Background
The Shoulder Pain and Disability Index (SPADI) is a self-administered questionnaire, designed by Roach and colleagues to measure the impact of shoulder pathology in terms of pain and disability, for both current status and change over time [
1]. The questionnaire consists of 13 items grouped into pain and disability subscales, the questions starting with "How severe is your pain...?" and "How much difficulty do you have...?", respectively. Items mainly deal with various activities of daily living (ADL) that may or may not be problematic to the patient. Items are rated on visual analogue scales to produce a score for each subscale, and the means of the two subscales are averaged to produce a total score ranging from 0 (best) to 100 (worst).
While responses to individual items are observable and concrete variables, the concepts they purport to assess when combined are abstract or latent variables, socalled "constructs" [
2]. For a meaningful comparison of scores, it is essential that scores reflect the same construct. To accommodate the interpretation of scores from multi-item questionnaires, health assessment scales are often divided, as is SPADI, into subscales with presumably separate constructs. In several cases, subsequent research has, however, failed to support the original structure of such scales [
3‐
5], demonstrating uncertainty regarding scale appropriateness in many settings.
SPADI is one of the shoulder rating scales that has been most extensively studied [
3,
6]. Construct validity (the extent to which a measure assesses the domain of interest [
2]) of the SPADI was investigated by the original developers through factor analysis. This is a statistical technique applied to a group of items to determine if the items form coherent subsets that are relatively independent from one another [
2]. In the beginning of the original SPADI article, the authors hypothesized that "Varimax rotation should produce two factors with the majority of items from each subscale primarily loading on different factors". However, the factors extracted did not delineate clearly between items of the pain and disability subscales in that original study, perhaps because the study was undersized [
1]. Three subsequent studies have come to diverging conclusions regarding SPADI factor structure. Only one factor was retained in studies by Roddey et al. [
7] and Placzek et al. [
4], but the original division of the questionnaire was supported in a study [
8] by MacDermid and colleagues.
The selection of an evaluative instrument should be based on evidence of reliability and validity in the target patient population [
9]. It has been argued that a shoulder rating scale should be equally valid across diagnostic categories. This may be far-fetched within the format of instruments like SPADI. Patients with different shoulder conditions may have functional limitations that need to be addressed in separate ways. Certain patient groups are mostly inhibited by pain while others experience e.g. limited strength or flexibility. Some conditions only become a problem when performing specific and highly demanding activities like throwing or weight-lifting.
Adhesive capsulitis is one of the most common disorders affecting the shoulder and SPADI has been employed in several clinical trials involving this patient category [
10‐
14]. The main findings associated with active adhesive capsulitis is a constricted glenohumeral joint capsule and a shoulder pain that becomes severe if the arm is passively or actively moved towards limits of range-of-motion. Symptoms indicate that both pain and disability parameters may apply when characterizing patients. SPADI items seem to address constructs of pain or disability in various situations, and in this sense, face validity is promising. However, it is uncertain whether the two can be assessed separately in this way since pain and disability may be very closely connected in these patients. The aim of the present study is to investigate if the underlying factor structure of the SPADI supports a separation of scores into different subscales when describing patients with adhesive capsulitis.
Discussion
The factor structure identified in this study does not support the original division of the SPADI since the two extracted factors do not seem to delineate clearly between pain and disability subscales. The questionnaire clearly asks the patient to report pain in the first section and difficulty in the last, but it is unclear if the difference is appreciated by the patients. The factor structure revealed in this study is in line with previous reports on the construct validity of the SPADI.
Region-specific scales used in rheumatology or orthopedics tend to include items that refer to pain and various aspects of limited functioning associated with activities of daily living. For some scales, the association between pain and function has been shown to be weak, while for others it seems to be stronger [
4]. It has been proposed that patients in some cases may have difficulties in separating the concepts because activities are essentially limited by pain [
5]. Others have proposed that pain and disability items in questionnaires may correlate because pain and disability items tend to address similar tasks [
22].
Critical to the analysis of the factor structure of a scale is deciding the number of factors to extract before rotation. In this study, two initial factors were extracted according to the eigenvalue criterion, a result that would seem to fit the number of constructs addressed in SPADI.
From a biomechanical perspective, it is tempting to label the first factor in the rotated solution "Pain interference" and the second factor "Functional limitation". Patients with adhesive capsulitis in the active stage experience an aggravation of pain when the arm is moved towards the limits of range-of-motion. The disability items that load on the first factor involve movements near (or beyond) end-range of shoulder motion in these patients. Hence it is not surprising that some "difficult" items in the disability section may load on pain. The interpretation would be that disability subscale scores depend on both pain interference and functional limitation.
The more pragmatic researcher might prefer to view both factors as "Pain interference" factors, the difference between them being pain interference with higher and lower demand activities, respectively. Variability in item differentiation may reflect clinical phenomena, but interpretational and psychological issues may be as relevant. Considerations regarding scoring procedures also apply.
Internal consistency was slightly lower than reported by previous researchers. Cronbach alpha was 0,90 for the total score, compared to Roach 0.95 [
1], Roddey 0.96 [
7], MacDermid 0.95 [
8] and Angst 0.95 [
23]. The value still indicates that the scale may be viewed as unidimensional.
Inter-subscale correlation (0.73) was in line with previous findings (Roach 0.87 [
1], Placzek 0.71 [
4], Roddey 0.77 [
7], MacDermid 0.66 [
8]). Investigation of agreement between the pain and disability subscales indicated that scores could vary by as much as 30 points for an individual patient, although the difference on average was only 3 points. A 30-point difference of this type is more than you would expect in a simple test-retest study of each subscale, indicating that subscale scores are not interchangeable for all patients.
The results of this study largely conform with the reports of Roach et al. [
1], Placzek et al. [
4] and Roddey et al. [
7], in the sense that we have identified a factor loading pattern that does not support the original subscale division of items. This result may appear to contrast with the factor structure reported in the study by MacDermid et al., where one factor loaded primarily on pain items and another factor primarily on disability items [
8]. However, the authors noted that higher demand activities tended to load on pain instead of disability. In MacDermid's study of community volunteers who self-identified as having shoulder pain, mean score of the disability section was 32. In our study it was 58. We find it reasonable that the "disability" items reflect pain interference to a higher degree in our study population of patients with adhesive capsulitis.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
EKT, NGJ and EB–H designed the study. EKT, OME, NGJ and EB–H collected the data. EKT and LS performed the statistical analysis. EKT drafted the manuscript with contributions from LS, OME, NGJ and EB–H. All authors read and approved the final manuscript.