Expectation of Recovery from Low Back Pain: A Longitudinal Cohort Study
Investigating Patient Characteristics Related to Expectations and
the Association Between Expectations and 3-month Outcome

This section is compiled by Frank M. Painter, D.C.
Send all comments or additions to:

FROM:   Spine (Phila Pa 1976). 2014 (Jan 1);   39 (1):   81–90 ~ FULL TEXT

Alice Kongsted, PhD, Werner Vach, PhD, Marie Ax0, MSc(Clin Biomech),
Rasmus Norgaard Bech, MSc(Clin Biomech), and Lise Hestbaek, PhD

Nordic Institute of Chiropractic and Clinical Biomechanics,
part of Clinical Locomotion Network,
Odense, Denmark

STUDY DESIGN:   A prospective cohort study conducted in general practice (GP) and chiropractic practice (CP).

OBJECTIVE:   To explore which patient characteristics were associated with recovery expectations in patients with low back pain (LBP), whether expectations predicted 3-month outcome, and to what extent expectations were associated with empirical prognostic factors.

SUMMARY OF BACKGROUND DATA:   Patients' recovery expectations have been associated with prognosis, but it is largely unknown why patients expect what they do, and how expectations relate to other prognostic factors.

METHODS:   A total of 1169 participants completed questionnaires at their first consultation due to LBP, and 78% were followed for 3 months. At baseline, recovery expectations were measured on a 0 to 10 scale. Outcome measures were LBP intensity and global perceived effect. Associations were tested in regression models, and the predictive capacity of expectations described in terms of adjusted R and area under the receiver operating characteristic curve. Correlations between predicted expectations and prognosis were quantified by the Spearman rho.

RESULTS:   Expectations were associated with known prognostic factors, mainly LBP history, but were only partly explained by measured factors (adjusted R, 35% [CP]/55% [GP]). Expectations had statistically significant associations with both outcomes after adjusting for other baseline factors, but explained only a little of the variance in LBP (adjusted R: 0.11 CP/0.32 GP) and did not add to the explained variance. The prediction of global perceived effect was limited in CP (area under the receiver operating characteristics curve, 0.59), but more substantial in GP (area under the receiver operating characteristics curve, 0.77) patients. Correlations between predicted expectations and predicted outcome were strong.

CONCLUSION:   Patients' recovery expectations were associated mainly with LBP history and were generally, but not consistently, similar to an empirically predicted prognosis. Expectations were significantly associated with outcome, and may, at least for some outcomes, be a relevant proxy for more complex models. Future studies should explore the effect of addressing negative recovery expectations.


From the FULL TEXT Article:


Low back pain (LBP) is an extremely common and burdensome condition [1] for which there are no truly effective treatments. Identification of potentially modifiable prognostic factors may assist in improving care, [2] and patients' expectations of recovery or return to work have been shown to be prognostic factors with relatively strong and very consistent associations with outcome in LBP [3-8] as well as in a number of other conditions. [9, 10]

However, it is not clear whether recovery expectations are associated with outcome because they reflect other prognostic factors, or if recovery expectations have a causal relationship with outcome, acting in a self-fulfilling manner. To explore causality and to improve our understanding of the potential for modifying expectations, it is essential to know which patient characteristics are associated with expectations.

Although the construct of expectations has been investigated, [6, 11, 12] little is known about the factors that form the expectations of patients with LBP. Moreover, the construct of expectations in an episodic and idiopathic condition such as LBP is likely to differ from conditions such as the Alzheimer disease [11] or after a traumatic injury. [12]

In LBP, a cohort study has demonstrated associations between return-to-work expectations in work-related LBP and pain, mood, prior back pain, job demands, functional limitation, and marital status. [6] In a qualitative study, multiple factors also emerged as important for expectations of patients with LBP, [13] but associations between patient characteristics and recovery expectations in those seeking care for their LBP are unexplored.

The objectives of this study were to explore which baseline characteristics were associated with LBP patients' expectations of recovery, whether recovery expectations predicted outcome after 3 months, and whether these expectations were similar to a prognosis predicted from other known prognostic factors.

These objectives were explored in a sample of Danish patients consulting for LBP in primary care. In Denmark, general practice (GP) and chiropractic practice (CP) are the means by which patients with LBP first gain access to the public health care system, and thus a combination of the 2 settings would represent first-line LBP care. There are several significant differences between patients in the 2 settings and therefore the objectives are explored separately in the 2 cohorts.


Participants were recruited between September 2010 and January 2012 from 17 CP clinics in the research network of the Nordic Institute for Chiropractic and Clinical Biomechanics in Denmark, and from Eebruary to May 2011 from GPs in the Region of Southern Denmark who accepted an invitation to participate in a quality assurance program. [14] Baseline questionnaires were completed in the reception area before the initial consultation in CPs, whereas in GPs, patients were handed a questionnaire and a prepaid envelope and asked to complete the survey at home. A follow-up questionnaire was mailed after 3 months, and nonresponders were reminded by phone.

Treatment was not affected by participation in the study, and the local ethics committee determined that the study did not need ethical approval according to Danish rules.'^ The study was conducted in agreement with the Declaration of Helsinki.


Clinicians at the recruiting practices were asked to invite consecutive patients to participate who were aged 18 to 65 years, attending the CP or GP for the first time because of their current episode of LBP, and who could read Danish. Exclusion criteria were nonresponse on the expectations question, suspicion of inflammatory or pathological pain, and nerve root involvement requiring acute referral to surgery. In CP, additional exclusion criteria were pregnancy and having had more than 1 health care consultation due to LBP within the previous 3 months.


At baseline, participants were asked: "How likely do you think it is that you will be fully recovered in 3 months?" The response was on a 0 to 10 numeric rating scale (0 = not at all likely; 10 = very likely). A time-based single-item 82 www.spinejournal.com tool specifically asking about expectations of return to work was previously found to be the optimal measure of recovery expectations. [3] Eor this study, a time-based, single-item question was used but not related to return to work because only a small proportion of participants were expected to be sick listed at the time of inclusion.

      Other Baseline Characteristics

The baseline characteristics investigated as potentially related to a patient's expectations were from the domains of LBP history, pain, activity limitation, psychology, general health, and social factors. The following variable definitions were used: previous LBP episodes (0,1-2,3 or more), LBP last year (^30 d, >30 d), duration (0-2 wk, 2-4 wk, 1-3 mo, or >3 mo), LBP intensity last week (0-10 numeric rating scale, 0 = no pain; 10 = worst imaginable pain [16, 17]), leg pain intensity (as for LBP), activity limitation (proportional score Roland-Morris Disability Questionnaires 0-100 [18]), depressive symptoms (sum score Major Depression Inventory 0-50 [19]), fear avoidance (Eear Avoidance Beliefs Questionnaire physical activity scale [0-24] [20]), self-rated general health (EQ5D VAS scale 0 = worst imaginable health state; 100 = best imaginable health state [21]) education (no qualification, vocational training, higher education <3 yr, higher education 3-4 yr, higher education >4 yr), heavy physical work (heavy physical work vs. nonheavy [sitting, sitting/walking, and light physical] work), and age (yr). The work scale of the Eear Avoidance Beliefs Questionnaire was not included because it also measures recovery expectations.

      Outcome Measures

The outcome measures were LBP intensity (0-10 numeric rating scale) and improvement ("much better" or "better" on a 7-point Global Perceived Effect Likert scale [22]) after 3 months.

      Data Analyses

Data were entered twice in EpiData. [23] Analyses were performed in STATA/SE 12.1 (STATA Corp, College Station, TX). GP and CP patients were analyzed separately because they differed on many factors, with GP patients having generally lower expectations and more complex profiles than CP patients.

Multiple imputations based on fully conditional specifications were used with 5 chained iterations to impute missing values. [24] Eor each analysis, the imputations were based on a model that included all variables, but without replacing missing values on outcome measures. All baseline variables except for age, sex, and expectations were affected by missing values, with a maximum prevalence of 7%.

      Cross-Sectional Analyses

To explore the associations between expectations and other baseline characteristics, univariate linear regression analyses were initially performed of all baseline measures, followed by multivariable linear regression. We also investigated whether the factors most strongly associated with expectations had that effect moderated by other characteristics using tests of interactions between the factor and the potential moderator.

      Longitudinal Analyses

Associations between expectations and outcome were tested by means of linear regression (for the outcome LBP intensity) and logistic regression (outcome improvement) and presented as β-coefficients and odds ratios with 95% confidence intervals (CIs), respectively. In addition, predictive models without the expectation variable were formed, that included all other baseline values, to investigate if adding recovery expectations substantially increased the explanatory power of the model in terms of

(1)   adjusted R- for LBP change and
(2)   area under the receiver operating characteristics curve (AUC) for improvement.

AUC was validated by means of a leave-oneout cross validation. Joint models for both cohorts with interaction terms between care setting and covariates were used to assess differences between settings.

Correlations (Spearman rho [p]) between observed recovery expectations, as well as expectations predicted from all other baseline characteristics, and the predicted 3-month outcomes from models also including all other baseline variables (hereafter denoted empirical prognosis) were calculated to investigate the degree to which patients' expectations were driven by the empirical prognosis.

Potential dependency between observations from the same clinics was taken into account using the STATA cluster option for robust variance estimation with clinics (CP) or clinicians (GP) as cluster level in all regression analyses.



In total, 1,169 participants were included, 928 from CP and 241 from CP (Figure 1). The baseline characteristics of the cohorts are summarized in Table 1. The response rates at follow- up were 76% from CP and 83% from GP. At baseline, the nonresponders had on average 0.7 points lower expectations (F = 0.001), slightly higher leg pain intensity (β = 0.05, P = 0.01), and depression scores (β = 1.7, P = 0.01), more frequently an episode duration more than 3 months (odds ratio = 1.7, P = 0.01), and were on average 5.1 years younger (P < 0.01) compared with the responders. The association between baseline characteristics and expectations did not differ between responders and nonresponders.


The median expectations score was 9 (interquartile range [IQR], 7-10) in CP and 6 (IQR, 3-9) in GP. In CP patients, 48% (95% CI, 45%-51%) had maxtmal recovery expectations (expectations score =10), whereas this was the case in 22% (95% CI, 16%-27%) of GP patients (Figure 2).

      Baseline Characteristics Associated With Recovery Expectations

Univariate Associations   Expectations were univariately associated with education, previous LBP episodes, LBP days last year, episode duration, leg pain intensity, depressive symptoms, and self-rated health in both cohorts (Tables 2, 3). In CP patients, age, sex, and heavy work were also associated with recovery expectations.

Multivariable Baseline Analyses   Previous episodes and duration were the only significant associations with recovery expectations in both cohorts in the multivariable models (Tables 2, 3). We tested if the effects of these factors were moderated by education, depression, LBP, or leg pain intensity. In CP patients, they were not. In GP patients, depression added to the negative effect of duration (0.1 additional decrease in expectations for each point on the depression scale in patients with a duration >1 mo, P = 0.03).

In the CP cohort, associations were also significant for age and LBP days last year (Table 2), whereas sex, education, and leg pain intensity were significantly associated with recovery expectations in GP patients (Table 3).

Measured parameters explained 35% (adjusted R2) of the variance in the recovery expectations measure in CP patients and 55% in GP patients.

      Expectations as Predictors of Outcome

LBP Intensity   After 3 months, median LBP intensity was 1 (IQR, 0-2) in CP and 2 (IQR, 1-5) in GP patients. Higher expectations at baseline were univariately associated with lower LBP intensity 3 months later (Table 4), with a stronger association in GP patients than in CP patients (interaction term P < 0.001) (Figure 3). Including all other baseline parameters in the model, associations were reduced but still statistically significant for both cohorts. The adjusted R- values were largely unaltered by adding expectations to a model consisting of all other predictors (Table 4).

Improvement   In the CP cohort, 83% (95% CI, 80%-86%) were improved at 3-month follow-up compared with 60% (95% CI, 53%- 67%) in the GP cohort. Higher baseline recovery expectations were associated with a higher probability of improvement in univariate (Figure 3, Table 4), and multivariable models (Table 4). Odds ratios differed between settings with stronger associations in GP patients (P = 0.02 crude and adjusted models).

In GP patients, a model consisting of all other baseline factors had higher predictive capacity for improvement (AUG = 0.71) than expectations alone (AUG = 0.59). Expectations alone (AUG = 0.77) predicted as well as the more complex model (AUG = 0.75) in GP patients. Adding recovery expectations to a model containing other baseline measures did not increase AUG values considerably in any of the cohorts (Table 4).

      Correlations Between Recovery Expectations and Empirical Prognosis

Taking the predicted outcomes from a model including all other prognostic factors except for expectations as an empirical measure for the prognosis, we could observe moderate correlations between prognosis and expectations: for LBP intensity p = -0.49 (P < 0.001) in the GP and p = -0.67 (P < 0.001) in the GP cohort, and correspondingly, for improvement, p = 0.47(P < 0.001) and p = 0.60 (P < 0.001). These moderate correlations suggest variability between empirical prognosis and expectations in individuals (Figure 4). To get some insight into a possible general difference between expectations and prognosis, we calculated the correlation between expectations predicted from other prognostic factors and the empirical prognosis, and found strong correlations for both LBP intensity (GP: p = -0.78, P < 0.001. GP: p = -0.87, P < 0.001) and improvement (GP: p = 0.64, P < 0.001, GP: p = 0.78, P < 0.001). This illustrated that, in general, the measured predictors affected patients' expectations and the empirical prediction in similar ways (Figure 4). Flowever, the correlations were below one, indicating the predictive model and patients' expectations weighted these factors differently.


Our main findings were:

(1)   that patients' recovery expectations were related to previous LBP experience more than to symptom severity or to the considered psychological factors,

(2)   that recovery expectations were associated with outcome independently of other measured factors,

(3)   that expectations were generally a good proxy for other measured prognostic factors but did not add predictive accuracy to that of the other factors, and

(4)   that patients' expectations generally were similar to an empirically derived prognosis. However, it should be recognized that the empirical model did not predict outcomes very accurately.

This means that patients generally had realistic expectations, although some considered their prognosis to be different from that which we would predict from known prognostic factors. It also means that in primary care, clinicians should pay attention to previous experience in patients with low expectations rather than focusing on psychological factors such as depressive symptoms and fear avoidance beliefs. However, factors measured in the study only explained part of the variance in expectations, suggesting that recovery expectations result from a more complex individual process and although we included factors covering a wide range of health domains, other measurable factors may better explain those aspects that inform expectations. Self-efficacy and illness beliefs are potentially relevant factors which theoretically relate to recovery expectations [25] and should be investigated as mediators of the association between recovery expectations and outcome.

It remains somewhat unclear if expectations are useful as a substitute for other known prognostic factors. In the prediction of improvement in GP patients expectations predicted outcome as well as a model with many other baseline factors, whereas this was not the case in CP patients or in relation to LBP intensity.

Associations with outcome were stronger in the GP than the CP cohort. This may be influenced by the GP patients having more previous episodes and a longer duration of LBP. In other words, GP patients' more profound experience with LBP may help them make a better prediction of their prognosis, and it may be that expectations are more important in more severely affected patients, who were more numerous in the GP cohort. Alternatively, the different effects of expectations may be a result of the interventions offered that could moderate the effects of expectations in different ways.

We did not investigate the possible effects of the interventions and cannot tell to what extent that played a role. Finally, the observed differences between settings could have been an effect of different timing of the expectations question. CP patients were asked about expectations prior to the first consultation whereas GP patients completed the questionnaire at home after the consultation. Having negative expectations after, rather than prior to, a consultation may be more important. This makes sense if clinicians generally are able to reassure patients during the consultation. Unfortunately, effects of setting and timing cannot be separated in this study.

Other limitations were that although the practices were instructed to include patients consecutively, the actual source population is unknown. Also, drop out may have introduced bias, but response rates were acceptable and baseline factors associated with expectations did not differ between responders and nonresponders. Finally, the reliability and validity of the expectations question was not established. It was deemed necessary to use a single-item question because the cohorts were asked a very high number of questions, and the wording of the recovery question was based on previous studies. [3]

The main strength of this study was the adequate sample size with complete data and enrollment from 2 settings. In an explorative study like this, we also considered the variability in duration a strength, as it allowed us to study the importance of symptom duration. Moreover, the choice of investigated patient characteristics was not restricted by a theoretical framework about the construct of expectations. This is needed to explore whether our thinking of recovery expectations should be broadened to encompass other elements. Conversely, as mentioned earlier important aspects may have been missed using this approach.


Recovery expectations were associated mainly with LBP history and were more complex than that which could be explained by other measured factors. Patients' expectations reflected that most patients assessed known prognostic factors in an empirically "reasonable" way. Nevertheless, some had expectations that seemed unrealistically low or high, although this could only be judged against a rather inaccurate model. Expectations were significandy associated with outcome, but causal pathways between expectations and outcomes are still to be explored. Given that other measured factors seemed to relate more strongly to prognosis than expectations, there is good justification for attempting to increase expectations in patients who have a good predicted prognosis but low expectations. Euture studies should explore the implications of having expectations that seem unrealistic, whether expectations can be modified, and if that alters outcome.

Key Points

  • Recovery expectations of patients witb LBP are a complex construct and associated mostly with pain history.

  • Patients' expectations were generally similarto an empirically predicted prognosis.

  • Recovery expectations were associated witb outcome independently of other measured factors.

  • Expectations did not increase tbe predictive accuracy when added to a model consisting of known prognostic factors, but they were a relevant proxy for some outcome measures.

  • It may be useful to investigate if expectations can be modified and if so, whether that would affect prognosis.


The authors thank the chiropractors in the KIP research network and the participating general practitioners for their participation in data collection. They also thank APO Odense for establishing the contact with general practitioners. Lastly, they thank the research assistants Jytte Johannesen and Orla Lund Nielsen for taking care of all the logistics.


  1. Vos T, Flaxman AD, Naghavi M, et al.
    Years Lived with Disability (YLDs) for 1160 Sequelae of 289 Diseases and Injuries 1990-2010:
    A Systematic Analysis for the Global Burden of Disease Study 2010

    Lancet. 2012 (Dec 15);   380 (9859):   2163–2196

  2. Hill JC, Whitehurst DG, Lewis M et al.
    Comparison of Stratified Primary Care Management for Low Back Pain
    with Current Best Practice (STarT Back): A Randomised Controlled Trial

    Lancet. 2011 (Oct 29); 378 (9802): 1560–1571

  3. lies RA, Davidson M, Taylor NF, et al.
    Systematic review of the ahility of recovery expectations to predict outcomes in non-chronic
    non-specific low hack pain.
    Occup Rehabil 2009;l9:25-40.

  4. Laisne F, Lecomte C, Corhiere M.
    Biopsychosocial predictors of prognosis in musculoskeletal disorders: a systematic review of the literature.
    Disabil Rehabil 2012;34:355-82.

  5. Reme SE, Hagen EM, Eriksen HR.
    Expectations, perceptions, and physiotherapy predict prolonged sick leave in subacute low hack pain.
    BMC Musculoskelet Disord 2009; 10:139.

  6. Kapoor S, Shaw WS, Pransky G, et al.
    Initial patient and clinician expectations of return to work after acute onset of work-related low hack pain.
    Occup Environ Med 2006;48:1173-80.

  7. Turner JA, Franklin G, Fulton-Kehoe D, et al.
    Worker recovery expectations and fear-avoidance predict work disability in a population-hased
    workers' compensation hack pain sample.
    Spine (Phila Pa 1976) 2006;31:682-9.

  8. Hallegraeff JM, Krijnen WP, van der Schans CP, et al.
    Expectations ahout recovery from acute non-specific low back pain predict absence from usual work
    due to chronic low hack pain: a systematic review.
    Physiother 2012;58:165-72.

  9. Carroll LJ.
    Beliefs and expectations for recovery, coping and depression in whiplash associated disorders:
    lessening the transition to chronicity.
    Spine (Phila Pa 1976) 2011;36(25 suppl): S250-6.

  10. Mondloch MV, Cole DC, Frank JW.
    Does how you do depend on how you think you'll do? A systematic review of the evidence for a
    relation between patients' recovery expectations and health outcomes.

  11. Janzen JA, Silvius J, Jacobs S, et al.
    What is a health expectation? Developing a pragmatic conceptual model from psychological theory.
    Health Expect 2006;9:37-48.

  12. Ozegovic D, Carroll LJ, Cassidy JD.
    Factors associated with recovery expectations following vehicle collision: a population-hased study.
    Rehabil Med 2010;42:66-73.

  13. lies RA, Taylor NF, Davidson M, et al.
    Patient recovery expectations in non-chronic non-specific low hack pain: a qualitative investigation.
    Rehabil Med 2012;44:781-7.

  14. Poulsen L, Munck A.
    Lotv Back Pain - from Primary Consultation to end of Treatment.
    1st Audit Registration Danish: Audit Project Odense; 2011.

  15. Danish National Commitee on Biomédical Research Ethics,
    [published online ahead of print October 4, 2011]
    Guidelines ahout Notification. 2011. Available at
    October 9, 2013.

  16. Jensen MP, Turner JA, Romano JM, et al.
    Comparative reliability and validity of chronic pain intensity measures.
    Pain 1999;83: 157-62.

  17. Bolton JE, Wilkinson RC.
    Responsiveness of pain scales: a comparison of three pain intensity measures in chiropractic patients.
    Manipulative Physiol Ther 199 8 ;21:1-7.

  18. Kent P, Lauridsen HH.
    Managing missing scores on the Roland Morris Disability Questionnaire.
    Spine (Phila Pa 1976) 2011;36:1878-84.

  19. Bech P, Rasmussen NA, Olsen LR, et al.
    The sensitivity and specificity of the Major Depression Inventory, using the Present State Examination
    as the index of diagnostic validity.
    Affect Disord 2001;66:159-64.

  20. Waddell G, Newton M, Henderson I, et al.
    A Fear-Avoidance Beliefs Questionnaire (FABQ) and the role of fear- avoidance beliefs in
    chronic low back pain and disability.
    Pain 1993;52: 157-68.

  21. Rabin R, de Charro F.
    EQ-5D: a measure of health status from the EuroQol Group.
    Ann Med 2001;33:337-43.

  22. Lauridsen HH, Hartvigsen J, Korsholm L, et al.
    Choice of external criteria in back pain research: Does it matter?
    Recommendations based on analysis of responsiveness.
    Pain 2007;131:112-20.

  23. Lauritsen JM.
    EpiData Data Entry. Data Management and basic Statistical Analysis System.
    Odense, Denmark, EpiData Association, 2008.

  24. Vergouw D, Heymans MW, van der Windt DA, et al.
    Missing data and imputation: a practical illustration in a prognostic study on low back pain.
    Manipulative Physiol Ther 2012;35:464-71.

  25. Main CJ, Foster N, Buchbinder R.
    How important are back pain beliefs and expectations for satisfactory recovery from back pain?
    Best Pract Res Clin Rheumatol 2010;24:205-17.



Since 9-04-2019

                       © 1995–2020 ~ The Chiropractic Resource Organization ~ All Rights Reserved