Development of a brief version of the Survey of Pain Attitudes

1. Introduction

Attitudes and beliefs have long been subject to psychological research because of their mediating influence on behavior (^{Ajzen and Fishbein, 1977}). In the case of an attitude, the influence stems from a person's feelings toward a _target or subject (^{Fishbein and Ajzen, 1975}), while beliefs impact behavior through information that a person possesses relevant to a _target. Attention to attitudes and beliefs in pain research has been relatively recent. Nevertheless, studies have shown these constructs to be important influences on response to treatment (^{Schwartz et al., 1985; Jensen et al., 1994a}) and patient adjustment to pain (^{Jensen and Karoly, 1991; Jensen et al., 1991; Strong et al., 1992; Jensen et al., 1994b}).

Not surprisingly, a number of assessment instruments have been developed to measure attitudes and beliefs about pain. One of the first, the Pain Information and Beliefs Questionnaire (PIBQ; ^{Schwartz et al., 1985}), was derived from pain patient responses to information about chronic pain and its treatment as presented in a brief educational videotape. The PIBQ has been found to correlate with outcomes of and satisfaction with pain treatment (^{Shutty et al., 1990}). The Pain and Impairment Relationship Scale (PAIRS; ^{Riley et al., 1988}) was developed to measure attitudes towards pain and pain-related disability. While it appears to do this reasonably well (^{Slater et al., 1991}), it is limited in that it only assesses attitudes concerning the pain-disability association. The Pain Beliefs and Perceptions Inventory (PBPI; ^{Williams and Thorn, 1989}) covers a broader range of beliefs than the PAIRS (including beliefs regarding pain stability over time, pain as a ‘mystery’, and self-blame for pain) but the research evidence is mixed regarding its factor structure and construct validity (^{Strong et al., 1992}).

The instrument that appears both to assess a range of attitudes and to do so with reliability and validity is the Survey of Pain Attitudes (SOPA), an instrument initially developed by ^{Jensen et al. (1987)} to assess patient attitudes toward five dimensions of the chronic pain experience: pain control, pain-related disability, medical cures for pain, solicitude of others, and medication for pain. The initial instrument included 24 items, but was subsequently revised to include 35 items and the additional dimension of emotionality (i.e., the influence of emotions on pain) (^{Jensen and Karoly, 1989}). While there was reasonable evidence supporting the psychometric properties of that version (^{Strong et al., 1992}), the authors subsequently added items to the Medication and Disability subscales and added a seventh dimension relating to pain as evidence of physical harm (^{Jensen et al., 1994b}). This version of the SOPA contained 57 items and the subscale scores demonstrated acceptable internal consistency, test-retest reliability, and convergent/discriminant validity (^{Jensen et al., 1994b}). Several studies with chronic pain patients have shown one version or another of the SOPA to be associated with psychosocial functioning (^{Jensen and Karoly, 1991; Jensen et al., 1994b}), physical activity levels (^{Jensen and Karoly, 1991}), utilization of professional medical services (^{Jensen and Karoly, 1992; Jensen et al., 1994b}), and physical disability (^{Jensen et al., 1994b}).

Although the SOPA shows considerable value as a research instrument, several questions remain. First, the factor structure of the 57-item version has not been verified. Although ^{Strong et al. (1992)} reported an acceptable factor structure for the SOPA that was consistent with its purported dimensions, that work was done on the 35-item SOPA. Since the SOPA is designed to yield seven subscale scores representing seven different facets of pain attitude, its factor structure is a critical part of its psychometric profile. Second, the utility of the current SOPA is diminished somewhat by its length. With 57 items and seven subscales, it can be cumbersome to administer and time-consuming to score. The present study was conducted to investigate the factor structure of the 57-item SOPA and to develop an alternative, shorter form of the instrument similar to the 35-item version studied by ^{Strong et al. (1992)}. Such a version of the SOPA, with acceptable psychometric properties, should facilitate more widespread use of the instrument, thus directing clinical and research attention toward the important constructs of pain attitudes and beliefs.

2. Methods

2.1. Subjects

The analyses presented here were based on data gathered from 395 consecutive chronic pain patients evaluated at a comprehensive pain service located in a major university medical center. The sample was 63% female and 37% male. The average age was 47.1 years (SD=14.6 years) and the average number of years of formal education was 12.4 (SD=2.4). The patients had been in pain for an average of 6.3 years (SD=8.7 years) and reported a usual pain intensity of 70.7 (SD=18.2) on a scale that ranged from 0 (no pain) to 100 (pain as bad as it could be). These patients exhibited predominantly musculoskeletal pain of varying distributions, including head/neck/upper extremity pain (11.4%); low back/lower extremity pain (34.4%); and widespread pain involving the low back, lower extremities, upper extremities, and the head or neck (47.8%). About 6% of the sample could not be classified as having musculoskeletal pain in any of the above distributions and pain type was unavailable for two patients.

2.2. Procedure

Patients completed a self-report assessment battery as a part of the pre-treatment screening protocol for the treatment facility. The assessment battery contained a number of measures in addition to the SOPA, administered in the following order: a measure of pain intensity, ranging from 0 (no pain) to 100 (excruciating pain); a pain drawing which yielded a measure of percentage of body surface in pain (PBS; ^{Margolis et al., 1986}); the Pain Disability Index (PDI; ^{Tait et al., 1990; Chibnall and Tait, 1994}); the Pain Behavior Checklist (PBCL; ^{Kerns et al., 1991}); the Vanderbilt Pain Management Inventory (VPMI; ^{Brown and Nicassio, 1987}); the SOPA; and the Center for Epidemiologic Studies Depression Scale (CES-D; ^{Ensel, 1986}). In addition to these standardized measures, patients also rated their pain severity on a 21-point visual analog scale and provided information on a variety of demographic characteristics. With the exception of the measure of pain intensity, each measure was standardized. In regard to pain intensity, where possible scores ranged from 0 to 100, anchors were provided at five-point intervals, yielding a modified, 21-point visual analog scale, a length that research has shown to be sufficient to yield reliable and useful data regarding pain intensity (^{Jensen et al., 1994c}).

2.3. Statistics

Four progressive principal components analyses were used to factor the SOPA items, eliminating redundant and/or extraneous items at each stage. Pearson product-moment correlation coefficients were computed to assess construct validity (i.e., relations among SOPA-B (SOPA-Brief; see Appendix A) and SOPA subscales) as well as relations among SOPA-B subscales and other standardized measures. Point-biserial correlations were computed to examine relations between SOPA subscales and categorical data (e.g., gender).

3. Results

3.1. Factor analyses

A principal components analysis of the 57 SOPA items yielded 15 factors with eigenvalues exceeding one. Together, these factors accounted for 59.6% of the total variance in the original items. Individual factors accounted for as little as 1.8% of the variance and as much as 13.0% of the variance. Inspection of the varimax rotated solution showed factors one through six to be reasonable representations of six of the seven SOPA subscales: Disability, Solicitude, Emotion, Control, Cure and Harm. However, only 31 of the 51 items included in these subscales loaded on these factors. Also, one item designated as Harm clearly loaded with the Disability items, while four items loaded highly on more than one factor. Factor seven contained a second set of Control items and factor nine was composed of three of the original six Medication items. Factor eight was composed of two Cure items reflecting the role of the physician in curing pain problems. Factors 10 through 15 were basically uninterpretable and did not represent in a coherent sense any SOPA subscale.

In a second principal components analysis, all items from factor eight and factors 10 through 15 were excluded. The subsequent solution yielded nine factors that accounted for 57.7% of the variance. Factors one through five represented Solicitude, Emotion, Control, Disability and Cure, respectively. Factor six contained the second set of Control items. Factors seven and eight were clearly Harm and Medication factors. Factor nine was basically uninterpretable, and no variable loaded higher than 0.44 on it. As in the first analysis, one Harm item clearly loaded with the Disability items and four variables in the first five factors loaded highly on more than one factor. In all, 38 items loaded on factors one through eight.

In the third principal components analysis, items from the second Control factor were excluded from the analysis. Of the nine factors generated in that solution (59.8% of the variance), the first seven factors reflected Solicitude, Emotion, Control, Cure, Disability, Harm and Medication, respectively. Only one item loaded on factor eight and factor nine was uninterpretable. The Harm item continued to load with the Disability items and four items continued to have multiple loadings.

For the final principal components analysis, all items that loaded on more than one factor were excluded from the analysis. The subsequent solution comprised seven clear factors with no cross-loading items (55.9% of the variance). The seven factors, moreover, were consistent with those constituting the original SOPA, with the exception of the one Harm item that loaded with the Disability items: Solicitude (15.2% of the variance), Emotion (12.6%), Cure (8.0%), Control (6.0%), Harm (5.2%), Disability (4.8%) and Medication (4.2%). Table 1 shows the factor structure for the final 30-item solution.

Alpha reliabilities were computed for the subscales for both the 30-item SOPA (hereafter referred to as the SOPA-B for SOPA-Brief) and the 57-item SOPA. The reliabilities are shown in Table 2 along with the scale reliabilities reported by ^{Jensen et al. (1994b)} in their recent article describing the 57-item SOPA. Inspection of the table shows the subscale reliabilities for the 57-item SOPA to be comparable between the current study and the ^{Jensen et al. (1994b)} study. The SOPA-B subscale reliabilities were equal to or greater than those reported here for the 57-item SOPA in four of seven cases, ranging from a high of 0.83 (Solicitude, five items) to a low of 0.56 (Medication, three items). While the latter alpha coefficient is relatively low, the other reliabilities were within a generally acceptable range from 0.70 to 0.83.

3.2. SOPA and SOPA-B subscale correlations

Correlation coefficients were computed between the SOPA-B subscales and the subscales of the 57-item SOPA to assess concurrent validity. As shown in Table 3, correlations between like subscales were consistently strong, ranging from a low of 0.79 for the Disability subscales to a high of 0.97 for the Solicitude subscales.

3.3. Demographics

Correlations between the SOPA-B subscales and selected demographic variables revealed several statistically significant relationships. Subjects with more years of formal education reported pain to be somewhat more controllable, r=0.19, P<0.001, and were less inclined to interpret pain as an indication of bodily harm, r=-0.19, P<0.001. Women were somewhat less likely than men to see pain as a disabling condition, r=-0.17, P<0.01. Older patients tended to see pain as more curable, r=0.14, P<0.01, and medicine as an effective treatment for pain, r=0.16, P<0.01. Finally, patients with pain of a longer duration were somewhat more likely to see a relation between emotional status and pain, r=0.14, P<0.01. While statistically significant, however, the largest effect size accounted for less than 4% of the variance shared between variables, so that the strength of these relations was uniformly weak.

3.4. Relations with other variables

To further assess the utility of the SOPA-B, correlations were computed between its subscales and other measures administered to patients as part of the assessment. As shown in Table 4, the pattern of correlations was generally consistent with the constructs represented by the various subscales. The Disability subscale correlated most highly with distorted ambulation and expressions of pain, passive coping behavior, and disability as measured by the PDI, consistent with a high degree of pain-related interference with function. The Medication subscale correlated positively with passive coping behaviors and help seeking, as would be expected of an intervention largely controlled by an external agent (the physician). The Control subscale correlated positively with active coping and negatively with passive coping. The Solicitude subscale correlated with depression, passive coping, affective distress, and help-seeking, a set of attitudes likely to solicit external support. Emotionality correlated with the above measures and with facial expressions of pain and pain extent, reflecting an array of dysfunctional responses to pain. The Harm subscale correlated negatively with active coping and positively with passive coping, usual pain intensity, and with disability, a pattern that reflected avoidance of pain-producing activities. Finally, the Medical Cure subscale (with lower scores indicating a lower expectation for a medical cure) was negatively correlated with indicators of depression and affective distress, potentially reflecting hope for a cure, and negatively with a measure of pain extent, suggesting that these patients presented with more circumscribed pain problems.

4. Discussion

The results of the analyses generally support the psychometric properties of a brief version of the Survey of Pain Attitudes. The Survey of Pain Attitudes-Brief (SOPA-B) contains seven subscales that are conceptually and empirically consistent with those described for the 57-item SOPA. Similarly, subscales from the SOPA-B correlated highly with their counterparts on the SOPA. Although the SOPA-B is almost half the length of the latter instrument, its subscales possessed adequate internal consistency, with the exception of the Medication subscale. Finally, correlations with a range of self-report measures were consistent with constructs represented by the SOPA-B subscales.

On the other hand, the support for the factor structure of the SOPA was mixed. While many items loaded on factors reflecting the appropriate subscale, other items loaded either on inappropriate or redundant factors. Hence, the 27 items eliminated from the SOPA to form the SOPA-B were removed without sacrificing subscales or significant reliability from most subscales.

The one subscale which did appear to suffer on the SOPA-B involved Medications. Internal consistency for this subscale was unacceptably low, partly due to its brevity (only three items). Even with double the number of items (the six-item Medication subscale for the SOPA), however, internal consistency was less than 0.70 in this study. Additional, more homogeneous items may be needed if this subscale is to become a reliable part of the assessment instrument.

There are several advantages associated with the availability of a brief measure of pain-related attitudes. Clearly, the primary one involves its greater ease of administration. In light of the importance that is attached to cognitive and attitudinal variables in regard to patient adjustment (^{Jensen and Karoly, 1991; Jensen et al., 1991; Strong et al., 1992; Jensen et al., 1994b}) and treatment outcomes (^{Schwartz et al., 1985; Jensen et al., 1994a}), there is value to any development that will facilitate more widespread use of such instruments. Our early experience with the SOPA-B indicates that the shortened version is easier to use and interpret in the context of pain patient evaluations.

Other advantages may accrue from easier administration of a brief attitude measure. With such a measure, it may be possible to assess attitudes of patients seen in a broad array of treatment settings (e.g., work hardening programs) so as to identify patients whose attitudes could interfere with a good response to that treatment. Whether the setting involves work hardening or other treatment venues, further assessment of attitudinal factors likely to affect/predict adjustment to pain and/or treatment outcomes could contribute useful information to the treatment regimen. Early evidence suggests that attitude assessment holds such possibilities.

While the present research supports the SOPA-B, limitations of the present study should be noted. Most importantly, the data presented in this paper regarding the SOPA-B were derived from an administration of the 57-item SOPA. Until cross-validation research is conducted with the SOPA-B itself, it is premature to assume that the SOPA-B is as reliable and valid as it appears to be in the present study. Similarly, it is premature to assume that the SOPA-B possesses the clinical and predictive value that the 35-item and 57-item versions of the SOPA have demonstrated in previous studies or the good test-retest stability (ranging from r=0.63 to r=0.68 over a 6-week time period) ascribed to the SOPA (^{Jensen et al., 1994b}).

Another shortcoming of the present study involves the fixed order in which the assessment instruments were administered. In fact, the order of the instruments was carefully chosen, with the assessment packet beginning with measures clearly related to the presenting problem of pain (e.g., intensity, distribution) and ending with measures of more clearly psychological phenomena (e.g., attitudes, depression). Hence, there may be order effects that influenced the responses that patients provided to the instruments, especially the latter instruments.

Despite the latter concerns, the SOPA-B appears to have considerable promise as part of an assessment battery used for the evaluation of patients with intractable pain. It can be recommended because of its brevity and preliminary evidence supporting its psychometric properties. With further study, it may emerge as a useful and practical measure for future research on pain attitudes and beliefs, and one which, because of its brevity, may lend itself to use beyond specialty, pain clinic setting.