State of the [Data] Science
Predictive Analytics World Healthcare
New York, NY October 31 2017
Ken Yale, DDS, JD
Chief Clinical Officer, Delta Dental
@docyale
Sophistication
Gain
Standard Reports What happened?
Source: Competing on Analytics, Davenport/Harris, 2007
Ad hoc reports How many, how often, where?
Query/drill down What exactly is the problem?
Alerts What actions are needed?
Statistical analysis Why is this happening?
Forecast/extrapolation What if these trends continue?
Predictive modeling What will happen next?
Optimization What’s the best that can happen?
Predictive
and
Prescriptive
Analytics
Descriptive
Analytics
DATA SCIENCE
@docyale
DATA SCIENCE APPLICATIONS
Healthcare Value
Consumer Retail
Business Process
Improvement
Health & Wellness
Source: https://www.veritasgenetics.com/
@docyale
* Lynch et al. Documenting Participation in a DM Program. JOEM 2006; 48(5)
1 Frazee et al. Leveraging the Trusted Clinician: Documenting Disease Management
Program Enrollment. Disease Mgmt 2007; 10:16-29
PROBLEM: CONNECTING WITH CONSUMERS
30%
15%
7%
3.5%
0
0.2
0.4
0.6
0.8
1
1.2
Eligible Contacted Participant Behavior
Improvement
Prepare
Data
Segment
Develop
Personas
- Gather demographics, health
history, risk assessments, health
attitudes, interaction data,
clinical analytics
- Organize and aggregate at the
Member, Activity and Condition
level
- Gather generally available, non-
specific information about
household behaviors and lifestyles
Gather Internal
Data
Gather External
Data
Combine
- Determine health
attitudes, behaviors
and lifestyles
- Put demographics
and behaviors in
context
- Convert all data into
numeric form for
statistical analysis
- Apply cluster
analysis algorithms
(e.g. K-Means, CART)
to determine
segments
- Analyze all Segment
characteristics
- Develop representative
example Persona Profiles
for each Segment
- Optimize products/
services for each
Segment
Example of population micro-segmentation: groups with uniform behaviors,
attitudes and lifestyles
SOLUTION: MICRO-SEGMENT
Source: Wiese K., 2014, Member Experience and Communications Segmentation Pilot, North Carolina State Health Plan
https://shp.nctreasurer.com/Board%20of%20Trustees%20Meeting%20Documents/BOT_4a_Segment_Pilot-8-1-2014.pdf
@docyale
PROBLEM: LOW PREDICTABILITY (≈ 20%)
Source: Society of Actuaries (2007) A Comparative Analysis of Claims-Based Tools for Health Risk Assessment.
@docyale
Traditional Financial Claims Data
Medical Conditions
Frequency
Psychosocial
Acuity or Chronicity
Complexity of Care Groupers
Clinical Decision Making
Financial Claims & Clinical Data
20% 25% 30% 35% 40% 45%
ACG
CDPS
Clinical Risk Group
DxCG (Verisk)
DxCG (Verisk)
Ingenix (Optum)
Medicaid Rx
Impact Pro
Ingenix ERG (Optum)
ACG Dx+Rx
DxCG UW Model
MEDai
Clin/Fin - Train
Clin/Fin - Test
SOLUTION: FINANCIAL AND CLINICAL INPUTS
Source: Wei H. “Prediction vs. Intervention (2014). Predictive Modeling Summit presentation. November 13, 2014,
Washington, DC.
@docyale
PROBLEM: COMPLEXITY OF MODERN MEDICINE
@docyale
@docyale
Current Market
50
@docyale
Payers/Providers:
• Care and Utilization Management
• Stratification and Identification
• Revenue Cycle Management
• Payment
• Risk Adjustment
• Claims Management
• Fraud and Abuse
• Actuarial and Underwriting
• Health Benefit Selection
• Treatment Options
• Medical Staffing
• Accountable Care / P4P
• Business & Clinical Process
Improvement
Pharma/Biotech/Device Manufacturers
• Research & Development
• Clinical Trials, Investigator Training
• Pragmatic Clinical Trials
• Health Economics/Outcomes Research
• Marketing/Product Launch
• Health Technology Assessment
• Protocol Development
• Regulatory Compliance
• Medical & Safety, FDA/EMA response
• Project Management, HR Planning
• Biostatistical Analysis
• Quality Assurance
• Coverage with Evidence Development
CURRENT USE OF ANALYTICS
@docyale
•Health analytics payer/provider (non-drug/device) market: $4.5 billion in 2013
•Growing at 25.2% CAGR from 2013 to 2020 to reach $21.5 billion, fueled mainly
by a growing need for predictive analytics for both payer and provider
•Growth fueled mainly by need for predictive analytics, a less saturated market
with new applications emerging across payer, provider, and life sciences.
•The market is considered “nascent” and unsaturated with new competitors and
emerging as technology advances and costs increase
•Fragmented, with major players having less than 15% share of market
•Major players include:
•Factors restraining market uptake: shortage of data scientists, provider resistance,
acute IT staff shortage, lack of standard data, operational gaps
 Cerner
 IBM/SPSS
 Elsevier/Medai
 McKesson/Medventive
 MedeAnalytics
 Optum/Humedica/Symmetry
 Oracle
 Truven Health
 Verisk/DxCG
 Other (SAS, Dell/Statsoft)
HEALTHCARE DATA ANALYTICS MARKET
@docyale12
NEW DATA SOURCES
@docyale
@docyale13
FRAGMENTED/CROWDED MARKET
@docyale
@docyale
▪ Improved quality of care
▪ Transparency for better care, products, and provider decisions
▪ Pharmacovigilance/comparative effectiveness
▪ Patient recruitment
▪ Medication adherence
▪ Quality measurement and improvement
▪ Evidence-based medicine
▪ New revenue sources (e.g., P4P, CIN, ACO)
▪ Value-based pricing, real-world outcomes
▪ Subpopulation coverage decisions
▪ Provider selection
Payors
Consumers
Hospitals/Physicians
Pharma
▪ Outcomes transparency
▪ Regulatory monitoring
▪ Understanding wellness
Employers/Government
 Value enabled by patient-level data
linking claims, Rx, lab, clinical, EMR
 New sources of data, such as social
media, give additional ability to create
an individual patient profile
 Given current data science tools, new
analytic services are an attractive
“adjacency”
 Investments in data lay groundwork
for long-term value both in direct data
opportunities and enhanced analytic
services
 New social media channels expand
patient engagement opportunities
NEW OPPORTUNITIES
@docyale
The Future
50
@docyale
PROBLEM: COMPLEXITY OF MODERN MEDICINE
@docyale
@docyale
ARTIFICIAL INTELLIGENCE?
WHAT IS AI?
Etc, etc, etc
@docyale
New medical “breakthroughs”
Tailor treatment and drugs to the individual – not “one-size-fits-all”
Better care, lower costs
“Precision medicine gives us one of the
greatest opportunities for new medical
breakthroughs that we have ever seen.”
President Barack Obama
January 30, 2015
https://www.whitehouse.gov/precision-medicine
https://www.genome.gov/images/content
Li-Pook-Than J, Snyder M. Chem Biol. 2013 May 23;20(5):662
PRECISION MEDICINE?
@docyale
1 2 3Identify Persons with Increased
Pre-disposed Risks
Genetic Test & Engagement in
Targeted Wellness Programs
Analyze Metrics, Refine
Approach
Feedback Loop
Identification & Screening
• Identify persons pre-disposed
to Metabolic Syndrome
• Provide counseling sessions to
help understand options and
answer questions
• While discussing options,
enroll in select wellness and
prevention programs
Measure ResultsGenetic Testing & Counseling
• Participation in targeted
wellness/prevention program
ID pre-
disposed risks:
• Surveys
• Participate in wellness
& prevention programs
• Claims analysis for
clinically recommended
preventative tests and
procedures (e.g., HEDIS)
GENETIC TESTING PILOT
Steinberg G., et al.. (2015). Reducing Metabolic Syndrome Risk Using a Personalized Wellness Program. JOEM. 57(12):
1269-1274.
@docyale
PILOT REACTION
@docyale
“DIGITIZATION” OF HEALTHCARE
1) Digital Inputs = Rapid Growth in
Sources of Digital Health Data
2) Data Accumulation = Proliferation
of Digitally-Native Data Sets
3) Data Insight = Generated Following
Accumulation & Integration of Data
4) Translation = Impact on
Therapeutics & Healthcare Delivery
5) Outcomes =
Measure Outcomes &
Iterate…
Innovation Cycle Times
Compressing
Source: Internet Trends 2017, by Mary Meeker, kpcb.com/InternetTrends
@docyale
Source: PBS, Propeller Health, TechCrunch, Livongo, Ayasdi, Flatiron, Xconomy, Kinsa, Omada
Patient Empowerment
& Health Management
Propeller Health + Bluetooth
Inhaler Sensor = Improved
Medication Adherence +
Insights
Livongo + Connected Glucose
Meter = Personalized Coaching
+ $100/Month Savings for
Payers
Improvements to
Clinical Pathways /
Protocol
Ayasdi AI + Mercy Health
System Patient Data = Clinical
Anomaly Detection + Improved
Clinical Pathway Development
Flatiron + Foundation Med (FMI)
= 20,000 Liked Cancer Patients
Records + Personalized
Medicine
Preventative Health
Kinsa + Crowdsourced
Temperature Data = Local Flu
Predictions + Proactive
Treatments for Populations
Omada + Preventative Program
= 4-5% Body Weight Reduction
+ Reduced Risk for Stroke and
Heart Disease
“DIGITIZATION” OF HEALTHCARE
Source: Internet Trends 2017, by Mary Meeker, kpcb.com/InternetTrends
@docyale@docyale
ALGORITHMIC MEDICINE
@docyale
HbA1C: >7
CONDITION: Diabetes
HISTORY: Co-morbid conditions
BEHAVIORAL
HEALTH:
Depression
Other
WEIGHT: Unexplained Loss
PHYSICAL:
AGE: 40-80
BP: >130/85
DIET
LAB VALUES:
TG: > 150 mg/dL
HDL: ♂ >
40 mg/dL
♀ >
50 mg/dL
LDL: <100 mg/dL
FPG: >100 mg/dL
GU:
-Polyurea
- UTI
- Yeast Infection
DERM: Dry, itchy skin
PREDISPOSITION:
Family History
Racial/Ethnic
Metabolic Syndrome
Pregnancy Diabetes
Symptomatic
VISION:
Blurred
NEUR: Numbness, Tingling
MEDICATION:
Metformin
Glitazones
Secretagogues
Combinations
Insulin
Other
CLAIMS:
ICD-9-CM: 250.xx,
357.2, 362.0x, 366.41,
648.0x
CPT: 9780x, 9920x,
9921x, 9930x, etc
OTHER
BMI: >30
a11 a12 a13 … a1n
a21 a22 a23 … a2n
:
am1 am2 am3 … amn
AGGREGATE DATA
@docyale
HEIGHT: 5’4”
WEIGHT: 175
AGE: 45
GU:
- Polyurea
- UTI
CITY:
New York
HDL: 55 mg/dL
LDL: 50
BMI: 33
TG: 250 mg/dL
FPG: 130mg/dL
MEDICAL Hx:
Pregnancy Diabetes
PHYSICAL:
BP: 185/35
Thirst, Hunger
Weight Loss
Blurred Vision
PROFESSION:
Senior Executive
BEHAVIORAL HEALTH:
Fatigue
Lethargy
FAMILY Hx:
Diabetes
Maternal Aunt
CLAIMS: 648.0
(12 years ago)
Otherwise healthy
b11 b12 b13 … b1n
b21 b22 b23 … b2n
:
bm1 bm2 bm3 … bmn
INDIVIDUAL DATA
@docyale
FAMILY
CARE TEAM
FRIENDS
NETWORK INTERESTS
DECISIONS
c11 c12 c13 … c1n
c21 c22 c23 … c2n
:
cm1 cm2 cm3 … cmn
OTHER DATA
@docyale
a11 a12 a13 … a1n
a21 a22 a23 … a2n
:
am1 am2 am3 … amn
z1(social determinants) + z2(personalized)
f (:)
b11 b12 b13 … b1n
b21 b22 b23 … b2n
:
bm1 bm2 bm3 … bmn
social- economic
health system
tx-related
condition-related
patient-
related
interest/social graph
spatial/temporal
media
c11 c12 c13 … c1n
c21 c22 c23 … c2n
:
cm1 cm2 cm3 … cmn
MACHINE LEARNING
1. Predictive Analytics
Health & Personal Data
2. Curated Medical Knowledge
3. Patient Relationships
4. Outcomes
Predict, Prescribe, Perform
UX INTERFACE
@docyale
RESOURCES
@docyale
QUESTIONS?
@docyale

1325 keynote yale_pdf shareable

  • 1.
    State of the[Data] Science Predictive Analytics World Healthcare New York, NY October 31 2017 Ken Yale, DDS, JD Chief Clinical Officer, Delta Dental
  • 2.
    @docyale Sophistication Gain Standard Reports Whathappened? Source: Competing on Analytics, Davenport/Harris, 2007 Ad hoc reports How many, how often, where? Query/drill down What exactly is the problem? Alerts What actions are needed? Statistical analysis Why is this happening? Forecast/extrapolation What if these trends continue? Predictive modeling What will happen next? Optimization What’s the best that can happen? Predictive and Prescriptive Analytics Descriptive Analytics DATA SCIENCE
  • 3.
    @docyale DATA SCIENCE APPLICATIONS HealthcareValue Consumer Retail Business Process Improvement Health & Wellness Source: https://www.veritasgenetics.com/
  • 4.
    @docyale * Lynch etal. Documenting Participation in a DM Program. JOEM 2006; 48(5) 1 Frazee et al. Leveraging the Trusted Clinician: Documenting Disease Management Program Enrollment. Disease Mgmt 2007; 10:16-29 PROBLEM: CONNECTING WITH CONSUMERS 30% 15% 7% 3.5% 0 0.2 0.4 0.6 0.8 1 1.2 Eligible Contacted Participant Behavior Improvement
  • 5.
    Prepare Data Segment Develop Personas - Gather demographics,health history, risk assessments, health attitudes, interaction data, clinical analytics - Organize and aggregate at the Member, Activity and Condition level - Gather generally available, non- specific information about household behaviors and lifestyles Gather Internal Data Gather External Data Combine - Determine health attitudes, behaviors and lifestyles - Put demographics and behaviors in context - Convert all data into numeric form for statistical analysis - Apply cluster analysis algorithms (e.g. K-Means, CART) to determine segments - Analyze all Segment characteristics - Develop representative example Persona Profiles for each Segment - Optimize products/ services for each Segment Example of population micro-segmentation: groups with uniform behaviors, attitudes and lifestyles SOLUTION: MICRO-SEGMENT Source: Wiese K., 2014, Member Experience and Communications Segmentation Pilot, North Carolina State Health Plan https://shp.nctreasurer.com/Board%20of%20Trustees%20Meeting%20Documents/BOT_4a_Segment_Pilot-8-1-2014.pdf
  • 6.
    @docyale PROBLEM: LOW PREDICTABILITY(≈ 20%) Source: Society of Actuaries (2007) A Comparative Analysis of Claims-Based Tools for Health Risk Assessment.
  • 7.
    @docyale Traditional Financial ClaimsData Medical Conditions Frequency Psychosocial Acuity or Chronicity Complexity of Care Groupers Clinical Decision Making Financial Claims & Clinical Data 20% 25% 30% 35% 40% 45% ACG CDPS Clinical Risk Group DxCG (Verisk) DxCG (Verisk) Ingenix (Optum) Medicaid Rx Impact Pro Ingenix ERG (Optum) ACG Dx+Rx DxCG UW Model MEDai Clin/Fin - Train Clin/Fin - Test SOLUTION: FINANCIAL AND CLINICAL INPUTS Source: Wei H. “Prediction vs. Intervention (2014). Predictive Modeling Summit presentation. November 13, 2014, Washington, DC.
  • 8.
    @docyale PROBLEM: COMPLEXITY OFMODERN MEDICINE @docyale
  • 9.
  • 10.
    @docyale Payers/Providers: • Care andUtilization Management • Stratification and Identification • Revenue Cycle Management • Payment • Risk Adjustment • Claims Management • Fraud and Abuse • Actuarial and Underwriting • Health Benefit Selection • Treatment Options • Medical Staffing • Accountable Care / P4P • Business & Clinical Process Improvement Pharma/Biotech/Device Manufacturers • Research & Development • Clinical Trials, Investigator Training • Pragmatic Clinical Trials • Health Economics/Outcomes Research • Marketing/Product Launch • Health Technology Assessment • Protocol Development • Regulatory Compliance • Medical & Safety, FDA/EMA response • Project Management, HR Planning • Biostatistical Analysis • Quality Assurance • Coverage with Evidence Development CURRENT USE OF ANALYTICS
  • 11.
    @docyale •Health analytics payer/provider(non-drug/device) market: $4.5 billion in 2013 •Growing at 25.2% CAGR from 2013 to 2020 to reach $21.5 billion, fueled mainly by a growing need for predictive analytics for both payer and provider •Growth fueled mainly by need for predictive analytics, a less saturated market with new applications emerging across payer, provider, and life sciences. •The market is considered “nascent” and unsaturated with new competitors and emerging as technology advances and costs increase •Fragmented, with major players having less than 15% share of market •Major players include: •Factors restraining market uptake: shortage of data scientists, provider resistance, acute IT staff shortage, lack of standard data, operational gaps  Cerner  IBM/SPSS  Elsevier/Medai  McKesson/Medventive  MedeAnalytics  Optum/Humedica/Symmetry  Oracle  Truven Health  Verisk/DxCG  Other (SAS, Dell/Statsoft) HEALTHCARE DATA ANALYTICS MARKET
  • 12.
  • 13.
  • 14.
    @docyale ▪ Improved qualityof care ▪ Transparency for better care, products, and provider decisions ▪ Pharmacovigilance/comparative effectiveness ▪ Patient recruitment ▪ Medication adherence ▪ Quality measurement and improvement ▪ Evidence-based medicine ▪ New revenue sources (e.g., P4P, CIN, ACO) ▪ Value-based pricing, real-world outcomes ▪ Subpopulation coverage decisions ▪ Provider selection Payors Consumers Hospitals/Physicians Pharma ▪ Outcomes transparency ▪ Regulatory monitoring ▪ Understanding wellness Employers/Government  Value enabled by patient-level data linking claims, Rx, lab, clinical, EMR  New sources of data, such as social media, give additional ability to create an individual patient profile  Given current data science tools, new analytic services are an attractive “adjacency”  Investments in data lay groundwork for long-term value both in direct data opportunities and enhanced analytic services  New social media channels expand patient engagement opportunities NEW OPPORTUNITIES
  • 15.
  • 16.
    @docyale PROBLEM: COMPLEXITY OFMODERN MEDICINE @docyale
  • 17.
  • 18.
  • 19.
    @docyale New medical “breakthroughs” Tailortreatment and drugs to the individual – not “one-size-fits-all” Better care, lower costs “Precision medicine gives us one of the greatest opportunities for new medical breakthroughs that we have ever seen.” President Barack Obama January 30, 2015 https://www.whitehouse.gov/precision-medicine https://www.genome.gov/images/content Li-Pook-Than J, Snyder M. Chem Biol. 2013 May 23;20(5):662 PRECISION MEDICINE?
  • 20.
    @docyale 1 2 3IdentifyPersons with Increased Pre-disposed Risks Genetic Test & Engagement in Targeted Wellness Programs Analyze Metrics, Refine Approach Feedback Loop Identification & Screening • Identify persons pre-disposed to Metabolic Syndrome • Provide counseling sessions to help understand options and answer questions • While discussing options, enroll in select wellness and prevention programs Measure ResultsGenetic Testing & Counseling • Participation in targeted wellness/prevention program ID pre- disposed risks: • Surveys • Participate in wellness & prevention programs • Claims analysis for clinically recommended preventative tests and procedures (e.g., HEDIS) GENETIC TESTING PILOT Steinberg G., et al.. (2015). Reducing Metabolic Syndrome Risk Using a Personalized Wellness Program. JOEM. 57(12): 1269-1274.
  • 21.
  • 22.
    @docyale “DIGITIZATION” OF HEALTHCARE 1)Digital Inputs = Rapid Growth in Sources of Digital Health Data 2) Data Accumulation = Proliferation of Digitally-Native Data Sets 3) Data Insight = Generated Following Accumulation & Integration of Data 4) Translation = Impact on Therapeutics & Healthcare Delivery 5) Outcomes = Measure Outcomes & Iterate… Innovation Cycle Times Compressing Source: Internet Trends 2017, by Mary Meeker, kpcb.com/InternetTrends
  • 23.
    @docyale Source: PBS, PropellerHealth, TechCrunch, Livongo, Ayasdi, Flatiron, Xconomy, Kinsa, Omada Patient Empowerment & Health Management Propeller Health + Bluetooth Inhaler Sensor = Improved Medication Adherence + Insights Livongo + Connected Glucose Meter = Personalized Coaching + $100/Month Savings for Payers Improvements to Clinical Pathways / Protocol Ayasdi AI + Mercy Health System Patient Data = Clinical Anomaly Detection + Improved Clinical Pathway Development Flatiron + Foundation Med (FMI) = 20,000 Liked Cancer Patients Records + Personalized Medicine Preventative Health Kinsa + Crowdsourced Temperature Data = Local Flu Predictions + Proactive Treatments for Populations Omada + Preventative Program = 4-5% Body Weight Reduction + Reduced Risk for Stroke and Heart Disease “DIGITIZATION” OF HEALTHCARE Source: Internet Trends 2017, by Mary Meeker, kpcb.com/InternetTrends
  • 24.
  • 25.
    @docyale HbA1C: >7 CONDITION: Diabetes HISTORY:Co-morbid conditions BEHAVIORAL HEALTH: Depression Other WEIGHT: Unexplained Loss PHYSICAL: AGE: 40-80 BP: >130/85 DIET LAB VALUES: TG: > 150 mg/dL HDL: ♂ > 40 mg/dL ♀ > 50 mg/dL LDL: <100 mg/dL FPG: >100 mg/dL GU: -Polyurea - UTI - Yeast Infection DERM: Dry, itchy skin PREDISPOSITION: Family History Racial/Ethnic Metabolic Syndrome Pregnancy Diabetes Symptomatic VISION: Blurred NEUR: Numbness, Tingling MEDICATION: Metformin Glitazones Secretagogues Combinations Insulin Other CLAIMS: ICD-9-CM: 250.xx, 357.2, 362.0x, 366.41, 648.0x CPT: 9780x, 9920x, 9921x, 9930x, etc OTHER BMI: >30 a11 a12 a13 … a1n a21 a22 a23 … a2n : am1 am2 am3 … amn AGGREGATE DATA
  • 26.
    @docyale HEIGHT: 5’4” WEIGHT: 175 AGE:45 GU: - Polyurea - UTI CITY: New York HDL: 55 mg/dL LDL: 50 BMI: 33 TG: 250 mg/dL FPG: 130mg/dL MEDICAL Hx: Pregnancy Diabetes PHYSICAL: BP: 185/35 Thirst, Hunger Weight Loss Blurred Vision PROFESSION: Senior Executive BEHAVIORAL HEALTH: Fatigue Lethargy FAMILY Hx: Diabetes Maternal Aunt CLAIMS: 648.0 (12 years ago) Otherwise healthy b11 b12 b13 … b1n b21 b22 b23 … b2n : bm1 bm2 bm3 … bmn INDIVIDUAL DATA
  • 27.
    @docyale FAMILY CARE TEAM FRIENDS NETWORK INTERESTS DECISIONS c11c12 c13 … c1n c21 c22 c23 … c2n : cm1 cm2 cm3 … cmn OTHER DATA
  • 28.
    @docyale a11 a12 a13… a1n a21 a22 a23 … a2n : am1 am2 am3 … amn z1(social determinants) + z2(personalized) f (:) b11 b12 b13 … b1n b21 b22 b23 … b2n : bm1 bm2 bm3 … bmn social- economic health system tx-related condition-related patient- related interest/social graph spatial/temporal media c11 c12 c13 … c1n c21 c22 c23 … c2n : cm1 cm2 cm3 … cmn MACHINE LEARNING
  • 29.
    1. Predictive Analytics Health& Personal Data 2. Curated Medical Knowledge 3. Patient Relationships 4. Outcomes Predict, Prescribe, Perform UX INTERFACE
  • 30.
  • 31.