Revision of the International Classification of Diseases to include standardized descriptions of multiple injuries and injury severity
Shinji Nakahara a & Junichiro Yokota b
a. Department of Preventive Medicine, St. Marianna University School of Medicine; 2-16-1 Sugao, Miyamae-ku, Kawasaki, Kanagawa, 216-8511, Japan.
b. Sakai Municipal Hospital, 1-1-1 Minamiyasui-cho, Sakai, Osaka, 590-0064, Japan.
Correspondence to Shinji Nakahara (e-mail: email@example.com).
(Submitted: 20 April 2010 – Revised version received: 28 December 2010 – Accepted: 11 January 2011.)
Bulletin of the World Health Organization 2011;89:238-240. doi: 10.2471/BLT.10.078964
The International Classification of Diseases (ICD) is widely used as a source of mortality statistics. However, two major difficulties arise when recording, presenting and analysing injury data using this diagnosis classification. First, due to the absence of standardized methods for describing multiple injuries, they are described in various ways in mortality and morbidity statistics. For example, designating the most severe injuries as the primary injuries or categorizing multiple injuries as such without further details.1 Second, an increasing need to describe injury severity for case-mix groups has led to the introduction of various additional severity-scoring methods, as the ICD itself, as a diagnosis classification for mortality, does not consider severity. To avoid the costs of additional severity scoring, methods have been developed to convert administrative ICD-based diagnosis codes into severity scores using computer software or to calculate survival probability for each diagnosis code from patient data. However, these approaches have respective disadvantages due to the need to track software updates or the need for large data sets to calculate the probabilities.2
The World Health Organization3 is currently advocating revision of the ICD to expand upon its largely administrative applications and allow more clinical uses. This provides an opportunity to address the issues associated with describing multiple pathologies and scoring the severity of injury data, which are also relevant to other non-injury diseases. In addition, a few low-income countries do not use the ICD or severity scores, even in the absence of vital registrations depending on periodical surveys,3 thus the revision process should facilitate their adoption of standardized methods. Here we discuss how the revised ICD system could standardize the description of multiple injuries to provide accurate statistics, incorporate severity scores to avoid additional resource input, and facilitate utilization in countries where it is not currently in use.
Describing multiple injuries
For mortality statistics, the one-dimensional principle of the ICD allows only one underlying cause of death to be selected and coded. The multi-dimensional phenomenon of multiple injury is thus usually reduced either to a single code reflecting the primary (most severe) injury or to one of a few multiple-injury codes, based upon an arbitrary decision.1,4 Selecting the primary injury when filling in death certificates, or the underlying cause from among several injuries reported in death certificates, is also an arbitrary practice that reflects the certifier’s or coder’s perception of which pathology is the most important. Choosing just one code results in a loss of information on the other, unselected, pathologies, so the resultant statistics underestimate the significance of each injury and inadequately depict the interactions between them.1 The limited number of multiple-injury codes included in the ICD cannot cover all possible patterns. For example, codes T00–T07 indicate injuries involving multiple body regions while S codes also include multiple injuries in the same body regions, (e.g. S52.7 indicates “multiple fractures of the forearm”). This arbitrariness, due to a lack of standardization, also applies to the presentation and analysis of morbidity statistics,1 although not to the way that they are recorded because clinical modifications of the ICD require the coding of each injury, thereby superseding the multiple-injury codes.
The shortcomings of one-dimensional coding have led some countries to introduce multiple coding systems for mortality statistics, in which all causes mentioned on a death certificate are coded and reported.4 It would be preferable to omit the multiple-injury codes from the revised ICD, and to code and record all injuries separately. This would allow all patients with a certain injury to be counted, even when it is not the primary injury, which is not the case with one-dimensional underlying-cause (or primary-injury) coding.1 When presenting data on multiple injuries, instead of simply listing all injuries sustained, it might be preferable to use two-dimensional coding that reflects the important attributes of the nature of the injury and the affected body region to characterize an individual’s injuries.
Proposed methods to describe multiple injuries while presenting statistics in a standardized way include the multiple injury profile, which combines information on the anatomy and the nature of the injury, using a body-region by injury-nature matrix.1 Each injury falls into one of the cells in the matrix. The multiple injury profile can summarize all of the individual injuries in one patient using cell combinations. The granularity of the categorizations used in the matrix can be changed by subdividing or collapsing the categories as needed. An abridged version of the matrix can be used as a shortlist in countries where the full list of the ICD is not used. By contrast, the matrix can be used as a supplement, in conjunction with listing all of the injuries to give complete descriptions in countries where multiple coding is done.
Multiple coding using standardized methods of presenting multiple pathologies, if applied to the whole ICD, would allow more accurate descriptions of each pathology and the interactions both within and between specific types of injury or internal cause.1,4 This would also help to clarify how underlying ailments contribute to the impact of injuries in ageing societies.
Describing injury severity
Various methods have been developed to score injury severity (Table 1).5 The abbreviated injury scale (AIS) describes the anatomical injury severity using consensus-based scores determined by experts. The revised trauma score (RTS) is based on physiological parameters independent of injury diagnoses. The injury severity score (ISS) consists of the square of the highest AIS scores in the three most severely injured body regions. The trauma and injury severity score (TRISS) predicts survival probabilities using logistic regression modelling that employs the ISS, RTS, age and injury mechanism as predictors. AIS-based methods, such as TRISS, are widely used because of their suitability and accuracy based on ample research findings. However, duplicate coding for injury diagnosis and severity carries additional costs in terms of human resources and training requirements to ensure accuracy, which is unaffordable in resource-constrained settings.2
To avoid the additional costs associated with duplicate coding, attempts have been made to assign a severity score to each ICD-based diagnosis. One successful example is a method that derives AIS severity scores from ICD-9 codes using computer software (ICDMAP).2 Although this is a validated tool, it also carries additional costs, albeit smaller ones than those associated with duplicate coding, and it notably fails to update using newer versions of the ICD and AIS, resulting in variability in the versions used in case-mix grouping methods.2
Another example is the ICD-based injury severity score (ICISS), which assigns an empirically derived severity score to each ICD code.2,5 Survival probabilities, called survival risk ratios (SRRs), are calculated for each code based on patient data (Table 1). The ICISS is a promising measure that performs as well as, or better than, AIS-based methods; however, it has some shortcomings that might hinder its use in low-income countries, particularly those with small populations. Large data sets are required to avoid large fluctuations occurring in the SRRs for rare injuries. Also, SRRs might differ across countries and over time, depending on health-care systems and improvements in treatment, thereby requiring countries to calculate and update their own data sets.2
Whereas code conversion and the ICISS operate outside the ICD framework and do not modify the diagnosis codes, an alternative approach would be to integrate consensus-based severity scores into the ICD. The revised ICD is expected to have wider coverage, including morbidity statistics and case-mix groupings.3 Integrating an AIS system into the revised ICD as a clinical modification or expansion would remove the need for duplicate coding or code conversion (and associated software updates). This would be facilitated by recent improvements in the compatibility between the ICD and the AIS.
None of the above-mentioned severity-scoring methods can be used in countries where a shortlist of ICD codes is required. The matrix-based approach can, however, be applied if the predominant AIS severity scores (because more than one code can fall in one cell) or ICISS-type survival probabilities are determined for each of the matrix cells based on empirical data.6 Assigning a consensus-based approximate severity score to each cell is also possible. This abridged method, with diminished diagnosis categories and the flexibility to handle both AIS-type and ICISS-type indices, can be used to create a short morbidity list with severity scores for resource-constrained settings.
The ICD revision process presents a good opportunity to standardize the description of multiple injuries and injury severities regardless of resource availabilities. We suggest that the revised ICD should have a multiple coding framework for individual pathologies, deactivating multiple-injury codes, so as to consider the significance of each injury or pathology and their interactions. The ICD should also incorporate consensus-based severity scores in its clinical modifications, so that case-mix groupings can be considered in resource-constrained settings without requiring duplicate coding or code-conversion software, while data-derived severity indices can be employed in less constrained settings. Matrix-based methods should also be considered, as they provide a simple basis for multiple injury description and case-mix groupings using fewer categories, making them suitable for countries where a shortlist of ICD codes is needed.
This work was supported by a grant for research on global health issues from the Ministry of Health, Labour and Welfare. The funding body did not play any role in this work.
- Aharonson-Daniel L, Giveon A, Peleg K. Gaps in injury statistics: multiple injury profiles reveal them and provide a comprehensive account. Inj Prev 2005; 11: 197-200 doi: 10.1136/ip.2005.008227 pmid: 16081744.
- Cryer C. Severity of injury measures and descriptive epidemiology. Inj Prev 2006; 12: 67-8 doi: 10.1136/ip.2006.011668 pmid: 16595417.
- WHO business plan for classifications: building blocks of health information. Geneva: World Health Organization; 2005. Available from: http://www.who.int/entity/classifications/BuisinessPlan.pdf [accessed 24 January 2011].
- Goldacre MJ, Duncan ME, Cook-Mozaffari P, Griffith M. Trends in mortality rates comparing underlying-cause and multiple-cause coding in an English population 1979-1998. J Public Health Med 2003; 25: 249-53 doi: 10.1093/pubmed/fdg058 pmid: 14575203.
- Chawda MN, Hildebrand F, Pape HC, Giannoudis PV. Predicting outcome after multiple trauma: which scoring system? Injury 2004; 35: 347-58 doi: 10.1016/S0020-1383(03)00140-2 pmid: 15037369.
- Clark DE, Ahmad S. Estimating injury severity using the Barell matrix. Inj Prev 2006; 12: 111-6 doi: 10.1136/ip.2005.010058 pmid: 16595426.