospitals through

CDRNs/PPRNs (PPRN pts. would need to connect through a CDRN to participate)

Pts. contacted electronically with trial information and eConsent;

treatment assignment will be provided directly to patient

ASA 81 mg QD ASA 325 mg QD

Electronic F/U Q 4 months;

supplemented with EHR/CDM/claims data

Duration: Enrollment over 24 months;

maximum f/u of 30 months

Primary Endpoint: Composite of all-cause mortality,

nonfatal MI, nonfatal stroke

Primary Safety Endpoint: Major bleeding complications23


ADAPTABLE: Streamlining Trial Operations

123 9 30

Call Center

• Contact those patients who

stop following up

Baseline

Data

ADAPTABLE

Patient

615

CMS and private health plans FOLLOW-UP

• Longitudinal health outcomes

Patient Portal

• Info/eConsent

• Randomization

• Medication use

• Patient Reported Outcomes

PCORNet Coordinating Center FOLLOW-UP

• Uses Common Data Model

• Longitudinal health outcomes

National Death Index


Learning to promote the rapid and complete uptake of clinical research

findings into routine practice, leading to improved quality of health care

and outcomes.

25

Using Clinical Data to Transform Care Practice

Bench Patients Populations

First Block:Translation from concept into first

human studies

Second Block:Translation from clinical trials into

practice

21


Using Data, Provider Feedback to

Drive Quality Improvement

Mehta RH, et al AHJ 2007

Co

mp

osit

e A

dh

ere

nce R

ate

s

60%

70%

80%

90%

Q1 '02 Q1 '03 Q2 '04 Q3 '05 Q4 '06

AcuteDischarge

Provider-led (QI) effortscan improve CV care! NRMI, CRUSADE

AHA GWTG

NCDR-ACTION ACS

QI Tools Motivated local champions

Timely, valued feedback

Simple toolsStandardized orders, Chart reminders

Collaborative Teams


Implementation Research

• Site Screening:

- Use National Clinical Registries

- Rapid screening of 100’s of centers

to find “outlier performance”

• Qualitative/Quantitative Research:

- Identify care processes linked

to better outcomes

• Empirical Evaluation and QI

- Formally test using cluster RCTs

- Disseminate what works through system!


Impact of Target Stroke: Care & Outcomes

All Rights Reserved, Duke Medicine 2007

Big Data and Population Health:The final frontier

• Top layer — concentrations of diabetes

patients.

• Next layer down — percentage single

female head of household.

• Below that in purple, another indicator of

economic status.

• The bottom layer maps the county

boundary and streets.

• Vertical green spines — longitude ad

lattitude coordinates of where diabetes

patients live and locations of key social

or commercial institutions, that can be

used to link all of these disparate data

sets together based on shared

geography.

Miranda, Ferranti, Strauss, Neelon, Califf. Health Affairs 2013;32:608-1615


Conclusion: The Future of Big Data

• We are entering a new era where expanded data sources

offer the potential to create learning health systems

• However, progress will not occur without the development

of novel IT and informatics, more efficient research, and

more rapid and effective learning health systems

• We need to harness the emerging data deluge to create

new knowledge and translate this into better care…

utilizing ‘big data’ in health care research · 2020-03-10 · utilizing ‘big data’ in...

Documents