(HLS305) Transforming Cancer Treatment: Integrating Data to Deliver on the Promise of Precision...
-
Upload
amazon-web-services -
Category
Technology
-
view
1.448 -
download
1
description
Transcript of (HLS305) Transforming Cancer Treatment: Integrating Data to Deliver on the Promise of Precision...
© 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.
November 13, 2014 | Las Vegas, NV
HLS305
Transforming Cancer Treatment:Integrating Data to Deliver on the Promise of Precision Medicine
Jonathan Hirsch, Syapse Kristen McCaleb, UCSF Nate Slater, AWS
75%OF CANCER
PATIENTS GET A
DRUG THAT IS
INEFFECTIVE FOR
THEM
Legacy oncology practice
“Nuclear bomb” therapies
Making cancer care
more effective through
precision medicine
Precision cancer care
“Smart bomb” therapies
UCSF’s challenge
Providing rich genetic
data and actionable
information to
physicians while
overcoming legacy
software infrastructure
Best of the 1980s:
EMR, PACS, LIS, CPOE, eMAR
Legacy software
The precision medicine workflow… …and barriers to adoption
Order test
Clinical workup &
Review clinical history
Lab generates MDx test report
View clinical & MDx data
Receive decision support based on
guidelines, clinical, molecular data
Order therapy or
enroll patient in clinical trial
Process drug procurement
Monitor patient outcome
& revise care strategy
Track cost & adherence
Obtain pre-authorization
Molecular Tumor Board reviews
clinical & MDx data; delivers
guidance to physician
Obtain off-label
reimbursement authorization
Assess health outcomes &
modify care pathways
decision support for MDx test orders
pre-authorization support
systematic decision support for
therapy or clinical trials
mechanism for sharing patient records
systematic capture of physician
decisions & patient outcomes
systematic capture of treatment costs
systematic update of care pathways
No
No
No
No
No
No
No
data integration and visualizationNo
The Precision Medicine Workflow… …and barriers to adoption.
Order test
Clinical workup &
Review clinical history
Lab generates MDx test report
View clinical & MDx data
Receive decision support based on
guidelines, clinical, molecular data
Order therapy or
enroll patient in clinical trial
Process drug procurement
Monitor patient outcome
& revise care strategy
Track cost & adherence
Obtain pre-authorization
Molecular Tumor Board reviews
clinical & MDx data; delivers
guidance to physician
Obtain off-label
reimbursement authorization
Assess health outcomes &
modify care pathways
decision support for MDx test orders
pre-authorization support
systematic decision support for
therapy or clinical trials
mechanism for sharing patient records
systematic capture of physician
decisions & patient outcomes
systematic capture of treatment costs
systematic update of care pathways
No
No
No
No
No
No
No
data integration and visualizationNo
EMR tabs
EMR records
Paper reports
Emails
Phone calls
XLS, PPT, DOC files
Mental steps
8
~50
9
4
5
12
4
Conservative estimate by users
UCSF “APeX” Electronic Medical Record
UCSF “APeX” Electronic Medical Record
Can’t handle
complex genomic data
No data mining,
visualization
Built for billing
& compliance
UCSF “APeX” Electronic Medical Record
But the EMR is central to clinical workflow, so we must figure out how to get
information in & out with minimal disruption to physicians.
Genomics InfoSec today
“Obtaining the information needed to develop treatment plans
and summaries…required 23 separate steps including visits to
various screens, tabs and other sources within the EMR.”
ASCO, The State of Cancer Care in America: 2014
Genomic data: EMR “Import”
How do we fix this?
A modern-day Tower of Babel
No standard schemas
No standard terminology
Unstructured or
semi-structured
Thousands of record types
Millions of property types
API for healthcare
The Syapse solution
Data integrationPhysician
Sequencing &
Analytics
Sendout
Labs
One-Time Migration
Data Ingestion
Excel
PowerPoint
Filemaker Pro
LIS
Data
WarehouseCPOE
EMRPACS
Drug
Administration
Interfaced Systems
Oncologist dashboard
5
4
3
2
1
2 Omics data
3 Drug procurement
5 Imaging metadata
1 Structured clinical data
4 Longitudinal data
* All data included in this chart is for informational purposes
only and does not include actual patient data
Cancer genomics workflow enabled by Syapse
Clinical
Workup
Patient
Consent
Test Order
in EMR
Specimen
Procurement
Sequencing
& Processing
FilteringSearchable
Database
Report
Delivery
Clinical
Data Review
Molecular
Tumor Board
Syapse
Clinical
Decision
Integrate molecular data into clinical workflow
Tailor decision support to organization best practices
Extend expertise to affiliate network
Introducing Syapse:Enterprise software to enable precision medicine
Adopted by large healthcare providers
DATA SOURCES
?
Modern data architecture enabling innovative features
DATA SOURCES
Industry-leading security & HIPAA compliance
ARCHITECTURE
Controlled Vocabs
ETL Process
Semantic Store
Data Model Config
Expressive, flexible data integration
DATA SOURCES
Industry-leading security & HIPAA compliance
TOOLS
Semantic Query
Rules Engine
Knowledge Base
Workflows
ARCHITECTURE
Controlled Vocabs
ETL Process
Semantic Store
Data Model Config
Interfaces with and augments existing systems
DATA SOURCES
Industry-leading security & HIPAA compliance
TOOLS
Semantic Query
Rules Engine
Knowledge Base
Workflows
ARCHITECTURE
Controlled Vocabs
ETL Process
Semantic Store
Data Model Config
APPLICATIONS
Analyst Application
Patient Dashboard
Decision Support
Interactive Reports
Intuitive user interfaces supporting complex processes
DATA SOURCES
Industry-leading security & HIPAA compliance
TOOLS
Semantic Query
Rules Engine
Knowledge Base
Workflows
ARCHITECTURE
Controlled Vocabs
ETL Process
Semantic Store
Data Model Config
APPLICATIONS
Analyst Application
Patient Dashboard
Decision Support
Interactive Reports
INTEGRATION
REST API Interface Engine External Connectors
Interfaces with and augments existing systems
DATA SOURCES
Syapse AWS architecture
Syapse technology stack
Interface Engine
Semantic Layer
Components
Web Application
Components
Application Data Stores
Application
RDBMS
RDF
Biomedical
Data Store
Customized
Components
Gateway
Web Server
Client Browser
Applications
IPsec Tunnel to
Customer Firewall
REST
API
Application Data
Access Layer
SPARQL Layer
Syapse Customer Deployment
Custom Apps
S3 File
Storage
Snapshot
• AWS enables creation of enterprise-grade web application architectures
• AWS does not constrain choice of database, application components, services, etc.
HL7
Syapse technology stack
• Single-page JavaScript applications
• RESTful HTTPS API
• Python / Django application server
• MySQL database
• “Bigdata” triple store
• Open-source (SYSTAP)
• AWS: Amazon EC2, Amazon S3, Amazon EBS
• Linux: Ubuntu Server
• nginx, Memcached, etc.
Web Application Stack
Semantic Data Layer
Deployment
• Custom server: Loosely coupled Python
component
• Integration messages (API)
• Workflows (API, Syapse UIs)
• Report configuration
• Open-source interface engine (Mirth
Connect)
• Standard message implementations
Configuration Facility
HL7 Messaging
Syapse semantic data platformComplexity
Flexibility
Customizability
Data Integration
Precision
Sufficient expressivity for medicine
Ontologies can be changed dynamically
Composable, extendable, reusable customer ontologies
Hub for integrating disparate data types
Expressive, exact queries
High availability
Amazon EC2 enables loose-coupled N-tier architecture
Tiers can scale independently and on demand
Elastic Load Balancing distributes traffic across EC2 instances / zones
Geographically distributed
3 main campuses,
19 other locations,
5 affiliates
UCLA Clinical Exome Sequencing
UCSF
Major
MDx laboratories
Segregation of Production, Dev, QC
Resource Tier
Client Layer
Service Tier
Customer 1
Production Server
Resource Tier
Client Layer
Service Tier
Customer 1
Dev Server
Resource Tier
Client Layer
Service Tier
Customer 1
QC Server • Easy deployment in cloud
• Templated instances
• Pipes to mirror configuration
• Necessary for integrating with
hospital systems
Web Server
HTTPS
Secure integration with customer systems
Interface Engine Gateway
Syapse Customer Deployment
Client
VPN
Syapse Application
SyMessage
FacilityInterface Engine
Customer Firewall
Application
Components
HL7 HL7
EMR CPOE Path.
IPsec Tunnel to
Customer FirewallSSLHTTPS
Real-time computation
Physicians Need:
• Point-of-care access to
large-scale data
• Surfacing of relevant data,
information, decisions
• Fast, complex, under-the-
hood queries
• Semantic rules engine for
real-time decision support
AWS Enables:
• Memory- / IO-intensive
application
• Indexing of data in real-time
& loading in memory
• Flexible node size, number
• Moving indexes between
nodes
AWS & HIPAA compliance
Backup, recovery, redundancy
Nightly
Snapshots
Customer
Instance
Backup tapes not required
Test recovery procedure
through AWS admin tools
Failover AZ #1
Region #1
AZ #2
AZ #3
Region #2
AZ #4
Security: standard
• Auth integration: SSO (SAML, OAuth), LDAP
• Best-in-class AWS security
• Regular audits against SAS70, ISO 27001, and others
• FedRAMP Compliant Cloud Service Provider (HHS)
Web Application Security
Data Center Security
Security: healthcare-grade
• Single-tenant deployment
• Dedicated hardware
• AWS CloudFormation templating
• Facilitated by AWS Management Console
• All data encrypted at rest and in transit
• Encryption of data in transit between each
layer of Syapse application stack
Customer Data Segregation
Data Encryption
• AWS Virtual Private Cloud
• Gateway to customer firewall
Firewall
HIPAA compliance
EC2
S3
EBS
Glacier
Application deployment
File storage
Snapshots
Long-term data retention
BAA-Covered Services AWS Platform
• Access logging
• Data integrity monitoring
• Flexibility to implement additional
security provisions (e.g., encryption of
data in transit between layers)
HIPAA compliance
Data can be segregated
within an AWS region
AWS does not move
customer content outside of
the customer’s chosen
AWS region
Syapse and AWS have signed a BAA
Syapse has signed BAAs and passed
InfoSec reviews with several of the largest
healthcare providers in the United States.
AWS & healthcare
Nate Slater
AWS Solutions Architect
The AWS philosophy
Eliminate “Undifferentiated Heavy Lifting”• Syapse and UCSF are able to focus on improving cancer treatment
“Shared Responsibility” Security Model• Provides a foundation that allows Syapse to meet the most stringent
security requirements in their technology stack
Customer Obsession• AWS is excited to partner with Syapse and UCSF to help them
deliver on the promise of precision medicine
Please give us your feedback on this
presentation
© 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.
Join the conversation on Twitter with #reinvent
HLS305
Join us outside for Q&A with Syapse
CTO Tony Loeser, PhD and VP of
Customer Solutions Andreas Heid.
www.syapse.com
gmi.ucsf.edu
Jonathan Hirsch
Kristen McCaleb