I am Tom Grissom, the Director of the Center for Medicare Management at the Centers for Medicare and Medicaid Services (CMS). I would like to thank the National Committee for Vital and Health Statistics for providing me this opportunity to speak on an issue that is extremely important to CMS as well as every entity concerned with the delivery and financing of health care in America. CMS relies on coded medical data for a number of our payment, quality, research, and other administrative activities.
Medicare, which spends over $100 billion a year for inpatient services, is the country's largest single payer for inpatient hospital care. The Center for Medicare Management is responsible for overseeing all aspects of making these payments. We are also responsible for developing and maintaining the HIPAA standard procedure coding system used for inpatient hospital claims, Volume 3 of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).
Of course, ICD-9-CM is not used solely for inpatient coding purposes. The diagnosis coding aspect of the ICD-9-CM is the HIPAA standard coding system used to identify patients' principal and secondary diagnoses in all clinical settings. While my remarks will concentrate on inpatient procedure coding, similar issues confront both the diagnosis and the procedure parts of ICD-9-CM.
Concerns about ICD-9-CM
I know the limitations of ICD-9-CM have been considered many times by this committee, and we appreciate your thoughtful attention to this very significant issue. In fact, the Committee in 1993 reported that ICD-9-CM was, even at that time, rapidly becoming outdated. The Committee recommended an immediate commitment to developing a migration to ICD-10 for morbidity and mortality coding. The Committee recognized similar problems with the procedure coding aspect of ICD-9-CM and called for creation of a new coding system with particular features and characteristics. In response to this report, CMS – then HCFA -- initiated efforts to develop a replacement system.
The problems associated with ICD-9-CM have grown worse since 1993 because of the increasing pace of changes in medical care and in the development of medical technology. In service for 24 years, ICD-9-CM does not accurately describe the diagnoses and inpatient procedures delivered in the 21st century. Furthermore, it cannot be readily and flexibly expanded to accommodate further changes. In our view, this coding system, in its current incarnation, is quickly approaching obsolescence.
ICD-9-CM, as many major code sets, has a hierarchical structure that groups elements together in order to facilitate its use by coders in the field and to contribute to analysis of claims for payment and quality purposes.
In the ICD-9-CM, procedures relating to specific body systems are first grouped together using the first two digits. Then, similar types of procedures are grouped together using the third digit. The fourth digit is then available to distinguish between at most ten specific procedures.
For example, codes for Operations on the Cardiovascular System are located within the range 35.00 and 39.99. Procedures involving removal of coronary artery obstruction and insertion of stents are assigned to codes 36.01 to 36.09. With the recent introduction of drug-eluting stents, CMS decided to establish separate and higher payments for drug-eluting stents compared to non-drug-eluting stents. As a result, it was necessary to create a new code to identify drug-eluting stents. As a result, only one unused code remains available to identify new developments in this fast-advancing area.
In many instances, the available codes for certain classes of procedures are either all assigned or nearly so. Increasingly, new procedures must be assigned to unused codes in unrelated series, with the result being an unstructured, haphazard coding system. The likely result of this is an increase in coding error, leading to inappropriate payments and inaccurate medical data.
Major coding systems have long used hierarchical structures. In fact, one of the characteristics included in the Committee's 1993 recommendation regarding characteristics of a replacement for ICD-9-CM was that it be a hierarchical system. A hierarchical structure assists in defining coding concepts by placing them into organized, distinct groupings. By starting with a major heading, such as the cardiovascular system, one can then subdivide the concepts into specific areas such as the heart and major vessels. This approach assists the coder in correctly locating a specific code. If the index does not precisely identify a concept, then the coder can examine terms within a general category to find the best fit to describe the concept.
Analysis for setting payment rates or doing review of care patterns also relies on the hierarchical structure. When analysts are trying to identify specific cases for review, they must develop a list of codes. They will then expand or narrow their focus after reviewing codes listed within specific categories. When some codes are not in the appropriate categories because of lack of space, analysts as well as coders may not find the most appropriate codes.
For example, when analysts are unaware that some cardiovascular procedure codes are not in the cardiovascular chapter, they may underreport the number and costs of patients receiving new services. CMS recently was forced to add new cardiovascular procedure codes to a miscellaneous chapter of the ICD-9-CM procedure codebook. If coders and analysts do not find them, it will appear that fewer of these procedures were performed than actually were performed, and third-party payers, regulators, and researchers may base their conclusions on inaccurate results.
Finally, in looking at trend data, it is frequently useful to have a tool that collapses the codes into related categories. Without this structure, a variety of independent approaches to aggregation of related groups could arise, which could lead to apparently inconsistent findings across analyses.
Accurate Payments
The space limitations of ICD-9-CM have direct implications upon Medicare's ability to make accurate payments to hospitals, particularly for providing new technologies. Original Medicare's fee-for-service program pays for inpatient services by grouping cases with similar diagnoses and procedures into diagnosis-related groups (DRGs). We rely upon ICD-9-CM diagnosis and procedure codes on the submitted bill in performing this grouping.
Assignment to a DRG begins with the ICD-9-CM principal diagnosis assigned to a patient. The principal diagnosis is the reason a patient was admitted to a hospital. Up to eight additional ICD-9-CM diagnosis codes and up to six procedure codes may also be included on the bill, and these codes may also affect DRG assignment. Once the DRG is assigned, Medicare uses a weight specific to that DRG to calculate the payment amount.
At present, Medicare uses about 520 DRGs. Every year, we are asked to reexamine the treatment of specific, generally new, procedures and technologies within the DRGs. The best way to insure our payments are reasonable is to identify cases in our data involving the specific technology, which generally requires a specific ICD-9-CM code associated with the technology.
As a result, the ICD-9-CM Coordination and Maintenance Committee, which is responsible for maintaining ICD-9-CM, is frequently asked to establish new procedure code primarily for the purpose of identifying cases involving a specific procedure. Such a code would enable analysis of the costs of those cases to determine whether increased payments are necessary, and if appropriate could lead to restructuring of DRGs to more accurately reflect costs of procedures.
Congress has encouraged Medicare to be more responsive in recognizing the higher costs of new technology, instructing the Secretary to establish a mechanism to pay an amount in addition to the regular DRG payment for technologies that would be inadequately paid otherwise. This provision has created even more interest in establishing ICD-9-CM procedure codes that are specific to individual technologies.
A recent example was the creation of procedure code 00.11 (Infusion of Drotrecogin Alfa (Activated)) in 2002. This code is used to identify the administration of the drug XigrisTM, used to treat patients with severe sepsis. Because CMS determined that this new drug would result in a substantial clinical improvement and that the regular DRG payment would be inadequate otherwise, Medicare makes an additional payment to hospitals for cases involving XigrisTM. This additional payment is triggered by the inclusion of procedure code 00.11 on the submitted bills.
Although we were able to create this new code in order to make additional payments, this instance provides a good example of how ICD-9-CM is not working as it was intended. In 2002 the ICD-9-CM Coordination and Maintenance Committee began assigning new codes to series 00.xx when no space was available in the system to appropriately categorize new procedures within the hierarchical structure.
As the number of ICD-9-CM codes available to assign to new technologies is reduced, trade-offs are inevitable between capturing more information about the use of new technologies and preserving the available codes as long as possible. Assigning new codes for even a fraction of new drug therapies alone, should that be considered useful, would quickly deplete the number of available procedure codes. The longer we delay a decision to replace ICD-9-CM, the more difficult these trade-offs will become. Eventually, we may not be able to separately identify and pay extra for a costly new medical advance because insufficient new codes are available.
Finally, we believe we need to reexamine the entire DRG system to see if payment accuracy can be substantially improved for the future. Because the DRGs are based on general patient characteristics and procedures, every DRG exhibits variation in the cost of treating individual patients assigned to that DRG. Some hospitals continue to argue the current DRG structure, which is based on national averages, does not adequately reflect the sicker mix of patients they treat, and urge that payments be adjusted to reflect the severity of their "case mix." Further, specific types of patients or particular procedures are identified as being so different from the average that it may be appropriate to move them to another DRG or to create a new DRG, steps which tend to reduce the variance of costs of cases within each DRG. While we take both of these steps periodically, the number of DRGs has only increased from 470 to 518 in the twenty-year history of the DRG system.
An alternative, however, would be to reflect varying levels of patient severity and/or resource use within existing DRGs. Collecting more detailed information on patient characteristics and procedure use would facilitate evaluation of whether such refinements to the DRGs are merited. However, at present the rigidity of the ICD-9-CM coding system inhibits our ability to adjust codes for this purpose.
Appropriate Payments
Medicare pays for covered services that are medically necessary. Services may occasionally only be covered or approved when specific medical conditions exist. For instance, a national coverage decision for a new technology could state specific medical conditions that must be present before the technology may be covered. Due to the limited specificity of ICD-9-CM, conditions that support medical necessity for a particular service may be classified to the same code as conditions that would not justify the service. This frequently leads to rejected claims or the need to provide additional documentation.
Medicare national coverage decisions may include limitations on the medical conditions for which a particular service or technology may be covered. New, more specific diagnoses codes are often needed to implement the coverage policy because the existing codes may include other diagnoses not covered by the payment policy. Frequently, available ICD-9-CM diagnosis codes are insufficient to identify the conditions of interest. As a result, it may be necessary to conduct manual reviews of the patients' medical records to ensure the treatment is covered under Medicare and, therefore, payable.
More precise information can also reduce the number of claims rejected as not medically necessary due to insufficient information available on the bill. When a bill is submitted for payment, it may be reviewed to assess whether the service appears to have been medically necessary. Initially, the only information available upon which to make such an assessment is the ICD-9-CM data on the bill. If the Medicare contractor processing the claim needs more information to make a determination that the service was medically necessary, it may be necessary to obtain the medical record associated with the bill.
An example would be a patient who fractures her left wrist and is seen for treatment by an orthopedic surgeon. A month later this same patient falls and breaks her right wrist and is seen by the same physician. There is only one ICD-9-CM code for a wrist fracture, so when the physician submits a bill, this code does not reflect the fact that a new fracture has occurred on the other wrist. The physician receives one payment for the global services of treating a single wrist fracture. When a second bill is submitted a month later for the new service, systems edits would not allow payment. The physician would have to provide additional documentation in the form of a written explanation that a new fracture was involved.
Monitoring Quality Outcomes
CMS has recently begun several initiatives to collect data intended to measure factors affecting quality outcomes. However, our ability to evaluate the outcome of new procedures and emerging health care conditions will be hampered by the lack of precise ICD-9-CM codes. For example, many procedures that were performed as open surgical procedures can now be performed percutaneously or transluminally. For basic quality measures such as mortality, rates can vary widely depending on the approach used during the procedure. A lack of such basic information as the surgical approach makes measures of the nature and quality of care difficult.
Increasingly, we do not have compatible data to compare internationally. As the United States prolongs the use of ICD-9 while many other countries are moving to ICD-10, data incomparability with the rest of the world will become a more significant issue. It may lead to problems identifying and tracking new health threats and in developing quick interventions for emerging diseases. It will also severely limit our ability to compare outcomes of new technologies as used in the US and abroad. As we become a global community, it becomes vital that our health care data represent current medical conditions and technologies and that it is compatible with the international version of ICD-10. Identifying and tracking diseases such as anthrax and SARS will be come increasingly important.
ICD-10-CM and ICD-10-PCS
While the World Health Organization developed ICD-10 as an international diagnosis coding system, the National Center for Health Statistics in the CDC has developed ICD-10-CM to replace the diagnosis section of ICD-9-CM, and CMS has developed the ICD-10-PCS as a possible replacement for the ICD-9-CM procedure codes. We believe these systems are consistent with the criteria recommended in the 1993 report by this committee.
The final rule adopting standards and code sets for electronic claim submission under the Health Insurance Portability and Accountability Act of 1996 designated ICD-9-CM as the standard code set for inpatient services. That rule also alerted providers, payers, and the public to the expectation that the standards would eventually need to be replaced. It was anticipated that the Designated Standard Maintenance Organizations and the NCVHS would carefully evaluate the need for any new standard, and that there would be opportunity for public comment.
It was also recognized that
". . . any major change in administrative coding systems involves significant initial costs and dislocations, as well as some level of discontinuity in data collected before and after the change. These factors must be weighed against expected improvements in the efficiency of data creation and in the accuracy and utility of the data collected." (Federal Register, May 7, 1998, page 25286)
We are currently evaluating the anticipated costs for CMS if we were to adopt ICD-10-CM and ICD-10-PCS. However, that analysis is not yet complete at this time. When those analyses are complete, we will be eager to share them with the Committee.
However, one thing is clear at this time: ICD-9-CM is no longer functioning as it was designed, and we must develop a plan to move forward in a way that provides the data necessary to ensure the viability of our health system and protects the health of our citizens.
Thank you for the opportunity to come before you today to discuss this important issue.