Making Predictive Modeling in Renewal Underwriting Work for You. Jeff Fluke Senior Consultant, Underwriting Services Reden & Anders

Size: px

Start display at page:

Download "Making Predictive Modeling in Renewal Underwriting Work for You. Jeff Fluke Senior Consultant, Underwriting Services Reden & Anders"

Linda Hicks
5 years ago
Views:

1 Making Predictive Modeling in Renewal Underwriting Work for You Jeff Fluke Senior Consultant, Underwriting Services Reden & Anders

2 Agenda 8:00 9:00AM Today s Renewal Approach Why use Predictive Modeling Benefits of Predictive Modeling Opportunities for Underwriting Integration Options Communication Ideas Case Study Q&A Ingenix, Inc. 2

3 Concerns with today s common renewal approaches Accuracy Many carriers base small group renewals on loss ratios even though they will agree that the loss ratio of an eight employee group is not credible Consistency Many carriers will have a medical underwriter estimate the ongoing claims; these estimates are rarely consistent between underwriters and sometimes will vary from day to day with the same underwriter Efficiency Many carriers will have a medical underwriter determine the diagnosis, prognosis, and projected ongoing amount for each large claimant. This process is frequently very manual and often involves going from one screen shot to the next to obtain the needed information Ingenix, Inc. 3

4 Why use Predictive Modeling in Underwriting Goal: set the right rate (improve accuracy) Determine underlying health risk of population Retain and attract good business Goal: set the right rate (improve consistency) Match premium revenue with expected costs promote stability and profit (consistency) Improve market/employer perception of ability to forecast and manage costs Increase productivity (improve efficiency) Value-added information produced on a systematic basis supports automation and standardization Value proposition Better information on health risk for individuals and groups can enhance the underwriting process Ingenix, Inc. 4

5 Why use Predictive Modeling in Underwriting Enhance the actuarial and underwriting process: Increase accuracy of forecasts for new and existing groups Improve market perception of ability to forecast and manage costs Improve efficiency and productivity of rating process Compliment or supplement existing tools Ingenix, Inc. 5

6 Predictive Modeling - Case Example Differentiating Between Members Patient A. Male, 50, diabetic Developed skin ulcers - last month Most recent HbA1c is 11.0; taken 9 months ago Documented hypertension, not refilling his prescription ER visit last month, also had increasing number of visits for the past 3 weeks and seen by 3 different specialists last week Prior Year s Cost $4,600 Patient B. Male, 50, diabetic Developed skin ulcers 9 months ago Most recent HbA1c is 6.3; taken 2 months ago Documented hypertension, refilling his prescriptions regularly No recent ER visit, also routine follow-up care 1 PCP and 1 Specialty visit in past 3 months Prior Year s Cost $5,500 Ingenix, Inc. 6

7 Case Example Predicted Risk Output for Patients A and B Predicted Risks Patient A Risk Score Patient B Risk Score The next 12 Months The next 3 Months Probability of an Inpatient Stay Predicted Cost (next 12 Months) Predicted Cost by Services 28 % 6 % $33,000 $10,500 Inp Out Rx Dr Dx Inp Out Rx Dr Dx 35% 25% 10% 20% 10% 5% 10% 35% 35% 15% Ingenix, Inc. 7

8 Mr. Wizard s Science Secrets: This Week Predictive Modeling Well Jimmy, the data goes in here, these lights flash on and off for a few minutes. We send the results to actuarial. After that, who knows? Let s go to a commercial. -- Don Herbert, TV s Mr. Wizard -- Ingenix, Inc. 8

9 How do I interpret the weights? A relative risk of 1.0 = the average person Therefore, a risk score of.70 means that the individual is only 70% as likely to use healthcare resources than the average person. A risk score of 37.0 means that the individual is 37 times more likely to use healthcare resources as the average person. Need to normalize scores and factors to the appropriate risk pool Ingenix, Inc. 9

10 Normalizing risk score to rating action Group Block of Business Relative risk score A/S factor Step 1. Normalize the group risk score to the block risk score (group rrs / block rrs) (.95 /.98 =.97) Step 2. Normalize the group A/S factor to the block A/S factor (group AS factor / block AS factor) (.90 / 1.03 =.87) Step 3. Develop the adjusted group risk score (#1 / #2) (.97 /.87 = 1.11) Adjusted group risk score: 11% higher than the block of business Ingenix, Inc. 10

11 Benefits of Predictive Modeling Streamline group renewal underwriting Automate large claim review process Improved data collection and case preparation More stable underwriting margins Automate reporting capabilities Improved communications with groups/agents Improved accuracy, consistency, and efficiency Ingenix, Inc. 11

12 Technical Approach Opportunities for Underwriting Small Groups (2 50) In states where a health status adjustment is allowed Increased accuracy area of greatest benefit Automate moving from a risk score to a rating action Medium Groups (51 150) Blended with historic claims to increase accuracy More credibility to the predicted risk vs. prior history Large Groups (150+) Some blending with historic claims can enhance accuracy Opportunity to determine risk drivers enhance account management function Ingenix, Inc. 12

13 Technical Approach Opportunities for Underwriting - Engaging Employers Better match premiums to risk fewer surprises More Transparent describe risk drivers Combine with Care Management programs based upon risk drivers Integration with current underwriting practices Predictive Modeling results must complement existing information, including prior experience, credibility assumptions, and other adjusters Ingenix, Inc. 13

14 Technical Approach Additional Underwriting/Actuarial Uses Measure risk for blocks of business Area Broker Product Watch trend/risk over time for book of business Proactive with future risk score Increasing/decreasing ability to change rating before impacting financial results Selection issues Monitor marketing, sales activities Ingenix, Inc. 14

15 Technical Approach Integration Options Added piece of data (macro and micro level) Confirm results from existing process/trends Support appeals Especially where predicted risk is less than experience Automated large claim reviews Enhanced ability to identify emerging claims Ease of researching groups and individuals Part of rating formula Revised credibility table Integrate into rating process Various levels of automation Ingenix, Inc. 15

16 Communication Ideas We are using current technology to improve our ability to assess risk and better match premium to claims We are using current technology to improve our understanding of medical risk of each renewing group This improved understanding will allow us to do a better job of setting rates that appropriately reflect the underlying medical risk for each employer This should increase our retention of lower risk groups and maximize renewal increases on higher risk groups The member level detail provided by the predictive modeling tool needs to be kept highly confidential No specific member or group risk scores should be communicated Ingenix, Inc. 16

17 Case Study: Integrating PM into Underwriting Process Integration with current underwriting practice Predictive modeling results must complement existing information, including prior experience, credibility assumptions, other adjusters. How can this be accomplished? Empirical Test Use different models based on prior experience and Impact Pro risk findings to simulate group premiums. Compare simulated rates with actual experience -- assess best models. Ingenix, Inc. 17

18 Case Study Details 2,557,137 members 85,166 groups 3 health plans Commercial population, mix of products Primarily non-elderly Different geographic census regions 30 months of claims and enrollment data ( ) Ingenix, Inc. 18

19 Case Study Details Group Size % of Members % of Groups Cost PMPM Risk Score % 73.8% $ and over Total 2.56 Million 85,166 $ Group size based on number of subscribers. Ingenix, Inc. 19

20 Case Study: Integrating Predictive Modeling into Underwriting Process Models Tested Model Description Features 1 Age/Sex Relative demographic risk for those members active as of the end of the base experience period. 2 1 year Experience Relative prior year s experience for group 3 Impact Pro Risk Impact Pro relative risk for group 4 Risk and Experience Combine Impact Pro risk and prior year s experience -- weighted Ingenix, Inc. 20

21 Predictive Accuracy Group R A/Sex Prior Cost Impact Pro 2 Tier credibility Group Size scenario using $50,000 threshold. Group R 2 describes the % variation in future costs across groups explained by a model. Ingenix, Inc. 21

22 Case Study: Integrating Predictive Modeling into Underwriting Process Weighting of Impact Pro Risk and Experience by Group Size Employer Group Size Weight for IPro Risk Weight for Prior Cost Ingenix, Inc. 22

23 Predictive Modeling is Very Powerful Information Ingenix, Inc. 23

Risk adjustment and the power of four

Risk adjustment and the power of four Ksenia Draaghtel, ASA, MAAA Diane Laurent For a long time, the healthcare industry has recognized the value of health status adjustments for predicting future healthcare