INTRODUCTION
Data collection in clinical practice occurs in 2 realms, the clinical notes and the codes. The clinical note organizes data from the history, physical examination, laboratory, and imaging studies into an assessment, diagnoses, procedures performed, and treatment plan. The coding translates the clinical note into numerical codes used for research, quality improvement, analysis of morbidity, mortality, and reimbursement. Therefore, it is important to standardize clinical data coding to ensure accuracy. Administrative data collection and accurate wound classification are problematic for chronic lower-limb wounds, including foot ulcers in people with diabetes, because of the current International Classification of Diseases, Ninth Edition (ICD-9) coding system for key wound variables. Reading this article will help clinicians to recognize the sensitivity and specificity of ICD-9-CM and ICD-10-CM codes for people with diabetes and foot ulcers.
Since 1995, the United States has used the International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) to (1) summarize clinical services, (2) compute morbidity statistics, and (3) calculate reimbursement. These diagnostic and procedural codes are periodically updated as medical technology, healthcare, and global classification efforts advance. The National Center for Health Statistics announced that ICD-9-CM will transition to ICD-10-CM by October 1, 2015, well over a decade after its European implementation.
Researchers recognize the importance of ICD codes, coding rules, and coding algorithms, as well as the limitations imposed by case definition criteria when using administrative data in research. Conversion to a new ICD system may result in data anomalies when codes in one system are not available in the other. An analysis of anticipated impact of the coding changes in lower-limb ulcers is lacking. Therefore, the purpose of this article is to describe and quantify the differences between the ICD-9 and ICD-10 systems for coding chronic lower-limb foot ulcers among very well-characterized study patients with diabetes.
RESEARCH DESIGN AND METHODS
Setting and Patients
The authors conducted a cohort study enrolling individuals with diabetes and foot ulcers to determine if organized wound care was acceptable, feasible, and safe in a rural environment. Rural residents reported an increased time and transportation burden when seeking healthcare. Veterans in rural areas have limited access to care, diminished health-related quality of life, and poorer physical and mental health than do veterans in urban areas.1 Inclusion criteria were veterans older than 20 years, a diabetes diagnosis, and 1 or more ulcer episodes below the malleoli during the period from October 1, 2006, to September 30, 2007. An ulcer episode was defined as the interval from the baseline visit for an incident (new onset) ulcer to 30 days following confirmed reepithelialization.2
The subject's foot ulcer data were uniformly collected using a medical records template embedded in the electronic medical record.3 The template included information on health history characteristics and ulcer-specific characteristics. This was augmented with wound photographs. These 3 combined data sources were considered the "criterion standard" for comparison. A certified wound care specialist and medical coder (J.R.L.) reviewed the medical record for the baseline and incident ulcer visits for the study population. The medical coder then selected ICD-9 and ICD-10 codes based on specific health history, diagnoses, and ulcer-specific clinical data. A multidisciplinary panel consisting of a dermatologist, physician's assistant, and certified wound care nurse reviewed and classified the diabetic foot ulcers by consensus. The authors' Veterans Affairs Health Services Research and Development-funded project (IBA 09 061) received Human Subjects Approval from Veterans Affairs Puget Sound (IRB no. 00253).
Considerations for Using ICD-9 and ICD-10
The authors recognize ICD-9 and ICD-10 have fundamentally different structures and are organized using different classification axes as shown in Table 1.4 The main axis for ICD-9 is the nature of the health condition, and the main axis for ICD-10 is the body region of the health condition, with the diagnostic code containing the highest degree of specificity always assigned in classification.
The ICD-9-CM has 3 to 5 numeric characters plus V (factors influencing health status) and E (external causes of injury and poisoning) codes, containing 14,000 alphanumeric diagnosis codes and 4000 5-character numeric procedure codes. In contrast, the ICD-10 contains 68,000 alphanumeric diagnosis codes that include the supplementary classifications V and E of ICD-9-CM as combination diagnosis/symptom codes. The inpatient procedural classification code for ICD-10 (ICD-10-PCS) contains an additional 73,000 3- to 7-character alphanumeric codes as illustrated in Table 1, which specify etiology, severity, and encounter type.5
Although both coding systems share organizational conventions and common formatting, codes that exist in ICD-10 may not correspond directly to codes in ICD-9, and vice versa. A consortium of government and professional organizations, including the National Center for Health Statistics and the Centers for Medicare & Medicaid Services, created tools to link data between the 2 code sets called General Equivalence Mappings (GEMs) in the absence of one-to-one matches between codes in ICD-9 and ICD-10.6,7 These GEMs assist health information data users translate codes forward and backward between ICD-9-CM and ICD-10-CM and logically organize the 2 code sets.
The 3 possibilities in translating codes between the 2 classification systems are (1) one-to-one mapping where a single code in 1 system links to a single code in the other; (2) GEM type single-entry codes, a one-to-many mapping, and (3) GEM type combination codes. With single-entry codes, unlike one-to-one mapping, codes in 1 system may be used multiple times in the other system, each time linked to different codes. For example, a pressure ulcer of the heel is coded as 707.07 in ICD-9, whereas there are more than 15 unique codes in ICD-10, given the separate codes for left, right, or unspecified heel combined with depth of injury to the skin, fat layer, muscle, or bone. The GEM type combination codes involve diagnoses and describe both the underlying etiology of the wound and its manifestation. Both codes are required to be valid. Diabetic foot ulcers are an excellent example of this coding convention in both coding systems. Table 2 illustrates ICD-10's additional sensitivity according to type of diabetes, type of complication, laterality of wound location, and depth of wound. Study definitions are found in Table 3.
Data Collection
A coding algorithm for lower-limb wounds in individuals with diabetes was created by translating the ICD-9 codes to ICD-10 and vice versa by individually cross-referencing each diagnosis backward and forward in each system. The algorithm was tested on the study patients with diabetes and foot ulcers, and foot ulcers in the study were coded for both systems. Codes in both systems were examined for their ability to capture select patient health history and wound care variables as shown in Table 4. Ulcer etiology was determined based on clinical testing and/or specialist assessment. In addition, most veterans in this study had multiple documented comorbidities.
Data Analysis
The authors compared ICD-9 and ICD-10 with the criterion-standard medical/research record data by first computing frequencies for health history and ulcer variables. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated for each ulcer characteristic variable in the ICD-9-CM data, and then the ICD-10-CM data. Kappa was used to assess agreement between each code's data and the medical/research record data.
Data were analyzed descriptively using Statistical Package for Social Sciences 17.0 software (SPSS, Inc, Chicago, Illinois). Frequencies were calculated for each variable for each ICD system. The unit of analysis for the patient characteristics was the episode of care. The unit of analysis for the ulcer characteristics was the incident ulcer. Mappings were created for each ulcer, and a coding algorithm for diabetic foot ulcers was developed (Table 4). For each ulcer characteristic variable identified in the medical/research record data, estimates of sensitivity and specificity of ICD-9-CM and ICD-10-CM were calculated (Table 5).
RESULTS
There were 49 veterans meeting the study criteria with 65 episodes of care and 81 incident ulcers. The average age for the cohort of 49 veterans was 66 (SD, 11) years. Average body mass index was 34 (SD, 7.4) kg/m2. The cohort was 100% male, and all members had type 2 diabetes. In 32% of patients, HbA1c levels greater than 8% were present, and prior amputations had occurred in 10 patients (20%) (data not shown).
Table 4 illustrates the mapping algorithm between ICD-9 and ICD-10. One-to-one mapping was applied to a limited number of health history and ulcer characteristics, such as corns and callosities. Single-entry mapping (one-to-many) pertained to most of the ulcer characteristics codes because of the limited number of descriptive codes in ICD-9 compared with the multiple codes in ICD-10 to specify ulcer location, laterality, and depth/severity. Combination code mapping was required for many of the health history characteristics, including neuropathy, angiopathy, arthropathy, and atherosclerosis, because 2 codes were required in ICD-9 to describe many of these conditions, whereas a single code described the condition in ICD-10. In ICD-10, the specific combination code for type 2 diabetes mellitus with foot ulcer is E11.621. It is important to note that the location of the ulcer must also be coded in both coding systems, but the descriptive capability of ICD-10 is much greater than ICD-9.
Table 5 reviews health history and wound variable data, showing the sensitivity for each variable coded by system. Both ICD-9 and ICD-10 systems had similar frequencies for health history variables. The ICD-9 captured only 69 of incident foot ulcers (85%) compared with ICD-10, which recorded 78 of the 81 incident ulcers (96%). For ulcers located on the heel or midfoot, 94% of ulcers were identified in ICD-9, and all were captured in ICD-10. For ulcers on other parts of foot (ie, toe or dorsal foot), the ICD-9 identified 82%, and the ICD-10, 95%.
The agreement between the medical/research record data and the coded data for ulcer characteristics was evaluated. A sensitivity analysis was used to determine whether the ulcer characteristics were depicted in the coded data when they were present in the medical/research record data. Ulcer location had lower sensitivity in ICD-9, 82.5% to 94.4% compared with 95.2% to 100% in ICD-10. Positive predictive value, the probability that the data actually represent the underlying condition, was 75% or greater for 6 ulcer characteristics in ICD-9 and 13 ulcer characteristics in ICD-10. Specificity was calculated to measure the absence of the ulcer characteristics in the coded data when they were also absent in the medical/research record data. Negative predictive value, probability that data absent from the coded data were also absent from the medical/research record, is 62.1% for 6 characteristics (ulcer location, gangrene, paronychia, cellulitis, abscess, and osteomyelitis) in ICD-9 compared with 85.7% or greater for all 13 measured ulcer characteristics in ICD-10. Measures of sensitivity and specificity for ulcer characteristics were consistently higher in ICD-10 versus ICD-9, ranging from 75.0% to 100%.
The score indicating agreement between the medical/research records and coded data was higher in ICD-10 for 10 ulcer characteristics (0.737-1.0) than in ICD-9. These characteristics included ulcer location (heel and midfoot/other part of foot), laterality (left/right), depth/severity (limited to breakdown of skin/with fat layer exposed/with necrosis of muscle/with necrosis of bone/unspecified), and infection (abscess). Two ulcer characteristics for infection, paronychia and cellulitis, had higher values in ICD-9. The values for the 2 sets of code data were equivalent for gangrene and osteomyelitis.
CONCLUSIONS
The authors analyzed ICD-9 and ICD-10 codes using a criterion-standard data set of prospectively collected data, photographic wound images, and the medical records. The authors found when mapping diabetic foot ulcers using ICD-9 and ICD-10 that there was improved classification of chronic lower-limb wounds in study participants using ICD-10. Other conditions may have fewer or more differences in coding details than the diabetic foot ulcer example.
Other strengths of ICD-10 include capturing data for a prior history of foot ulcers, a strong risk factor for further ulcerations or lower-limb amputation,8,9 and inclusion of ulcer location (laterality) and severity, which will improve capture of multiple ulcers in ICD-10. In the authors' study, the added precision of ICD-10 allowed more incident ulcers (criterion standard) to be identified than were identified using ICD-9. The authors' study shows ICD-10 will improve coding sensitivity and specificity for patients with diabetes and foot ulcers; however, care and attention to documentation in the patient's record are still needed.
There are limitations in classifying chronic lower-limb wound data using ICD codes. Coding systems are only as good as the documentation provided. If a provider fails to adequately document an encounter for a chronic lower-limb wound, medical coders will not code it, and researchers will be unable to collect data on that wound.
The ICD-9 codes do not capture multiple ulcers for the same patient if they occur on the same body part. For pressure ulcers, ICD-9 allows for the assignment of only 1 code for a patient with multiple ulcers on the same body part.10,11 Severity of ulcer is captured only for pressure ulcers and does not apply to other chronic wounds, thereby further limiting data capture for diabetic foot ulcers. Multiple ulcers and history of ulcers are not captured by coded data because each unique ICD-9-CM diagnosis code may be reported only once per patient encounter. Studies intending to capture these data would be limited to prospective research extracting data directly from the medical records.
Some studies12-15 in countries that have already adopted ICD-10 find very few data loss in the transition from ICD-9; however, other studies16 show that while both systems captured major procedures relatively well, neither system properly coded less invasive or minimal procedures. Quan et al13 found little difference in validity between the ICD-9-CM and ICD-10 systems for capturing chronic clinical conditions in a Canadian study; however, the authors also attributed this apparent lack of difference to coders' unfamiliarity with the new coding system and predicted that validity of ICD-10 would improve as coders gained experience with the system. A study from Australia, where ICD-10 has been routinely used since 1998, supports the supposition that data coding quality improves the longer the system has been used. Henderson et al12 audited hospital discharge data and found agreement between diagnostic codes in 1998 to 1999 at 85% and at 87% from 2000 to 2001,12 suggesting a slight learning curve during the transition period for the new coding systems. However, the development of clinically validated coding algorithms will facilitate the transition.
Recommendations
Improving data capture with ICD-10-CM will enhance chronic wound databases. To take full advantage of the features of ICD-10, training and support for clinicians and medical coders are needed.
Merging or comparing data from different time periods by mapping between coding systems can facilitate longitudinal research on patients with diabetic foot ulcers. Mappings and algorithms for all types of chronic lower-limb wounds need to be tested to improve data capture in administrative databases.
Future Research
Risk factors for foot ulcers in patients with diabetes are well established, yet interventions to prevent and treat them have been limited in part because of the difficulty collecting accurate foot ulcer data. For chronic lower-limb wounds, including diabetic foot ulcers, data collection, and correct wound classification have been restricted by the ICD-9 lack of specificity. The ICD-10 removes much of this barrier and offers new benefits of classification for studies on improving outcomes for patients with diabetes.
PRACTICE PEARLS
* Risk factors for foot ulcers in patients with diabetes are wellestablished, yet interventions and treatment have been limited by difficulty in collecting accurate data.
* Data collection and correct wound classification in the United States have been restricted by the lack of specificity of ICD-9.
* ICD-10 will improve coding sensitivity and specificity for patients with diabetes and foot ulcers; however, accurate and detailed documentation in the patient record is required.
* Implementation of ICD-10 on October 1, 2015 in the United States will offer new benefits of classification for studies on improving outcomes for patients with wounds and diabetes.
* Training and support for clinical documentation and medical coding is needed to facilitate full advantage of the new coding system.
REFERENCES