Introduction

It is well known that damage to the neural elements in the spinal canal (spinal cord and cauda equina) resulting in resolving or permanent neurological deficit can arise from many different causes other than trauma. Although there is no universally accepted term for spinal cord damage not because of trauma1 (often referred to as non-traumatic spinal cord injury (NTSCI)—but better alternatives may be ‘spinal cord damage’ or ‘spinal cord myelopathy’) this field is an important aspect of spinal cord medicine. NTSCI can occur at any age. It has been reported that the incidence of NTSCI is greater than that of traumatic spinal cord injury (SCI).2 It is proposed that this is quite likely the case in many countries, especially developed countries, and there is evidence emerging that this is the case.3 It is anticipated that the incidence of NTSCI will increase greatly in the coming decades because of population aging.4

There are many aetiologies of NTSCI reported in the literature.5, 6 Case series that have been published over the years have described the aetiologies of NTSCI using different classification systems.7, 8, 9, 10, 11, 12, 13, 14, 15 It is likely that these authors formulated the classification systems used themselves because there is no consensus international classification system for NTSCI. A consensus classification is essential in order to have good epidemiological studies, prevention, treatment and outcomes research programs and especially to facilitate comparative studies across different settings.

The International SCI Data Sets have standardised the collection and reporting of information regarding SCI patients in order to facilitate comparison of results between different studies.16 The International SCI Core Data Set describes standardised methods of collection and reporting of minimal information necessary to describe a SCI study population to allow meaningful comparison of study results and an assessment of the study population characteristics and potential sources of bias.17 One of the items in the core data set is the aetiology of the SCI. The core data set currently classifies the causes of SCI into seven categories. One of these is NTSCI. There is, however, no further sub-categorisation of the numerous possible causes of NTSCI within the core data set. The absence of an agreed classification system for NTSCI has been recognised by ISCoS as an important deficiency.

The aim of this project was to develop an International Data Set for NTSCI, including a classification system for the causes of NTSCI.

Materials and methods

This project commenced in 2005 when the ISCoS International Data Set committee discussed the need for a system of classifying the causes of NTSCI. Over the ensuing years, a number of processes and activities occurred to develop and refine the classification presented here. These included a consultative process involving over 120 key researchers and senior clinicians involved with SCI, focusing on those active in NTSCI and the International Data Set process. This was accomplished using e-mail in three rounds of communication during 2008 and 2009 through an iterative process. Invited participants were informed that they could confer with interested colleagues, or forward on the e-mail to any other potentially interested clinician they were aware of to share an interest or expertise in the topic.

Complementing the above, presentations and discussions occurred at ISCoS conferences in 2006 (Boston), 2007 (Reykjavik), 2008 (Durban) and 2009 (Florence). A half-day workshop was held on 1 September 2008 at the ISCoS conference and a detailed presentation given at a meeting on the International SCI Data Sets in October 2008.18 Numerous discussions between the authors were held over the course of this project.

A near final draft of the data set was circulated in December 2011 to participants in the development process and the ISCoS and American Spinal Injury Association Executive Committees for final feedback and suggestions.

The consultation processes described above aimed to clarify the following aspects of the data set: overall structure of the classification; placement of certain groupings (for example, vertebral column degenerative disorders, metabolic disorders) and certain specific aetiologies; missing aetiologies; definitions and data variables. During the course of the development of the data set, every effort was made to consider feedback and suggestions from participants.

International classification of external causes of injury

The World Health Organisation (WHO) has developed an International Classification of External Causes of Injury (ICECI)19 Table 1. This is based on the International Statistical Classification of Diseases (ICD) and Related Health Problems 10th edition codes of external causes (chapter 20).20 The ICECI directs that iatrogenic injury be considered traumatic. This includes complications of health care, including ‘medical or surgical care, unintentionally leading to injury or other harm, acts of omission as well as acts of commission’. As a result of the approach taken by the WHO regarding classification of iatrogenic injury, it was decided that this project should adopt the ICECI as a guiding framework.

Table 1 International classification of external causes of injury: complications of heath care

Results

The International NTSCI Data Sets are presented in both basic and extended versions, with the common element including a classification of aetiology of NTSCI. The elements of the NTSCI data sets are presented below, and then the components of the basic and extended data sets are outlined.

Classification of NTSCI aetiology

In the extended data set, a detailed classification for NTSCI aetiology was developed that consists of two axis. Axis 1 provides a hierarchical classification of NTSCI using clinical classes and pathophysiological mechanisms in two major tiers: ‘congenital–genetic’ and ‘acquired’ Table 2. Five levels are provided within the axis 1 to allow for an increasing level of detail to be recorded about the aetiology. In some circumstances, NTSCI can be caused by a cascade of events. It is intended that the axis 1 be used to classify the final aetiological process responsible for the NTSCI.

Table 2 Classification of the aetiology of non-traumatic spinal cord injury–axis 1

A number of classification principles were developed to assist those using this data set. In addition to the case studies that are available on the ISCoS website, which provide useful guidance and an explanation of approaches to dealing with a number of scenarios, the following principles for classifying the aetiology are offered:

  • Only one aetiology is coded for each case.

  • If a patient has an NTSCI lesion that occurs as a result of different causes during the course of the same admission then the condition that causes the most severe neurological impairment is the condition that should be classified. For example, a patient that had a mild thoracic myelopathy from a meningioma, then while in rehabilitation the patient had an epidural bleed from subcutaneous heparin for deep venous thrombosis prevention. The myelopathy was much worse after the bleed. In this case, because the epidural bleed was the more severe causative factor it should be the one coded as the NTSCI aetiology, although the meningioma was indirectly responsible (could be coded as axis 2, if extended data set is used, see below).

  • If a patient has a single-cause of NTSCI that could possibly be placed in two (or more) different aetiological groups in the classification then the more specific aetiology should be selected. For example, Achondroplasia that leads to compression of multiple regions of the thoracic cord is classified as ‘CONGENITAL: skeletal malformations—Achondroplasia’ and not ‘ACQUIRED: vertebral column degenerative disorders—spinal cord compression because of combination of multiple developmental and/or acquired factors’ as a matter of routine, unless there are ‘acquired factors’ that are also responsible for cord compression.

  • If a patient has transverse myelitis from an aetiology listed in the classification, then this specific cause is coded. The transverse myelitis aetiology is only selected where the cause is idiopathic.

Where NTSCI is caused by a cascade of events and if it is desired to collect this level of detail, a second axis, axis 2, is available to document the triggering diseases or processes using the ICD.20 It is suggested that up to three disease processes could be recorded, if relevant, to code the initiating event(s) that resulted in the NTSCI. Currently, most countries use the 10th edition of the ICD, but a few, most notably the USA, are still using the 9th edition. An 11th edition is also in the planning phase. Therefore, the version of the ICD used should also be indicated in the data collection process.

Timeframe of onset of NTSCI

In NTSCI, the date of onset is not always clear, which is different to the situation in most cases of traumatic SCI. In many cases of NTSCI, the timeframe of the onset of neurological damage is not instantaneous as typically is the case with traumatic SCI. Some diseases that cause NTSCI can have an onset of symptoms over minutes (for example, cord infarction), hours (for example, transverse myelitis), days (for example, spinal abscess) or weeks to months (for example, spinal canal stenosis). The pattern of the timeframe of NTSCI onset should be documented in the extended NTSCI data set. It is recommended that the timeframe over which clinical symptoms developed, which are attributable to the onset of the NTSCI be coded as either: acute (1 day), sub-acute (>1 day but 7 days), prolonged (>7 days but 1month) or lengthy (>1 month). This information has relevance for prevention and treatment programs and clinical trials.

No specific internationally recognised classification for the timeframes of disease symptom onset has been located. The time periods are included along with descriptor terms so that there will be consistency in classification of this item. They were selected to match the typical time periods of onset seen in various NTSCI aetiologies. The terms chosen are descriptive only and it is suggested that they would be useful in reporting results in the text of the manuscripts that choose to report this item.

Iatrogenic role in aetiology of NTSCI

Iatrogenic SCI occurring in the setting of health care is not uncommon, with one recent study reporting the proportion was 10% of all patients with SCI.21 These cases of iatrogenic SCI should be coded to indicate the presence of this factor in the aetiological process (see Discussion section). The potential to prevent NTSCI is likely to be different in these cases, and the approaches to prevention required will also be quite different.

Basic and extended NTSCI data set

Both the basic and extended data sets are presented here because of the relatively small amount of detail needed to describe them. It is suggested that the classification of the aetiology of NTSCI can be performed at either a Basic or Extended level of detail Table 3.

Table 3 Summary of components in the NTSCI data set

The basic data set codes the aetiology of NTSCI using levels 1 and 2 of axis 1. The extended data set can code the aetiology through a possible further three levels, that is, levels 3–5. It is suggested that the number of levels used in any project be determined by the investigators based on the aims of the project. Researchers should list the most common aetiologies in their Table 1 of results. Aetiology categories can be left out of research reports if there are no participants with these disorders. For congenital or genetic categories, when there are small numbers in a particular group, they can be collapsed to level 1 (that is, reported simply as ‘genetic’ or ‘congenital’, if the researchers deem this appropriate. Acquired disorders should always be described, however, to level 2 at a minimum. If there are very small numbers of some disorders then these can be collapsed into an ‘other’ category, if the researchers deem this appropriate.

The extended data set, in addition, includes axis 2 coding of triggering diseases or processes indirectly responsible for the NTSCI using the ICD. This coding is only used where the axis 1 classification does not include the process or disease that had a role in the aetiology. The coding of the timeframe of the pattern of onset of NTSCI symptoms and presence of iatrogenicity are also included in the extended data set. A copy of the data collection forms for the data set appears in the Appendix (Tables A1 and A2).

Discussion

Data sets that include a classification for NTSCI have been developed. It is proposed that this should be used in all research involving this group of patients. During the development of the data sets, and particularly of the classification, it became apparent that the needs of potential users varied widely. In particular, some users wanted a detailed system while others indicated the need for a simple approach. For this reason, we have developed the classification system in such a way that it allows a degree of flexibility for potential users regarding the amount of specificity with which the aetiology is coded that we feel achieves the necessary balance.

Classification of NTSCI

In developing the classification system, it was realised that there were a number of approaches that could be utilised for classifying certain groups of conditions that constitute NTSCI, in particular, a number of conditions in the genetic and motor neurone disease groups. We recognise that a number of different classification systems for these conditions are in use. The approach taken here was to develop a classification that is believed to be the most appropriate for spinal cord medicine settings. It was also desired to develop a classification that could also be used in countries with constraints on resources, particularly regarding genetic testing. In addition, for the various causes we identified, it was decided to group conditions in a way that would help direct preventative strategies, where these are possible. Furthermore, some level 3 and 4 conditions could potentially be further sub-divided, but this was deliberately not done for those that are rare in spinal cord medicine settings.

It is suggested that consideration be given to including the basic data set for acquired NTSCI classification into the core data set in future revisions. It is also acknowledged that the extended data set classification is comprehensive and presents a challenge for data collection. It is planned, therefore, to develop a database proforma for the classification that could be made available via the ISCoS and spinal cord web sites to facilitate data collection.

Date of onset of NTSCI

Information on the date of onset of SCI is vital to an understanding of some research findings and is a component of the core data set.17 In the core data set if the exact date is not known, then the year, and if known, the month, should be recorded. In recording the date of onset for patients with NTSCI, it is proposed that the date of onset should be considered to be that date, which based on the patients’ history, was consistent with the onset of the neurological symptoms that were shown subsequently to be due to NTSCI. Rather than have a separate element in the NTSCI data set, it is suggested that the onset date in the core data set be modified to include the above proposed qualification. This suggestion will need to be considered by the data set committee of ISCoS, and included in a revision of the core data set.

Iatrogenic SCI

A major challenge in developing the data sets was dealing with iatrogenic causes of SCI. There were two schools of thought regarding this issue. One group, including us, felt that these conditions should be included in the NTSCI classification system if there was no direct external force involved. It was suggested that a prefix be used to indicate the occurrence of an iatrogenic component—either in the axis 1 classification or as a separate data field—for when iatrogenicity was a contributing factor in SCI. The other group highlighted the approach taken by the ICECI, and advocated strongly the need to follow their standard. Given that ICECI system was already being utilised in other areas of injury research, ultimately the decision was made to use the ICECI approach.

There are a number of concerns, however, that have been expressed regarding the use of the ICECI framework to classify all iatrogenic SCIs as traumatic. The following three clinical scenarios illustrate these concerns.

  1. 1)

    In patients with a tumour-causing NTSCI, with neurological symptoms and signs because of spinal cord compression (from the tumour), the spinal damage can worsen as a result of radiotherapy, chemotherapy or surgery carried out in an attempt to treat the tumour.

  2. 2)

    Patients presenting with canal stenosis-causing cord myelopathy can have spinal cord damage symptoms pre-surgery and worsening of these signs post-operatively without any direct operative trauma to the cord.

  3. 3)

    In cases of a bacterial epidural abscess-causing spinal cord compression, there may be a known incident such as an animal scratch, trauma or surgical wound that results in an infection, immune suppression for a medical condition, which is a contributing factor, or cases in which a cause for the infection is never found.

The first and second examples could be classified as iatrogenic, but this is misleading in terms of aetiology if they were considered to be traumatic. SCI trauma prevention would have no impact on the occurrence of these types of SCI. The pathological process is not traumatic in these. Furthermore, most SCI intervention trials would not include these patients as having a traumatic SCI.

In the third example, there are few if any differences in outcome between an epidural abscess case for which no cause is found; a case that occurred following immune suppression for a medical condition; or a surgical wound, a fall or animal scratch (that is, a trauma) that resulted in a known skin break, which became infected and progressed to septicaemia and a subsequent epidural abscess?

Feedback from participants in the data set development process indicated that there is a wide range in what different people consider to be traumatic or non-traumatic in the above three examples. The enormous challenge faced in developing an approach for dealing with iatrogenic SCI has resulted in the belief that it is almost impossible to develop a framework to completely standardise the classification of these conditions because of the nuances of clinical cases, the subjective nature of how clinicians interpret key events and contributing factors in the NTSCI cases, and the influence of legal and cultural factors. It was believed that it would not be appropriate to use the Delphi process to attempt a definitive consensus approach. It is recommended that, in using this data set classification, the ICECI be used as the overall guiding framework but clinicians make the final decision regarding whether they consider the SCI to be traumatic or non-traumatic. It is suggested that, when the iatrogenic component is a direct ‘cause’ involving an ‘unintentional cut, puncture, perforation during a surgical intervention (ICECI 20.4), the case should be considered a traumatic SCI. If the iatrogenic component involves medication (that is, iatrogenic but no direct external force), or is only a factor in an already established clinical case of NTSCI even if there is some progression in severity of SCI as a result of the iatrogenic component, then these should be classified as NTSCI and in the extended data set the iatrogenic component is indicated as being present.

The authors recommend that further work on refining the approach to the iatrogenic issue is needed in order to refine the classification of NTSCI (and, by implication, traumatic SCI). This process should involve key stakeholders, including the ISCoS, WHO ICECI subcommittee, traumatic SCI researchers, taxonomistis and nosologists. During the early phase of this project, including the workshop at ISCoS 2008 and the preceding round of e-mail consultation, it had been proposed to include an item in the Data Sets on whether the NTSCI was progressive or not. There were very strong and opposing opinions expressed, both in favour and against this. It was decided to omit classifying NTSCI as progressive or not from this version of the Data Set in order to assist in facilitating acceptance. The need to consider classifying NTSCI in this way has recently been highlighted.22 This is an additional issue that will need to be revisited in future revisions of these Data Sets.

In addition, a validation study of the classification system and refinement of key criteria and principles used in the classification may also be appropriate.

Limitations

Limitations to these data sets are the lack of basis is classify some22 NTSCI conditions, in particular the subjectivity involved in considering what is an iatrogenic SCI. There will be some limitations in certain settings regarding the availability of certain investigations necessary to determine the precise aetiology of SCI. This is not a limitation of the data sets per se, but it does create a restriction in settings with resource limitations on the precision with which the aetiology can be specified, and thereby use the extended data set levels.

Conclusion

It is suggested that the basic data set be used for routine data collection and the extended version for comprehensive research projects, at the discretion of the study investigators. It is proposed that these data sets be reviewed three years after publication. The major advantage of this data set is that it provides a common language for clinicians and researchers in SCI medicine. The developers hope that it will be useful for both outcomes research and preventive programs.

DATA ARCHIVING

There were no data to deposit.