My name is William Marder and I am a Vice President and General Manager with The MEDSTAT Group, Inc. ("MEDSTAT"). I have overall responsibility for our business with federal government research organizations and our commercial research clients. It is a pleasure to be here to discuss medical/clinical coding and classification issues in connection with the recent enactment of the Health Insurance Portability and Accountability Act of 1996 (PL104-191).
Prior to becoming a manager of research activities, I was an active health economist with interests in a variety of health services research topics, especially topics around physician supply. I mention this background because my training as an economist is similar to the type of experience that MEDSTAT brings to this panel. Economists are trained to analyze data sets that typically are designed and implemented by others. Training in economics contains large doses of complaints about data quality and hard work associated with doing the best you can with the data that are at hand.
Early in my career, I became convinced that timely investments of effort in the design and implementation of data collection strategies would pay large dividends, especially when combined with sophisticated approaches to analyzing the resulting data sets. I joined MEDSTAT because the company operates in a way that is broadly consistent with this mixed strategy: try to collect the right data, but be sure to make the most of the data at hand.
MEDSTAT specializes in the strategic application of health information, and offers knowledge-based systems, consulting and research for improving the quality and total value of healthcare. Our company services in excess of 1,000 healthcare purchasers, managed care organizations, policy researchers and providers in the public and private sectors. MEDSTAT has been providing healthcare information products and services to our clients since 1981. With headquarters in Ann Arbor, Michigan, we have over 725 employees in offices across the United States.
MEDSTAT is in the healthcare information business focusing on what could be termed post-encounter decision support. Our customers are trying to take action on important issues but they are not trying to affect patient care as it is being provided. We are well known for developing claims and encounter databases that help purchasers or managed care and insurance companies monitor and evaluate their impacts on covered populations.
I want to focus on four different MEDSTAT projects that form the basis for my comments on coding issues. The four projects are:
· standardizing private health insurance claims and encounter data for self-insuring employers,
· standardizing Medicaid claims data for the Health Care Financing Administration (HCFA),
· standardizing hospital discharge abstracts for the Agency for Health Care Policy and Research (AHCPR), and
· reviewing and reporting HEDIS data from managed care plans.
In each of these projects, MEDSTAT staff encountered significant issues which are relevant to this panel. These issues are concerned with the comparability of data from different sources.
Standardized private insurance claims data. MEDSTAT was founded in 1981 to help large, self-insuring employers understand and better manage the costs and quality of the healthcare utilized by their employees and their employee's dependents. National employers contract with different insurance companies around the country. Are the local managed care companies providing equal access and quality at comparable cost? MEDSTAT builds analytic databases and decision support software combines these products with consulting services to answer this question.
To build our decision support tools we needed to deal with the large, local variation in coding practices. Comparing access, quality, and cost, after all, depends on an understanding of how the healthcare needs of local populations might differ. Our goal was to provide appropriate adjustments to the data before advising our clients about alternative actions. We invested considerable time and energy to understand the coding practices that were acceptable to private insurers around the country. We did this to distinguish between real differences in disease and treatment and variations in coding practices.
One of the first lessons learned from this exercise was that when the national coding system does not closely meet the needs of those paying the bills, there will be a proliferation of local solutions. MEDSTAT training is like the training I described for economists. We learned how to cope with local coding variations and to derive as much value as we could from the imperfectly comparable data that are available.
Standardized Medicaid claims data. The Research and Policy Division of MEDSTAT that I head, was built from SysteMetrics, Inc. a California-based company that provided research database building services to the federal government. One of the long-standing contracts that was started by SysteMetrics and that was only recently completed was known as the Tape-to-Tape project. Tape-to-Tape was funded by the HCFA to provide comparable data from a limited number of state Medicaid programs.
We encountered in the Tape-to-Tape project the same issues described above for private insurance. Even in the context of federal-state partnership to administer a single program, we incurred significant costs imposing the standardization that was needed by HCFA-funded researchers.
As part of Tape-to-Tape, we learned all about the difficulties of mapping ICD-9-CM procedure codes into comparable CPT codes. We grappled with crosswalks among HCPCS, UB82 and UB92 revenue codes and other difficult problems. In general, we learned that no one system will meet user needs and that users will adapt any system in order to meet the perceived requirements of their situation.
The costs associated with the Tape-to-Tape project are a matter of public record. This historical record can provide a beginning estimate of the costs associated with standardized claims data systems. HCFA's current experience with the successor system to Tape-to-Tape, the Medicaid Statistical Information System, can provide more recent data about the challenges of standardized administrative data systems.
Within the past two years we have continued to provide technical assistance to HCFA and its regional offices as they grapple with the encounter data systems that Medicaid managed care plans are asked to develop. Our extensive work in this area depends on our ability to work with any classification system in use within the healthcare industry. As the move into managed care continues, there will be a growing number of information systems that use medical/clinical coding systems. Introducing change into this pluralistic system will be challenging and inevitably costly.
Standardized Hospital Discharge Abstracts. We have worked under contract to the Agency for Health Care Policy and Research and its predecessors to construct the Healthcare Cost and Utilization Project (HCUP) database. HCUP is a research-ready file of UB92 hospital discharge abstracts from a representative sample of acute care hospitals around the country. ICD-9-CM coding seems to fit this inpatient discharge data fairly well. The level of abstraction provides reasonable diagnostic and procedural detail without swamping the analytic files with minor differences that could either represent differences in care or differences in coding practice. Even in the hospital environment with significant coding expertise and a well-adapted system, we incur substantial costs standardizing data from different hospitals.
When HCUP expanded to include ambulatory surgery data, we encountered on this project the same problem that could be noted above for claims data. ICD-9-CM and CPT contain different levels of detail for surgical procedures. There is no one-to-one correspondence between the two systems. If the institutions doing the coding need the detail for reimbursement purposes, then CPT will be used, and used correctly. If no reimbursement depends on it, the detail of CPT is not likely to be consistently implemented.
In the context of the HCUP project, we developed the Disease Staging approach to grouping diseases and examining complications. Disease Staging abstracted further from the ICD-9-CM view of diagnostic information to provide a more manageable number of categories and a hierarchy of disease severity. We continue to use Disease Staging as a core method that we apply to data from many sources to assist in the process of fairly comparing providers and plans.
HEDIS data audits. MEDSTAT has been auditing self-reported HEDIS data since the National Committee on Quality Assurance (NCQA) implemented the program. Even with the relatively straightforward indicators included in the HEDIS data sets, we have found that there is enough ambiguity in administrative systems that self-reported data do not provide a reliable basis to compare managed care plans.
This year, for the third year in a row, we are providing technical support to the California Healthcare Reporting Initiative (CCHRI). CCHRI is a coalition of healthcare purchasers that provides a comparison of the major managed care plans in the state. As a result, we develop samples of medical records to investigate the status of each HEDIS indicator. We continue to use medical records abstraction as the gold standard for HEDIS reporting. The diversity of information systems in place (and the difficulties encountered when plans use their own administrative systems for quality measurement) lead CCHRI and others to continue this relatively costly procedure.
In a variety of settings we have found that the information systems of managed care plans and providers can provide useful information. As a provider of software, we have developed techniques that compute the HEDIS indicators from the administrative data systems. This software coupled with samples of medical records to validate accurate reporting has considerable promise for the future.
1. What medical/clinical codes and classifications do we use?
We have the capacity to work with any classification system in use by the healthcare industry. We handle data that has been coded not only with the major classification systems - ICD, CPT, DSM-IV, HCPCS, and UB-92 Revenue codes - but also with user-defined systems such as state-specific codes for Medicaid (HCPCS Level III) and SNOMED. We have developed many crosswalks in order to analyze data that comes from disparate sources. We also work with secondary classification systems, such as DRGs and APGs.
In our experience, what appears to be a strength of a classification system for one application may be a weakness for another application. For example, ICD-9-CM procedure codes are appropriate to identify most categories of surgeries and major procedures performed, but do not do as well classifying ambulatory interventions such as lab, x-ray, and office visits - a strength of the CPT system. Yet the CPT system may be overly complex and too detailed for the level of use in the hospital setting, where the goal may be only to get the proper DRG reimbursement or to maintain the required surgical indices.
2. What medical/clinical codes and classifications do we recommend for initial standards for administrative transactions?
For diagnostic reporting, the ICD system has certain advantages, especially with regard to public health reporting at the international level. This system, developed and maintained through the World Health Organization, is well accepted and understood by practitioners. Even with the modifications made for use in the United States, we can translate back to the basic WHO classification system to allow epidemiological comparisons between the U.S. and other countries. There is also a mechanism in place for the training and education of users of the ICD system. The well-established maintenance structure through the Cooperating Parties and the ICD-9-CM Coordination and Maintenance Committee should not be disregarded. For procedure coding, the differences between ICD-9-CM and CPT are well known. ICD-10-PCS may offer a suitable replacement for both systems, with its increased rigor and specificity.
Even more important than the classification system(s) chosen for administrative transactions is the imperative that ALL who submit data to any system for any purpose should use the system as intended. Codes should all be reported to their most specific level. Transactional systems should allow enough data fields, for both diagnosis and procedure codes, so that each record can include all of the codes that fully describe each patient's condition and treatment rendered. Standards for inclusion or exclusion of codes should be rigorously defined.
Totality of coding to include representation of patient risk factors should be encouraged so that electronic data rather than expensive medical record review can be used to adjust for outcomes of care.
With the move toward computer based patient records, the codification and standardization of terminology becomes even more important.
3. If ICD is chosen for administrative transactions, should it be ICD-9-CM, or ICD-10 and ICD-10-PCS, assuming field evaluations are generally positive?
Regardless of the field evaluations, there should also be a careful evaluation or cost-benefit analysis of the shift. We have heard that some countries that have implemented ICD-10 have done so with much difficulty. There are probably more major implications in the US than in other countries. There are many more and different players in our health care industry - providers, managed care organizations, federal and state governments, vendors, transactional and fiscal intermediaries, data brokers, educators, and researchers. What is the impact on the economy of the expenditures required to switch to a new system? Can we cope using the existing systems, with one standardized and sanctioned translation for the international requirements? A thorough evaluation would be valuable.
4. How should we deal with current coding environment to improve simplification, reduce administrative burden, and obtain medically meaningful information?
The key to simplification would be to have more standardized implementation practices. This would obviate the need for expansive and expensive manipulation of data. If all payors used the same system and rules, there would be less rejection of claims, less paperwork involved with those claims, and less confusion by staff supplying codes for claims purposes. However, it needs to be determined if more rigor would exacerbate the underreporting that currently is apparent in data, especially for diagnosis codes in the ambulatory setting. It is not enough to just submit a code that will pass the edits and get the claim paid. Data are used for many other purposes and need to be accurate and complete, to produce a total picture of the patient. Look at the move toward using administrative data to determine rates of compliance with medical practice guidelines, such as for mammography screening. The HEDIS measures now allow women with bilateral radical mastectomies to be excluded from the population at risk. How can this be measured if the coding is incomplete, or if there is no standardized longitudinal patient-level database?
5. How should ongoing maintenance (responsibility, intellectual input, funding) be addressed in the standards?
The current system of the ICD-9-CM Coordination and Maintenance Committee works well for intellectual input, with annual updates published in the Federal Register. Changes more on a less than annual basis would be difficult to monitor, track, and update in the myriad of systems available, and would impose a financial burden on users of these systems. For the CPT system, if it is retained as a reporting system for physician billing, there should be a more structured mechanism for updates, similar to the ICD system.
4. What issues do we encounter linking data coded with different classification systems?
On both the HCUP and the Tape-to-Tape projects we encountered mismatches in nomenclature that are difficult to resolve; result is wider groupings that include widely variant entities, such as procedures, where the coding systems differ on approach, excision vs biopsy, etc.
5. What is the impact of the emerging computerized patient record?
Nomenclature of any system implemented in the CPR needs to be much more structured than any of the systems currently in widespread use. Most of them have sections by body system that were developed for the specialties to which they pertain.
8. Is this practical to move to a single procedure classification system on the timetable required for initial implementation?
The timetable seems quite aggressive for any major change.
9. If a system is selected as a standard, should providers be able to use all the available codes?
All providers/payors should use the full set of codes in the classification to provide accurate data that are undoubtedly going to be used for purposes other than payment. All should also be held to the same standards of coding rules.
Thank you for the opportunity to address these important issues. Quantification is an important step in understanding and managing complex processes. Quantifying health care needs and services in a pluralistic system such as ours depends on standardizing information. The medical/clinical coding systems that you are examining will provide the information that companies like mine will use in the next century to help purchasers, managed care and insurance companies, providers, and researchers do their jobs more effectively. We hope these comments have been helpful to you.