Agile Analytics: The Key to Improving Everything from Surgical Services … · 2019-03-18 ·...
Transcript of Agile Analytics: The Key to Improving Everything from Surgical Services … · 2019-03-18 ·...
Session #9
Agile Analytics: The Key to Improving Everything from Surgical Services to Genomic Personalized Medicine
S. Mark Poler, MDPhysician Informaticist for Enterprise Data Strategy,
Division of Informatics, Geisinger Health System Vice Chairman, Geisinger Health System Anesthesiology
Learning Objectives
• Identify technically and financially advantageous options to replace traditional data systems.
• Illustrate elements of agile healthcare transformation with examples from the surgical suite.
• Define the expansive requirements for non-structured data and genomic-based precision medicine.
• Recognize the role and utility of integrated healthcare information and agile analytics for innovation.
Provision of surgical care is one of the most complex and
expensive of human endeavors.
Geisinger Health believes it must provide its patients with high-quality, safe, and cost-
effective surgical care supported by well-integrated
data and analytics.
Background
Over 50% of health system costs and
revenues are related to surgical
procedures.
More than forty-five million procedures are performed each year.
Variation at least four to five-fold across hospital referral regions.
Nationally: How Big Is the Challenge?
Poll Question #1How effectively do your information systems minimize costs and optimize outcomes (clinical, patient satisfaction, staff engagement, revenue contribution) for surgical operating rooms?
1) Not effective2) Somewhat effective3) Moderately effective4) Very effective5) We do not have such capability6) Unsure or not applicable
Geisinger Health An Integrated Health Service Organization
• Mostly rural and small cities.• PA• NJ: AtlantiCare
• 13 Hospitals.• 3 Level I trauma centers.• 81 community practice sites.• 8 outpatient surgery centers.• 2 nursing homes.• Home health & hospice services covering
25 counties in PA and 3 counties in NJ.• >138K admissions, OBS & SORUs.• 2,663 licensed inpatient beds.
• Multispecialty group.• ~1,300 physician FTEs.• ~790 advanced practitioners.• ~215 primary & specialty clinic sites.• ~3.4 million outpatient visits. • ~520 resident & fellow FTEs.
• Geisinger Commonwealth School of Medicine.
~365 medical students.
• Research: basic, health systems and population health, genomics.
• ~500,000 members.• ~84,000 Medicare Advantage.• ~153,000 Medicaid.
• Diversified products offered on public & private exchanges.
• 45 PA counties.• 5 states.
• ~56,000 contracted providers & facilities.
Moody’s Aa2/Stable Standard & Poor’s AA/Stable
Provider Facilities$3,147M
Physician Practice Group$1,130M
Managed Care Companies$2,395M
>100K procedures/year~45K surgeries with anesthesia/year
www.geisinger.org
Turning PointsTransitional opportunities• Newer technologies.
• Versatility.• Fewer limitations.
• Unstructured data (e.g., natural language processing [NLP]).• Larger scope (search, genomics, data science).
Cost and performance• Storage demand and cost (proprietary vs. commodity).• Modular upgrades (CPU, RAM, storage).
Surgical arena: largest risk and opportunity• Resource management.
• Cost versus revenue.• People, material, and time.
Results
Significant stakeholder
engagement, triggered by visual
presentation of data.
Substantial cost savings realized
through new approach to
analytics.
New analytics capability vastly more efficient, flexible, and
powerful.
Detailed information. Drill-down to
individual cases.
Beyond opinions…changes conversations.
Enterprise data warehouse (EDW)
storage: $2K as opposed to $50K
per terabyte.
Near real time access to data—continually
updated.
Search over 200M clinical notes (two decades) in 5
seconds!
Improvement.
Innovation.
Finance.
Operating room (OR) operations:
Decreased cost variation.Improved time utilization.
How Geisinger Achieved Its Results
12
3
45
Installed next generation hardware and software replacements for EDW:
achieved data integration, vendor independence. Employed analytics in
education, awareness, and engagement.
Changed approach to governance.
Replaced and improved existing analytics capabilities. (Supported and migrated legacy systems.)
Executed systemwide integration of enterprise
information management.
Replace and Improve Existing Analytics CapabilitiesTimeline
EHR Reporting Datbazse
Milestones
CDIS Business
Plan Endorsed
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
CDIS IT Formed
CDIS Go-Live
Research Data BrokerProcedures
Begin
Analytic Direct Access
Begins
Transition to EHR
Reporting Database
360 Degree Patient-Centric
View
Interactive BI Begins
Leadership Dashboard Released
Unstructured Data Analytics
Begin
Analytics Hardware Platform Go-Live
CDIS Needs Assessment
Endorsed
CDIS PMO Established Enterprise
Data Model PoC
EDS Steering Comte
Chartered
Big Data Platform Go-Live
CDIS User Community
• Data at Go-Liveo EHR Clinical Orderso ADT and Schedulingo Physician & Hospital Billingo Clinical Enterprise Costingo Net Revenueo GHP Medical & Rx Claims
• HR System • COPD and Asthma
• Patient Satisfaction
• Cardiology
• Patient Entered Data
• Referrals
• Quality Advisor
• Oncology
Early CDISPlanning & Implementation
EHR Reporting Database
CDIS 1.0 CDIS 1.1 CDIS 2.0 CDIS 3.0
Data Sources
Advantages of HadoopAn Open-sourced Late-binding Software Framework
Less expensive due to
commodity hardware.
Frequent and faster data ingestion.
Supports multiple views
of the data.
Accepts structured and unstructured
data.
Able to use open source and proprietary
analytic tools.
Commodity hardware
allows easy expansion by nodes.
Data integration:de-siloed
“data lake.”
The Role of Analytics
Integration of diverse data from incompatible sources to support analysis across clinical and business domains.
Key Design Criteria. • Larger capacity, faster, legacy (SQL), and NoSQL.
• Integrate diverse data types, including NLP.
• Support transparency, access by operational users.
• Be vendor independent.
• Support rapid and massive expansion including genomic data. • Full exomes linked to 20 years of electronic health records.
Key Design Criteria:
Production Footprint Traditional vs. Open Source vs. Hybrid
• Original cost.• Replacement cost.• Implementation cost.• Incremental nodes.• Incremental storage.• Cloud storage.
Consider and Compare:
Major Data Source Comparison
• EHR clinical data.• Financial.• Claims. • Pulmonary.• Pathology.• Oncology.
All EDW sources, plus:• EHR data x10.• Cardiology.• RIS (Radiology).• Lab and Microbiology.• Supply chain.• Phone systems.• Mobile message delivery logs.• IT, e.g., wireless AP IDs.• PA Health Info Exchange.• Text analytics.• And more…
TRADITIONAL EDW HADOOP
Big Data Architecture
Cross platform
SQL query tool
Data visualization tool
Open source enterprise
search platform
Data system 1
Data system 2
Data system 3
Data system 4 Other
InternalHIE
Rapid Storage and Hadoop Growth
0
5000
10000
15000
20000
25000
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
TeraBytes
Year
Storage Growth2006 - 2018
Hadoop System Growth
Genomic(Projection,
Cloud)
Non-Genomic(local, SSD)
Nodes & RAM (TB) Storage (TB)
2015 2016 2017Nodes 8 23 70RAM (TB) 2 9.5 31.25
01020304050607080
Year
2015 2016 2017Storage (TB) 100 300 1100
0200400600800
10001200
Year
Poll Question #2
How is your organization managing genomic data for personalized medicine?
a) We have not started or we are just starting to think about itb) We have started implementationc) We have a small-scale implementation (terabytes)d) We have a large-scale implementation (petabytes)e) We had no idea about the magnitude of requirementsf) Unsure or not applicable
Enterprise Information Management and Governance
• Enable agility, tracking, and flexibility to support rapid improvement and innovation.
• Process.
• Democratize data (hub and spokes).
• Use existing data better!
Data Management
BI & Analytics
Data Governance
Education
Clinical Innovation
GHP*
Admin/HRWorkforce
Research*
Finance
Spoke Project List
Spoke Project List
Spoke Project List
Spoke Project List
Spoke Project List
Master Project List
* EDS Guidance
SCRUM 1
SCRUM 3
SCRUM 2
SCRUM 1
QualitySpoke Project List
Spoke Project List
Spoke Project List
Market Strategy& DevelopmentSpoke Project List
Analytics Center of Excellence (CoE) FrameworkSupporting Process Improvement
CE
GHP
ClinicalOperations(Institutes)
Education, Awareness, and EngagementIn This Context
Engage all interested parties and embrace grassroots approaches.• Show data in unfamiliar ways; open new vistas.• Reveal formerly veiled concepts.
• Make tools and data freely available.• Invite exploration.• Expect insights.
“Given enough eyeballs, all bugs are shallow.”
Source: Raymond, E. (1997). The Cathedral and the Bazaar. Retrieved from https://archive.org/details/CathedralAndTheBazaar
Example: OR Engagement
New analytic visualizations were
conceived, designed, and
implemented over ~ four months.
Initial presentations to physicians,
administrators, and other staff.
As new analytics capabilities were put
into place, those involved in surgical
services were educated about the
approach and concept and
feedback was sought.
This approach proved to be a
powerful engagement
process across disciplines.
Big Picture Surgical Overview to Details
Video-“RetrospectOR”
Implementation of OR Dashboards—1 Month
Old School
vs. New
School
Griffin
At a restaurant, you know how much your meal cost, why not procedures?
Video-“Check Please!”
A small, talented team can be very
agile.
Synthesis of the How
Enterprise Data Management needs to be engaged in support and transition.
Innovative data management and
analytics are necessary.
New tools and techniques require new
work paradigms.
Lessons Learned
Hadoop isversatile.
Pictures are worth more than
1000 words! Aha! Generic
hardware and software can
replace proprietary.
Integrated data enables new
analytic insights. Daily huddles
expose issues not captured in standard data.
• Structured SQL is accommodated —unstructured data as well.
• Traditional skills are necessary, plus new skills.
• Illustrate experiences staff and leadership can relate to, like timelines and cost variance.
• Proactively report whatever needs improvement.
• Costs about one-tenth the anticipated proprietary next-generation.
• Equivalent or better performance, with some translational challenges.
• Easier incremental upgrades (CPU and storage).
A Cycle of Innovation for Viability
12
3
45
Innovate: do differently.
Financial margin: despite decreased reimbursement.
Financial viability: opportunity to repeat.
Creativity: new approaches enabled by innovation.
Do it again!
Future Plans for Surgical Services
• Replace opinion with data.• Open access to data.• Engagement of insights,
bottom-up and top-down.• Continued process improvements.• Staff engagement, feedback.• Continue integration of data sources.• Real-time location system (RTLS)
integration.
• Provide immediate post-procedure feedback to clinicians:• Usage of supplies, anesthetics,
pharmaceuticals.• Improved data quality and scope.
• Incorporate clinical outcomes into assessment of value.
• Expand to other clinical domains.
Next steps? Where do we want to be in a year or two?
Thank You