- Email: [email protected]
Assessing pain behaviour of low-back pain patients in real time: concurrent validity and examiner sensitivity
Behaviour Research and Therapy 40 (2002) 595–607 www.elsevier.com/locate/brat
Behavioural assessment
Assessing pain behaviour of low-back pain patients in real time: concurrent validity and examiner sensitivity K.M. Prkachin
a, d,*
, I. Schultz b, J. Berkowitz b, E. Hughes a, D. Hunt
c
a
Department of Psychology, University of Northern British Columbia, 3333 University Way, Prince George, BC, Canada V2N 4Z9 b Department of Educational and Counselling Psychology and Special Education, University of British Columbia, 2329 West Mall, Vancouver, BC, Canada V6T 1Z4 c Department of Family Practice, University of British Columbia, 2329 West Mall, Vancouver, BC, Canada, V6T 1ZT d Workers’ Compensation Board of British Columbia, 6951 Westminster Highway, Richmond, BC Canada V7C 1C6 Received 18 April 2001; accepted 5 June 2001
Abstract Several systems for measuring pain behaviour have been developed for clinical settings. The present study reports on a real-time system for coding five categories of pain behaviour for low-back pain patients: guarding, touching, sounds, words, and facial expression. Unique features of the system are the use of refined measures of facial expression and integration of the measurements with a standardized physical examination. 176 sub-acute and chronic low-back pain patients underwent a physical examination while their pain behaviour was coded. Concurrent measures of subjective pain, medically-incongruent signs, and independent global ratings of pain behaviour were taken. Analyses indicated that the pain behaviours, particularly guarding and facial expression, varied systematically with the alternative measures, supporting the concurrent validity of the behaviour observation system. While pain behaviours, especially use of words and facial expressions, were significantly associated with the examiners’ independent ratings, the strength of the associations suggested that, in the absence of direct training, examiners’ performance was relatively poor. Implications for training of clinicians in detecting pain behaviour are discussed. 2002 Elsevier Science Ltd. All rights reserved. Keywords: Pain; Behaviour; Facial expression; Guarding; Low-back pain; Patients; Validity
* Corresponding author. Tel.: +1-250-960-6633; fax: +1-250-960-5536. E-mail address: [email protected] (K.M. Prkachin).
0005-7967/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 0 1 ) 0 0 0 7 5 - 4
596
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
1. Introduction The purpose of this paper is to describe properties of a system for observing pain behaviour in real time. The Pain Behaviour Observation System (PBOS; Prkachin, Hughes, Schultz, & Joy, 2001) integrates observations of pain behaviour with a standardized physical examination for lowback pain patients (Hunt et al., in press). Observers code five behaviours: guarding, touching, and the use of words, sounds, and facial expression to communicate pain. People in pain often show behavioural changes that are quite distinctive to observers whose attention happens to be drawn to the sufferer. In turn, changes in sufferers’ behaviour lead observers to engage in a number of processes (Prkachin & Craig, 1994), among which one is to draw inferences about the existence, or the intensity, of suffering. It can be argued that all of pain science and management is based on the process of observers drawing inferences about sufferers’ experiences based on observations of their (the sufferers’) behaviour. Consequently, it is important to be as precise and sophisticated as possible when trying to understand the nature of pain behaviour and how to assess it. In the behavioural literature, a number of different systems for recognizing and measuring pain behaviour have been developed (Jensen, 1997). Among these, probably the most widely used was developed and refined by Keefe (Keefe & Block, 1982; Keefe & Williams, 1992). Keefe’s system recognizes five categories of pain behaviour: guarding, bracing, rubbing, sighing, and grimacing. Several other systems have been developed to quantify pain behaviours associated with different types of pain (Cinciripini & Floreen, 1983; Follick, Ahern, & Aberger, 1985; Richards, Nepomuceno, Riles, & Suer, 1982; Vlaeyen et al., 1990). Generally speaking, these systems vary in the number of behaviours incorporated into the observational procedure and the settings in which they are applied. However, they tend to reflect some common underlying assumptions, among which are that the behaviours are broadly communicative of pain intensity, that an additive accumulation of behaviours reflects greater pain, and that the behaviours resemble one another in their metric properties. A different approach to the assessment of pain behaviours has been developed by Waddell and colleagues (Waddell, 1991; Waddell, McCulloch, Kummel, & Venner, 1980). This technique, which has become increasingly popular in the pain assessment and rehabilitation field, involves evaluation of a variety of symptoms, such as reports of pain that do not follow recognized distributions based on understanding of neuroanatomy, and behavioural ‘signs’. The behavioural signs (also commonly referred to as ‘nonorganic signs’, ‘Waddell signs’, or the more neutral ‘medicallyincongruent pain’; Reesor & Craig, 1988) require exposure of the patient to manoeuvres such as testing for tenderness, exposure to tests that generally do not cause pain but give the appearance of doing so, or observing for inconsistencies in pain expression when patients’ attention is distracted. Scoring of behaviours within this system is based on classification of a sign as present or absent. There has been substantial discussion in the literature about the meaning and interpretation of these signs (Karas, McIntosh, Hall, Wilson, & Melles, 1997; Main & Waddell, 1998; Scalzitti, 1997; Werneke, Harris, & Lichter, 1993). Current opinion seems to favour the suggestion that they are markers of psychological distress. One of the objectives of the present study was to evaluate the concurrent validity of the PBOS in low-back pain patients by determining the relationship between pain behaviours and alternative indicators of patients’ pain. One set of indicators was the patient’s own reports of pain experienced
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
597
during the physical examination. Another was the technique for observing medically-incongruent signs developed by Waddell and colleagues (1980). The term ‘pain behaviour’ tends to be used as a unidimensional construct. This is probably attributable, in part, to Fordyce (1976) who first popularized its use. In Fordyce’s formulation, pain behaviour could be affected by injury or tissue damage on the one hand, or its personal and social consequences on the other. The former he referred to as ‘respondent’ and the latter ‘operant’ pain. Although this view recognized two categories of pain behaviour based on their hypothesized determinants, clinical use of the term tends not to make even this degree of differentiation. As long ago as 1977, however, Liebeskind and Paul drew attention to the fact that pain behaviours are organized in different ways, at different levels of the nervous system, implying that a more complex analysis might be necessary (Liebeskind & Paul, 1977). Turk, Wack, and Kerns (1985) used multidimensional scaling and hierarchical clustering analysis to identify the structure evident in clinicians’ ratings of 20 different pain behaviours. Their results appeared to identify four broad categories of behaviour: distorted gait/posture, negative affect, audible/visible behaviour, and activity avoidance. Vlaeyen, Van Eek, Groenman, and Schuerman (1987) identified nine components of observed pain behaviour (anxiety, attention seeking, pain complaint, general complaint, medication use, distorted posture, fatigue, insomnia, and depressive mood) loading on three overall dimensions: withdrawal–approach, arousal, and visible/audible. These observations are consistent with a more multidimensional view of pain behaviour. Different categories of pain behaviour are likely to serve different functions (Prkachin, 1986). Guarding, for example, seems likely to be a self-management behaviour reflecting, to a significant degree, active efforts on the sufferer’s part to minimize discomfort. Grimacing, on the other, is a member of a broader category of behaviour that has evolved to serve a communicative function (Prkachin, 1986; van Hoof, 1972). Thus, although these and other definable types of pain behaviour are observable and provide evidence about pain in a sufferer, there is no reason to expect that they will yield the same information, nor that they will be affected by precisely the same controlling variables. The construct of pain behaviour has permeated the professional community. Clinicians, for better or worse, use observations of pain behaviour to draw inferences about pain experience and its causal variables. Most systems for assessing pain behaviour require special examinations focused exclusively on pain behaviour. For example, Keefe and Block’s (1982) approach is based on alternating two-minute periods during which back pain patients are observed while standing, sitting, and walking. Another setting that is likely to provide information about pain and that may have advantages over contrived behavioural observation tests is the clinical examination. During clinical examinations of back pain patients, clinicians expose the sufferer to a series of challenges designed to yield diagnostic evidence. Many of these challenges will, of necessity, evoke pain and pain behaviour. The second, and chief, objective of the present study was to examine the relationship between pain behaviours as assessed by our system, and clinicians’ ratings of the same behaviours based on their impressions of the examinations. As noted above, clinicians are increasingly making use of the construct of pain behaviour, and it is not uncommon in clinical or medico-legal settings for such behaviour to be noted and used as a basis for recommendations. Frequently, however, it is not clear how systematically such observations have been made or the degree of training the clinician has had in making them. This raises the question of the validity of such judgements.
598
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Consequently, we were interested in determining how closely relatively sophisticated clinicians’ ratings corresponded with more objective measurements performed by highly trained experts. 2. Method 2.1. Participants Patients were 176 Workers’ Compensation Board of British Columbia claimants who took part in a larger study of predictors of disability. They were categorized into two groups based on the time that had elapsed since the injury leading to their currently active claim. Patients were designated as sub-acute (n=148) if their claim had taken place within the preceding 4–6 weeks, or as chronic (n=28) if their claim was between 6 and 12 months’ duration. Mean age of the sub-acute patients was 40.5 (SD=10.2), and of the chronic patients was 43.3 (SD=11.6). 126 patients were male and 50 were female. For purposes of the present study, patients were collapsed across pain type and sex. Patients were recruited from the Board’s computerized claim system based on rosters generated weekly. They were allocated at random to one of six examiners who were either Board physicians or physiotherapists. Physical examinations were observed by one of five female observers, each of whom held baccalaureate degrees in Psychology (one PhD). 2.2. Procedure Patients underwent a standardized physical examination prepared by a Board physician. Examination procedures were selected based on an extensive literature review to identify tests that the scientific evidence suggested were reliable and prognostically valid. Anthropometric measures, physical signs, flexion and extension, medically-incongruent signs, reflexes, strength, muscle bulk, sensation, tenderness, and range of motion in a variety of tests were recorded. All examiners underwent extensive training to be able to perform the examination in a standardized manner (Hunt et al., in press). 2.2.1. Pain behaviour observation system (PBOS) Each of the examinations was observed by a coder who had been trained in the PBOS. The system was designed to be applied in real time during clinical conditions. Consequently, it was necessary that observations be performed without resorting to videotape recording. We also wished to take advantage of the clinical examination as a setting for behavioural observations. The categories of the PBOS were inspired by, but modified from, Keefe and Block’s (1982) technique, to meet several further requirements or to match the conditions of the study. First, two of Keefe and Block’s categories — guarding and bracing — were collapsed into one. This was done because bracing is considered to be a form of guarding that takes place when the sufferer is not moving (Keefe, Crisson, & Snipes, 1987). Second, a category was included for pain-related verbalization. Third, a category for facial expression was developed. Although Keefe and Block included a category for grimacing in their system, their operational definition differs in significant ways from the pain-related facial expression that has been described in empirical research (Craig, Prkachin, & Grunau, 2001; Prkachin, 1992; Prkachin & Craig, 1994). Finally, in Keefe and
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
599
Block’s system, behaviours must be present for at least three seconds in order to be coded. As the duration of some pain behaviours, particularly facial expressions, is frequently substantially shorter, this rule was eliminated. To learn the system, observers participated in a day-long workshop after reading a manual describing the behaviours and recording principles. The workshop consisted of didactic instruction accompanied by a training video that displayed each of the categories of behaviour to be observed, and special considerations associated with their scoring. It also provided opportunity for observers to practice and receive feedback on their coding. A separate video was used to train observers in the detection of facial expressions of pain. The training procedure was based on the system developed and evaluated by Solomon, Farewell, and Prkachin (1997). Observers were trained to identify five categories of pain behaviour, as follows: Guarding: behaviour that prevents or alleviates pain, including stiffness, hesitation, limping, bracing, and flinching. Touching: any contact between the patient’s hands and the lower back, the hips, buttocks, and the outer aspect of the thighs. Words: any spontaneous utterance that relates to the patient’s pain. Sounds: moaning, sighing, puffing or slow exhalation of breath. Facial expression: movements of the facial musculature that narrow the palpebral fissure, tighten the eyelids and raise the cheeks, lower the eyebrows, and raise the nose. When the training workshop was complete, observers were given a further training videotape to practice coding. They were provided with feedback about their coding in the form of how many behaviours they had got correct, proportion of false positives (coding a behaviour as present when it was not) and misses (coding a behaviour as absent when it was present). Finally, observers underwent a final test of proficiency in which they coded 40 vignettes taken from patients undergoing the procedures of the physical examination. All coders achieved a score exceeding 78% on this test, with the average percent correct being 81%. Based on extensive observations of prototypes of the physical examination, a system for parsing it was developed. The entire examination was broken into 57 segments based on the tests being performed by the examiner or transitions taking place within the examination. Of those segments, behavioural coding took place in 36. Observers were trained in parsing the examination in realtime before beginning the study. Table 1 lists observed and unobserved coding epochs. Guarding, touching, words and sounds were recorded as present or absent during each of the recording epochs. Facial expressions were rated on a three-category scale (0=none, 1=some, 2=much) because evidence indicates that the information contained in facial expression is graded and that discriminating power is lost if this is not taken into consideration (Prkachin & Mercer, 1989). Coders positioned themselves inside the examination room where they could maintain a complete view of the patient’s face while not interfering with the progress of the examination. The observers made notations on a recording sheet. 2.2.2. Evaluation of medically-incongruent signs Medically-incongruent signs were evaluated according to the techniques described by Waddell and colleagues (Waddell et al., 1980; Waddell, Pilowsky, & Bond, 1989; Waddell & Turk, 1992).
600
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Table 1 Discrete epochs of the physical examination (coding epochs in italics) Interval 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.
to sitting 1 introduction to scale weight/height to landmarks landmarks lordosis heel raise 1 and 2 forward/back rotation side-to-side stand for measurement lumbar extension 1 lumbar flexion 1 lumbar extension 2 lumbar flexion 2 lumbar extension 3 lumbar flexion 3 lumbar extension 4 lumbar flexion 4 lumbar extension 5 lumbar flexion 5 lateral flexion left 1 lateral flexion right 1 lateral flexion left 2 lateral flexion right 2 lateral flexion left 3 lateral flexion right 3 lateral flexion left 4
30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57.
lateral flexion right 4 lateral flexion left 5 lateral flexion right 5 axial compression simulated trunk rotation to kneeling ankle reflexes to sitting 2 knee reflexes knee extension 1 knee extension 2 muscle strength 1 muscle strength 2 to supine 1 ankle dorsiflexion toe extensor muscle bulk measure sensation passive straight leg raise R passive straight leg raise L to prone palpation McKenzie push-up prone active extension to supine 2 active situp bilateral straight leg raise to standing
Their assessment was based on the following tests, which were performed by the examiners as part of the physical examination: Axial loading: the examiner applied 0.5–1 kg pressure with his or her hands to the top of the patient’s head. The test was considered to show a positive sign if the patient complained of resultant pain in the lumbar region. Simulated trunk rotation: from behind the patient, the examiner held the participant’s arms to the side, while rotating their hips, making sure the pelvis and shoulders rotated together. A positive sign was recorded if the patient complained of pain in the lumbar region. Superficial tenderness: the examiner applied light pinching in a standardized fashion across both sides of the lower back. A positive sign was recorded if the patient complained of tenderness over a wide area of the lumbar skin. Distraction on straight-leg raising: the examiner raised the left and right leg in the supine and
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
601
seated positions, noting the maximum range of motion. During the seated test, the examiner distracted the patient. An improvement in range of motion during distraction was taken as a positive sign. Sensory disturbance: the examiner tested for sensation to light touch or pinprick on both legs. Altered sensation involving the whole upper or lower leg, or a stocking-like distribution were taken as a positive sign. Motor disturbance: the examiner noted the presence of ‘jerky’ giving way of the lower limbs during testing. 2.2.3. Pain drawing, timed walk, and visual analogue scale measures As part of the examination, participants also completed a pain drawing (Ransford, Cairns, & Mooney, 1976). Participants were presented with an anatomical drawing, with front and back views, of the human form. They were asked to mark on the drawing areas of numbness, pins and needles, burning, stabbing, and aching. The examiner evaluated the drawing for evidence of poorly localized, widespread, or non-anatomical sensations and evidence of written comments documenting severity, and rendered a decision as to whether the drawing was largely ‘anatomical’ or ‘non-anatomical’. At the end of the examination, patients underwent a timed walk. Instructions for the test indicated that the participant was to walk as briskly as possible. The examiner accompanied the patient, taking care not to provide encouragement or to give feedback as to the elapsed time. The walk was conducted over a distance of 500 m on a flat surface. Time to complete the walk was recorded in seconds. At the end of the examination, the patients also made ratings of their pain on 100 cm visual analogue scales (VAS), anchored with the words no pain and worst possible pain. Patients provided two ratings, one of their current pain and one of the maximum pain experienced during the examination. VAS ratings were measured to the nearest 10 cm. 2.2.4. Examiner’s rating of pain behaviour At the end of each examination, the examiner made an independent rating of the intensity of each of the five categories of pain behaviour represented on the PBOS. Examiners had not been given training in the detection of pain behaviours; however, they were all familiar with the concept and had been provided with the same operational definitions that the observers worked from when making their observations.
3. Results Table 2 presents intercorrelations among the total PBOS pain behaviour scores and their means and standard deviations. All the pain behaviours were positively associated with one another at high levels of significance. The magnitudes of the intercorrelations ranged from moderate to high. A principal components analysis (with Varimax rotation) of these intercorrelations resulted in a single factor accounting for 46% of the common variance. Factor loadings are also presented in Table 2.
602
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Table 2 Intercorrelations among PBOS total behaviour scores, means and standard deviations, and loadings on the single principal componenta Guard Guard Touch Words Sound Face Mean SD
6.34 5.85
Touch
Words
Sound
Face
PC
0.45
0.44 0.31
0.64 0.34 0.55
0.41 0.36 0.37 0.58
0.71 0.48 0.64 0.82 0.67
1.47 2.01
4.12 4.14
9.76 7.11
12.95 10.62
a Note: guard, guarding; touch, touching; sound, sounds; face, facial expression; PC, principal component. All values reflect counts of coding intervals in which the behaviour occurred, except for facial expression, which represents the sum of three-point intensity codes across coding intervals. All p⬍0.001.
Correlations between patients’ subjective pain ratings, walking time, the pain drawing measure, and observed pain behaviours are presented in Table 3. Observers’ ratings of each of the five categories of pain behaviour were significantly and positively related to patients’ ratings of both their current pain and the worst pain experienced during the examination. Of the different pain behaviours observed, words bore the weakest and guarding the strongest relationship with patients’ pain reports. All the observed pain behaviours were significantly positively correlated with the time they took to complete the 500 m walk. Whether the patient’s pain drawing appeared to be anatomical or nonanatomical was only significantly related to facial expression. In most cases, the relationships between pain behaviours and the concurrent measures of pain report, walking time, and nature of pain drawing, though significant, were weak to modest. Facial expression bore the most consistent pattern of relationships with the other indices while guarding showed the strongest overall relationships with pain reports. Point biserial correlations were calculated between the examiners’ binary ratings of the six signs of medically-incongruent pain and the observed pain behaviours. These correlations are
Table 3 Correlations between patients’ verbal pain ratings and observers’ pain behaviour recordingsa
Current pain Worst pain Walk time (s) Drawing
Guard
Touch
Words
Sound
Face
0.43*** 0.41*** 0.26** 0.07
0.34*** 0.30*** 0.34*** 0.09
0.24** 0.29*** 0.24** 0.00
0.33*** 0.33*** 0.25** 0.11
0.39*** 0.35*** 0.34*** 0.18*
Note: guard, guarding; touch, touching; sound, sounds; face, facial expression. Worst pain = VAS rating of most pain experienced during the physical examination; drawing = examiner’s rating of whether pain drawing revealed an anatomical (1) or a non-anatomical (2) pattern. * p⬍0.05; ** p⬍0.01; *** p⬍0.001. a
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
603
Table 4 Correlations between examiners’ ratings of medically-incongruent signs and pain behavioursa
Axial loading Simulated rotation Superficial tenderness Distracted SLR Sensory disturbance Motor disturbance Total signs a *
p⬍0.05;
**
p⬍0.01;
Guard
Touch
Words
Sound
Face
0.07 0.13 0.29*** 0.17* 0.04 0.32*** 0.31***
0.20** 0.12 0.25** 0.14 ⫺0.03 0.30*** 0.29***
0.03 0.17* 0.17* 0.18* 0.07 0.25** 0.27***
0.09 0.10 0.19* 0.09 ⫺0.05 0.17* 0.18*
0.14 0.21** 0.23** 0.20** 0.08 0.31*** 0.35***
***
p⬍0.001.
presented in Table 4. Superficial tenderness and motor disturbance showed consistently significant associations with each of the pain behaviours, with correlations ranging from low to moderate in size. Of the observed pain behaviours, use of pain words and facial expression showed the most frequent significant relationships to medically-incongruent signs, with five of the seven being significant, while pain sounds showed the least frequent, with only three of the seven being significant. Table 4 also presents the Pearson correlations between the individual pain behaviours and an index consisting of the sum of positive medically-incongruent signs. All of these correlations were significant. Table 5 presents Pearson correlations between the observers’ ratings of the five pain behaviours and the examiners’ global ratings of the intensity of patients’ display of each of the behaviours. All were statistically significant, with the highest intercorrelations for facial expression and guarding. Despite this, the majority of the relationships were moderate in magnitude, with only those for facial expression sharing 25% or more common variance. 4. Discussion The present study had two objectives. One was to evaluate the concurrent validity of a technique for quantifying pain behaviours in real time. In the course of this analysis it was also expected that the findings would shed light on the utility and nature of different components of pain behaviour. The second was to examine the degree to which judgements of clinicians familiar with the concepts of pain behaviour, but lacking training in its observation, would be comparable to measurements provided by trained observers. Table 5 Correlations of examiners’ ratings and observers’ coding of pain behavioursa Guard
Touch
Sound
Words
Face
0.36***
0.33***
0.18*
0.46***
0.52***
a *
p⬍0.05;
***
p⬍0.001.
604
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Reliability and validity are fundamental properties of a useful measurement system. In a previous study, we have demonstrated that the inter-observer reliability of the PBOS was acceptable (Prkachin et al., 2001). We also examined the test–retest reliability of the five categories of pain behaviour over a two- to three-day interval. Although all the behaviours were correlated significantly over the study period, only facial expression (r=0.71), guarding (r=0.54), and sounds (r=0.53) showed a sufficient degree of stability to be considered acceptable. The findings of the present study support the concurrent validity of the PBOS. To the extent that components of the PBOS reflect variations in pain, one would expect them to be correlated with alternative measures that are also likely to reflect pain. All of the PBOS pain behaviours were reliably correlated in the expected direction with two self-report measures of patients’ pain. In addition, higher rates of all of the PBOS behaviours were associated with the amount of time it took patients to complete the 500 m timed walk, a finding one would expect if pain, or pain behaviour, reduces one’s ability to walk at a brisk pace. With respect to the pain drawing, only facial expression was reliably associated with higher scores, reflecting a judgement that the patient’s depiction of their pain was ‘non-anatomical’. Although it is to be expected that this correlation would be attenuated, given that it represented an association between a binary and a continuous variable, it was nevertheless too low to allow for any substantial conclusions to be drawn. In general, then, PBOS pain behaviours varied in a systematic and expected manner with other widely used measures of pain. Each PBOS behaviour was correlated significantly with the sum of medically-incongruent pain signs. The individual signs that were correlated with pain behaviours most reliably occurred during testing for superficial tenderness and motor disturbances and the pain behaviour that was most consistently associated with medically-incongruent signs was facial expression. In light of the common interpretation of these signs as markers of psychological distress, it would be tempting to speculate that the behaviours measured by the PBOS also reflect similar processes. Such an interpretation would, however, be problematic. The overall magnitude of the correlations with medically-incongruent signs was modest. The strongest relationship, between facial expression and the sum of incongruent signs, suggested a shared variance of 12%. Though statistically significant due to the relatively large sample size, this value does not seem large enough to justify strong conclusions about conceptual overlap. The most appropriate conclusion would seem to be that these findings simply support the concurrent validity of the PBOS measures. It might be argued that the present findings suggest that the PBOS reflects a single overall dimension of pain behaviour and, thus, support a unitary pain behaviour construct. Indeed, the data in Table 2, indicating that all of the PBOS behaviours were significantly intercorrelated and load on a single principal component, are consistent with a unitary view. Simple intercorrelations of pain behaviours are not, however, sufficient to confirm such an interpretation. For one thing, our previous study (Prkachin et al., 2001) revealed substantial differences between the behaviours in their test–retest reliability, suggesting that the different behaviours have different determinants. For another, the patterns of relationships between PBOS behaviours and concurrent measures suggest that there are differences in the properties of the component behaviours. For example, facial expression showed more consistent relationships to the concurrent measures, while facial expression and guarding tended to show stronger relationships to the concurrent measures than the other pain behaviours. While these differing properties of the component behaviours of the PBOS suggest that they have different determinants, only further research can clarify what those
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
605
might be. As noted above, Prkachin (1986) has emphasized that facial expression may be understood in terms of adapted communicative functions. Vlaeyen and Linton (2000) have suggested that guarding may be understood as a consequence of fear of pain. Studies of the implications of these interpretations, and further analysis of the functions of pain words, sounds, and other forms of pain behaviour may add depth to our understanding of their organization. The present findings suggest that facial expression and guarding show more promise to yield useful information in clinical and epidemiological studies than the other behaviours. It should be noted that facial expression displayed more robust associations with concurrent measures than has typically been the case for a comparable measure, grimacing, in previous research, whereas the findings for guarding were relatively consistent with previous studies (Keefe & Block, 1982; Keefe et al., 1986; Keefe, Wilkins, & Cook, 1984; McDaniel et al., 1986). This probably reflects the adjustments made in the PBOS to improve measurement of facial expressions of pain — the use of an operational description that is consistent with the results of fine-grained analyses of facial behaviour, recognition of the relatively short duration of pain expressions, and use of a scale to capture the intensity dimension. As Table 5 shows, with the exception of sounds, PBOS behaviours were associated with the physical examiners’ independent ratings of the same behaviours at high levels of statistical significance. This also supports the concurrent validity of the PBOS. Examination of the pattern of correlations suggests that facial expression and words were the principal cues to which examiners were sensitive when formulating their ratings. The findings shed light on the ability of clinicians who have an understanding of the concept of pain behaviour to provide accurate ratings. On the one hand, the consistency and high degree of statistical significance of the relationships between direct measurements and clinicians’ judgements might be taken to imply that the clinicians were quite adept at grading patients’ pain behaviour. The degree of statistical significance, however, was largely a consequence of the relatively large sample size. If amount of shared variance is taken as the criterion, a more sobering conclusion must be reached, because only 3–27% of the variance in observers’ ratings was predictable from the direct measurements. Thus, although the findings demonstrate a certain sensitivity to variations in pain behaviour among examiners, the absolute level leaves considerable room for improvement. It is noteworthy that the examiners were highly trained, followed a standardized protocol, were familiar with the concept of pain behaviour and sensitized to its definitions, and were participating in a research investigation. These highly favourable conditions are unlikely to apply to the same degree outside research settings and, thus, the present findings probably approximate the best that could be expected in terms of examiners’ abilities to gauge pain behaviours. The performance of the average clinician, regardless of discipline, is likely to be poorer as conditions depart from those employed in this study. The latter observation is of some importance, given the large number of examinations of lowback pain patients that take place every year and their common use in medico-legal contexts. Evidence about pain behaviour is often noted in such settings. Nevertheless, it is frequently not clear that examiners have training that would allow them to gauge pain behaviour reliably or that they perform examinations with sufficient standardization to meet standards of acceptable of measurement. Consequently, the findings of the present study suggest that the development of training protocols to enhance clinicians’ abilities to assess pain behaviours reliably would be an important goal for future research.
606
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Acknowledgements Supported by a grant from the Workers’ Compensation Boards of British Columbia and Alberta (I. Schultz, Principal Investigator). Portions of this paper were presented at the 2001 Annual Meeting of the Society of Behavioral Medicine, Seattle, WA. We would like to extend our thanks to the behavioural coders who collected data for the present study. References Cinciripini, P. M., & Floreen, A. (1983). An assessment of chronic pain behavior in a structured interview. Journal of Psychosomatic Research, 27, 117–123. Craig, K. D., Prkachin, K. M., & Grunau, R. V. E. (2001). The facial expression of pain. In D. C. Turk, & R. Melzack (Eds.), Handbook of pain assessment. New York: Guilford. Follick, M. J., Ahern, D. K., & Aberger, E. W. (1985). Development of an audiovisual taxonomy of pain behavior: Reliability and discriminant validity. Health Psychology, 4, 555–568. Fordyce, W. E. (1976). Behavioral methods for chronic pain and illness. St. Louis: Mosby. Hunt, D. G., Zuberbier, O. A., Kozlowski, A. J., Robinson, J., Berkowitz, J., Schultz, I., Milner, R., Crook, J. & Turk, D. C. (in press). Reliability of the lumbar flexion, lumbar extension and passive straight leg raise test in normal populations embedded within a complete physical examination. Spine. Jensen, M. P. (1997). Validity of self-report and observational measures. In T. S. Jensen, J. A. Turner, & Z. WiesenfeldHanlin (Eds.), Proceedings of the Eighth World Congress on Pain: Progress in Pain Research and Management, 8 (pp. 637–661). Seattle: IASP Press. Karas, R., McIntosh, G., Hall, H., Wilson, L., & Melles, T. (1997). The relationship between nonorganic signs and centralization of symptoms in the prediction of return to work for patients with low back pain. Physical Therapy, 77, 354–360. Keefe, F. J., & Block, A. R. (1982). Development of an observational method for assessing pain behavior in chronic low back pain patients. Behavior Therapy, 13, 363–375. Keefe, F. J., & Williams, D. A. (1992). Assessment of pain behaviors. In D. C. Turk, & R. Melzack (Eds.), Handbook of pain assessment (pp. 277–292). New York: Guilford. Keefe, F. J., Crisson, J. E., Maltbie, A., Bradley, L., & Gil, K. M. (1986). Illness behavior as a predictor of pain and overt behavior patterns in chronic low back pain patients. Journal of Psychosomatic Research, 30, 543–551. Keefe, F. J., Crisson, J. E., & Snipes, M. T. (1987). Observational methods for assessing pain: A practical guide. In J. A. Blumenthal, & D. C. McKee (Eds.), Applications in behavioral medicine and health psychology: A clinician’s source book (pp. 39–65). Sarasota, FL: Professional Resource Exchange. Keefe, F. J., Wilkins, R. H., & Cook, W. A. (1984). Direct observation of pain behavior in low back pain patients during physical examination. Pain, 20, 59–68. Liebeskind, J. J., & Paul, L. A. (1977). Psychological and physiological mechanisms of pain. Annual Review of Psychology, 28, 41–60. Main, C. J., & Waddell, G. (1998). Behavioral responses to examination: A reappraisal of the interpretation of “nonorganic signs”. Spine, 23, 2367–2371. McDaniel, L. K., Anderson, K. O., Bradley, L. A., Young, L. D., Turner, R. A., Agudelo, C. A., & Keefe, F. J. (1986). Development of an observation method for assessing pain behavior in rheumatoid arthritis patients. Pain, 24, 165–184. Prkachin, K. M. (1992). The consistency of facial expressions of pain: A comparison across modalities. Pain, 51, 297–306. Prkachin, K. M. (1986). Pain behaviour is not unitary. Behavioral and Brain Sciences, 9, 754–755. Prkachin, K. M., & Mercer, S. (1989). Pain expression in patients with shoulder pathology: Validity, coding properties and relation to sickness impact. Pain, 39, 257–265. Prkachin, K. M., & Craig, K. D. (1994). Expressing pain: The communication and interpretation of facial pain signals. Journal of Nonverbal Behavior, 19, 191–205.
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
607
Prkachin, K. M., Hughes, E. L., Schultz, I., & Joy, P. (2001). Real-time assessment of pain behavior during clinical assessment of low-back pain patients. Pain (in press). Ransford, A. O., Cairns, D., & Mooney, V. (1976). The pain drawing as an aid to the psychological evaluation of patients with low back pain. Spine, 1, 127–134. Reesor, K. A., & Craig, K. D. (1988). Medically incongruent chronic back pain: Physical limitations, suffering and ineffective coping. Pain, 32, 35–45. Richards, J. S., Nepomuceno, C., Riles, M., & Suer, Z. (1982). Assessing pain behavior: The UAB Pain Behavior Scale. Pain, 14, 393–398. Scalzitti, D. A. (1997). Screening for psychological factors in patients with low back problems: Waddell’s nonorganic signs. Physical Therapy, 77, 306–312. Solomon, P. E., Prkachin, K. M., & Farewell, V. (1997). Enhancing sensitivity to facial expressions of pain. Pain, 71, 279–284. Turk, D. C., Wack, J. T., & Kerns, R. D. (1985). An empirical examination of the “pain-behavior” construct. Journal of Behavioral Medicine, 8, 119–130. van Hoof, J. A. R. A. M. (1972). A comparative approach to the phylogeny of laughter and smiling. In R. A. Hinde (Ed.), Non-verbal communication (pp. 209–241). Cambridge, UK: Cambridge University Press. Vlaeyen, J. W. S., & Linton, S. J. (2000). Fear-avoidance and its consequences in chronic musculoskeletal pain: A state of the art. Pain, 85, 317–332. Vlaeyen, J. W. S., Pernot, D. F., Kole-Snijders, A. M., Schuerman, J. A., van Eek, H., & Groenman, N. H. (1990). Assessment of the components of observed chronic pain behavior: The Checklist for Interpersonal Pain Behavior (CHIP). Pain, 43, 337–347. Vlaeyen, J. W., Van Eek, H., Groenman, N. H., & Schuerman, J. A. (1987). Dimensions and components of observed chronic pain behavior. Pain, 31, 65–75. Waddell, G. (1991). Occupational low-back pain, illness behavior and disability. Spine, 16, 683–685. Waddell, G., Pilowsky, I., & Bond, M. R. (1989). Clinical assessment and interpretation of abnormal illness behavior in low back pain. Pain, 39, 41–53. Waddell, G., McCulloch, J. A., Kummel, E., & Venner, R. M. (1980). Nonorganic physical signs in low back pain. Spine, 5, 117–125. Waddell, G., & Turk, D. C. (1992). Clinical assessment of low back pain. In D. C. Turk, & R. Melzack (Eds.), Handbook of pain assessment (pp. 15–36). New York: Guilford. Werneke, M. W., Harris, D. E., & Lichter, R. L. (1993). Clinical effectiveness of behavioral signs for screening chronic low-back pain patients in a work-oriented physical rehabilitation program. Spine, 18, 2412–2418.
Behavioural assessment
Assessing pain behaviour of low-back pain patients in real time: concurrent validity and examiner sensitivity K.M. Prkachin
a, d,*
, I. Schultz b, J. Berkowitz b, E. Hughes a, D. Hunt
c
a
Department of Psychology, University of Northern British Columbia, 3333 University Way, Prince George, BC, Canada V2N 4Z9 b Department of Educational and Counselling Psychology and Special Education, University of British Columbia, 2329 West Mall, Vancouver, BC, Canada V6T 1Z4 c Department of Family Practice, University of British Columbia, 2329 West Mall, Vancouver, BC, Canada, V6T 1ZT d Workers’ Compensation Board of British Columbia, 6951 Westminster Highway, Richmond, BC Canada V7C 1C6 Received 18 April 2001; accepted 5 June 2001
Abstract Several systems for measuring pain behaviour have been developed for clinical settings. The present study reports on a real-time system for coding five categories of pain behaviour for low-back pain patients: guarding, touching, sounds, words, and facial expression. Unique features of the system are the use of refined measures of facial expression and integration of the measurements with a standardized physical examination. 176 sub-acute and chronic low-back pain patients underwent a physical examination while their pain behaviour was coded. Concurrent measures of subjective pain, medically-incongruent signs, and independent global ratings of pain behaviour were taken. Analyses indicated that the pain behaviours, particularly guarding and facial expression, varied systematically with the alternative measures, supporting the concurrent validity of the behaviour observation system. While pain behaviours, especially use of words and facial expressions, were significantly associated with the examiners’ independent ratings, the strength of the associations suggested that, in the absence of direct training, examiners’ performance was relatively poor. Implications for training of clinicians in detecting pain behaviour are discussed. 2002 Elsevier Science Ltd. All rights reserved. Keywords: Pain; Behaviour; Facial expression; Guarding; Low-back pain; Patients; Validity
* Corresponding author. Tel.: +1-250-960-6633; fax: +1-250-960-5536. E-mail address: [email protected] (K.M. Prkachin).
0005-7967/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 5 - 7 9 6 7 ( 0 1 ) 0 0 0 7 5 - 4
596
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
1. Introduction The purpose of this paper is to describe properties of a system for observing pain behaviour in real time. The Pain Behaviour Observation System (PBOS; Prkachin, Hughes, Schultz, & Joy, 2001) integrates observations of pain behaviour with a standardized physical examination for lowback pain patients (Hunt et al., in press). Observers code five behaviours: guarding, touching, and the use of words, sounds, and facial expression to communicate pain. People in pain often show behavioural changes that are quite distinctive to observers whose attention happens to be drawn to the sufferer. In turn, changes in sufferers’ behaviour lead observers to engage in a number of processes (Prkachin & Craig, 1994), among which one is to draw inferences about the existence, or the intensity, of suffering. It can be argued that all of pain science and management is based on the process of observers drawing inferences about sufferers’ experiences based on observations of their (the sufferers’) behaviour. Consequently, it is important to be as precise and sophisticated as possible when trying to understand the nature of pain behaviour and how to assess it. In the behavioural literature, a number of different systems for recognizing and measuring pain behaviour have been developed (Jensen, 1997). Among these, probably the most widely used was developed and refined by Keefe (Keefe & Block, 1982; Keefe & Williams, 1992). Keefe’s system recognizes five categories of pain behaviour: guarding, bracing, rubbing, sighing, and grimacing. Several other systems have been developed to quantify pain behaviours associated with different types of pain (Cinciripini & Floreen, 1983; Follick, Ahern, & Aberger, 1985; Richards, Nepomuceno, Riles, & Suer, 1982; Vlaeyen et al., 1990). Generally speaking, these systems vary in the number of behaviours incorporated into the observational procedure and the settings in which they are applied. However, they tend to reflect some common underlying assumptions, among which are that the behaviours are broadly communicative of pain intensity, that an additive accumulation of behaviours reflects greater pain, and that the behaviours resemble one another in their metric properties. A different approach to the assessment of pain behaviours has been developed by Waddell and colleagues (Waddell, 1991; Waddell, McCulloch, Kummel, & Venner, 1980). This technique, which has become increasingly popular in the pain assessment and rehabilitation field, involves evaluation of a variety of symptoms, such as reports of pain that do not follow recognized distributions based on understanding of neuroanatomy, and behavioural ‘signs’. The behavioural signs (also commonly referred to as ‘nonorganic signs’, ‘Waddell signs’, or the more neutral ‘medicallyincongruent pain’; Reesor & Craig, 1988) require exposure of the patient to manoeuvres such as testing for tenderness, exposure to tests that generally do not cause pain but give the appearance of doing so, or observing for inconsistencies in pain expression when patients’ attention is distracted. Scoring of behaviours within this system is based on classification of a sign as present or absent. There has been substantial discussion in the literature about the meaning and interpretation of these signs (Karas, McIntosh, Hall, Wilson, & Melles, 1997; Main & Waddell, 1998; Scalzitti, 1997; Werneke, Harris, & Lichter, 1993). Current opinion seems to favour the suggestion that they are markers of psychological distress. One of the objectives of the present study was to evaluate the concurrent validity of the PBOS in low-back pain patients by determining the relationship between pain behaviours and alternative indicators of patients’ pain. One set of indicators was the patient’s own reports of pain experienced
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
597
during the physical examination. Another was the technique for observing medically-incongruent signs developed by Waddell and colleagues (1980). The term ‘pain behaviour’ tends to be used as a unidimensional construct. This is probably attributable, in part, to Fordyce (1976) who first popularized its use. In Fordyce’s formulation, pain behaviour could be affected by injury or tissue damage on the one hand, or its personal and social consequences on the other. The former he referred to as ‘respondent’ and the latter ‘operant’ pain. Although this view recognized two categories of pain behaviour based on their hypothesized determinants, clinical use of the term tends not to make even this degree of differentiation. As long ago as 1977, however, Liebeskind and Paul drew attention to the fact that pain behaviours are organized in different ways, at different levels of the nervous system, implying that a more complex analysis might be necessary (Liebeskind & Paul, 1977). Turk, Wack, and Kerns (1985) used multidimensional scaling and hierarchical clustering analysis to identify the structure evident in clinicians’ ratings of 20 different pain behaviours. Their results appeared to identify four broad categories of behaviour: distorted gait/posture, negative affect, audible/visible behaviour, and activity avoidance. Vlaeyen, Van Eek, Groenman, and Schuerman (1987) identified nine components of observed pain behaviour (anxiety, attention seeking, pain complaint, general complaint, medication use, distorted posture, fatigue, insomnia, and depressive mood) loading on three overall dimensions: withdrawal–approach, arousal, and visible/audible. These observations are consistent with a more multidimensional view of pain behaviour. Different categories of pain behaviour are likely to serve different functions (Prkachin, 1986). Guarding, for example, seems likely to be a self-management behaviour reflecting, to a significant degree, active efforts on the sufferer’s part to minimize discomfort. Grimacing, on the other, is a member of a broader category of behaviour that has evolved to serve a communicative function (Prkachin, 1986; van Hoof, 1972). Thus, although these and other definable types of pain behaviour are observable and provide evidence about pain in a sufferer, there is no reason to expect that they will yield the same information, nor that they will be affected by precisely the same controlling variables. The construct of pain behaviour has permeated the professional community. Clinicians, for better or worse, use observations of pain behaviour to draw inferences about pain experience and its causal variables. Most systems for assessing pain behaviour require special examinations focused exclusively on pain behaviour. For example, Keefe and Block’s (1982) approach is based on alternating two-minute periods during which back pain patients are observed while standing, sitting, and walking. Another setting that is likely to provide information about pain and that may have advantages over contrived behavioural observation tests is the clinical examination. During clinical examinations of back pain patients, clinicians expose the sufferer to a series of challenges designed to yield diagnostic evidence. Many of these challenges will, of necessity, evoke pain and pain behaviour. The second, and chief, objective of the present study was to examine the relationship between pain behaviours as assessed by our system, and clinicians’ ratings of the same behaviours based on their impressions of the examinations. As noted above, clinicians are increasingly making use of the construct of pain behaviour, and it is not uncommon in clinical or medico-legal settings for such behaviour to be noted and used as a basis for recommendations. Frequently, however, it is not clear how systematically such observations have been made or the degree of training the clinician has had in making them. This raises the question of the validity of such judgements.
598
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Consequently, we were interested in determining how closely relatively sophisticated clinicians’ ratings corresponded with more objective measurements performed by highly trained experts. 2. Method 2.1. Participants Patients were 176 Workers’ Compensation Board of British Columbia claimants who took part in a larger study of predictors of disability. They were categorized into two groups based on the time that had elapsed since the injury leading to their currently active claim. Patients were designated as sub-acute (n=148) if their claim had taken place within the preceding 4–6 weeks, or as chronic (n=28) if their claim was between 6 and 12 months’ duration. Mean age of the sub-acute patients was 40.5 (SD=10.2), and of the chronic patients was 43.3 (SD=11.6). 126 patients were male and 50 were female. For purposes of the present study, patients were collapsed across pain type and sex. Patients were recruited from the Board’s computerized claim system based on rosters generated weekly. They were allocated at random to one of six examiners who were either Board physicians or physiotherapists. Physical examinations were observed by one of five female observers, each of whom held baccalaureate degrees in Psychology (one PhD). 2.2. Procedure Patients underwent a standardized physical examination prepared by a Board physician. Examination procedures were selected based on an extensive literature review to identify tests that the scientific evidence suggested were reliable and prognostically valid. Anthropometric measures, physical signs, flexion and extension, medically-incongruent signs, reflexes, strength, muscle bulk, sensation, tenderness, and range of motion in a variety of tests were recorded. All examiners underwent extensive training to be able to perform the examination in a standardized manner (Hunt et al., in press). 2.2.1. Pain behaviour observation system (PBOS) Each of the examinations was observed by a coder who had been trained in the PBOS. The system was designed to be applied in real time during clinical conditions. Consequently, it was necessary that observations be performed without resorting to videotape recording. We also wished to take advantage of the clinical examination as a setting for behavioural observations. The categories of the PBOS were inspired by, but modified from, Keefe and Block’s (1982) technique, to meet several further requirements or to match the conditions of the study. First, two of Keefe and Block’s categories — guarding and bracing — were collapsed into one. This was done because bracing is considered to be a form of guarding that takes place when the sufferer is not moving (Keefe, Crisson, & Snipes, 1987). Second, a category was included for pain-related verbalization. Third, a category for facial expression was developed. Although Keefe and Block included a category for grimacing in their system, their operational definition differs in significant ways from the pain-related facial expression that has been described in empirical research (Craig, Prkachin, & Grunau, 2001; Prkachin, 1992; Prkachin & Craig, 1994). Finally, in Keefe and
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
599
Block’s system, behaviours must be present for at least three seconds in order to be coded. As the duration of some pain behaviours, particularly facial expressions, is frequently substantially shorter, this rule was eliminated. To learn the system, observers participated in a day-long workshop after reading a manual describing the behaviours and recording principles. The workshop consisted of didactic instruction accompanied by a training video that displayed each of the categories of behaviour to be observed, and special considerations associated with their scoring. It also provided opportunity for observers to practice and receive feedback on their coding. A separate video was used to train observers in the detection of facial expressions of pain. The training procedure was based on the system developed and evaluated by Solomon, Farewell, and Prkachin (1997). Observers were trained to identify five categories of pain behaviour, as follows: Guarding: behaviour that prevents or alleviates pain, including stiffness, hesitation, limping, bracing, and flinching. Touching: any contact between the patient’s hands and the lower back, the hips, buttocks, and the outer aspect of the thighs. Words: any spontaneous utterance that relates to the patient’s pain. Sounds: moaning, sighing, puffing or slow exhalation of breath. Facial expression: movements of the facial musculature that narrow the palpebral fissure, tighten the eyelids and raise the cheeks, lower the eyebrows, and raise the nose. When the training workshop was complete, observers were given a further training videotape to practice coding. They were provided with feedback about their coding in the form of how many behaviours they had got correct, proportion of false positives (coding a behaviour as present when it was not) and misses (coding a behaviour as absent when it was present). Finally, observers underwent a final test of proficiency in which they coded 40 vignettes taken from patients undergoing the procedures of the physical examination. All coders achieved a score exceeding 78% on this test, with the average percent correct being 81%. Based on extensive observations of prototypes of the physical examination, a system for parsing it was developed. The entire examination was broken into 57 segments based on the tests being performed by the examiner or transitions taking place within the examination. Of those segments, behavioural coding took place in 36. Observers were trained in parsing the examination in realtime before beginning the study. Table 1 lists observed and unobserved coding epochs. Guarding, touching, words and sounds were recorded as present or absent during each of the recording epochs. Facial expressions were rated on a three-category scale (0=none, 1=some, 2=much) because evidence indicates that the information contained in facial expression is graded and that discriminating power is lost if this is not taken into consideration (Prkachin & Mercer, 1989). Coders positioned themselves inside the examination room where they could maintain a complete view of the patient’s face while not interfering with the progress of the examination. The observers made notations on a recording sheet. 2.2.2. Evaluation of medically-incongruent signs Medically-incongruent signs were evaluated according to the techniques described by Waddell and colleagues (Waddell et al., 1980; Waddell, Pilowsky, & Bond, 1989; Waddell & Turk, 1992).
600
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Table 1 Discrete epochs of the physical examination (coding epochs in italics) Interval 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.
to sitting 1 introduction to scale weight/height to landmarks landmarks lordosis heel raise 1 and 2 forward/back rotation side-to-side stand for measurement lumbar extension 1 lumbar flexion 1 lumbar extension 2 lumbar flexion 2 lumbar extension 3 lumbar flexion 3 lumbar extension 4 lumbar flexion 4 lumbar extension 5 lumbar flexion 5 lateral flexion left 1 lateral flexion right 1 lateral flexion left 2 lateral flexion right 2 lateral flexion left 3 lateral flexion right 3 lateral flexion left 4
30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57.
lateral flexion right 4 lateral flexion left 5 lateral flexion right 5 axial compression simulated trunk rotation to kneeling ankle reflexes to sitting 2 knee reflexes knee extension 1 knee extension 2 muscle strength 1 muscle strength 2 to supine 1 ankle dorsiflexion toe extensor muscle bulk measure sensation passive straight leg raise R passive straight leg raise L to prone palpation McKenzie push-up prone active extension to supine 2 active situp bilateral straight leg raise to standing
Their assessment was based on the following tests, which were performed by the examiners as part of the physical examination: Axial loading: the examiner applied 0.5–1 kg pressure with his or her hands to the top of the patient’s head. The test was considered to show a positive sign if the patient complained of resultant pain in the lumbar region. Simulated trunk rotation: from behind the patient, the examiner held the participant’s arms to the side, while rotating their hips, making sure the pelvis and shoulders rotated together. A positive sign was recorded if the patient complained of pain in the lumbar region. Superficial tenderness: the examiner applied light pinching in a standardized fashion across both sides of the lower back. A positive sign was recorded if the patient complained of tenderness over a wide area of the lumbar skin. Distraction on straight-leg raising: the examiner raised the left and right leg in the supine and
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
601
seated positions, noting the maximum range of motion. During the seated test, the examiner distracted the patient. An improvement in range of motion during distraction was taken as a positive sign. Sensory disturbance: the examiner tested for sensation to light touch or pinprick on both legs. Altered sensation involving the whole upper or lower leg, or a stocking-like distribution were taken as a positive sign. Motor disturbance: the examiner noted the presence of ‘jerky’ giving way of the lower limbs during testing. 2.2.3. Pain drawing, timed walk, and visual analogue scale measures As part of the examination, participants also completed a pain drawing (Ransford, Cairns, & Mooney, 1976). Participants were presented with an anatomical drawing, with front and back views, of the human form. They were asked to mark on the drawing areas of numbness, pins and needles, burning, stabbing, and aching. The examiner evaluated the drawing for evidence of poorly localized, widespread, or non-anatomical sensations and evidence of written comments documenting severity, and rendered a decision as to whether the drawing was largely ‘anatomical’ or ‘non-anatomical’. At the end of the examination, patients underwent a timed walk. Instructions for the test indicated that the participant was to walk as briskly as possible. The examiner accompanied the patient, taking care not to provide encouragement or to give feedback as to the elapsed time. The walk was conducted over a distance of 500 m on a flat surface. Time to complete the walk was recorded in seconds. At the end of the examination, the patients also made ratings of their pain on 100 cm visual analogue scales (VAS), anchored with the words no pain and worst possible pain. Patients provided two ratings, one of their current pain and one of the maximum pain experienced during the examination. VAS ratings were measured to the nearest 10 cm. 2.2.4. Examiner’s rating of pain behaviour At the end of each examination, the examiner made an independent rating of the intensity of each of the five categories of pain behaviour represented on the PBOS. Examiners had not been given training in the detection of pain behaviours; however, they were all familiar with the concept and had been provided with the same operational definitions that the observers worked from when making their observations.
3. Results Table 2 presents intercorrelations among the total PBOS pain behaviour scores and their means and standard deviations. All the pain behaviours were positively associated with one another at high levels of significance. The magnitudes of the intercorrelations ranged from moderate to high. A principal components analysis (with Varimax rotation) of these intercorrelations resulted in a single factor accounting for 46% of the common variance. Factor loadings are also presented in Table 2.
602
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Table 2 Intercorrelations among PBOS total behaviour scores, means and standard deviations, and loadings on the single principal componenta Guard Guard Touch Words Sound Face Mean SD
6.34 5.85
Touch
Words
Sound
Face
PC
0.45
0.44 0.31
0.64 0.34 0.55
0.41 0.36 0.37 0.58
0.71 0.48 0.64 0.82 0.67
1.47 2.01
4.12 4.14
9.76 7.11
12.95 10.62
a Note: guard, guarding; touch, touching; sound, sounds; face, facial expression; PC, principal component. All values reflect counts of coding intervals in which the behaviour occurred, except for facial expression, which represents the sum of three-point intensity codes across coding intervals. All p⬍0.001.
Correlations between patients’ subjective pain ratings, walking time, the pain drawing measure, and observed pain behaviours are presented in Table 3. Observers’ ratings of each of the five categories of pain behaviour were significantly and positively related to patients’ ratings of both their current pain and the worst pain experienced during the examination. Of the different pain behaviours observed, words bore the weakest and guarding the strongest relationship with patients’ pain reports. All the observed pain behaviours were significantly positively correlated with the time they took to complete the 500 m walk. Whether the patient’s pain drawing appeared to be anatomical or nonanatomical was only significantly related to facial expression. In most cases, the relationships between pain behaviours and the concurrent measures of pain report, walking time, and nature of pain drawing, though significant, were weak to modest. Facial expression bore the most consistent pattern of relationships with the other indices while guarding showed the strongest overall relationships with pain reports. Point biserial correlations were calculated between the examiners’ binary ratings of the six signs of medically-incongruent pain and the observed pain behaviours. These correlations are
Table 3 Correlations between patients’ verbal pain ratings and observers’ pain behaviour recordingsa
Current pain Worst pain Walk time (s) Drawing
Guard
Touch
Words
Sound
Face
0.43*** 0.41*** 0.26** 0.07
0.34*** 0.30*** 0.34*** 0.09
0.24** 0.29*** 0.24** 0.00
0.33*** 0.33*** 0.25** 0.11
0.39*** 0.35*** 0.34*** 0.18*
Note: guard, guarding; touch, touching; sound, sounds; face, facial expression. Worst pain = VAS rating of most pain experienced during the physical examination; drawing = examiner’s rating of whether pain drawing revealed an anatomical (1) or a non-anatomical (2) pattern. * p⬍0.05; ** p⬍0.01; *** p⬍0.001. a
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
603
Table 4 Correlations between examiners’ ratings of medically-incongruent signs and pain behavioursa
Axial loading Simulated rotation Superficial tenderness Distracted SLR Sensory disturbance Motor disturbance Total signs a *
p⬍0.05;
**
p⬍0.01;
Guard
Touch
Words
Sound
Face
0.07 0.13 0.29*** 0.17* 0.04 0.32*** 0.31***
0.20** 0.12 0.25** 0.14 ⫺0.03 0.30*** 0.29***
0.03 0.17* 0.17* 0.18* 0.07 0.25** 0.27***
0.09 0.10 0.19* 0.09 ⫺0.05 0.17* 0.18*
0.14 0.21** 0.23** 0.20** 0.08 0.31*** 0.35***
***
p⬍0.001.
presented in Table 4. Superficial tenderness and motor disturbance showed consistently significant associations with each of the pain behaviours, with correlations ranging from low to moderate in size. Of the observed pain behaviours, use of pain words and facial expression showed the most frequent significant relationships to medically-incongruent signs, with five of the seven being significant, while pain sounds showed the least frequent, with only three of the seven being significant. Table 4 also presents the Pearson correlations between the individual pain behaviours and an index consisting of the sum of positive medically-incongruent signs. All of these correlations were significant. Table 5 presents Pearson correlations between the observers’ ratings of the five pain behaviours and the examiners’ global ratings of the intensity of patients’ display of each of the behaviours. All were statistically significant, with the highest intercorrelations for facial expression and guarding. Despite this, the majority of the relationships were moderate in magnitude, with only those for facial expression sharing 25% or more common variance. 4. Discussion The present study had two objectives. One was to evaluate the concurrent validity of a technique for quantifying pain behaviours in real time. In the course of this analysis it was also expected that the findings would shed light on the utility and nature of different components of pain behaviour. The second was to examine the degree to which judgements of clinicians familiar with the concepts of pain behaviour, but lacking training in its observation, would be comparable to measurements provided by trained observers. Table 5 Correlations of examiners’ ratings and observers’ coding of pain behavioursa Guard
Touch
Sound
Words
Face
0.36***
0.33***
0.18*
0.46***
0.52***
a *
p⬍0.05;
***
p⬍0.001.
604
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Reliability and validity are fundamental properties of a useful measurement system. In a previous study, we have demonstrated that the inter-observer reliability of the PBOS was acceptable (Prkachin et al., 2001). We also examined the test–retest reliability of the five categories of pain behaviour over a two- to three-day interval. Although all the behaviours were correlated significantly over the study period, only facial expression (r=0.71), guarding (r=0.54), and sounds (r=0.53) showed a sufficient degree of stability to be considered acceptable. The findings of the present study support the concurrent validity of the PBOS. To the extent that components of the PBOS reflect variations in pain, one would expect them to be correlated with alternative measures that are also likely to reflect pain. All of the PBOS pain behaviours were reliably correlated in the expected direction with two self-report measures of patients’ pain. In addition, higher rates of all of the PBOS behaviours were associated with the amount of time it took patients to complete the 500 m timed walk, a finding one would expect if pain, or pain behaviour, reduces one’s ability to walk at a brisk pace. With respect to the pain drawing, only facial expression was reliably associated with higher scores, reflecting a judgement that the patient’s depiction of their pain was ‘non-anatomical’. Although it is to be expected that this correlation would be attenuated, given that it represented an association between a binary and a continuous variable, it was nevertheless too low to allow for any substantial conclusions to be drawn. In general, then, PBOS pain behaviours varied in a systematic and expected manner with other widely used measures of pain. Each PBOS behaviour was correlated significantly with the sum of medically-incongruent pain signs. The individual signs that were correlated with pain behaviours most reliably occurred during testing for superficial tenderness and motor disturbances and the pain behaviour that was most consistently associated with medically-incongruent signs was facial expression. In light of the common interpretation of these signs as markers of psychological distress, it would be tempting to speculate that the behaviours measured by the PBOS also reflect similar processes. Such an interpretation would, however, be problematic. The overall magnitude of the correlations with medically-incongruent signs was modest. The strongest relationship, between facial expression and the sum of incongruent signs, suggested a shared variance of 12%. Though statistically significant due to the relatively large sample size, this value does not seem large enough to justify strong conclusions about conceptual overlap. The most appropriate conclusion would seem to be that these findings simply support the concurrent validity of the PBOS measures. It might be argued that the present findings suggest that the PBOS reflects a single overall dimension of pain behaviour and, thus, support a unitary pain behaviour construct. Indeed, the data in Table 2, indicating that all of the PBOS behaviours were significantly intercorrelated and load on a single principal component, are consistent with a unitary view. Simple intercorrelations of pain behaviours are not, however, sufficient to confirm such an interpretation. For one thing, our previous study (Prkachin et al., 2001) revealed substantial differences between the behaviours in their test–retest reliability, suggesting that the different behaviours have different determinants. For another, the patterns of relationships between PBOS behaviours and concurrent measures suggest that there are differences in the properties of the component behaviours. For example, facial expression showed more consistent relationships to the concurrent measures, while facial expression and guarding tended to show stronger relationships to the concurrent measures than the other pain behaviours. While these differing properties of the component behaviours of the PBOS suggest that they have different determinants, only further research can clarify what those
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
605
might be. As noted above, Prkachin (1986) has emphasized that facial expression may be understood in terms of adapted communicative functions. Vlaeyen and Linton (2000) have suggested that guarding may be understood as a consequence of fear of pain. Studies of the implications of these interpretations, and further analysis of the functions of pain words, sounds, and other forms of pain behaviour may add depth to our understanding of their organization. The present findings suggest that facial expression and guarding show more promise to yield useful information in clinical and epidemiological studies than the other behaviours. It should be noted that facial expression displayed more robust associations with concurrent measures than has typically been the case for a comparable measure, grimacing, in previous research, whereas the findings for guarding were relatively consistent with previous studies (Keefe & Block, 1982; Keefe et al., 1986; Keefe, Wilkins, & Cook, 1984; McDaniel et al., 1986). This probably reflects the adjustments made in the PBOS to improve measurement of facial expressions of pain — the use of an operational description that is consistent with the results of fine-grained analyses of facial behaviour, recognition of the relatively short duration of pain expressions, and use of a scale to capture the intensity dimension. As Table 5 shows, with the exception of sounds, PBOS behaviours were associated with the physical examiners’ independent ratings of the same behaviours at high levels of statistical significance. This also supports the concurrent validity of the PBOS. Examination of the pattern of correlations suggests that facial expression and words were the principal cues to which examiners were sensitive when formulating their ratings. The findings shed light on the ability of clinicians who have an understanding of the concept of pain behaviour to provide accurate ratings. On the one hand, the consistency and high degree of statistical significance of the relationships between direct measurements and clinicians’ judgements might be taken to imply that the clinicians were quite adept at grading patients’ pain behaviour. The degree of statistical significance, however, was largely a consequence of the relatively large sample size. If amount of shared variance is taken as the criterion, a more sobering conclusion must be reached, because only 3–27% of the variance in observers’ ratings was predictable from the direct measurements. Thus, although the findings demonstrate a certain sensitivity to variations in pain behaviour among examiners, the absolute level leaves considerable room for improvement. It is noteworthy that the examiners were highly trained, followed a standardized protocol, were familiar with the concept of pain behaviour and sensitized to its definitions, and were participating in a research investigation. These highly favourable conditions are unlikely to apply to the same degree outside research settings and, thus, the present findings probably approximate the best that could be expected in terms of examiners’ abilities to gauge pain behaviours. The performance of the average clinician, regardless of discipline, is likely to be poorer as conditions depart from those employed in this study. The latter observation is of some importance, given the large number of examinations of lowback pain patients that take place every year and their common use in medico-legal contexts. Evidence about pain behaviour is often noted in such settings. Nevertheless, it is frequently not clear that examiners have training that would allow them to gauge pain behaviour reliably or that they perform examinations with sufficient standardization to meet standards of acceptable of measurement. Consequently, the findings of the present study suggest that the development of training protocols to enhance clinicians’ abilities to assess pain behaviours reliably would be an important goal for future research.
606
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
Acknowledgements Supported by a grant from the Workers’ Compensation Boards of British Columbia and Alberta (I. Schultz, Principal Investigator). Portions of this paper were presented at the 2001 Annual Meeting of the Society of Behavioral Medicine, Seattle, WA. We would like to extend our thanks to the behavioural coders who collected data for the present study. References Cinciripini, P. M., & Floreen, A. (1983). An assessment of chronic pain behavior in a structured interview. Journal of Psychosomatic Research, 27, 117–123. Craig, K. D., Prkachin, K. M., & Grunau, R. V. E. (2001). The facial expression of pain. In D. C. Turk, & R. Melzack (Eds.), Handbook of pain assessment. New York: Guilford. Follick, M. J., Ahern, D. K., & Aberger, E. W. (1985). Development of an audiovisual taxonomy of pain behavior: Reliability and discriminant validity. Health Psychology, 4, 555–568. Fordyce, W. E. (1976). Behavioral methods for chronic pain and illness. St. Louis: Mosby. Hunt, D. G., Zuberbier, O. A., Kozlowski, A. J., Robinson, J., Berkowitz, J., Schultz, I., Milner, R., Crook, J. & Turk, D. C. (in press). Reliability of the lumbar flexion, lumbar extension and passive straight leg raise test in normal populations embedded within a complete physical examination. Spine. Jensen, M. P. (1997). Validity of self-report and observational measures. In T. S. Jensen, J. A. Turner, & Z. WiesenfeldHanlin (Eds.), Proceedings of the Eighth World Congress on Pain: Progress in Pain Research and Management, 8 (pp. 637–661). Seattle: IASP Press. Karas, R., McIntosh, G., Hall, H., Wilson, L., & Melles, T. (1997). The relationship between nonorganic signs and centralization of symptoms in the prediction of return to work for patients with low back pain. Physical Therapy, 77, 354–360. Keefe, F. J., & Block, A. R. (1982). Development of an observational method for assessing pain behavior in chronic low back pain patients. Behavior Therapy, 13, 363–375. Keefe, F. J., & Williams, D. A. (1992). Assessment of pain behaviors. In D. C. Turk, & R. Melzack (Eds.), Handbook of pain assessment (pp. 277–292). New York: Guilford. Keefe, F. J., Crisson, J. E., Maltbie, A., Bradley, L., & Gil, K. M. (1986). Illness behavior as a predictor of pain and overt behavior patterns in chronic low back pain patients. Journal of Psychosomatic Research, 30, 543–551. Keefe, F. J., Crisson, J. E., & Snipes, M. T. (1987). Observational methods for assessing pain: A practical guide. In J. A. Blumenthal, & D. C. McKee (Eds.), Applications in behavioral medicine and health psychology: A clinician’s source book (pp. 39–65). Sarasota, FL: Professional Resource Exchange. Keefe, F. J., Wilkins, R. H., & Cook, W. A. (1984). Direct observation of pain behavior in low back pain patients during physical examination. Pain, 20, 59–68. Liebeskind, J. J., & Paul, L. A. (1977). Psychological and physiological mechanisms of pain. Annual Review of Psychology, 28, 41–60. Main, C. J., & Waddell, G. (1998). Behavioral responses to examination: A reappraisal of the interpretation of “nonorganic signs”. Spine, 23, 2367–2371. McDaniel, L. K., Anderson, K. O., Bradley, L. A., Young, L. D., Turner, R. A., Agudelo, C. A., & Keefe, F. J. (1986). Development of an observation method for assessing pain behavior in rheumatoid arthritis patients. Pain, 24, 165–184. Prkachin, K. M. (1992). The consistency of facial expressions of pain: A comparison across modalities. Pain, 51, 297–306. Prkachin, K. M. (1986). Pain behaviour is not unitary. Behavioral and Brain Sciences, 9, 754–755. Prkachin, K. M., & Mercer, S. (1989). Pain expression in patients with shoulder pathology: Validity, coding properties and relation to sickness impact. Pain, 39, 257–265. Prkachin, K. M., & Craig, K. D. (1994). Expressing pain: The communication and interpretation of facial pain signals. Journal of Nonverbal Behavior, 19, 191–205.
K.M. Prkachin et al. / Behaviour Research and Therapy 40 (2002) 595–607
607
Prkachin, K. M., Hughes, E. L., Schultz, I., & Joy, P. (2001). Real-time assessment of pain behavior during clinical assessment of low-back pain patients. Pain (in press). Ransford, A. O., Cairns, D., & Mooney, V. (1976). The pain drawing as an aid to the psychological evaluation of patients with low back pain. Spine, 1, 127–134. Reesor, K. A., & Craig, K. D. (1988). Medically incongruent chronic back pain: Physical limitations, suffering and ineffective coping. Pain, 32, 35–45. Richards, J. S., Nepomuceno, C., Riles, M., & Suer, Z. (1982). Assessing pain behavior: The UAB Pain Behavior Scale. Pain, 14, 393–398. Scalzitti, D. A. (1997). Screening for psychological factors in patients with low back problems: Waddell’s nonorganic signs. Physical Therapy, 77, 306–312. Solomon, P. E., Prkachin, K. M., & Farewell, V. (1997). Enhancing sensitivity to facial expressions of pain. Pain, 71, 279–284. Turk, D. C., Wack, J. T., & Kerns, R. D. (1985). An empirical examination of the “pain-behavior” construct. Journal of Behavioral Medicine, 8, 119–130. van Hoof, J. A. R. A. M. (1972). A comparative approach to the phylogeny of laughter and smiling. In R. A. Hinde (Ed.), Non-verbal communication (pp. 209–241). Cambridge, UK: Cambridge University Press. Vlaeyen, J. W. S., & Linton, S. J. (2000). Fear-avoidance and its consequences in chronic musculoskeletal pain: A state of the art. Pain, 85, 317–332. Vlaeyen, J. W. S., Pernot, D. F., Kole-Snijders, A. M., Schuerman, J. A., van Eek, H., & Groenman, N. H. (1990). Assessment of the components of observed chronic pain behavior: The Checklist for Interpersonal Pain Behavior (CHIP). Pain, 43, 337–347. Vlaeyen, J. W., Van Eek, H., Groenman, N. H., & Schuerman, J. A. (1987). Dimensions and components of observed chronic pain behavior. Pain, 31, 65–75. Waddell, G. (1991). Occupational low-back pain, illness behavior and disability. Spine, 16, 683–685. Waddell, G., Pilowsky, I., & Bond, M. R. (1989). Clinical assessment and interpretation of abnormal illness behavior in low back pain. Pain, 39, 41–53. Waddell, G., McCulloch, J. A., Kummel, E., & Venner, R. M. (1980). Nonorganic physical signs in low back pain. Spine, 5, 117–125. Waddell, G., & Turk, D. C. (1992). Clinical assessment of low back pain. In D. C. Turk, & R. Melzack (Eds.), Handbook of pain assessment (pp. 15–36). New York: Guilford. Werneke, M. W., Harris, D. E., & Lichter, R. L. (1993). Clinical effectiveness of behavioral signs for screening chronic low-back pain patients in a work-oriented physical rehabilitation program. Spine, 18, 2412–2418.