Introduction
Persons sustaining a spinal cord injury or disorder (SCI/D) are at increased risk to experience reduced quality of life (QoL) [
1]. SCI/D is a severe impairment caused by physical trauma or disease [
2] and the corresponding damage to the neurological tissue typically leads to an immediate, sometimes recovering, sometimes permanent complete or partial loss of body functions, mainly sensor motor and autonomic nerve function below the lesion level [
3]. These functional impairments and the risk for developing various secondary health conditions such as chronic pain, fatigue, or muscle spasms [
4] as well as resulting restrictions in daily activities and social participation [
5] may additionally negatively affect QoL in persons with SCI/D. Thus, besides improving physical functioning and fostering community reintegration, increasing QoL is a key goal of first inpatient rehabilitation following SCI/D [
6‐
8].
Despite general agreement on the clinical relevance of QoL, there is no consensus on the definition of QoL within the SCI/D literature and as a result various measures have been used to assess it [
6]. This limits comparability of findings from different studies and creates difficulties to draw any firm conclusions [
9]. Therefore, an international group of experts developed the international spinal cord injury quality of life basic data set (SCI-QoL-BDS) [
10] with the goal of standardizing the collection and reporting of QoL data in individuals with SCI/D. In addition, the measure aimed to be brief by assessing only a minimal amount of information to facilitate implementation in daily clinical practice. A brief QoL instrument is indeed especially useful in the inpatient rehabilitation phase to efficiently evaluate the evolution of patients’ QoL and the effectiveness of care. The SCI-QoL-BDS is based on a definition of QoL as individual’s subjective evaluation of how things are in their life [
8]. This is operationalized with three items asking respondents to indicate to what degree they are satisfied with their life as a whole, their physical health, and their psychological health. Though the authors acknowledged that perceived QoL is a multifaceted construct covering also other domains, these were considered as the most relevant ones in the SCI/D context [
10].
Since its development in 2012, the SCI-QoL-BDS has been employed by different research groups around the globe. Accordingly, the original English version was translated into other languages including Dutch [
11], Brazilian Portuguese [
12], German, French, Italian [
13], and Thai [
14]. Preliminary evidence from these studies revealed promising psychometric properties of the instrument as indicated by good internal consistency, convergent, and divergent validity among individuals with SCI/D during inpatient rehabilitation [
15‐
17] as well as among community-dwelling individuals [
11,
14,
15,
18]. Two longitudinal studies further demonstrated acceptable to good test–retest reliability across a two-week interval [
14,
18]. Nevertheless, more examination of the psychometrics properties of the SCI-QoL-BDS during first inpatient rehabilitation are still needed.
Aiming to extend previous research on the psychometric properties of the SCI-QoL-BDS, the objective of the present study was to test the SCI-QoL-BDS’s internal consistency and longitudinal measurement invariance from admission to discharge from SCI/D first inpatient rehabilitation. Testing longitudinal measurement invariance allows to determine whether the instrument assesses the same construct on the same metric at different points in time [
19]. This is an important aspect for determining the instrument’s validity, consistency and a fundamental prerequisite to calculate change in the QoL construct and compare its structural relationships with other constructs over time [
19,
20]. More specifically, if measurement invariance is not achieved, this indicates that respondents interpret the specific questions and/ or the underlying construct differently at different points in time [
21]. As such, changes in scores over time do not necessarily represent quantitative differences in the construct itself. Instead, they may be the result of changes in the meaning of the construct over time (i.e., response shift) or they might be caused by different response styles over time [
22,
23]. In sum, examining longitudinal measurement invariance of the SCI-QoL-BDS across first inpatient rehabilitation lays the ground for future research and clinical practice efforts aiming to evaluate the success of rehabilitation practices by measuring changes in the QoL of individuals with SCI/D from admission to discharge.
Results
Descriptive statistics of the study participants are depicted in Table
1. To determine the representativeness of the study sample, participants were compared to non-participants with available data (Supplementary Table 1). No significant differences emerged with regard to physical functioning at T1, level and completeness of the injury. In contrast, participants included in the present study were significantly more likely male, younger, having a traumatic cause of the injury and a longer time to discharge from inpatient rehabilitation. However, the effect sizes in these comparisons were small (Cohen’s
d between 0.07 and 0.27; Cramer’s
V between 0.06 and 0.08) indicating only a slight selection bias in the study sample.
Table 1
Descriptive statistics of the study sample (n = 218)
Age at injury | | 0 (0.0) | 51.8 | 17.1 |
Sex | | 0 (0.0) | | |
Male | 162 (74.3) | | | |
Female | 56 (25.7) | | | |
Language | | 0 (0.0) | | |
German | 177 (81.2) | | | |
French | 37 (17.0) | | | |
Italian | 1 (0.5) | | | |
Other | 3 (1.4) | | | |
Time from injury to T1 (days) | | 10 (4.6) | 36.4 | 6.7 |
Time from injury to T2 (days) | | 5 (2.3) | 83.1 | 7.8 |
Time from injury to T4 (days) | | 8 (3.7) | 174.7 | 58.6 |
Cause of SCI/D | | 0 (0.0) | | |
Traumatic | 142 (65.1) | | | |
Non-traumatic | 76 (34.9) | | | |
Level of injury at T1 | | 8 (3.7) | | |
Paraplegia | 135 (61.9) | | | |
Tetraplegia | 72 (33.0) | | | |
Unable to determine | 3 (1.4) | | | |
Completeness of injury (AIS) at T1 | | 9 (4.1) | | |
A | 48 (22.0) | | | |
B | 28 (12.8) | | | |
C | 29 (13.3) | | | |
D | 103 (47.3) | | | |
E | 0 (0.0) | | | |
Unable to determine | 1 (0.5) | | | |
SCIM at T1 | | 3 (1.4) | 36.7 | 19.3 |
SCIM at T4 | | 5 (2.3) | 71.7 | 20.5 |
Descriptive statistics of the SCI-QoL-BDS item and total scores are displayed in Table
2. The distributions of the item scores and of the total score were slightly non-normal (Skewness between -0.6 and 0.2; Kurtosis between 2.2 and 2.7) at each of the three measurement time points. Overall, item and total scores tended to increase between inpatient rehabilitation admission and discharge. T-test comparing the SCI-QoL-BDS item and total scores between German-speaking participants and participants speaking other languages showed no statistical difference at any of the three time point.
Table 2
Descriptive statistics and internal consistency of the SCI-QoL-BDS items and total score (n = 218)
Satisfaction with life | | | | | | | | |
T1 | .74 | – | – | 5.2 [4.8; 5.6] | 2.7 | 0.1 | 2.2 | 0–10 |
T2 | .82 | – | – | 5.6 [5.3; 5.9] | 2.5 | − 0.2 | 2.3 | 0–10 |
T4 | .83 | – | – | 6.5 [6.2; 6.8] | 2.5 | − 0.6 | 3.0 | 0–10 |
Satisfaction with physical health | | | | | | | | |
T1 | .68 | – | – | 4.3 [3.9; 4.6] | 2.6 | 0.2 | 2.4 | 0–10 |
T2 | .73 | – | – | 5.3 [5.0; 5.7] | 2.6 | − 0.2 | 2.2 | 0–10 |
T4 | .82 | – | – | 5.9 [5.6; 6.2] | 2.4 | − 0.5 | 2.7 | 0–10 |
Satisfaction with psychological health | | | | | | | | |
T1 | .63 | – | – | 6.3 [5.9; 6.6] | 2.5 | − 0.5 | 2.5 | 0–10 |
T2 | .71 | – | – | 6.4 [6.1; 6.7] | 2.6 | − 0.6 | 2.4 | 0–10 |
T4 | .72 | – | – | 6.9 [6.6; 7.2] | 2.4 | − 0.6 | 2.7 | 0–10 |
SCI-QOL-BDS total score | | | | | | | | |
T1 | – | .83 | .82 | 5.2 [4.9; 5.5] | 2.2 | − 0.3 | 2.4 | 0–10 |
T2 | – | .88 | .87 | 5.8 [5.5; 6.1] | 2.3 | − 0.3 | 2.5 | 0–10 |
T4 | – | .90 | .89 | 6.4 [6.2; 6.7] | 2.1 | − 0.5 | 2.8 | 0–10 |
Internal consistency
Both McDonald’s Omega (between .83 and .90) and Cronbach’s alpha (between .82 and .89) of the SCI-QoL-BDS total score were good at each of the three measurement time points. Item-total correlations were at least
r = .63 and therefore in the acceptable range (Table
2). Pearson correlations among the SCI-QoL-BDS items are depicted in Table
3. They are exclusively in the moderate to large range [
35]. In particular the correlations among the three SCI-QoL-BDS items measured at a specific time point were large.
Table 3
Correlations among the SCI-QoL-BDS items (n = 218)
1. Satisfaction with life at T1 | – | | | | | | | |
2. Satisfaction with physical health at T1 | .68 | – | | | | | | |
3. Satisfaction with psychological health at T1 | .61 | .53 | – | | | | | |
4. Satisfaction with life at T2 | .48 | .49 | .50 | – | | | | |
5. Satisfaction with physical health at T2 | .45 | .58 | .46 | .76 | – | | | |
6. Satisfaction with psychological health at T2 | .38 | .34 | .51 | .73 | .61 | – | | |
7. Satisfaction with life at T4 | .49 | .50 | .47 | .58 | .54 | .44 | – | |
8. Satisfaction with physical health at T4 | .42 | .45 | .41 | .54 | .55 | .45 | .82 | – |
9. Satisfaction with psychological health at T4 | .37 | .30 | .49 | .58 | .45 | .58 | .70 | .69 |
Longitudinal measurement invariance
The fit of the nested models testing different aspects of longitudinal measurement invariance of the SCI-QoL-BDS are shown in Table
4. Except for the significant
χ2, the fit of the configural invariance model tested in the first step was good. All items loaded significantly (
p’s < .001) on the latent QoL factor at each of the three measurement time points. The corresponding standardized factor loadings were large [
35] and ranged between .72 (satisfaction with psychological health at T1) and .95 (satisfaction with life at T2). Taken together, these results indicate that there is configural invariance across inpatient rehabilitation.
Table 4
Longitudinal measurement invariance of the SCI-QoL-BDS (n = 218)
1 | Configural invariance | 25.31 | 15 | .046 | .989 | .056 [.014; .091] | .027 | – | – | – | – | – | Accept |
2 | Loading invariance | 27.71 | 19 | .089 | .991 | .046 [.000; .079] | .029 | 2 vs 1 | 1.79 | 4 | .774 | .002 | Accept |
3 | Intercept invariance | 57.52 | 23 | < .001 | .964 | .083 [.057; .109] | .047 | 3 vs 2 | 31.37 | 4 | < .001 | − .027 | Reject |
4 | Partial intercept invariancea | 34.39 | 22 | .045 | .987 | .051 [.012; .081] | .034 | 4 vs 2 | 6.77 | 3 | .080 | − .004 | Accept |
After having established configural invariance, we tested the loading invariance model. In this second step, adding the equality constraints on the factor loadings over time did not significantly worsen model fit, as indicated by a scaled Δ
χ2(4) = 1.79,
p = .774 and ΔCFI = .002. This suggests full loading invariance of the measure over time. We then proceeded with testing the intercept invariance model. As can be seen in Table
4, the intercept invariance model showed a significantly worse fit than the loading invariance model with a scaled Δ
χ2(4) = 31.37,
p < .001 and ΔCFI = − .027. This indicates that full intercept invariance was not achieve. To identify the source of misfit, we examined the freely estimated intercepts of all items more closely. The largest discrepancy in the unstandardized intercepts emerged for the satisfaction with physical health item at T1 (4.26), which was substantially lower than the ones at later occasions (T2: 5.31; T4: 5.93). Hence, we ran a partial intercept invariance model, releasing the equality constraint on the T1 intercept of the satisfaction with physical health item. This model achieved a fit which was not significantly worse than the one of the loading invariance model, as indicated by a scaled Δ
χ2(3) = 6.77,
p = .080 and ΔCFI = − .004. Hence, the variability in the measurement of the latent QoL construct can be attributed to the instability of the intercept of the satisfaction with physical health item (see Supplementary Table 2 for the parameter estimates of the partial intercept invariance model).
Discussion
The present study was the first to examine the internal consistency and different aspects of longitudinal measurement invariance of the SCI-QoL-BDS during first SCI/D inpatient rehabilitation. Supporting results from previous research [
11,
14‐
18], we found good internal consistency of the measure administered at one and three months after injury and at discharge from first inpatient rehabilitation. Using longitudinal factor analyses, we additionally demonstrated that the factorial structure and the factor loadings of the measure were invariant during first inpatient rehabilitation. This suggests that the three items of the SCI-QoL-BDS equally represent a latent QoL construct and that the meaning of this latent QoL factor seems to be stable across time [
36,
37]. However, we also found some reasons for caution when using the instrument since only two out of the three items additionally proved to have invariant intercepts over time.
The item which was non-invariant at the intercept level is the respondents’ satisfaction with physical health. For this item, the intercept at one month post injury had to be freely estimated because it was substantially lower than at the two later measurement occasions. This means that the position on the latent QoL construct does not equally transfer to the observed level on this item over time. In other words, there is a systematic tendency for individuals with SCI/D to indicate lower satisfaction with physical health at one month post injury as compared to later measurement occasions and this tendency is not attributable to the concurrent position on the latent QoL variable [
36,
37]. This might have resulted from a change in (most of) the respondents internal standards of measurement, so-called recalibration response shift [
23,
38]. Individuals might have “recalibrated” their interpretation of the response options for the satisfaction with physical health item over the course of SCI/D inpatient rehabilitation. For example, at the beginning of inpatient rehabilitation, being
totally satisfied with physical health might have required for individuals to have full physical functioning including the ability to walk and the physical health status before SCI/D onset might be used as reference framework to appraise satisfaction. At later stages of inpatient rehabilitation, individuals may have gained a better understanding of the primary physical consequences of the SCI/D as well as secondary health conditions and thus establish a new reference framework to evaluate their health. At this stage, positive responses to the item
total satisfaction with physical health may reflect appreciation of absence of secondary health conditions and gains in physical well-being and functioning achieved during inpatient rehabilitation. Further longitudinal research, particularly qualitative studies examining the temporal differences in how individuals with SCI/D evaluate their satisfaction with physical health are needed to understand this potential response shift across inpatient rehabilitation.
As a side note, reevaluating one’s values and criteria for what constitutes good QoL can be an adaptive response to a life-changing event such as SCI/D. Studies of individuals who experienced cancer, loss, or other potentially traumatic events have documented changes in sense of self [
39], appreciation of life [
40], and life priorities [
41]. This raises the possibility that the partial invariance observed in the present study actually might indicate good validity in the sense of reflecting a dynamic cognitive process rather than a static self-assessment, which would not capture the complexity of the psychological adaptation to potentially traumatic events such as the onset of an SCI/D.
The finding of partial intercept invariance has important methodological implications for future longitudinal studies. The response shift seems to take place in the very early phase of inpatient rehabilitation and intercept invariance can be observed in the two later time points. This indicates that the SCI-QoL-BDS might be used without longitudinal measurement issues in studies focusing on the later stages of inpatient rehabilitation spanning from the third month post SCI/D to rehabilitation discharge. Moreover, establishing partial intercept invariance is sufficient to allow for a meaningful interpretation of differences in the latent QoL factor means and their structural relations to other constructs across time [
19,
21,
29]. Nevertheless, caution is needed when calculating and interpreting changes in the observed SCI-QoL-BDS total score (i.e., mean score) and in particular in the satisfaction with physical health during the early phase post SCI/D. Comparing observed means would require full intercept invariance [
37]. Consequently, a latent variable framework such as latent change score models [
42] may be best suited for examining the longitudinal course of QoL and its relationship with other constructs during the early rehabilitation phase [
42].
Limitations
The present study is subject to several limitations. First, it should be noted that we examined the longitudinal measurement invariance of the SCI-QoL-BDS across the inpatient rehabilitation period of individuals with SCI/D. Therefore, it remains unclear whether similar results would emerge when including also measurement time points in the community setting. Hence, future studies are needed to examine the longitudinal invariance of the SCI-QoL-BDS across the whole life span of individuals with SCI/D.
Second, the post hoc investigation of partial invariance at the intercept level was a data-driven approach. As such, it is subject to capitalization on chance [
43]. Thus, the present findings should be replicated in future studies with different samples to increase confidence.
Third, a comparison of the characteristics of participants and non-participants with available data indicated a minor selection bias in the present study’s sample. However, individuals who completely refused data collection could not be compared to participants. Hence, some uncertainty regarding the representativeness of the present study sample remains and findings should therefore be interpreted cautiously.
Fourth, a sample size of
N = 200 is considered to be sufficient for running structural equation models [
44]. Nonetheless, with an
N = 218 our sample size can be considered as modest. Hence, we might have lacked the power to detect weak violations of measurement invariance. As such, future studies with a larger sample size are required to validate the present findings.
Conclusion
Brief QoL instruments with good psychometrics properties are dearly needed in inpatient rehabilitation settings for an efficient evaluation of the care provided and the recovery of patients. In general, the present study revealed preliminary evidence that the SCI-QoL-BDS is a consistent and valid measure to assess QoL among individuals with SCI/D in clinical research and practice focusing on the inpatient rehabilitation setting. However, our results revealed that the measure might not be fully invariant at the intercept level indicating a recalibration response shift with satisfaction with physical health being comparatively evaluated more negatively in the early phase of SCI/D inpatient rehabilitation. Consequently, we recommend using latent variable frameworks instead of mean scores when examining longitudinal changes from the early stage of SCI/D inpatient rehabilitation to discharge.
Acknowledgements
We thank the SwiSCI Steering Committee with its members Xavier Jordan, Fabienne Reynard (Clinique Romande de Réadaptation, Sion); Michael Baumberger, Hans Peter Gmünder (Swiss Paraplegic Center, Nottwil); Armin Curt, Martin Schubert (University Clinic Balgrist, Zürich); Margret Hund-Georgiadis, NN (REHAB Basel, Basel); Laurent Prince (Swiss Paraplegic Association, Nottwil); Heidi Hanselmann (Swiss Paraplegic Foundation, Nottwil); Daniel Joggi (Representative of persons with SCI); Mirjana Bosnjakovic (Parahelp, Nottwil); Mirjam Brach, Gerold Stucki (Swiss Paraplegic Research, Nottwil); Armin Gemperli (SwiSCI Coordination Group at Swiss Paraplegic Research, Nottwil).
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.