Medicare Payment: CMS Methodology Adequate to Estimate National
Error Rate (24-MAR-06, GAO-06-300).
The Centers for Medicare & Medicaid Services (CMS) estimated that
the Medicare program paid approximately $20 billion (net) in
error for fee-for-service (FFS) claims in fiscal year 2004. CMS
established two programs--the Comprehensive Error Rate Testing
(CERT) Program and the Hospital Payment Monitoring Program
(HPMP)--to measure the accuracy of claims paid. The Medicare
Prescription Drug, Improvement, and Modernization Act of 2003
directed GAO to study the adequacy of the methodology that CMS
used to estimate the Medicare FFS claims paid in error. GAO
reviewed the extent to which CMS's methodology for estimating the
fiscal year 2004 error rates was adequate by contractor type for
(1) the CERT Program, (2) the HPMP, and (3) the combined national
error rate (including both the CERT Program and the HPMP). GAO
reviewed relevant CMS documents and reports related to the CERT
Program and the HPMP. In addition, GAO reviewed work performed by
the Department of Health and Human Services (HHS) Office of
Inspector General (OIG) and its contractor that evaluated CMS's
fiscal year 2004 statistical methods and other aspects of the
error rate estimation process. GAO also conducted interviews with
officials from CMS, HHS's OIG, and their contractors.
-------------------------Indexing Terms-------------------------
REPORTNUM: GAO-06-300
ACCNO: A49880
TITLE: Medicare Payment: CMS Methodology Adequate to Estimate
National Error Rate
DATE: 03/24/2006
SUBJECT: Claims processing
Claims settlement
Erroneous payments
Errors
Medicare
Statistical data
Statistical methods
Cost estimates
Fee-for-Service Plans
Comprehensive Error Rate Testing Program
Hospital Payment Monitoring Program
******************************************************************
** 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-06-300
* Results in Brief
* Background
* CMS Programs to Monitor the Payment Accuracy of Medicare FFS
* CERT Program
* HPMP
* Estimation of the National Medicare FFS Error Rate
* CMS Methodology Adequate for Estimating the Error Rates in t
* Sampling Methods
* Medical Record Collection Process
* Identification and Categorization of Payment Errors
* Statistical Methods
* CMS Methodology Adequate for Estimating the Error Rate in th
* Sampling Methods
* Medical Record Collection Process
* Identification and Categorization of Payment Errors
* Statistical Methods
* CMS Methodology Adequate for Estimating the National Error R
* Concluding Observations
* Agency Comments
* GAO Contact
* Acknowledgments
* GAO's Mission
* Obtaining Copies of GAO Reports and Testimony
* Order by Mail or Phone
* To Report Fraud, Waste, and Abuse in Federal Programs
* Congressional Relations
* Public Affairs
Report to Congressional Committees
United States Government Accountability Office
GAO
March 2006
MEDICARE PAYMENT
CMS Methodology Adequate to Estimate National Error Rate
Medicare Error Rate
GAO-06-300
Contents
Letter 1
Results in Brief 5
Background 6
CMS Methodology Adequate for Estimating the Error Rates in the CERT
Program 16
CMS Methodology Adequate for Estimating the Error Rate in the HPMP 22
CMS Methodology Adequate for Estimating the National Error Rate 28
Concluding Observations 29
Agency Comments 30
Appendix I Scope and Methodology 32
Appendix II Fiscal Year 2004 Error Rate Information by Contractor
Type-Carriers, DMERCs, FIs, and QIOs 35
Appendix III Comments from the Department of Health and Human Services 40
Appendix IV GAO Contact and Staff Acknowledgments 42
Tables
Table 1: Medicare FFS Error Rates and Dollars of Claims Paid in Error,
Fiscal Year 2004 8
Table 2: National Medicare FFS Error Rate by Category of Error, Fiscal
Year 2004 14
Figures
Figure 1: Medicare FFS Error Rates Estimated through the CERT Program 9
Figure 2: Medicare FFS Error Rates Estimated through the HPMP 11
Figure 3: Medicare FFS Error Rates That Produce the National Error Rate 13
Abbreviations
CDAC Clinical Data Abstraction Center CERT Comprehensive Error Rate
Testing CMS Centers for Medicare & Medicaid Services DMERC durable medical
equipment regional carrier DRG diagnosis-related group FFS fee-for-service
FI fiscal intermediary GPRA Government Performance and Results Act of 1993
HHS Department of Health and Human Services HPMP Hospital Payment
Monitoring Program IPIA Improper Payments Information Act MAC Medicare
administrative contractor OIG Office of Inspector General OMB Office of
Management and Budget PPS prospective payment system QIO quality
improvement organization
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.
United States Government Accountability Office
Washington, DC 20548
March 24, 2006
The Honorable Charles E. Grassley Chairman The Honorable Max Baucus
Ranking Minority Member Committee on Finance United States Senate
The Honorable Joe L. Barton Chairman The Honorable John D. Dingell Ranking
Minority Member Committee on Energy and Commerce House of Representatives
The Honorable William M. Thomas Chairman The Honorable Charles B. Rangel
Ranking Minority Member Committee on Ways and Means House of
Representatives
The Centers for Medicare & Medicaid Services (CMS), the agency that
administers the Medicare program, monitors the accuracy of claims paid for
services provided to Medicare beneficiaries. Each fiscal year, CMS reports
an estimate of the claims paid in error based on a sample of claims from
previous years. In fiscal year 2004, CMS reported an error rate of 9.3
percent, which represented approximately $20 billion in error out of the
approximately $214 billion in fee-for-service (FFS) payments.1 The fiscal
year 2004 error rate estimated the percentage of FFS payments that did not
comply with Medicare's payment rules for a sample of claims that included
inpatient discharges that occurred from July 1, 2002, through June 30,
2003, as well other services that were paid in 2003. The fiscal year 2004
Medicare FFS error rate was significantly higher than the goal of 4.8
percent for that fiscal year, which CMS set under the Government
Performance and Results Act of 1993 (GPRA).2
1Unless otherwise specified, dollars paid in error and error rates
discussed in this report are net amounts. Net dollars paid in error were
calculated by subtracting dollars paid in error that were due to
underpayments from those that were due to overpayments. The net dollars
paid in error were then used to estimate the error rate. CMS also reported
gross dollars paid in error and error rates in fiscal year 2004. Gross
dollars paid in error were calculated by adding dollars paid in error that
were due to underpayments to those that were due to overpayments.
CMS uses several types of contractors to ensure the payment accuracy of
Medicare claims,3 including carriers,4 durable medical equipment regional
carriers (DMERC),5 fiscal intermediaries (FI),6 and quality improvement
organizations (QIO).7 Using contractor-specific error rate information,
CMS estimates an error rate for each type of contractor; the agency
produces a national Medicare error rate by aggregating the four
contractor-type error rates. In its fiscal year 2004 Medicare error rate
report, CMS stated that it planned to use error rate information to help
determine the underlying reasons for claim errors, such as incorrect
coding, and implement corrective actions.8 In a congressional testimony in
July 2005, the Director of CMS's Office of Financial Management stated
that CMS plans to create performance incentives for contractors.9 CMS is
also implementing a multiyear contractor reform initiative, which will
reduce the number of contractors responsible for paying claims.
2GPRA requires agencies to develop multiyear strategic plans, annual
performance goals, and annual performance reports. See Pub. L. No. 103-62,
107 Stat. 285.
3In a few cases, program safeguard contractors are responsible for
ensuring the payment accuracy of Medicare claims. Program safeguard
contractors are Medicare contractors that conduct activities to address or
prevent improper payments.
4Carriers are health insurers and pay claims submitted by physicians,
diagnostic laboratories and facilities, and ambulance service providers.
5DMERCs are health insurers and pay claims submitted by durable medical
equipment suppliers. In fiscal year 2004, a program safeguard contractor,
TriCenturion, was responsible for medical review and for lowering the
error rates in its region.
6FIs are almost exclusively health insurers and pay claims submitted by
home health agencies, non-prospective payment system (PPS) hospitals,
hospital outpatient departments, skilled nursing facilities, and hospices.
PPS is a reimbursement method used by Medicare where the payment is made
based on a predetermined rate and is unaffected by the provider's actual
costs.
7QIOs (formerly known as peer review organizations) are responsible for
ascertaining the accuracy of coding and payment of paid Medicare FFS
claims for acute care inpatient hospital stays-generally those that are
covered by PPS-for Medicare beneficiaries in all 50 states, the District
of Columbia, and Puerto Rico. Unlike carriers, DMERCs, and FIs, however,
QIOs do not process and pay claims. These activities are conducted by FIs.
8See Department of Health and Human Services, Centers for Medicare &
Medicaid Services, Improper Medicare Fee-for-Service Payments Report
Fiscal Year 2004 (Baltimore, Md.: December 2004).
To monitor the accuracy of Medicare FFS claims paid by contractors, CMS
established two programs-the Comprehensive Error Rate Testing (CERT)
Program and the Hospital Payment Monitoring Program (HPMP).10 Through the
CERT Program, CMS monitors payment decisions made by three types of
contractors-carriers, DMERCs, and FIs. It does this through a review of
the claims and submitted medical record documentation to ensure that there
is support for the payment based on the information reviewed for a sample
of paid claims. CMS uses a similar process for the HPMP for a sample of
claims that are reviewed by QIOs for accuracy of payment.
The Department of Health and Human Services (HHS) Office of Inspector
General (OIG)11 estimated the error rate for each fiscal year from 1996
through 2002. CMS made significant changes to the methodology, including
substantially increasing the size of the sample used to estimate the error
rate, when it assumed responsibility for estimating the Medicare error
rate in fiscal year 2003. CMS has continued to make changes to the
methodology in subsequent years.
The Medicare Prescription Drug, Improvement, and Modernization Act of 2003
requires that we study the adequacy of the methodology that CMS used to
estimate Medicare error rates and to make recommendations as deemed
appropriate.12 Specifically, we report on the extent to which the
methodology used by CMS to estimate the fiscal year 2004 Medicare error
rates was adequate (1) by contractor type (carrier, DMERC, and FI) for the
CERT Program, (2) by contractor type (QIO) for the HPMP, and (3) for the
combined national error rate (including the CERT Program and the HPMP).
9See Senate Homeland Security and Governmental Affairs Subcommittee on
Federal Financial Management, Government Information and International
Security, Hearing on Medicare and Medicaid Improper Payments, Statement of
the Director of Office of Financial Management, Centers for Medicare &
Medicaid Services, 109th Congress, July 12, 2005.
10Each program monitors the accuracy of paid claims that constitute
approximately 50 percent of Medicare's FFS payments annually.
11OIG regularly conducts audits, evaluations, and investigations
pertaining to HHS programs.
12Pub. L. No. 108-173, S: 921(b)(3), 117 Stat. 2066, 2388-89.
To conduct our analysis of the adequacy of the methodology that CMS used
to estimate fiscal year 2004 Medicare error rates, we reviewed relevant
documents, including CMS's Medicare error rate reports for fiscal years
2003 and 2004, CERT and HPMP program documentation, and HHS OIG reports
evaluating the fiscal year 2004 CERT Program and HPMP.13 In addition, we
reviewed work performed by an OIG contractor that evaluated CMS's
statistical sampling and estimation methodology for the fiscal year 2004
Medicare error rate, including the contractor's report and supporting
workpapers. We interviewed OIG officials; OIG contractor staff; CMS
officials; and staff of the CERT subcontractor responsible for calculating
the error rates for carriers, DMERCs, and FIs and the national error rate
for fiscal year 2004. Commenting on the adequacy of the methodology used
in any other years was beyond the scope of our work. However, it is
important to note that changes in the methodology may affect the
estimation of the error rates and thus the comparability of these rates
over time.
As part of our assessment of the adequacy of the methodology that CMS used
to estimate the Medicare error rates for fiscal year 2004, we reviewed the
reliability of these estimates by examining the precision of the
contractor-specific error rates, the contractor-type error rates, and the
national error rate. Precision is the amount of variation between an
estimate (such as the error rate for a sample of Medicare FFS paid claims)
and the result that would be obtained from measuring the entire population
(such as the error rate for all Medicare FFS paid claims). We examined
precision of the error rate estimates by assessing relative precision,
which is the standard error14 of the error rate estimate divided by the
estimate itself. Estimates with lower relative precision are more
reliable. For the purposes of this report, we established that a relative
precision of no greater than 15 percent was within the acceptable
statistical standard for precision.15 We chose relative precision because
it allows for better comparison of the reliability of the range of error
rates across contractors.16
13Since the creation of the CERT Program and the HPMP in 2003, OIG has
conducted annual reviews of the programs as part of its oversight of work
performed for HHS by contractors.
14The standard error is a measure of variation around the estimate, in
this case, the error rate.
15See, for example, M.H. Hansen, W.N. Hurwitz, and W.G. Madow, Sample
Survey Methods and Theory, vol. I (New York, N.Y.: John Wiley & Sons,
Inc., 1953), 130.
During the course of our work, CMS published a report in November 2005
that included its fiscal year 2005 error rates.17 While an evaluation of
the methodology used to estimate the fiscal year 2005 error rates was
outside the scope of our work, we reviewed the report and included
references in this report where appropriate.
For more information on our scope and methodology, see appendix I. We
performed our work from April 2005 through March 2006 in accordance with
generally accepted government auditing standards.
Results in Brief
We found that the methodology used by CMS for the CERT Program was
adequate to estimate the fiscal year 2004 error rates by contractor type
(carrier, DMERC, and FI). CMS's sample of 120,000 claims was sufficiently
large to reliably estimate the error rate and was appropriately selected.
Further, CMS used systematic sampling with a random start, a method that
is designed to ensure that the sample is representative of the population.
CMS also had appropriate procedures in place to collect medical records
from providers, such as physicians, durable medical equipment suppliers,
and hospital outpatient departments, which supported the paid claims.
Additionally, the processes used by CMS to identify and categorize payment
errors were adequate because they ensured that the reviews conducted of
the medical records supporting the paid claims were performed according to
the established procedures for the CERT Program. This included adequate
qualifications and training of those individuals conducting the medical
record reviews. Further, CMS used valid statistical methods to estimate
the fiscal year 2004 carrier, DMERC, and FI contractor-type error rates
and standard errors.
16To provide illustration, consider that one contractor has an error rate
of 11.9 percent with a standard error of 2.1 percent and a second
contractor has an error rate of 20.4 percent also with a standard error of
2.1 percent. The standard errors are the same, but relative precision,
which is calculated by dividing the standard error by the error rate
estimate, illustrates that the reliability of the estimates is different.
Relative precision of the error rate estimate for the first contractor is
17.6 percent, while the relative precision of the error rate estimate for
the second contractor is 10.3 percent. This indicates that the second
contractor's error rate estimate is more reliable.
17See Department of Health and Human Services, Centers for Medicare &
Medicaid Services, Improper Medicare FFS Payments Long Report (Web
Version) for November 2005. 2005.
https://www.cms.hhs.gov/apps/er_report/preview_er_report.asp?from=public&which=long&reportID=3
(downloaded Jan. 26, 2006).
We found also that the methodology used by CMS for the HPMP to calculate
the fiscal year 2004 contractor-type error rate for QIOs was adequate to
reliably measure claims paid in error. We found the sampling methods to be
adequate because CMS's sample of approximately 40,000 claims was
sufficiently large to estimate the QIO contractor-type error rate. It was
also representative of the population from which it was drawn in terms of
average dollar amount per claim. Based on our review of oversight work of
the HPMP conducted by OIG, we also found that the process used in the HPMP
to collect the medical records that support the claims selected for review
was adequate. Additionally, the processes CMS used to identify and
categorize payment errors were adequate because they ensured that the
reviews conducted of the medical records supporting the paid claims were
performed according to established procedures for the HPMP. This included
adequate qualifications and training of those individuals conducting the
medical record reviews. CMS also used valid statistical methods to
estimate the QIO contractor-type error rate and standard error.
The fiscal year 2004 error rates by contractor type (carrier, DMERC, FI,
and QIO) were appropriately aggregated to determine the national Medicare
error rate through the use of a valid statistical method. CMS estimated
the national Medicare error rate by averaging the error rates of the four
contractor types (carrier, DMERC, FI, and QIO), weighted by each
contractor type's proportion of total Medicare FFS payments.
In written comments on a draft of this report, HHS noted that we found the
CMS methodology adequate for estimating the fiscal year 2004 national
Medicare FFS error rate. HHS also noted that CMS is continually committed
to refining the processes to estimate, as well as lower, the level of
improper payments in the Medicare FFS program.
Background
In fiscal year 2003, CMS assumed responsibility for estimating the
national Medicare error rate, a responsibility that had previously been
held by HHS OIG. OIG began estimating the national Medicare error rate in
fiscal year 1996,18 and continued doing so for each subsequent fiscal year
through 2002. The transfer of responsibilities for estimating the national
Medicare error rate to CMS coincided with the implementation of the
Improper Payments Information Act of 2002 (IPIA). The IPIA requires
federal agencies to estimate and report annually on the extent of
erroneous payments in their programs and activities.19 The IPIA defines an
improper payment as any payment that should not have been made or that was
made in an incorrect amount, including both under- and overpayments. All
agencies that identify a program as susceptible to significant improper
payments, defined by guidance from the Office of Management and Budget
(OMB) in 2003 as exceeding both 2.5 percent of total program payments and
$10 million,20 are required to annually report to Congress and the
President an estimate of improper payments and report on corrective
actions.
In addition to estimating the national Medicare error rate for purposes of
compliance with the IPIA, CMS also began producing contractor-specific
error rate estimates beginning in fiscal year 2003 to identify the
underlying causes of errors and to adjust action plans for carriers,
DMERCs, FIs, and QIOs. To produce these contractor-specific error rate
estimates for fiscal year 2004, CMS sampled approximately 160,000 claims.
The contractor-specific error rate information was then aggregated by the
four contractor types (carrier, DMERC, FI, and QIO), which were ultimately
combined to estimate the national Medicare error rate. Under the
methodology previously used by OIG to estimate the national Medicare error
rate, 6,000 claims were sampled. While the sample size used by OIG was
sufficient to estimate the national Medicare error rate, it was not
sufficient to reliably estimate the contractor-specific error rates.
Additionally, the increased sample size improved precision of the national
Medicare error rate estimate.
18According to OIG testimony in February 2000, OIG began estimating the
national Medicare error rates in fiscal year 1996 as part of its audit of
CMS's financial statements. See House Committee on the Budget, Statement
of Inspector General, Department of Health and Human Services, Hearing on
Medicare and Medicaid: HHS High-Risk Programs, 106th Congress, February
17, 2000.
19Pub. L. No. 107-300, 116 Stat. 2350 (codified at 31 U.S.C. S: 3321
note).
20OMB Mem. M-03-13 (2003).
CMS Programs to Monitor the Payment Accuracy of Medicare FFS Claims
The objective of the CERT Program and the HPMP is to measure the degree to
which CMS, through its contractors, is accurately paying claims. Through
the CERT Program, CMS monitors the accuracy of Medicare FFS claims that
are paid by carriers, DMERCs, and FIs. In fiscal year 2004, the Medicare
error rates by contractor type as estimated through the CERT Program were
10.7 percent for the carrier contractor type, 11.1 percent for the DMERC
contractor type, and 15.8 percent for the FI contractor type. (See table
1.)
Table 1: Medicare FFS Error Rates and Dollars of Claims Paid in Error,
Fiscal Year 2004
Error rate Dollars paid in error (in
CMS program Contractor type (percentage) billions)
CERT Program Carrier 10.7 $6.5
DMERC 11.1 1.0
FI 15.8 9.3
HPMP QIO 3.6 3.1
National Medicare All contractor 9.3 $19.9
FFS error rate types
Source: CMS.
Notes: This table reflects net Medicare FFS error rates and dollars of
claims paid in error. Based on data provided in CMS's fiscal year 2005
error rate report, we calculated the net Medicare FFS error rates and net
dollars paid in error for fiscal year 2005 by contractor type as follows:
carriers- 6.0 percent and $4.1 billion; DMERCs-8.6 percent and $0.8
billion; FIs-3.2 percent and $2.0 billion; and QIOs-3.8 percent and $3.5
billion. The national Medicare FFS error rate and dollars paid in error
were 4.4 percent and $10.3 billion. See Department of Health and Human
Services, Centers for Medicare & Medicaid Services, Improper Medicare FFS
Payments Long Report (Web Version) for November 2005. 2005.
https://www.cms.hhs.gov/apps/er_report/preview_er_report.asp?from=public&which=long&reportID=3
(downloaded Jan. 26, 2006).
Through the HPMP, CMS monitors the accuracy of paid Medicare FFS claims
for acute care inpatient hospital stays-generally those that are covered
under the prospective payment system (PPS). For fiscal year 2004, the
Medicare error rate for the QIO contractor type, as estimated through the
HPMP, was 3.6 percent. (See table 1.)
CERT Program
To estimate contractor-specific Medicare FFS error rates for the CERT
Program, CMS reviews a sample of claims from each of the applicable
contractors, which included 25 carriers, 4 DMERCs, and 31 FIs for the
fiscal year 2004 error rates. These error rates are then aggregated by
contractor type. (See fig. 1.) For fiscal year 2004, CMS contracted with
AdvanceMed to administer the CERT Program. AdvanceMed sampled
approximately 120,000 claims submitted from January 1, 2003, through
December 31, 2003, to estimate the fiscal year 2004 contractor-specific
and contractor-type error rates for the CERT Program.
Figure 1: Medicare FFS Error Rates Estimated through the CERT Program
For each of the approximately 120,000 sampled claims, AdvanceMed requested
the medical records from the provider that rendered the service or from
the contractor that processed the related claim, if the contractor
previously performed a medical review on the claim. If a provider did not
respond to the initial request for medical records after 19 days,
AdvanceMed initiated a series of follow-up procedures in an attempt to
obtain the information. The follow-up procedures with nonresponding
providers for fiscal year 2004 included three written letters and three
contacts by telephone. Additionally, in fiscal year 2004, OIG followed up
directly with nonresponders on claims over a certain dollar amount. If
medical records were not received within 55 days of the initial request,
the entire amount of the claim was classified by AdvanceMed as an
overpayment error.
When medical records were received from the provider or from the
contractor, CERT medical review staff reviewed the claim (which billed for
the services provided) and the supporting medical records (which detailed
the diagnosis and services provided) to assess whether the claim followed
Medicare's payment rules and national and local coverage decisions.21
Claims that did not follow these rules were classified by AdvanceMed as
being in error. Providers whose claims were reviewed were allowed to
appeal these claims, and if the error determination for a claim was
overturned through the appeals process, AdvanceMed adjusted the error rate
accordingly. For the fiscal year 2004 error rate, AdvanceMed notified
individual carriers, DMERCs, and FIs of their respective payment errors.22
HPMP
For the HPMP, CMS analyzes a sample of claims across QIOs to estimate
Medicare error rates by state, because QIOs are organizations with
state-based service areas. CMS estimated the QIO contractor-type error
rate by aggregating the QIO error rate estimates for each of the 50
states, the District of Columbia, and Puerto Rico. (See fig. 2.) Through
the HPMP, CMS sampled approximately 40,000 claims for acute care inpatient
hospital discharges that occurred from July 1, 2002, through June 30,
2003, to estimate the fiscal year 2004 state-specific and contractor-type
error rates for QIOs.
21In Medicare, decisions about whether and under what circumstances new
procedures or devices are covered are made nationally by CMS or locally by
Medicare contractors for beneficiaries in their service areas.
22In the fiscal year 2005 error rate report, CMS reported that carriers,
DMERCs, and FIs collected overpayments identified during the November 2005
error rate reporting period. Further, CMS reported that the agency will
instruct carriers, DMERCs, and FIs to make payments to providers in
underpayment cases identified for the November 2006 and later reports. See
Department of Health and Human Services, Centers for Medicare & Medicaid
Services, Improper Medicare FFS Payments Long Report (Web Version) for
November 2005.
Figure 2: Medicare FFS Error Rates Estimated through the HPMP
For fiscal year 2004, CMS contracted with two organizations known as
Clinical Data Abstraction Centers (CDAC)-AdvanceMed and DynKePRO-that were
responsible for requesting medical records from providers for each of the
approximately 40,000 sampled claims. Each CDAC was responsible for
reviewing the sampled claims, which were assigned on the basis of the
geographic location where the discharge occurred. Upon receipt of the
medical records, CDAC admission necessity reviewers screened the related
claims for the appropriateness of the hospitalization and, with the
exception of claims from Maryland, coding specialists independently
recoded diagnosis-related groups (DRG) based on the records submitted.23
Because Maryland does not use DRG coding, nonphysician reviewers screened
claims from Maryland to determine whether the length of the acute care
inpatient hospital stay was appropriate.24, 25 Claims that failed the
screening process, including those where the admission was determined to
be unnecessary or where an inappropriate DRG code was used, were forwarded
to the QIO responsible for the state where the discharge occurred for
further review. Records not received by the CDACs within 30 days of the
request for information were "canceled" and referred to the QIO to be
processed as overpayment errors caused by nonresponse. The QIO referred
these claims to the FI responsible for paying the claim for the necessary
payment adjustments.
At the QIO, claims forwarded from the CDACs underwent further review,
primarily medical necessity admission reviews and DRG validations.
Determinations of error were made by QIO physician reviewers. Providers
whose claims were reviewed were given the opportunity to provide comments
or discuss the case and pursue additional review, which could result in an
appeal to an administrative law judge. After the matter was resolved,
resulting in a determination that a provider was either underpaid or
overpaid, the QIO forwarded the claim to the FI for payment adjustment.
23DRG coding is the classification system used by Medicare to group
patients according to diagnosis, type of treatment, age, and other
criteria. Under PPS, hospitals are paid a predetermined rate for treating
patients based on the specific DRG category, regardless of the actual cost
of care for the individual.
24Maryland is the only state that does not use the PPS system for acute
care inpatient hospitals. Maryland instead has an alternative payment
system, known as an all-payer system, in which the state decides each
hospital's level of reimbursement and requires that all payers be charged
the same rate for the same service. Medicare and Medicaid pay the
state-approved rates.
25Claims from Maryland with length of stay errors are considered medically
unnecessary services. Length of stay reviews identified cases of potential
delayed discharge. For example, the patient was medically stable, and
continued hospitalization was unnecessary.
Estimation of the National Medicare FFS Error Rate
CMS estimated the national Medicare FFS error rate by combining the three
contractor-type error rates (carrier, DMERC, and FI) from the CERT Program
and the one contractor-type error rate (QIO) from the HPMP. (See fig. 3.)
Figure 3: Medicare FFS Error Rates That Produce the National Error Rate
Medicare FFS claims that were paid in error as identified by the CERT
Program and the HPMP for the fiscal year 2004 error rates were sorted into
one of five categories of error:
o Insufficient documentation: Provider did not submit sufficient
documentation to support that the services billed were actually
provided.
o Nonresponse: Provider did not submit any documentation to
support that the services billed were actually provided.
o Medically unnecessary services: Provider submitted sufficient
documentation, but the services that were billed were deemed not
medically necessary or the setting or level of care was deemed
inappropriate.
o Incorrect coding: Provider submitted documentation that
supported a different billing code that was associated with a
lower or higher payment than that submitted for the services
billed.
o Other: Provider submitted documentation, but the services
billed did not comply with Medicare's benefit or other billing
requirements.
See table 2 for the national Medicare FFS error rate by category
of error for fiscal year 2004.
Table 2: National Medicare FFS Error Rate by Category of Error,
Fiscal Year 2004
Source: CMS.
Notes: This table reflects net Medicare FFS error rates generated
by both the CERT Program and the HPMP. Numbers do not sum to total
because of rounding.
As reported in CMS's fiscal year 2004 Medicare error rate report,
the agency planned to use the error rates to help determine the
underlying reasons for claim errors, such as incorrect coding or
nonresponse, and implement corrective action plans for carriers,
DMERCs, FIs, and QIOs.26 Draft statements of work, dated February
and April 2005, for carriers, DMERCs, and FIs set goals for
contractors to achieve a paid claims error rate of less than a
certain percentage, to be determined by CMS. According to the
standards for minimum performance on QIO statements of work that
ended in 2005 for some QIOs and 2006 for other QIOs,27 QIOs are
evaluated on 12 tasks, one of which is the HPMP. QIOs have to meet
the performance criteria standards on 10 tasks set forth by CMS to
be eligible for a noncompetitive contract renewal.
CMS's use of the error rates is being done in the context of the
agency's current effort to significantly reform its contracting
efforts for the payment of Medicare claims.28 By July 2009, CMS
plans to reduce the total number of contractors responsible for
paying Medicare claims to 23 total contractors, which the agency
refers to as Medicare administrative contractors (MAC). CMS also
plans to institute performance incentives in the new contracts,
which will be based on a number of different factors, including
the Medicare error rates. According to CMS's report to Congress on
Medicare contracting reform, CMS believes that the consolidation
of Medicare contractors and the integration of processing for
Medicare claims29 will lead to a reduced Medicare error rate.30
The methodology used by CMS in the CERT Program to estimate error
rates by contractor type (carrier, DMERC, and FI) in fiscal year
2004 was adequate. We found that the sample size and the use of
systematic sampling with a random start were adequate to reliably
estimate the Medicare error rates by contractor type. The CERT
Program also had adequate processes in place to collect medical
records and to accurately identify and categorize payment errors.
The statistical methods that CMS used to estimate the
contractor-type error rates were valid.
The sample size that CMS used in the CERT Program, approximately
120,000 claims, was sufficiently large to produce reliable
estimates of the fiscal year 2004 Medicare error rates by
contractor type (carrier, DMERC, and FI). CMS selected 167 claims
each month on a daily basis from each of the 60 contractors,
including 25 carriers, 4 DMERCs, and 31 FIs.31 This sample
generated error rate estimates by contractor type within
acceptable statistical standards, such as relative precision of no
greater than 15 percent.32 Specifically, the error rate for the
carrier contractor type was 10.7 percent with a relative precision
of 3.7 percent, the error rate for the DMERC contractor type was
11.1 percent with a relative precision of 13.5 percent, and the
error rate for the FI contractor type was 15.7 percent with a
relative precision of 4.5 percent.
Further, we found that the sampling methods were adequate because
CMS used a systematic sample with a random start.33 Sampling
methods that employ a random start are designed to ensure that the
sample selected is representative of the population from which it
is drawn. We reviewed CERT Program documentation, which described
the use of a systematic sample with a random start. The OIG
contractor reviewed the computer program used for the CERT Program
sample selection and verified that the claims were selected
according to the documentation. CMS officials told us that the
CERT Program conducts tests to compare the sampled claims to the
population of claims. For example, CMS compared the percentage of
claims sampled in each category of Medicare-covered service to the
percentage of claims in the population by category of
Medicare-covered service. CMS provided us with an example of this
test for one contractor's claims from January 2003 through June
2003.
While the relative precision of the fiscal year 2004 error rate
estimates by contractor type for the CERT Program was within
acceptable statistical standards of no greater than 15 percent,
the relative precision of half of the contractor-specific error
rate estimates was not. (See app. II for contractor-specific error
rate information, including the estimates and corresponding
relative precision, for carriers, DMERCs, and FIs.)
Thirty of the 60 contractor-specific error rates had relative
precision that were not within acceptable statistical standards.34
Additionally, the relative precision of the contractor-specific
error rates showed wide variation within each contractor type.
Relative precision among carriers ranged from 8.9 percent to 17.0
percent; among DMERCs, relative precision ranged from 12.3 percent
to 20.7 percent; and among FIs, relative precision ranged from
10.3 percent to 42.5 percent. As demonstrated by the range in
relative precision among FIs, for example, the error rate estimate
for one FI was nearly four times more reliable than the error rate
estimate for another.
The variation in relative precision among the contractor-specific
error rate estimates was due, in part, to the sampling method CMS
used for the CERT Program. CMS took an equal sample size from each
contractor despite the fact that individual contractors accounted
for varied amounts of Medicare claim volumes and total payments.
For example, the claim volume for carriers in 2003 ranged from a
minimum of 5.3 million claims to a maximum of 206 million claims;
total payments for carriers in 2003 ranged from a minimum of $168
million to about $6.7 billion.
CMS officials told us that they plan to reallocate the CERT
Program sample at the contractor level by increasing the sample
size for those contractors that are not reaching CMS's targeted
precision and by decreasing the sample size for those contractors
that are reaching targeted precision and achieving low error
rates. In September 2005, CMS officials reported that this change
to the methodology is expected to be implemented for the fiscal
year 2007 error rate estimation, which will be based on claims
processed in parts of 2006 and 2007. We support CMS's planned
changes to its sampling methodology. We believe that reallocation
of the sample as planned by CMS will improve the relative
precision of these estimates. If future samples were based on the
volume of claims or total payments of each contractor and the
relative precision of the contractor-specific error rate rather
than on the current basis of an equal allocation across
contractors, relative precision would likely be improved for the
contractor-specific error rates of those targeted contractors that
were allocated a larger sample. This is because relative precision
improves with increased sample size. There would also likely be
decreased variation in relative precision across all
contractor-specific error rates.35 These results could be achieved
without increasing the overall sample size for the CERT Program.
Based on our review of oversight work conducted by OIG, we found
that the process CMS used to collect medical records from
providers for the CERT Program was adequate. Staff of AdvanceMed,
the CMS contractor responsible for administering the CERT Program,
were responsible for requesting medical records for each of the
approximately 120,000 sampled claims used to estimate the fiscal
year 2004 error rates. According to an OIG review of CMS's
corrective actions to improve nonresponse in the CERT Program for
fiscal year 2004, AdvanceMed conducted a timely and systematic
follow-up with providers that did not respond to initial requests
for medical records.36 For the medical records collection process
for the fiscal year 2004 error rates, CMS implemented corrective
actions in the CERT Program to address the factors associated with
the high rate of nonresponse experienced during the medical
records collection process for the prior fiscal year. According to
the CMS fiscal year 2003 error rate report, for example, the
agency found that some nonresponse in fiscal year 2003 was due to
providers' lack of familiarity with AdvanceMed.37 In previous
years when OIG had responsibility for estimating the Medicare
error rate, OIG requested medical records directly from providers;
providers were familiar with OIG and understood the importance of
complying with the requests. However, when the responsibility for
estimating the Medicare error rate was transferred to CMS, many
providers were unfamiliar with AdvanceMed and may have been
reluctant to submit medical records to an unknown company. Another
factor that caused provider nonresponse in fiscal year 2003,
according to the CMS report, was providers' confusion about the
submission of medical records within the constraints of the
privacy regulations issued by HHS under the Health Insurance
Portability and Accountability Act of 1996,38 which limit the use
and release of individually identifiable health information.
According to the CMS report, CMS found that providers were
sometimes unaware that sending medical records to the CERT Program
contractor was permissible under the regulations. As reported in
the OIG review cited previously, CMS implemented corrective
actions that increased provider compliance with medical record
requests in fiscal year 2004. According to the OIG report, CMS
conducted educational efforts to clarify the role of AdvanceMed.
Additionally, OIG further reported that CMS took action to address
providers' concerns about compliance with the privacy regulations
by revising its request letters to providers to highlight
AdvanceMed's authorization, acting on CMS's behalf, to obtain
medical records as requested. OIG told us that CMS instructed
carriers, DMERCs, and FIs to refer certain claims for
nonresponding providers to OIG for follow-up.39
These improvements in the process used to collect medical records
in the CERT Program helped reduce nonresponse. According to
information provided to us by CMS, the percentage of error caused
by nonresponse in the CERT Program decreased from 61 percent for
fiscal year 2003 to 34 percent in fiscal year 2004.40 According to
CMS's fiscal year 2005 error rate report, the agency continued
several corrective actions to address nonresponse for sampled
claims for the fiscal year 2005 error rates.41 Further, beginning
with claims sampled to estimate the fiscal year 2006 Medicare
error rates, CMS transferred the medical record collection duties
to a second contractor, Lifecare Management Partners, which the
agency refers to as the CERT Program documentation contractor. CMS
officials told us that the CERT Program documentation contractor
is automating the medical record collection process and
eliminating paper copies of documentation.
Based on our review of OIG's fiscal year 2004 CERT Program
evaluation, we concluded that the processes used in the CERT
Program to identify and categorize payment errors for fiscal year
2004 were adequate because the medical record reviews were
performed appropriately and the CERT Program staff conducting the
reviews were adequately trained and qualified.42 Staff of the CERT
Program contractor, AdvanceMed, reviewed the medical records to
verify that claims were processed according to Medicare payment
rules; if not, a claim was found to be in error and assigned to
one of five categories of error (insufficient documentation,
nonresponse, medically unnecessary, incorrect coding, or other).
We reviewed work conducted by OIG that found AdvanceMed, the CMS
contractor responsible for administering the CERT Program, had
appropriate controls in place to ensure that the medical record
reviews were performed in accordance with established CERT Program
procedures. We also reviewed work by OIG, which examined the
educational and training requirements for medical record reviewers
as established in the CERT Program and assessed selected training
files for selected medical record reviewers. OIG officials told us
that they found these selected CERT Program medical record
reviewers to be adequately trained and qualified.
OIG found that AdvanceMed did not complete all required quality
assurance reviews within the designated time frame. CMS told OIG
that it planned to reduce AdvanceMed's workload. AdvanceMed
conducts quality assurance reviews on a sample of medically
reviewed claims to validate the initial reviewer's decision on
whether a claim was paid in error. OIG found that for the fiscal
year 2004 CERT Program, AdvanceMed completed 984 of the required
2,587 quality assurance reviews by the required date. To determine
whether these quality assurance reviews ensured the reliability of
the CERT Program claims review process, OIG randomly sampled 45 of
the 2,587 claims selected for quality assurance reviews. Of these
45 claims, AdvanceMed had completed a quality assurance review on
5 claims. OIG reported that the results of the 5 quality assurance
reviews confirmed the results of the initial medical record
reviews. Further, OIG reported that AdvanceMed stated that a
backlog of medical reviews prevented the completion of the
required quality assurance reviews within the prescribed time
frame. In response to the OIG report on the fiscal year 2004 CERT
Program evaluation, CMS commented that with Lifecare Management
Partners assuming responsibilities for medical record collection
for the fiscal year 2006 Medicare error rate estimation,
AdvanceMed's workload would be reduced. As a result, CMS commented
that this will free up the necessary resources for AdvanceMed to
comply with the quality assurance requirements. Further, in its
response to the OIG report, CMS commented that both AdvanceMed and
Lifecare Management Partners are required to report to the agency
on the results of the quality assurance activities conducted.
According to OIG's evaluation of the fiscal year 2005 CERT
Program, OIG found that AdvanceMed completed all of the required
quality assurance reviews.43
We found that the statistical methods used to estimate the error
rates and standard errors by contractor type (carrier, DMERC, and
FI) for the CERT Program were adequate. Based on our review of the
computer programming code that generated the error rate estimates
and standard errors by the CERT Program subcontractor responsible
for calculating the contractor-type error rates, The Lewin Group,
we found that the statistical methods were based on standard
statistical principles and were used appropriately. For each
contractor type, the stratified combined ratio estimation method
was used to calculate the error rate by taking the difference
between the overpaid dollars and the underpaid dollars divided by
the total dollars paid by Medicare for FFS claims of each
contractor type.44 The payment errors from the sample were then
extrapolated to the population for each contractor type to
estimate total payment errors. Further, The Lewin Group used a
standard statistical method to calculate the standard errors of
each of the contractor-type error rates.45 This method is
appropriate for obtaining the standard error of an estimate when
the stratified combined ratio estimation method is used and is
valid for large sample sizes, such as that used for the CERT
Program.
We found that the methodology used by CMS was adequate to produce
a reliable estimate of the fiscal year 2004 Medicare error rate
for the one contractor type (QIO) in the HPMP. We found the
methodology adequate because the sample size was large enough to
produce a reliable error rate estimate. Additionally, the sample
was representative of the population. We found also that the
methodology was adequate because the HPMP contractors responsible
for collecting the medical records for the sampled claims, as well
as for identifying and determining errors, had appropriate
controls in place to ensure that established procedures were
followed. Further, the statistical method that CMS used to
calculate the contractor-type error rate was valid.
The sample size that CMS used for the HPMP, about 40,000 claims,
was sufficiently large to produce a reliable estimate of the
fiscal year 2004 error rate for the QIO contractor type. Using a
systematic sample, CMS selected 62 discharge claims per month for
the District of Columbia, Puerto Rico, and each state except
Alaska. CMS selected 42 claims per month for Alaska. The QIO
contractor-type error rate was 3.6 percent with a relative
precision of 5.6 percent. The relative precision for the QIO
contractor-type error rate estimate is within acceptable
statistical standards (a relative precision of no greater than 15
percent).
For the QIO contractor-type error rate to be a reliable estimate,
it was necessary that the sample of discharge claims from which
the error rate was estimated be representative of the population
from which it was drawn. CMS's documentation stated that the HPMP
used a systematic sample selection process with a random start,
which is a generally accepted method of sampling that is designed
to ensure that the sample drawn is representative of the
population. Our review of the computer programming code that
selected the sample, however, found that a random start was not
used.46 To determine whether the HPMP sample was compromised by
the lack of a random start and whether it represented the
population from which it was drawn, we examined the OIG
contractor's comparison of the June 2003 sample to a re-created
version of the June 2003 population file from which the sample was
drawn.47 Based on our review, we found that the HPMP sample was
representative of the population from which it was drawn in terms
of average dollar amount per claim.
While relative precision of the fiscal year 2004 QIO
contractor-type error rate estimate was within acceptable
statistical standards, relative precision of most of the
state-specific QIO error rate estimates was not. (See app. II for
state-specific QIO error rate information, including the error
rate estimates and corresponding relative precision.) Only three
states' error rate estimates-Kentucky, Massachusetts, and New
Mexico-had relative precision of less than 15 percent.
Additionally, there was wide variation in relative precision of
the state-specific QIO error rate estimates. Relative precision of
the state-specific QIO error rates ranged from 10.5 percent in
Massachusetts to 83.3 percent in Mississippi. The differences in
relative precision of these state-specific QIO error rate
estimates indicate that the error rate estimate for the QIO that
served Massachusetts was eight times more reliable than the error
rate estimate for the QIO that served Mississippi. The variation
in relative precision was due, in part, to the sampling methods
used by CMS for the HPMP. CMS took an equal sample size for each
state except Alaska, despite the fact that there was significant
variation between states in the overall volume of discharge claims
and total payments. The number of discharges per state varied from
a low of 15,166 in Wyoming to a high of 825,845 in Florida.48
Similarly, total dollars paid for acute-care inpatient hospital
stays varied from less than $100 million in Wyoming to a high of
$7.5 billion in California.
Although in February 2006 a CMS official told us the agency has no
plans to reallocate the HPMP sample, CMS could adopt a similar
sampling strategy as it plans to do for the CERT Program. If
future state samples were based on the volume of discharge claims
or total payments per state and the relative precision of the
state-specific QIO error rates, rather than on the current basis
of an equal allocation per state, relative precision would likely
be improved for the state-specific QIO error rates in those states
that were allocated a larger sample since relative precision
improves as sample size increases. There would also likely be
decreased variation in relative precision across all
state-specific QIO error rates.49 These results could be achieved
without increasing the overall sample size for the HPMP.
In addition to issues with the wide variation of relative
precision of the state-specific QIO error rate estimates, we also
found large differences in the average dollar amount per claim
between the state-specific samples for some states and the
respective state populations. These differences suggest that the
samples drawn for more than half of the states were not
representative of each state's population. Based on our
examination of the OIG contractor's comparison of the state
samples and the state populations for June 2003, we found that the
ratio of the average dollar amount per claim in a state's sample
to the average dollar amount per claim in a state's population
varied from 62 percent in Maryland to 143 percent in Kentucky.
Twelve states had a ratio above 110 percent, and 16 states had a
ratio below 90 percent.50 It is still possible for the national
HPMP sample to be representative of the national HPMP population
even if all of the state-specific samples are not representative
of their state populations. The larger size of the HPMP sample
overall mitigates the problems identified in the smaller
state-specific samples.
Based on our review of oversight work of the HPMP conducted by
OIG,51 we found that the process CMS used for collecting medical
records from providers was adequate. OIG selected 46 discharge
claims that were sampled for the HPMP to determine if the CDACs,
AdvanceMed and DynKePRO, followed established HPMP procedures for
obtaining and reviewing medical records to identify payment
errors. OIG found that the CDACs generally had appropriate
controls in place to ensure that the medical records were obtained
and reviewed according to established HPMP procedures. Of the 46
discharge claims reviewed, OIG found that in two instances a
required follow-up letter to the provider was not sent due to an
error by a substitute CDAC employee. However, the medical records
for these two discharge claims were obtained within 30 days of the
original request, which resulted in no adverse effect on the error
rate estimates. Overall, nonresponse for fiscal year 2004
represented approximately 5.1 percent of the total QIO
contractor-type error rate of 3.6 percent, or 0.2 percent of all
discharge claims reviewed through the HPMP.
The issue with providers not responding to requests for medical
records was not as significant an issue for the HPMP as it was for
the CERT Program. According to the CMS report on the fiscal year
2005 error rate, nonresponse was less problematic in the HPMP
because of several factors, including the following: (1) providers
were more likely to respond to requests from the HPMP since the
average claim value was higher than the average claim value in the
CERT Program;52 (2) providers were more familiar with the HPMP
than with the CERT Program; and (3) providers were paid the cost
of providing medical records by the HPMP, but not by the CERT
Program.53
Based on our review of OIG's fiscal year 2004 HPMP evaluation,54
we concluded that the CDACs (AdvanceMed and DynKePRO) generally
had processes in place to adequately identify and categorize
claims paid in error in the HPMP for fiscal year 2004. OIG
officials told us that they found the medical record reviewers,
both admission necessity reviewers and DRG coding specialists, at
the two CDACs met CMS's qualifications for these positions.55 As
part of its review of the fiscal year 2004 HPMP, OIG reviewed 46
discharge claims that were part of the sample for estimating the
QIO contractor-type error rate. Based on that review, OIG reported
that the CDACs generally had appropriate controls in place to
ensure that admission necessity and DRG validation reviews were
performed in accordance with CMS established procedures and that
the results of those reviews were adequately maintained, updated,
and reported.
As part of the internal HPMP quality control process, two
activities were conducted regularly to ensure the reliability and
accuracy of CDAC reviews both within each CDAC and across the two
CDACs. Each CDAC randomly chose 30 claims per month to be reviewed
by two of its medical record reviewers for intra-CDAC tests. Each
CDAC compared the results of the two medical record reviews to
determine the reliability of reviews within the CDACs and reported
the results of the comparisons to CMS. The CDACs performed
inter-CDAC tests to assess the reliability of the reviews between
the two CDACs. For these tests, an additional 30 claims were
chosen at random per quarter by each of the CDACs for review by a
medical records reviewer at the other CDAC. As part of its
evaluation of the fiscal year 2004 HPMP, OIG selected 45 claims
that went through the intra-CDAC process and 42 claims that went
through the inter-CDAC process to determine if these quality
control activities ensured the reliability of the CDAC review
process. OIG reported that the quality control reviews were
generally operating effectively to ensure the reliability of the
review process and the consistency of the error rate determination
decisions.56
From the same evaluation of the fiscal year 2004 HPMP, OIG found
that the CMS contractor tasked with calculating the dollar amounts
paid in error, Texas Medical Foundation, used a method that
produced an amount of dollars in error that in some cases differed
from what OIG found to be the amount of dollars in error. For
claims identified by a QIO as having errors caused by changes in
DRG codes, Texas Medical Foundation used a method that produced
different dollar amounts in error than would have been produced if
it had used the software that FIs used to pay the original
discharge claims.57 The Texas Medical Foundation calculated a
different amount in error for about 76 percent of 200 incorrectly
coded claims that OIG reviewed. However, OIG reported that the
differences did not have a significant effect on the QIO
contractor-type error rate estimate. A CMS official told us that
the agency has not invested in modifying the software for use by
the Texas Medical Foundation for technical and financial reasons.
For example, the software requires modifications using a specific
programming language for which CMS has limited personnel with the
needed expertise.
We verified the statistical methods CMS used to estimate the QIO
contractor-type error rate and standard error in the HPMP by
reviewing the computer programming code that produced this
information. We found that the methods CMS used were adequate
because they were based on standard statistical methods and were
applied appropriately. To estimate the QIO contractor-type error
rate, CMS weighted58 each state-specific QIO error rate according
to that state's share of the total Medicare FFS payments for
acute-care inpatient hospital claims nationwide. This method is
referred to as a stratified mean per unit estimation.59 Like the
CERT Program, CMS used a standard statistical method to calculate
the standard error of the estimate.60 In our review of the
computer programming code that generated the QIO contractor-type
error rate estimate, we found that CMS used annual instead of
monthly weights in its estimate of the annual total dollars paid
in error.61 It would have been more appropriate for CMS to have
used monthly weights because the HPMP sample was drawn on a
monthly, not an annual, basis. However, when we reviewed the OIG
contractor's comparison of the estimate of annual dollars paid in
error using annual weights to what the estimate would have been
had CMS used monthly weights, we concluded that the use of annual
weights did not significantly affect the QIO contractor-type error
rate estimate. A CMS official told us and provided us with
documentation that beginning with the HPMP's fiscal year 2005
error rate estimation process, monthly weights are being used.
CMS appropriately combined the error rates under the CERT Program
and the HPMP to estimate the fiscal year 2004 national Medicare
error rate. CMS estimated the national Medicare error rate by
averaging the error rates of the four contractor types (carrier,
DMERC, FI, and QIO), weighted by each contractor type's share of
total Medicare FFS payments. Likewise, CMS calculated the standard
error, or precision, of the national error rate based on the
standard error of each of the four types of contractors' error
rate estimates, weighted by each contractor type's proportion of
total Medicare FFS payments. The methods CMS used to calculate the
national error rate and the standard error were statistically
valid, since the units of measurement, which in this case were
Medicare claims, of the four error rates that were combined were
mutually exclusive (independent) among contractor types.62 Each
contractor type consisted of multiple individual contractors.
These contractors were independent in that one contractor's
estimated error rate or standard error did not affect the
estimates of other contractors, since the claims in the population
and in the sample were not overlapping among contractors.
Since assuming responsibility for estimating the national Medicare
error rate in fiscal year 2003, CMS has made changes to the
methodology, which have provided CMS with more detailed
information about the error, thereby allowing the agency to better
identify the underlying causes of error and implement corrective
action plans to address them. For example, CMS significantly
increased the size of the sample used to estimate the Medicare FFS
claims paid in error. The increased sample size allowed the agency
to estimate not only the error rate at the national level, but
also more detailed error rates at the contractor-type and
contractor-specific levels. Further, CMS has made changes in the
way it collects medical records from providers in an effort to
reduce the rate of error caused by nonresponse and insufficient
documentation. These changes may affect the error rate estimates
and thus the comparability of the estimates over time.
Consequently, users of the error rate information should exercise
caution when making year-to-year comparisons.
Our work focused on the methodology CMS used to estimate the
national Medicare error rate and contractor-type error rates for
fiscal year 2004. For these error rates, we found the methodology
adequate for that year. Under CMS's contracting reform initiative,
there will be fewer individual contractors (carriers, DMERCs, and
FIs). If CMS maintains the same overall sample size, the sample
sizes of the remaining individual contractors would be increased.
Reliability of the contractor-specific error rate estimates is
likely to improve with the larger sample sizes. Until then, the
wide variation in reliability of the contractor-specific error
rate estimates may preclude meaningful comparisons across
individual contractors.
We received written comments from HHS (see app. III.) In
responding to our draft report, HHS noted that we found the CMS
methodology adequate for estimating the fiscal year 2004 national
Medicare FFS error rate. HHS also noted that CMS is continually
committed to refining the processes to estimate, as well as lower,
the level of improper payments in the Medicare FFS program.
In its comments, HHS noted improvement in the national Medicare
error rate from fiscal years 2004 to 2005. The department
attributed the decline in the error rate to marked improvement in
the nonresponse (which CMS now calls "no documentation") and the
insufficient documentation error rates. Commenting on the adequacy
of the fiscal year 2005 methodology was beyond the scope of our
work; however, as we noted in the draft report, changes in the
methodology may affect the estimation of the error rates and thus
the comparability of these error rates over time. For example, we
discussed in the draft report that CMS has made changes in the way
it collects medical records from providers in an effort to reduce
the rate of error caused by nonresponse and insufficient
documentation. These changes primarily affected HHS's processes
for calculating an annual error rate estimate for the Medicare FFS
program. This may represent a refinement in the program's
estimation methodology rather than improved accountability over
program dollars.
The national Medicare error rates for fiscal years 2004 and 2005
provided by HHS in its comments are not comparable to the error
rates cited in this report for fiscal years 2004 and 2005. HHS
provided gross error rates, which were calculated using gross
dollars paid in error. Gross dollars paid in error were calculated
by adding dollars paid in error that were due to underpayments to
those that were due to overpayments. As noted in the draft report,
we reported net error rates. Net error rates were calculated using
net dollars paid in error. Net dollars paid in error were
calculated by subtracting dollars paid in error that were due to
underpayments from those that were due to overpayments.
HHS also provided technical comments, which we have addressed as
appropriate.
We are sending copies of this report to the Secretary of Health
and Human Services, the HHS Inspector General, the Administrator
of CMS, and appropriate congressional committees. We will also
provide copies to others upon request. In addition, the report is
available at no charge on the GAO Web site at http://www.gao.gov .
If you or your staff have any questions about this report, please
contact me at (202) 512-7101 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
IV.
A. Bruce Steinwald Director, Health Care
We reviewed the following components of the Centers for Medicare &
Medicaid Services's (CMS) methodology for estimating the fiscal
year 2004 error rate:
o Sampling methods, including sample size, sample selection,
sample representation, and precision of the estimates.
o The medical records collection process.
o Identification and categorization of claims payment error,
including the medical record review process and quality assurance
reviews.
o Statistical methods used to estimate the error rates and
precision.
To conduct our analysis of CMS's sampling methods, we reviewed
work performed by the Department of Health and Human Services
(HHS) Office of Inspector General (OIG) contractor that assessed
these methods and CMS documentation for the fiscal year 2004
Medicare error rate. For the Comprehensive Error Rate Testing
(CERT) Program, we reviewed the program manual, which described
the CERT Program sampling methods as well as CMS's Medicare error
rate reports for fiscal years 2003 and 2004.1 For the Hospital
Payment Monitoring Program (HPMP), we reviewed the program manual
and the HPMP computer programming code that generated the sample
to verify that the sample was taken in accordance with the
procedures outlined in the manual. Additionally, we reviewed the
OIG contractor's comparison of the June 2003 sample and a
re-created version of the June 2003 sampling frame, or population,
for the HPMP. It was not possible for the OIG contractor to obtain
the exact June 2003 population file because the file is
continuously updated and previous versions are not retained. We
did not believe it was necessary to compare every month's sample
to the population from which it was drawn because of the large
size of the sample (approximately 40,000 discharge claims) and
population (approximately 11.5 million discharge claims), and the
fact that the sample was drawn in the same manner each month.
To conduct our analysis of CMS's medical record collection and
review processes and identification and categorization of payment
errors, we relied primarily on reports published by OIG. Since
2003, OIG has conducted annual reviews of the CERT Program and the
HPMP as part of its review of work performed for HHS by
contractors. These annual reviews examine whether the CERT Program
and HPMP contractors have appropriate controls in place to ensure
that the medical record reviews and quality assurance reviews were
performed in accordance with established procedures. We reviewed
OIG's annual reviews of the CERT Program and the HPMP for fiscal
year 2004.2 Our analysis of provider nonresponse within the CERT
Program relied on two OIG studies of CMS's actions to reduce
nonresponse implemented for the CERT Program for fiscal year
2004.3 For the HPMP, we also reviewed four intra-Clinical Data
Abstraction Center (CDAC) reports and two inter-CDAC reports,
which were quality assurance reviews intended to assess the
consistency of review decisions both within and across CDACs.
To conduct our analysis of CMS's statistical methods, we reviewed
the OIG contractor's computer programming code, which replicated
CMS's estimation of the error rates for carriers, durable medical
equipment regional carriers (DMERC), and fiscal intermediaries
(FI), as calculated by the CERT Program subcontractor responsible
for statistical analysis of the error rates for fiscal year 2004.
We reviewed CMS's computer programming code, which calculated the
HPMP error rate for quality improvement organizations (QIO). In
conducting these reviews of the computer programming codes for
both the CERT Program and the HPMP, we verified that each code
appropriately implemented a methodology that employed standard
statistical principles and was used appropriately.
To inform all aspects of our study, we interviewed OIG officials
with oversight responsibility for the error rate estimation, OIG
contractor staff who conducted the evaluation of the statistical
methodology, CMS officials with programmatic responsibilities for
the CERT Program and the HPMP, and staff of the CERT Program
subcontractor for statistical analysis.
We performed our work from April 2005 through March 2006 in
accordance with generally accepted government auditing standards.
Source: GAO analysis of CMS data.
Note: This table reflects net paid claims error rates.
aFor carriers, DMERCs, and FIs, sample size was the targeted
number of claims drawn for the fiscal year 2004 error rate
estimates for each contractor. While CMS selected an equal sample
from each of these contractors, the final sample sizes varied
among contractors. Some selected claims were excluded from the
final sample because the claims were missing information, such as
dates of service and provider or patient information. For QIOs,
the targeted sample size was the actual sample size.
bWe calculated total Medicare fee-for-service payments by dividing
the projected improper payment by the paid claims error rate.
According to the CMS fiscal year 2004 error rate report, CMS did
not adjust projected improper payments data to exclude beneficiary
co-payments, deductibles, and reductions to recover previous
overpayments. This means that the improper payment amounts appear
larger than they would otherwise. However, error rates are
unaffected.
cStandard error is a measure of variation around the estimate, in
this case the error rate.
dRelative precision equals the contractor's standard error divided
by the contractor's paid claims error rate.
eCarriers are health insurers and pay claims submitted by
physicians, diagnostic laboratories and facilities, and ambulance
service providers.
fDMERCs are health insurers and pay claims submitted by durable
medical equipment suppliers.
gFor the fiscal year 2004 error rate,TriCenturion, a program
safeguard contractor, was responsible for medical review in one of
the four DMERC regions. Program safeguard contractors are Medicare
contractors that conduct activities to address or prevent improper
payments. As such, it was TriCenturion, not the DMERC, which was
responsible for lowering the error rates in its region.
hFIs are almost exclusively health insurers and pay claims
submitted by home health agencies, non-prospective payment system
(PPS) hospitals, hospital outpatient departments, skilled nursing
facilities, and hospices. PPS is a reimbursement method used by
Medicare where the payment is made based on a predetermined rate
and is unaffected by the provider's actual costs.
iQIOs (formally known as peer review organizations) are
responsible for ascertaining the accuracy of coding and payment of
paid Medicare FFS claims for acute care inpatient hospital
stays-generally those that are covered by PPS-for Medicare
beneficiaries in all 50 states, the District of Columbia, and
Puerto Rico. Unlike carriers, DMERCs, and FIs, however, QIOs do
not process and pay claims. These activities are conducted by FIs.
A. Bruce Steinwald, (202) 512-7101 or [email protected]
In addition to the contact named above, Debra Draper, Assistant
Director; Lori Achman; Jennie Apter; Dae Park; and Ann Tynan made
key contributions to this report.
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.
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."
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
Contact:
Web site: www.gao.gov/fraudnet/fraudnet.htm E-mail:
[email protected] Automated answering system: (800) 424-5454 or
(202) 512-7470
Gloria Jarmon, Managing Director, [email protected] (202) 512-4400
U.S. Government Accountability Office, 441 G Street NW, Room 7125
Washington, D.C. 20548
Paul Anderson, Managing Director, [email protected] (202)
512-4800 U.S. Government Accountability Office, 441 G Street NW,
Room 7149 Washington, D.C. 20548
Net errors as a percentage of total
dollar amount sampled (fiscal year 2004)
Category of error
Insufficient documentation 4.1
Nonresponse 2.8
Medically unnecessary 1.6
Incorrect coding 0.7
Other 0.2
National Medicare FFS error rate 9.3
26See Department of Health and Human Services, Centers for Medicare &
Medicaid Services, Improper Medicare Fee-for-Service Payments Report
Fiscal Year 2004.
27CMS entered into multiyear contracts with QIOs divided into three
groups. Each of the three groups had different contract end dates.
28See GAO, Medicare Contracting Reform: CMS's Plan Has Gaps and Its
Anticipated Savings Are Uncertain, GAO-05-873 (Washington, D.C.: Aug. 17,
2005).
29Under the current contracting structure, Medicare Part A and Part B
claims are paid by different types of contractors. Part A covers inpatient
hospital care, skilled nursing facility care, some home health care
services, and hospice care, which are paid by FIs. Part B services include
physician and outpatient hospital services, diagnostic tests, mental
health services, outpatient physical and occupational therapy, ambulance
services, some home health services, and medical equipment and supplies,
which are paid by carriers and DMERCs. Under the reformed structure, MACs
will be responsible for both Part A and B claims.
30See Department of Health and Human Services, Medicare Contracting
Reform: A Blueprint for a Better Medicare (Washington, D.C.: Feb. 7,
2005).
CMS Methodology Adequate for Estimating the Error Rates in the CERT Program
Sampling Methods
31While CMS selected an equal sample from each contractor, the final
sample sizes among contractors varied. Some selected claims were excluded
from the final sample because the claims were missing information, such as
dates of service and provider or patient information.
32Relative precision of no greater than 15 percent is considered to be
within acceptable statistical standards. See, for example, Hansen,
Hurwitz, and Madow, 130.
33Systematic sampling is a selection procedure by which the sample is
selected from the population (Medicare claims) on the basis of a uniform
interval, such as every fifth claim, between sampling units (claims),
after a random starting point has been determined. The uniform interval is
determined by dividing the given sample size into the population size and
dropping decimals in the result. The random start is determined by using
an acceptable method of selecting random numbers and is a number between 1
and the uniform interval.
34Of the 30 contractor-specific error rates with relative precision above
acceptable statistical standards, 25 were FIs.
Medical Record Collection Process
35See, for example, W.G. Cochran, Sampling Techniques, 3rd Ed. (New York,
N.Y.: John Wiley & Sons, 1977), 96-99.
36See Department of Health and Human Services, Office of Inspector
General, Review of Corrective Actions to Improve the Comprehensive Error
Rate Testing Process for Obtaining Medical Records, A-03-04-00005
(Washington, D.C.: June 2004). See also Department of Health and Human
Services, Office of Inspector General, Review of Providers' Responsiveness
to Requests for Medical Records Under the Comprehensive Error Rate Testing
Program, A-01-04-00517 (Washington, D.C.: September 2004).
37See Department of Health and Human Services, Centers for Medicare &
Medicaid Services, Improper Medicare Fee-for-Service Payments Fiscal Year
2003 (Baltimore, Md.: December 2003).
3845 C.F.R. Parts 160 and 164 (2005).
39Claims greater than $40 were referred to OIG for follow-up.
Identification and Categorization of Payment Errors
40In fiscal year 2003, 54.7 percent of the national Medicare error rate
was due to nonresponse. In fiscal year 2004, nonresponse decreased to 29.7
percent of the national Medicare error rate.
41According to CMS's fiscal year 2005 error rate report, the CERT Program
reduced error caused by nonresponse in fiscal year 2005 through several
corrective actions, including educating providers about the CERT Program
and encouraging providers to submit medical records by fax. Unlike fiscal
year 2004, for which CMS reported net error rates, in fiscal year 2005,
CMS reported only gross error rates and gross dollars paid in error;
therefore we can only compare gross figures for nonresponse for fiscal
years 2004 and 2005. As a percentage of the total gross Medicare error
rate, nonresponse decreased from 30.7 percent in fiscal year 2004 to 13.5
percent in fiscal year 2005. See Department of Health and Human Services,
Centers for Medicare & Medicaid Services, Improper Medicare FFS Payments
Long Report (Web Version) for November 2005.
42See Department of Health and Human Services, Office of Inspector
General, Oversight and Evaluation of the Fiscal Year 2004 Comprehensive
Error Rate Testing Program, A-03-04-00007 (Washington, D.C.: November
2004).
43See Department of Health and Human Services, Office of Inspector
General, Oversight and Evaluation of the Fiscal Year 2005 Comprehensive
Error Rate Testing Program, A-03-05-00006 (Washington, D.C.: November
2005).
Statistical Methods
CMS Methodology Adequate for Estimating the Error Rate in the HPMP
Sampling Methods
44See, for example, Cochran, 164-166.
45The Lewin Group used the Taylor series approximation method to calculate
the standard errors. See, for example, Cochran, 319.
46A CMS official told us and provided documentation that beginning with
the fiscal year 2006 error rate estimation, the HPMP will move to a simple
random sample in which all records are chosen at random within each state,
thus eliminating the need for systematic sampling. Simple random sampling
is also an accepted method of sampling to achieve a sample that is
representative of the population from which it was drawn.
47It was not possible for the OIG contractor to obtain the exact June 2003
population file because the file is continuously updated and previous
versions are not retained. We did not believe it was necessary to compare
every month's sample to the population from which it was drawn because the
large size of the annual sample (approximately 40,000 claims) and
population (approximately 11.5 million claims) would make the task too
burdensome, and the fact that the sample was drawn in the same manner each
month meant the results from one month should not differ significantly
from the results from any other month.
48The range reported here does not reflect the claims or total payment
volume in Alaska since CMS takes a smaller sample from Alaska than from
all other states, the District of Columbia, and Puerto Rico.
49See, for example, Cochran, 96-99.
Medical Record Collection Process
50A ratio of 100 percent would mean that the average claim amount in the
sample was equal to the average claim amount in the population.
51See Department of Health and Human Services, Office of Inspector
General, Oversight and Evaluation of the Fiscal Year 2004 Hospital Payment
Monitoring Program, A-03-04-00008 (Washington, D.C.: November 2004).
Identification and Categorization of Payment Errors
52For example, according to our analysis of data provided by the HPMP, the
average claim value for claims reviewed for the fiscal year 2004 error
rate for QIOs was approximately $7,500. According to our analysis of
Medicare claims data from the Part B Extract Summary System Carrier Data
File, the average claim value for carriers in 2003 was $32.
53See Department of Health and Human Services, Centers for Medicare &
Medicaid Services, Improper Medicare FFS Payments Long Report (Web
Version) for November 2005.
54See Department of Health and Human Services, Office of Inspector
General, Oversight and Evaluation of the Fiscal Year 2004 Hospital Payment
Monitoring Program.
55According to OIG, CMS requires that CDACs employ admission necessity
reviewers who are licensed practical nurses with utilization review
experience. CMS requires that coding specialists be registered health
information administrators, registered health information technicians, or
certified coding specialists.
Statistical Methods
56See Department of Health and Human Services, Office of the Inspector
General, Oversight and Evaluation of the Fiscal Year 2004 Hospital Payment
Monitoring Program.
57FIs, which are responsible for paying acute-care inpatient hospital
claims, use a software program available on the CMS Web site, PRICER, to
calculate the Medicare payment amount. The program calculates the Medicare
payment amount using information supplied on the provider claim and
current national and hospital-specific factors related to the payment
amount. CMS stated that the PRICER program does not consider all of the
factors used by FIs when pricing acute-care inpatient hospital claims.
CMS Methodology Adequate for Estimating the National Error Rate
58To better ensure that data from a sample represent data from the
population from which they are drawn, the sample data are often adjusted
to reflect the probability of a specific data point, in this case an
acute-care inpatient hospital discharge claim, being chosen. This process
is called weighting. Sample weights reflect the different probabilities
that each claim has of being chosen as part of the sample. The less likely
a claim is to be selected, the larger its sample weight.
59See, for example, Hansen, Hurwitz, and Madow, 172-173.
60CMS used a Taylor series linear approximation method.
61To estimate the total annual dollars paid in error for QIOs, CMS
projects the dollar amounts found in error in the sample to the broad
population.
Concluding Observations
62Statistical theory demonstrates that combining the estimates based on
independent samples is a valid estimate of the aggregate of the samples.
See, for example, Hansen, Hurwitz, and Madow, 190.
Agency Comments
Appendix I: Scope and Methodology Appendix I: Scope and Methodology
1See Department of Health and Human Services, Centers for Medicare &
Medicaid Services, Improper Medicare Fee-for-Service Payments Fiscal Year
2003 (Baltimore, Md.: December 2003). See also Department of Health and
Human Services, Centers for Medicare & Medicaid Services, Improper
Medicare Fee-for-Service Payments Report Fiscal Year 2004 (Baltimore, Md.:
December 2004).
2See Department of Health and Human Services, Office of Inspector General,
Oversight and Evaluation of the Fiscal Year 2004 Comprehensive Error Rate
Testing Program, A-03-04-00007 (Washington, D.C.: November 2004). See also
Department of Health and Human Services, Office of Inspector General,
Oversight and Evaluation of the Fiscal Year 2004 Hospital Payment
Monitoring Program, A-03-04-00008 (Washington, D.C.: November 2004).
3See Department of Health and Human Services, Office of Inspector General,
Review of Corrective Actions to Improve the Comprehensive Error Rate
Testing Process for Obtaining Medical Records, A-03-04-00005 (Washington,
D.C.: June 2004). See also Department of Health and Human Services, Office
of Inspector General, Review of Providers' Responsiveness to Requests for
Medical Records Under Comprehensive Error Rate Testing Program,
A-01-04-00517 (Washington, D.C.: September 2004).
Appendix II: Fiscal Year 2004 Error Rate Information by Contractor
Type-Carriers, DMERCs, FIs, and QIOs Appendix II: Fiscal Year 2004 Error
Rate Information by Contractor Type-Carriers, DMERCs, FIs, and QIOs
Total Medicare
fee-for-service CMS CMS
CMS payments in estimated estimated
targeted fiscal year paid claims standard
sample 2004b(in error rate errorc Relative
Contractor sizea dollars) (percentage) (percentage) precisiond(percentage)
QIO by statei
Carriere
Triple S, Inc. 2,004 $689,224,693 17.9 1.6 8.9
PR/VI
BCBS AR NM/OK/LA 2,004 2,293,083,008 12.7 1.2 9.4
NHIC CA 2,004 6,837,462,204 10.8 1.1 10.2
NHIC MA/ME/NH/VT 2,004 3,323,197,031 9.6 1.0 10.4
TrailBlazer TX 2,004 5,169,066,589 14.1 1.5 10.6
BCBS RI 2,004 232,458,933 13.5 1.5 11.1
GHI NY 2,004 372,383,958 14.3 1.6 11.2
Palmetto GBA 2,004 4,226,979,481 10.6 1.2 11.3
OH/WV
First Coast 2,004 7,367,509,907 9.7 1.1 11.3
Service Options
FL
BCBS UT 2,004 287,713,078 10.2 1.2 11.8
First Coast 2,004 1,106,082,763 7.6 0.9 11.8
Service Options
CT
TrailBlazer 2,004 4,158,091,772 9.2 1.1 12.0
MD/DC/DE/VA
BCBS AR AR/MO 2,004 2,292,786,396 10.6 1.4 13.2
HGSA PA 2,004 3,606,318,041 9.7 1.3 13.4
WPS WI/IL/MI/MN 2,004 8,126,245,486 11.1 1.6 14.4
Cahaba GBA 2,004 3,868,072,306 11.1 1.6 14.4
AL/GA/MS
BCBS KS 2,004 1,581,255,014 6.9 1.0 14.5
KS/NE/Kansas City
Palmetto SC 2,004 1,189,260,267 13.1 1.9 14.5
Noridian 2,004 1,865,892,800 9.5 1.4 14.7
CO/ND/SD/WY/IA
Empire NY/NJ 2,004 7,268,107,083 10.8 1.6 14.8
Noridian 2,004 4,981,083,701 10.7 1.6 15.0
AZ/HI/NV/AK/OR/WA
AdminaStar IN/KY 2,004 2,708,331,380 10.0 1.5 15.0
HealthNow NY 2,004 1,358,023,183 8.2 1.3 15.9
CIGNA ID/TN/NC 2,004 4,830,134,495 10.9 1.8 16.5
BCBS MT 2,004 193,432,019 5.3 0.9 17.0
All carriers 50,100 $79,932,195,591 10.7 0.4 3.7
DMERCf
TriCenturion 2,004 $1,364,899,356 7.3 0.9 12.3
Region Ag
AdminaStar 2,004 2,241,150,409 6.6 0.9 13.6
Federal-Region B
CIGNA-Region D 2,004 1,800,134,845 11.6 2.1 18.1
Palmetto GBA 2,004 4,928,003,571 14.0 2.9 20.7
Region C
All DMERCs 8,016 $10,334,188,182 11.1 1.5 13.5
FIh
UGS 2,004 $6,003,110,480 20.4 2.1 10.3
CA/HI/AS/GU/NMI
Palmetto GBA SC 2,004 6,194,956,951 10.3 1.1 10.7
Mutual of Omaha 2,004 11,797,457,474 26.8 3.2 11.9
First Coast 2,004 2,472,517,626 23.0 2.9 12.6
Service Options
FL
TrailBlazer 2,004 4,556,783,468 14.1 2.0 14.2
TX/CO/NM
Cahaba GBA AL 2,004 705,028,658 15.5 2.2 14.2
Trispan MS/LA/MO 2,004 1,675,273,646 15.8 2.6 16.5
BCBS RI 2,004 781,806,244 19.3 3.2 16.6
Empire NY/CT/DE 2,004 5,811,286,709 17.2 2.9 16.9
UGS VA/WV 2,004 1,449,840,434 16.6 2.8 16.9
COSVI PR/VI 2,004 158,822,429 11.9 2.1 17.6
Medicare 2,004 711,126,486 14.6 2.6 17.8
Northwest
OR/ID/UT
Palmetto GBA NC 2,004 3,190,067,317 16.7 3.0 18.0
Veritus PA 2,004 2,079,132,007 14.7 2.7 18.4
UGS MI/WI 2,004 4,952,538,415 13.5 2.5 18.5
Anthem NH/VT 2,004 641,811,111 9.0 1.7 18.9
BCBS WY 2,004 83,003,027 14.7 2.8 19.0
BCBS AZ 2,004 325,070,959 7.3 1.4 19.2
CareFirst MD/DC 2,004 2,159,553,514 25.3 4.9 19.4
Cahaba GBA IA 2,004 4,273,518,964 5.6 1.1 19.6
Noridian MN/ND 2,004 1,309,949,370 16.2 3.3 20.4
BCBS AR 2,004 481,442,284 26.1 5.5 21.1
BCBS NE 2,004 299,081,984 12.8 2.7 21.1
Anthem MA/ME 2,004 2,852,313,346 10.4 2.2 21.2
BCBS GA 2,004 2,105,558,870 6.9 1.5 21.7
BCBS KS 2,004 512,584,700 10.0 2.2 22.0
AdminaStar 2,004 9,610,571,631 12.2 2.7 22.1
IN/IL/KY/OH
BCBS MT 2,004 229,695,544 6.8 1.7 25.0
BCBS OK 2,004 1,109,256,221 8.6 2.2 25.6
Riverbend TN/NJ 2,004 3,622,031,691 9.7 3.0 30.9
Premera WA/AK 2,004 1,004,968,329 7.3 3.1 42.5
All FIs 62,124 $83,160,159,889 15.7 0.7 4.5
Massachusetts 744 $2,135,744,081 8.6 0.90 10.5
Kentucky 744 1,482,350,516 9.3 1.10 11.8
New Mexico 744 324,592,033 6.1 0.90 14.8
Maine 744 417,801,848 4.6 0.70 15.2
Louisiana 744 1,388,303,707 5.8 0.90 15.5
Arkansas 744 851,144,822 4.5 0.70 15.6
Illinois 744 3,864,432,432 4.4 0.70 15.9
Delaware 744 271,799,810 4.2 0.70 16.7
Maryland 744 2,067,187,033 3.0 0.50 16.7
Iowa 744 812,196,278 3.6 0.60 16.7
Indiana 744 1,784,654,000 4.1 0.70 17.1
Nevada 744 402,837,978 4.6 0.80 17.4
New Hampshire 744 328,223,324 3.4 0.60 17.6
Florida 744 5,696,783,961 5.1 0.90 17.6
Michigan 744 3,467,564,282 3.9 0.70 17.9
West Virginia 744 729,042,409 4.4 0.80 18.2
Vermont 744 164,700,697 3.3 0.60 18.2
South Dakota 744 232,787,316 3.8 0.70 18.4
Ohio 744 3,469,584,344 3.2 0.60 18.8
Alabama 744 1,603,881,531 3.2 0.60 18.8
Rhode Island 744 269,904,786 4.2 0.80 19.0
Virginia 744 1,933,408,829 3.5 0.70 20.0
Oklahoma 744 964,748,057 3.5 0.70 20.0
Alaska 504 100,985,029 3.5 0.70 20.0
North Dakota 744 216,246,500 2.0 0.40 20.0
South Carolina 744 1,408,487,704 5.4 1.10 20.4
New Jersey 744 3,595,399,138 2.9 0.60 20.7
Puerto Rico 744 376,450,167 4.8 1.00 20.8
Utah 744 389,527,711 3.8 0.80 21.1
Connecticut 744 1,295,269,906 3.2 0.70 21.9
New York 744 6,522,717,692 2.6 0.60 23.1
Idaho 744 237,198,385 2.6 0.60 23.1
Texas 744 5,573,613,357 4.2 1.00 23.8
North Carolina 744 2,720,223,476 2.1 0.50 23.8
Washington 744 1,253,681,476 2.1 0.50 23.8
Oregon 744 689,865,040 2.5 0.60 24.0
Pennsylvania 744 4,290,842,680 2.5 0.70 28.0
Nebraska 744 500,351,357 1.4 0.40 28.6
District of 744 393,305,231 1.3 0.40 30.8
Columbia
Kansas 744 762,382,857 2.8 0.90 32.1
Georgia 744 2,215,263,714 2.1 0.70 33.3
Arizona 744 1,081,388,500 2.4 0.80 33.3
Tennessee 744 2,093,513,706 1.7 0.60 35.3
Missouri 744 1,935,671,182 1.1 0.40 36.4
Wyoming 744 99,863,364 1.1 0.40 36.4
California 744 7,517,783,935 4.6 1.70 37.0
Wisconsin 744 1,575,519,000 1.0 0.40 40.0
Minnesota 744 1,412,860,400 1.0 0.50 50.0
Colorado 744 703,166,846 1.3 0.70 53.8
Hawaii 744 203,010,800 0.5 0.30 60.0
Montana 744 226,885,429 0.7 0.50 71.4
Mississippi 744 884,792,083 1.2 1.00 83.3
All QIOs 38,448 $84,939,940,736 3.6 0.20 5.6
Appendix III: Comments from the Department of Health and Human Services
Appendix III: Comments from the Department of Health and Human Services
Appendix IV: GAO Contact and Staff Acknowledgments
GAO Contact
Acknowledgments
(290355)
GAO's Mission
Obtaining Copies of GAO Reports and Testimony
Order by Mail or Phone
To Report Fraud, Waste, and Abuse in Federal Programs
Congressional Relations
Public Affairs
www.gao.gov/cgi-bin/getrpt? GAO-06-300 .
To view the full product, including the scope
and methodology, click on the link above.
For more information, contact A. Bruce Steinwald at (202) 512-7101 or
[email protected].
Highlights of GAO-06-300 , a report to congressional committees
March 2006
MEDICARE PAYMENT
CMS Methodology Adequate to Estimate National Error Rate
The Centers for Medicare & Medicaid Services (CMS) estimated that the
Medicare program paid approximately $20 billion (net) in error for
fee-for-service (FFS) claims in fiscal year 2004. CMS established two
programs-the Comprehensive Error Rate Testing (CERT) Program and the
Hospital Payment Monitoring Program (HPMP)-to measure the accuracy of
claims paid.
The Medicare Prescription Drug, Improvement, and Modernization Act of 2003
directed GAO to study the adequacy of the methodology that CMS used to
estimate the Medicare FFS claims paid in error. GAO reviewed the extent to
which CMS's methodology for estimating the fiscal year 2004 error rates
was adequate by contractor type for (1) the CERT Program, (2) the HPMP,
and (3) the combined national error rate (including both the CERT Program
and the HPMP).
GAO reviewed relevant CMS documents and reports related to the CERT
Program and the HPMP. In addition, GAO reviewed work performed by the
Department of Health and Human Services (HHS) Office of Inspector General
(OIG) and its contractor that evaluated CMS's fiscal year 2004 statistical
methods and other aspects of the error rate estimation process. GAO also
conducted interviews with officials from CMS, HHS's OIG, and their
contractors.
The methodology used by CMS for the CERT Program was adequate to estimate
the fiscal year 2004 error rates by contractor type-carrier, durable
medical equipment regional carrier (DMERC), and fiscal intermediary (FI).
Carriers pay claims submitted by physicians, diagnostic laboratories and
facilities, and ambulance service providers. DMERCs pay claims submitted
by durable medical equipment suppliers. FIs pay claims submitted by
hospitals, home health agencies, hospital outpatient departments, skilled
nursing facilities, and hospices. The methodology was adequate because CMS
used a large sample-about 120,000 claims-and an appropriate sample
selection strategy. For these fiscal year 2004 error rate estimates, CMS
made improvements in the collection of medical records that supported the
sampled claims. These medical records were appropriately reviewed to
determine whether there were errors in payment. CMS used valid statistical
methods to estimate the fiscal year 2004 error rates for the carrier,
DMERC, and FI contractor types.
The methodology used by CMS for the HPMP was adequate to estimate the
fiscal year 2004 error rate by quality improvement organizations (QIO),
which are responsible for ascertaining the accuracy of coding and payment
of Medicare FFS paid claims for acute care inpatient hospital stays. CMS's
sampling methods were adequate because the agency used a large sample,
approximately 40,000 claims, that was representative of the population
from which it was drawn in terms of average dollar amount per claim. Also,
the HPMP had adequate processes in place to ensure appropriate
determinations of error. CMS used valid statistical methods to estimate
the fiscal year 2004 error rate for the QIO contractor type.
The fiscal year 2004 contractor-type error rate estimates for the CERT
Program and the HPMP were appropriately combined to determine the national
Medicare error rate through the use of a valid statistical method. CMS
estimated the national Medicare error rate by averaging the carrier,
DMERC, and FI contractor-type error rates in the CERT Program and the QIO
contractor-type error rate in the HPMP, weighted by each contractor type's
share of total Medicare FFS payments.
In written comments, HHS noted that GAO found CMS's methodology adequate
for estimating the fiscal year 2004 national Medicare FFS error rate. HHS
also noted that CMS is continually committed to refining the processes to
estimate, as well as lower, the level of improper payments in the Medicare
FFS program.
Medicare Net FFS Error Rates and Dollars of Claims Paid in Error, Fiscal
Year 2004
Error rate Dollars paid inerror (in
CMS program Contractor type (percentage) billions)
CERT Program Carrier 10.7 $6.5
DMERC 11.1 1.0
FI 15.8 9.3
HPMP QIO 3.6 3.1
National Medicare All contractor 9.3 $19.9
FFS error rate types
Source: CMS.
*** End of document. ***