Pessimistic back beliefs and lack of exercise: a longitudinal risk study in relation to shoulder, neck, and back pain

(1)

Article

Reference

Pessimistic back beliefs and lack of exercise: a longitudinal risk study in relation to shoulder, neck, and back pain

ELFERING, Achim, et al.

Abstract

This cross-lagged-panel study tested the validity of the German version of the back beliefs questionnaire (BBQ) in predicting pain in the shoulders, neck, and back. A random sample of 2860 individuals participated at baseline, and 73% responded at one-year follow-up.

Structural equation modeling was used to carry out a model comparison to evaluate whether paths differed between individuals with and without initial back pain and between those who exercised at baseline and those who did not. Factor analysis showed eight of the nine original items loaded on the expected common factor. High BBQ scores at baseline significantly predicted an increase in shoulder, neck, and back pain in individuals with current back pain (β

= .11, p < .05), but not in other respondents (β = .02, p = .259). Similarly, baseline BBQ scores predicted the increase in shoulder, neck, and back pain among those who did not exercise (β = .15, p < .05), but not in those who did (β = .04, ns). The risk of negative back beliefs preceding an increase in shoulder, neck, and back pain was greatest for those with current back pain who did not exercise (β = [...]

ELFERING, Achim, et al . Pessimistic back beliefs and lack of exercise: a longitudinal risk study in relation to shoulder, neck, and back pain. Psychology, Health & Medicine , 2015, vol. 20, no. 7, p. 767-780

DOI : 10.1080/13548506.2015.1017824

Available at:

http://archive-ouverte.unige.ch/unige:97987

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=cphm20

Download by: [Université de Genève] Date: 10 October 2017, At: 05:01

Psychology, Health & Medicine

ISSN: 1354-8506 (Print) 1465-3966 (Online) Journal homepage: http://www.tandfonline.com/loi/cphm20

Pessimistic back beliefs and lack of exercise: a longitudinal risk study in relation to shoulder, neck, and back pain

Achim Elfering, Urs Müller, Cornelia Rolli Salathé, Özgür Tamcan & Anne F.

Mannion

To cite this article: Achim Elfering, Urs Müller, Cornelia Rolli Salathé, Özgür Tamcan & Anne F. Mannion (2015) Pessimistic back beliefs and lack of exercise: a longitudinal risk study in relation to shoulder, neck, and back pain, Psychology, Health & Medicine, 20:7, 767-780, DOI:

10.1080/13548506.2015.1017824

To link to this article: http://dx.doi.org/10.1080/13548506.2015.1017824

Published online: 02 Mar 2015.

Submit your article to this journal

Article views: 204

View related articles

View Crossmark data

Citing articles: 5 View citing articles

(3)

Pessimistic back beliefs and lack of exercise: a longitudinal risk study in relation to shoulder, neck, and back pain

Achim Elfering^a,b*, Urs Müller^c, Cornelia Rolli Salathé^a, Özgür Tamcan^cand Anne F. Mannion^d

aDepartment of Psychology, Institute for Psychology, University of Bern, Fabrikstrasse 8, 3012 Bern, Switzerland;^bNational Centre of Competence in Research, Affective Sciences, University of Geneva, CISA, Geneva, Switzerland;^cInstitute for Evaluative Research in Orthopaedic Surgery, University of Bern, Bern, Switzerland;^dSpine Center, Schulthess Klinik, Zurich, Switzerland

(Received 11 April 2014; accepted 21 January 2015)

This cross-lagged-panel study tested the validity of the German version of the back beliefs questionnaire (BBQ) in predicting pain in the shoulders, neck, and back. A random sample of 2860 individuals participated at baseline, and 73% responded at one-year follow-up. Structural equation modeling was used to carry out a model comparison to evaluate whether paths differed between individuals with and without initial back pain and between those who exercised at baseline and those who did not.

Factor analysis showed eight of the nine original items loaded on the expected common factor. High BBQ scores at baseline significantly predicted an increase in shoulder, neck, and back pain in individuals with current back pain (β= .11,p< .05), but not in other respondents (β= .02,p= .259). Similarly, baseline BBQ scores predicted the increase in shoulder, neck, and back pain among those who did not exercise (β= .15,p< .05), but not in those who did (β= .04, ns). The risk of negative back beliefs preceding an increase in shoulder, neck, and back pain was greatest for those with current back pain who did not exercise (β= .29, p< .05). The findings confirmed the validity of the German BBQ. Cognitive behavioral interventions should address pessimistic back beliefs in high-risk groups.

Keywords:pessimistic back beliefs; German version BBQ; cross-lagged-panel

Introduction

The biopsychosocial model of pain postulates that psychological factors are important with respect to the susceptibility, severity, and course of illness (Borrell-Carrió, Suchman, & Epstein, 2004; Engel, 1977; Rolli Salathé et al., 2013; Rolli Salathé &

Elfering, 2013). Psychological risk factors for the development of chronic back pain include individual beliefs about back problems and its management (Dionne et al., 2005; Elfering & Mannion, 2008; Kendall, Linton, & Main, 2004). Beliefs regarding the inevitability of the future as a consequence of having back trouble are associated with pain maintenance (Elfering, Mannion, Jacobshagen, Tamcan, & Müller, 2009), back pain-related absenteeism and presenteeism (Mannion et al., 2009), and back pain- related use of the healthcare system (Mannion, Wieser, & Elfering,2013). Hence, modi- fying individuals’ beliefs about back problems has become an important goal of disabil-

*Corresponding author. Email:[email protected]

Vol. 20, No. 7, 767–780, http://dx.doi.org/10.1080/13548506.2015.1017824

Downloaded by [Université de Genève] at 05:01 10 October 2017

(4)

ity prevention strategies (Buchbinder & Jolley, 2005; Woby, Watson, Roach, &

Urmston,2004).

One questionnaire that has been used to evaluate the success of media campaigns is the Back Beliefs Questionnaire (BBQ, Symonds, Burton, Tillotson, & Main, 1996). A recent large, cross-sectional study provided strong evidence to support the present structure, scoring, and construct validity of the BBQ (Bostick, Schopﬂocher, & Gross, 2013). However, although frequently used (Gross et al., 2012), no longitudinal evaluations of its construct validity have been carried out.

The aim of this study was therefore to validate the newly created German version of the BBQ in a prospective population study. Since pessimistic back beliefs impede recovery from back pain (Symonds et al., 1996), we hypothesized that the prognostic value of BBQ scores in predicting future pain would be greatest in those with back pain at baseline. Pessimistic back beliefs are also associated with (ineffective) passive pain management, especially when the individual’s habitual level of physical activity is already low (Larsson & Nordholm, 2008). Thus, we hypothesized that the association between pessimistic back beliefs and future pain would be the greatest in those who did not exercise at baseline. Finally, we anticipated that, in those reporting back pain at baseline, exercise would buffer the association between pessimistic beliefs and future pain, compared with those who reported back pain but were inactive.

Methods Participants

In 2003, a population-based survey of health and prevalence of back pain was carried out in Switzerland (N= 16,674, Figure 1). In 2005, individuals from this cohort were selected for participation in this study, after their pre-stratiﬁcation for the presence/

absence of low back pain ‘in the last month’ indicated both in the original survey in 2003 and in a short telephone interview in 2005 just prior to this study. In many individuals, low back pain episodes are either recurrent or persistent (Melloh et al., 2014), and so the presence vs. absence of low back pain was deﬁned in relation to a consistent back status (pain/no pain) in both the four-week periods preceding the two surveys.

From those two groups (pain/no pain), 2860 individuals were randomly selected. Two hundred of these were invited to complete the BBQ twice within a period of approxi- mately 2 weeks, to assess the reproducibility of BBQ scores. Participant characteristics at baseline (in 2005) and at one-year follow-up (in 2006) are shown in Table 1. The research was approved by the local Research Ethics Committee.

Self-report measures

The German version of the BBQ was used (Elfering et al., 2009). The BBQ contains nine single-construct content items and ﬁve distractor items with a 7-point scale (completely disagree = 1, completely agree = 7). The items address the inevitability of the future as a consequence of having back pain. In this study, a higher score indicated more negative beliefs about back pain. Pain was assessed using the Pain Standard Eval- uation Questionnaire (SEQ Pain; Müller et al.,2008). Four questions enquired about the intensity of pain in various regions of the back:‘In the lastfour weeks, how much pain did you experience in the following body regions?’The regions included: (1) left shoulder; (2) right shoulder; (3) neck and upper back; and (4) lower back. Response options ranged from 0 (no pain) to 6 (unbearable pain). Present back pain at baseline was

(5)

assessed by asking, ‘Do you have back pain now?’ with a binary response format (0 = no, 1 = yes). A second question enquired,‘If yes, how long has the present episode lasted?’ with the response format being the number of weeks. A single item assessed exercise habits: ‘Do you participate in sport or carry out any sporting activities/exercise (e.g. cycling to work, hiking, yoga, etc.)?’ The response format was binary (0 = no, 1 = yes).

.

SET MXLOOP=nnn.

2003: Population based cross-sectional baseline:

n = 16,634, 4046 report Low Back Pain

2700 (94%) consented to participate :

No questionnaire returned:

n = 193 (6.7%)

Baseline 2005: 2507 (88%) returned questionnaire, 1071 reported current low

back pain (43%) Sampling for present study on random sample of n = 2860 from prestratified

groups with and without back pain

Follow-up 2006:

1833 (73%) returned questionnaire, 733 reported current low back pain (40%)

No questionnaire returned:

n = 674 (27%)

Refused: n = 160 (5.6%)

Figure 1. Flow of subjects through the study.

(6)

Table 1. Participant characteristics.

Baseline frequency or mean

(%) (SD)

Follow-up frequency or mean

(%) (SD)

Sample size 2507 1833 (73.1)

Women 1142 (45.6) 840 (45.8)

Age (mean [SD]) (years) 53.0 (16.4) 53.9 (15.8)

Exercise (0 = no, 1 = yes) 2076 (83.3) 1518 (84.4)

Current low back pain (0 = no, 1 = yes) 1071 (42.7) 733 (40.0)

Education

Compulsory education 262 (10.4) 155 (8.4)

Apprenticeship 1224 (49.0) 901 (49.2)

Apprenticeship and further education 264 (10.6) 191 (10.4)

Advanced technical education 485 (19.4) 365 (19.9)

University 265 (10.6) 219 (12.0)

Current pain intensity (Mean [SD]) (0 = no pain–6 = unbearable pain)

Left shoulder .5 (1.1) .5 (1.0)

Right shoulder .6 (1.3) .6 (1.2)

Neck and upper back 1 (1.4) .9 (1.3)

Lower back 1.3 (1.6) 1.1 (1.5)

BBQ score (Mean [SD]) (9 items, range 9–63)

30.4 (10.1) 29.4 (10.1)

Q1. There is no real treatment for back trouble^a

2.3 (1.6) 2.3 (1.6)

Q2. Back trouble will eventually stop you from working^a

2.7 (1.8) 2.6 (1.8)

Q3. Back trouble means periods of pain for the rest of one’s life^a

3.7 (2.0) 3.5 (1.9)

Q4. Doctors cannot do anything for back trouble

2.6 (1.7) 2.5 (1.6)

Q5. A bad back should be exercised 2.7 (1.8) 2.7 (1.8)

Q6. Back trouble makes everything in life worse^a

4.4 (1.9) 4.3 (1.9)

Q7. Surgery is the most effective way to treat back trouble

2.6 (1.6) 2.6 (1.6)

Q8. Back trouble may mean you will end up in a wheelchair^a

3.4 2.0 3.4 1.9

Q9. Alternative treatments are the answer to back trouble

3.6 1.6 3.7 1.5

Q10. Back trouble means long periods of time off work^a

3.7 1.8 3.5 1.7

Q11. Medication is the only way of relieving back trouble

2.4 1.6 2.4 1.6

Q12. Once you have had back trouble there is always a weakness^a

3.7 2.0 3.5 1.9

Q13. Back trouble must be rested^a 2.8 1.7 2.7 1.7

Q14. Later in life back trouble gets progressively worse^a

3.9 1.8 3.7 1.8

aContent items used by the developers in the scoring of the BBQ.

(7)

Statistical analyses

The reproducibility of BBQ scores was described by the intraclass correlation coefficient (ICC) (≥.60 considered substantial (Landis & Koch, 1977)). Cronbach’s alpha indicated the internal consistency of the BBQ (≥.70 considered good, Kline, 2000). The factorial structure of the BBQ was tested using an exploratory factor analysis (EFA). AMOS 18.0 software was used to run confirmatory factor analysis (CFA). Measurement equivalence, that is, whether each item loaded on the same factor with the same loading at baseline and follow-up (Sharma, Mukherjee, Kumar, & Dillon, 2005), was tested by constraining the loadings of each indicator to be equal at baseline and follow-up. No significant difference in fit (Chi squared) between free and restrained models confirmed measurement equivalence.

The hypotheses were tested using cross-lagged-panel models (Clegg, Jackson, &

Wall, 1977). Cross-lagged-panel correlation is a method for identiﬁcation of both the source and direction of a causal effect when no true experimental studies can be done.

It is better suited than is multiple regression to many questions in panel studies. The longitudinal directional path between BBQ scores at baseline predicting increased pain at follow-up was entitled the ‘prospective risk path.’A number of prospective risk path models were compared: Model 1a and model 1b included all original 9 questions of the BBQ. In model 1a (free loadings) loadings of single questions on the latent BBQ factor were allowed to differ at baseline and follow-up (e.g. question number 2 might have a loading of .58 at baseline and .60 at follow-up). In model 1b (restrained loadings) the loading of question 2 was forced to be the same at baseline and follow-up (e.g. .58).

Model 1b requested the factor structure to be the same at baseline and follow-up (measurement equivalence). In model 1c, not all original 9 questions of the BBQ were included, but only 8 questions. Question 1 was omitted in model 1c (as in Bostick et al., 2013 question 1 did not show the main loading on the common BBQ factor in EFA). The free 8-question structure (model 1c) had to be compared with the free 9-question structure (model 1a) to test whether the 8-question structure was better than the 9-question structure. Separate models evaluated whether the prospective risk path was (a) greater in those with back pain at baseline than in those with no back pain, (b) greater in those not exercising at baseline than in those who exercised, and (c) greater in those with back pain and not exercising at baseline than in those with back pain who exercised. These comparisons were made by constraining the paths to be the same for all groups. When the restrained model was significantly worse infit than the model that allowed the paths to vary across groups, the test of group-difference was considered significant. Path coefficients were considered small (.10), moderate (.30), and large (.50) in terms of effect size classification of Cohen (1977).

Results

In total, 151/200 (76%) individuals who were sent a second BBQ returned it after a median of nine days (range 4–13 days). Reliability was good (ICC .89; 95%

CI = .85–.92).

The response rate at one-year follow-up was 73% (Figure1). There were no signiﬁ- cant differences between the drop-outs and the completers for baseline BBQ, presence of back pain at baseline, or pain in shoulders, neck, or back. The BBQ showed accept- able item-total correlation coefﬁcients (>.48) at baseline and follow-up and adequate internal consistency (Cronbach alpha was .80 at baseline and .80 at follow-up). Exercise

(8)

was associated with lower BBQ scores (Table 2). Indicators of muscular pain were associated with higher BBQ scores.

The EFA yielded three factors (Table 3); these explained a greater proportion of the total variation in items (48%) than did a two-factor solution (38%). The ﬁrst factor in the three-factor solution contained the same content items used in the original scoring of the BBQ, with the exception that Q1 was not included. Two content items (Q3 and Q12) loaded on both factor 1 and factor 2 (factor 2 loadings being stronger). The distractor items loaded on the second or third factor.

CFA and measurement models

First a CFA with all 9 content items at baseline BBQ and follow-up BBQ was created (Table 4, model 1a). Next, a restrained CFA was performed that assumed the same loadings of indicators on BBQ at baseline and follow-up (model 1b). Model 1a and model 1b showed equivalent fit (p= .341); hence, model 1b was favored. Q1 was then excluded from the CFA (model 1c) and this showed a significantly betterfit than model 1a (χ²diff(30) = 132.56, p< .001). When model 1c was compared with a restrained 8-item CFA, model 1d, the two showed equivalentfit (p= .360). Thus, the more strin- gent model 1d was favored as the best model and the constraints on factor loadings were retained in measurement model 1e where pain was added. Thefit of model 1e was good (Table4).

Cross-lagged-panel models

Figure 2 shows the cross-lagged-panel model (model 2a) with good ﬁt (Table 4). The prospective risk path from BBQ at baseline to pain at the one-year follow-up was significant (β= .06, p< .05). In contrast, shoulder, neck, and back pain at baseline was not a signiﬁcant predictor of the increase in BBQ at follow-up (β=–.01, ns).

Back pain at baseline

The cross-sectional association between BBQ and pain at baseline was large in those who reported back pain at baseline (model 2b, Figure 3; β= .38, p< .001) and non- existent in those who did not (β= .01,ns). The prospective risk path was significant in those with back pain at baseline (β= .11, p< .05), but not in those with no back pain (β= .02, ns). A model that forced the cross-sectional association and longitudinal path between BBQ and pain to be the same for those with and without present back pain at baseline (model 2c) revealed a significantly worse fit than model 2b (χ²diff(2) = 37.69, p< .001, Table4).

Exercise at baseline

The cross-sectional association between BBQ and pain at baseline was moderate in those who did not exercise at baseline (model 2d, Figure 4; β= .32, p< .001) while it was small in those who exercised (β= .12,p< .01). The prospective risk path was significant only in the group that did not exercise (β= .15, p< .05). Restrained and free models in relation to exercise were significantly different (χ²diff(2) = 9.23, p< .05, Table 4). In both groups, pain in the shoulders, neck, and back at baseline was not a significant predictor of the increase in BBQ at one-year follow-up.

(9)

Table2.Pearsoncorrelationcoefﬁcientsshowingtherelationshipsbetweenstudyvariables. BaselineFollow-up Current backpainExerciseBBQLeft shoulderRight shoulderNeckand upperbackLower backBBQLeft shoulderRight shoulderNeckand upperbackLower back Baseline Currentbackpain Exercise−.03 BBQ.08−.18 Painintensity: Leftshoulder.22−.08.13 Rightshoulder.23−.08.12.43 Neckandupper back.40−.04.10.36.41 Lowerback.57−.04.10.29.27.42 Follow-up BBQ.02−.13.65.08.09.05.05 Painintensity: Leftshoulder.21−.06.15.42.32.32.24.14 Rightshoulder.20−.08.13.29.43.34.25.17.50 Neckandupper back.31−.10.12.30.34.57.34.11.44.46 Lowerback.40−.05.11.24.25.35.53.14.31.35.46 Notes:Boldvaluesshowthestabilityofthevariablesacross1year.Coefﬁcients>¦.04¦p<.05,two-tailed.

(10)

Combined risk of back pain and no exercise

Coefficients of cross-sectional (β= .48, p< .01) and longitudinal associations (β= .29, p< .05) were highest in individuals with back pain at baseline who did not exercise (Figure5), while doing exercise seemed to weaken these associations, even in individuals reporting pain at baseline (cross-sectional path, β= .08, p< .05; longitudinal path, β= .03,ns). Model 2f, in which all path coefficients were allowed to be freely estimated in the four groups of combined risk factors (in relation to back pain and exercise at baseline), was clearly a betterfit than a model that forced cross-sectional and longitudinal associations to be the same for all four groups (model 2g, χ²diff(2) = 7.72,p< .05, Table4).

Discussion

The newly translated German version of the BBQshowed good internal consistency and test-retest reliability. Its factor structure was replicated and the common variance explained (48%) was higher than that reported (28%) in a previous population study (Bostick et al., 2013). Eight of the nine items loaded most strongly on the one common factor proposed in the original BBQ (Symonds et al., 1996). As previously reported (Bostick et al., 2013) the ﬁrst item did not load on this common factor. The item loadings of the remaining eight BBQ items on their common factor were invariant across time (absence of‘beta change,’Golembiewski, Billingsley, & Yeager, 1976).

The study validates a questionnaire instrument that assesses pessimistic back pain beliefs. Pain in the shoulders, neck, and back is common and most individuals recover within weeks (Elfering et al., 2009). The study ﬁndings show that pessimistic pain beliefs increase the risk than pain persists, pain episodes frequently recur, or pain becomes chronic. Pessimistic pain beliefs correspond to passive pain management such Table 3. BBQ score factor loadings for the three-factor solution derived by principal axis extraction and varimax rotation (baseline sample,n= 2225).

Items

Factor loadings

Factor 1 Factor 2 Factor 3 h²

Q1^a .06 .78 .15 .64

Q2^a .52 .39 −.05 .43

Q3^a .38 .60 −.14 .52

Q4 −.01 .82 −.04 .68

Q5 −.06 .10 .64 .42

Q6^a .50 .16 −.40 .44

Q7 .59 −.11 .31 .46

Q8^a .67 .12 −.16 .44

Q9 −.01 −.11 .66 .49

Q10^a .76 .06 −.22 .62

Q11 .51 .23 .41 .49

Q12^a .36 .43 −.05 .31

Q13^a .55 .15 .09 .34

Q14^a .52 .41 −.03 .45

Notes: Salient loadings are generally considered to be > ¦.30¦. Rotation method = Varimax with Kaiser normalization.h²= communality coefﬁcient.

aContent items used by the developers in the scoring of the BBQ.

(11)

Table4.Structuralequationmodelsoftherelationshipbetweennegativebackbeliefsandpainintheshoulders,neck,andbackacrossone-yearfollow-up. Modelχ2 dfχ2 /dfpSRMRRMSEAAGFIAICModelcomparisonΔχ2 dfp ConﬁrmatoryFactorAnalysesincludingBBQatbaselineandfollow-up (1a)OriginalBBQ9items816.801256.53.00.07.06.92908.80 (1b)Original9items;sameitem loadingsatbaselineandfollow-up825.821336.21.00.07.06.92901.82(1a)vs.(1b)9.028.341 (1c)ExcludingQ1684.24957.20.00.06.06.92766.24(1a)vs.(1c)132.5630.000 (1d)ExcludingQ1;sameitem loadingsatbaselineandfollow-up691.941026.78.00.06.06.93759.94(1c)vs.(1d)7.707.360 BBQandpaininshoulders,neckandbackatbaselineandfollow-up (1e)Measurementmodel910.602323.92.00.06.04.941046.60freevs.restraineda 10.1910.424 (2a)Cross-lagged-panelmodel920.792423.80.00.06.04.941036.79 Riskgroupcomparisons (2b)Currentbackpainatbaseline (n=656)vs.nocurrentbackpain (n=885) 1440.874972.92.00.05.04.911652.87Comparisonofpathsb model2bwithrestrained model2c

37.692.000 (2d)Exerciseatbaseline(n=1332) vs.noexerciseatbaseline (n=206)

1200.864962.43.00.08.03.921412.86Comparisonofpathsb model2dwithrestrained model2e

9.232.010 (2f)Backpain&exerciseatbaseline (n=562),backpain&noexercise atbaseline(n=92),nobackpain& exerciseatbaseline(n=770),no backpain&noexerciseatbaseline (n=114)

1664.2210021.66.00.07.02.902060.22Comparisonofpathsb model2fwithrestrained model2g

7.722.021 Notes:ThefitbetweenempiricalcorrelationsandstructuralmodelswasassessedbytheStandardizedRootMeanSquareResidual(SRMR;0–.05indicatedgoodfit (Schermelleh-Engel,Moosbrugger,&Müller,2003))andtheRootMeanSquareErrorofApproximation(RMSEA;≤.06indicatedgoodfit(Schermelleh-Engeletal.,2003)).Two furthercommonindicatorswereassessed:AdjustedGoodness-of-fitIndex(AGFI;≥.90indicatedgoodfit)andAkaikeInformationCriterion(AIC;shouldbeaslowaspossible) (Schermelleh-Engeletal.,2003).χ2=Chi-squarevalueindicatestheminimumdiscrepancybetweenempiricalcovariancestructuresandthoseimpliedbythemodel;df=degrees offreedom;χ2/df=minimumdiscrepancydividedbyitsdegreesoffreedom,asanindicatoroffit;p=probabilityofthediscrepancytodifferfromzero(shouldbenon-significant inagoodmodel). aComparisonoffreeitemloadingvs.sameitemsloadingsatbaselineandfollow-upinmodel1e. bCross-sectionalassociationbetweenBBQandpainatbaselineandlongitudinalriskpathfrombaselineBBQtopainatfollow-uprestrainedtobesameinbothgroupsinmodels 2c,2e,and2gvs.freelyestimatedinmodels2b,2d,and2f.

(12)

as resting and the avoidance of normal activities (Kendall et al.,2004). A vicious cycle may arise with inactivity, whereby the musculature weakens, pain persists and– in turn – pessimistic beliefs about pain increase (Rolli Salathé & Elfering, 2013). Pessimistic back pain beliefs can be modiﬁed by cognitive behavioral pain therapy (Rolli Salathé, Elfering, & Melloh, 2012). In cognitive behavioral therapy, individuals agree to participate in a graded exercise program and in doing so realize that their inappropriate beliefs, Figure 2. Cross-lagged structural equation panel model of association between BBQ and pain over the one-year study period (model 2a).

*p< .05, **p< .01, ***p< .001, two-tailed.

Figure 3. Cross-lagged structural equation panel model for those who reported no back pain at baseline and those who did report back pain at baseline (model 2b).

*p< .05, **p< .01, ***p< .001, two-tailed.

(13)

for example, about unsafe movements that will result in persistent harm are not substan- tiated. Many of their pessimistic back beliefs are revised after experiencing that activity does not result in persistent pain, but– gradually–in the reduction of pain. The passive lifestyle that is associated with pessimistic back beliefs can be changed by activity management training (i.e. pacing). Other techniques address attention control and motivation (goal setting, self-reinforcement, etc.). Cognitive behavioral therapy is interdisciplinary, tries to increase self-efﬁcacy in pain management and is skill-oriented (Gatchel, McGeary, McGeary, & Lippe, 2014). The key factor of a good interdisciplinary approach is communication and coordination among physicians, nurses, clinical psychol- ogists, physical therapists, and occupational therapists (Gatchel et al., 2014). Another point is that health care providers often have those pessimistic beliefs, too (Linton, Figure 4. Cross-lagged structural equation panel model for those who exercised at baseline and those did not exercise at baseline (model 2d).

*p< .05, **p< .01, ***p< .001, two-tailed.

Figure 5. Cross-lagged structural equation panel model for those who reported back pain and exercise at baseline and those who reported back pain and no exercise at baseline (model 2f ).

(14)

Vlaeyen, & Ostelo, 2002). Any attempts to modify beliefs should, therefore, target all users of and providers within the health care system to prevent mixed messages from negating any positive effects of campaigns that otherwise lead to better-educated patients. Insurers and workers compensation insurance companies should rely primarily on public health education programs and the development, communication, and use of adequate clinical practice guidelines. Insurers should favor interdisciplinary treatments even if they are more expensive at the start, since they pay off in the long term. A case management approach is recommended. Mass media campaigns may help.

In the cross-lagged-panel models, pessimistic back beliefs predicted an increase in pain in the shoulders, neck, and back at 1-year follow-up, but only in those with back pain at baseline and those who did not exercise at baseline. The BBQ scores were very stable over time. Perhaps this explains why the modiﬁcation of beliefs via media campaigns is so challenging and time-consuming (Bostick et al., 2013). Nonetheless, efforts to do so should not be discouraged, since they are likely to pay off with respect to health care utilization and other important outcomes, especially in risk groups of inactive individuals and those already suffering from pain. Analysis of combined risk factors showed that the presence of pain and no exercise was a high-risk combination. In this sense, exercise is perhaps preventive, especially in the presence of back pain.

Strengths and limitations

This is the ﬁrst population-based, longitudinal validation study to evaluate the stability of the BBQ and reverse causation, that is, pain predicting an increase in BBQ over time.

The results clearly showed that pain did not trigger pessimistic back beliefs because the reverse causation path was not significant. However, only a randomised controlled trial can prove causality and further studies should include randomised controlled trials to evaluate whether specific BBQ-directed public health interventions can modify back beliefs and subsequent pain reporting. The large study sample allowed sufficient power in longitudinal analysis (almost 1, Preacher & Coffman, 2006). Participation rates were high, in both the baseline and follow-up surveys. Limitations include the fact that sampling was restricted to 2860 individuals from an earlier population-based study (prestratified regarding back pain, prior to random selection) and to those who werefluent in German, in the German-speaking part of the country only. Because all measures were self-reported, bias from common source variance may have boosted correlations (cf.

Semmer, Grebner, & Elfering,2004).

Conclusion

The German version of the BBQ showed good reliability and a robust factorial structure. Its availability allows wide-spread assessment of back beliefs and their inﬂuence on the reporting of back pain and should promote further studies aimed at reducing pessimistic back beliefs in the in German-speaking population by mass media campaigns (analogous to e.g. Buchbinder & Jolley, 2005). It can be used to evaluate pessimistic back beliefs in health care providers and their affect on the patient’s recovery (as done by Silcock, Moffett, Edmondson, Waddell, & Burton, 2007 and Briggs et al., 2013, respectively). Theﬁndings of this study suggest that pessimistic back beliefs should be included in screening for those at risk of developing persistent pain.

(15)

Funding

The research reported in this article was supported by a grant from the Swiss National Science Foundation, National Research Program NRP53‘Health –Chronic Pain’ entitled‘Deﬁning norm values on the natural history of acute and chronic low back pain in the Swiss population and prognostic indicators for chronic low back pain’ (Project 405340-104826/1) and project “From norm values and predictive items in low back pain to operational tools for health care professionals”(405340-104826/2).

References

Borrell-Carrió, F., Suchman, A. L., & Epstein, R. M. (2004). The biopsychosocial model 25 years later: Principles, practice, and scientiﬁc inquiry.The Annals of Family Medicine, 2, 576–582.

doi:10.1370/afm.245

Bostick, G. P., Schopﬂocher, D., & Gross, D. P. (2013). Validity evidence for the back beliefs questionnaire in the general population. European Journal of Pain, 17, 1074–1081.

doi:10.1002/j.1532-2149.2012.00275.x

Briggs, A. M., Slater, H., Smith, A. J., Parkin-Smith, G. F., Watkins, K., & Chua, J. (2013). Low back pain-related beliefs and likely practice behaviours amongﬁnal-year cross-discipline health students.European Journal of Pain, 17, 766–775. doi:10.1002/j.1532-2149.2012.00246.x Buchbinder, R., & Jolley, D. (2005). Effects of a media campaign on back beliefs is sustained 3

years after its cessation.Spine, 30, 1323–1330. doi:10.1097/01.brs.0000164121.77862.4b Clegg, C. W., Jackson, P. R., & Wall, T. D. (1977). The potential of cross-lagged correlation anal-

ysis inﬁeld research. Journal of Occupational Psychology, 50, 177–196. doi:10.1111/j.2044- 8325.1977.tb00374.x

Cohen, J. (1977). Statistical power analysis for the behavioral sciences. San Diego, CA: Aca- demic Press.

Dionne, C. E., Bourbonnais, R., Fremont, P., Rossignol, M., Stock, S. R., & Larocque, I. (2005).

A clinical return-to-work rule for patients with back pain. Canadian Medical Association Journal, 172, 1559–1567. doi:10.1503/cmaj.1041159

Elfering, A., & Mannion, A. F. (2008). Epidemiology and risk factors of spinal disorders. In N. Boos & M. Aebi (Eds.), Spinal disorders – Fundamentals of diagnosis and treatment (pp. 153–173). Berlin: Springer.

Elfering, A., Mannion, A. F., Jacobshagen, N., Tamcan, O., & Müller, U. (2009). Beliefs about back pain predict the recovery rate over 52 consecutive weeks.Scandinavian Journal of Work, Environment & Health, 35, 437–445. doi:10.5271/sjweh.1360

Engel, G. L. (1977). The need for a new medical model: A challenge for biomedicine. Science, 196, 129–136.

Gatchel, R. J., McGeary, D. D., McGeary, C. A., & Lippe, B. (2014). Interdisciplinary chronic pain management: Past, present, and future. American Psychologist, 69, 119–130. doi:10.103 7/a0035514

Golembiewski, R. T., Billingsley, K., & Yeager, S. (1976). Measuring change and persistence in human affairs: Types of change generated by OD designs.The Journal of Applied Behavioral Science, 12, 133–157. doi:10.1177/002188637601200201

Gross, D. P., Deshpande, S., Werner, E. L., Reneman, M. F., Miciak, M. A., & Buchbinder, R.

(2012). Fostering change in back pain beliefs and behaviors: When public education is not enough.The Spine Journal, 12, 979–988. doi:10.1016/j.spinee.2012.09.001

Kendall, N. A. S., Linton, S. J., & Main, C. J. (2004).Guide to assessing psychosocial yellow ﬂags in acute low back pain: Risk factors for long term disability and work loss. Wellington:

Accident Rehabilitation and Compensation Insurance Corporation of New Zealand and the National Health Committee. Retrieved from http://www.acc.co.nz/PRD_EXT_CSMP/groups/

external_ip/documents/internet/wcm002131.pdf

Kline, P. (2000).The handbook of psychological testing(2nd ed.). London: Routledge.

Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data.

Biometrics, 33, 159–174.

Larsson, M. E. H., & Nordholm, L. A. (2008). Responsibility for managing musculoskeletal disorders – A cross-sectional postal survey of attitudes. BMC Musculoskeletal Disorders, 9, 110.

doi:10.1186/1471-2474-9-110

(16)

Linton, S. J., Vlaeyen, J., & Ostelo, R. (2002). The back pain beliefs of health care providers:

Are we fear-avoidant? Journal of Occupational Rehabilitation, 12, 223–232. doi:10.1023/

a:1020218422974

Mannion, A. F., Horisberger, B., Eisenring, C., Tamcan, O., Elfering, A., & Müller, U. (2009).

The association between beliefs about low back pain and work presenteeism. Journal of Occupational and Environmental Medicine, 51, 1256–1266. doi:10.1097/JOM.0b013e3181- beac69

Mannion, A. F., Wieser, S., & Elfering, A. (2013). Association between beliefs and care-seeking behavior for low back pain.Spine, 38, 1016–1025. doi:10.1097/BRS.0b013e31828473b5 Melloh, M., Elfering, A., Stanton, T., Barz, T., Aghayev, E., Röder, C., & Theis, J.-C. (2014).

Low back pain risk factors associated with persistence, recurrence, and delayed presentation.

Journal of Back and Musculoskeletal Rehabilitation, 27, 281–289. doi:10.3233/BMR-130445 Müller, U., Tänzler, K., Bürger, A., Staub, L., Tamcan, O., Roeder, C., … Trelle, S. (2008). A

pain assessment scale for population-based studies: Development and validation of the pain module of the standard evaluation questionnaire. Pain, 136, 62–74. doi:10.1016/j.pain.

2007.06.014

Preacher, K. J., & Coffman, D. L. (2006, May).Computing power and minimum sample size for RMSEA[Computer software]. Retrieved fromhttp://quantpsy.org/

Rolli Salathé, C., & Elfering, A. (2013). A health- and resource-oriented perspective on NSLBP.

ISRN Pain, 2013, 1–19. doi:10.1155/2013/640690

Rolli Salathé, C., Elfering, A., & Melloh, M. (2012). Wirksamkeit, Zweckmässigkeit und Wir- tschaftlichkeit des multimodalen Behandlungsansatzes bei chronisch lumbalen Rückenschmer- zen [Efﬁcacy, utility and cost-effectiveness of multidisciplinary treatment for chronic low back pain].Der Schmerz, 26, 131–149. doi:10.1007/s00482-012-1148-2

Rolli Salathé, C., Kälin, W., Semmer, N. K., Roth, M., Müller, U., Melloh, M., & Elfering, A.

(2013). Comparison of a pain-resilient group of working individuals to population-based case controls with and without acute low back pain. European Pain Journal, 17, 1411–1421.

doi:10.1002/j.1532-2149.2013.00319.x

Schermelleh-Engel, K., Moosbrugger, H., & Müller, H. (2003). Evaluating the fit of structural equation models: Tests of significance and descriptive goodness-of-fit measures. Methods of Psychological Research Online, 8, 23–74.

Semmer, N. K., Grebner, S., & Elfering, A. (2004). Beyond self-report: Using observational, physiological, and event-based measures in research on occupational stress. In P. L. Perrewé

& D. C. Ganster (Eds.), Emotional and physiological processes and positive intervention strategies.Research in occupational stress and well-being(Vol. 3, pp. 205–263). Amsterdam:

JAI. doi:10.1016/S1479-3555(03)03006-3

Sharma, S., Mukherjee, S., Kumar, A., & Dillon, W. (2005). A simulation study to investigate the use of cutoff values for assessing model ﬁt in covariance structure models. Journal of Busi- ness Research, 58, 935–943. doi:10.1016/j.jbusres.2003.10.007\

Silcock, J., Moffett, J. K., Edmondson, H., Waddell, G., & Burton, A. K. (2007). Do community pharmacists have the attitudes and knowledge to support evidence based self-management of low back pain?BMC Musculoskeletal Disorders, 8, 10. doi:10.1186/1471-2474-8-10

Symonds, T. L., Burton, A. K., Tillotson, K. M., & Main, C. J. (1996). Do attitudes and beliefs inﬂuence work loss due to low back trouble?Occupational Medicine, 46, 25–32. doi:10.1093/

occmed/46.1.25

Woby, S. R., Watson, P. J., Roach, N. K., & Urmston, M. (2004). Are changes in fear-avoidance beliefs, catastrophizing, and appraisals of control, predictive of changes in chronic low back pain and disability?European Journal of Pain, 8, 201–210. doi:10.1016/j.ejpain.2003.08.002