Medicaid Expenditures Todd Gilmer, PhD Rick Kronick, PhD - - PowerPoint PPT Presentation

Inter- and Intrastate Variation in Medicaid Expenditures Todd Gilmer, PhD Rick Kronick, PhD University of California, San Diego SUPPORTED BY A GRANT FROM HCFO/RWJ

Research Questions  Does interstate variation in Medicaid spending result primarily from variation in the volume of services or in the price per unit of service?  How do inter- and intrastate variation in Medicaid utilization and spending compare to variation in Medicare spending and utilization?  Is more better for Medicaid beneficiaries? 2

Data  Medicaid Analytic eXtract (MAX) data for CY 2001- 2005 for all 50 states and DC  MAX data starts with data from the Medicaid Statistical Information System (MSIS), and is then massaged by CMS to create person-level analytic files  Complete claims and eligibility data on approximately 280 million beneficiaries (not necessarily unique) over five years 3

Methods  Exclude partial benefits beneficiaries (SLMBs, QMBs, family planning-only, etc)  Focus on Cash-Assistance, Medicaid-Only, fee-for-service, beneficiaries with Disabilities (CAMODs) - Restrict to cash disabled because uniform national eligibility standard for SSI increases comparability of the analysis sample across states - Restrict to Medicaid-only (eliminate dual eligibles) to get a complete view of utilization and expenditures - Restrict to FFS because encounter data are incomplete for beneficiaries in managed care  In analyses of spending on CAMODs, exclude five states (AL, AZ, DE, MD, and ND) because managed care penetration is too high or other data anomalies 4

Distribution of Medicaid Beneficiaries and Expenditures, 2001-2005 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Beneficiaries Total Expenditures Acute LTC 47.2 million $234.6 billion $149.2 billion $74.8 billion Cash Assistance, Medicaid-only, Disabled (CAMOD Other disabled Aged Adults Children 5

Correlation Coefficients, State-level Expenditures per Beneficiary and Expenditures per CAMOD, 2001-2005 Standardized Acute LTC Expenditures per Expenditures expenditures expenditures Beneficiary per CAMOD per CAMOD per CAMOD Standardized expenditures per beneficiary 1.00 — — — Expenditures per CAMOD 0.86 1.00 — — Acute expenditures per CAMOD 0.81 0.96 1.00 — LTC expenditures per CAMOD 0.81 0.93 0.81 1.00 Source: 2001-2005 MAX data. N=46 (excludes AL, AZ, DE, MD, and ND). 6

Measuring Volume and Price  Inpatient (PH & MH) - Volume = inpatient days - Price = expenditure per day  Outpatient (PH & MH) - Volume = unique days - Price = expenditure per day  Pharmacy - Volume = number of fills - Volume = average expenditure per fill at mean prices - Price = Ratio of actual expenditure to predicted expenditure at mean prices 7

Calculating Incremental Effects  Decomposition of price and volume effects  Incremental volume is the difference between actual and mean volume multiplied by mean price - Vinc = (Vi – Vm) * Pm  Incremental Price is the difference between actual and mean price multiplied by actual volume - Pinc = (Pi – Pm) * Vi  Combined incremental effect is the product - Inc Effect = Vinc + Pinc 8

Results 9

10

Distribution of State-level per Beneficiary Acute and LTC Spending on CAMODs, 2001 – 2005 11 Source: 2001-2005 MAX data. Note: Data exclude AL, AZ, DE, MD, and ND.

12

13

14

15

16

17

18

19

Robust Effect of Primary Care  Multiple indicators of access to primary care are related to reduced admissions: - Supply of primary care physicians - Average number of outpatient visits - Average Price Per Visit 20

The relationship between Medicare and Medicaid utilization and spending 21

Distribution of state-level 2004 Medicare spending per beneficiary, and 2001-2005 acute care Medicaid spending per CAMOD 22 Source: Medicare, Dartmouth Atlas; Medicaid, MAX data, 2001-2005 AL, AZ, DE, MD, and ND are excluded.

23

2004 Medicare spending per beneficiary and 2001-2005 acute care Medicaid spending per CAMOD 24 Source: Medicare, Dartmouth Atlas; Medicaid, MAX data, 2001-2005 AL, AZ, DE, MD, and ND are excluded.

2004 Medicare admissions/1,000 and 2001-2005 Medicaid admissions per CAMOD Source: Medicare, Dartmouth Atlas; Medicaid, MAX data, 2001-2005AL, AZ, DE, MD, and ND are excluded. Admissions to psychiatric hospitals and admissions to acute care hospitals with a primary mental health 25 diagnosis are excluded from the Medicaid data.

2004 Medicare Part B spending, and 2001-2005 Medicaid 'Part B' spending Source: Medicare, Dartmouth Atlas; Medicaid, MAX data, 2001-2005AL, AZ, DE, MD, and ND are excluded. Medicaid 'Part B' spending includes MD/OPD/Clinic spending, and expenditures for laboratory and 26 radiology services.

2004 Medicare spending per beneficiary and 2001-2005 acute care Medicaid spending per CAMOD, by HRR, selected states 27 2004 Medicare Part B spending, and 2001-2005 Medicaid 'Part B' spending Source: Medicare, Dartmouth Atlas; Medicaid, MAX data, 2001-2005AL, AZ, DE, MD, and ND are excluded.

2004 Medicare spending per beneficiary and 2001-2005 acute care Medicaid spending per CAMOD, by HRR, California 28

Exhibit 3. 2001 – 2005 Medicare and Medicaid Acute Care Spending per Beneficiary by Hospital Referral Region, Deviations from State Means 6000 5000 4000 3000 Medicare 2000 1000 0 -1000 -2000 -4000 -3000 -2000 -1000 0 1000 2000 3000 4000 Medicaid Source ce: Medicare data from the Dartmouth Atlas of Health care. Medicaid from MAX data. Note: Medicaid data limited to non-dually eligible FFS disabled beneficiaries receiving cash assistance. Excludes AL, AZ, DE, MD, ND. Each data point calculated as the difference from the state mean. 29

Conclusions 30

There is wide variation across states in spending per Medicaid beneficiary  For example, NY spends more than twice as much per beneficiary than CA on acute care  Spending is generally lower in the South, and higher in the Mid-Atlantic, New England, and the upper Midwest  Inpatient utilization only partially follows the contours of acute care spending - Low in New England; high in FL, LA, TX, and OK  There is much more interstate variation in mental health and ‘ other acute ’ spending than in inpatient, MD/OPD/Clinics, or Rx, and much more variation in LTC than in acute spending 31

Volume of services drives relative positioning, unit price is secondary  High-spending and low-spending states are different from the national average primarily because of volume (2/3), and only secondarily because of price (1/3) - Inpatient, mental health, and other spending contribute approximately equally, while variation in MD/OPD/Clinic has very little effect - Inpatient spending varies approximately equally because of volume and price, while MH and drugs variation is driven almost entirely by volume 32

At the state level, Medicaid and Medicare spending are unrelated  There is a weak relationship between Medicare and Medicaid inpatient admissions/1,000  There is no relationship between Medicare Part B and Medicaid outpatient spending  Inpatient hospital spending is a much larger component of Medicare spending than of Medicaid spending 33

Within most states, Medicaid and Medicare spending are strongly related at the HRR level  Inpatient admissions strongly related  Outpatient spending very weakly related  California a notable exception, with no relationship between Medicare and Medicaid within state 34

Making sense of the Medicare-Medicaid relationship  Virtually zero state-level correlation in spending, and very small correlation in inpatient admissions suggest that Medicaid policy variables mediate the supply-utilization relationship suggested by the Dartmouth Atlas  Relatively strong within-state relationship at the HRR level for inpatient admissions suggests that, holding Medicaid policy constant, supply of resources affects Medicare and Medicaid utilization similarly  Very weak within-state relationship on outpatient spending requires more investigation 35

Is more better in Medicaid?  At the state level, some suggestion that more physician visits are associated with lower readmission and lower ACS rates  Strong association between larger fraction of primary care physicians and lower hospitalization, and ACS rates  At the state level, little indication that a higher volume of mental health services or more prescription drug fills are associated with lower hospitalization rates 36