Article Text
PDF
Abstract
Although the Abbreviated Injury Scale (AIS) is the most widely used severity scoring system for traumatic injuries, hospitals are required to document and bill based on the International Classification of Diseases (ICD). An expert panel recently developed a map between ICD-9-CM and ICD-10-CM to AIS 2005 Update 2008. This study aimed to validate the recently developed map using a large trauma registry. The map demonstrated moderate to substantial agreement for maximum AIS (MAIS) scores per body region based on expert chart review versus map-derived values (range: 44%–86%). Injury Severity Scores (ISSs) calculated from expert coders versus map-derived values were also compared and demonstrated moderate agreement (ICD-9-CM: 48%, ICD-10-CM: 54%). Although not a perfect conversion tool, the new ICD-AIS map provides a systematic method to assign injury severity for datasets with only ICD-9-CM and ICD-10-CM codes available and can be used for future injury-related research and data analysis.
- coding systems
- severity scales
- registry
Statistics from Altmetric.com
Introduction
Classifying injuries based on severity is critical for clinical decision-making, quality improvement of trauma centres and for injury-related research. The Abbreviated Injury Scale (AIS), created by the Association for the Advancement of Automotive Medicine (AAAM), is the most widely used severity scoring system and can be applied to both vehicular and non-vehicular trauma.1 AIS codes designate the specific body region involved as well as the severity of an injury on a scale of 1 (minor) to 6 (maximal). These codes can then be used to calculate the Injury Severity Scores (ISS), defined as the sum of the squares of the maximum AIS (MAIS) severity of the three most severely injured body regions. However, hospitals are required to document and bill based on the International Classification of Diseases (ICD). An expert-based map was recently developed between the 9th and 10th revisions of the ICD to the AIS 2005 Update 2008.2 3 The goal of creating this map was to provide a comprehensive solution to bridge the gap between hospital-coded injuries using ICD-9-CM and ICD-10-CM and AIS ISSs. Prior to this updated map, the most widely used map between ICD-9-CM and AIS-85 was validated and, as of spring 2017, has been cited more than 250 times in the peer-reviewed literature.4 The objective of this study was to validate the most recent expert-based map between both ICD-9-CM and ICD-10-CM and AIS codes using a large trauma registry. We hypothesised that the agreement between MAIS and ISS as determined by coders and values derived from the 2016 ICD-AIS map would be comparable with a previously validated map using an older version of AIS.4
Methods
Data sources
A retrospective analysis was conducted using data from paediatric and adult trauma patients in the Harborview Medical Center (Seattle, Washington, USA) trauma registry. Patients were included if they were treated between 1 January 2010 and 30 December 2015 and had ICD-9-CM or ICD-10-CM injury codes as their primary source of diagnoses; however, ICD-10-CM codes were not assigned until 1 June 2015 and thereafter. Each patient encounter had up to 20 injury-related ICD-9-CM and/or 20 ICD-10-CM codes assigned by medical providers and billing. In addition, each patient was assigned up to 20 AIS codes (and a composite ISS score) based on chart reviews by individuals trained and certified in AIS coding. No patient identifiers were used in the data set, and the project protocol was deemed exempt by our institutional review board.
Mapping and creating new variables
AIS codes assigned by the trained coders based on chart review were separated into ISS body regions and AIS severity score based on the AIS manual.1 The six ISS body regions included (1) head/neck, (2) face, (3) chest, (4) abdominal and pelvic contents, (5) extremities and pelvic girdle and (6) external. Injuries to the spine were coded to either head/neck, thorax or abdomen body regions based on location of the spine injury. AIS severity and ISS body regions were assigned for each ICD-9-CM and ICD-10-CM code based on the AAAM expert-based map. MAIS severity was calculated for each body region involved. ISS was then calculated using the sum of the squares of the MAIS for the three most severely injured body regions (values from 1 to 75, with an automatically assigned score of 75 if the patient had at least one MAIS of 6 in any body region). By deriving MAIS for each body region and ISS based on ICD-9-CM and ICD-10-CM scores using the expert map, statistical comparisons could then be made between the map-derived variables and coder-assigned MAIS and ISS.
Statistical analysis
Using the abstracted, trained coder-based AIS scores as the gold standard and compared with the mapped AIS codes, summary statistics were used to calculate the percent agreement for MAIS for the ISS body regions, as well as the percent agreement for exact ISS, ISS within ±5 points, and ISS for patients whose abstracted ISSs were between the following ranges: 1–4, 5–14, 15–24 and 25+.
Results
Table 1 shows the number, age, sex and ISS range distribution (based on abstracted codes) for patients with ICD-9-CM and ICD-10-CM codes. There were 34 989 patients included in the ICD-9-CM analysis with a mean age of 42.9 years (SD 23.3, range 0–105) and 68% were male. There were 1990 patients included in the ICD-10-CM analysis with a mean age of 44.4 years (SD 23.3, range 0–96) and 66% were male. Table 2 describes the number of ICD-9-CM and ICD-10-CM codes assigned to each body region by expert chart review and the per cent agreement in MAIS between the abstracted AIS codes (based on expert chart review) and MAIS calculated from map-assigned severity scores.
Number, age, sex and ISS ranges of patients with ICD-9-CM codes and ICD-10-CM codes
Agreement in MAIS scores
Non-matching most frequently occurred due to mapped codes being lower in severity than the abstracted codes. Additionally, 4.2% of the ICD-9-CM codes and 12.4% of the ICD-10-CM codes in the sample were assigned an ‘unspecified’ ISS body region and/or AIS severity.
Table 3 shows the per cent agreement between ISS based on abstracted AIS (from expert chart review) and ISS calculated from MAIS per body region derived from the 2016 ICD-AIS map. For ICD-9-CMs, 48% of cases had exact agreement between abstracted ISS and ISS calculated from map-derived values. For ICD-10-CMs, 54% had exact agreement. For both data sets, the majority of cases had map-derived ISSs that either exactly matched or fell within 5 points of the abstracted ISSs (ICD-9-CM: 76%, ICD-10-CM: 71%). Table 4 shows the percentage of cases with exact agreement between abstracted ISS and map-derived ISS for abstracted ISSs within each range. Agreement percentages decreased as abstracted ISS ranges increased.
Agreement in ISS
Agreement for exact ISS
Discussion
The results demonstrate moderate to substantial agreement for both MAIS as well as ISS when comparing expert-based chart review and ICD-AIS map-derived values. Compared with results of the Mackenzie et al study used to validate a prior ICD-9-CM to AIS-85 map used before creation of the 2016 updated map,4 MAIS agreement for the ICD-9-CM map was comparable or higher in this analysis for nearly all body regions (head/neck: 49% vs 48%, face: 53% vs 54%, thorax: 59% vs 63%, abdomen: 58% vs 54%, extremities: 80% vs 74%, external/burns: 79% vs 70%). In addition, a larger patient sample size was used for the ICD-9-CM analysis (n=34 989) compared with that of the Mackenzie et al map (n=1120).
It is interesting to consider the discrepancy between MAIS agreement percentages in the different body regions. Head/neck had the lowest reported agreement (ICD-9-CM: 49%, ICD-10-CM: 44%) between abstracted MAIS and MAIS based on the map. This is likely due to the fact that many ICD-9-CM and ICD-10-CM codes for the head are ambiguous in detailing the injury, making it difficult to assign an accurate severity score. For example, ICD-9-CM scores do not indicate relative sizes of subdural or epidural haematomas. Additionally, a single ICD-9-CM or ICD-10-CM score may include descriptions of multiple head injuries (eg, a haemorrhage, fracture and coma), all of which might indicate different levels of severity. However, the agreement between abstracted MAIS and MAIS based on the map for the face region is much greater for the ICD-10-CM map compared with ICD-9-CM (86% vs 53%), perhaps indicating a greater level of detail and specificity of ICD-10-CM codes for facial injuries compared with ICD-9-CM. Of note, the non-matching most frequently occurred secondary to mapped codes being lower in severity than abstracted codes, which aligns with the ‘conservative coding’ philosophy of AIS and of the map itself.
Similarly, the agreement between scores based on abstracted AIS codes and ICD-AIS mapped codes that fell within exact ISS ±5 was 76% (ICD-9-CM) and 71% (ICD-10-CM), both exceeding the agreement of 70% reported by MacKenzie et al.4 Our data also demonstrate a higher performance of the map for exact ISS agreement in less severely injured patients. One possible explanation is that lower ISS scores are more likely to be based on an injury in a single body region, with an increased likelihood of exactly matching the abstracted versus mapped codes.
Future analysis should be considered for the ICD-10-CM map, given its much smaller sample size compared with that used to validate the ICD-9-CM map (n=1990 vs n=34 989) and to account for possible inconsistencies in its initial use soon after its implementation. In addition, validation could be considered for specific patient populations such as paediatric or geriatric patients.
There are limitations to this study that should be considered. First, the results are only from a single institution. However, it was a very large sample size from a regional trauma centre with a large catchment area, and all AIS scores were assigned by individuals trained and certified in AIS coding. Additionally, many ICD-9-CM and ICD-10-CM codes in the sample were assigned ‘unspecified’ ISS body region and/or AIS severity (ICD-9-CM: 4.2%, ICD-10-CM: 12.4%). However, this was lower than the percentages of unmapped codes based on all ICD injury codes as described in the original paper by Loftis et al (ICD-9-CM: 12.1%, ICD-10-CM: 21.2%).2 This suggests that, while there are some codes that cannot be mapped, in a large sample, this is a relatively small proportion of the total number of injuries. This is primarily driven by the limitation in code descriptions, which often do not have enough detail to confidently assign an AIS code. For example, a majority of the ICD-10-CM codes that could not be assigned severities or body regions were T07, ‘Unspecified multiple injuries’. Finally, not all possible codes in the map were actually tested, as some are extremely rare even in a large trauma registry.
Conclusion
This data analysis provided validation of a recently developed map between ICD-9-CM and ICD-10-CM and AIS codes. While the use of AIS scores assigned by formally trained coders through medical record abstraction is ideal, this map can be used to assign MAIS and ISS when AIS codes are not available.
What is already known on the subject
AIS is the most widely used severity scoring system for traumatic injuries, but hospitals are required to document and bill based on a different coding system, the International Classification of Diseases (ICD).
The most widely used ICD-to-AIS map was created and validated in 1989 by MacKenzie et al between ICD-9-CM and AIS-85.
An updated map between ICD-9-CM and ICD-10-CM and AIS 2005 Update 2008 was created recently by an expert panel of trained coders.
What this study adds
Validation of the most recent, expert-based map between ICD-9-CM, ICD-10-CM and AIS codes with agreement statistics comparable with that of the previous map’s validation.
Although not a perfectly precise conversion tool, this ICD-to-AIS conversion map provides a systematic method to assign severity and ISSs for large administrative datasets that only have ICD-9-CM or ICD-10-CM codes available.
Acknowledgments
The authors would like to thank Janet Price, RN, MSA, Kathy Cookman, BS, and Pat Gillich, MS, and all the members of the original expert panel that created the ICD-AIS map used for this validation for their assistance. The authors would also like to thank Dr Eileen Bulger and Patricia Klotz from Harborview Medical Center for use of their data.
Footnotes
Contributors KMG and MRZ designed the research, conducted the research, analysed the data, drafted the manuscript and critically reviewed and approved the final manuscript; MRZ had primary responsibility for final content.
Competing interests None declared.
Ethics approval Institutional review board.
Provenance and peer review Not commissioned; externally peer reviewed.