The Financial Health of a Wellness Program Brent Jensen FSA, MAAA Consulting Actuary 1 November 2017
The Financial Health of a Wellness ProgramBrent Jensen FSA, MAAA
Consulting Actuary
1 November 2017
2
Data Provides Direction
3
Data GOOD Data Provides Direction
4
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Background
▪ National employer – grocery and wholesaler chain
▪ Low margin industry
▪ Competitive benefits
▪ Focus on Employee Health
▪ Wanted a wellness program
▪ Improved employee health
▪ Potential H&W savings
▪ Wanted to give an incentive to participate – credit savings to employees
5
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Wellness Vendor
▪ Considered simplicity of program for employees.
▪ Incentive structure for qualifying for lower employee contributions.
▪ Vendor guarantees related to the savings generated by the program
▪ Used propensity score matching methodology
6
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Milliman Review of ROI
▪ Independent review of vendor ROI plus estimating ROI with different methodology
▪ Data expertise in risk adjustment using Milliman Advanced Risk Adjustment (MARA)
▪ Access to integrated claims and wellness data via MedInsight
▪ ROI tied results to actual medical costs for entire group
7
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ How to use claims data to measure the ROI of the wellness program?
▪ Need a way to accurately compare participants with non-participants
▪ Claims data housed in MedInsight
▪ Milliman Advanced Risk Adjuster (MARA) run on detailed claims data
Two column list
8
Participants Non-Participants
–Typically Higher Cost
–Higher Risk
–Lower Participation Rate that Expected
–Typically Lower Cost
–Lower Risk
StrategyProvider
SelectionThe Process Results Action Steps
9
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Two drivers of savings
▪ Reduced Trend (short-term)
▪ Reduced Overall Risk – Health Risk Reduction (long-term)
10
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
2016 Risk Adjusted Allowed Trends by Wellness Activity
2016 ROI Analysis
No Condition
Management in
2016
Condition
Management in
2016
No Lifestyle or
Condition
Management in
2016
Lifestyle
Management in
2016
Total
2016 Member Months 119,424 2,148 44,616 74,808 121,572
2015 Rx Risk Score (Normalized) 0.980 2.094 0.906 1.024 1.000
2016 Rx Risk Score (Normalized) 0.998 2.073 0.913 1.048 1.017
2015 Allowed PMPM 479 1,446 436 505 497
2016 Allowed PMPM 527 1,240 511 536 539
2016/2015 Trend 9.9% -14.2% 17.2% 6.1% 8.6%
2015 Risk Adjusted Allowed PMPM 489 690 481 493 497
2016 Risk Adjusted Allowed PMPM 528 598 560 512 531
2016/2015 Risk Adjusted Trend 8.0% -13.3% 16.3% 3.7% 6.9%
Risk Adjusted Trend Difference 0.0% -21.3% 0.0% -12.5%
Only includes members who were present all 2015-2016
11
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Reduced Trend Savings
12
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
Lifestyle Management Risk Shifting from 2015 to 2016 by Risk Score Bucket
% of Total Members 2016 Risk Score Bucket
2015 Risk Score Bucket a: Low b: MedLow c: MedHigh d: High Grand Total
a: Low 40% 7% 3% 2% 52%
b: MedLow 7% 9% 3% 2% 21%
c: MedHigh 3% 3% 4% 2% 12%
d: High 2% 2% 3% 9% 15%
Grand Total 52% 20% 13% 15% 100%
13
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
Lifestyle Management Risk Shifting from 2015 to 2016 by Risk Score Bucket
% of Total Members 2016 Risk Score Bucket
2015 Risk Score Bucket a: Low b: MedLow c: MedHigh d: High Grand Total
a: Low 40% 7% 3% 2% 52%
b: MedLow 7% 9% 3% 2% 21%
c: MedHigh 3% 3% 4% 2% 12%
d: High 2% 2% 3% 9% 15%
Grand Total 52% 20% 13% 15% 100%
Non Participant Risk Shifting from 2015 to 2016 by Risk Score Bucket
% of Total Members 2016 Risk Score Bucket
2015 Risk Score Bucket a: Low b: MedLow c: MedHigh d: High Grand Total
a: Low 52% 5% 3% 2% 63%
b: MedLow 6% 5% 2% 2% 15%
c: MedHigh 3% 2% 2% 2% 10%
d: High 2% 2% 2% 8% 13%
Grand Total 63% 14% 9% 14% 100%
14
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Reduced Overall Risk Savings – Health Risk Reduction
15
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Savings on a Per Participant per Month (PPPM) Basis
16
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Savings vs Expenses on a Per Participant per Month (PPPM) Basis
17
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Bottom-line ROI results
18
Data for a Wellness Program
StrategyProvider
SelectionThe Process Results Action Steps
▪ Year to year decision of how to adjust and continue
▪ Effectiveness of wellness program, from independent source
▪ Condition management targeting to achieve potential savings from high cost participants
Thank you
Brent Jensen
TURNING DATA INTO ACTION:
EXPLORING INTEGRATED DATA
PREDICTIVE ANALYTICS
20
mailto:[email protected]
MONEYBALL MEASURES
TRADITIONAL MEASURES
MONEYBALL AND SABERMETRICS
Indicators of Offensive Success
Home Runs
Batting Average
Stolen Bases
RBI’s
On-Base %
Slugging %
Pitch Data
Expected Future Runs Scored in an inning given certain conditions. (1961-77 data set)
21
“People operate with beliefs and biases. To the extent you can eliminate both and replace them with data, you gain a clear advantage.” Michael Lewis, Moneyball: The Art of Winning an Unfair Game
WHAT IS INTEGRATED DATA PREDICTIVE ANALYTICS?
Applies mathematical & statistical techniques:
▪ Predict future outcomes.
▪ Improve ability to segment a population of data on the basis of
future probability or outcome.
▪ Provides more objective reasoning based on mathematical/statistical techniques.
▪ Leverages internal and external data.
▪ Improves understanding of risk characteristics that influence future outcomes.
An Objective
Approach to
Analyze Risk
▪ Segment low risk and high risk.
▪ Objective guidance for more efficient, consistent decisions and resource allocation.
▪ Creates opportunities to enhance traditional work processes.
A Tool for
Improved
Efficiency &
Consistency
▪ Model itself provides only a relative risk index.
▪ The rest of the value comes from operationalizing model results in business practices and actions.
A Tool…
Not a
Silver Bullet
HIGHER than
Average Severity
LOWER than
Average Severity
Supports important
business decisions
and yields
efficiency.22
INTEGRATED DATA PREDICTIVE ANALYTICS USE CASE
Data Cleansing and Aggregation
Create Variables
Develop Models
Build Reason Codes & Business Rules
Non-
Traditional
External Data
Sources
Non-traditional
data
introduces new
risk characteristics
into model.
INTEGRATED
DATA SOURCES
ADVANCED
DATA ANALYTICSBUSINESS
IMPLEMENTATION
Claims:
▪ Segment high-risk and low-risk
▪ Claim Triaged and Assigned.
▪ Resource Allocation.
▪ Develop consistent approach for managing claim.
▪ Cost-effective use of clinical resources and legal services.
“Right claim, right resource, right time.”
Data
Specifications
Traditional
Data
Sources
23
For claims that have never been prescribed or exceeded a specified range of Opioid
consumption, predict the likelihood of specific
consumption ranges in specific time ranges in the
future
▪ Test various dosage ranges and prediction periods
▪ Claim data set - @1 million annually
BUSINESS CHALLENGE
Individuals that are most likely to consume greater than 50
MED
Individuals that are most likely to consume greater than 90
MED
1. Understanding that risk is pervasive
across the enterprise, without
boundaries.
2. Using risk and business intelligence to
drive performance metrics and business
processes is critical.
3. Leveraging and harnessing the power
of “big data” to gain a competitive
advantage and improve decision
making.
KEY TRENDS FOR ORGANIZATIONS
27
The main requirement for predictive modeling process is data. This technique requires
extensive data mining.
WHAT SHOULD YOU DO?
Before you begin…
Effectiveness of any modeling is based on data integrity and quality.
Check internal resources to effectively manage data quality control.
The key is to make sure your data is
accurate before you begin!28