Introduction

Longstanding pain is a major complication after spinal cord injury (SCI) affecting approximately 80% of patients,1 which is associated with lower levels of psychological well-being2, 3 and decreased daily function.4 Therefore, a reduction in pain’s effects on functioning is an important goal of all pain interventions. Consequently, a comprehensive pain assessment taking multiple aspects of the psychosocial impact into account is integral to designing optimal treatments.

The West Haven-Yale Multidimensional Pain Inventory is a self-report questionnaire measuring the impact of pain on an individual’s life, how others respond to that person’s pain and frequency at which the individual engages in specific daily activities.5 The MPI has been used in numerous pain populations and has been found to have good psychometric properties including sensitivity to a variety of treatments. The IMMPACT group6 has recommended this instrument for the assessment of individuals suffering from chronic pain and as an outcome measure in clinical trials.

In the original Spanish validation of the West Haven-Yale Multidimensional Pain Inventory,7 the authors also concluded that the Spanish MPI was acceptable to measure important domains related to chronic pain such as perceptions of impact of pain on daily life, social support, self-control and activity levels. However, that version7 was developed for Spanish chronic pain patients in general, and not for individuals with chronic pain and physical impairments such as SCI. Thus, it is inappropriate to assume that measures developed to be used with other chronic pain populations can be readily used in people with SCI.

Based on exploratory and confirmatory factor analyses, Widerström-Noga and colleagues revised the MPI for use in the SCI chronic pain population.4 The reliability and validity of the MPI-SCI for most subscales were later demonstrated in a sample of individuals with SCI and chronic pain.2, 8 Despite the widespread use of the MPI in clinical pain practice in Spain to assess pain impact,7 the psychometric properties of a Spanish version of the MPI-SCI (MPI-SCI-S) have not yet been evaluated. The primary purposes of the present investigation were to: (1) confirm the factor structure of the MPI-SCI-S; (2) test its internal consistency, and (3) construct validity.

Materials and methods

Individuals who received an annual assessment at the outpatient SCI clinic (April 2005–July 2007) were informed about the study. Those who agreed to participate were given a questionnaire package including the MPI-SCI-S (Appendix) that was sent back by mail. Demographic and injury information was collected from patient’s medical records. Participants were: (1) over 18 years old, (2) had chronic pain (>1 year), (3) chronic SCI (>2 years) and (4) average pain intensity of three or more on a Numerical Rating Scale (NRS). The Ethics Committee of the Hospital of Neurorehabilitation Institut Guttmann approved the study.

MPI-SCI

The MPI is a 60-item questionnaire5 based on the cognitive-behavioral perspective on chronic pain answered on a 7-point Likert scale. It comprises Section 1 (pain impact), Section 2 (responses by significant others) and Section 3 (common activities) with subscales assessing pain severity, pain interference, affective distress, control over life, support from significant others, responses by significant others (negative, distracting and solicitous responses) and the performance of common, general activities (Table 1). The MPI-SCI2, 8 is a modified version of the MPI developed to be used in persons with SCI where Section 3 asks about pain-specific interference.

Table 1 Scales, subscales used in the present study and their abbreviations

Translation of the MPI-SCI

The development of the MPI-SCI-S and evaluation of its psychometric properties were performed according to recommendations for adaptation and validity of health questionnaires and diagnostic tests.9 The original English version of the MPI-SCI was translated by a co-author (Yenisel Cruz-Almeida). The translation was reviewed by three experts including two specialists in pain management and a clinical pain researcher. As the original version of the questionnaire was well defined and structured, the expert panel did not consider it necessary to redefine its sections or reformulate any of the original questions. No cultural bias that could be equivocal or non-translatable was detected in the original instrument. This intermediate version was then tested in a sample of seven patients to assess initial feasibility and other potential comprehension problems. The final version was back-translated into English by two other professional translators (different from the first translator and English natives) and again reviewed and approved by the panel experts. The MPI-SCI-S is presented in the Appendix.

NRS

Participants rated their average pain intensity during the past week on a 0–10 NRS, with anchors 0 (no pain) and 10 (pain as bad as could be). The NRS10 was recommended by the IMMPACT group for use in pain clinical trials6 and by the 2006 NIDRR SCI Pain outcome measures group.11

Brief Pain Inventory (BPI) interference

The 12-item subscale measures the interference with general activity, sleep, mood and enjoyment of life, walking ability, ability to work and perform daily tasks, and relationship with other people. The BPI was adapted for people with physical impairments and SCI, and it has shown excellent psychometric properties in this population.12

Beck Depression Inventory (BDI)

The BDI is a 21-item scale measuring symptoms indicative of clinical depression. The measure is considered to be reliable in the SCI population.8, 13

Multidimensional Health Locus of Control (MHLC)

The MHLC14 consists of three subscales: (1) the internal health locus of control subscale that assesses the extent to which one believes that internal factors are responsible for health and illness; (2) the chance health locus of control (CHLC) subscale that assesses the extent to which one believes that health and illness are a matter of fate, luck or chance; and (3) the powerful other health locus of control subscale assessing the belief that one’s health is determined by powerful others. Previous research has supported its use in SCI.14

Functional Independence Measure (FIM)

The FIM15 quantifies severity of activity limitation by assessing performance in six areas: self-care, locomotion, mobility, sphincter control, communication and cognition. In the current study, only FIM scores related to motor independence were analyzed. This subscale has shown excellent internal consistency8 and can be administered in-person or via telephone format.16

Duke-UNC

The Duke-UNC Functional Social Support Questionnaire17 is a self-administered instrument designed for use in primary care settings. It measures two components of perceived emotional support: confidant and affective support. Moderate-to-excellent reliability and validity of the scale are supported by a previous study in Spain.17

Psychological Global Well-being Index (PGWBI)

The PGWBI was developed to measure subjective psychological well-being or distress in the general population. The Spanish version of the PGWBI has shown satisfactory psychometric properties.18 The questionnaire contains 22 items grouped into six dimensions, but for the present study the ‘positive well-being’ dimension was used for analyses.

Statistical analysis

Using SPSS 20.0, Pearson correlations and paired t-tests were used for continuous variables and χ2 tests were used for dichotomous variables. All tests were two-tailed and a P-value less than 0.05 was considered statistically significant. Cronbach’s alpha correlations were used to assess internal reliability. To assess the ability of the MPI-SCI-S to predict positive well-being, two separate stepwise multiple regression analyses were performed with positive well-being as the dependent variable. In order to confirm the factor structure of the MPI-SCI-S, a confirmatory factor analysis (CFA) was performed for each subsection of the MPI-SCI-S (that is, pain impact, interpersonal support and activities). The CFA was conducted using analysis of moment structures (AMOS)19 as previously described.8

Results

Participants

The study postal packages containing consent forms and questionnaires were given to a total of 558 subjects with a 22.6% response rate (n=126). Detailed demographic and injury-related characteristics are presented in Table 2. No significant differences were found between responders and non-responders with the exception of educational level.

Table 2 Demographic and injury characteristics of participants with chronic pain duration greater than 6 months who were invited to participate in the study (n=558)

Reliability internal consistency

The Cronbach’s alpha of the MPI subscales averaged 0.81 and ranged from 0.66 (LC) to 0.94 (LI). The validation instruments displayed coefficients ranging from 0.61 (internal health locus of control) to 0.92 (BPI; Table 3).

Table 3 Internal consistencies of the MPI-SCI subscales and validation instruments

Convergent validity

All subscales, except the NR and the SR, were strongly correlated with the hypothesized-related construct (Table 4). The PS subscale was highly (r=0.67) correlated with the NRS, whereas LI was strongly (r=0.75) correlated with the BPI. Although the S (r=0.36) and DR subscales (r=0.35, P<0.001) were significantly correlated with the Duke-UNC, the NR and the SR subscales were not significantly correlated with the Duke-UNC.

Table 4 Construct validity of the MPI-SCI subscales and validation instruments

Discriminant validity

To examine discriminant validity, the LC, S, DR, NR and the SR subscales were compared with the MHLC chance orientation, whereas all other MPI subscales were compared with the powerful other orientation of the MHLC, a construct hypothesized to correlate only moderately or minimally with the MPI subscales. There were trivial correlations between the MPI subscales and the MHLC (Table 4).

Predictive validity

To examine the ability of the MPI-SCI-S to predict a person’s perception of positive well-being, all MPI-SCI-S subscales were entered as independent variables in a stepwise multiple regression analysis with the well-being subscale of the PGWB score as the dependent variable (Table 5). High levels of S (P<0.01), low levels of AD (P<0.001) and a high degree of GA (P<0.01) were significantly (P<0.001) associated with higher scores on the well-being subscale of the PGWB. Similarly, when all the validation measures were entered in a second regression, overall perception of well-being was significantly (P<0.001) predicted by low scores on the BDI (P<0.01), and higher scores on the Duke-UNC (P<0.01) (Table 5).

Table 5 Stepwise regression analysis predicting a person’s perception of well-being

CFA

In order to assess the fit of the hypothesized model in each section of the MPI, fit indices greater than 0.75 were deemed appropriate similar to criteria used in previous studies using the MPI-SCI.6, 8 All indexes supported adequate fit of the hypothesized models in Section 1 (NFI=0.81, CFI=0. 89) and Section 2 (NFI=0.77, CFI=0.86). However, fit indices of the 18 items in Section 3 suggested that the model could be significantly improved (NFI=0.72, CFI=0.73). After re-inspecting the data, four items did not apply to many participants. These were: ‘How often do you mow the lawn?’ (17.4%); ‘How often do you work in the garden?’ (31.4%), ‘How often do you wash the car?’ (60%) and ‘How often do you work on the car?’ (60%). Therefore, these items were removed to reassess model fit within Section 3 and the new model indices supported an improved and adequate fit (NFI=0.88, CFI=0.89).

Discussion

The results of the present study suggest that the MPI-SCI-S is a reliable and valid measure for use in the Spanish SCI chronic pain population with the exception of the Negative and Solicitous responses subscales. The subscales of the MPI-SCI-S demonstrated acceptable reliability coefficients (0.66 to 0.94). High Cronbach’s alpha coefficients indicate that the items of the MPI-SCI-S are consistent in the domains they measure. Coefficients below 0.60 indicate inadequate reliability, and coefficients greater than 90 indicate excellent reliability useful for making individual treatment decisions. Our results are also similar to those obtained for the original MPI-SCI, which were reported to be consistently greater than 0.60.8

The present study also demonstrated that the MPI-SCI-S has acceptable construct validity across the pain intensity, pain interference, locus of control, social support and functional independence domains with the exception of the negative responses and solicitous responses. Unlike reliability, it is uncommon for a correlation (that is, validity) coefficient to be greater than 0.50, and rarely exceeding 0.50. Moreover, a recent review of depression and anxiety measures in the SCI population,13 used the following criteria for validity coefficients: (1) excellent (0.60); (2) adequate (0.30–0.59); and (3) ‘poor’ (0.29). According to these criteria the PS and LI subscales had excellent validity, the LC, AD, S, DR, GA and PA had adequate validity, whereas the NR and SR subscales had poor validity. The poor validity coefficients for the NR and SR subscales using the Duke-UNC scale might be related to the wording of the items. Although the MPI-SCI-S significant other subscales ask specifically about the perceptions of the person who suffers from pain regarding social support from one person identified as the ‘significant other’, the Duke-UNC items are concerned with the perceived social support network. It is also possible that this result indicates cultural and socio-demographic differences between Spanish and American people. For example, in the original version developed by Widerström-Noga,8 only 31% of the subjects were married, whereas the marriage frequency was doubled (62%) in our sample. Having high levels of social support does not necessarily imply high levels of support from spouses or significant others or conversely having high levels of support from significant others does not guarantee high levels of social support.

The present results also support the discriminant validity of the MPI-SCI-S subscales. It was hypothesized that the internal health locus of control (IHLOC) would correlate more highly with a similar construct, namely, life control, and lower with the less related subscales of the MPI-SCI. Consistent with previous research we found only minimal to no relationships between MPI-SCI-S subscales and the MHLC.8

The CFA of the activity subscales of the MPI-SCI-S suggested that several items had to be removed to improve the factor structure. In particular, items infrequently endorsed, such as, activities involving work in the garden or on the car were removed. This may reflect cultural differences relating to different ways of life.

In the original MPI-SCI, the authors hypothesized that the subscales of the MPI-SCI and the set of measures used for testing the convergent validity should be able to predict satisfaction with life in a person with SCI. In the present study, we used a person’s perception of well-being, which is a dimension of the PGWBI and another measure of quality of life. Similar to the study by Widerström-Noga et al.,8 we confirmed the hypothesis that having a combination of lower levels of affective distress, higher levels of general activity and lower levels of negative support predicted positive well-being. A previous study involving SCI patients with and without chronic pain, showed statistically significant differences between the pain prevalence and the perception of psychological well-being; those who suffered chronic pain were the ones with more psychological distress.3

Several limitations to this study should be noted. At the time of study design, we did not include measures to analyze test–retest stability of the MPI-SCI-S. Future research is needed to test the stability of the MPS-SCI-S over time in the Spanish population. However, the MPI-SCI-S showed excellent internal consistency indicating adequate reliability. Another potential limitation is that the instruments used in this study were administered via postal surveys and the return rate was very low 22.6%. A possible explanation for this low response rate could be the lack of motivation of the participants to complete the questionnaires. Unlike many other studies, subjects did not receive any kind of financial reward to answer the questions. In addition, the set of questionnaires was quite long and required a relatively long time to be completed. Another possible explanation could be that the educational level was a limitation to understand the content of the questions. Many participants who never returned the questionnaires (60.4%) reported an educational level of elementary school or less compared with our participants (46.8%). Finally, future psychometric studies in different Spanish speaking populations with SCI should be performed to assess transcultural validation.

In conclusion, the MPI-SCI-S subscales with the exception of the NR and SR subscales were found to have satisfactory criterion-related validity and internal consistency confirming its usefulness as a measure for assessing multidimensional pain in individuals with SCI. Future studies should include additional measures of social support to adequately assess this domain.

DATA ARCHIVING

There were no data to deposit.