Critical Illness Insurance: An Introduction to Pricing. Chris Jewson Head of Protection Pricing, UK & Ireland

Transcription

1 Critical Illness Insurance: An Introduction to Pricing Chris Jewson Head of Protection Pricing, UK & Ireland 22 October 2015

2 Agenda The Evolution of Critical Illness Pricing Setting The Long-Term Trend Assumption Product Innovation Adding New CI Conditions Severity-based Payments Multi-pay Critical Illness Lessons Learned Recent Research Risk Management Controls 2

3 The Evolution of Critical Illness Pricing 3

4 The Evolution of Critical Illness Pricing 1980 s: The Early Pioneers Variety of product designs, influenced by pricing uncertainty and with support of reinsurers Accelerated term assurance Typically 6 or 7 CI conditions Flexible Whole of Life (Universal Life) Unit-linked with investment underpin Mortgage endowment assurance Pays full sum assured on expiry, if not already claimed With-profits allows for retrospective pricing via bonuses 4

5 The Evolution of Critical Illness Pricing 1990: Dread Disease Cover, An Actuarial Perspective Research paper presented in January 1990, more commonly known as Dash & Grimshaw after its authors Recognised the lack of reliable data on post-ci mortality from non-ci causes, which is needed to price an accelerated term assurance Proposed the heroic assumption that post-ci mortality from non-ci causes is the same as the mortality of healthy lives Accelerated incidence rate = i x + (1 - k x ) q x i x = critical illness incidence rate q x = mortality rate k x = proportion of deaths that are due to a critical illness 5

6 The Evolution of Critical Illness Pricing 2000: A Critical Review Actuarial research paper presented in March 2000 Developed the first UK Critical Illness Base Table (CIBT93) from population data Also presented an analysis of insured lives experience for the period

7 The Evolution of Critical Illness Pricing 2003: Reinsurers get Cold Feet Reinsurers started to worry about future trend and reduced or withdrew capacity at guaranteed rates Guaranteed capacity reserved for key clients & distribution channels Reviewable rates for everyone else GE continued to offer guaranteed CI in volume Prudent margins: priced for >3% p.a. deterioration in experience Guaranteed rates ~40% higher than reviewable Significant profits 7

8 The Evolution of Critical Illness Pricing 2006: Exploring the Critical Path Actuarial research paper presented in December 2006 A major report (over 200 pages) with substantial commentary on individual conditions, trends etc. Developed the second UK Critical Illness Base Table (CIBT02) from population data Presented as a work-in-progress and never fully completed 8

9 The Evolution of Critical Illness Pricing 2011: AC04 Standard Tables The first full UK industry tables using insured lives data Graduated from the industry experience submitted by CMI members Covers experience during More stable mix of contributing offices than previous industry data Still immature by age and duration 9

10 The Evolution of Critical Illness Pricing Data Challenges Consistent reporting of claim dates Date of Diagnosis / Reporting / Claim / Settlement Consistent reporting of claim amounts Mortgage business Child CI, other part payments, proportionate settlements Cause of claim Lack of consistency, free-form entries Recording of deaths caused by CI Incurred But Not Settled (IBNS) Can be lengthy delay from claim to settlement 10

11 The Evolution of Critical Illness Pricing Industry Study Challenges Mix of products has changed Old: Endowments, light underwriting New: Pure protection (term assurances), stringent underwriting Mix of contributing offices has changed Different underwriting standards Different distribution models Inconsistent from one industry study to the next 11

12 The Evolution of Critical Illness Pricing Industry Study Challenges 12

13 The Evolution of Critical Illness Pricing History of Industry Studies 2000 CIBT93 Graduated population tables insured lives experience insured lives experience 2006 CIBT02 Graduated population tables 2007 CIIT00 Unofficial GenRe graduation of CMI data 2011 AC04 First official graduated tables based on insured lives experience insured lives, further analysis by major cause of claim 13

14 The Evolution of Critical Illness Pricing Comparison of UK Data Sources Data size Reinsurer data Single client Small but directly relevant to insured population Includes rating factors such as age, gender, product, duration, cause of claim etc. Industry data CMI, pooled Larger database, most major insurers contribute data Analysis Own choice of methodology Industry body analyses the data and summarises the results, saving us time Problems - Small sample size - High degree of volatility when analysing a subset of data in more detail Credibility - No data for new business - Past claims experience may not be appropriate for new business Changes over time - Conditions covered, definition of illness, underwriting standards, corporate re-organisation - Data aggregated across multiple insurers - Limited to the published analyses only - Aggregated over insurers with different illnesses covered and underwriting standards - Mix of insurers and products - Definition of illnesses Population data Massive data set for England & Wales general population - No underwriting - Does not match insured definition of illness - Not insured lives population - HES: hospital visits, not 1st incidence - Lots of adjustments required for insured lives, underwriting, 1 st incidence etc. Can monitor trends as the population becomes more health conscious or obese 14

15 The Evolution of Critical Illness Pricing UK Selection Pattern by Duration and Risk Class Source: Continuous Mortality Investigation (CMI) 15

16 The Evolution of Critical Illness Pricing Cause of Claim, UK: Males Source: Continuous Mortality Investigation (CMI) 16

17 The Evolution of Critical Illness Pricing Cause of Claim, UK: Females Source: Continuous Mortality Investigation (CMI) 17

18 The Evolution of Critical Illness Pricing History of RGA Pricing Derived from population data with adjustments to estimate an insured population Insured lives experience starting to emerge at early durations Basket of major conditions calibrated to CMI industry data and own experience GLM analysis of insured data for established CI conditions with traditional methods for more recent conditions 18

19 The Long-Term Trend Assumption 19

20 The Long-Term Trend Assumption Influences on Trends Population Real changes in population morbidity Earlier detection of disease Medical definition drift Insured Lives Real changes in assured population morbidity Weakness in insured definition Insured definition drift Anti-selective lapsation 20

21 The Long-Term Trend Assumption Objective To estimate future incidence-rate trends for a standard acceleration critical illness product Particular emphasis on adverse scenarios appropriate to pricing of guaranteed business Not a tail scenario investigation 21

22 The Long-Term Trend Assumption Conditions Covered Heart Attack Stroke MS CABG Angioplasty Colorectal Cancer Pancreatic Cancer Stomach Cancer Prostate Cancer Bladder Cancer Lung Cancer Breast Cancer Ovarian Cancer Uterus Cancer Melanoma Non-CI deaths 22

23 The Long-Term Trend Assumption Who should we call for the crystal ball gazing? Options Pros Cons Academics Ahead of the game Narrow, unworldly Consultants, Practitioners CMO s Underwriters Broader view Understand insurance Understand insurance Stuck in the present, don t understand insurance Stuck in the present Not enough expertise Dr. Google Available 24/7 None 23

24 The Long-Term Trend Assumption Expert Sessions CMO s with relevant specialism Briefing packs provided Insured definition Chart of past trends (from HES, Cancer Registration Statistics) Suggested rationale for past trends Telephone discussions, strictly moderated, 1 hour per condition Discussions must reach consensus on future trends, which is documented as the Expert Opinion Statement 24

25 The Long-Term Trend Assumption Expert Opinion Statement Level and direction of recent population trends, with rationale Projected trend for next 10yrs and subsequent 10yrs, with rationale: No change Slightly up/down Sharply up/down Potential shocks in next 10yrs and subsequent 10yrs, with rationale 25

26 The Long-Term Trend Assumption Expert Opinion: Shocks A shock is an event which the experts expect to occur in the future with some level of likelihood A shock gives rise to a one-off increase in incidence rates above the current trend, to a sustained higher level A shock is described by three parameters: The assumed date of occurrence The assumed likelihood of occurrence The assumed impact on incidence rates A shock is NOT an unpredicted or random event! 26

27 The Long-Term Trend Assumption Converting Expert Opinion Statements into Numbers Actuaries interpreted the choices into numbers, by age band, starting with the trend over most recent 5 year period Expert opinion statements slightly up etc. converted into numeric trend assumptions for 20 years, in blocks of 5 years Further adjustment for expected differences between general population and insured population, using insured v. population incidence trends as a clue For non-ci deaths, used trends in available death by cause data 27

28 The Long-Term Trend Assumption Converting Shocks into Numbers Starting point is Expert Opinion for time, likelihood and impact Adjust for deemed differences between population and insured lives Socio-economic Effect of underwriting Grade impact variation in 10-year age groups Further discussion and refinements following discussions with claims team 28

29 The Long-Term Trend Assumption Significant Future Trends Predicted Heart Attack, Stroke Deteriorate for 5-10yrs at young ages, Improve at older ages and after 10yrs Multiple Sclerosis Prostate cancer and melanomas Bladder, stomach and ovarian cancers Breast cancers Deteriorate, although slowing Rapid deterioration, flattening over time Improve First deteriorate, then level off and improve Uterus cancer Deteriorate 29

30 The Long-Term Trend Assumption Trends in Other Critical Illnesses Other Diseases Assumed flat future trend Based on past trends for all other cancers Non-CI Deaths No obvious past trend below age 60 (males) or 50 (females) For the older ages, a small future improvement trend 30

31 The Long-Term Trend Assumption Significant Future Shocks Predicted Stroke Exploitation of weak insured definition for symptoms (and for MS) Inclusion of some transient ischaemic attacks (TIAs) 20% likelihood, up to 20% impact at youngest ages for weak definition 10% likelihood, up to 20% impact at youngest ages for TIAs Heart Attack Risk of having to accept lower clinical Troponin threshold 0.1 instead of 1.0 in the insured definition Various Cancers Earlier detection of malignancies; but earlier detection of pre-malignant cancers would be beneficial and work in our favour 31

32 The Long-Term Trend Assumption Trend Scenarios: Principles for Best Estimate Best Estimate (Reviewable Premium Business) Include <100% of the shock impacts Allow for improvements where expected for some conditions 32

33 The Long-Term Trend Assumption Trend Scenarios: Principles for Guaranteed Business Guaranteed Premium Business Aim to be P% (?) certain of claims costs within pricing Various scenarios from optimistic to pessimistic Cautious interpretation of past trends Reduced allowance for long-term improvements in some for conditions Include >100% of the shock impacts Keep a sense of balance: there are beneficial environmental and technological drivers, as well as adverse ones Shocks limited to small set of predicted causes, not a 1 in 200 event or wildcard unknown 33

34 The Long-Term Trend Assumption Combining Trends and Shocks for All Conditions Trends and Shocks weighted appropriately to produce a two-way table by age and duration for each risk class Further variations by age, smoker status, policy term 34

35 Product Innovation 35

36 Marketing CMO Product Innovation Process and Stakeholders Insurer Insurer s proposed definition Refine definition and price Insurer challenge Definition Review Refine Price Pricing Refine Definition 36

37 Product Innovation Pricing New Conditions Identify what data sources are available Often no insured data Population data (eg. HES) use CMO s, Dr. Google etc. Identify the most relevant ICD codes Subjective adjustments Under-reporting or multiple treatments for the same patient Severity of insurance definition v. medical definition Insured population v. general population Overlaps with existing cover 37

38 Product Innovation CI in South Africa Began selling CI in 1985 Sophisticated market Complex CI products and little standardisation Highly competitive between providers Driven by definitions numbers of conditions covered, and the scope of each definition Insurers are pushing the boundaries to differentiate Sold by broker as a rider to death or standalone cover 38

39 Product Innovation CI in South Africa: Plain Vanilla to The Works CI in 1985 CI in

40 Product Innovation CI in South Africa Heart Attack, Stroke, CABG, Cancer Accelerated (part of Life policy) > 50 impairments covered Stand Alone and Accelerated Scaled severity levels, paying out more for more severe disease 40

41 Product Innovation: Severity-based Payments Key Features of CI Products South Africa Australia Singapore Limited Premium guarantees (10years) Accelerated rider or Standalone Fully reviewable rates Premium guarantees 50+ conditions 50+ conditions 37 conditions Reinstatement / buyback Staged / multi-pay Early Stage Reinstatement / buyback Staged /multi-pay (developing) Reinstatement / buyback Staged / multi-pay Early Stage Whole of Life No WOL No WOL 41

42 Product Innovation: Severity-based Payments Arguments for Scaled CI Better meets client needs Matches payment to disease severity Improved public image. fewer declined claims More resilient to medical advances as long as definitions well controlled Insurable interest Affordable insurance Multiple events Windfall payments on non-scaled products: Is it fair to pay someone who has had a mild heart attack and back at work next week the same as a paraplegic? 42

43 Product Innovation: Severity-based Payments Challenges with Scaled CI Product complexity Consumer understanding Agent sells the 100% level Marketed as being not as comprehensive Pricing statistics Country variations in stage distribution? Stage progression? Multiple claims? Impact of future screening? Lacking credible experience or population data 43

44 Product Innovation: Severity-based Payments Example: Cancer Definition Cancer is a malignant tumour positively diagnosed with histological confirmation and characterised by the uncontrolled growth of malignant cells and invasion of tissue. The term malignant tumour includes leukaemia, lymphoma and sarcoma. The following conditions are excluded from this definition: -All cancers in situ and all pre-malignant conditions -All tumours of the prostate unless histologically classified as having a Gleason score greater than 6 or having progressed to at least clinical TNM classification T2N0M0 -All skin cancers, other than, malignant melanoma that has been histologically classified as having caused invasion beyond the epidermis (outer layer of skin) Non-Tiered 100% The cancer must require treatment by surgery, radiotherapy, or chemotherapy. Tiered (except Prostate, leukemia and lymphoma) 25% TNM Stage I malignant tumour 50% TNM Stage II malignant tumour 75% TNM Stage III malignant tumour 100% TNM Stage IV malignant tumour 44

45 Product Innovation: Severity-based Payments Scaled Cancer Pricing Stage IV Stage III Progression Cancer Stage II Stage I 45

46 Product Innovation: Severity-based Payments What is required to price Scaled CI Initial Incidence rates for each Tier Often estimated from non-scaled incidences Early pay / Smaller tiers not covered by Non-scaled product need to estimate incidence Assumptions on Progression to higher Tiers proportion of cases How long will it take to progress Adjust for sum assured paid out at each stage Scaled CI typically around 80% - 85% of Non-Scaled CI Rates 46

47 Product Innovation: Multi-pay Critical Illness On standalone CI can claim more than once for unrelated conditions Product paying up to 5 times the sum assured or more on multiple CI events Usually CI events are grouped by related conditions and only 1 payment per group is allowed Require waiting period between claims to be considered unrelated 47

48 Product Innovation: Multi-pay Critical Illness What is required to price Multi-pay CI Incidence Rates Base single event incident rates by condition, age, gender, smoking habit Discount rates Lapses (pre and post first claim) Issue Age and Policy Term Risk Relativities Probability of Second/Subsequent Event given First Event has occurred Product Groupings to allow for implicit price adjustment Survival Probabilities Initial Extra Mortality over one year waiting period following the first event Extra mortality by duration for further four years following the first event Expenses, capital, profits etc. 48

49 Lessons Learned 49

50 Lessons Learned Evolution of Pricing Basis Smoker/non-smoker differential initially too wide Early incidence rates were derived from aggregate population data with smoker/non-smoker differentials derived from mortality data Now understand that smoking diseases mainly occur at older ages Females <40 years old, most claims are cancers unrelated to smoking Evidence of greater anti-selection than term life insurance Reduced/shorter selection discount (typically 10-20% for 2 years max.) Sum Assured is less significant than expected Channel differentials are less pronounced than expected But client-specific factors are more evident 50

51 Lessons Learned Variation in Duration 0 Selection Discount by Insurer 51

52 Lessons Learned Population Trend (Male aged 40-60, , England & Wales) 52

53 Lessons Learned Product Developments and New Conditions Insurer and reinsurer agree timelines at outset Requires careful management of all stakeholders Marketing Underwriting Claims Pricing May require multiple iterations before settling on the final product specification Package price easier to determine than individual conditions 53

54 Recent Research 54

55 Recent Research The Hospital Episodes Statistics (HES) Dataset The full HES dataset covers 1989/90 to 2009/10 Unique patient identifier from 1997/98 to 2009/10 and ICD10 codes Potential uses include: Pricing new CI conditions where insured data is not available More accurate adjustments for repeat visits Understanding correlations and overlaps between conditions Socio-economic studies to assist in predictive modelling? No longer available due to privacy concerns 55

56 Recent Research The Hospital Episodes Statistics (HES) Dataset Patient Identifier Basic Patient Information Basic Episode Information Diagnosis Information 47m unique patient identifiers Age, gender Date started, date finished, admission method, current status etc. Up to 20 different diagnoses per patient Procedure Information Geographic Information Up to 20 different operations, with date of operation Postal district, local super output area, deprivation index ranking 56

57 Risk Management Controls 57

58 Risk Management Controls We are all risk managers Product Design Limit volumes of single-condition cover Maximum term (for guaranteed rates) Alignment of Interest Insurer retention, if using reinsurance Pricing Mis-estimation Increased availability of insured data Regular experience studies eg. when quoting for new business Trend risk Future-proof definitions, regular review CMO approval 58

59 Thank You Any Questions? 59

60 2015 RGA. All rights reserved. No part of this publication may be reproduced in any form without the prior permission of RGA. The information in this publication is for the exclusive, internal use of the recipient and may not be relied upon by any other party other than the recipient and its affiliates, or published, quoted or disseminated to any party other than the recipient without the prior written consent of RGA.