and Precision Medicine for Health Systems Enabling the Transformation of Healthcare Systems November 14 th , 2016 CONFIDENTIAL
Key Messages 1. A variety of data types are needed to enable precision medicine 2. These data types include: a. Clinical data b. Lab and genomics data c. Imaging data d. Sensor data e. Patient reported data 3. The amount and size of these data sets will require them to be collected in a cloud computing environment 4. Hospital systems need a well thought out, systematic approach to developing the infrastructure to collect and analyze this data 5. If done properly, this comprehensive data set can be used to drive insights and better clinical outcomes and improve drug development through both traditional analytics and machine learning 2
Enterprises are experiencing a Digital Transformation 2010 2020 Individual Collective Productivity Intelligence IT Silos Distributed Computing Data on premise, hard to Data stored in cloud, simple to query access, analyze and use Machine learning drives deep, actionable insights Productivity tools built for individual, local usage Collaborative, cloud based productivity applications IT focusing on where it computes IT changing how it computes.
Enterprises are experiencing a Digital Transformation ^ 2010 2020 Individual Collective Productivity Intelligence IT Silos Distributed Computing Data on premise, hard to Data stored in cloud, simple to query access, analyze and use Machine learning drives deep, actionable insights Productivity tools built for individual, local usage Collaborative, cloud based productivity applications IT focusing on where it computes IT changing how it computes. 4
The same data and technology can be used for both clinical research and patient care 5
Creating the Infrastructure to Support Precision Medicine Sources Settings Data Processing / Google Cloud Based Platform Solutions & Apps Epic and Cerner Community Based Health Records Care IMAGING SENSORS SELF- Acute Care Imaging Systems REPORTED MOLECULAR DATA CLINICAL Port-Acute Care Lab and Genomic Data RESEARCH Ambulatory Surgery Center Sensor and PRO data Patients
Healthcare Capabilities
Standing on the shoulders of the Web erily (2013) (2008) (1998) Building on Google's core infrastructure, data analytics, and machine learning. 8
Platform Vision CLINICAL STUDY DISEASE CUSTOM QUALITY & MANAGEMENT MANAGEMENT APPLICATIONS REIMBURSEMENT API Analysis Tools SELF- IMAGING SENSORS CLINICAL MOLECULAR REPORTED DATA shared infrastructure (workflows, frameworks,…) COLLABORATIVE INVESTOR PUBLIC DATA COLLABORATIVE DATA PRIVATE DATA PUBLIC DATA DATA DATA Images courtesy of Verily Life Sciences 9
Clinical Capabilities CLINICAL STUDY DISEASE CUSTOM QUALITY & MANAGEMENT MANAGEMENT APPLICATIONS REIMBURSEMENT API Analysis Tools SELF- IMAGING SENSORS CLINICAL MOLECULAR REPORTED DATA shared infrastructure (workflows, frameworks,…) COLLABORATIVE INVESTOR PUBLIC DATA COLLABORATIVE DATA PRIVATE DATA PUBLIC DATA DATA DATA Images courtesy of Verily Life Sciences 10
Mapping Clinical Data Sources Settings Processing/ Warehousing Registry Standards Based Epic and Cerner Community Based EHR Adapter Health Records Care HISP Registry 1 DIRECT Authentication CCDA Imaging Systems Acute Care Warehouse Smart Registry 2 Cleansing/ FHIR Translating i2b2 Lab and Genomic Data Registry 3 NLP Port-Acute Care Registry 4 Ambulatory Surgery Center Sensor and PRO data Registry 5
Getting Data Mapping Right - Core Suite Tools
Getting Data Mapping Right - Risk Assessment Process
Provider analytics Your hospital had a 10.1% longer Areas for Highlights improvement length-of-stay for Knee Joint Replacement (127 bed-days which costs $549,004.20). This change may be driven by severity 2 cases, which are higher by 10.3%. The longer stays in severity 2 could account for 49.5% (63 bed-days) of the total increase.
Genomics Capabilities CLINICAL STUDY DISEASE CUSTOM QUALITY & MANAGEMENT MANAGEMENT APPLICATIONS REIMBURSEMENT API Analysis Tools SELF- IMAGING SENSORS CLINICAL MOLECULAR REPORTED DATA shared infrastructure (workflows, frameworks,…) COLLABORATIVE INVESTOR PUBLIC DATA COLLABORATIVE DATA PRIVATE DATA PUBLIC DATA DATA DATA Images courtesy of Verily Life Sciences 15
Genomics workflow PI/Biologist Bioinformatics Scientist Bioinformatics Programmer Web Access R, Python, SQL SSH DNA Sequencer Share API Reads Google Cloud Storage Google Genomics Google BigQuery & Variants Store Process Explore 16
PrecisionFDA Truth Challenge 17
PrecisionFDA Truth Challenge
Imaging Capabilities CLINICAL STUDY DISEASE CUSTOM QUALITY & MANAGEMENT MANAGEMENT APPLICATIONS REIMBURSEMENT API Analysis Tools SELF- IMAGING SENSORS CLINICAL MOLECULAR REPORTED DATA shared infrastructure (workflows, frameworks,…) COLLABORATIVE INVESTOR PUBLIC DATA COLLABORATIVE DATA PRIVATE DATA PUBLIC DATA DATA DATA Images courtesy of Verily Life Sciences 20
21
Other research possibilities ... OTHER IMAGING RETINA SKIN CONDITIONS EYE DISEASES Moles Glaucoma Skin cancer Age-related macular Infections degeneration Acne/rosacea Dermatitis Hair/nail SYSTEMIC DISEASES Stroke & heart attack risk Diabetic nephropathy, neuropathy EAR, NOSE, THROAT Vascular dementia, Alzheimer’s Ear infections Mortality? Hospitalizations? Sore throat 22
Sensor Data Capabilities CLINICAL STUDY DISEASE CUSTOM QUALITY & MANAGEMENT MANAGEMENT APPLICATIONS REIMBURSEMENT API Analysis Tools SELF- IMAGING SENSORS CLINICAL MOLECULAR REPORTED DATA shared infrastructure (workflows, frameworks,…) COLLABORATIVE INVESTOR PUBLIC DATA COLLABORATIVE DATA PRIVATE DATA PUBLIC DATA DATA DATA Images courtesy of Verily Life Sciences 23
Architecture Overview Analysis Sensor Data Data Secure End-user Devices Upload Pipeline Storage Website End-user Mobile Apps send the right data to the right algorithms at the right times 24
Pipeline Example: pulse data computation ppg data Pulse estimate 15 minute Summaries Raw Firehose Collect 1 Pulse Recovery Regions hr acc data Collect 1 Idle pulse Active pulse hr Pulse Recovery activity Model 25
Pipeline Example: REM Truth (Blue) sleep quantity and quality NREM Conf Model (Dash) (Red) Extract Nighttime PPG Sleep Using Sleep Onset/Offset Stager Activity Detections Classification (~18 hrs) Heart Rate Variability Sleep Onset/Offset Sleep Onset Sleep Offset Detection HR Raw Firehose Estimation (~18 hours) 26
Solutions: Baseline Study CLINICAL STUDY DISEASE CUSTOM QUALITY & MANAGEMENT MANAGEMENT APPLICATIONS REIMBURSEMENT API Analysis Tools SELF- IMAGING SENSORS CLINICAL MOLECULAR REPORTED DATA shared infrastructure (workflows, frameworks,…) COLLABORATIVE INVESTOR PUBLIC DATA COLLABORATIVE DATA PRIVATE DATA PUBLIC DATA DATA DATA Images courtesy of Verily Life Sciences 27
“ Google has embarked on what may be its most ambitious and difficult science project ever: a quest inside the human body. ” Wall Street Journal | July 2014 28
Broad and Deep Molecular, Device, and Clinical Phenotyping Data for Each Participant Clinical Device Imaging -Omics Immunoprofiling Data Data Data Data Data Monocytes B cells CD4 T cells CD8 T cells other 29
Example: Supported Verily Analyses ? Ingestion, preprocessing, QC: Import data at LIMS-level. Automatically survey data quality and highlight areas of concern. Determine pre-analytical, analytical, and biological variability. Clustering: unbiased or hypothesis-weighted clustering of multi-omics data to reveal unique patterns. Regression: supervised or semi-supervised methods that import known biological information. Longitudinal: analysis and sequence prediction in longitudinal data. CC eQTL/mQTL: integrative analysis combining multiple genetic data types. CA GRS: genomic predisposition; advanced modeling across multiple population data. Advanced machine learning: for integrative pathway discovery & analysis, data annotation, quality control, and phenotype-*omics associations. 30
Solution: Quality Improvement / MACRA CLINICAL STUDY DISEASE CUSTOM QUALITY & MANAGEMENT MANAGEMENT APPLICATIONS REIMBURSEMENT API Analysis Tools SELF- IMAGING SENSORS CLINICAL MOLECULAR REPORTED DATA shared infrastructure (workflows, frameworks,…) COLLABORATIVE INVESTOR PUBLIC DATA COLLABORATIVE DATA PRIVATE DATA PUBLIC DATA DATA DATA Images courtesy of Verily Life Sciences 31
The MACRA Quality Payment Program Consolidates key aspects of three existing physician-based programs PQRS Medicaid Medicare EP MU EP MU Value-based Separate Requirements & Modifier Separate Submission Merit-based Advanced Alternative Incentive Payment Payment Models System (MIPS) (Advanced APM) Eligible Clinicians Qualifying APM Participant Composite Performance Score Risk-based Payment arrangements
Recommend
More recommend
