CANCER LINQ Peter Paul Yu, MD, FACP, FASCO Washington State Medical - - PowerPoint PPT Presentation

CANCER LINQ

Peter Paul Yu, MD, FACP, FASCO Washington State Medical Oncology Society March 27, 2015

SITUATION LEARNING IS SLOW, LIMITED AND OFTEN NOT RELEVANT

Today, most information is lost

Only3% enroll in clinical trials.

3%

1.7

people diagnosed with cancer in the US

MM

less diverse…

healthier… younger… and less diverse…

…than most of the patients we care for every day.

Clinical trial patients tend to be…

• 1. Lewis JH, et al. Participation of patients 65 years of age or older in cancer clinical trials. J Clin Oncol. 2003;21:1383-1389. http://jco.ascopubs.org/content/21/7/1383.full.pdf.
• 2. Mitchell AP, et al. Clinical trial subjects compared to "real world" patients: generalizability of renal cell carcinoma trials. J Clin Oncol. 2014;32(suppl):6510.
• 3. Taking action to diversify clinical cancer research. National Cancer Institute Web site. http://www.cancer.gov/ncicancerbulletin/051810/page7. Accessed July 23, 2014.

Credit: Dan Masys

Information overload

In 2013, Medline added 734,052 citations Assume just 1% of that new literature is relevant to a doctor's practice If a doctor reads 2 articles per night…. ….they would still be 10+ years behind

1986 One disease 2014 7 molecular drivers …and more to be discovered

From one cancer to many…

1.7

people diagnosed with cancer in the US

MM

97%

• f patient data

locked away in unconnected files and servers

will unlock a universe of practical insights to improve the care of every patient with cancer.

Data into Learning

Data Knowledge Base Learning

12

The future of cancer care relies on big data and health IT

The ability to mine large repositories of data in order to: – Evaluate treatment quality using large populations of similar patients – Identify long-term patient outcomes – Validate treatments/outcomes – Test treatment/outcome hypotheses

With Big Data… In a clinical trial…

Hypothesis* genera/on* Data$ collec)on$&$ analysis$ **Adap/ve** *****clinical*trial******* design* Tes/ng* Insights$

Hypothesis development Study design Data collection & analysis Insights Clinical intervention

Real-world patients Da Data$ a$ ag aggr greg ega(o a(on n Population health

• utcomes

Imp mproved$ he healthc althcar are$

• p
• per

era(on

• ns

Improved quality of care Clinical decision support

What is CancerLinQ?

CancerLinQ

• A rich data repository
• Not a registry for predefined questions
• Allows data exploration of the real world
• Generates new insights into warranted

variation

• Accelerates best practices, guideline

development and CDS.

Example: ESA Usage in 8,300 Breast Cancer Cases

Percentage of Cycles Median Hb level

Data Sources

• 12-15 Vanguard Practices:

cancer centers, hospital systems and community practices

• CancerLinQ will extract information from

EMR and Practice Management Systems

• The entire medical record will be captured
• Both structured fields and unstructured

data (such as physician notes)

20

We want all the data associated with oncology patients…

We do not have a defined set of fields, a specific data extract, or an implementation guide— we work with each data source to extract and load.

Clinical data

• Demographics, clinical notes, procedures, practice/provider information, eRX/drug

administration, ROS/physical exams, allergies, history of present illness Practice management data

• Scheduling, billing, inventory , payer information

Other data

• Labs, pathology, tests, measurements, immunizations, encounters, medical equipment

We Have Laid Out Our Principles That Underpin CancerLinQ

• Stewardship

– Robust standards; ethical procedures; adapting to changes in legal and ethical standards

• Protection

– Prevent harm; minimize risk; strong data security

• Transparency and Accountability

– Ethical duty to respect all participants in the system

22

CancerLinQ data requirements

Data are the fuel that drives CancerLinQ

Our focus is on ensuring minimal impact to sources of data

Institute fully automated processes, including process monitoring No data entry requirements for participating data sources Data provided from the practices to CancerLinQ are fully identifiable and will contain PHI and PII CancerLinQ will de-identify the data as part of its standard processing CancerLinQ will generate reports or redacted data sets necessary to address 
 specific questions raised by interested users

Board-appointed 


• versight committees

Compliance policies 
 and procedures Routine data 
 quality assessments Robust data request 
 management process

CancerLinQ will be governed by a robust and multifaceted data stewardship program

Regulatory Compliance

We have extensively studied the relationship of CancerLinQ to the Common Rule and to HIPAA. Our findings were carefully laid out in a JCO paper in August

• f this year.

25

• There is a sound legal basis for CancerLinQ
• All patient identifiable data in CLQ will be

maintained and used in compliance with HIPAA

• New England IRB has determined that CLQ

participation is not human subjects research

• Patients will be informed about CLQ and the

system will accommodate patient opt-out

Regulatory Underpinnings of CLQ

26

• Practices can disclose patient identifiable

health information (“PHI”) to CLQ without patient authorization

– for quality improvement and other health care

• perations

– under a HIPAA “Business Associate Agreement”

• This is done in ASCO’s QOPI program and

many other registries and quality initiatives

HIPAA

27

• CLQ can use or disclose PHI only for the

health care operations agreed to in advance by the oncology practices

– quality improvement and reporting – benchmarking – clinical decision support – clinical trials indexing – creating limited data sets and de-identified information

HIPAA

Protecting personal health information

• Physicians and practices will be able to

access PHI from their patients only

• Learning, trend-analysis, research, and

guideline development will be done on redacted data sets

29

Data Ingestion

How will we bring data into CancerLinQ?

• Extract

→ Read data from source and push data into DB1 in CLQ

• Transform

→ Standardize, normalize, and ontologize/ conceptualize the raw practice data

• Load

→ Load transformed data into DB2

Data Source DB2 DB1

Extraction Storage Transformation Data Load

Clinical Data Practice Management Data Payor Data Integration Gateway

Raw Data (as received) Processed Data

Canonical Format

Use Case: Quality Improvement

• Physicians and practice

managers can look at reports of adherence to quality performance measures

• Reports will be ‘near real-

time’ and not involve manual chart abstraction

• Trends will be longitudinal,

control charts mapping non- random variation

CancerLinQ measures analysis—Preliminary design

My favorites (1) All measurements (143) Colon (23) Breast (32) Lung (12) Prostate (67) Non-patient related (9) Pathology report Antiemetic therapy administered for moderately emetogenic chemotherapy risk Fertility preservations discussed Antiemetic therapy administered for moderately emetogenic chemotherapy risk Trastuzumab not received— her-2/neu negative Pain assessed, 2 most recent visits Central-line catheter— associated blood stream infection rate fro ICU and high-risk nursery (HRN) patients KRAS gene mutation testing Chemotherapy plan GCSF administered to patients who received chemotherapy for metastatic breast cancer

88.3

%

91.5

%

91.4

%

84.3

%

92.0

%

87.3

%

1.42

%

72.6

%

90

% %

Use Case: Clinical Decision Support

• Personalized diagnostic

and treatment guidance based on the best available evidence

• Prompts to improve

quality, such as oncology drug related interactions

• Iterative machine learning

driven CDS development

Types of CDS

Information Management Info Button, Up ToDate Situational Awareness Alerts, Dashboards Patient Decision Making Logic-based guidance

Healthcare Systems KB Systems Biology KB CPGs KB Clinical Decision Support Systems Shared Decision Making Outcomes Patient Data

Integrating Guidelines & Measures into CancerLinQ

Guidelines Measures QOPI/eQOPI CancerLinQ CDS = Clinical Decision Support

Outcome Data Measure Data

CancerLinQ Platform Powered by SAP HANA

Ambulatory EHRs Inpatient EHRs Biobank Lifestyle Practice Management Systems

‘Omics Clinical

EHR Fin

Biometric

SAP HANA Health Platform

DB-oriented Logic Text Mining SQL Scripts Decision Tables Extended App Services (Web Server) Procedural App Logic Rules Engine R Integration Unstructured Predictive

Financial

Applications Analytics

HANA Healthcare Platform

Leading Organizations Trust SAP’s Industry Expertise

Health Insurance Life Sciences &Suppliers D i s c r e t e M a n u f a c t u r i n g Health Insurance Life Sciences &Suppliers D i s c r e t e M a n u f a c t u r i n g

26

SAP Snapshot

$22.2B+

SAP revenue worldwide

#1

Enterprise Software

68,000+

employees worldwide

261,000

customers in 190 countries

74%

World’s transaction revenue

86%

Fortune 500 market share 25

CancerLinQ Co-Innovation

ASCO’S CancerLinQ, LLC SAP

! Overall development of CancerLinQ ! Control over the data, services and products that stem from CancerLinQ including clinical decision support tools and analyses ! Access to HANA ! Customized tools unique to CancerLinQ’s needs ! Engineering and other technical support

41

The Build-Out

San Francisco Office (April 1) Construction in HQ (Alexandria)

CancerLinQ’s Potential