shooting the moon: it infrastructure for data-sharing networks · 1 shooting the moon: it...
Post on 20-Jul-2020
2 Views
Preview:
TRANSCRIPT
1
Shooting the Moon: IT Infrastructure for Data-Sharing NetworksSession PM5, February 19, 2017
Jonathan Hirsch, Founder & President, Syapse & Paul Tittel, Systems Director, Providence St. Joseph’s
2
Speaker Introduction
Jonathan Hirsch, MSci
Founder & President, Syapse
3
Conflict of Interest
Employed by and equity in Syapse Inc.
4
Agenda
•Trends in Cancer Care
• Overview and Aims of the Oncology Precision Network (OPeN)
•IT Requirements for Data-Sharing
• OPeN Membership and Traction
• Why Data Sharing is Critical
5
Cancer Care is Entering a New Era
Cancer patients actively seek out care personalized for them
90% of cancer drugs in late phase trials target a
molecular pathway
Real-world evidence will improve outcomes and justify reimbursement
6
Pooled, Real-World Evidence Leads to Better Treatment Decisions
• Aggregating real-world evidence on molecularly-defined cohorts can inform treatment decisions for precision medicines.
• Using molecular data to stratify the populations leads to small samples, limiting our ability to improve care.
• It is critical to pool real-world data across multiple institutions to draw large-scale, statistically powerful treatment insights.
7
IT Solutions Enable Precision Medicine
Understand
Integrated Clinical and
Molecular Data
Decide
Decision Support and
Best Practice Automation
Improve
Clinical Analytics &
Learning Health System
8
What is the Oncology Precision Network (OPeN)?
• OPeN is a trusted network of renowned community health systems and academic medical centers
• Members share aggregated clinical, molecular, treatments and outcomes data
• Access insights from aggregated data to improve cancer care
9
IT Requirements for Data-Sharing
In order to enable data sharing, an IT platform must:
• Integrate and aggregate data from individual health systems
• Standardize and normalize data for comparisons across multiple
health systems
• Maintain data privacy and security to build a trusted network
• Provide a point-of-care application so health systems providing
data can learn and improve
10
Step 1: Source System Integration• Health systems use the IT platform to integrate data
across multiple systems and labs
Data-Sharing Platform
11
Step 2: Semantic Normalization Across Systems
• Choose a set of data elements that are clinically actionable and meaningful
• Emphasize data elements that can be automatically captured from existing
systems, except for data elements that require re-engineering data capture
workflows
– i.e. tumor histology
• Use vocabulary standards
• Automate the normalization process after the schema and standards have
been established
12
Step 3: Federated Architecture Allows for a Secure, Trusted Network
Step 4: Filter Real-World Data to View Insights on Clinically and
Molecularly-Similar Patients
14
OPeN Data Scope• Demographics: age, sex, gender, race, ethnicity
• Cancer diagnosis: primary site, histological diagnosis, stage
• Tumor genomics: gene, alteration
• Tumor markers: biomarker tests
• Treatments: next line of treatment after tumor genomic profile (chemo, targeted therapies)
• Outcomes: duration of therapy, survival, quality of life
15
OPeN Membership
Founding Members
Anticipated Future Members
16
Anticipated Reach of OPeN
136,000 New Cancer Cases Per Year
598Oncologists
241Hospitals
17
OPeN is Part of VP Biden’s Cancer Moonshot
• Vice President Biden announced the OPeN Network in his address at the Cancer Moonshot Summit in June 2016
• The effort aims to double the rate of progress in clinical care and cancer research over the next 5 years
• The initiative encourages health systems to come together in a national effort to share data
• VP Biden acknowledged the importance of OPeN to the future of cancer care
18
Benefits of Data-Sharing
• Provide clinicians with real-world, aggregated patient data to support treatment decisions and quality improvement
• Develop real-world evidence for existing therapies in new indications
• Support payer reimbursement efforts by referencing multi-institutional outcomes data
19
Why Share Data Now?
• The future of providing cancer care will be highly collaborative, evidence-based, and individualized
• Real-world evidence will increasingly guide treatment decisions and support payer reimbursement
• Join a national effort of innovator health systems to share insights and improve cancer care for all
20
Speaker Introduction
Paul Tittel, MHA
Systems Director, Providence St. Joseph
21
Conflict of Interest
Consulting Fees: Providence Health & Services
Swedish Health Services
Immunexpress Inc.
No other conflicts to report.
22
Agenda
• Challenges to data sharing
• Strategies for mitigating data-sharing challenges
• OPeN governance & data use provisions
• Legal, privacy, & compliance considerations
23
Challenges to Data Sharing
Technical & informatics challenges
• EHR landscape & IT ecosystems – lots of complexity
• Data standardization & semantic harmonization
• Lab heterogeneity & genomic data complexity
Legal, privacy & compliance considerations
• Legal framework: Consortium data sharing agreement
• Data “ownership” & use provisions
• Privacy & compliance
24
Challenge: Clinical Data HarmonizationDisparate clinical systems:
• Intermountain: HELP2 & Cerner EMRs
• Stanford: Epic EMR
• Swedish / Providence: Epic EMRs 5 different instances, 3 distinct “builds”
Mitigating strategies:
1. Leverage enterprise data warehouse (EDW) sources
• EDW data already partially normalized & harmonized within member orgs.
2. Focus on discrete, structured data elements
3. Map to standardized clinical data ontologies
25
Leveraging EDW Assets
26 distinct
cancer
registry
systems
Swedish
Ent. Data
Warehouse
Providence
WA/MT
Providence
OR/CA
Providence
AK
Kadlec
26
Example: Medications Data Harmonization
RxNorm: standardized ontology for medications from UMLS / NLM
• Normalized drug names for automated decision support, system interoperability, quality reporting, healthcare research & reimbursement analyses
• Supports multiple levels of descriptions & relationships
• Links to 11 distinct external drug vocabularies
• National Drug File Reference Terminology (NDF-RT) integration
– Metadata on clinical uses, therapeutic categories, mechanism of action, contraindications, known drug interactions, etc.
27
RxNorm Example: Nivolumab
NIVOLUMAB
100 mg/10ml - IV Soln.
Epic med. ID: 142344
RxNorm CUI: 1597876
Thera. class: Antinoplastics
Pharm. class: Antineoplastic;
Anti-Programmed Death-1
(PD-1) mAb
Nivolumab
Monoclonal antibodies
Other Antineoplastic Agents
Antineoplastic Agents
Antineoplastic & Immunomod. AgentsIM/MIN Ingredient
BN Brand Name
nivolumab
Opdivo
SCDC Clinical Drug Component
nivolumab 10 MG/ML
SBDC Branded Drug Component
nivolumab 10 MG/ML [Opdivo]
SCD/GPCK Clinical Drug or Pack
10 ML nivolumab 10 MG/ML Injection
SCDG Clinical Dose Form Group
nivolumab Injectable product
EMR-specific content
Standard ontology references & cross-platform metadata
Programmed Death Receptor-1
Blocking Antibody
28
Attribute Value
Display_Name NIVOLUMAB
code C50.416^4972^
FDA_UNII 31YO63LBSN
label NIVOLUMAB
Level Ingredient
NUI N0000191289
RxNorm_CUI 1597876
RxNorm_Name nivolumab
UMLS_CUI C3657270
VANDF_Record50.416^4972^Active/Master50.4164972Active/Ma
ster
VUID 4034032
Drug Name Interaction Description
Acetyldigitoxin Acetyldigitoxin may decrease the cardiotoxic activities of Nivolumab.
belimumabThe risk or severity of adverse effects can be increased when Nivolumab
is combined with Belimumab.
bevacizumab Bevacizumab may increase the cardiotoxic activities of Nivolumab.
cabazitaxelThe risk or severity of adverse effects can be increased when
Cabazitaxel is combined with Nivolumab.
Cyclophosphamide Cyclophosphamide may increase the cardiotoxic activities of Nivolumab.
Ouabain Ouabain may decrease the cardiotoxic activities of Nivolumab.
PaclitaxelThe risk or severity of adverse effects can be increased when Paclitaxel
is combined with Nivolumab.
trastuzumab Trastuzumab may increase the cardiotoxic activities of Nivolumab.
RxNorm Example: Nivolumab
Cross-references to other
standard drug dictionaries
Curated content on drug-drug
interactions (via DrugBank)
29
Example: Cancer Case Characterization Codified primary site & histopathology ontologies
• Developed at Swedish Cancer Institute by MD clinical informatics lead on Precision Medicine Program
• Aligned with World Health Organization (WHO; ICD-O-3) & College of American Pathologists (CAP) standards
Examples:Central Nervous System (Brain / Spinal Cord)
Astrocytic Tumors
Glioblastoma (WHO grade IV)
Giant cell glioblastoma (WHO gr. IV)
Ovary
Carcinoma
Clear cell carcinoma
30
Challenge: NGS Data/Lab Standardization
Again, disparate clinical sequencing & LIS / LIMS solutions
• Many NGS data management systems are “home-grown”
Mitigating strategies:
1. Rigorous enforcement of Syapse Lab Certification Program standards
• Focus: Complete, high-quality, & well-curated genomic data
• Codification of genomic metadata
2. Leverage Human Genome Variation Society (HGVS) standards
• Standardized nomenclature & descriptions for sequence variants
• Well-defined approach to reference sequences
31
Example: HGVS-Compliant Variant Desc. Genome build:
HUGO gene name:
RefSeq:
HGVS genomic change: :
HVGS coding change: :
HGVS protein change:
• External DB variation IDs populated whenever possible:
TP53
NM_000546
tumor protein P53
NC_000017.10
NM_000546
g.C7577058A
c.880G>T
GRCh37 / hg19
NP_000537.3 p.E294*
dbSNP dbVAR COSMIC ClinVar
32
OPeN Data Sharing AgreementCritical foundations:
• Clinical executive sponsor alignment
• Shared vision & aligned objectives
Data Sharing Agreement – legal codification
• Months of work; “working group” of institutional attorneys
– Key considerations: IP, data ownership; dissolution / exit provisions
• Each participating institution retains data “ownership” (stewardship)
• OPeN repository – fully de-identified; HIPAA risks markedly reduced
33
OPeN DSA: Data Use Provisions
1. Research projects: OPeN Steering Committee must review & approve project requests involving consortium data
2. Grant development: similarly, Steering Committee review & approval
3. Publications: all publications must cite the consortium in methodology & acknowledgement sections; authorship determined by contribution of individual authors
4. IRB requirements: OPeN participants responsible for institutional IRB-approval for specific research projects
34
Privacy & Compliance: Providence Perspective
Full review from privacy, compliance, & information security standpoints:
• Chief Privacy Officer, research compliance lead, & IT security analyst
• Key considerations:
– Full HIPAA de-identification
– OPeN inclusion only with patient consent (IRB-approved protocol)
Best practices:
• Transparent engagement, from the outset
• Engage Risk Mgmt. / Privacy as partners
35
Questions
jon@syapse.com
www.syapse.com
@syapse
paul.tittel@providence.org
www.providence.org
@prov_health
top related