1. Felipa de Mello Sampayo ISCTE-IUL BRU-IUL
Spatial Heterogeneity in the Effects of Quality on Elderly’s Medicare Spending
Felipa de Mello Sampayo
ISCTE-IUL
BRU-IUL 1

2. Overview of the Presentation
Goals and Motivation
Elderly’s Medicare Spending
Econometric Methodology
Results
Conclusions

3. Goals and Motivation Motivation:
Populations are rapidly aging and the elderly account for the majority of medical spending.
Governments are under increasing pressure to contain medical spending.
An extensive literature exists on health spending variation, but few studies have examined the elderly population.
This paper helps to fill this gap by examining the Medicare market

4. Goals and Motivation Goals:
Examine the Medicare market, in which the healthcare spending drastically varies across the country.
Using regional data, we first examine the impact of the heterogeneous effects of the quality of care on elderly´s Medicare spending at the aggregate level.
We then delve into details and examine whether the quality effects are heterogenous by service type.
Such information will help policy makers implement better healthcare policies for the elderly.

5. Elderly’s Medicare Spending
Our data are from the Centers for Medicare & Medicaid Services (CMS) which has developed a public use file, the Geographic Variation Public Use File (GV PUF). Our data cover the year 2014.
We found that Medicare spent an average of $9050 per beneficiary over 65 in Considerable variation in spending occurred among the 304 US continent hospital referral regions (HRR).
The highest-spending regions were the Bronx, New York ($14,506) and Miami, Florida ($14,436), compared to the lowest-spending regions of Grand Junction, Colorado ($6,552) and Missoula, Montana ($6,791).

6. Elderly’s Medicare Spending
Medicare spending can vary for two reasons:
First, Medicare often pays different amounts for the same service in different areas. The payment rates were standardized to account for local wages or input prices.
Second, the health of Medicare beneficiaries also varies geographically. The data were adjusted to account for beneficiaries’ health status.
By standardizing payment amounts and adjusting for differences in beneficiaries’ health status, these data provide a more accurate picture of how resource use varies for Medicare beneficiaries across the country.

7. Figure 1

8. In Figure 1 we show three maps showing the distribution of per beneficiary Medicare spending and use across the United States in 2014.
The first map shows regional variation of the original Medicare spending (MS)
The second map shows standardized (to account for local wages or input prices) Medicare spending (SMS).
The third map shows standardized and risk adjusted Medicare spending, i.e. Medicare Service Use (MU)
Service use has less regional variation than MS and SMS, but substantial variation remains.
There are clusters of high MS regions largely concentrated in Florida, the deep South, and urban areas on the East and West coasts, whereas in terms of MU the regions of high usage are the South, the East coast and central West.

9. Econometric Methodology

10. Results
The results are presented for two separate cases. In the first case, presented in Table 1, we control for health risk status (HS) of Medicare beneficiaries proxied by the HCC score.
Table 2 includes in our analysis hospital readmissions rate (HR).
The variable HR enters the model so as to characterize the quality of continuity of care. The higher the HR, the lower the quality of continuity of care and the higher SMS is expected to become.

11. HS Test for Bandwidth Dependent Variables: I. Aggregated ResultsHS
Test for Bandwidth
Dependent Variables:
I. Aggregated Results
Total Standardized
1.259***
(0.205)
4.983***
II. By Treatment Spending
II. A. Inpatient Setting
Inpatient hospital care (IPHC)
1.281***
(0.333)
3.961***
Skilled nursing facilities (SNFs)
1.252***
(1.239)
3.088***
II. B. Outpatient Setting
Hospital Outpatient Services (HOP)
-0.185*
(0.738)
Clinics
-4.925***
(12.384)
1.917***
Outpatient dialysis facilities (OPD)
3.283*
(2.103)
Home Health services (HH)
3.622***
(2.061)
Ambulatory Surgical Centers (ASCs)
-0.204*
(1.722)
Evaluation and Management Services (E&M)
1.491*
(1.116)
Physician Procedures (PP)
0.532**
(1.121)
Imaging
1.528*
(0.929)
Durable Medical Equipment (DME)
-0.113*
(0.526)
Tests
1.770*
(2.750)
Ambulance
1.791***
(1.213)
III. Prescription Drugs
Drugs
0.498***
(1.581)

12. HS HR Test for Bandwidth I. Aggregated Results Total StandardizedHS HR
Test for Bandwidth
I. Aggregated Results
Total Standardized
0.867*
0.179**
4.983***
(0.243)
(0.100)
II. By Treatment Spending
II. A. Inpatient Setting
Inpatient hospital care (IPHC)
0.465**
0.375**
(0.175)
(0.043)
3.961***
Skilled nursing facilities (SNFs)
0.360***
0.434*
(1.416)
(0.351)
3.088***
II. B. Outpatient Setting
Hospital Outpatient Services (HOP)
-0.322*
0.095***
(0.990)
(0.403)
Clinics
-5.699*
0.852*
(13.964)
(3.775)
1.917***
Outpatient dialysis facilities (OPD)
3.349*
-0.112*
(2.210)
(0.567)
Home Health services (HH)
3.309***
0.138***
(1.971)
(0.687)
Ambulatory Surgical Centers (ASCs)
-0.382*
0.941*
(2.934)
(1.100)
Evaluation and Management Services (E&M)
1.299*
0.126***
(0.653)
(0.347)
Physician Procedures (PP)
0.663*
-0.057*
(0.744)
(0.279)
Imaging
1.345**
0.055*
(1.043)
(0.396)
Durable Medical Equipment (DME)
-0.525*
0.225*
(1.115)
(0.330)
Tests
1.474**
-0.003***
(1.887)
(0.694)
Ambulance
0.587*
0.593*
(0.888)
(0.325)
III. Prescription Drugs
Drugs
0.814**
-0.157*
(1.298)
(0.433)

13. Figure 2: Estimated Local Parameters Relative to Table 2
The spatial distribution of GWR estimates for the SMS at aggregate level shown in Table 2 is presented in the maps of Figure 2.
We see that the relationship between SMS and HS was always positive and was stronger mainly in the Mountains’ region, the West Central, and South Atlantic regions.
In relation to the mapping of the coefficients of HR, we observe that the stronger positive effects were in counties in the Southwest region, with the exception of some HRR, like in Texas, in the West and East North Central, and New England regions. In these regions there is evidence that lower quality increases the use of Medicare services. However, in the West and the South Atlantic regions, the effect of HR becomes negative, which may give support to the hypothesis that the higher SMS is due to higher quality of continuity of care.

14. Figure 2

15. Conclusion
We find that at the aggregate level poor quality of the healthcare increases Medicare spending and thus costs per beneficiary in 2014.
Second, quality effects are heterogeneous, and their impact varies both spatially and by the type of medical service. In particular, Clinics and Ambulatory Surgical Centers’ services stand out for their high positive effect of hospital readmissions, a proxy for poor quality of continuity of care.
Third, we find that the effect of health risk status on medical spending differs by treatment type. Specifically, healthier elderly use more Hospital Outpatient Services, Clinics, Ambulatory Surgical Centers, and Durable Medical Equipment, i.e. supply sensitive types of services.
Reducing geographic variation in healthcare spending would not necessarily improve the overall quality of medical practice. Reducing payments to high-spending areas and increasing payments to low spending areas reduces spending variation, but the results on the quality of care will be ambiguous.