Background

The history of ICD dates back at least to 1763 when French physician Francois Bossier de Lacroix, seeking to aid his fellow physicians in making diagnoses, published a classification system listing 10 major classes of diseases and 2400 individual diseases.22

In 1853, the first International Statistical Congress appointed William Farr and Jacob Marc d'Espine to create an internationally acceptable classification of causes of mortality.23 Consequently, the general arrangement proposed by Farr was selected as the basis of the International List of Causes of Death (ILCD). The first ILCD was developed by a committee chaired by Jacques Bertillon (Chief Statistical Services of the City of Paris) in 1893, and was introduced at the International Statistical Institute in Chicago.23 In 1898 it was adopted by Canada, Mexico and the USA.23 Subsequently, ILCD classifications for mortality reporting were officially published in 1900, 1910, 1920, 1929 and 1938.23

The WHO became the custodian of ICD in 1948, and in 1949 adopted the ICD, which was expanded to include morbidity coding.23 As the WHO assumed responsibility for preparing and publishing the revisions in 1948, revisions have been performed every 10 years; thus WHO sponsored the seventh and eighth revisions in 1957 and 1968, respectively. The ICD-9 revision provided additional, detailed four digit level categories and some optional five digit subdivisions. ICD-9 was published in 1977 by the WHO Department of Knowledge Management and Sharing. However, the WHO no longer publishes or distributes the ICD-9 as it is now in the public domain.

Work on ICD-10 began in 1983 and was completed in 1992. The code set allows more than 155 000 different codes and permits tracking of many new diagnoses and procedures, a significant expansion on the 17 000 codes available in ICD-9.6 ,19 ,23 The unchanged international version of ICD-10 is used in about 110 countries for cause of death reporting and statistics. Nonetheless, many countries have not adopted it for mortality and morbidity coding.6 ,20

In 2003, the Health Insurance Portability and Accountability Act (HIPAA) of 199624 named ICD-9 as the code set for reporting diagnoses and procedures in electronic administrative transactions. On 16 January 2009, the Department of Health and Human Services published a regulation requiring the replacement of ICD-9 with ICD-10 as of 1 October 2013. The ICD-10-CM will be used for diagnoses coding and the procedure coding system (ICD-10-PCS) for inpatient procedure coding.6 ,25 All HIPAA ‘covered entities’ must make the change; a prerequisite to ICD-10 is the adoption of electronic billing (EDI) V.5010 by 1 January 2012.

Evolution of ICD-10

The ICD-9-CM, the clinical modification of ICD-9, developed by the National Center for Health Statistics for reporting morbidities, was adopted for use in the US in 1979. In 1983, the Inpatient Prospective Payment System was adopted, and ICD-9-CM volumes 1, 2 and 3 were used for assigning cases to diagnoses related groups. As a result of the advances in medicine since ICD-9-CM was implemented, the system has been updated and revised periodically with the establishment of regular updates annually via a coordination and maintenance committee. Responsibility for maintenance of the ICD-9-CM lies with the National Center for Health Statistics and the Centers for Medicare and Medicaid Services (CMS).

HIPAA24 has identified 10 standard transactions for electronic data interchange for the transmission of healthcare data, claims and encounter information, payment and remittance advice, and claims status and inquiry. Code sets are the codes used to identify specific diagnoses and clinical procedures on claims and encounter forms. Such examples of code sets for procedures, diagnoses and drugs with which providers are familiar include Healthcare Common Procedures Coding System (HCPCS), Current Procedural Terminology (CPT), ICD-9 and National Drug Code (NDC).

Structure of ICD-10

The ICD-10-CM has the same type of hierarchy in its structure as ICD-9-CM. All codes have the same first three digits describing common traits, with each character beyond the first three providing more specificity. However, ICD-10-CM is alphanumeric with up to seven digits of specificity. It also has the same organization and use of notes and instructions. When a note appears under a three character code, it applies to all codes within that category, and notes under a specific code apply to that single code.26

However, there are substantial differences between ICD-9 and ICD-10. These are described as improvements in the content and format of the ICD-10-CM, which include:26

1. The addition of information relevant to ambulatory and managed care encounters.

2. Expanded injury codes in which ICD-10-CM groups injuries by the site of the injury, as opposed to grouping in ICD-9-CM by type of injury or type of wound.

3. Creation of combination diagnoses and symptom codes, which reduces the number of codes needed to fully describe a condition.

4. Greater specificity in code assignment.

5. V and E codes being incorporated into the main classification in ICD-10-CM.

6. ICD-10-CM codes being alphanumeric and including all letters except U.

7. The length of codes in ICD-10-CM being a maximum of seven characters, as opposed to five digits in ICD-9.

8. Some vacant, three character codes in ICD-10 to allow for revision and future expansion.

Overall, ICD-10-CM far exceeds its predecessors in the number of codes available. Table 1 illustrates the comparison of chapter numbers and titles in ICD-9-CM with those in ICD-10-CM.26 However, diseases of the eye and its adnexa and diseases of the ear and mastoid process will have their own chapters in ICD-10-CM.

ICD-10-CM characteristics are as follows:

• Tabular lists containing cause of death titles and codes (volume 1).

• Inclusion and exclusion terms for cause of death titles (volume 1).

• Alphabetic index to diseases and nature of injury.

• External causes of injury.

• Table of drugs and chemicals (volume 3).

• Description, guidelines and coding rules (volume 2).

Table 2 illustrates the differences between ICD-9 and ICD-10.22

Lack of need for ICD-10 and its complex mechanism

ICD-10 proponents argue that in ICD-9 the number of diagnostic codes based on alphanumeric name badges the diagnoses carry for insurance billing and other purposes will increase from 14 000 to 69 000. In addition, the number of codes for procedures that can be performed on an inpatient basis in hospitals will jump from about 3800 to 72 000. The shift will affect just about every aspect of clinical and business operations as the codes document what clinicians do with patients and are embedded in nearly all clinical information and billing operations nationwide. Proponents also agree that in effect, switching to ICD-10 means that every diagnosis and piece of claims activity will operate under a different language than it does now.19

It has been argued that ICD-10 implementation requirements are much more extensive than any encountered with HIPAA to date, or even with Y2K.27 The most valuable lesson from both HIPAA and Y2K is that successful ICD-10 implementation will cost more than present estimations, potentially by a large amount. While it is guaranteed that the costs will be a part of ICD-10, the benefits are only potential. Understandably, physicians, hospitals, health plans, software vendors, government agencies and nearly all other healthcare entities bracing for the implementation of the inevitable ICD-10 coding system are anxious as ICD-10 introduces a monumental new set of diagnostic and procedural codes. The anxiety is most severe for small physician practices, hospitals and health plans which may not be ready for ICD-10 by the October 2013 deadline. An additional concern is that some organizations will not be ready for the transaction system conversion on 1 January 2012.19 In fact, ICD-10, similar to various healthcare regulations in the past, including implementation of HIPAA and the Occupational Safety and Health Administration (OSHA) compliance regulations, has created a cottage industry of consultants. While there is no proof of savings or improved efficiency, there are estimations of overriding costs and inconvenience for practitioners. These include overhauling computer software systems, training staff and making other needed changes. The Department of Health and Human Services projects a total cost to the US healthcare system to be in the range of $2.3 billion to $2.7 billion over a period of 15 years. Furthermore, other cost estimates are highly variable, with one consulting firm estimating $3.2 billion to as much as $8.3 billion and another consulting firm estimating $15.2 billion to $34.1 billion.21 America's Health Insurance Plans (AHIP), the Health Plan Trade Group, projects that the cost to health plans alone will be $2 billion to $3 billion.28 Aetna's estimation is approximately $50 million to $70 million a year through 2013 to make the conversion.19 Finally, estimated costs for physician practices will vary from $83 000 for a three doctor practice to $2.7 million for a 100 doctor practice.21 Hospitals are estimating the costs to be approximately $15 million to $20 million for each hospital.

A Research and Development study estimated that cost, as well as lost productivity and time to resolve coding issues, could range between $5 million and $25 million nationally for the first year and would decline to $16 million after 6 months. However, the amount of productivity lost in a practice depends on the level of understanding of ICD-10 and training obtained by the physicians and staff. The Research and Development study also estimated that software costs could range from $500 million up to $1.6 billion nationally, with smaller or solo practices expected to spend $180 million overall. The American Medical Association, in a guide to preparing for ICD-10, has presented cost projections29 for a small surgical practice of three physicians to be $44 000 whereas for a typical medium sized practice with 10 physicians the cost will be $93 000.

ICD-10 changes for practice

Conversion of ICD-9-CM codes to ICD-10-CM codes will likely be very complicated. Supplementary table 1 (available online only) illustrates the select examples of codes utilized in neurointerventional practices, showing codes variable from 2 to 20 ICD-10 codes for each ICD-9 code. This example illustrates the complicated nature of conversion, creating further confusion. To understand the transition and mapping, one would need to expand multiple precious resources, including time and financial resources. Furthermore, as illustrated in table 3, one ICD-10 code represents multiple ICD-9 codes compounding the complexity and complications. The entire argument of granularity increasing and simplicity are lost with further restrictions than ICD-9.

Practical impact

To help facilitate care and commerce, the government has invested in providing mappings between ICD-9 and ICD-10 and vice versa. There are two such mappings endorsed by CMS: the general equivalence mappings (GEMs) (for both ICD-9 to ICD-10 and ICD-10 to ICD-9) and the reimbursement maps (for ICD-10 to ICD-9 only). The GEMs established links among codes that are generally equivalent in each code set. The reimbursement maps were created after the GEMs maps and are more specific, identifying the top candidate mappings from within the GEMs.

Some published statistics30 may illuminate the challenges inherent in linking across the code sets. In the GEMs maps for procedures from ICD-9 to ICD-10, multiple examples are provided, although these are not specific to neurointerventional radiology or interventional pain management.6

• There are 255 instances where a single ICD-9 code can map to more than 50 ICD-10 codes.

• There are 119 instances where a single ICD-9 code can map to more than 100 ICD-10 codes.

Some GEMs maps of ICD-9 to ICD-10 are not specific:

• There are 7239 instances in the mappings for diseases where a single ICD-10 code can map to more than one ICD-9 code.

• There are 7241 instances in the mappings for procedures where a single ICD-10 code can map to more than one ICD-9 code.

In the reimbursement maps from ICD-10 to ICD-9:

• There are 3684 instances in the mappings for diseases where a single ICD-10 code can map to more than one ICD-9 code.

• There are 2135 instances in the mappings for procedures where a single ICD-10 code can map to more than one ICD-9 code.

Furthermore, different rules exist for different purposes. While CMS has tried to create clarity with GEMs and reimbursement mappings, the results are extremely disappointing. It has been shown that GEMs ICD-10 to ICD-9 mappings have a 5.1% exact match for diseases and only 0.1% exact match for procedures. In contrast, GEMs ICD-9 to ICD-10 mappings have an approximately 20.1% exact match for diseases and 1.2% exact match for procedures. With so few exact matches, it may be impossible for struggling practices to continue.

The troubles do not seem to stop with the different rules and different purposes and multitude of codes. They are also related to adopting software with multiple crosswalk variations. Independent package software vendors will have different offerings and divergent approaches to cross walking. Some may support sophisticated rules and others will not.30 Either way, if medical systems, claims systems and financial systems have different tools, things will become extremely difficult, causing hardship. On some occasions, it appears the data may have to be entered into multiple programs. Wollman30 describes that any business rules for mappings would need to be entered and stored in at least five systems, plus any analytic systems that source data from the applications. Thus with cross walking systems the potential for errors and rework is astronomical, due to the over 250 GEMs mappings, approximately 150 000 reimbursement mappings and over 160 000 ICD-9/ICD-10 codes used to manage a total of approximately 600 000 records and potentially tens of thousands of overrides in addition to the GEMs and reimbursement maps. Most prudent practices and providers require at least 3 years of historical data for trending and analysis purposes.

On September 30 2013, all of this history will be encoded in ICD-9 nomenclature. On the following day and going forward, the ‘new history’ will start to be encoded in ICD-10. Consequently, any type of trending will either require a migration of all of the history to ICD-10 or some mechanism for stepping up ICD-9 codes to ICD-10 or stepping back ICD-10 codes to ICD-9 for analysis. However, it may be necessary to use both. Considering other major issues, this may be a minor problem.

Preparing for the inevitable

Appropriate preparation about switching to ICD-10 is of paramount importance. Some of the most important concerns are as follows:

1. Whether organizations will be able to undertake the huge array of changes to make the conversion possible.

2. Whether organizations can do so in time to meet the government imposed deadline of 1 October 2013 for the transition.

3. Where does the money come from and is US healthcare ready for more expenses.

Timing is crucial to manage practices in the USA. Most provider systems, health plans and software vendors, despite reporting that they are on schedule in their conversions to ICD-10 and the new electronic transaction format, it is evident that some have not even started their preparations. As an example, the American Hospital Association survey in October 2010 found that only one-half to two-thirds of hospitals had taken the recommended planning and assessment steps.19 Not surprisingly, many providers are not even aware of ICD-10 changes, let alone the implementation date. Very few physicians have completed the impact survey. This is a major issue for smaller practices.