Introduction

Being able to compare and aggregate data on all aspects of spinal cord injury (SCI), including its causes, pathology, and lived experience of people with SCI, is important at the service level for e.g. monitoring outcomes and planning of rehabilitation, as well as at policy level for e.g. monitoring policy changes. To ensure the comparability of data collected and to enhance the aggregation of various data sets locally, nationally, and internationally, the application of information standards is essential [1].

Existing information sources, such as databases and cohort studies, commonly applied in SCI can be grouped broadly into two groups. First, health condition-specific information standards, such as the International SCI Data Sets [2]. Second, generic or etiologically neutral information standards such as the International Classification of Functioning, Disability and Health (ICF) [3]. Health condition-specific standards allow clinicians and researchers to collect valid information that is most relevant in a given health condition, thus, enabling detailed monitoring and comparison within a health condition. In contrast, generic standards enable comparisons across health conditions, which is useful for instance in light of multi-morbidities or when comparing rehabilitation facilities with complex case mixes [4, 5]. Nevertheless, generic standards may only pick up some variation of each health condition. Hence, these standard groups and their respective standards are complementary and are ideally applied in this manner in practice.

Both, generic and health condition-specific standards can be reviewed from the perspective of their origin. Some are standards released by international organizations such as the International Classification of Diseases (ICD) or ICF of the World Health Organization (WHO) to be subsequently implemented in practice, others are developed “out of practice for practice”, such as the Model Systems in the United States. The ICF, representing an etiological neutral information standard released by WHO and endorsed by all its member states in 2001, serves as the standard for describing the functioning of people with any health condition [3]. The ICF consists of two parts, each comprising two components: First, Functioning and Disability with the components Body Functions and Body Structures, and Activities and Participation, and second the Contextual factors with the components Environmental and Personal Factors. The comprehensiveness of the ICF is its strength, and yet, it challenges the practicability in clinical practice and research. Thus, ICF generic Sets and health condition-specific ICF Core Sets have been rigorously developed to enhance the utility of the ICF for clinical practice and research [6, 7]. For SCI, ICF Core Sets for the post-acute [8] and the long-term [9] settings are available. In SCI, the ICF Core Sets have been used as a framework to set up national [10] and international cohort studies [11], and also as a reference framework for examining existing datasets. To ensure comparability across health conditions and settings, two generic ICF Sets of 7 [12] and 30 [13] ICF categories exist. These Generic Sets have been developed statistically based on data collected on all available health condition-specific ICF Core Sets.

The National Spinal Cord Injury Database (NSCID) is the longest active and largest SCI research database and the most extensive source of available information about the life course and characteristics of people with SCI. It is a prominent example for an “out of practice for practice” data collection standard and was therefore chosen as a basis for this study [14, 15]. It is part of the Spinal Cord Injury Model Systems (SCIMS) program and is in place since 1970 as a network of rehabilitation facilities supported by federal funding of the U.S. Department of Health and Human Services. The SCIMS has the goal to provide a comprehensive system of care for persons with SCI and to conduct research in order to improve the health and quality of life of persons with SCI. In 1973, the NSCID as a part of the SCIMS has been established to collect, manage, and analyze the data being generated at SCIMS sites [14, 15]. The NSCID has captured data from 29 SCIMS centers over the years, estimated to be 6% of new SCI cases that occur each year in the United States [16]. The strength of the NSCID is its longitudinal follow-up, large sample size, and geographical distribution [17].

Though each of the described data standards has a merit in its own, they ought to be complementary. Knowing the content of the NSCID covered by the ICF, in particular, the respective ICF Core Sets for Spinal Cord Injury, would be an important first step to facilitate the comparability of the NSCID with other data sets in the field of SCI. The NSCID version of 2006–2011 was already linked to the ICF [18]. Since then, the ICF Linking Rules have been revised and an updated version of the NSCID exists. Therefore, the objective of this paper was to link the content of the latest versions of the NSCID to the ICF. We specifically aim to (1) to compare the content of the current NSCID 2016–2021 version to its predecessor (NSCID 2011–2016 version), using the ICF as a neutral reference framework, and (2) to compare the content contained in the current NSCID 2016–2021 version with the relevant ICF Sets.

Methods

Study design

The ICF Linking Rules established and revised by Cieza et al. [19,20,21]. were applied to link the contents contained in the NSCID, both 2011–2016 and 2016–2021 version, to the ICF. In a first step, corresponding to aim 1, we compared the content of the current NSCID 2016–2021 version to its predecessor, using the ICF as a neutral reference framework. In a second step, corresponding to aim 2, the ICF categories identified in the current NSCID 2016–2021 version, were compared to the ICF categories of the relevant ICF Core Sets. Figure 1 outlines the methodological steps conducted in this study.

Fig. 1
figure 1

Overview of the methodology

NSCID

The data collected by the NSCID are arranged into personal data, record status, registry data, form I, and form II [14]. Form I comprises data from inpatient rehabilitation and thus is comparable to the ICF Core Sets for SCI in the post-acute context. Form II refers to sociodemographic and outcome data at follow-up, therefore comparable to the ICF SCI Core Sets for the long-term context. The NSCID is revised on a five-year basis. The two latest versions of NSCID data collection forms (2011–2016 and 2016–2021) were used in this study [22, 23].

Linking process and analysis

The meaningful concepts extracted from the items contained in the NSCID, both the 2011–2016 and the 2016–2021 versions, were linked to the ICF categories using the revised ICF Linking Rules [21]. ICF categories are the units of the classification and are hierarchically ordered. In this order, the first level refers to chapters. Each chapter contains categories at the second and third level, some also on the fourth level, whereby the granularity increases at each level. In this study, we only applied ICF categories at the second level in the linking process. Since no classification for personal factors is available, concepts in relation to demographic or person characteristics (e.g., age, attitude, coping strategy, lifestyle) were linked broadly as ‘personal factors’. In line with the ICF Linking Rules the meaningful concepts of the NSCID that were not contained in the ICF, (e.g., cause of injury), were coded as ‘not covered’. The linking was conducted by two researchers independently (RM & MS for NSCID 2011–2016, RM & KP for NSCID 2016–2021). Disagreements of linking were discussed between the two researchers. When no agreement was found, a third researcher was involved (BP). As a last step, the research team, i.e. the authors of this study, representing expertize in the area of the ICF, SCI and the NSCID reviewed the results of the linking.

In response to aim 1, we compared the content of the NSCID 2016–2021 version to the NSCID 2011–2016 version based on the identified ICF categories using an excel spread sheet. Descriptive statistics were applied to examine the differences between the two NSCID versions by comparing the ICF categories covered.

In response to aim 2, a comparison between the ICF categories covered by the NSCID current version 2016–2021 and six ICF Sets was performed. Six ICF Sets including the four existing ICF Core Sets for SCI and the two generic ICF Sets were used as a basis for comparison: the brief and the comprehensive ICF Core Sets for SCI in the post-acute context [8] and the brief and comprehensive Core Sets for SCI for in the long-term context [9]. Their validation studies confirmed that these Core Sets represent what is important for SCI patients from the perspective of different health care professionals [24, 25]. Furthermore, the two existing ICF Generic Sets, the ICF Generic-30 (also referred to as the ICF Rehabilitation Set) [13], and the ICF Generic-7 (also referred to as the ICF generic set) [12], were included.

The second level ICF categories derived from the first step, covering the NSCID contents, were compared to the ICF second level categories of the ICF generic Sets and SCI Core Sets and the percentage of ICF category coverage was calculated to give some sort of quantification. We focused specifically on the comparison of the NSCID form I that is collected during the inpatient setting with the ICF SCI Core Sets for the post-acute setting, and the comparison of the NSCID form II that is collected at the yearly follow-up visits with the ICF SCI Core Sets for the long-term setting. To provide a succinct overview, we subsequently summarized this comparison on the level of the forms, as well as all the NSCID contents not covered by the ICF.

Results

Comparison between the current NSCID version (2016–2012) and the prior NSCID version (2011–2016)

The variables from the NSCID 2011–2016 were linked to 54 ICF categories at the second level, and the variables from the NSCID 2016–2021 version were linked to 62 ICF categories as shown in Table 1. The current NSCID version contains 11 ICF categories more than its predecessor does: 5 in the component of Body functions and 6 in the component of Activities and Participation (represented in bold in Table 1), and 3 categories less in the component of Body functions, resulting in a total addition of 8 ICF categories.

Table 1 Comparison of the NSCID 2011–2016 and the NSCID 2016–2021 versions using the ICF as a neutral reference

Comparison of the NSCID 2016–2021 version with the relevant ICF Core Sets

Table 2 presents the coverage of the relevant ICF Core Set by the NSCID 2016–2021 version. Overall, the NSCID 2016–2021 variables covered more than 70% of the categories in the two ICF generic sets and more than 55% of the categories in the two Brief ICF Core Sets for SCI. However, the NSCID coverage of the two Comprehensive ICF Core Sets for SCI was low with less than 40%, especially in the post-acute context. For instance, the NSCID form I covers only 26% of the related post-acute Comprehensive ICF Core Set. A detailed linking table showing the detailed linking of all NSCID items to the respective ICF Core Sets is provided in Supplementary Appendix 1. Those items that could not be linked to the ICF are outlined in Supplementary Appendix 2.

Table 2 Coverage of the ICF categories from the NSCID 2016–2011 when mapping with the predefined ICF sets

The following 14 ICF categories identified in the NSCID 2016–2021 version were not included in any of the ICF Core Sets: b140 Attention functions, b147 Psychomotor functions, b199 Mental functions, unspecified, d175 Solving problems, d310 Communicating with-receiving-spoken messages, d315 Communicating with-receiving-nonverbal messages, d335 Producing nonverbal messages, d398 Communication, other specified, d449 Carrying, moving and handling objects, other specified and unspecified, d499 Mobility, unspecified, d730 Relating with strangers, d740 Formal relationships, d855 Non-remunerative employment, and e560 Media services, systems, policies.

Discussion

This study aimed to compare the content of the current NSCID 2016–2021 version to its predecessor based on the ICF as a neutral reference framework, and to identify the coverage of ICF categories of relevant ICF sets for SCI. These standards-the NSCID data model and the ICF-represent a health condition “out of practice for practice” standard and a generic, international standard in the field of health care. This study makes the NSCID comparable with other data sets based on the ICF as a neutral reference. At the same time, the items identified in this study may serve as the foundation for operationalizing the ICF categories contained in the ICF through the NSCID.

By extracting and linking all outcome variables reported in the NSCID to the ICF, the present result showed that the latest NSCID 2016–2021 version covered a broader spectrum of functioning than the prior NSCID 2011–2016. In particular, items were changed related to chapter b1 Mental functions, including items on sleep, attention, and psychomotor functions, as well as d4 Mobility, including items referring to handling objects, hand and arm use. From the perspective of the ICF, these additions reflect changes toward the realization of a biopsychosocial model of functioning informing the NSCID. By linking both NSCID versions they become comparable based on the ICF as a neutral and international reference. Furthermore, the linking and comparison to older NSCID versions but also other relevant SCI datasets could potentially inform future developments of the NSCID.

With regard to strengthening the relevant and internationally comparable data on functioning information as well as data aggregation from various databases of SCI, the second aim of this comparison study used the ICF Core Sets to specify what is important to document to gain a comprehensive picture of functioning in persons with SCI. The comparison of the ICF categories included in the NSCID 2016–2021 and the respective ICF Sets (Table 2) revealed that the NSCID covers most categories of the ICF Generic Sets (ICF Generic 7 and 30), which represent the minimum standard of functioning across clinical populations and across settings. The NSCID furthermore covers more than 50% of ICF categories of the Brief ICF Core Sets for SCI. The categories contained in the NSCID had, however, less correspondence with the Comprehensive ICF Core Sets for SCI both in post-acute and long-term settings. The detailed analysis revealed that in particular items related to Body Structures are hardly included in the NSCID.

At the same time, 14 ICF categories were identified in the NSCID that were not included in any ICF Core Set. These ICF categories refer to ICF categories of the chapter b1 Mental functions, 6 ICF categories of chapters d3 Communication, d4 Mobility, d7 Interpersonal interactions and relationships and d8 Major life areas, as well as e5 Services, systems, and policies. Since these categories are missing in the ICF Core Sets, it would be valuable to further review to what extent such categories, represented in a well-established SCI dataset, should be recommended to complement the existing ICF Core Sets. At the same time, it is noteworthy that particularly in the post-acute context, the NSCID revealed a lack of items in some of the following chapters, e.g. d7 Interpersonal interactions and relationships to d9 Community, social and civic life, and e3 Support and relationships and e4 Attitudes. It is only in the Body functions and selected Activity and Participation chapters—d4 Mobility and d5 Self-care—where the number of categories identified was almost comparable. In the long-term context, the comparison revealed more overlap across all Activity and Participation chapters; also, the Environmental Factors are better represented in the NSCID, though the ICF Core Sets contain consistently more ICF categories across all Environmental Factors chapters.

To date, there are other existing SCI databases built upon the conceptual model of the ICF, such as the Rick Hansen Spinal Cord Injury Registry [26, 27], International Spinal Cord Injury Data Sets [28], and the Swiss Spinal Cord Injury Cohort Study [10]. Studies using these databases have contributed greatly to the understanding of the epidemiology of functioning in SCI. To ensure that various datasets can be aggregated, compared and contrasted, all data needs to be reported in a standardized manner [29,30,31]. The comparison presented in this study exemplifies the process toward making existing data sets comparable by using the ICF as a reference.

An important limitation of this study is that the personal factors are not yet classified in the ICF. Thus, aspects related to the particular background of an individual’s life and living, such as social and educational background, ethnicity, thoughts and beliefs, etc. could not be compared and presented in our study even though these factors are collected in most datasets and seem to play a role in disability at any level. Nevertheless, personal factors are an important interacting factor in the understanding of functioning from a comprehensive perspective. Furthermore, depending on the level of granularity needed the ICF linking methodology comes with limitations. It can be difficult to directly compare two items linked to the same ICF codes. To address this issue, we have applied the refined Linking Rules, which imply that the perspective adopted in the item is considered in the linking process, as well as a differentiation between a main and additional concept of an item. Moreover, in order to be able to fully aggregate data and conduct quantitative comparisons a number of analytical steps beyond linking the content of a dataset to the ICF are required [21]. More specifically, the ICF linking process serves as the foundation of comparing the content of different data sets based on the ICF. The results of this process, however, do not provide any information about the score or metric equivalence of different items linked to a given ICF category. To establish such metric equivalence, the application of psychometric methods is needed [32]. Furthermore, ICF categories have not been developed as operational variables or items. Thus, further specification of an ICF category is needed when moving from classification, e.g. an ICF category, to measurement, e.g. an item of an instrument related to or based on an ICF category [33].

In conclusion, the recent version of the NSCID 2016–2021 covers functioning as classified in the ICF more comprehensively than its predecessor. Differences in coverage of the different functioning components identified in the NSCID 2016–20121 in comparison with relevant ICF Core Sets still exist. This study emphasizes how the ICF can serve as a reference framework to compare existing data sets from both clinical practice and research, seen as a first step towards data comparability and aggregation.

Data archiving

The materials used in this study are all public domain. Detailed linking sheets can be requested by the authors.