Predictive Analytics Milliman Risk Score and Hannover Re. Eric Carlson & Sean Conrad November 18, 2016

Transcription

1 Predictive Analytics Milliman Risk Score and Hannover Re Eric Carlson & Sean Conrad November 18, 2016

2 Agenda Why Risk Score What is Risk Score Retrospective Study Results

3 The Future of Underwriting Increasing Decreasing Electronic requirements (Rx, MIB, MVR, Medical, Credit ) Decision engines driven by data Predictive Models Automation APS Labs Interviews Cycle times Costs Better Customer Experience

4 How does Rx work? Clients Health Plan PBM Data Sources Obtain the authorization Submit the query Review the results Clearing House Retail Pharmacy 4

5 What is a rule engine? Data Input Rx Application data MIB, MVR Medical Data Credit Data Other? RxRules Underwriting Decisions 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 5

6 RxRules Timing and Duration matter Corticosteroids 105% relative mortality Low Frequency / Duration High Frequency / Duration 99% 201% Corticosteroids are very common among insurance applicants

7 RxRules Drug Combinations matter Spironolactone 209% relative mortality With 2 out of 3 of: Thiazide Diuretics (102%) Ace / Angio II (ARBS) (116%) Beta Blocker (122%) 328% Without 2 out of 3 of: Thiazide Diuretics (102%) Ace / Angio II (ARBS) (116%) Beta Blocker (122%) 166%

8 Milliman Risk Score RxRules-driven Predictive Model Predicts relative mortality of a life or group of lives Used in PopulationRx Delivers results within RxRules

9 The Milliman Risk Score is built on RxRules. 6,800 GPI codes Hundreds of RxRules 1.27 Milliman Risk Score 200,000 NDC codes

10 Milliman s Risk Score effectively predicts mortality Milliman Mortality Study: 13M lives, 8M Rx hits, 230K deaths

11 Hannover Re Analysis Independent Validation of Milliman Risk Score

12 Milliman RxRules & Risk Score Model Validation Over 13 million records representing IntelliScript Rx queries Entry years Exposure period ,000+ deaths Over 52 million exposure years 12 6 lines of business: Life Health LTC Final Expense Med supp DI Milliman included risk score, hit status, and most severe drug (red/yellow/green)

13 Milliman RxRules & Risk Score Total Model Results High level results consistent with expectations Lower Risk Scores = Lower Mortality This relationship was consistent regardless of whether we looked at all lines of business (LOB) together or Life only 13

14 Milliman RxRules & Risk Score By Drug Severity Relative Mortality by Drug Severity Green Yellow Red 600% 500% 400% 300% Predictive value of risk score is consistent regardless of whether most severe drug is red/yellow/green Multivariate nature of the risk score is better way to stratify risk 200% 100% 0% Milliman Risk Score 14

15 Milliman RxRules & Risk Score Distribution of IntelliScript Risk Scores - Life Only by Drug Severity Distribution of Risk Scores - Life Only by Drug Severity % of Exposure 18% Green Yellow Red 16% 14% 12% 10% 8% 6% Red/yellow/green mix varies by risk score Not all green have low risk scores and not all red have high risk scores 4% 2% 0% Milliman Risk Score 15

16 Milliman RxRules & Risk Score Benefits & Potential Use Cases Quick, efficient, and consistent method for evaluating Rx Synthesize complex aspects of Rx history (# of refills, who prescribed?, etc.) Triage Rx info identify clean Rx histories vs. those to review in more detail Increase acceptance rates and/or decrease mortality expectation

17 Case Study Background Simplified Issue writer Cases are either issued or declined no rate classes Decisions are almost exclusively automated Used application, Rx / RxRules, plus MIB and MVR as decision elements 230,000 lives 1,200 deaths deaths for issued policies AND declined policies

18 Case study Risk Score distribution Accepted Cases Declined Cases

19 Relative mortality by decision

20 Thresholds can be adjusted to achieve desired business results Low ThresholdHigh Threshold

21 Case Study Management Option #1 Set Risk Score threshold to issue the same amount of business Some issued cases now get declined Equal number of declined cases now get issued # of Apps 60,000 50,000 40,000 30,000 20,000 10,000 0 Threshold # of Apps New - Accepts New # Declines of Apps - Decline Issued Cases Relative A/E Before Risk Score After Risk Score 76% 70% Same amount of business issued 8% Mortality Improvement $22 Million increase in profit

22 Case Study Management Option #2 Set Risk Score threshold to maintain the same mortality A/E Some issued cases now get declined More declined cases now get issued # of Apps 60,000 50,000 40,000 30,000 20,000 10,000 0 Threshold # of Apps New - Accepts New # Declines of Apps - Decline Number of Cases Issued Before Risk Score After Risk Score 174K 199K Same mortality A/E 14% More issued business $37 Million increase in profit

23 Integration with RxRules 1.27 Milliman Risk Score RxRules Overall Results Risk Score Exceeds Threshold of 1.20 Hypertension First Line Lipid Rx Run standardized rule set to calculate Risk Score Pass Risk Score to client calibrated RxRules engine Determine underwriting decision based on most severe of RxRules decisions and Risk Score

24 Contact Info Eric Carlson, FSA, MAAA Life Actuary (262) Sean Conrad, FSA, MAAA, FIA, CFA VP & Actuary (704) Proprietary and Confidential.