Consumer Driven Health Plans Early Findings from the

Consumer Driven Health Plans: Early Findings from the Field and Future Directions Stephen T. Parente, Roger Feldman, Jon B. Christianson University of Minnesota Presented at the Wharton School, University of Pennsylvania May 7, 2004 Funded by the Robert Wood Johnson Foundation Health Care Organization and Financing Initiative For more information: sparente@csom. umn. edu

Presentation Objectives • Describe the CDHP business model. • Illustrate the mechanics of a CDHP using Definity Health as an example. • Provide an Overview of our RWJ evaluation of Definity. • Present current analysis results. • Opportunities and conundrums of CDHPs.

Issues Driving CDHP Creation ØPatients § § Dissatisfaction with provider access Patient incentives are to consume Limited choices of benefits and providers Combative relationship with managed care companies ØProviders § Loss of autonomy § Erosion of physician/patient relationship § Misalignment of physician reimbursement and incentives ØEmployers § Plan costs are increasing § Employees are not happy § Increase of employer administration burdens

CDHP Business Enablers – ‘Ready to Lease’ Components of Health Insurance: • • • – Electronic claims processing National panel of physicians National pharmaceutical benefits management firms Consumer-friendly health data web portals Disease management vendors Internet • Transaction medium for claims processing • 2 -way communication with members – ERISA-exemption • Lack of state oversight • Half the US commercial health insurance market is self-insured.

Early CDHPs in Operation – Definity • Concept developed in 1998, Funded in April, 2000 • Minnesota based • Clear first mover & dot-bomb survivor – Lumenos • Started in 2000 • Based in Atlanta • Harvard B-School inspired (Regina Herzlinger) – Destinty • Operating as Medical Savings Account model • In operation for 10 years in South Africa

Definity Health Component Details Health Tools and Resources $$ Health Coverage • Preventive care covered 100% • Annual deductible Definity • Expenses beyond the PCA Health Care • Nationwide provider access Advantage • No referrals required Health Tools and Resources • Care management program • Extensive easy-to-use information and services 1 2 Annual Deductible PCA Employer selects which expense apply toward the Health Coverage annual deductible. Paid out of employer’s general assets. Preventive Care 100% Health Covera ge Annual Deductible Personal Care Account (PCA) • Employer allocates PCA 1 • Member directs PCA • Section 213(d) “scope” • Roll over at year-end • Apply toward deductible 2 Web- and Phone. Based Tools

New RWJ-Funded Research Key Research Questions 1. 2. Is there an ‘adverse selection’ problem? Traditionally, adverse selection is defined as the situation when healthy individuals choose Definity leaving the sick in a traditional plan that will soon implode its premiums because of disproportionate share of sick individuals in the insurance pool. What is the impact on cost and utilization? Definity has been chosen as a response to rising premium prices in an attempt to make the consumer ‘drive the market’ be examining price variations and constraining their personal consumption, if possible.

Research Design – 2 Year study (11/1/2002 - 10/31/2004) – Six employers examined: • • • University of Minnesota, MN Medtronic, National Ridgeview Medical Center, MN Hannaford Bros, New England Welch-Allyn, Upstate NY (tentative) Raytheon (New England or South Atlantic firm) – Data collected • Claims data of all utilization for all health plan choices, pre (2001) and post (2002 -2003) Definity. • Employer info on flexible spending accounts and employee income • Survey information on Definity choices in 2002 & 2003 from U of M.

Early Results #1: Employee Choice of a Consumer Driven Health Plan in a Multi-Plan, Multi-Product Setting

Conceptual Model to Address Selection Question We use a choice model based on utility maximization, where utility is considered to be a function of personal attributes such as health status, health plan attributes such as price, and the interaction of price and health status, formally stated as: Uij = f(Zj, Yi, Xij), where, • i is the decision-making employee choosing among, • j health plan alternative choices, • Yi = employee personal attributes, • Zj = health plan alternative attributes and • Xij = interactions between alternative-specific and personal attributes, Xij. Follows methods used by Harris, Schultz and Feldman (2002).

Multivariate Analysis of Plan Choice – Focus on the University of Minnesota 2002 survey – Combine survey data with HR information including: • • • After tax income Contract type Age and Gender Location Medical premium choice set – Run Conditional Logistic Regression Model to predict the effects of premium price, employee characteristics and health plan feature preferences for early adopters of Definity compared to other health plans.

Health Plan Choices 1. 2. 3. 4. Health Partners: Staff model HMO with direct capitation contracting at a limited number of group practices. Patient Choice: A ‘Tiered-direct contracting’ descendent of Minnesota’s Buyers Health Care Action Group health benefit design experiment. Definity Health: Consumer-driven Health Plan Preferred One: Preferred Provider Organization

UPlan Options/Enrollment (2002)

Early UM Definity Experience Year 2002

Definity Age/Gender Distribution 2002 University of Minnesota

All Respondents Satisfaction with Plan

Health Plan Features Most Preferred

Econometric Issues • How to group health plan choices? • Ideally, estimate separate choice models for: 1. Single employees with no dependents 2. Families who have no other source of health insurance 3. Families who have multiple choices of health insurance • • • Practically, we can’t identify (2) and (3), so we combine single and family contracts into one choice model through the use of planinteracted dummy variables (Feldman & Schultz, 2001). Considered a nested logit, but the Definity next, if weighted, was not large enough. Correction for oversampling Definity and undersampling the other plans. Lerman-Manski correction was used obtain appropriate standard errors.

Impact of price, employee characteristics and health plan feature preferences on health plan choice

Continued: Impact of employee characteristics and health plan feature preferences on health plan choice

Results: Premium Sensitivity • Employees are sensitive to out-of-pocket premiums, and surprisingly, employees with chronic conditions are more premium-sensitive • If Definity raised its premium by 1% it would lose 4. 6 % of healthy single enrollees and 5. 4% of healthy families • 1% premium boost would cause 6. 9% of singles and 10. 7% of families with chronic condition to leave Definity • The results depend on 100% of the premium hike being passed along to the employee (i. e, defined contribution), as is the case for the UM

Results: Health Status and Other Employee Characteristics • Employees and families with chronic conditions prefer the PPO, but otherwise, there is no evidence of adverse selection Ø Having a chronic condition is associated with a 3. 2% increase in the probability of choosing Preferred. One vs. Health. Partners Ø Note that Preferred. One had the highest premiums ($189. 51 for single coverage and $448. 40 for family coverage per pay period), suggesting that the plan is experiencing adverse selection • Higher income employees chose Definity or Choice Plus, suggesting these plans may evolve as favorites of the ‘well-to-do’ • Older employees chose Preferred. One or Choice Plus

Early Results #2: Consumer-Driven Health Plans: Early Evidence about Utilization, Spending and Cost

Study Setting • Health plan choices by employees: – HMO, 2000 -2002 – PPO, 2000 -2002 – CDHP, 2001 -2002 • Variation in cost sharing by contract • Take-up of CDHP approximately 15%. • General caveat: Each of the six employers’ experience can be quite different due to: – – Alternatives offered Plan design Communications with employees Sponsor’s objectives for the plan

Presentation of Results • Results are limited to two groups of employees who worked for their firm continuously for three years (2000 -2002) where: 1. • • Employee chose the CDHP in 2001 and 2002. 2. Employee chose another health plan in 2001 and 2002. This limitation removed 40% to 50% of all employees from the analysis. Why make this limitation? We want to see both adoption and maturing impact of CDHP while controlling for prior spending. – – – 2000: Pre-CDHP experience controls for prior spending 2001: CDHP adoption year 2002: CDHP ‘maturation’ year

Statistical Approach • Used difference-in-difference approach • Generate unadjusted and regression-adjusted comparisons. • Regression adjustment based on two-part model • Regressors included: age, gender, baseline illness burden (ADGs), contemporaneous health shock, number of dependents, FSA election and income. • Subsequent tests for regression to the mean found the problem to be present, but not to a degree that would influence our results by the last year of our findings.

Baseline Demographics

Case-mix Differences based on a Weighted ADG Index

CDHP, HMO versus PPO PMPM Differences for Continuously enrolled sample

What was the ADJUSTED impact on provider and patient payment? NOTE: These are results from a restricted continuously enrolled sample of 50% to 60% of the total employee population and are not a reflection of the plans’ full PMPM expenditures. Also note: 1) Patient expenditures from the Personal Care Account (PCA) are included in the employer payment category. 2) Consumer payment reflects deductibles, copayments, and coinsurance expenses.

What was the ADJUSTED impact on provider & patient payment by different services? NOTE: These are results from a restricted continuously enrolled sample of 50% to 60% of the total employee population and are not a reflection of the plans’ full PMPM expenditures.

Was ADJUSTED service use different for CDHPs? NOTE: These are results from a restricted continuously enrolled sample of 50% to 60% of the total employee population and are not a reflection of the plans’ full admissions and prescription drug experience.

Distribution of CDHP Population by PCA Usage Levels Continuously enrolled population

Conclusions • The most important factor affecting choice is income. • The consumer drive health plan was not disproportionately chosen by the young and the healthy (for the U of MN population). • For the non-U of MN pop, in adjusted dollars, the patterns remains, despite the likely event that the CDHP experienced favorable expenditure selection. • Year 3 of CDHP experience will reveal if they can stem high cost growth trajectory from years 1 & 2.

Critical Caveat: Benefit Design will Drive Expenditure Results ‘Draconian’ benefit design for family contract: • $1, 0000 PCA, $4, 000 Deductible, 20% co-insurance after deductible Current industry standard design for family contract: • $2, 000 PCA, $4, 000 Deductible, 10% to 15% co-insurance after deductible Generous benefit design for family contract: • $2, 000 PCA, $3, 000 Deductible, 0% co-insurance after deductible

Policy Opportunities What is the Upside to CDHPs? • Innovative means to bring consumer choice into the medical marketplace as well as consumer awareness of the trade-offs of liberal medical insurance coverage policies. • Creates infrastructure for personal, portable health care coverage. • Hybrid variants could be crafted to serve low income, part time workers and possibly the uninsured through tax credits and vouchers.

Policy Conundrums What is the Downside to CDHPs? • What if CDHPs accelerate the consumer’s burden of health care spending ‘too’ quickly? • Not much incentives for managed care’s proven assets (e. g. , disease management) to play a role.

Next Steps within the RWJ Grant Period • • 2 nd Year U of M Survey Examination of switching behavior Multi-employer/multi-year results Examine the effect of online tools on medical care demand • More detailed examination of service use differences from claims data

Next Steps in a (soon to be) Proposed RWJ Grant Extending the Analysis • Involve 12 employers – 6 original employers in study – 6 new employers using non-Definity CDHPs • Three year continuation grant • Develop five year cohorts for six original employers to see long terms effects on plan selection and the resulting cost and utilization impact. • Develop optimal choice model of plan selection for a consumer. • Empirically test optimal choice to examine rationality of consumers as the market evolves.

CDHP’s Legislative Cousins: Health Savings Accounts ØIntroduced in the 2003 Medicare Reform Law ØWhat it enables: • • • Any U. S. citizen can create a ‘qualified’ HSA account. Must have ‘catastrophic health insurance’ with minimum deductible of $2, 000. Max is $10, 000 for a family contract. Individuals or employers can make annual pre-tax contribution to an HSA, separate from the insurance policy, of 100% of the deductible (max of $5, 150).

Compare & Contrast: The CDHP Model versus…. Health Tools and Resources $$ Health Coverage • Preventive care covered 100% • Annual deductible Definity • Expenses beyond the PCA Health Care • Nationwide provider access Advantage • No referrals required Health Tools and Resources • Care management program • Extensive easy-to-use information and services 1 2 Annual Deductible PCA Employer selects which expense apply toward the Health Coverage annual deductible. Paid out of employer’s general assets. Preventive Care 100% Health Covera ge Annual Deductible Personal Care Account (PCA) • Employer allocates PCA 1 • Member directs PCA • Section 213(d) “scope” • Roll over at year-end • Apply toward deductible 2 Web- and Phone. Based Tools

…The HSA Model Annual Deductible Health Coverage • Purchased by ‘Qualified’ Plans • Annual deductible • Expenses beyond the HSA • No managed care provisions • Nationwide provider access • No referrals required Health Covera ge Annual Deductible Pre ve n t iv e C ar e e 1 00 % % Health Care Account (HSA) • Consumer/Employer allocates HSA • Consumer directs HSA • Owned by consumer and portable • Roll over at year-end $$ • Many deposited pre-tax • Consumer can withdrawal with penalty • Can apply toward deductible HSA

Why Would Anyone Want such a Wacky Thing? Compare & Contrast Old Way for Family to Buy Coverage: ØBuy family policy from BCBS ØPolicy cost: $8, 460 ØPlan has $500 deductible ØDeductible applied person ØDeductible capped at $1, 000 ØCost if healthy: $8, 460 ØCost if 1 person sick: $8, 960 ØCost if 2+ people sick: $9, 460 The New HSA Way: Ø Buy ‘qualified’ plan from BCBS Ø Qualified plans costs: $3, 936 Ø Plan has $2, 500 deductible Ø Deductible applied person Ø Deductible capped at $5, 000 Ø Cost of healthy: $3, 936 Ø Cost if 1 person sick: $6, 436 Ø Cost if 2+ people sick: $8, 936

HSA Policy Questions we can answer • What type of person chooses the plans? • If they save money, how? • What would be the cost of filling the uninsured gap with this type of plan?

What unique data do we have? • For all CDHP-offered employees, a combination of: – – – Plan choice Benefit design Income and other demographics FSA election Claims history (medical, hospital, pharmacy) • For CDHP enrollees: – Use of Internet tools – Use of preventive service benefit – Behavior at ‘key’ points of inflection

What We Want to Do if Someone Pays the Tab • Combine our data with MEPS to cost out HSA plan for different segments of the population through micro-simulation. – Focus on low-income because we could price a voucher/tax credit. – We know how much low-income people would spend because they are present in our data. – We know how preventive care in HSA would work with actual data for this population too.

Could We Estimate a Choice Model to Look at Possible Take-Up? • What we know: – Choices of individuals in firms with multiple health plans – Their consumption given their choices • What we don’t know: – Probability of uninsured choosing an HSA – If employers will offer HSAs as a choice with other plans or as a replacement – Individual insurance market take-up rate

Econometric Issues • How to group health plan choices? • Ideally, estimate separate choice models for: 1. Single employees with no dependents 2. Families who have no other source of health insurance 3. Families who have multiple choices of health insurance • • • Practically, we can’t identify (2) and (3), so we combine single and family contracts into one choice model through the use of planinteracted dummy variables (Feldman & Schultz, 2001). Considered a nested logit, but the Definity next, if weighted, was not large enough. Correction for oversampling Definity and undersampling the other plans. Lerman-Manski correction was used obtain appropriate standard errors.

Impact of price, employee characteristics and health plan feature preferences on health plan choice

Continued: Impact of employee characteristics and health plan feature preferences on health plan choice