decision analytic products in us life insurance underwriting · proprietary and confidential | ©...

Proprietary and Confidential | © General Reinsurance Corporation

Decision Analytic Products in US

Life Insurance Underwriting

Thomas Ashley, MD, FACP

Vice President and Chief Medical Director


• CRL SmartScore, ExamOne Risk IQ

• Mine historical customer results of medical exam,

blood, urine

Decision Analytics Products

ACSW Dallas, TX | Thomas Ashley | November 7, 2014 2

Industry

Lab Vendors

• Synthesize results across many clinical literature studies

into unified mortality risk equation

BioSignia

• Generate risk prediction from consumer behavior data Deloitte


• Dataset of all lab customers who applied for insurance in

past 15 years

• Many millions of records with height, weight, blood

pressure plus results of blood and urine tests

• Social Security Death Master File to infer mortality outcome

• Construct integrated mortality risk prediction model

SmartScore and Risk IQ Common Threads



Risk IQ SmartScore

Method • Generalized linear model

• Created synthetic variables such

as ratio of test results

• Excluded special selective tests

• Univariate relationship for

each test

• Assigned relative risk along

each curve

• Adjusted for age/sex

• Summed variables

• Added score for special tests PSA,

NT-proBNP, HCV

Output • Integer score, 0-99

• Approximates %ile mortality risk

by age/sex

• Score normed to approximate

credits (better than standard) or

debits (substandard)

Explanation • Both vendors report subscores that estimate contribution of single tests

to total score

SmartScore and Risk IQ Differences



• Applicants only, no knowledge of underwriting result,

medical history

• Model inaccurate to extent that lab data duplicates known

medical risk (unless use model as substitute for other

underwriting)

• SSDMF incomplete

• Thus, each algorithm would look different if derived from

issued cases, adjusted for underwriting risk class, claims

Lab Models Common Weaknesses



• Mortality of unplaced cases is invisible

• Use SSDMF to infer deaths

• Comparison to in-force mortality experience

• Measure accuracy of SSDMF against Gen Re claims

SSDMF Accuracy

Byproduct of facultative unplaced analysis


In-force All deaths observed

Unplaced Incomplete reporting, but by how much?


SSDMF Accuracy

Age at Death


40%

45%

50%

55%

60%

65%

70%

75%

80%

85%

90%

0-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90+


• Claim analysis allows us to adjust for undetected deaths in

Facultative unplaced analysis

• Unclaimed property application of SSDMF

• Annuity surveillance

Implications



• Goal is to integrate typical preferred underwriting criteria

(ht, wt, bp, family history, cholesterol, MVR, occupation)

• Appended select lab tests (glucose, liver enzymes)

• Meta-analysis: digest clinical literature to derive

relationship between each parameter and mortality risk

• Synthesize results across many studies into unified

mortality risk equation

• Output normed to approximate mortality % 2001VBT

Biomedical


BioSignia


• Deloitte Consulting

• Ignore conventional underwriting evidence

• Mine electronic databases of consumer history

‐ Credit card purchases

‐ Warranty registration

‐ Survey responses

• Relate this profile to risk of disease and mortality

• Hundreds of parameters available for inclusion in model

• Construct unique model for each client company

‐ Choice of parameters to include / exclude

‐ Tune to customers of each company

Consumer Behavior



• Multiple criteria for preferred considered separately

distorts overall measure of risk

• Prediction from integrated model might outperform

conventional underwriting of each variable separately

‐ More efficient risk classification

‐ Less overlap among risk classes

‐ Recognition of interactions that represent different risk than

sum of the parts

Upside


Biomedical

• Faster, cheaper, automated underwriting without need for

blood, urine, exam

Deloitte


• Demonstrate that score

corresponds to mortality

experience

• Industry labs

‐ Published performance on own data

‐ Unpublished trials for individual customers

• BioSignia

‐ Obtained large experience study data with

underwriting evidence

Validation


Mo

rta

lity A

/E

SCORE

Industry Labs,

BioSignia


• Demonstrate that score corresponds to risk class

assignment from existing underwriting process

• Replication of underwriting action immediate – no need for

experience to develop or retrospective study

Validation


Conventional UW Class

Mo

de

l U

W C

lass

1 2 3

1

2

3

Deloitte


• Hypothesis

‐ Refine preferred / STD risk and reclassify more consistently

‐ Qualify more applicants or adjust prices for risk classes

• Demonstrate efficacy

‐ Direct company could implement it

‐ Reinsurer could reflect it in pricing

‐ Regulator / producer could accept it

• Lab vendors derived model from insurance applicants /

SSDMF

• How does it perform on underwritten population?

• Single company study lacks power to measure low

risk groups

Gen Re Lab Score Validation Project



• Measure performance of SmartScore and RiskIQ on

issued policies

Laboratory Mortality Risk Score Validation


Goal

• Risk score construction used applicants and SSDMF

• Performance on issued lives and observed deaths will differ

• Decision on effective use of a score needs

inforce experience

Rationale

• Assemble underwriting evidence

• Obtain RiskIQ and SmartScore

• Assemble mortality experience

• Compare mortality risk prediction to mortality experience

Process


Study Population Statistics


Lives Claims Maximum Duration Average Duration

1,211,741 2,348 8 Year

1.9 Year


65.5% 73.3%

78.1% 84.1%

95.2%

0%

20%

40%

60%

80%

100%

40-69 70-74 75-79 80-83 84-110

Diastolic Blood Pressure

Conventional Underwriting Criteria


64.3% 70.8%

83.0% 81.4%

101.6%

0%

20%

40%

60%

80%

100%

120%

80-110 111-118 119-123 124-132 133-198

Systolic Blood Pressure

69.5% 69.3% 74.3%

93.8%

118.9%

0%

20%

40%

60%

80%

100%

120%

140%

16-23 24-26 27-29 30-34 35-50

BMI

81.6%

69.6% 75.6% 73.8%

86.3%

0%

20%

40%

60%

80%

100%

53-169 170-192 193-210 211-237 238-622

Cholesterol


81.6%

69.6% 75.6% 73.8%

86.3%

0%

20%

40%

60%

80%

100%

53-169 170-129 193-210 211-237 238-622

Cholesterol

Cholesterol


68.5% 72.4%

78.1% 77.3%

98.4%

0%

20%

40%

60%

80%

100%

1.0-2.8 2.9-3.5 3.6-4.1 4.2-5.2 5.2-12.0

Cholesterol Ratio


0%

20%

40%

60%

80%

100%

120%

40-69 70-74 75-79 80-83 84-110

Diastolic Blood Pressure

20-40

41-59

60+

Age Bands


0%

20%

40%

60%

80%

100%

120%

140%

160%

80-110 111-118 119-123 124-132 133-198

Systolic Blood Pressure

0-40

41-59

60+

0%

20%

40%

60%

80%

100%

120%

140%

160%

16-23 24-26 27-29 30-34 35-50

BMI

0-40

41-59

60+

0%

20%

40%

60%

80%

100%

120%

1.0-2.8 2.9-3.5 3.6-4.1 4.2-5.2 5.3-12.0

Cholesterol Ratio

0-40

41-59

60+


Cholesterol Ratio Charts


68.5% 72.4%

78.1% 77.3%

98.4%

0%

20%

40%

60%

80%

100%

1.0-2.8 2.9-3.5 3.6-4.1 4.2-5.2 5.2-12.0

Cholesterol Ratio Group

0%

20%

40%

60%

80%

100%

120%

1.0-2.8 2.9-3.5 3.6-4.1 4.2-5.2 5.3-12.0

Cholesterol Ratio Group

0-40

41-59

60+


49%

64%

73%

87%

126%

57% 59%

71%

81%

115%

0%

20%

40%

60%

80%

100%

120%

140%

160%

1 2 3 4 5

ExamOne

CRL

Lab Score Mortality Correlation

21 ACSW Dallas, TX | Thomas Ashley | November 7, 2014


Comparison – Gender


0%

20%

40%

60%

80%

100%

120%

140%

160%

1 2 3 4 5

ExamOne

Female

Male

0%

20%

40%

60%

80%

100%

120%

140%

160%

1 2 3 4 5

CRL

Female

Male


Comparison – Age


0%

20%

40%

60%

80%

100%

120%

140%

160%

1 2 3 4 5

ExamOne

0-46

47-57

58-99

0%

20%

40%

60%

80%

100%

120%

140%

160%

1 2 3 4 5

CRL

0-46

47-57

58-99


• Multivariate score superior to single criteria

• Risk IQ and SmartScore identify highest risk

• High mortality rate sufficient to see at individual company

level

• Performance in low risk segments less striking but

meaningful

• Additional analysis–for participants only

- Greater detail, especially on low risk

- Additional stratification by gender, tobacco, duration,

underwriting risk class

Discussion



Visit genre.com for more info.

Thank you

Thomas Ashley, MD, FACP

Vice President and Chief Medical Director

decision analytic products in us life insurance underwriting · proprietary and confidential | ©...

Documents