Page 1
The Product Development Section Presents
Underwriting Issues & Innovation Seminar July 31-August 1, 2017 | The Westin O’Hare | Chicago, IL
Risk Assessment Tools – Part 1
Moderator:
Cindy Mitchell
Presenters: Robert Stout
Brian Lanzrath Dianne Schuetz
Eric J. Carlson, FSA, MAAA
Page 2
Underwriting Issues and InnovationEric CarlsonJuly 2017
Page 3
Agenda
What is Risk Score
Simplified Issue Case Study
Implementation Options
Rx and RxRules Background
Fully Underwritten Case Study
Page 4
The Future of Underwriting
Electronic requirements (Rx, MIB, MVR, Medical, Credit …)
Decision engines driven by data
Predictive Models
Automation
Increasing
APS, Labs
Cycle times
Costs
Decreasing
Better Customer Experience
Page 5
Rx and mortality
4
2015Milliman mortality study 53M exposure years 13M lives 231,000 deaths Created Milliman Risk Score
Health Plan
PBM
Clearing House
RetailPharmacy
Access (with authorization) to Rx Histories on more than
200 million Americans.
Milliman has accumulated a large Rx and mortality data set.
Page 6
Rx Histories
1 PrescriptionBrand and generic name | Dosage and quantity | Date of fill
2 PhysicianSpecialty | Contact information
3 PharmacyContact information
4 Dates of eligibilityWith or without prescriptions
5 Underwriting significance indicatorRed, yellow, green
Page 7
RxRules interprets prescriptions and other inputs.
RxRulesData Input Rx data Application data MIB / MVR Medical data
UW Guidance Conditions Severity Decisions
Rule Variables Indication / Therapeutic class Drug combinations Fill timing(date or duration ranges) Fill counts / patterns Dosage / quantity Physician specialty / count Gender / Age Other variables
Page 8
RxRules – Dosage Matters
Trazodone147% relative mortality
Low Dose
132%
High Dose
224%
Page 9
RxRules – Drug Combinations Matter
Spironolactone209% relative mortality
Thiazide Diuretics (102%)
Ace / Angio II (ARBS) (116%)
Beta Blocker (122%)
With 2 out of 3 of:
328%
Thiazide Diuretics (102%)
Ace / Angio II (ARBS) (116%)
Beta Blocker (122%)
Without 2 out of 3 of:
166%
Page 10
Milliman Risk Score
RxRules-driven predictive model
Predicts relative mortality of a life or group of lives
Holistic, multi-variate Rx model of mortality risk
Delivered within RxRules
Page 11
The Milliman Risk Score is built on RxRules.
Milliman Risk Score
250,000NDC codes
7,500GPI codes
Hundreds of RxRules
1.27
Page 12
Milliman’s Risk Score effectively predicts mortality.
0%
100%
200%
300%
400%
500%
600%
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40
400,000
800,000
1,200,000
1,600,000
2,000,000
2,400,000
Milliman Risk Score
Rel
ativ
e M
orta
lity
Indi
vidu
als
Relative Mortality and Lives by Milliman Risk Score
Lives Relative Mortality
Source: 2015 Milliman Mortality Study: 13M lives, 8M Rx hits, 230K deaths
Page 13
Risk Score Applications
1 Individual Underwriting
2 Group Underwriting
3 Inforce Analysis
4 Market Segmentation
5 Pension Risk Transfer
Page 14
SI Case Study Background
Auto-decision via RxRules
Risk Score as of time of underwriting
Have deaths on issued and declined cases
Page 15
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 > 4.0
# of
Liv
es
Risk Score
Risk Score Distribution by UW DecisionSI Case Study #1 - Hits Only
Lives - Issue Lives - Decline
SI Case Study – Distribution of Lives
Issue
Decline
Average Score (Hits Only)
Issue 0.96
Decline 1.52
Page 16
0%
100%
200%
300%
400%
500%
600%
0.00 ≤ x < 1.00 1.00 ≤ x < 1.50 1.50 ≤ x < 2.00 2.00 ≤ x < 3.00 3.00 ≤ x
Rela
tive
Mor
talit
y
Risk Score Range
Relative Mortality by Risk Score and UW DecisionSI Case Study #1 - (Hits Only)
Issue Decline
SI Case Study – Relative Mortality
Decline
Issue
Page 17
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 > 4.0
# of
Liv
es
Risk Score
Lives - Issue Lives - Decline
Thresholds can be adjusted to achieve desired business results.
Low ThresholdHigh Threshold
Page 18
02,0004,0006,0008,000
10,00012,00014,00016,00018,000
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 > 4.0
# of
App
s
Lives - Issue Lives - Decline
Some issued premium now gets declined
Equal amount of declined premium now gets issued
Set Risk Score threshold to issue the same amount of business.
02,0004,0006,0008,000
10,00012,00014,00016,00018,000
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 > 4.0
# of
App
s
New Issues New Declines
02,0004,0006,0008,000
10,00012,00014,00016,00018,000
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 > 4.0
# of
App
s
New Issues New Declines
Threshold
83%
Before Risk Score
75%
After Risk Score
Issued Cases Relative A/ESame amount of business issued
$4 Million increase in profit
9% Mortality improvement
Page 19
Set Risk Score threshold to maintain the same mortality A/E.
Some issued premium now gets declined
More declined premium now gets issued
02,0004,0006,0008,000
10,00012,00014,00016,00018,000
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 > 4.0
# of
App
s
New Issues New Declines
02,0004,0006,0008,000
10,00012,00014,00016,00018,000
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 > 4.0
# of
App
s
New Issues New Declines
02,0004,0006,0008,000
10,00012,00014,00016,00018,000
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 > 4.0
# of
App
s
Lives - Issue Lives - Decline
Threshold
$56.1 M
Before Risk Score
$66.0 M
After Risk Score
Premium IssuedSame mortality A/E
$2.9 Million increase in profit
18% More issued business
Page 20
Fully Underwritten Case Study
More difficult to have sufficient claims
Generally distribution of score and highlighted cases only
Page 21
Fully Underwritten Opportunities to Improve Risk Selection
Issued CasesStatus = Active or Death
# Lives # Deaths Total Face Claimed Face
Score > 2.0 1265 18 $1.2 B $37 M
Score > 3.0 204 7 $217 M $11 M
Page 22
Risk Score adds value to fully underwritten policies.
Issued
Declined
Issued policies that should have been declined
Page 23
Implementation considerations and options
Fully automated – SI and Final Expense
Accelerated Underwriting
Fully Underwritten
Reinsurance support
Page 24
Thank youEric Carlson, Life [email protected] 262-641-3537
Page 25
examone.com
Risk Identifier
Medical History-Driven Mortality Risk Assessment
SOA Underwriting Issues and Innovations
July 31st, 2017
Page 26
???
???
Potential Causal(Indirect Relationship) Mortality Risk
Correlation & Causation in Predictive Modeling for Mortality
ClinicalHistory
PrescriptionHistory
Medical Risk
Self-Disclosure
Tobacco Risk
Behavioral Risk
MVR’s
Criminal History
InsuranceTesting
Vocational Risk
Occupation
Socioeconomic Status
CreditHistory
ZipCode
Education Income
Personality Traits(conscientiousness, risk aversion)
???
Lifestyle
SocialNetworks
PurchaseHistories
Subscriptions
???
???
Known Causal(Direct Relationship)
Page 27
3 CONFIDENTIAL – For internal circulation only
ScriptCheckPrescription Histories and Mortality
Page 28
4 CONFIDENTIAL – For internal circulation only
Leveraging the most comprehensive prescription history dataExclusive access to Pharmacy Benefit Manager data
ExamOne works with the largest Pharmacy Benefit Managers in the country
Exclusive relationship with the 2nd largest PBM in the country, OptumRx
Includes data from Catamaran acquisition Expected lift in hit rates of 5-10 points
40% of all maintenance drugs are ordered directly from PBMs
Continues to be one of the most reliable tools to protect from applicant non-disclosure
PBM
Page 29
5 CONFIDENTIAL – For internal circulation only
Selected ScriptCheck univariate risk profiles
Drug Class Male Mean Age Cotinine+ Hazard Score Risk IQ 99
Insulins 57.3% 53.2 8.0% 328.7 9.9%
Antihypertensives 60.0% 58.9 5.2% 325.8 8.8%
Smoking deterrent 62.8% 45.8 51.6% 134.4 1.1%
Acne agents 47.5% 32.6 4.2% 111.3 0.5%
Topical hypertrichotics, eyelashes* 2.0% 49.2 5.2% 88.8 0%
Demographics and Risk IQ Scoring
*N=101
Page 30
6 CONFIDENTIAL – For internal circulation only
Abnormal flags
Ordering Physician
160M+ Unique lives
Test results and
reference ranges
45% Repeat encounters
>400M encounters
50% of hospitals
QuestCheck™
Quick access to physician ordered laboratory testing results
Verifies self-reported medical disclosure
Reduce costs
70% of decisions based off lab results
Leveraging Quest Diagnostics extensive clinical laboratory database
Page 31
7 CONFIDENTIAL – For internal circulation only
Ordering Account Specialty Applicants % HitsGENERAL/FAMILY PRACTICE 1140 43.1%INTERNAL MEDICINE 633 23.9%OBGYN 617 23.3%OTHER HOSPITAL 138 5.2%UROLOGIST 101 3.8%MULTI-SPECIALTY GRP PRACT 90 3.4%ENDOCRINOLOGIST 77 2.9%CARDIOLOGIST 66 2.5%GASTROENTEROLOGY 66 2.5%RHEUMATOLOGIST 44 1.7%SURGEON 40 1.5%NEUROLOGIST 32 1.2%ONCOLOGIST 24 0.9%HEMATOLOGIST 13 0.5%GERIATRIC 11 0.4%NEPHROLOGIST 11 0.4%PRISON/JAIL 1 0.04%
Selected account specialties
Page 32
Cardiologist diagnoses adrenal tumorApplicant B: 32-year-old male Account name: BMG Stern CardiologyAccount specialty: CardiologyPanel(s) ordered: Catecholamines,Vanillymandelicacid, metanephrinesResults of note: Elevated adrenal hormones suggest pheochromocytomaICD Codes:
• 786.50 – Chest pain, unspecified• 401.9 – Essential hypertension• 250.0 – Diabetes Mellitus w/o complications• 785.1 – Palpitations• 786.05 – Shortness of breath
Page 33
9 CONFIDENTIAL – For internal circulation only
Risk Identifier
Page 34
10 CONFIDENTIAL – For internal circulation only
Risk Identifier scoring: Prescription history risk evaluation scoring
268
589
139
116
Individual drugs scored
HIC3 therapeutic classes scored
Drug-to-drug interactions scored
Class-to-class interactions scored
1,112Total variable inputs
Page 35
11 CONFIDENTIAL – For internal circulation only
Model Definition Hit rate
Complete Blood Count (CBC)
Unique and only available to Quest Diagnostics 19 component tests:
White blood cell count Hemoglobin Other cell type counts (neutrophils, monocytes, etc.) Red cell distribution width
38%
Comprehensive metabolic panel (CMP)
18 component tests: eGFR BUN Protein panel (ALB, PROT, GLOB) Glucose LFTs (AST, ALT, ALP)
33%
Lipid
6 component tests: Total cholesterol HDL cholesterol Triglycerides CHOL/HDL ratio LDL cholesterol Non-HDL cholesterol
29%
Risk Identifier scoring: 5 interacting component models using QuestCheck
Page 36
12 CONFIDENTIAL – For internal circulation only
Model Definition Hit rate
Order
Applicant Quest Check history evaluated for presence or absence of 305 common clinical panels.
Most common panels: CBC CMP Lipid panel (Cholesterol, HDL, Triglycerides, etc.) Thyroid stimulating hormone (TSH) Hemoglobin A1c
81%
Physician specialty
Applicant Quest Check history evaluated for presence or absence of 26 common ordering physician specialties
Most common specialties: General/Family Practice OBGYN Internal Medicine Other Hospital
81%
Risk Identifier scoring: 5 interacting component models using QuestCheck
Page 37
13 CONFIDENTIAL – For internal circulation only
Experience
75 Expertise
Structure of a Risk Identifier scoreCombining analytics insights with expert knowledge
Quantitative score Color score
Derived from a statistical analysis of historical mortality
Equal to the % increase in mortality risk (debit/credit)
May be unavailable for some rare drugs or tests (risk unquantifiable)
Categorization of risk based on expert knowledge and clinical literature
Qualitative (Green-Yellow-Red)
Will be available even for rare results
Page 38
14 CONFIDENTIAL – For internal circulation only
Predictive value by available data sources
204
336
633
761
1120
0
200
400
600
800
1000
1200
QuestCheck Ordersand Specialties
ScriptCheck Only QuestCheck Orders,Specialties, and CMP
ScriptCheck, allQuestCheck Panels
Risk IQ (full labpanel)
Haz
ard
Scor
e (M
edia
n=10
0)
Data Sources
Relative Risk of 99th Model Percentile
Page 39
15 CONFIDENTIAL – For internal circulation only
Thank You!
Page 40
Dynamic Risk SelectorSM
Dianne Schuetz
July 31, 2017
VP, Business Initiatives, U.S. Markets
Accelerated, nimble underwriting
Page 41
2
A Customer’s Perspective
Page 42
3
What Accelerated Underwriting Enables You To Do
Issue near-fully underwritten retail
rates
Identify applicants who qualify for fluid-
less underwriting
Page 43
4
Acceleration Solution Overview
Full ApplicationTele-Interview & Age/ Amount Requirements
Does applicant meet the requirements?
Does applicant meet the requirements?
Apply Full Underwriting with Fluids
Accelerated offer
No Fluids Preferred
No Fluids Super Preferred
Page 44
5
Acceleration Solutions Can Vary
Different designs, approaches, and definitions in the market
Most target elimination of some underwriting elements in the age/amount grid
Can result in a wide range of mortality outcomes
Page 45
6
Accelerate and balance
mortality and throughput rates
Use keys that fit to meet the
challenge Find the right balance
Acceleration Challenge
Page 46
7
Dynamic Risk Selector Addresses These Challenges
Applicants want …
• Affordability
• Better process
Carriers want …
Reliable underwriting evidence
New forms of evidence and advanced data analysis techniques are starting to balance the needs of both groups
Page 47
8
Dynamic Risk Selector
Path to new data and models
Automated acceleration solution built by RGA mortality, underwriting, and data science experts
Streamlined access to expert acceleration rules, evidence, and predictive models
Page 48
9
Dynamic Risk Selector Components
Predictive Model
TransUnion TrueRisk®
Life
Accelerated Underwriting
Rules
Underwriting acceleration rules Includes application data, disclosures, MIB,
MVR, Rx Targeted rules support the model
Predictive Model Identifies preferred classes for applicants eligible for
acceleration Custom calibration to control mortality slippage
TrueRisk® Life leveraged in addition to
other underwriting evidence to enhance
protective value
Page 49
10
Process Flow
Dynamic Risk Selector
(Accelerated Rules, Predictive Model)
Client System or RGA Portal
TransUnion TrueRisk®
Life
MIB
MVR
Rx
Future Data Sources
Case Data
Evidence Result/Accelerated Decisions
Page 50
11
Decision Output
Page 51
12
Risk Selection Methodology Is Continuously Validated
Parallelunderwriting
Investigatedifferences
Quantifyeffect
Refine andrevise