Ambulance Providers: Costs and Expected Medicare Margins Vary
Greatly (23-MAY-07, GAO-07-383).
In 2002, Medicare implemented a national fee schedule designed to
standardize payments for ambulance services. The Medicare
Prescription Drug, Improvement and Modernization Act of 2003
(MMA) required GAO to study ambulance service costs. GAO examined
providers' costs of ground ambulance transports in 2004 and
factors that contributed to cost differences; average Medicare
ambulance payments expected under the national fee schedule in
2010 and how those payments will relate to providers' costs per
transport; and changes that occurred in Medicare beneficiaries'
use of ambulance transports from 2001 to 2004. GAO estimated
costs of ambulance transports based on a nationally
representative survey of 215 ambulance providers that did not
share costs with nonambulance services. Providers that shared
costs with other institutions or services and could not report
their costs for ambulance services separately, such as fire
departments, were excluded because their reported costs appeared
unreliable. GAO used its survey, Medicare claims, and other data
for its analyses.
-------------------------Indexing Terms-------------------------
REPORTNUM: GAO-07-383
ACCNO: A69849
TITLE: Ambulance Providers: Costs and Expected Medicare Margins
Vary Greatly
DATE: 05/23/2007
SUBJECT: Beneficiaries
Cost analysis
Cost sharing (finance)
Fees
Financial analysis
Health care services
Medicare
Surveys
Transportation rates
Cost estimates
******************************************************************
** This file contains an ASCII representation of the text of a **
** GAO Product. **
** **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced. Tables are included, but **
** may not resemble those in the printed version. **
** **
** Please see the PDF (Portable Document Format) file, when **
** available, for a complete electronic file of the printed **
** document's contents. **
** **
******************************************************************
GAO-07-383
* [1]Results in Brief
* [2]Background
* [3]Medicare Ambulance National Fee Schedule
* [4]MMA Temporary Payment Provisions
* [5]Medicare Margins and Costs
* [6]Costs per Transport Were Highly Variable, Reflecting Differe
* [7]Providers' Reported Costs per Transport Were Highly Variable
* [8]Certain Provider Characteristics Contributed to Differences
* [9]Average Payments under the National Fee Schedule Will Be Gre
* [10]Average Payments under the National Fee Schedule Will Be Gre
* [11]Expected Medicare Margins Will Vary Greatly
* [12]MMA Provisions Resulted in Greater Average Payments for High
* [13]Payment Increases Were Targeted to Higher-Cost Super-Rural T
* [14]Regional Payment Adjustments Required by the MMA Were Not Ju
* [15]Medicare Beneficiaries' Use of Ambulance Transports Increase
* [16]Conclusions
* [17]Recommendation for Executive Action
* [18]Agency and External Comments and Our Evaluation
* [19]Appendix I: Data and Methods
* [20]National Survey of Ground Ambulance Providers' Costs
* [21]Survey Instrument Development
* [22]Sample Design
* [23]Survey Administration
* [24]Survey Data Validity and Reliability
* [25]Interpretation of Confidence Intervals and Analysis of
Nonre
* [26]Modeling Ambulance Costs per Transport
* [27]Average Cost Regression Analysis--Methods and Results
* [28]Estimating Average Medicare Ambulance Payments and Use of Tr
* [29]Medicare Payment Calculations and Simulation
* [30]Analysis of Medicare Margins
* [31]Estimating Use of Ambulance Transports by Medicare
Beneficia
* [32]Medicare Claims Data Reliability
* [33]Appendix II: Comments from the Centers for Medicare & Medica
* [34]Appendix III: GAO Contact and Staff Acknowledgments
* [35]GAO Contact
* [36]Acknowledgments
* [37]Related GAO Products
* [38]Order by Mail or Phone
Report to Congressional Committees
United States Government Accountability Office
GAO
May 2007
AMBULANCE PROVIDERS
Costs and Expected Medicare Margins Vary Greatly
GAO-07-383
Contents
Letter 1
Results in Brief 6
Background 7
Costs per Transport Were Highly Variable, Reflecting Differences in
Certain Provider Characteristics 17
Average Payments under the National Fee Schedule Will Be Greater Than
Average Historical Payments, but Providers' Expected Medicare Margins Will
Vary Greatly 22
MMA Provisions Resulted in Greater Average Payments for Higher-Cost
Super-Rural Transports and Adjusted Payments Regionally Where No
Significant Cost Differences Were Observed 27
Medicare Beneficiaries' Use of Ambulance Transports Increased from 2001 to
2004, Except in Super-Rural Areas 29
Conclusions 30
Recommendation for Executive Action 31
Agency and External Comments and Our Evaluation 31
Appendix I Data and Methods 34
Appendix II Comments from the Centers for Medicare & Medicaid Services 52
Appendix III GAO Contact and Staff Acknowledgments 54
Related GAO Products 55
Tables
Table 1: Summary of MMA Temporary Payment Provisions, Implemented July 1,
2004 16
Table 2: Estimated Average Cost per Transport for Provider Characteristics
That Affect Costs 20
Table 3: Payments Prior to and under the National Fee Schedule after MMA
Provisions Expire 23
Table 4: Expected Average Medicare Margins under the National Fee Schedule
for Providers without Shared Costs in 2004 Dollars 24
Table 5: Average Payments prior to MMA Implementation and after
Implementation 28
Table 6: Percentage Changes in Average Payments prior to MMA
Implementation and after Implementation, by Region 29
Table 7: Ambulance Transports per 1,000 Beneficiaries in Urban, Rural, and
Super-Rural Areas 30
Table 8: Provider and Local Area Characteristics Included in Analysis of
Average Cost per Transport, 2004 41
Table 9: Results for Average Cost of an Ambulance Transport
Regression--Estimated Effects of Selected Provider and Local Area
Characteristics on the Average Cost of Ambulance Transports for Providers,
Not Including Impact of Productivity and Community Tax Support 43
Table 10: Results for Average Cost of an Ambulance Transport
Regression--Estimated Effects of Selected Provider and Local Area
Characteristics on the Average Cost of Ambulance Transports for Providers,
Including Impact of Productivity and Community Tax Support 44
Table 11: Census Divisions 47
Table 12: Average Mile per Transport, First Half of 2004 49
Table 13: Mileage Rates 50
Figures
Figure 1: Medicare Ambulance Payment Formula under the National Fee
Schedule 13
Figure 2: National Fee Schedule and Regional Fee Schedule 15
Figure 3: Distribution of Cost per Transport for Providers without Shared
Costs in 2004 18
Figure 4: Expected Medicare Margins for Urban, Rural, and Super-Rural
Providers without Shared Costs 26
This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.
Abbreviations
ALS advanced life support
BLS basic life support
CF conversion factor
CMS Centers for Medicare & Medicaid Services
CPI-U Consumer Price Index for All Urban Consumers
EMS emergency medical services
EMT emergency medicaltechnician
GPCI geographic practice cost index
MMA Medicare Prescription Drug, Improvement and Modernization Act of
2003
MSA metropolitan statistical area
NECMA New England county metropolitan area
NFS national fee schedule
RVU relative value unit
United States Government Accountability Office
Washington, DC 20548
May 23, 2007
The Honorable Max Baucus
Chairman
The Honorable Charles E. Grassley
Ranking Member
Committee on Finance
United States Senate
The Honorable John D. Dingell
Chairman
The Honorable Joe Barton
Ranking Member
Committee on Energy and Commerce
House of Representatives
The Honorable Charles B. Rangel
Chairman
The Honorable Jim McCrery
Ranking Member
Committee on Ways and Means
House of Representatives
In 2005, ambulance service providers completed more than 12.6 million
ground transports for Medicare beneficiaries.^1 Medicare paid more than $4
billion for ground ambulance transports and is likely the largest single
payer of ambulance services in the United States.
As part of a series of Medicare payment reforms in 1997, Congress required
the Health Care Financing Administration, now the Centers for Medicare &
Medicaid Services (CMS), to develop a national fee schedule for Medicare
ambulance services, which was implemented in 2002.^2 Historically, CMS had
used two methods to pay for ambulance services, which resulted in wide
variations in payment for the same service among different types of
ambulance service providers. In particular, CMS had used one
method--reasonable costs^3--to pay hospital-based providers. It used
another method--reasonable charges^4--to pay other, nonhospital-based
types of ambulance service providers. This meant that hospital-based and
nonhospital-based providers were paid different amounts for the same
ambulance services.
^1We use the term providers to refer to all types of organizations that
provide ambulance transports for Medicare beneficiaries.
^242 U.S.C. S 1395m(l).
In 2002, CMS began phasing in a national fee schedule that established a
single payment method for all ambulance services regardless of the type of
provider.^5 This fee schedule standardized Medicare payments for ambulance
services. In general, providers strive to keep their costs of delivering a
service at or below the standard fee schedule rate for that service. Under
the Medicare ambulance national fee schedule, providers that have costs of
delivering ambulance services above the fee schedule payment lose the
difference between the payment amount and their costs, while providers
with costs below the fee schedule payment are able to keep the difference
between the payment amount and their costs. In aggregate, these
differences are known as Medicare margins and express whether the provider
makes a profit or loss on its Medicare transports. Some providers rely
heavily on Medicare revenues and adequate Medicare margins help ensure the
continuing availability of beneficiaries' access to ambulance services.
CMS phased in the ambulance national fee schedule from April 2002 through
December 2005. During this transition, the new fee schedule payments were
blended with the previous reasonable-cost payments for hospital-based
providers and reasonable-charge payments for nonhospital-based providers.
In 2003, Congress passed the Medicare Prescription Drug, Improvement and
Modernization Act (MMA), which introduced several temporary payment
provisions, including a regional fee schedule that overlapped with the
transition to the national fee schedule.^6 Beginning in July 2004, these
temporary payment provisions were expected to add about $840 million to
Medicare payments for ambulance services through December 2009, when the
last of these provisions are set to expire.
^3Reasonable-cost payments were based on the provider's cost of providing
ambulance services as reported on cost reports.
^4Reasonable-charge payments were based on the bill from the ambulance
service provider but were subject to an upper limit.
^5The national fee schedule applies to ground and air ambulance services,
but this report and our analysis are limited to ground ambulance services
only, which include water ambulance services and account for nearly all
ambulance services. We use the terms service, transport, and ambulance
service to refer to ground ambulance transport services only.
^6Pub. L. No. 108-173, S 414, 117 Stat. 2066, 2278-80 (2003).
The Medicare, Medicaid, and SCHIP Benefits Improvement and Protection Act
of 2000^7 and the MMA required GAO to study ambulance service costs. As
discussed with the congressional committees of jurisdiction, in this
report we (1) examined the differences in providers' costs of ambulance
transports in 2004 and the factors that contributed to these cost
differences, (2) assessed how the ambulance national fee schedule in 2010
is expected to affect average ambulance payments and how those payments
will relate to providers' costs per transport, (3) determined the effect
of MMA temporary payment provisions on ambulance payments, and (4)
described the change that occurred in Medicare beneficiaries' use of
ambulance transports from 2001 to 2004.
To examine differences in costs of providing ambulance transports, we
conducted a national survey of ambulance providers in 2005. In our survey,
we requested information about providers' costs, revenues, transports, and
organizational characteristics for their most recently completed fiscal
year.^8 We selected a stratified,^9 random sample of 500 eligible
providers that billed Medicare, and we received 321 completed
questionnaires for a response rate of 64 percent. We used this nationally
representative sample of 321 providers to describe the ambulance
industry.^10 However, after excluding two cost outliers, our analysis was
further limited to a subgroup of providers that (1) did not share costs
with other institutions or services or (2) shared costs but reported costs
of ambulance services separately from the costs of their other services.
Our analysis and findings are nationally representative of this subgroup
of ambulance providers, which we refer to as providers without shared
costs.^11 We excluded 104 providers that shared costs of ambulance
services with other institutions or nonambulance services and could not
distinguish their costs for providing ambulance services from other costs,
including but not limited to all fire departments. We excluded these
providers because their reported costs appeared to be unreliable. The
resulting sample size for our analysis was 215 providers without shared
costs. Although our sample is nationally representative of an estimated
5,200 providers without shared costs, the small sample size along with the
variability of responses reduces the precision of our estimates,
increasing the range of the 95 percent confidence intervals we report. A
95 percent confidence interval is the range within which we expect the
true population estimate to fall 95 percent of the time, and it is the
range of the confidence interval that expresses the precision of our
estimates.
^7Pub. L. No. 106-554, appendix F, S 436, 114 Stat. 2763, 2763A-527
(2000).
^8The GAO Survey of Ambulance Services asked providers to report on their
organizations' costs of providing ground ambulance services. As such, the
survey measures organizations' expenses, or spending, for ground ambulance
services.
^9The sample was stratified by five types of ambulance providers:
hospital-based, volunteer, fire-based, government, and freestanding.
^10To make our survey sample representative of all Medicare ambulance
providers, we computed a sample weight for each respondent provider.
To examine factors that contributed to differences in costs, we used our
survey data and Medicare data supplemented by data from two other sources.
The Oil Price Information Service was our source for the average annual
retail price of fuel by zip code, and the United States Postal Service
supplied building rents because it tracks its facility costs in each zip
code. We used regression analysis to analyze the relationships between
various provider and local area characteristics and cost per transport
among providers without shared costs. We also compared Medicare claims
data for all nonrespondents with those of respondent providers without
shared costs and determined that our cost estimates were not biased by
nonresponse. See appendix I for details regarding our survey, other data
sources, data limitations, and the analytic methods we employed.
To assess the effect of the ambulance national fee schedule on payments,
we used Medicare claims data to compute average payments for ambulance
transports in 2001, before the implementation of the ambulance national
fee schedule, and in 2004, 2 years after the phase-in of the fee schedule
had begun. For Medicare payment analyses, payments were expressed in 2004
dollars to exclude the effects of inflation. We also compared average
payments for urban, rural, and super-rural transports.^12 We used Medicare
claims data and payment formulas as specified in federal regulations to
simulate average payments under the national fee schedule in 2010, after
all of the MMA provisions expire, but computed these payments in 2004
dollars, the year that best reflects the cost data collected in our
survey. To compare the simulated Medicare payments under the national fee
schedule for providers without shared costs with the costs per transport
of those providers, we computed providers' Medicare margins--the
percentage difference between average Medicare payments and providers'
costs per transport.^13 All costs per transport and provider margins are
based solely on our sample of providers without shared costs, and for this
reason, these estimates are reported with their confidence intervals.
^11We applied our sample weights to the subgroup so the providers were
nationally representative of all ambulance service providers without
shared costs.
We also assessed the effect of the MMA temporary payment provisions on
payments in 2004 using Medicare claims data by examining the change in
payments from the first half of the year, before the MMA changes went into
effect, with the second half of the year, when MMA payment provisions had
their maximum effect. To assess the change in Medicare beneficiaries' use
of ambulance transports from 2001 to 2004, we used Medicare claims and CMS
enrollment data, which contain information about beneficiaries, to compute
transports per 1,000 beneficiaries for both years.
We tested the internal consistency and reliability of our survey data and
all non-Medicare data sources and determined that all data sources were
adequate for our purposes. We conducted our work from July 2004 through
April 2007 in accordance with generally accepted government auditing
standards.
^12To define urban, rural, and super-rural transports, CMS uses the
metropolitan statistical areas (MSA) definitions established by the Office
of Management and Budget. During the period of our study, CMS defined
urban transports as those that originate within MSAs and New England
county metropolitan areas (NECMA), rural transports as those that
originate in rural counties that are outside of MSAs and NECMAs as well as
small towns and rural areas within MSAs or NECMAs that are isolated from
central areas by distance or other features, such as mountains. CMS
defines super-rural transports as those that originate in the bottom 25
percent of rural areas as defined by population density.
^13This comparison assumes that providers' cost structures under the fee
schedule would be the same as they were in 2004.
Results in Brief
Costs of ground ambulance transports were highly variable across providers
without shared costs, reflecting differences in certain provider
characteristics. Costs per transport for providers without shared costs
averaged $415, but varied from $99 to $1,218 per transport--a range of
more than $1,100. Contributing to the variability were differences in
providers' volume and mix of transports; service area (urban, rural, or
super-rural); productivity, which we defined as the number of transports
per staffed hour; and the percentage of total revenue derived from local
tax support. Providers without shared costs had higher costs per transport
when they had fewer transports per year, a greater percentage of
transports in which more than a basic medical assessment or intervention
occurred, and more transports from super-rural areas than providers
without shared costs that did not have these characteristics. In addition,
providers without shared costs that had lower productivity and those
receiving a greater percentage of revenues from local tax support had
higher costs per transport than providers without shared costs that had
higher productivity and less local tax support. Other provider and local
area characteristics, such as type of provider, region as determined by
census division, building rent, and price of fuel, did not significantly
affect average costs per transport among providers without shared costs.
Average payments under the national fee schedule in 2010, after all of the
MMA temporary payment provisions are set to expire, are expected to be
higher than payments in 2001, but Medicare margins for providers without
shared costs will vary greatly. In 2010, average ambulance national fee
schedule payments are estimated to be 3 percent higher overall than
payments in 2001, after adjusting for inflation and assuming that
providers bill the maximum amounts allowed. Further, a greater percentage
than the overall increase in payments will accrue to rural and super-rural
transports, on average, while urban transports will receive a decrease in
payments under the national fee schedule in 2010. We could not assess
whether providers without shared costs will break even, lose, or profit,
on average, under the ambulance national fee schedule after the MMA
temporary payment provisions expire, because the 95 percent confidence
interval for the average expected Medicare margin for providers without
shared costs spanned from negative 14 percent to positive 2 percent when
we took into account the number of respondents in our sample and the range
of their reported costs. However, across all providers without shared
costs, we estimated that 39 to 56 percent will have average Medicare
payments above their average costs per transport under the ambulance
national fee schedule in 2010.
The MMA's temporary payment provisions, which included base- and
mileage-rate increases as well as the introduction of a regional fee
schedule, resulted in raised average ambulance payments overall,
particularly for super-rural transports, which we determined were
typically more costly to provide. However, regional payment adjustments by
census division under the MMA did not appear to be warranted on the basis
of regional cost differences. The regional fee schedule increased payments
substantially for some regions but not others. After controlling for
various characteristics that affected the costs of providing ambulance
transports, we did not discern any significant differences in average cost
per transport across regions.
From 2001 to 2004, Medicare beneficiaries' use of ambulance transports
increased overall, even in those regions that had a decrease in average
payments after the MMA was implemented. However, beneficiaries' use in
super-rural areas decreased by 8 percent over the same period. The decline
in the use of super-rural ambulance transports did not appear to be
related to any significant change in the population of Medicare
beneficiaries residing in super-rural areas.
In light of the variability in ambulance providers' Medicare margins and
the potential for negative margins to have an impact on beneficiary
access, we recommend that the Administrator of CMS monitor utilization of
ambulance transports to ensure that Medicare payments are adequate to
provide for beneficiary access to ambulance services, particularly in
super-rural areas. In its comments on a draft of this report, CMS stated
that it agreed with our recommendation. External commenters generally
agreed with our findings.
Background
Ground ambulance services are provided by a wide range of organizations
that differ in their organizational structure, types of services offered,
staffing, and revenue sources. Local conditions--including whether
providers are affiliated with other organizations; whether their service
areas are predominately urban, rural, or super-rural; and the amount of
community-dedicated revenues--contribute to this diversity.
In addition, communities have few, if any, tools to help them decide the
optimal organizational structure, staffing, or amount of funding for
ambulance services, given local conditions. Although there have been
efforts to establish national performance or quality standards,^14 there
is limited information about how to best evaluate the costs of providing
ambulance services in a community. In recent years, industry associations
and federal agencies have worked to develop a data system and mechanisms
for measuring the performance of emergency medical services (EMS);
however, these tools cannot yet be applied to measure performance and
evaluate the efficiency of ambulance services.^15 The lack of data and
performance standards makes it difficult to assess whether any given
provider is delivering quality care or whether services are being provided
efficiently.
Organizational structures differ in that some ambulance providers are
affiliated with another institution, such as a hospital or fire
department, and in that providers may or may not offer other types of
services, such as hospital services, fire suppression, rescue, or
wheelchair transportation. Providers affiliated with another institution
or that offer other types of services may share resources and operational
costs, such as building space, administrative support, or personnel, with
these other entities and services. About two-fifths of the ambulance
industry shared operational costs with other institutions or services in
2004.^16 Nine percent of the ambulance industry was affiliated with a
hospital, while 37 percent was affiliated with a fire department and 21
percent was affiliated with another government agency. Other communities
(33 percent of the ambulance industry in 2004) were served by
freestanding, for-profit or not-for-profit provider organizations.
The types of services offered and the staff employed also vary among
providers. Some providers perform only emergency transports, in response
to a 911 or equivalent call, while other providers offer nonemergency
transports, which are typically transfers from one facility to another and
may be scheduled in advance. In 2004, 45 percent of the ambulance industry
performed only emergency transports; 55 percent performed emergency and
nonemergency transports. Some providers perform only basic life support
(BLS)^17 transports because their staff are not certified to perform more
intensive medical assessments and interventions, such as advanced life
support (ALS)^18 and other more complex services. In 2004, 9 percent of
the ambulance industry specialized in only ALS and more complex Medicare
transports, such as those requiring one or more ALS services or
respiratory care; 14 percent performed only BLS Medicare transports; and
86 percent provided a mix of BLS, ALS, and more complex Medicare
transports.^19 Fire departments are more likely to be specialized in ALS
and more complex services. According to the fire departments that
responded to our survey, 70 percent of their Medicare transports, on
average, required ALS or more complex services; for other providers ALS
and more complex transport services constituted only 49 percent of their
Medicare transports.
^14The Commission on Accreditation of Ambulance Services and the
Commission of Accreditation of Medical Transport Systems have established
ambulance service accreditation standards.
^15The National Association of State EMS Officials has worked with other
industry associations and federal partners to develop tools for measuring
the performance of EMS systems: (1) EMS system indicators, which were
released for public comment and are being finalized; (2) the National EMS
Information Management System; and (3) a cost framework for cost analysis
in EMS research. The American Ambulance Association published its
Community Guide to Ensure High-Performance Emergency Ambulance Services in
2004.
^16Unless otherwise noted, all background information on the ambulance
industry is from the GAO 2005 Survey of Ambulance Services. These
estimates are based on the 321 total providers that responded to our
survey.
Ambulance providers use of a variety of staff to deliver services to their
communities. First, providers and communities determine what proportion of
their staff will be emergency medical technicians (EMT) trained to perform
BLS services, and what proportion will be EMT-intermediates or paramedics,
who have training to perform more intensive ALS interventions.^20 In 2004,
77 percent of providers that completed our survey reported having at least
one staff member trained as a paramedic, nurse, or physician, but the
remaining 23 percent of respondent providers relied on staff with less
training than a paramedic. Other staffing choices include whether to
employ career-oriented paid staff, rely on volunteers, or use a mix of
paid and volunteer staff. Some providers choose to employ cross-trained
staff.^21 In 2004, two-fifths of the ambulance industry relied
substantially on volunteer staff.^22 In addition, providers and
communities make different choices about whether to maintain backup
vehicles and staff or to rely to a greater extent on nearby providers for
backup assistance.
^17BLS services include basic, noninvasive interventions to reduce
morbidity and mortality associated with acute out-of-hospital medical and
traumatic emergencies.
^18ALS services include advanced, invasive, and pharmacological
interventions to reduce morbidity and mortality associated with acute
out-of-hospital medical and traumatic emergencies.
^19Under the Medicare program, there are seven levels of ambulance
transports. BLS and ALS transports each constitute two levels, emergency
or nonemergency. The remaining levels of service--ALS Level 2 and
specialty care transport (both of which involve invasive or specialized
care) and paramedic ALS intercept (when a paramedic provides ALS services
but does not transport the patient)--made up less than 2 percent of all
Medicare transports in 2004.
^20EMT-Basic personnel are trained in BLS services only. EMT-Intermediate
personnel are qualified to perform essential advanced techniques and to
administer a limited number of medications. Paramedics have the
competencies of EMT-Intermediate personnel in addition to other enhanced
skills and can administer additional interventions and medications.
Providers' service areas can be urban, rural, super-rural, or a mix of
areas. During 2004, CMS defined urban transports as those that originate
within metropolitan statistical areas (MSA) and New England county
metropolitan areas (NECMA), rural transports as those that originate in
rural counties that are outside of MSAs and NECMAs,^23 and super-rural
transports as those that originate in the bottom 25 percent of rural areas
as defined by population density. About half of the ambulance providers
served predominately urban areas, and the other half served predominately
rural and super-rural areas.^24 However, three-fourths of all Medicare
transports originated in urban areas. Therefore, rural and super-rural
providers performed fewer transports on average than urban providers. In
addition, rural and super-rural transports were longer than urban
transports, on average, requiring more time and resources per transport.
In 2004, urban Medicare transports averaged 7 miles, while rural Medicare
transports averaged 13 miles and Medicare transports from super-rural
areas averaged 20 miles.
Providers have several potential revenue sources depending on their
communities and their choices about funding ambulance services. Revenue
sources can include community tax support (such as revenue from local
governments); charitable donations; state and federal grants; subscription
programs;^25 and reimbursements from Medicare or Medicaid patients, and
private health insurance companies. Not all providers receive revenues
from all sources, and the mix and amount of revenues available may
vary.^26 For example, 48 percent of the ambulance industry indicated that
a state or local government approves the fees they may choose to charge.
In communities that limit ambulance providers' fees, providers may bill
Medicare (and other payers) less than the allowed amount and therefore do
not receive the maximum Medicare payment allowed for their services. In
2004, Medicare payments for ambulance transports accounted for 31 percent
of the providers' revenues, on average, while Medicare beneficiaries
accounted for about 40 percent of their transports during that same
year.^27 However, the percentage of Medicare revenues among providers
ranged from less than 1 percent to 80 percent. Furthermore, different
communities provide different levels of tax support to address specific
issues, such as ensuring a minimum level of service in remote areas or
being equipped with more sophisticated transport vehicles or having more
highly trained staff.
^21Cross-trained staff typically refers to firefighters who are also
trained as EMTs.
^22We defined substantial use of volunteer staff as 20 percent or more of
staff hours spent providing ambulance services, exclusive of
administration.
^23Rural areas can also be small towns and rural areas within large
metropolitan counties that are isolated from central areas by distance or
other features, such as mountains.
^24From this point forward in the report, we refer to providers that
served predominately urban, rural, and super-rural areas as urban, rural,
and super-rural providers, respectively. We classified providers as
super-rural if 60 percent or more of their Medicare transports in 2004
originated in a super-rural zip code. We classified providers as rural if
they did not meet the super-rural definition and 60 percent or more of
their Medicare transports in 2004 originated in rural or super-rural zip
codes. We defined providers as urban if they did not meet the rural or
super-rural classifications.
^25A subscription program is an arrangement in which an ambulance service
provider is paid an annual fee for providing emergency transportation for
a community.
^26In 2004, the annual spending of providers that responded to our survey
ranged from less than $10,000 to more than $70 million.
^27This estimate is based on a smaller sample of 209 providers that
reported Medicare revenues and total revenues on the GAO 2005 Survey of
Ambulance Services.
Medicare Ambulance National Fee Schedule
The ambulance national fee schedule was part of a series of payment
reforms to make Medicare a more equitable and prudent purchaser of health
care services. Phased in from April 2002 through December 2005, the
national fee schedule standardized payment rates and reduced wide
variations in payments for the same service.
Medicare ambulance payments under the fee schedule have two components: a
base-rate payment and a mileage payment.^28 The base-rate component of
ambulance payments under the fee schedule consists of the relative value
unit (RVU), the conversion factor (CF), and a geographic adjustment
factor. Ambulance RVUs account for the relative resources needed to
provide services during an ambulance transport.^29 The ambulance CF
converts the RVU into a payment expressed in dollars and is set by CMS
annually. Ambulance base-rate payments are also adjusted by a geographic
practice cost index (GPCI), which is intended to account for regional
differences in the cost of providing ambulance services.^30 The mileage
component consists of the number of miles traveled during an ambulance
transport multiplied by the applicable mileage rate. (See fig. 1 for an
example of the payment formula.^31)
^28Medicare ambulance transports must be deemed medically necessary in
order for Medicare payments to be disbursed.
^29There is an RVU for each of the seven levels of service defined by
Medicare.
^30The practice expense portion of the physician fee schedule's GPCI is
used to adjust the ambulance national fee schedule.
^31See 42 C.F.R. S 414.601 et seq.
Figure 1: Medicare Ambulance Payment Formula under the National Fee
Schedule
Note: Payments are expressed in 2004 dollars.
aThe GPCI is applied to only 70 percent of the unadjusted base rate
payment, which is the product of the RVU and the CF. For example, the base
rate payment equals 1.90 x $176.04 x [0.30 + (0.70 x 1.166)]. The RVU, CF,
GPCI, and mileage rate are based on 2004 values.
The ambulance national fee schedule was phased in from April 2002 through
December 2005 by blending new fee schedule payments with historical
payments. During this transition, the national fee schedule portion
constituted a greater share of the total blended ambulance payment each
year until January 2006, when the historical payment portion of the blend
was discontinued. For example, in the latter part of 2002, total ambulance
payments were a blend of 20 percent under the national fee schedule and 80
percent under the historical payment system. In 2004, the blend was 60
percent national fee schedule and 40 percent historical payment system.
(See fig. 2 for further details on the blending of historic payments and
national fee schedule payments.)
MMA Temporary Payment Provisions
The MMA introduced several temporary ambulance payment provisions that
were implemented in 2004, the last of which expires at the end of 2009.
CMS estimates that these payment adjustments will add $840 million to
Medicare ambulance services over the 5 years they are in effect. The MMA
provisions increased payment rates for urban and rural transports and for
transports 51 miles or greater. The MMA also provided a significant base
rate increase for transports originating in super-rural areas and provided
for a new regional fee schedule based on the nine census divisions.32
The regional fee schedule was designed to ease the transition from the
historical payment system to the national fee schedule. The introduction
of the regional fee schedule overlapped with the phase-in of the national
fee schedule. The regional fee schedule gave temporarily higher ambulance
payments than what would generally be paid under the national fee schedule
to ambulance providers in census divisions that had historically higher
payments. If the regional base-rate payment was determined to be greater
than the national base-rate payment for a particular region, then the
region received the more advantageous blend of the regional fee schedule
base-rate payment and the national fee schedule base-rate payment. For
example, in the second half of 2004 under the regional fee schedule,
affected regions received a blend of 80 percent of their regional fee
schedule base-rate payment and 20 percent of the national fee schedule
base payment rate. This base-rate payment was then further blended with
historical payments as a part of the gradual phase-in of the ambulance
national fee schedule.33 The regional fee schedule component of the base
rate blend was reduced each year from 2005 through 2007, and expires at
the end of 2009. (See fig. 2 for further details on the blending of
historical, national fee schedule, and regional fee schedule payments.)
^32In this report we use the term regions to refer to the nine census
divisions as defined by the U.S. Census Bureau. See table 11 in app. I for
a description of the nine census divisions.
^33The national fee schedule portion blend with historical payments
includes mileage payments. The regional fee schedule is only applicable to
the base rate payment.
Figure 2: National Fee Schedule and Regional Fee Schedule
Note: All years refer to calendar years unless otherwise specified.
aJanuary 1, 2001 through March 31, 2002.
bApril 1, 2002 through December 31, 2002.
cThe national fee schedule payment portion includes MMA increases of 1
percent and 2 percent for urban and rural transports, respectively, and an
approximate 23 percent base rate increase for super-rural transports.
dThe national fee schedule payment portion includes an MMA base rate
increase of approximately 23 percent for super-rural transports only.
eJuly 1, 2004 through December 31, 2004.
In addition to providing for a regional fee schedule, the MMA temporarily
required higher payment rates for super-rural transports. As a result,
base rate payments for transports originating in super-rural areas
increased about 23 percent. The MMA also provided for a 25 percent
increase in the mileage rate for every ambulance mile traveled exceeding
50. Finally, the MMA required an increase in payment rates for mileage and
transports originating in urban and rural areas by 1 percent and 2
percent, respectively. (See table 1 for MMA temporary payment provisions
and their expiration dates.)
Table 1: Summary of MMA Temporary Payment Provisions, Implemented July 1,
2004
Payment provision Expiration date
The regional fee schedule December 31, 2009
An increase in the base rate for super-rural transportsa December 31, 2009
A 25 percent increase in the urban and rural mileage December 31, 2008
rate for every ambulance mile exceeding 50
A 1 percent and 2 percent increase for urban and rural December 31, 2006
transports and mileage, respectively
Source: 42 U.S.C. S 1395m(l)(10)-(13).
aCMS determined that this increase would be approximately 23 percent. 69
Fed. Reg. 40288 (July 1, 2004).
Medicare Margins and Costs
A provider's Medicare margin under a fee schedule generally depends on
whether the provider's costs of delivering a service are below its
Medicare payments for the service. Under the ambulance national fee
schedule, providers with costs per transport less than the Medicare
payment for that transport are able to retain the difference between the
fee schedule payment and their costs per transport. Likewise, providers
with costs per transport above the national fee schedule payment will lose
the difference between the Medicare payment and their costs per transport.
Therefore, ambulance providers that can control their costs per transport
may have an advantage over those that cannot control their costs per
transport.
A 2003 GAO study found that transport costs are likely to be higher in
less densely populated rural areas because rural providers furnish fewer
transports and because fewer transports were linked to higher costs per
transport.34 As a result, we recommended that CMS adjust payments for
transports in rural counties with particularly low population density to
help ensure Medicare beneficiaries' access to ambulance services in those
areas. Subsequently, the MMA increased payments for super-rural transports
from July 1, 2004, through December 31, 2009. The report also found that
the majority of ambulance providers' costs were related to readiness--the
availability of ambulance and crew for immediate emergency response--and
were fixed costs. Fixed costs, such as staff on call, vehicles, building
space, and administration, generally do not increase as the number of
transports increases. Fuel costs and supplies are not fixed costs because
they increase with the number of transports.
Costs per Transport Were Highly Variable, Reflecting Differences in Certain
Provider Characteristics
Costs of ground ambulance transports were highly variable across providers
without shared costs; an average ambulance transport ranged from a low of
$99 to a high of $1,218 during 2004, the year for which we gathered data.
The variability of costs per transport reflected differences in certain
characteristics--volume and mix of transports; service areas (urban,
rural, and super-rural); productivity, which we defined as transports per
staffed hour; and amount of local tax support.35 As expected, low volume,
a greater percentage of ALS and more complex transports, and more
transports from super-rural areas were key characteristics that helped
explain why some providers without shared costs had higher costs per
transport. Two other provider characteristics--productivity and amount of
local tax support--were also associated with higher costs per transport
for providers without shared costs. Other provider and local area
characteristics--such as type of provider, region, building rent, and
price of fuel--did not significantly affect average costs per transport
among providers without shared costs.
^34GAO, Ambulance Services: Medicare Payments Can Be Better Targeted to
Trips in Less Densely Populated Rural Areas, [39]GAO-03-986 (Washington,
D.C.: Sept. 19, 2003).
^35Transports per staffed hour is the total number of transports divided
by the total number of hours that an ambulance and crew were staffed and
available to respond to an emergency call.
Providers' Reported Costs per Transport Were Highly Variable
Providers' average costs for a ground ambulance transport varied from $99
to $1,218--a range of more than $1,100--across providers without shared
costs in 2004. Figure 3 shows the wide variation in the reported costs per
transport among providers without shared costs. Five percent of providers
without shared costs had average costs per transport that were less than
$152, while 5 percent of providers had average costs per transport more
than $913. From our sample of providers without shared costs, we estimated
the average cost per transport at $415, with a 95 percent confidence
interval--the range within which we expect the population average cost per
transport to fall 95 percent of the time--of $381 to $450.36 This means
that the actual average cost per transport across ambulance providers in
the United States without shared costs was from $381 to $450 in 2004.
Figure 3: Distribution of Cost per Transport for Providers without Shared
Costs in 2004
Note: Based on a sample of 215 providers, weighted to represent more than
5,200 providers in the United States that did not share costs with
nonambulance services.
^36Because our cost information is estimated from a sample of providers,
all costs per transport are reported with confidence intervals. The range
of the confidence interval is affected by the variability of the responses
within the sample and the size of the sample. See app. I for a full
discussion of our sample, methods, and computations.
When we categorized providers without shared costs by service area and
compared the average costs per transport across the groups, average cost
per transport among super-rural providers was statistically significantly
different from that of urban providers, but rural providers' average cost
per transport was not statistically significantly different from that of
urban providers. The average cost per transport for super-rural providers
without shared costs was $538,37 statistically significantly different
from the $370 average cost per transport for urban providers without
shared costs. The 95 percent confidence interval for average costs per
transport among super-rural providers without shared costs ranged from
$448 to $628 and among urban providers without shared costs ranged from
$326 to $414. The average cost per transport among rural providers without
shared costs was $409 within a confidence interval spanning $354 to $465,
an interval that overlapped with the average costs per transport estimates
for both urban and super-rural providers without shared costs.
Certain Provider Characteristics Contributed to Differences in Costs per
Transport
The variability of costs per transport among providers without shared
costs reflected differences in certain provider characteristics. The
provider characteristics that contributed to significant differences in
costs per transport were volume, mix of transports, service area,
productivity, and amount of local tax support.38 (See table 2 and app. I
for a full description of our methods.)
^37The average cost per transport of super-rural providers without shared
costs was based on the average costs of survey respondents without shared
costs that served predominantly super-rural areas, and reflects the
different characteristics of those respondents, including the volume of
trips they provided. It is different from the estimated average cost of a
super-rural transport based on our regression analysis.
^38For each characteristic, we measured its effect on providers' average
cost per transport, independent of other characteristics, by assuming the
national average value for the other variables.
Table 2: Estimated Average Cost per Transport for Provider Characteristics
That Affect Costs
Estimated average
cost per
Provider characteristics Assigned values transporta
Volume of transports 2,000 or less $464
2,001 - 3,000 388
3,001 - 4,000 327
4,001 - 5,000 330
5,001 - 6,000 276
6,001 or more 330
Mix of transports Only BLS 360
ALS or more intensive 476
services
Service area Urban only 358
Rural only 420b
Super-rural only 545
Productivity--transports per 1 transport per 8 437
staffed hour hours (0.12)
5 transports per 8 386
hours (0.64)
Local tax support No tax support 392
Local tax support as 632
81% of revenues
Sources: GAO analysis of 2005 GAO Survey of Ambulance Services and 2004
Medicare claims.
aThe estimated average cost per transport is for providers without shared
costs. It was created by assigning a value to the provider characteristic
of interest for all cases and using the national average value for the
other characteristics. See app. I. for detailed information about our
methods.
bThe estimated cost of a rural transport was not significantly different
from the estimated cost of an urban transport.
Providers without shared costs that had lower transport volumes generally
had higher average costs than providers without shared costs with higher
transport volumes. Our analysis affirms the finding of our prior work,
that volume of transports was the main characteristic affecting providers'
costs per transport.39 Because most ambulance costs are fixed, and
therefore do not increase significantly when a provider completes more
transports, it is expected that as the number of transports provided
increases, associated costs per transport will be lower. In 2004, the
volume of transports completed by a provider without shared costs ranged
from 21 to more than 50,000. Estimated average cost per transport is
reduced, from $464 for providers without shared costs completing 2,000 or
fewer transports a year to $327 for those completing from 3,001 to 4,000
transports. Although estimated average cost per transport is slightly
higher for 4,001 to 5,000 transports, rising to $330, and again above
6,000 transports, rising to $330, every other volume category of provider
without shared costs had lower estimated average costs per transport than
the lowest volume group of 2,000 transports or less.
^39 [40]GAO-03-986 .
Also, as expected, we observed that average costs per transport were
higher for providers without shared costs that also had a greater
percentage of ALS and more complex transports (compared with BLS
transports) and those with a greater percentage of super-rural transports
(compared with urban transports)--two characteristics incorporated into
the national fee schedule to account for the additional costliness
associated with more intensive services and isolated service areas. ALS
and more complex transports completed by providers without shared costs
ranged from 0 to 100 percent of transports provided. We estimated that
providers without shared costs specializing in ALS and more complex
transport services had average costs of $476, which was 32 percent higher
than providers without shared costs specializing in BLS transport
services. We estimated that the average cost of a super-rural transport
was $545,40 while the average cost of an urban transport was $358. Rural
transports were not significantly higher cost than urban transports.
Two other provider characteristics--productivity and amount of local tax
support--were also associated with higher average costs per transport.
Costs were higher when providers without shared costs had lower
productivity or a lower ratio of transports per staffed hour. The average
level of productivity for providers without shared costs was 0.12, or
about one transport per 8 staffed hours, and had an estimated average cost
per transport of $437. The second highest level of productivity for these
providers was 0.64, or more than five transports per 8 staffed hours, and
had an estimated average cost per transport of $386. The impact of
productivity on average costs may be explained by the fixed costs incurred
in maintaining readiness--having an ambulance and crew available to
respond to emergency calls. Although providers may have discretion about
staffing and the ability to make backup arrangements to substitute for
additional staff, not all providers can increase productivity by
increasing the number of transports they provide or reducing the number of
staffed ambulance hours. For example, providers that operate in small or
isolated communities with one ambulance on call may serve only their own
community's needs and may not be able to expand their service area or
increase their volume of ambulance transports.
^40The estimated average cost of a super-rural transport is different from
the average cost per transport of super-rural providers without shared
costs that was explained in footnote 37. The estimated average cost of a
super-rural transport was calculated using regression analysis that
assumed all providers without shared costs only performed super-rural
transports and had the national average for the other variables in our
analysis. See our app. I for more details on our methodology.
Average costs were also higher for providers without shared costs that
derived a larger percentage of their total revenues from local tax
support. Among providers without shared costs, those with the largest
percentage of revenues from local tax support (81 percent of revenues) had
estimated average costs that were $240, or 61 percent, above those with no
local tax support. Again, this effect was independent of volume and mix of
transports, service area, cost of labor, use of volunteers, productivity,
and other provider and local area characteristics. The relationship
between greater local tax support and higher average costs may be
explained as the income effect: if an organization has more money, it is
able to and likely to spend more. Moreover, if costs increase without
resulting in additional transports, the average cost per transport will
increase.
Characteristics that did not significantly contribute to the variability
of average costs per transport among providers without shared costs
included type of provider; a provider's region, as measured by the nine
census divisions that defined the regional fee schedule; building rent;
and the price of fuel.
Average Payments under the National Fee Schedule Will Be Greater Than Average
Historical Payments, but Providers' Expected Medicare Margins Will Vary Greatly
Average ambulance national fee schedule payments in 2010 are estimated to
be 3 percent higher overall than payments in 2001, after adjusting for
inflation and assuming that providers bill the maximum amounts allowed.
However, the Medicare margins of providers without shared costs--whether
they make a profit or a loss on Medicare transports--will vary under the
national fee schedule. We cannot assess whether providers without shared
costs will break even, lose, or profit on average under the ambulance
national fee schedule in 2010 after all of the MMA temporary payment
provisions have expired, because the 95 percent confidence interval
surrounding the average Medicare margin spans from negative 14 percent to
positive 2 percent. However, across all providers without shared costs, we
estimate that 39 to 56 percent will have average Medicare payments above
their average costs per transport under the ambulance national fee
schedule even after all of the MMA provisions expire.
Average Payments under the National Fee Schedule Will Be Greater Than Average
Historical Payments and Will Be Redistributed from Urban Transports to Rural and
Super-Rural Transports
Compared with average historical payments in 2001, average payments under
the national fee schedule in 2010 will be 3 percent higher, after
adjusting for inflation and assuming providers will bill the maximum
amount allowed under the national fee schedule. According to our analysis,
urban transports will experience a decrease in payments, on average, while
rural and super-rural transports will receive an increase that is greater
than the overall increase. Average payments for rural and super-rural
transports will increase 20 percent and 15 percent, respectively, while
average payments for urban transports will decline 3 percent compared with
average payments prior to the fee schedule. (See table 3.)
Table 3: Payments Prior to and under the National Fee Schedule after MMA
Provisions Expire
Payment prior to Payments under
Transports national fee schedule national fee schedule Percentage change
Urban $309 $301 -3
Rural 303 363 20
Super-rural 379 437 15
National 309 319 3
Sources: GAO analysis of 2001 and 2004 Medicare claims.
Notes: All payments are in 2004 dollars. Payments under the national fee
schedule assume that providers charge the maximum allowed amount.
Expected Medicare Margins Will Vary Greatly
After all of the MMA temporary payment provisions expire, expected
Medicare margins under the national fee schedule will vary greatly among
providers without shared costs. When we compared expected payments under
the national fee schedule in 2010 with providers' costs per transport, the
resulting Medicare margins ranged from negative 194 percent for one
provider to positive 76 percent for another provider. This wide difference
is related to the great variability in reported costs among providers
without shared costs.
Among providers without shared costs, we estimated that the average
Medicare margin, or the average percentage difference between these
providers' Medicare payments and their costs, will be about negative 6
percent with a 95 percent confidence interval from negative 14 percent to
positive 2 percent. (See table 4.) This span in the confidence interval
means we cannot assess whether providers without shared costs would break
even, lose, or profit, on average, under the national fee schedule in 2010
after all of the MMA temporary payment provisions have expired. Similarly,
we estimated that the average Medicare margin for urban, rural, and
super-rural providers without shared costs will be negative under the
national fee schedule, but each estimate will fall within a broader
confidence interval range that includes positive Medicare margins. The
estimated Medicare margin for an urban provider without shared costs will
be negative 18 percent to positive 6 percent, while the estimated Medicare
margin for a rural provider without shared costs will be negative 13
percent to positive 12 percent. We estimated that a super-rural provider
without shared costs will have an estimated Medicare margin from negative
35 percent to positive 2 percent, making it more likely that the average
Medicare margin for any given super-rural provider without shared costs
would be negative rather than positive. However, given the confidence
intervals surrounding the estimated average margin for each subset of
providers without shared costs and the lack of statistical difference
between them, we cannot conclude with certainty that any subset of
providers would have significantly better or worse financial experience
under Medicare's national fee schedule than another. Rather, we can
conclude only that Medicare margins are likely to vary even among urban,
rural, and super-rural providers without shared costs.
Table 4: Expected Average Medicare Margins under the National Fee Schedule
for Providers without Shared Costs in 2004 Dollars
Average cost (95
Providers' Payment under percent Providers' average Medicare
predominate national fee confidence margins in percentage (95
service area schedule interval) percent confidence interval)
Urban $350 $370 ( 326 to -6(-18 to 6 )
414)
Rural 408 409 (354 to 465) -1(-13 to 12)
Super-rural 471 538 ( 448 to 628) -17(-35 to 2)
All 394 415 ( 381 to 450) -6(-14 to 2)
Sources: GAO analysis of 2005 GAO Survey of Ambulance Services and 2004
Medicare claims.
Notes: All payments and costs are in 2004 dollars. Payments under the
national fee schedule assume that providers charge the maximum allowed
amount. The range of the confidence interval is affected by the
variability of costs per transport within the sample and the size of the
sample. Providers' average Medicare margin is the average margin across
all providers in the sample.
When we assessed the likely experiences of all providers without shared
costs under the national fee schedule after all of the MMA temporary
payment provisions expire, we estimated that 39 to 56 percent of them will
have positive Medicare margins. Among urban and rural providers, 39 to 65
percent and 34 to 64 percent, respectively, will have positive Medicare
margins, according to our estimations. Among super-rural providers,
however, we estimate that 18 to 51 percent will have positive Medicare
margins, while 49 to 82 percent would have zero or negative Medicare
margins. (See fig. 4.) The breadth of these confidence intervals reflects
the variability of providers' costs in 2004 and expected financial
experience under the national fee schedule after MMA temporary payment
provisions expire.
Figure 4: Expected Medicare Margins for Urban, Rural, and Super-Rural
Providers without Shared Costs
Notes: The range of the confidence interval is affected by the variability
of costs per transport within the sample and the size of the sample.
Percentages and confidence intervals are rounded.
MMA Provisions Resulted in Greater Average Payments for Higher-Cost Super-Rural
Transports and Adjusted Payments Regionally Where No Significant Cost
Differences Were Observed
The MMA temporary payment provisions, which were implemented by CMS in the
second half of 2004, resulted in raised ambulance average payments
overall, particularly for super-rural transports, which we found typically
more costly to provide. Payment adjustments under the MMA's regional fee
schedule were not justified on the basis of regional cost differences, as
we did not find significant differences in average cost per transport
across regions.
Payment Increases Were Targeted to Higher-Cost Super-Rural Transports
When we compared ambulance payments in the first half of 2004, prior to
the implementation of the MMA provisions, with ambulance payments in the
second half of 2004, after the temporary payment provisions were
implemented and had their maximum effect, we found that payments, on
average, increased by 5 percent overall. Super-rural transports received
more substantial payment increases than urban or rural transports. (See
table 5.) After MMA temporary payment provisions were implemented, average
payments for urban and rural transports increased by 5 and 3 percent,
respectively, while average payments for super-rural transports rose by 12
percent, compared with average payments before the MMA provisions were
implemented. Increased payments for super-rural transports under the MMA
were in keeping with our finding that super-rural transports were more
costly than urban transports, independent of other characteristics that
affected ambulance costs.
Table 5: Average Payments prior to MMA Implementation and after
Implementation
Average payment per Average percentage
transport prior to Average payment per change in payment
Transports MMA transport under MMA under MMA
Urban $306 $322 5
Rural 358 370 3
Super-rural 442 497 12
National 322 338 5
Source: GAO analysis of 2004 Medicare claims.
Notes: The period prior to MMA implementation for which payments were
computed was January 1, 2004 through June 30, 2004. The period after
implementation of the MMA for which payments were computed was July 1,
2004 through December 31, 2004.
Regional Payment Adjustments Required by the MMA Were Not Justified on the Basis
of Regional Cost Differences
Regional payment adjustments under the MMA were not warranted on the basis
of regional cost differences. The MMA required a regional fee schedule,
which resulted in ambulance payments for similar services that differed
based on the region where they were provided. When comparing average
regional payments before the implementation of MMA provisions to payments
after implementation of the MMA, when the regional fee schedule had its
greatest effect, we found that average payments increased substantially
for some regions but not others. (See table 6.) However, we found no
significant differences in costs by region, after controlling for
differences in volume and mix of transports, cost of labor, service area,
and other characteristics that may have affected costs. The regional fee
schedule is due to expire on December 31, 2009.
Table 6: Percentage Changes in Average Payments prior to MMA
Implementation and after Implementation, by Region
Region Percentage change in average payments
Pacific 19
New England 12
Mountain 8
West South Central 7
West North Central 3
Middle Atlantic 3
South Atlantic 1
East North Central 1
East South Central 0
Source: GAO analysis of 2004 Medicare claims.
Notes: East South Central received a decrease in average payments of -0.04
percent, which rounds to 0 percent. The period prior to MMA implementation
for which payments were computed was January 1, 2004 through June 30,
2004. The period after implementation of the MMA for which payments were
computed was July 1, 2004 through December 31, 2004. See table 11 in app.
I for a listing of the regions, or census divisions, and their
corresponding states.
Medicare Beneficiaries' Use of Ambulance Transports Increased from 2001 to 2004,
Except in Super-Rural Areas
Nationally, the use of ambulance transports by Medicare beneficiaries
increased by 16 percent, from 2001, the year before the transition to the
national fee schedule began, to 2004, the year we studied. (See table 7.)
Medicare beneficiaries' use of ambulance transports in urban areas
experienced the greatest growth, 19 percent, while rural areas experienced
a modest increase of 6 percent. However, Medicare transports per 1,000
beneficiaries in super-rural areas decreased by 8 percent. The decrease in
Medicare beneficiaries' use of ambulance transports in super-rural areas
was driven mostly by a decline in the volume of transports rather than any
significant change in the number of beneficiaries or the demographic
characteristics of beneficiaries residing in super-rural areas. For
example, factors such as age, race, and gender remained stable in the
super-rural Medicare population. Meanwhile, Medicare beneficiaries' use of
ambulance transports increased in all regions from 2001 to 2004, including
one region that had a decrease in average payments under the MMA compared
with before the implementation of MMA payment provisions.
Table 7: Ambulance Transports per 1,000 Beneficiaries in Urban, Rural, and
Super-Rural Areas
Area Transports, 2001 Transports, 2004 Percentage change, 2001-2004
Urban 371 443 19
Rural 372 396 6
Super-rural 264 244 -8
Total 364 420 16
Sources: GAO analysis of 2001 and 2004 Medicare claims and CMS enrollment
data.
Conclusions
The diversity of the ambulance industry is reflected in its range of
organizations, services offered, staffing, revenue sources, and costs per
transport. We found that certain ambulance provider characteristics, such
as volume, mix of transports, service area, productivity, and amount of
local tax support, affected the cost per transport of providers without
shared costs. For some providers and communities, these characteristics
may be self-determined and may reflect those communities' preferences for
readiness, quality standards, and ambulance services offered. For example,
some communities may prefer to fund the greater costs of operating at a
higher level of readiness or being equipped with more sophisticated
transport vehicles and more highly trained staff. Other providers and
communities have little or no control over the characteristics that affect
providers' cost per transport. These communities, particularly more rural
areas with low population density, may be constrained by local conditions,
including their financial resources. Therefore, local conditions and
community preferences may explain some of the predicted variability in the
financial experience of providers without shared costs under the Medicare
national fee schedule.
We are unable to discern whether providers without shared costs would be
compensated appropriately under the national fee schedule for two reasons.
First, when providers experience the national fee schedule payments in
2010 after all of the MMA temporary payment provisions expire, they may
make changes to control or reduce their costs. The cost data we collected
were from 2004 and may not reflect any changes providers may make to
control or reduce their costs in response to the national fee schedule.
Second, we did not assess if Medicare beneficiaries are receiving quality
care that is delivered efficiently. There are no national performance
standards to use as benchmarks for determining quality and efficiency of
services or for assessing whether providers could increase productivity by
increasing the number of transports they provide or by reducing the number
of staffed ambulance hours. However, current efforts to develop a national
data system and indicators for EMS systems may, in the future, yield
useful tools for measuring efficiency and quality of ambulance services
under the Medicare program.
Based on our survey of ambulance costs, we were able to estimate that some
providers without shared costs would have positive Medicare margins under
the national fee schedule after the MMA provisions expire, while others
would have negative Medicare margins. Among super-rural providers, we
estimated that 18 to 51 percent would have positive Medicare margins.
However, 49 to 82 percent would have zero or negative Medicare margins.
Ideally, Medicare payments should be adequate to ensure beneficiary access
to services while using the program's resources judiciously. The decline
in use of super-rural ground ambulance transports from 2001 to 2004, a
time when payments for super-rural transports were increased, suggests
that Medicare payment levels may not be linked to the decreased
utilization of transports in super-rural areas. However declining
utilization coupled with potentially negative Medicare margins in
super-rural areas, which could be exacerbated when the MMA provisions
expire, raise questions as to whether Medicare payments will be adequate
to support beneficiary access in super-rural areas.
Recommendation for Executive Action
In light of the variability in ambulance providers' Medicare margins and
the potential for negative margins to have an impact on beneficiary
access, we recommend that the Administrator of CMS monitor utilization of
ambulance transports to ensure that Medicare payments are adequate to
provide for beneficiary access to ambulance services, particularly in
super-rural areas.
Agency and External Comments and Our Evaluation
We provided a draft of this report to CMS and to five associations that
represent the ambulance industry: the American Ambulance Association, the
National Association of State EMS Officials, the National Ambulance
Coalition, the National Volunteer Fire Council, and the International
Association of Fire Chiefs. CMS's written comments are reprinted in
appendix II.
CMS stated that, for the most part, the report reinforces its findings.
CMS also stated that it agreed with our recommendation that the agency
monitor utilization of ambulance transports to ensure that Medicare
payments are adequate to provide for beneficiary access to ambulance
services, particularly in super-rural areas. CMS noted that it would
continue to monitor ambulance rates and would make adjustments should the
original assumptions made during the development of the ambulance fee
schedule need to be changed. In addition, CMS also highlighted its
implementation of a refinement in the definition of rural areas that
should enable rural areas within urban areas to receive the benefit of
higher rural payments under the ambulance fee schedule.
CMS noted that we should have discussed in our conclusions the
implications of omitting "shared services" providers from our analysis, as
these providers tend to have higher costs. As we discussed in the report,
ambulance providers that could not separately report the costs of the
ambulance portion of their business were excluded because their cost data
were determined to be unreliable. Consequently, we have no basis or
information to suggest that providers with shared services have higher or
lower costs than other providers.
Ambulance industry associations generally agreed with our findings.
However, the associations raised various concerns regarding our
calculations and assumptions. Two associations questioned the inclusion in
our analysis of ambulance providers that used unpaid staff and suggested
that it might have been more appropriate to focus on providers who bear
the full cost of providing ambulance services. As we note in the report,
use of unpaid staff by ambulance providers is widespread with an estimated
two-fifths of the industry relying substantially on volunteers in 2004.
Thus, in order for our analysis to be representative of ambulance
providers, we included those that used volunteer staff. We recognize that
use of volunteer staff affects ambulance providers' costs and included the
percentage of volunteer hours as a control variable in our cost model.
Two associations were concerned that we did not allow for the effect of
Medicare bad debt in our analysis and may have therefore overestimated
payments. We acknowledge that bad debt will affect the percentage of costs
recoverable for providing ambulance services. However, we explicitly state
that our payment estimates assume providers are paid the full Medicare
payment amounts. It was beyond the scope of our study to estimate the
effect of Medicare bad debt on ambulance payments or to determine the
extent to which payments should be adjusted to reflect bad debt.
Two associations expressed concern that our analysis showed a 3 percent
increase in payments to ambulance providers with the transition from the
historical payment system to the fully implemented national fee schedule
and thought payments should have been relatively level. We note that our
analysis incorporated increases in mileage rates over time that likely
accounted for some of this increase. In addition, when simulating payments
under the national fee schedule, we assumed that providers would bill
Medicare for all services they were entitled to bill. However, as we noted
in the report, nearly half of the industry indicated that a state or local
government approves the fees they may charge, and some providers are
required to bill Medicare less than the allowed amount and therefore do
not receive the maximum Medicare payment allowed under the fee schedule.
This discrepancy between the state or local allowed amount and the actual
Medicare payment could also account for some of the 3 percent increase.
CMS and the associations also provided technical comments and
clarifications, which we incorporated as appropriate.
We are sending copies of this report to other interested congressional
committees and the Administrator of CMS. We will also provide copies to
others upon request. The report will also be available at no charge on the
GAO Web site at http://www.gao.gov .
If you or your staff have any questions, please contact me at (202)
512-7119 or [email protected]. Contact points for our Offices of Congressional
Relations and Public Affairs may be found on the last page of this report.
GAO staff who made major contributions to this report are listed in
appendix III.
Kathleen M. King
Director, Health Care
Appendix I: Data and Methods
This appendix describes, in detail, the data and methods we used to
respond to our research objectives and evaluate ambulance providers' costs
per transport and Medicare payments under the national ambulance fee
schedule. We conducted a survey of ambulance costs to collect cost data.
We relied on these survey data for much of our analyses of costs and
supplemented our survey results with information from other sources,
including Medicare claims data, as appropriate. We also analyzed Medicare
claims data to determine the effect of the national fee schedule and the
Medicare Prescription Drug, Improvement and Modernization Act of 2003
(MMA) temporary payment provisions on payments, as well as to describe
changes in beneficiaries' use of ambulance transports over time. We tested
the internal consistency and reliability of the data from our survey and
other non-Medicare data sources and determined that all data sources were
adequate for our purposes. We conducted our work from July 2004 through
April 2007 in accordance with generally accepted government auditing
standards.
National Survey of Ground Ambulance Providers' Costs
To collect data on ground ambulance providers' costs, revenues,
transports, and organizational characteristics for their most recently
completed fiscal year, we mailed a survey to a nationally representative
sample of 500 eligible ambulance service providers that billed Medicare in
2003.1 We used a two-stage sampling process to select a stratified,2
random sample of providers. In the first stage, we selected a preliminary
sample of hospital and nonhospital-based providers for screening. In the
second stage, we conducted telephone screening interviews to confirm
eligibility for the study and, for nonhospital-based providers, to
identify provider type. We obtained 321 completed surveys for a response
rate of 64 percent. We excluded two cost outliers. We also excluded from
our analysis providers that reported sharing ambulance costs with other
institutions or other nonambulance services, including, but not limited
to, all fire departments, after preliminary analysis revealed problems
with the reliability of their reported costs. The resulting sample size
was 215 providers. The results from our analysis are nationally
representative of all Medicare ambulance providers that can distinguish
their costs for providing ambulance services from the costs of other
services they provide.3 However, the small sample size and the variability
of reported costs reduce the precision of our estimates.
^1Our survey asked providers to report on their organizations' costs of
providing ground ambulance services.
^2The sample was stratified by types of ambulance providers:
hospital-based, volunteer, fire departments, government, and freestanding.
Survey Instrument Development
To develop our survey instrument, we reviewed other survey instruments and
analyses of ambulance cost data, consulted with experts in survey methods
and the ambulance industry, and tested our survey instrument. We reviewed
cost data collected in 1999 by Project HOPE Center for Health Affairs,
which was a nonprofit health policy research organization, during a survey
effort sponsored by the American Ambulance Association. We also reviewed
other surveys of emergency medical services, as well as industry and
association guidelines about emergency medical and ambulance services. In
addition to surveying ambulance providers on their costs and revenues, we
included questions to identify organizational and local area
characteristics that might affect ambulance costs, such as the number of
emergency transports and number of volunteer hours.
Industry experts and a survey specialist reviewed and commented on the
draft survey instrument. We conducted a pilot test of the survey with 104
ambulance service providers as well. We redesigned and refined the
instrument based on the experience of the pilot test. Then, to further
refine the wording of our survey questions, we asked four types of
ambulance service organizations and a former volunteer fire chief to
pretest the instrument and point out any issues they noted. These pretests
were conducted mostly by telephone--one pretest was in person.
Sample Design
We developed separate lists for all hospital-based and nonhospital-based
providers from information maintained by the Centers for Medicare &
Medicaid Services (CMS), the agency that oversees the Medicare program.
Next, we used a two-stage sampling process to select a stratified, random
sample of providers. In the first stage, we sorted each list by census
division and predominant service area (urban, rural, or super-rural),
began sampling the list at a random starting point, and chose providers at
regular intervals from their respective lists. This method of sampling
implicitly included representation across census divisions and service
areas. In the second stage, telephone screening interviews were conducted
with each sampled provider to confirm eligibility for the study and, for
nonhospital-based providers, to identify their type--volunteer, fire
department, government, or freestanding. Once a provider's type was
established, providers were stratified by type and randomly selected to
ensure somewhat equal representation among all provider types. We
developed initial sample rates for each type of provider and adjusted the
rates midway through the screening process. Each provider representative
was told at the close of the telephone screening call whether the provider
had been selected to participate in our survey. Finally, a survey
instrument was mailed to the selected providers.
^3Our cost analysis sample includes only those providers that could
apportion the ground ambulance component of any cost shared between their
ambulance services and other services. Thus, our results only reflect the
experiences of providers that did not share costs or that could feasibly
apportion them.
To identify nonhospital-based ambulance providers that billed Medicare, we
contacted Medicare carriers4 for a list of ambulance providers and matched
this list to 2003 carrier claims for ambulance services by provider
identification numbers. Our nonhospital-based sample frame included 12,082
unique provider identification numbers.5 We later learned that a number of
provider identification numbers in the nonhospital-based sample frame were
duplicate entries for ambulance providers that had more than one Medicare
provider identification number for the same location.6 We analyzed the
extent of duplication in the sample frame and reduced the estimated
population size to 7,968. Although our sample frame included a substantial
number of duplicate entries, there were only 28 duplicates among our
first-stage sample of 900 nonhospital-based providers. The sample weights
were adjusted to account for the duplicate entries in the sample frame.
To identify the total number of hospital-based ambulance providers that
billed Medicare, we matched Medicare's Provider of Service file with
Medicare Part A inpatient and outpatient claims for ambulance services. We
excluded skilled nursing facilities for a total of 828 hospital-based
ambulance service providers. We then selected a first-stage sample of 150
hospital-based ambulance providers.
^4Carriers are private companies that have contracts with Medicare to
administer payment of providers' bills for covered Medicare Part B
services.
^5We excluded provider identification numbers that had been selected to
participate in the pilot study.
^6Duplicate entries for the same nonhospital provider were identified by
name, address, and other information.
To make our survey sample representative of all Medicare ambulance
providers, the population from which the sample was drawn, we computed a
sample weight for each respondent provider. The computation of the sample
weight took into account the type of provider, sample rate, and the
response rate for the type of provider--hospital, volunteer, fire
department, government, or freestanding.
Survey Administration
We contracted with CODA Inc., an independent survey research firm, to
perform the telephone screening and administer the mailed survey
instrument.7 The contractor screened the 900 nonhospital-based and 150
hospital-based providers we selected for our first-stage sample, and
mailed the survey instrument to 500 eligible organizations, randomly
sampled by provider type. In order to properly select our sample with even
representation across all types of providers nationally, we designed our
sampling strata using data collected from our pilot test.
CODA Inc., administered the mailed surveys and conducted all follow-up and
data coding in coordination with us. Our survey period began in April 2005
and ran through September 2005. The survey instruments were mailed using
Federal Express and 2-day Priority Mail. Telephone contact was initiated 3
business days after the instrument was mailed to ascertain when the
respondent could return the completed instrument. Prompting and follow-up
requests for data were conducted by phone and occurred whenever providers'
returned instruments were incomplete, vague, or included conflicting
responses to key items. In some cases, follow-up requests involved
multiple contacts, faxing survey instruments, and spending an hour or more
on the phone with the respondent. On average, more than 4 phone calls were
made for each respondent with as many as 20 calls made to one respondent.
All requests for data were conducted by CODA Inc. staff following strict
protocols that we developed. Respondents were encouraged to contact CODA
Inc. and GAO via toll-free numbers, so that any questions or problems
could be resolved. All survey data were double-key entered into an
electronic file, and computer programs were checked for keying
discrepancies and data inconsistencies. In all, we received 321 completed
surveys; this represents a response rate of 64 percent.
^7CODA Inc. also administered the 1999 ambulance survey for Project HOPE
and our pilot survey. During the course of the contract, CODA Inc. became
Survey and Epidemiology Services Division of Social & Scientific Systems,
Inc.
Survey Data Validity and Reliability
In addition to the survey administration procedures described above, we
took several measures to ensure that the data reported on the survey were
valid and reliable.8 First, the survey instrument included items intended
to validate the reported cost data. We also used strict protocols during
follow-up to validate the reported cost data. For example, if respondents
could not provide cost breakdowns by the categories listed in our survey
instrument, a separate phone protocol was used to verify the reported cost
data.
Second, we tested the data for internal consistency and excluded cases
when necessary. Computer analyses were performed to identify and, where
possible, correct any inconsistencies in responses or other errors. We
also excluded 2 providers that appeared to be cost outliers. These
providers had costs per transport that were at least three standard
deviations above the mean of the standard statistical distribution (the
lognormal), and no other variables explained their extraordinary costs.
Through our analyses, we determined that the costs reported by providers
that shared costs with other institutions or offered other services
appeared to be unreliable. We found the costs reported by these "shared
costs" providers to be highly variable, which may reflect inconsistent
methods for separating staff time and other resources across different
services. Therefore, we excluded these providers--including but not
limited to all fire departments--from our analysis. The resulting sample
size was 215 providers, representing a population of more than 5,200
providers without shared costs.9
^8For a randomly selected subsample of 30 providers--6 of each type of
ambulance provider--we requested providers' financial records to match to
their survey responses. We were only able to validate costs, revenues, and
transport volumes for 2 providers in the subsample because providers
either did not return supporting documentation or the
categories--particularly those for costs--on providers' supporting
documentation were different from those on the survey. As a result, we
concluded that this method for validating survey data was not viable
because of differences in data reporting, particularly cost categories, on
providers' supporting documents.
^9We could not determine a separate response rate for our subgroup of
providers without shared costs because we could not identify which
nonrespondents in our original sample were providers without shared costs.
However, we have no reason to believe that the response rate for providers
without shared costs would be different from the overall survey response
rate.
Third, we compared information reported on the survey to information on
Medicare claims submitted by respondents, such as the number of Medicare
transports and percentage of emergency Medicare transports. All computer
syntax was peer reviewed and verified by separate programmers to ensure
that the syntax was written and executed correctly.
We used providers' total costs and total transports reported on the survey
to compute providers' average costs per transport. This cost information
and other information about revenues and provider characteristics were
used to model ambulance costs per transport. Although these survey data
were self-reported and had not been audited, based on efforts to validate
the data, computer testing, and corrections and comparisons with Medicare
data, we have concluded that they were sufficiently valid and reliable for
our purposes.
Interpretation of Confidence Intervals and Analysis of Nonrespondents
All sample surveys are subject to sampling error--that is, the extent to
which the survey results differ from what would have been obtained if we
had collected responses from every ambulance provider in the country.
Because we used a sample, it is only one of a large number of samples that
we might have drawn. As each sample could have provided different
estimates, we express our confidence in the precision of our particular
sample's results as a 95 percent confidence interval.10 This is the
interval that would contain the actual value for all providers for 95
percent of the samples we could have drawn. As a result, we are 95 percent
confident that the reported confidence intervals based on the mailed
survey include the true values for all providers. For this reason, all
costs per transport and provider margins are reported with their
confidence intervals.
We also analyzed 2004 Medicare claims data for survey nonrespondents and
compared this information with similar claims information for providers
without shared costs in our sample. Nonrespondents served predominantly
urban areas rather than rural or super-rural areas. On average,
nonrespondents completed about half the number of Medicare transports in
2004 compared with providers without shared costs. Nonrespondents also had
a higher percentage of basic life support (BLS) transports, as opposed to
advanced life support (ALS) and more complex transports, and about the
same percentage of nonemergency transports compared with providers without
shared costs.
^10The range of the confidence interval is affected by the variability of
the responses within the sample and the size of the sample.
It is unclear whether nonrespondents had higher or lower costs, on
average, than providers without shared costs. In our regression analysis
of the cost information for providers without shared costs, we found that
those with fewer transports per year generally had higher costs per
transport than those with more transports. Providers without shared costs
that served predominately urban areas had lower costs compared with
providers serving super-rural areas. We also found that providers with
higher percentages of BLS transports had lower costs compared to providers
with no BLS transports. Although nonrespondents' characteristics differed
from those of providers without shared costs, these differences were
associated with both higher and lower costs among providers without shared
costs. Therefore, we have no basis for concluding that nonresponse has
biased our cost estimates in any particular direction.
Modeling Ambulance Costs per Transport
We analyzed the relationship between providers' average costs of ambulance
transports and the provider and local area characteristics that may have
affected their average costs. We used regression analysis to examine the
effect of these characteristics on providers' costs per transport. We then
used the results from this regression analysis to predict the average
costs per transport across all providers without shared costs, based on
five key provider characteristics: (1) total transports per year, (2)
percentage of BLS Medicare transports, (3) percentage of Medicare
transports in rural and super-rural areas, (4) number of ambulance
transports per staffed hour, and (5) amount of revenue derived from local
tax support.
To perform these analyses, we identified measures and data sources for
each of the provider and local area characteristics that we identified as
potentially contributing to differences in costs per transport. A summary
of these characteristics, measures we used to assess their potential
relationship to costs per transport, and data sources used is presented in
table 8.
Table 8: Provider and Local Area Characteristics Included in Analysis of
Average Cost per Transport, 2004
Characteristic Measure Source of data
Transport volume Indicator of volume group: GAO Survey of
<=2,000, 2001-3000, Ambulance Services
3001-4,000, 4001-5000,
5001-6000, 6001+
Type of provider Staffing and organizational GAO Survey of
structure Ambulance Services
Productivity Transports per staffed GAO Survey of
ambulance houra Ambulance Services
Price of fuel Annualized average retail Oil Price
price of fuel in the Information Service
provider's zip code
Building rent Price per square foot of United States Postal
office space in the Service
provider's zip code
Cost of labor used by Geographic practice cost Medicare Ambulance
providers index (GPCI)b Fee Schedule Public
Use File
Percentage of volunteer GAO Survey of
hours, including on-call time Ambulance Services
Revenue derived from Community tax support as a GAO Survey of
local tax support percentage of total ambulance Ambulance Services
provider revenuec
Regulation of fees State or local government GAO Survey of
approval of fees charged Ambulance Services
Mix and intensity of Percentage of emergency GAO Survey of
transports transports Ambulance Services
Percentage of Medicare Medicare claims
transports that are BLS
Service area Percentage of Medicare Medicare claims
transports that are rural
Percentage of Medicare Medicare claims
transports that are
super-rural
Census division in Census Division 1 - New Medicare claims
which the ambulance England
provider was located Census Division 2 - Middle Medicare claims
Atlantic
Census Division 3 - East Medicare claims
North Central
Census Division 4 - West Medicare claims
North Central
Census Division 5 - South Medicare claims
Atlantic
Census Division 6 - East Medicare claims
South Central
Census Division 7 - West Medicare claims
South Central
Census Division 8 - Mountain Medicare claims
Census Division 9 - Pacific Medicare claims
Sources: 2005 GAO Survey of Ambulance Services, Oil Price Information
Service, United States Postal Service, Medicare Ambulance Fee Schedule
Public Use File, and Medicare claims.
aThis measure was computed as the total number of transports in the year
divided by the total number of hours that an ambulance and crew were
staffed and available to respond to an emergency call. This measure is
also known as unit hours of utilization.
bThe practice expense portion of the physician fee schedule's GPCI is used
to adjust the ambulance national fee schedule.
cThis measure was computed as the total amount of revenues derived from
community tax support divided by total revenues.
Average Cost Regression Analysis--Methods and Results
Our regression analysis modeled the average cost of a transport at the
provider level as a function of the provider and local area
characteristics described above. We modeled the lognormal distribution of
average cost per transport for a provider, which was calculated as the log
of total costs divided by the total number of transports for that
provider. We estimated the model using least squares11 and applied the
appropriate sample weights. We used the statistical program SUDAAN(R) in
SAS to estimate this model, which takes account of the sample
stratification and weighting to obtain appropriate parameter estimates and
standard errors. We tested for and found no specification problems in the
model resulting from heteroscedasticity, misspecification, or evidence of
a particular observation having undue influence.
Tables 9 and 10 show the regression results for estimating the
determinants of cost per transport. We ran two sets of average cost
regressions. The first regression results, in table 9, do not include two
provider characteristics, (1) transports per staffed ambulance hour and
(2) community tax support as a percentage of total ambulance provider
revenue, because data on these variables were not available for a number
of providers in our sample. Including these variables as explanatory
independent variables reduces the number of observations in the regression
from 205 to 157. Nevertheless, we wanted to measure the impact of these
policy variables, in addition to maintaining a larger sample size, in
order to have a more precise estimate of the impact of the other
explanatory variables. Therefore, we ran the same regression model with
and without these two explanatory variables.
^11Least squares is a common method of regression analysis.
Table 9: Results for Average Cost of an Ambulance Transport
Regression--Estimated Effects of Selected Provider and Local Area
Characteristics on the Average Cost of Ambulance Transports for Providers,
Not Including Impact of Productivity and Community Tax Support
Variable used to measure Parameter
Characteristic characteristic estimate t-value
Transport volumea 2,001-3,000 per year -0.18 -1.70*
3,001-4,000 per year -0.35 -2.98***
4,001-5,000 per year -0.34 -3.33***
5,001-6,000 per year -0.52 -2.71***
6,001+ per year -0.34 -2.86***
Cost of labor used by GPCI 1.34 2.27**
providers
Percentage of volunteer -0.28 -2.73***
hours, including on-call time
Regulation of fees State or local government -0.08 -1.00
approval of fees charged
Mix and intensity of Percentage of emergency 0.17 0.99
transports transports
Percentage of Medicare -0.28 -2.39**
transports that are BLS
Service areab Percentage of Medicare 0.16 1.61
transports that are rural
Percentage of Medicare 0.42 3.11***
transports that are
super-rural
Census division in which Census Division 1 - New -0.27 -1.66*
the ambulance provider England
was locatedc Census Division -0.24 -1.66*
2 - Middle
Atlantic
Census Division 0.23 1.61
3 - East North
Central
Census Division 5 - South 0.07 0.41
Atlantic
Census Division 6 - East -0.07 -0.43
South Central
Census Division 7 - West -0.16 -0.85
South Central
Census Division 8 - Mountain 0.15 0.97
Census Division 9 - Pacific -0.27 -1.39
Intercept 4.83 8.73***
R-squared 0.33
Observations 205
Sources: GAO analysis of 2005 GAO Survey of Ambulance Services, Medicare
Ambulance Fee Schedule Public Use File, and Medicare claims.
*** significant at the 1 percent level.
** significant at the 5 percent level.
* significant at the 10 percent level.
aTransports less than or equal to 2,000 per year is the excluded category.
In order for the regression model's parameters to be estimated, we needed
to exclude one of the transport volume categories.
bThe percentage of Medicare transports that are urban was the excluded
category. In order for the regression model's parameters to be estimated,
we needed to exclude one of the service area categories.
cThe West North Central census division was the excluded category. In
order for the regression model's parameters to be estimated, we needed to
exclude one of the census divisions.
Table 10: Results for Average Cost of an Ambulance Transport
Regression--Estimated Effects of Selected Provider and Local Area
Characteristics on the Average Cost of Ambulance Transports for Providers,
Including Impact of Productivity and Community Tax Support
Variable used to measure Parameter
Characteristic characteristic estimate t-value
Transport volumea 2,001-3,000 per year -0.17 -1.84*
3,001-4,000 per year -0.30 -2.69***
4,001-5,000 per year -0.34 -3.40***
5,001-6,000 per year -0.31 -1.72*
6,001+ per year -0.22 -1.92*
Cost of labor used by GPCI 1.29 2.62***
providers
Percentage of volunteer hours, -0.10 -0.79
including on-call time
Regulation of fees State or local government -0.11 -1.49
approval of fees charged
Mix and intensity of Percentage of emergency 0.06 0.29
transports transports
Percentage of Medicare -0.31 -2.25**
transports that are BLS
Service areab Percentage of Medicare 0.06 0.67
transports that are rural
Percentage of Medicare 0.26 1.93*
transports that are super-rural
Productivity Transports per staffed -0.24 -2.76***
ambulance hour
Community tax support Community tax support as a 0.59 3.81***
percentage of total ambulance
provider revenue
Census division in Census Division 1 - New England -0.24 -1.55
which the ambulance Census Division 2 -0.24 -1.42
provider was locatedc - Middle Atlantic
Census Division 3 0.17 1.04
- East North
Central
Census Division 5 - South 0.09 0.48
Atlantic
Census Division 6 - East South -0.16 -1.00
Central
Census Division 7 - West South -0.30 -1.48
Central
Census Division 8 - Mountain 0.20 1.38
Census Division 9 - Pacific 0.02 0.12
Intercept 4.95 9.81***
R-squared 0.51
Observations 157
Sources: GAO analysis of 2005 GAO Survey of Ambulance Services, Medicare
Ambulance Fee Schedule Public Use File, and Medicare claims.
*** significant at the 1 percent level.
** significant at the 5 percent level.
* significant at the 10 percent level.
aTransports less than or equal to 2,000 per year is the excluded category.
In order for the regression model's parameters to be estimated, we needed
to exclude one of the transport volume categories.
bThe percentage of Medicare transports that are urban was the excluded
category. In order for the regression model's parameters to be estimated,
we needed to exclude one of the service area categories.
cThe West North Central census division was the excluded category. In
order for the regression model's parameters to be estimated, we needed to
exclude one of the census divisions.
We estimated the average cost per transport associated with a range of
values for each of five provider characteristics by using our regression
result parameter estimates and fixed values for five provider
characteristics. For each of the five provider characteristics, we assumed
that all providers in the sample had the same value for one provider
characteristic and used the national average value for all other
characteristics. The five provider characteristics of interest were (1)
total transports per year, (2) percentage of BLS Medicare transports, (3)
percentage of Medicare transports in rural and super-rural areas, (4)
number of transports per staffed ambulance hour, and (5) amount of revenue
derived from local tax support. We estimated the cost per transport
associated with two or more values for each of five provider
characteristics. Our estimates are presented in table 2 of this report.
In order to create the estimated costs per transport for one value of a
provider characteristic, we performed the following steps. First, we
modified the data set we used in the regression model by assigning one
fixed value for one provider characteristic, such as the percentage of
super-rural Medicare transports, for all observations in the data set. For
this characteristic, we fixed the percentage of Medicare super-rural
transports to 100 percent and the percentage of Medicare rural transports
to 0 percent for all observations in the data set.12 Second, we used our
regression result parameter estimates for all the other characteristics in
the regression model and calculated an estimated average cost per
transport for each provider. Third, we computed the mean of these
estimated costs per transport. We report this hypothetical value as the
estimated average cost per transport associated with the fixed value of
the provider characteristic of interest. In the example above, we computed
our estimated average cost per transport of a super-rural Medicare
transport assuming that every provider in the sample provided only
super-rural Medicare transports and their other characteristics--volume
and mix of transports, productivity, and amount of local tax support--were
the national average.
^12The percentage of super-rural and rural Medicare transports were two
separate variables in our model and the percentage of urban Medicare
transports was our reference category. To assume that 100 percent of a
provider's Medicare transports were rural, we set the percentage of rural
Medicare transports to 100 percent and the percentage of super-rural
Medicare transports to 0 percent. To assume that 100 percent of a
provider's Medicare transports were urban, we set the percentage of
super-rural Medicare transports to 0 percent and the percentage of rural
Medicare transports to 0 percent.
Estimating Average Medicare Ambulance Payments and Use of Transports with Claims
Data
To assess the effect of the ambulance national fee schedule on Medicare
payments, we used Medicare claims data to compute average payments in
2001, before the implementation of the fee schedule, and in 2004, 2 years
after the phase-in of the national fee schedule had begun. For all
analyses of Medicare payments, we expressed payments in 2004 dollars. We
also used Medicare claims data and payment formulas as specified in
federal regulations to simulate average payments under the national fee
schedule in 2010, after all the MMA provisions are due to expire on
December 31, 2009, but computed these payments in 2004 dollars, the year
that best reflects the cost data collected in our survey. We used the
simulated payments to compute providers' Medicare margins, a comparison
that assumes that providers' cost structures under the fee schedule would
be the same as they were in 2004.13 We also assessed the effect of the MMA
temporary payment provisions on payments in 2004 using Medicare claims
data by examining the change in payments from the first half of the year,
before the MMA changes were implemented, with the second half of the year,
when MMA payment provisions had their maximum effect.
To estimate average Medicare ambulance payments in 2001 and 2004, we used
Medicare Part A and Part B claims data from Medicare's National Claims
History files. We constructed summary data sets with the number of
ambulance transports, miles, and their Medicare payments, total14 and by
level of service,15 at the zip code level for 2001 and 2004 to compare
payments over time. These data sets were classified by three geographic
levels: national; census divisions; and urban, rural, and super-rural
areas. Census divisions were categorized according to the U.S. Census
Bureau's guidelines. (See table 11 for a listing of census divisions.) CMS
designates urban, rural, and super-rural areas according to population
density.
^13The Medicare margin reflects the percentage difference between the
average Medicare payment and a provider's cost per transport. It was
computed as the difference between the average Medicare payment and the
cost per transport, divided by the payment.
^14Total ambulance payments in 2001 and 2004 included any payments made
for supplies in addition to mileage payments and service-level payments.
^15Under the Medicare program, there are seven levels of ambulance
transports. BLS and ALS transports each constitute two levels, emergency
and nonemergency. The remaining levels of service are ALS Level 2 and
specialty care transport (both of which involve invasive or specialized
care) and paramedic ALS intercept (when a paramedic provides ALS services
but does not transport the patient).
Table 11: Census Divisions
Census divisions States
New England Connecticut, Maine, Massachusetts, New Hampshire, Rhode
Island, Vermont
Middle Atlantic New Jersey, New York, Pennsylvania
East North Central Illinois, Indiana, Michigan, Ohio, Wisconsin
West North Central Iowa, Kansas, Minnesota, Missouri, Nebraska, North
Dakota, South Dakota
South Atlantic Delaware, District of Columbia, Florida, Georgia,
Maryland, North Carolina, South Carolina, Virginia,
West Virginia
East South Central Alabama, Kentucky, Mississippi, Tennessee
West South Central Arkansas, Louisiana, Oklahoma, Texas
Mountain Arizona, Colorado, Idaho, Montana, Nevada, New Mexico,
Utah, Wyoming
Pacific Alaska, California, Hawaii, Oregon, Washington
Source: U.S. Census Bureau.
Medicare Payment Calculations and Simulation
We chose to examine payments in three distinct time periods: 2001, the
first half of 2004, and the second half of 2004. The year before the
national fee schedule was implemented was 2001. The first half of 2004
represented a period after the implementation of the national fee schedule
but before the introduction of MMA temporary payment changes. It also
coincided with the year for which we have cost data. The second half of
2004 represented a period after the introduction of MMA provisions when
the regional fee schedule had its greatest effect.
In general, average Medicare payments were estimated as the sum of the
average base rate payment--that is, the amount paid for the level of
service16--and the average mileage payment. Base rate payments and mileage
payments for each period were computed separately. Average Medicare
payments were calculated nationally; by census division; and by urban,
rural, and super-rural areas. All payments are expressed in 2004 dollars.
^16Estimated average base rate payments for 2001 and 2004 included
ambulance supply payments. Under the fee schedule, the base rate payment
was meant to incorporate the transport and any necessary services or
supplies used during the transport.
For the first half of 2004, the average base rate payment was calculated
by dividing total base rate payments by the total number of transports
provided in the first half of 2004. For 2001 and the second half of 2004,
the average service-level base payment rate was calculated by dividing
total base rate payments for each level of service by the total number of
transports provided at each level of service. Then we computed a weighted
average base rate payment for 2001 and for the second half of 2004 by
applying the percentage of transports for each level of service (also
known as the mix of services) for the first half of 2004 to the
service-level base payment rates in 2001 and the second half of 2004,
respectively. The 2001 payments were adjusted to 2004 dollars by
multiplying weighted average base rate payment by the ratio of the 2004
Consumer Price Index for All Urban Consumers (CPI-U) over the 2001 CPI-U.
For simulated 2010 base rate payments, we multiplied the 2004 conversion
factor (CF) and applicable relative value unit (RVU) for each level of
service to compute the service level payments.17 Then, we computed a
weighted average base rate payment by applying the service mix from the
first half of 2004 to the 2004 service-level payments. The weighted
average base-rate payment was adjusted for regional differences in the
cost of providing ambulance services by the GPCI.18 In accordance with
CMS's payment methodology, only 70 percent of the average base rate
payment was adjusted by the GPCI.
Average mileage payments were calculated by multiplying the average miles
per transport in each geographic level from the first half of 2004 by
mileage rates in the other periods. (See tables 12 and 13.) For example,
when estimating average 2001 mileage payments for urban transports, the
2001 mileage rate applicable to urban areas was used,19 but the average
mile per transport in urban areas for the first half of 2004 was applied.
For simulated payments in 2010, mileage rates from the first half of 2004
were applied. For average national and census division mileage payments,
weighted mileage rates were applied reflecting the applicable percentage
of urban and rural miles in the first half of 2004. Because the average
mile per transport for super-rural areas was above 17, simulated payments
for super-rural areas were computed using the two rural mileage
rates--$8.48 for miles 1 through 17 and $5.65 for miles 18 through 50.
^17Ambulance RVUs account for the relative resources needed to provide
services during an ambulance transport. There is one RVU for each of the
seven levels of service.
^18In many cases, more than one GPCI would apply to our average base rate
payment. For these cases we applied an average GPCI, which was weighted by
the total number of transports in each geographic area of
interest--national; census division; or urban, rural, or super-rural area.
^19The 2001 mileage payments were eventually adjusted to 2004 dollars.
Table 12: Average Mile per Transport, First Half of 2004
Area Average mile per transport
National 9
Urban 7
Rural 13
Super-rural 20
East North Central 9
East South Central 12
Middle Atlantic 7
Mountain 9
Pacific 7
New England 7
South Atlantic 10
West North Central 14
West South Central 11
Source: 2004 Medicare claims.
.
Table 13: Mileage Rates
Period Urban Rural (1-17) Rural (18-50)
2001 $5.33 $5.33 $5.33
1/1/04 - 6/30/04 5.65 8.48 5.65
7/31/04 - 12/31/04 5.71 8.65 5.76
Simulated fee schedule 5.65 8.48 5.65
Sources: 67 Fed. Reg. 9100 and 2004 Medicare Ambulance Fee Schedule Public
Use File.
Notes: All mileage rates are expressed in 2004 dollars. The 2001 mileage
rate is based on the mileage rate for ground ambulance services used by
CMS when creating the national fee schedule. Rural mileage rates also
apply to super-rural transports.
Analysis of Medicare Margins
We used simulated Medicare payments and costs per transport to compute
Medicare margins under the national fee schedule after MMA temporary
payment provisions are due to expire, to determine how much providers
stood to lose or gain. To do this, we merged the survey results with
providers' Medicare payment information from 2004. For each provider, we
computed the mix of services and the average miles per transport for
January 2004 through June 2004. This information was used to simulate each
provider's average Medicare payment under the national fee schedule in
2010 after all MMA temporary payment provisions are set to expire on
December 31, 2009.20 We subtracted the provider's 2004 cost per transport
from the simulated average Medicare payment under the national fee
schedule, and then divided by the simulated average payment to compute the
provider's Medicare margin.
Estimating Use of Ambulance Transports by Medicare Beneficiaries
To estimate transports per 1,000 Medicare beneficiaries, we supplemented
our Medicare claims data with CMS enrollment data from 2001 and 2004,
which contained information regarding Medicare beneficiaries. The numbers
of Medicare beneficiaries in 2001 and 2004 were measured as the number of
months beneficiaries were covered by Medicare in each year divided by 12.
A ratio of the number of transports over the number of Medicare
beneficiaries was estimated in each year and multiplied by 1,000.
^20We simulated the average payments under the national fee schedule in
the same way described earlier in this appendix.
Medicare Claims Data Reliability
Medicare claims data, which are used by the Medicare program as a record
of payments made to health care providers, are closely monitored by both
CMS and Medicare carriers--contractors that process, review, and pay
claims for Medicare Part B-covered services, including ambulance services.
The data are subject to various internal controls, including checks and
edits performed by the carriers before claims are submitted to CMS for
payment approval. Although we did not review these internal controls, we
found the data to be sufficiently reliable for the purposes of this
report. We also assessed the reliability of CMS's enrollment data. We
found these data to be sufficiently reliable for the purposes of this
report.
Appendix II: Comments from the Centers for Medicare & Medicaid Services
Appendix III: GAO Contact and Staff Acknowledgments
GAO Contact
Kathleen M. King, (202) 512-7119 or [email protected]
Acknowledgments
In addition to the contact named above, Christine Brudevold, Assistant
Director; Jennie F. Apter; Carl S. Barden; Stella Chiang; Kevin J. Dietz;
James M. Fields; Leslie V. Gordon; Michael Kendix; and Jessica Cobert
Smith made key contributions to this report.
Related GAO Products
Ambulance Services: Medicare Payments Can Be Better Targeted to Trips in
Less Densely Populated Rural Areas. [43]GAO-03-986 . Washington, D.C.:
September 19, 2003.
Ambulance Services: Changes Needed to Improve Medicare Payment Policies
and Coverage Decisions. [44]GAO-02-244T . Washington, D.C.: November 15,
2001.
Emergency Medical Services: Reported Needs Are Wide-Ranging, With A
Growing Focus on Lack of Data. [45]GAO-02-28 . Washington, D.C.: October
12, 2001.
Rural Ambulances: Medicare Fee Schedule Payments Could Be Better Targeted.
[46]GAO/HEHS-00-115 . Washington, D.C.: July 17, 2000.
(290371)
GAO's Mission
The Government Accountability Office, the audit, evaluation and
investigative arm of Congress, exists to support Congress in meeting its
constitutional responsibilities and to help improve the performance and
accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.
Obtaining Copies of GAO Reports and Testimony
The fastest and easiest way to obtain copies of GAO documents at no cost
is through GAO's Web site ( www.gao.gov ). Each weekday, GAO posts
newly released reports, testimony, and correspondence on its Web site. To
have GAO e-mail you a list of newly posted products every afternoon, go to
www.gao.gov and select "Subscribe to Updates."
Order by Mail or Phone
The first copy of each printed report is free. Additional copies are $2
each. A check or money order should be made out to the Superintendent of
Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more
copies mailed to a single address are discounted 25 percent. Orders should
be sent to:
U.S. Government Accountability Office 441 G Street NW, Room LM Washington,
D.C. 20548
To order by Phone: Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061
To Report Fraud, Waste, and Abuse in Federal Programs
Contact:
Web site: www.gao.gov/fraudnet/fraudnet.htm E-mail:
[50][email protected] Automated answering system: (800) 424-5454 or (202)
512-7470
Congressional Relations
Gloria Jarmon, Managing Director, [email protected] (202) 512-4400 U.S.
Government Accountability Office, 441 G Street NW, Room 7125 Washington,
D.C. 20548
Public Affairs
Paul Anderson, Managing Director, [email protected] (202) 512-4800
U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548
www.gao.gov/cgi-bin/getrpt?GAO-07-383 .
To view the full product, including the scope
and methodology, click on the link above.
For more information, contact Kathleen M. King at (202) 512-7119 or
[email protected].
Highlights of [54]GAO-07-383 , a report to congressional committees
May 2007
AMBULANCE PROVIDERS
Costs and Expected Medicare Margins Vary Greatly
In 2002, Medicare implemented a national fee schedule designed to
standardize payments for ambulance services. The Medicare Prescription
Drug, Improvement and Modernization Act of 2003 (MMA) required GAO to
study ambulance service costs. GAO examined providers' costs of ground
ambulance transports in 2004 and factors that contributed to cost
differences; average Medicare ambulance payments expected under the
national fee schedule in 2010 and how those payments will relate to
providers' costs per transport; and changes that occurred in Medicare
beneficiaries' use of ambulance transports from 2001 to 2004. GAO
estimated costs of ambulance transports based on a nationally
representative survey of 215 ambulance providers that did not share costs
with nonambulance services. Providers that shared costs with other
institutions or services and could not report their costs for ambulance
services separately, such as fire departments, were excluded because their
reported costs appeared unreliable. GAO used its survey, Medicare claims,
and other data for its analyses.
[55]What GAO Recommends
GAO recommends that the Administrator of CMS monitor utilization of
ambulance transports to ensure that Medicare payments are adequate to
provide for beneficiary access to ambulance services, particularly in
super-rural areas. CMS agreed with GAO's recommendation.
Costs of ground ambulance services were highly variable across providers
that did not share costs with nonambulance services in 2004, reflecting
differences in certain provider and community characteristics. Costs per
transport among these providers varied from $99 per transport to $1,218.
Providers without shared costs that had higher costs per transport
typically had fewer transports per year, a greater percentage of
transports in which more than a basic medical intervention occurred, more
transports in super-rural areas (rural counties with lowest population
density), lower productivity--measured as number of transports furnished
per staffed hour, and a greater percentage of revenues from local tax
support.
Average payments under the national fee schedule in 2010 are expected to
be higher than historical payments, but providers' Medicare margins will
vary greatly. GAO could not assess whether, on average, providers without
shared costs would break even, lose, or profit under the national fee
schedule, because the average Medicare margin for providers without shared
costs was estimated to fall from negative 14 percent to positive 2
percent. However, GAO estimated that approximately 39 to 56 percent of
providers without shared costs would have average Medicare payments above
their average cost per transport under the national fee schedule in 2010.
From 2001 to 2004, utilization of ambulance transports per beneficiary
increased 16 percent overall. However, use declined by 8 percent in
super-rural areas.
Declining utilization coupled with potentially negative Medicare margins
in super-rural areas, which could be exacerbated when the MMA temporary
payment provisions expire, raise questions as to whether Medicare payments
will be adequate to support beneficiary access in super-rural areas.
Distribution of Cost per Transport for Providers without Shared Costs,
2004
Note: Based on a sample of 215 providers, weighted to represent more than
5,200 providers in the United States that did not share costs with
nonambulance services.
References
Visible links
39. http://www.gao.gov/cgi-bin/getrpt?GAO-03-986
40. http://www.gao.gov/cgi-bin/getrpt?GAO-03-986
43. http://www.gao.gov/cgi-bin/getrpt?GAO-03-986
44. http://www.gao.gov/cgi-bin/getrpt?GAO-02-244T
45. http://www.gao.gov/cgi-bin/getrpt?GAO-02-28
46. http://www.gao.gov/cgi-bin/getrpt?GAO/HEHS-00-115
54. http://www.gao.gov/cgi-bin/getrpt?GAO-07-383
*** End of document. ***