Steven Lieberman Assistant Director Health and Human Resources Division Congressional Budget Office Lowering Medicare Costs: Regions or Beneficiaries? National Health Policy Conference January 28, 2004

Contents 1.Empirical observations 2.Potential payoffs from two “naïve” interventions –High-cost beneficiaries –High-cost regions 3.Next steps –Policy questions to lower spending –Criteria for evaluating policy interventions

Empirical Observations –More is not better  Patients in high-spending regions received 60 percent more care  No better quality of or access to care  If spending in all regions were lowered to levels in the lowest decile, Medicare spending would fall by 29 percent –Medicare spending is concentrated  Most expensive 5 percent of beneficiaries accounted for 47 percent of total Medicare spending during –Degree of concentration is similar across HRRs –Many high spenders have persistently high spending  18 percent of beneficiaries were in the top quartile of spenders in each year for at least two consecutive years during

Distribution of Medicare Spending and Beneficiaries Notes: Data from a 5 percent random sample of fee-for-service (FFS) beneficiaries between 1995 and 1999.

Concentration of Medicare Spending Notes: Data from a 5 percent random sample of fee-for-service (FFS) beneficiaries between 1995 and Spending reported in 1999 dollars. Medicare Spending Beneficiaries% of TotalMean Top Quintile84%$24,161 Fourth Quintile 12%$3,353 Third Quintile3%$943 Bottom Two Quintiles 1%$155 Total100%$5,753 Medicare Spending Beneficiaries% of TotalMean Top 1 %17%$98,074 Top 5 %47%$53,538 Top 10 %66%$37,855

Distribution of Medicare Spending among HRRs Notes: Data from a 5 percent random sample of fee-for-service (FFS) beneficiaries between 1995 and Excluded are beneficiaries who could not be assigned an HRR.

Persistence of Medicare Spending High Cost 27% of Benes 75% of Costs Low Cost 73% of Benes 25% of Costs Persistent 18% of Benes 57% of Costs Not Persistent 9% of Benes 18% of Costs 60% Alive End of Year 5 40% Dead End of Year 5 Year 1 Cohort of Beneficiaries Notes: CBO analysis in “Concentration and Persistence of Expenditures among Medicare Beneficiaries.” Persistently high-cost group is defined as those who were in top 25 % of spending for at least 2 consecutive years during the 5-year period,

Profiles of Beneficiaries by Spending Group Spending Group (Mean Spending) Top 5 Percent ($53,538) Top 20 Percent ($24,161) Bottom 40 Percent ($155) Inpatient PPS Spending / Beneficiary $25,309$11,119None Physician Spending / Beneficiary $8,872$4,934$129 Chronic Conditions Congestive heart failure Diabetes Cognitive impairment 47% 35% 20% 36% 29% 16% 6% 10% 3% Notes: Data from a 5 percent random sample of fee-for-service (FFS) beneficiaries between 1995 and Spending reported in 1999 dollars.

Contents 1.Empirical observations 2.Potential payoffs from two “naïve” interventions –High-cost beneficiaries –High-cost regions 3.Next steps –Policy questions to lower spending –Criteria for evaluating policy interventions

Potential Payoffs from Targeting High-Cost Beneficiaries vs. High-Cost Regions By BeneficiaryBy HRR % of Beneficiaries % of Spending% of Beneficiaries% of Spending Top 1%16.7 %Top 0.7 %1.2 % Top 5%45.8 %Top 4.9 %7.0 % Top 10%65.1 %Top 9.8 %13.1 % Top Quintile83.5 %Top 19.4 %24.4 % Fourth Quintile12.0 %Next 20.3 %21.9 % Third Quintile3.4 %Next 20.1 %19.6 % Bottom Two Quintiles 1.2 %Bottom 40.2 %34.0 % Notes: Data from a 5 percent random sample of fee-for-service (FFS) beneficiaries between 1995 and Excluded are beneficiaries who could not be assigned an HRR.

Potential Payoffs from Targeting High-Cost Beneficiaries vs. High-Cost Regions

Possible Policy Interventions Intervention TargetPossible Examples High-Cost Regions Establish and fund Comprehensive Centers for Medical Excellence (CCMEs) Revise FFS payments to reward efficient care and reduce payments for HRRs with excessive supply-sensitive care High-Cost Beneficiaries Introduce case management/ disease management programs Use risk screening/ predictive modeling High-Cost Beneficiaries in High-Cost Regions Reduce the number of teaching hospital beds in high-cost HRRs Case manage high-cost beneficiaries in high- cost HRRs Possible interventions through reducing capacity, changing reimbursements and creating new programs

Challenges of the High-Cost Beneficiaries Approach Generating savings from high-cost beneficiaries requires: –Identifying who is, or will be, high cost  Who survives but remains high cost?  Could interventions change cost or outcome? –Developing and implementing effective interventions  Requires achieving lower costs or improved outcomes  Should interventions be provider or patient focused? –Rigorously demonstrating net savings  Possibly use randomized control groups  Capture costs of intervention and changes in benefit spending

Challenges of High-Cost Regions Approach Generating savings from high-cost HRRs requires: –Identifying who/ what will change practice patterns/ utilization  What to target?  Could interventions cause changes in cost or outcome? –Developing and implementing effective interventions  Requires achieving lower costs or improved outcomes  Where or on what should interventions be focused? –Rigorously demonstrating net savings  Possibly use randomized control groups  Capture costs of intervention and changes in benefit spending

1.Empirical observations 2.Potential payoffs of two “naïve” interventions –High-cost beneficiaries –High-cost regions 3.Next steps –Policy questions to lower spending –Criteria for evaluating policy interventions Contents

Policy Questions to Lower Spending 1.How best to identify excess spending? 2.What interventions would reduce excess spending? 3.How do we evaluate results of different policy interventions?

Criteria for Evaluating Policy Interventions CriteriaRegionsBeneficiaries Target efficiencyAlso include beneficiaries with low spending in high-cost regions Target a small group of beneficiaries with very high spending Number of people disrupted ManyRelatively few Return on intervention Possibly generate broad delivery system changes in utilization & savings within high-cost HRRs Focus on persistently high spenders could lower their spending, but effect on system cost unknown