Acomprehensive assessment of patients suffering. Internal Structure and Validity of the Multidimensional Pain Inventory, Italian Language Version

Transcription

1 PAIN MEDICINE Volume 1 Number Internal Structure and Validity of the Multidimensional Pain Inventory, Italian Language Version Renata Ferrari,* Caterina Novara, Ezio Sanavio, PhD, and Federica Zerbini *Servizio di Terapia Antalgica e Cure Palliative, A.S.L. n.6 Vicenza, Italy Department of General Psychology, University of Padova, Italy ABSTRACT Objective. The aim of the study is an investigation of the psychometric characteristics of the Italian translation of the Multidimensional Pain Inventory and a comparision with the American, German, Swedish and Dutch versions of the MPI. Method. The Italian translation of the MPI was administered together with Melzack McGill Pain Inventory, Beck Depression Inventory, Spielberger State-Trait Anxiety Inventory, and Visual Analog Scales. Confirmatory factor analyses were accomplished on the MPI scores. Furthermore, reliability, intercorrelations, and convergent validity of MPI were evaluated. Patients. Participants were 220 patients suffering from a variety of chronic pain syndromes (cephalalgia 45.8%; low-back pain 30.5%). Results. Confirmatory factor analyses suggest changes to all 3 sections of the MPI-IV. Factor structure, after having excluded several items sorted according to the 3 sections of the questionnaire, is basically the same as in other versions of the MPI. Internal consistency analyses yielded acceptable reliability (Cronbach coefficients) for 11 out of 13 scales. Conclusions. After making appropriate changes in all 3 sections of the inventory, the MPI is substantially suitable for use in cross-cultural and international research. Key Words. Chronic Pain; Multidimensional Pain Inventory; Confirmatory Factor Analysis; Reliability Reprint requests to: Caterina Novara, Dipartimento di Psicologia Generale, Università degli Studi di Padova, Via Venezia, 8, Padova, Italy. Fax: ; cnovara@ psico.unipd.it Introduction Acomprehensive assessment of patients suffering from chronic pain should include physical, psychosocial and behavioral information. Integration of this kind of information is clinically useful in making decisions regarding therapy, especially to identify a pain management program and assess how the results of pain treatments may change in time [1,2] The (West Haven-Yale) Multidimensional Pain Inventory (MPI) [3] is theoretically related to cognitive behavioral conditions and takes into account a number of individual pain perception aspects, including the patient s perception of the reactions of others towards one s own pain symptoms, the perception of behavioral limitations, and the impact caused by pain on the patient s lifestyles. In particular, the 61-item MPI (which includes 9 experimental, unscaled items) consists of 3 sections, each comprising a certain number of scales. Section 1 consists of 28 items, distributed across 5 scales, namely: (1) reports of pain severity (PS scale); (2) perception of the extent to which pain interferes with one s own lifestyle (IN scale); (3) perceived life control (LC scale); (4) affective distress (AD scale); and (5) appraisal of the amount of support from significant others (SU scale). Section 2 asks patients to assess the behavioral responses that significant others reveal in reaction to their own indications of pain (i.e. pain behaviors). This section comprises 14 items, classified according to 3 scale levels, namely: (6) punishing responses (PR scale); (7) solicitous responses (SR scale); and (8) distracting responses (DR scale). Section 3 asks for an assessment of the frequency with which patients perform various everyday tasks, and comprises 19 Blackwell Science, Inc /00/$15.00/

2 124 Ferrari et al. items classified according to a 4-scale level, namely: (9) household chores (HC scale); (10) outdoor work (OW scale); (11) activity away from home (AH scale); and (12) social activities (SA scale). The internal structure of the American version was originally assessed by means of a confirmatory factor analysis (Section 1) and a exploratory factor analysis (Sections 2 and 3) on a sample of 120 patients (81.5% male) suffering from chronic pain and affected by several different pain syndromes [3]. The internal validity of each of the 12 scales was satisfactory (i.e., Cronbach coefficient ranged from 0.70 to 0.90) and likewise for the 2-week testretest stability (with Pearson s r coefficient ranging from 0.62 to 0.91). Since its first presentation, MPI has undergone some minor changes. Through the years, a number of items have been added or removed for specific requirements related to the various studies, especially with reference to Section 3. In particular, due to the scarce number of items in Section 3 (i.e., 19 items, distributed according to 4 scale levels) and their similarity to each other, it was proposed to consider all these items grouped together as an expression of a wider scale (the thirteenth) and to call it General Activity (GA). The MPI has been translated into German [4], Dutch [5] and Swedish [6]. A confirmatory factor analysis on the German, Swedish and Dutch versions has confirmed correspondence with the structure of the American version in the first 2 sections of the questionnaire, although slight differences have been found in the composition of the scales. In the third section, in both the German and the Swedish versions researchers found that a 3-factor solution (i.e., household chores, activity away from home, and a combined social and recreational activity factor) provided a better description of their data. The aim of this research has been to assess the properties of the Italian version of the Multidimensional Pain Inventory (MPI-IV). The study involves the following operations: (a) to assess the internal structure of the questionnaire and compare it to the original structure devised by Kerns et al. [3], in order to examine cultural diversities; (b) to examine the reliability of the scales and their mutual correlation; (c) to assess the concurrent validity of several scales by simultaneous administration of other standard tests. Method Participants Participants are 220 clinical patients (75.1% females) attending the Pain Therapy Service at Vicenza Hospital for the first time. The majority of the examined subjects had achieved a 13-year school education level (37.6%), followed by those who had finished a 5-year school education (33.5%) and those who had graduated from university (5.9%). All subjects were native Italians. Seventy percent were married or living with a partner. Most of the participants had a regular job (52.1%). The main criteria for inclusion of patients in this study were: a) suffering from pain for more than 6 months; b) the absence of oncologic pain; c) an age ranging between 18 and 65; and d) the absence of psychotic disorders. The average age of the patients was years (SD 12.33). The patients suffered from a variety of pain syndromes, including cephalalgia (45.8%), lowback pain (30.5%) and other pathological conditions such as myofascial pain, atypical facial algia, fibromyalgic syndrome, and osteoarthrosis (23.7%). The average duration of pain symptoms was 8.95 years (SD 9.88; ranging from 6 months to 31 years) Procedure In the original version of the MPI [3], patients were asked to give replies by rating their responses according to a 7-point numerical scale. The instructions given to the subject varied for each of the 3 sections of the questionnaire. The MPI was translated into 2 independent versions by 2 Italian psychologists, who then met to agree on a final common version. This newly translated questionnaire was then translated back into English by an American translator for comparison proposes. The original version and the translated version of the MPI were found to be basically the same. Only Item 16, in Section 3: Go to the park or to the beach in the original version was changed into Go to the park or go for a walk downtown to adapt it to the environment in which the research was carried out. Each patient had to undergo medical examination by a pain medicine specialist, as well as a semistructured interview by a psychologist. Assessment parameters included frequency and duration of pain symptoms referring to the past month and the total duration of painful symptoms expressed in terms of hours per month. Each patient was informed about the aim of this research. Volunteering patients who had signed an agreement giving their informed consent received the questionnaires to be administered with the assistance of a psychologist. The participants then had to fill in the following questionnaires: McGill Melzack Pain Questionnaire () [7,8]. State Trait Anxiety Inventory (STAI X1 and STAI X2) [9], in its Italian version [10], has an in-

3 Validity of the Multidimensional Pain Inventory, Italian 125 ternal consistency coefficient ranging between 0.91 and 0.92, and a test-retest stability of 0.68 after a time lapse of 7 days in its state version; whereas it has an internal consistency of 0.90 and a test-retest stability of 0.78 after 7 and also after 30 days in its trait version. Beck Depression Inventory (BDI) [11]; Italian translation by Ranchetti [12]. Visual Analog Scales (VAS) [13]. These include four scales, each consisting of a 10-cm line segment with the phrases no pain and strongest possible pain indicated at the 2 opposite ends of the line segment. Subjects were asked to rate the following: (a) the strongest degree of pain felt during the past month prior to assessment (VAS max), (b) the lightest degree of pain felt during the past month prior to assessment (VAS min), (c) the habitual degree of pain felt during the past month prior to assessment (VAS hab), and (d) the present degree of pain felt at the moment of assessment (VAS cur). Results Factor Structure In order to assess the factor structure of the MPI- IV, a confirmatory factor analysis was performed by using the LISREL-8 program [14]. Each one of the 3 sections was analyzed separately. Factors (latent variables) were defined on the basis of their corresponding items (observed variables), whereas the correlation coefficients between various factors, factor loading on the related latent variable, and the residual variances of the items were set as free parameters. As far as stand-alone indices were concerned (which determine how well the structural equation model fits to the data), the chi-square index, the goodness-of-fit index (GFI) [15], and the root mean square error of approximation index (RMSEA) [16] were used. In terms of incremental (relative fit) indices, this study involved the use of the comparative fit index (CFI) [17] and the nonnormed fit index (NNFI) [18]. In the event of an unsatisfactory fit with the model, the following parameters were examined: modification indices for factor loadings, standard errors, standard residuals, the statistical significance of each parameter and square multiple correlation [14]. With reference to Section 1, the index of fit for the first solution was fairly moderate. The value of chi-square was with 340 degrees of freedom (P 0.0). The GFI was 0.82, RMSEA was 0.06, CFI was.88 and NNFI was It was therefore concluded that the hypothesized model was incapable of providing an adequate description of the correlation in the data. The high standardized residuals achieved suggested the opportunity to eliminate item number 1 on the PS scale: Rate the level of your pain at the present moment and item number 10 on the IN scale How much has your pain changed your ability to take part in recreational and other social activities? However, this new model still did not prove satisfactory: chi-squared was , with 242 degrees of freedom (P 0.0). The GFI was 0.86, RMSEA was 0.08, CFI was 0.91 and NNFI was In this case, the largest variations of lambda-x were determined by item number 22 of the LC scale How much control do you feel that you have over your pain? A new LISREL analysis was performed after having excluded item 22. The model improved as a result of the decrease of the chi-square to , with 179 degrees of freedom (P 0.0). The GFI was 0.89, RMSEA was 0.05, CFI was 0.95 and NNFI was The 21 items that were retained and their corresponding factor loadings are shown in Table 1. In Section 2 of the MPI-IV, the first solution achieved with the LISREL analysis yielded a chisquare of , with 74 degrees of freedom (P 0.0). The GFI was 0.92, RMSEA was 0.06, CFI was 0.92 and NNFI was Although the model appeared satisfactory, it was decided that it would be appropriate to remove item number 1 Ignores me and item number 7 Gets frustrated with me from the PR scale due to their low factor loading value, and to remove item 2 Asks me what he/she can do to help and item number 14 Switches on the television to distract me from my pain due to the high standardized residuals. Another LISREL analysis was run after having omitted these 4 items. Under these conditions, chi-square was 41.43, with 32 degrees of freedom (P 0,12). The GFI was 0.96, RMSEA was 0.035, CFI was 0.99 and NNFI was This was considered an acceptable fit. The 10 retained items and their corresponding factor loadings are listed in Table 2. The first solution achieved with the LISREL analysis with reference to the third section of the MPI-IV yelded a chi-square of , with 146 degrees of freedom (P 0.0). The GFI was 0.82, RMSEA was 0.11, CFI was 0.72 and NNFI was It was therefore decided to remove item number 10 Work on the car and item 14 Wash the car from the OW scale due to its extremely high modification index values, whereas item number 18 Work on a needed household repair was removed from the same scale because of its low factor loading. Item number 12 Visiting relatives was removed from the SA scale due to its high standard-

4 126 Ferrari et al. Table 1 The 21 retained items and their corresponding factor loading Scales and Items Factor Loading Pain severity (8) On the average, how severe has your pain been during the last week?.62 (16) How much suffering do you experience because of your pain?.70 Interference (2) In general, how does your pain interfere with your day-to-day activities?.75 (3) Since the time your pain began, how much has your pain changed your ability to work?.79 (4) How much has your pain changed the amount of satisfaction or enjoyment you get from taking part in social and recreational activities?.64 (11) How much do you limit your activities in order to keep your pain from getting worse?.64 (12) How much has your pain changed the amount of satisfaction or enjoyment you get from family-related activities?.68 (18) How much has your pain changed your relationship with our spouse, family, or significant other?.47 (19) How much has your pain changed the amount of satisfaction or enjoyment you get from your work?.57 (23) How much has your pain changed your ability to do household chores?.73 (25) How much has your pain interfered with your ability to plan activities?.73 (27) How much has your pain changed or interfered in your friendships with people other than your family?.55 Life control (14) During the past week how much control do you feel that you have had over your life?.49 (21) During the past week how much do you feel that you have been able to deal with your problems?.71 (24) During the past week how successful were you in coping with stressful situations in your life?.65 Affective distress (6) Rate your overall mood during the past week..77 (26) During the past week how irritable have you been?.66 (28) During the past week how tense or anxious have you been?.83 Support (5) How supportive or helpful is your spouse (significant other) to you in relation to your pain?.76 (13) How worried is your spouse (significant other) about you because of your pain?.89 (20) How attentive is your spouse (significant other) to you because of your pain?.93 ized residuals. A new LISREL analysis was run after having omitted these 4 items. This analysis brought about a general improvement, generating a chi-square of , with 84 degrees of freedom (P 0.0). The GFI was 0.94, RMSEA was 0.03, CFI was 0.97 and NNFI was This was considered an acceptable fit. An additional LISREL analysis was performed in order to test the 3-factor model with the items of the Social Activity (SA) and Activity Away From Home (AH) scales merged into a single factor called Leisure Activity (LA). This produced a chi-square of , with 87 degrees of freedom (P 0.0). The GFI was 0.93, RMSEA was 0.03, CFI was 0.97 and NNFI was Although this operation appears to have produced nothing more than a repetition of the 4-factor solution, the 3-factor solution appears to be more parsimonious. Table 3 shows a list of the 15 items and their corresponding factor loadings, while the values in parentheses represent the factor loadings for the 3-factor model. Reliability Table 4 presents a list of the average scores, standard deviations and internal consistency (Cronbach ) for the 13 MPI scales. For 11 out of the total 13 scales, the relative internal consistency remains within the range of values from 0.68 to However, the remaining 2 scales (i.e., AD and SA) provide lower values, namely, 0.49 and 0.50, respectively. The Leisure Activity scale generated by the merging of items from the AH and the SA scales yielded an internal consistency of Intercorrelations Pearson s r correlations were calculated to assess the correlation among MPI-IV scales (Table 5). The correlation values between PS-IN (r 0.63, P 0.01), LC-AD (r 0.38, P 0.01), SR-SU (r 0.60, P 0.01), PR-DR (r 0.79, P 0.01), AH-SR Table 2 The 10 retained items and their corresponding factor loading Scales and Items Factor Loading Punishing responses (4) Gets irritated with me 1.14* (10) Gets angry with me.70 Solicitous responses (5) Takes over my jobs or duties.66 (8) Tries to get me to rest.78 (11) Gets me pain medication.68 (13) Gets me something to eat or drink.73 Distracting responses (3) Reads to me.48 (6) Talks to me about something else to take my mind off the pain.68 (9) Tries to involve me in some activity.73 (12) Encourages me to work on a hobby.88 *The correlation between item number 4 and item number 10 is fairly high (0.71). This may justify a factor loading 1. The theta-delta value is not significant (1.46).

5 Validity of the Multidimensional Pain Inventory, Italian 127 Table 3 The 15 retained items and their corresponding factor loading Scales and Items (r 0.64, P 0.01) and AH-SA (r 0.64, P 0.01), are fairly high, but the alpha coefficients are high enough to suggest that each of these scales possesses a discriminating capacity [19]. In general, the correspondence between the Italian MPI and the American, German, Swedish and Dutch versions is confirmed. These 4 versions of MPI have shown a relatively high inter-scale correlation on 3 pairs of scales (PS-IN, LC-AD, SR-SU). Convergent validity In order to examine the agreement between the MPI-IV and other validated questionnaires, the correlation coefficient was calculated between the 13 scales plus the one (LA) added in the MPI-IV and the parameters VAS max, VAS min, VAS hab, and VAS cur,, STAI X1 and STAI X2, and BDI. Table 6 lists the correlation values that have emerged. Discussion Factor Loading Household chores (1) Wash dishes.73 (.73) (5) Go grocery shopping.72 (.72) (9) Help with the house cleaning.79 (.79) (13) Prepare meal.85 (.85) (17) Do the laundry.87 (.87) Outdoor work (2) Mow the lawn.46 (.45) (6) Work in the garden 1.25* (1.28) Activity away from home (3) Go out to eat.60 (.60) (7) Go to a movie.47 (.47) (11) Take a ride in a car or bus.59 (.60) (15) Take a trip.74 (.75) Social activity (4) Play cards or other games.28 (.31) (8) Visit friends.62 (.65) (16) Go to a park or go for a walk downtown.51 (.52) (19) Engage in sexual activities.36 (.37) * The correlation between item number 2 and item 6 is fairly high (0.77). This may justify a factor loading 1. The theta-delta value is not significant (1.97). The results of this research establish that the psychometric properties of the MPI-IV are satisfactory with reference to both factor structure and reliability. Current results are similar to those achieved by researchers using the other versions of the MPI questionnaire, namely the MPI-DLV, MPI-S and the MPI-D: high intercorrelations were found between the same scale (PS-IN and LC-AD). The issues that previous studies had brought out in favor of a 5-factor solution of MPI Section 1 may also apply to the present study. Table 4 Means, standard deviations and Cronbach coefficients for MPI-IV scales Scale Number of Questions Mean SD Cronbach Pain severity Interference Life control Affective distress Support Punishing responses Solicitous responses Distracting responses Household chores Outdoor work Activities away from home Social activities Leisure activity General activity From a clinical diagnostic point of view, it appears that it is important to distinguish pain severity from pain interference. Although it is fairly commonly found that high pain severity is associated with the feeling of interference in one s daily lifestyle, this is not necessarily always true. If the 2 scales were merged together into a single Pain-Interference scale, the 2 types of information would be lost altogether. The same could be said about the Affective Distress and Life Control scales. It seems important to be able to distinguish between affective information on one hand and cognitive information on the other. Another reason for maintaining the original factor structure is in order to compare the clusters analysis as described in a series of articles by Turk et al. [20 22]. To compare the different characteristics of the patients taken into consideration for the various different studies, it would be appropriate to maintain the same number of factors. The confirmatory structure analysis and the reliability analysis of Section 1 would suggest the removal of 3 items. The different characteristics of a group of subjects, including differing semantic and cultural factors or an unclear translation, may produce an unsatisfactory fit, as proposed by the LISREL analysis. For instance, although item number 1 ( Rate the level of your pain at the present moment ) of the PS scale had a factor loading value of 0.65, its omission produced an increase of the alpha coefficient from 0.21 to The 2 separate time intervals to which reference is made (i.e., in this moment in item number 1 and during the last week in item 8) may have caused a decrease in internal validity. Due to the central importance of this issue, it is recommended that Italian users use this item as a qualitative index in the case that no

6 128 Ferrari et al. Table 5 Intercorrelations between scales: PS Pain Severity; IN Interference; LC Life Control; AD Affective Distress; SU Support; PR Punishing Responses; SR Solicitous Responses; DR Distracting Responses; HC Household Chores; OW Outdoor Work; AH Activity Away From Home; SA Social Activity; LA Leisure Activity; GA General Activity PS IN LC AD SU PR SR DR HC OW AH SA LA IN.63* LC.10.20* AD.31**.27**.38** SU.24**.27** PR.20*.23**.12.19*.08 SR.15*.16* **.05 DR.19*.24** **.09 HC * *.04 OW AH.16*.16*.15* **.29**.20**.01 SA.29** *.06.26**.32**.20**.03.64** LA * **.34**.23**.01.91**.89** GA ** **.15.79**.48**.61**.67**.67** * p ** p other individual pain assessment instruments are used. Item 10 of the IN scale, How much has your pain changed your ability to take part in recreational and other social activities? had a factor loading value of 0.31; and item 22 of the LC scale, How much control do you feel that you have over your pain? had a factor loading value of Their elimination brought about an increase in the alpha coefficient from 0.75 to 0.88 and from 0.60 to 0.68 respectively. With reference to Section 2 of the MPI-IV, the removal of item 1 Ignores me is in agreement with the idea of Flor et al. [4], as well as with the Dutch and Swedish versions, which have found that this item does not fit into the PR scale. With regard to both item 7 and item 2 alike, it is believed that there may have been some difficulty of interpretation. Item 7 Gets frustrated with me appears to be interpreted by patients in a sense that is not so clearly negative as for the other items of this scale (e.g. Ignores me or Gets angry with me ). Item 2 of the SR scale Asks me what he/she can do to help differs from other items of the scale as it implies a response involving an offer of help of a general, nonconcrete nature, and for this reason probably is subject to various different interpretations. Table 6 Correlation between MPI-IV scales and Visual Analog scale for maximum intensity of pain (VAS max), Visual Analog scale for minimum intensity of pain (VAS min), Visual Analog scale for habitual intensity of pain (VAS hab), Visual Analog scale for current intensity of pain (VAS cur), sensorial (-S), affective (-A), Evaluative (-E), mixed sensorial (-MS), affective evaluative (-AE), mixed (-M), Pain Rating Index (-TOT), State anxiety Inventory (STAI X1), Trait anxiety inventory (STAI X2), Beck depression inventory (BDI) VAS max VAS min VAS hab VAS cur -S -A -E PS.31**.36**.35**.37**.21**.30**.30**.08.45**.33**.18**.24**.20**.27** IN.28**.34**.17*.31**.21**.30**.28**.05.29**.31**.10.19*.26**.44** LC.06.16* * **.35**.36** AD.22**.25**.19*.14* **.42**.44** SU *.19**.15*.03.17*.18**.04.16* PR.15*.14* **.23**.17* SR.25** * *.05.16* DR.19** * HC OW ** AH.04.14* **.15*.17* SA **.20**.28** LA.09.15* **.18**.25** GA.00.20** * * p ** p MS -AE -M TOT STAI X1 STAI X2 BDI

7 Validity of the Multidimensional Pain Inventory, Italian 129 Item 14 Turns on the TV to take my mind off my pain does not entirely appear to fit in this domain. Rather than an offer for help, it appears to be a response of distraction, like item 6 Talks to me about something else to take my mind off my pain in the DR scale. Section 3 of the MPI-IV has been changed, particularly with reference to the OW scale, removing item number 10 Work on the car, item number 14 Wash the car and item 18 Work on a needed household repair, which did not load on into the scale. In actual fact, these appeared to be jobs typically performed by males only. As our sample population involved mostly women (75.1%), these items may have been somewhat too specific for them. Furthermore, item 12 Visit relatives may be interpreted by Italians not so much as a leisure activity like the other items in this scale, but rather as a social responsibility. The LISREL confirmatory analysis did not bring about any improvement in the third part of a three-factor solution. On the basis of economical criteria, it is suggested that Italian users should consider either the 3-factor solution or Section 3 as a single factor belonging to the General Activity scale. With regard to the correlation values between the various MPI scales, the results obtained in previous studies are confirmed [3 6]. In agreement with the expected outcome, a positive correlation was found between the Pain Severity and Interference scales, between the Support and Solicitous Response scales, between Activity Away From Home and Solicitous Response scales, and Activity Away From Home and Social Activity, whereas a negative correlation was found between the Life Control and Affective Distress scales. However, the high correlation value found between the Punishing Responses and Distracting Responses scales is a rather unexpected result. This suggests that the subjects under examination appear to interpret punishing responses in a more positive manner. Correlation factors with external measurements indicate the presence of significant positive correlation values between the PS and Visual Analogscales. Results also highlight the positive correlation between the PS scale and the sub-scale of the, and with STAI and BDI. The agreeing positive correlation between the measurements of anxiety and depression with the Affective Distress scale and the negative correlation with the scales of Life Control, Activity Away From Home and Social Activity confirms the capability of these scales to capture the extent of emotional uneasiness. This paper supports the adoption of the MPI-IV, both as a research instrument and as a clinical assessment device. Changes have been made to all 3 sections of the questionnaire in order to achieve an acceptable fit for the hypothesized factor structure. After having made these changes, we discovered that the MPI-IV had the same factor structure as that of the American, German, Swedish and Dutch versions of the MPI. The psychometric properties are basically similar, the scales contain more or less the same items and the correlation values between the various scales are fairly much the same. The MPI-IV also features a satisfactory reliability, and for this reason it may well be suited for an Italian population suffering from chronic pain. However, based on the small sample size and the prevalence of females (75.1%) in our sample, results of this research must be taken with caution. Validation of the MPI-IV by means of the confirmatory factor analyses to distinguish between the 2 sexes, and consideration of the validity and sensitivity to change of the scales may be a prospective area for future research. Acknowledgments This study was conducted while the second author was supported by Fondazione S. Maugeri, Clinica del Lavoro e della Riabilitazione, IRCCS. We would like to express our gratitude to Alberto Voci for his statistical assistance and to Manuela Partinico for data-management. Also our gratitude to the staff of Servizio di Terapia Antalgica e Cure Palliative dell Ospedale S. Bortolo di Vicenza for their help with the data collection. References 1 Turk DC, Rudy TE. Towards a comprehensive assessment of chronic pain patients. Behav Res Ther 1987;25: Wall PD, Melzack R. Textbook of Pain. Edinburgh: Churchill Livingstone; Kerns RD, Turk DC, Rudy TE. The West Haven- Yale Multidimensional Pain Inventory (WHYMPI). Pain 1985;23: Flor H, Rudy TE, Birbaumer N, Streit B, Schugens MM. Zur Anwendbarkeit des West-Haven-Yale Multidimensional Pain Inventory im deutschen Sprachraum. Der Schmerz 1990;4: Lousberg R, Van Breukelen GJ, Groenman NH, Schmidt AJ, Arntz A, Winter FA. Psychometric properties of the Multidimensional Pain Inventory, Dutch language version (MPI-DLV). Behav Res Ther 1999;37: Bergström G, Jensen IB, Bodin L, Linton SJ, Nygren AL, Carlsson SG. Reliability and factor structure of Multidimensional Pain Inventory-Swedish Language Version (MPI-S). Pain 1998;75:

8 130 Ferrari et al. 7 Melzack R. The McGill Pain Questionnaire: major properties and scoring methods. Pain 1975;1: Majani G, Sanavio E. Semantics of Pain in Italy. The Italian version of the McGill Pain Questionnaire. Pain 1985;22: Spielberger CD, Gorsuch RL, Lushene RE. STAI Manual for State-Trait Anxiety Inventory. Palo Alto, Calif: Consulting Psychologists Press; Lazzari R, Pancheri P. STAI Questionario di autovalutazione dell ansia di stato e di tratto. Firenze: Organizzazioni Speciali; Beck AT, Rush AJ, Shaw BF, Emery G. Cognitive therapy of depression. New York: Guilford Press; Ranchetti C. Terapia cognitiva della depressione. Boringhieri; Scott PJ, Huskisson EC. Graphic representation of pain. Pain 1976;2: Jöreskog KG, Sörbom D. LISREL 8: Structural equation modeling with the SIMPLIS Command Language. Hillsdale, NJ: Lawrence Erlbaum; Jöreskog KG, Sörbom D. LISREL 7: A guide to the program and applications. Chicago: SPSS; Steiger J. Structure model evaluation and modification: An interval estimation approach. Multivariate Behav Res 1990;25: Bentler PM. Comparative fit indexes in structural models. Psychol Bull 1990;107: Bentler PM, Bonnet DG. Significance tests and goodness of fit in the analysis of covariance structures. Psychol Bull 1980;88: Campbell DT, Fiske DW. Convergent and discriminant validation by the multitrait-multimethod matrix. Psychol Bull 1959;56: Turk DC, Rudy TE. Towards an empirically derived taxonomy of chronic pain patients: integration of psychological assessment data. J Consult Clin Psychol 1988;56: Rudy TE, Turk DC, Zaki HS, Curtin HD. An empirical taxometric alternative to traditional classification of temporomandibular disorder. Pain 1989;36: Jamison RN, Rudy TE, Penzien DB, Mosley TH. Cognitive-behavioral classifications of chronic pain: replication and extension of empirically derived patient profiles. Pain 1994;57: