Journal of Occupational & Environmental Medicine:
The Pain Recovery Inventory of Concerns and Expectations: A Psychosocial Screening Instrument to Identify Intervention Needs Among Patients at Elevated Risk of Back Disability
Shaw, William S. PhD; Reme, Silje Endresen PhD; Pransky, Glenn MD, MOccH; Woiszwillo, Mary Jane BS; Steenstra, Ivan A. PhD; Linton, Steven J. PhD
From the Liberty Mutual Research Institute for Safety (Drs Shaw, Reme, and Pransky and Ms Woiszwillo), Hopkinton, Mass; University of Massachusetts Medical School (Drs Shaw and Pransky), Worcester, Mass; Harvard School of Public Health (Dr Reme), Boston, Mass; Institute for Work and Health (Dr Steenstra), Toronto, Ontario, Canada; and Center for Health and Medical Psychology (Dr Linton), School of Law, Psychology and Social Work, Örebro University, Örebro, Sweden.
Address correspondence to: William S. Shaw, PhD, Liberty Mutual Center for Disability Research, 71 Frankland Rd, Hopkinton, MA 01748 (firstname.lastname@example.org).
This study was an intramural research project of the Liberty Mutual Research Institute for Safety, with travel support from a 2006 International Association for the Study of Pain research grant funded by the Scan/Design Foundation by Inger and Jens Bruun (Linton and Shaw).
Disclosure: The authors declare no conflict of interest.
Objective: To reduce a full psychosocial test battery to a brief screening questionnaire to triage return-to-work strategies among patients with low back pain (LBP).
Methods: Workers (N = 496) with acute, work-related LBP completed multiple psychosocial measures at intake, then a 3-month follow-up of pain, function, and work status. A sensitivity analysis was conducted to reduce the number of items while maintaining scale reliability, preserving associations with outcomes, and maintaining separation between patient subgroups.
Results: The pool of items was trimmed from 129 to 46 items, describing elements of emotional distress, pain beliefs, organizational support, and activity limitation. A confirmatory cluster analysis replicated previous findings of three risk subgroups: distressed, avoidant, and lacking employer support.
Conclusions: The reduced measure is a reliable and valid screening measure that can be used to identify early intervention needs among working adults with LBP.
There is a growing body of evidence supporting the importance of psychosocial and workplace concerns (“yellow flags”) as prognostic factors in low back pain (LBP) recovery and return to work.1–11 Early assessment of psychosocial factors is recommended in many treatment guidelines,7,12–14 and this issue has been highlighted as an important component of quality care in “best practices” summaries.6,15–21 But, despite a widespread acceptance of the biopsychosocial framework for medical management of LBP,22 psychosocial issues are rarely raised by patients with LBP,23 and frontline providers often lack referral options for dealing with psychosocial issues.24 Thus, there is a continuing need to develop feasible psychosocial assessment and intervention methods in occupational medicine and other clinical settings to more effectively triage efforts to prevent chronic back disability.
An essential question is whether psychosocial factors can be reliably and feasibly assessed and incorporated into patient education and clinical decision making.25 A number of self-report screening measures have been developed,26–28 but scores typically indicate only whether a patient is at elevated risk, with no ability to discern specific problems that might be the target of intervention. When used to triage patients in clinical trials, at least one of these measures (the STarTBack screening tool) (Keele University, Keele, Staffordshire, UK)29 has shown modest improvements in patient outcomes and a significant reduction in treatment costs overall (primarily by reducing unnecessary treatments to low-risk patients).30 In that study, primary care patients at elevated risk received psychologically informed physiotherapy sessions. Although the STarTBack trial provides preliminary support for psychosocial triage of patients with LBP, more research is needed to refine psychosocial screening methods and evaluate them in a variety of clinical settings.
Cluster analysis of patient self-report data is one method that has shown preliminary support for delineating meaningful LBP subgroups, especially those with and without emotional distress.31–36 In a previous patient cohort, we identified four preliminary patient subgroups on the basis of disability risk factors that are as follows: those with few or no concerns (“low risk”), those lacking employer support, those with high emotional distress, and those with physical restriction without distress31; and these results were replicated in the current patient cohort by using a full psychosocial test battery.32 Three-month functional outcomes and return to work were poorest in the emotional distress group and intermediate in the other two risk groups, when compared with low-risk patients. In an occupational medicine setting, such classifications may provide a basis for triaging patients for additional intervention beyond conservative care and reassurance. The aim of this study was to build on our prior work by reducing the number of questionnaire items necessary to produce a feasible and reliable LBP screening tool (ie, the Pain Recovery Inventory of Concerns and Expectations [PRICE] measure) that could guide early interventions for high-risk patients.
For development of the screening measure, data were taken from a prospective cohort of 496 working adults (58% men) who completed a psychosocial test battery before an initial medical evaluation for work-related, acute LBP. The study was conducted in cooperation with a private network of occupational health clinics in the United States, with the eight participating clinics located in Massachusetts, Rhode Island, or Texas. Study eligibility requirements were (1) nonspecific sacral or lumbar back pain, (2) acute onset or exacerbation (fewer than 14 days), (3) pain self-reported as occupational in origin, (4) aged 18 years or older, and (5) fluency in English or Spanish. Of the 496 patients enrolled in this study, 359 (72%) could be reached for the 3-month follow-up assessment. Detailed demographic characteristics of the study sample have been published elsewhere,32 but the sample was generally representative of younger, blue-collar workers (mean age, 37 years) employed by medium to large companies. Comparisons of responders and nonresponders at 3-month follow-up showed that responders were older, more likely women, and of lower educational status (P < 0.05). The significant difference on psychosocial predictor variables was only for responders to report more organizational support (t = 2.96; P = 0.003).
Patients were recruited from the consecutive caseload of patients reporting LBP, and volunteer patients completed a brief demographic questionnaire and a 10-page psychosocial test battery (approximately 10 to 15 minutes) after providing written consent. Patients then received a standard medical history and physical examination in adherence to current occupational medicine guidelines.13 Three months after the initial medical evaluation, participants were contacted to complete a brief follow-up questionnaire assessing pain, functional limitation, and work status. The 3-month time frame was chosen for its significance in distinguishing acute from subacute phases of LBP. All study procedures were approved by the institutional review board of the Liberty Mutual Research Institute for Safety.
Predictor Measures and Established Patient Clusters
Rationale for scale inclusion, existing psychometric data, and initial cluster analysis results for the full scales used in the patient cohort study have been presented in an earlier report.32 The findings from these original analyses were that eight of the psychosocial scales were reliable and valid predictors of disability outcomes and that a four-cluster solution provided the best fit in a K-means cluster analysis. The four patient clusters were given the labels of “minimal risk” (29%), “workplace concerns” (26%), “activity limitation” (27%), and “emotional distress” (19%). The eight scales that were included in the original cluster analysis were the Quebec Back Pain Disability Scale,37 a single-item rating of pain intensity,38 the Center for Epidemiologic Studies–Depression scale,39 the Pain Catastrophizing Scale,40 the Tampa Scale of Kinesiophobia,41 the Survey of Perceived Organizational Support,42 and two new measures developed or adapted by the authors: a nine-item Life Impact of Pain Scale and a three-item rating of recovery expectations.
The two new measures were developed for the study to expand on existing disability screening constructs. The Life Impact of Pain Scale was designed to assess the belief that a single episode of LBP has lifelong health and career implications, a negative viewpoint expressed in qualitative studies43–45 but not explicit in existing constructs such as pain catastrophizing. Respondents record level of agreement for nine statements (eg, “This pain episode will affect my future”) on a four-point scale from “strongly disagree” to “strongly agree.” In this initial cohort, the internal consistency (Cronbach α) was 0.82 and the 1-week test–retest correlation 0.88. Although many previous studies have included a single question pertaining to recovery expectation,46 we expanded this to three questions assessing different aspects of recovery (symptoms, function, and work status),47 with recovery time as the primary reference. Patients were asked to estimate the length of time (0 to 2, 3 to 7, 8 to 14, 15 to 30, 31 to 60, and more than 60 days) they expected for pain resolution, resumption of normal activities, and return to full-duty work. The internal consistency of the three-item scale in this cohort was 0.90, and the 1-week test–retest correlation was 0.65.
In addition to reassessing pain intensity (Visual Analog Scale) and functional limitation (Quebec Back Pain Disability Scale) at 3-month follow-up, participants provided details about current work status, any temporary modifications or physician restrictions, and the cumulative duration of work absences and work modifications. To provide an overall outcome measure that would distinguish the patients with a possible need for further monitoring, treatment, or referral, a composite clinical case rating was also used. Caseness was defined by problems in any of the three outcome domains: work status (unable to resume full duty work), pain rating (5 or greater), or physical dysfunction (greater than 50% items endorsed). Rationale for these cutoff scores can be found in an earlier cohort study.48
A sensitivity analysis was conducted separately for each of the eight psychosocial variables to reduce the overall length while still maintaining a sufficient level of reliability and validity. Although a conventional approach to psychometric testing of a new measure might suggest that the full sample be randomly partitioned into “training” and “validation” samples, our stepwise sensitivity analysis and clustering methods were not conducive to split-sample validation. Also, we hoped to minimize error by including a large sample size when making individual item discriminations. Thus, the full sample was included in all data analyses. With the exception of the single-item pain intensity rating, items were dropped one by one until any of the following criteria were reached: (1) the internal consistency (α) of the individual scale was diminished by more than 10% from the full-scale version; (2) the Pearson correlations of the measure with 3-month pain or disability scores were decreased by more than 10%; (3) the odds ratio of the measure with 3-month return to work was decreased by more than 10% in magnitude; or (4) only two items remained in the scale. Items were chosen for deletion in order of lowest to highest correlation with the total scale score (to maintain the highest level of internal consistency). Scale scores were coded as “missing” if more than two items on the scale were not completed.
After item reduction, the original K-means cluster analysis from the full set of items32 was repeated using scale scores computed only from the reduced items. The K-means cluster analysis is a statistical procedure for identifying relatively homogenous groups of cases based on selected characteristics.49–51 The computational algorithm starts with k random clusters, and then moves subjects iteratively between those clusters with the goal of minimizing variability within clusters while maximizing variability between clusters. The number of clusters, k, can be designated by the user to produce different cluster solutions. The relative locations of cluster centers were used to provide descriptive labels. Our criterion for the correct number of clusters was to determine the maximum number of patient clusters that still showed a noticeable improvement in the percentage of total variance explained (the “elbow” method)52 while also maintaining a Squared Euclidian distance between neighboring cluster centers of at least 2.0 standard deviation units (an “information criterion” method).51 We chose these combined criteria to consider both the explanatory strength and informativeness of cluster results. The analysis was repeated in sequential steps, each time increasing the number of clusters by one.
The reduced questionnaire was reevaluated in terms of total score reliability (internal consistency [α]) and validity (prospective correlation with 3-month pain and disability outcomes). A total severity index (a mean of the eight z scores) was computed to assess the new measure's overall correlation with 3-month outcomes. The area under the curve (AUC) was computed for the dichotomous outcome measures of return to work and clinical case rating.
Descriptive data for the eight psychosocial scales (both full scale and reduced scale) are shown in Table 1. The total pool of items was trimmed from 129 to 46 items without reducing the internal consistency or correlation of individual scales with outcome measures by more than 10%. The scales requiring that more items be preserved were the measures of depressive symptoms (Center for Epidemiologic Studies–Depression) and activity limitation (Quebec Back Pain Disability Scale). In contrast, only two items were retained for assessment of pain catastrophizing (Pain Catastrophizing Scale) and four items for fear of movement (Tampa Kinesiophobia Scale). The final 46 items were distributed as follows: depressive symptoms, 12 items; pain catastrophizing, 2 items; lack of organizational support, 7 items; activity limitation, 15 items; fear of movement, 4 items; perceiving grave life impacts, 3 items; poor expectations for recovery, 2 items; and pain intensity, 1 item.
The eight shortened-scale scores were standardized (z scores), then subjected to a K-means cluster analysis. When five clusters were specified, the Euclidean distance fell below 2.0; thus, the four-cluster solution was chosen, which converged after 18 iterations, with the Euclidian distances between cluster centers varying from 2.03 to 4.90 (Table 2). Analysis of variance results comparing group means for the four-cluster solution (Table 3) showed the relative contribution of each variable to the separation of groups overall. The omnibus test of group differences was statistically significant (P < 0.05) for all eight measures, but this would be expected on the basis of the aim of cluster analysis to maximize group differences. Depressive symptoms had the lowest F value for discriminating groups, and organizational support, the highest.
Interpreting and labeling of clusters were accomplished by noting the largest deviation of cluster means from the grand mean by using a radar graph (Fig. 1). On each of the spokes of the radar graph, the mean standardized scores are plotted by cluster (a score of “0” representing the grand mean). Cluster 1 consisted of 108 patients (22%) who reported few concerns on any of the predictor variables. Cluster 2 consisted of 115 patients (23%) who reported problems in nearly all areas, but the greatest elevations were related to emotional distress (catastrophizing, depression, poor recovery expectations, and life impact of pain). Cluster 3 consisted of 125 patients (25%) who reported lack of organizational support as a more pronounced problem relative to other groups. Cluster 4 consisted of 147 patients (30%) who had elevations with regard to pain intensity and functional limitation but were average in other respects. On the basis of these characteristics, we retained the four cluster labels from the prior analysis of the full-scale measures: “organizational concerns,” “emotional distress,” “activity limitation,” and “minimal risk.”
Individual cases were then “assigned” to the nearest cluster grouping and compared on demographic variables and outcomes of pain, functional limitation, return to work, and the composite clinical case rating at 3-month follow-up (Tables 4 and 5). Demographic comparisons (one-way analysis of variance or chi-square tests) showed no differences between clusters on sex or marital status (P > 0.05). Nevertheless, those in the “activity limitation” cluster were older (mean, 39.2 vs 35 to 37 years in the other groups), and those in the “emotional distress” cluster had lower education, income, and job tenure (P < 0.05).
There was a significant main effect for the four clusters on pain outcomes at 3 months (F[df = 3] = 21.64; P < 0.001). Post hoc comparisons using the Tukey honestly significant difference test indicated that the mean score of the emotional distress group (cluster 2) was significantly different from that of all the other groups (P < 0.001). There was also a significant main effect for the four clusters on functional limitations after 3 months (F[df = 3] = 28.1; P < 0.001), with the post hoc comparisons showing significant differences between all the clusters, except for no difference between the workplace concerns group and the physical limitation group. Controlling for age, education, income, and job tenure did not attenuate these group effects.
In terms of return to work (Table 5), the emotional distress group was more than seven times less likely to have resumed normal work responsibilities after 3 months (P < 0.001). The workplace concerns and activity limitation groups showed some increased risk of not returning to work (odds ratio = 2.09 and 1.82, respectively), but these effects did not reach statistical significance (P > 0.05). The composite clinical case rating, however, showed statistically significant differences (P < 0.05) between the minimal risk group and the other three groups. There was approximately a 2-fold risk of caseness in the workplace concerns and activity limitation groups and a 12-fold increase in the emotional distress group. The AUC for the overall PRICE severity index was 0.75 for predicting 3-month disability (Fig. 2) and 0.73 for predicting 3-month clinical case rating (Fig. 3).
This study provides initial psychometric evaluation of a new patient-screening tool that can be used for subgroup classification of patients with LBP, and this research direction has been ranked as a high priority among back pain specialists and researchers.53 Like several existing psychosocial screening measures,28 the PRICE provides a total score to estimate the overall prognosis for a quick recovery and resumption of normal daily activities. In addition, the PRICE screening questionnaire provides an indication of whether attention should be focused on workplace coordination, physical activation, or psychological coping, and this may improve the ability to provide more patient-centered strategies for early disability prevention. Further research is needed to determine whether the PRICE produces similar findings in other clinical settings and whether clinical application might improve disability outcomes; nevertheless, this study provides further support for the triage of patients with LBP, using a condensed psychosocial screening inventory.
One finding from the sensitivity analysis was that while only two to seven items were retained from most of the original scales, this was not the case for the Center for Epidemiologic Studies–Depression measure (where 12 of 20 items were retained) and the QPS physical function measure (15 of 20 items retained). Further deletion of items from either of these scales would have substantially diminished their associations with 3-month outcomes. Our interpretation is that the complexity of these constructs required more items to reliably quantify their effect on longitudinal outcomes. Although shorter psychosocial screening questionnaires are available to assess overall patient risk,29 a total of 46 items were necessary on the PRICE measure to reliably classify patients by problem area domains. At 46 items, the PRICE measure requires approximately 5 minutes to complete, and this seems feasible for waiting room administration.
Results of the study also provide evidence that a lengthy psychosocial test battery can be reduced to a feasible clinical screening instrument (46 items) without sacrificing reliability or validity of individual constructs necessary to draw meaningful contrasts between patients. Reduction of the total item pool from 129 to 46 items preserved the internal consistency of individual psychosocial constructs, while not confounding the four-group classification of patients found in an earlier cluster analysis using the full-length measures.31 This subgroup classification (derived from the 46-item scale) was further supported by significant group differences in outcome measures at 3 months. Although specific cutoff scores for choosing and assigning patients to early intervention may depend on other factors (cost of intervention, anticipated effect size, etc), the normative data from this initial cohort provide a preliminary basis for making such determinations for applying the PRICE questionnaire in clinical practice or in subsequent research trials. One possible strategy would be to establish a cutoff score on the total severity index, then assign early intervention strategies on the basis of the nearest cluster centroid.
One challenge of applying the PRICE in clinical decision making is that although specific patterns of psychosocial factors can be discerned, there remains a substantial correlation among these variables, and it may be difficult to judge which factors are the root cause to be targeted for early intervention. In particular, emotional distress seems to be an issue of at least moderate importance across all three risk groups. Interpreting a depression screening measure in the presence of acute pain is challenging, and clearly more theoretical work is necessary to understand whether such elevations represent a comorbid mental health condition, an affective component of pain experience, or an indication that pain has exceeded available coping resources. There is some evidence that emotional distress might have a different conceptualization for acute rather than chronic LBP.54,55 Although it might benefit all at-risk patients to receive multidisciplinary interventions addressing workplace, emotional, and physical function issues, this high level of clinical intervention is neither feasible nor cost-effective in most settings. When taking into account limited resources and referral options, measures such as the PRICE might provide an evidence-based strategy for using limited available resources for case management, workplace communication, and patient education and support.
Validity of the PRICE screening was supported by its prospective association with the 3-month disability outcome measures (return to work, functional limitation, and clinical case rating). The AUC for predicting 3-month outcome measures can be rated as “acceptable.”56 This moderate level of predictive strength might be expected for a measure that was limited to modifiable factors within a circumscribed conceptual model and with no inclusion of demographic, medical history, or physical examination variables that also contribute incrementally to disability risk.5 Other self-report measures that have been developed solely for the purpose of maximizing prediction have reported similar or slightly better prediction,28 but these measures provide minimal guidance in selecting appropriate intervention strategies, and many questionnaire items represent nonmodifiable factors (eg, age, sex, education, job tenure).5 Given the complexity of psychosocial and workplace factors that contribute to LBP disability,5 it seems unlikely that future screening questionnaires will add substantially to the predictability of LBP outcomes from psychosocial measures in general patient cohorts. Combining the PRICE results with other patient information (eg, LBP history, physical examination findings, pain patterns, demographics, comorbid health conditions, job type) may further improve its accuracy of prediction.
One limitation of the PRICE measure is its focus on workplace and psychosocial factors to the exclusion of any biomedical data beyond a simple pain intensity rating. Obviously, such a measure should not be used in lieu of a standard medical history, and physical and psychosocial screening should not eliminate the need for careful screening of medical red flags, use of diagnostic triage and treatment guidelines, and biomechanical explanations for pain and prognosis. For most cases of uncomplicated LBP, however, a focus on workplace and psychosocial factors is clearly supported by epidemiologic evidence,5 and contemporary models of work disability have been clear to illustrate the breadth of individual, medical, and societal factors affecting work disability.18 Psychosocial screening of patients with LBP does not imply a psychogenic cause for LBP, but rather that the ability of an individual to adapt, intepret, and solve pain-related problems is influenced by pain beliefs, coping strengths, and social and environmental factors, and these factors influence musculoskeletal rehabilitation and healing.20 Although inclusion of physical examination results may have altered patient clusters, there is little evidentiary support that psychosocial factors in back disability are significantly confounded by symptom patterns or type of injury.10,57,58
Incorporating psychosocial and workplace factors into routine clinical decision making is consistent with providing quality, patient-centered health care, but health care providers need feasible and effective treatment or referral options to respond promptly and appropriately to patient needs. Providing more individualized treatment can be challenging in health care systems that are struggling to control costs and with treatment guidelines discouraging unnecessary diagnostics and referrals. Nevertheless, a number of (low- to moderate-intensity) intervention strategies for LBP have been helpful to address psychosocial and workplace concerns (Table 6), and their effectiveness may be improved if matched to patient subgroups. For patients lacking organizational support, interventions might include participatory ergonomic interventions, facilitated communication with supervisors, or problem solving to address workplace barriers.59,60 For patients with severe emotional distress, interventions might include group or individual sessions applying cognitive–behavioral strategies to address unhelpful pain beliefs, strengthen coping resources, and provide instruction in pain self-management.61–63 For patients with severe pain and activity restriction (but without emotional distress), intervention might focus on pain education, graded exercise, and gradual activity exposure.64–66 Combining patient screening with targeted early intervention may prove to be the best all-around strategy for reducing disability while also containing treatment costs, but more research is needed to support these claims.
Limitations of the study include the narrowly defined study population (mostly younger, blue-collar workers), an a priori theoretical model, and a focus on the acute phase of LBP, when a sizable number of patients will experience spontaneous improvement and many will be able to resume normal work activities with little or no support. Nevertheless, we found evidence of patients describing severe emotional distress, poor expectations for resuming work, pain catastrophizing, and beliefs that LBP had life-altering consequences within just hours or days after pain onset. These extreme views of patients are difficult to ignore, even during this very early stage of medical evaluation and monitoring. In contrast with chronic pain cohorts, we have found little effect of LBP history on psychosocial factors in this occupational medicine setting;67 so psychosocial factors do not seem to be the result of experience. Future studies might evaluate whether such disability risk factors vary with respect to physical job demands, job flexibility, or duration of pain. Although statistical techniques for determining the reliability and validity of cluster results are not well developed, stability of cluster membership over time would be another important psychometric property to evaluate when applying the PRICE in clinical decision making. Test–retest reliability for most of the underlying root measures is well established, but additional studies are needed to support reliability of the 46-item version of the PRICE measure in different clinical settings and disability contexts. Although we were unable to randomly subdivide the sample to assess reproducibility, the relatively narrow 95% confidence bands around the AUC estimates suggest these associations are not because of chance.
This study provides initial support for the reliability and validity of the PRICE, a new self-report questionnaire intended to provide a brief (46-item) screen of psychosocial and workplace factors that might be the target of return-to-work efforts for working adults with LBP. Although most questionnaire items in the PRICE were taken from existing psychosocial measures, the reduced scale is feasible for routine administration and designed to triage patients to the most appropriate forms of early intervention that might prevent long-term back disability. Such efforts to classify patients into meaningful subgroups for clinical decision making may improve the efficacy and cost-effectiveness of LBP treatment in occupational medicine and other clinical settings.
1. Pincus T, Burton AK, Vogel S, Field AP. A systematic review of psychological factors as predictors of chronicity/disability in prospective cohorts of low back pain. Spine. 2002;5:E109–E120.
2. Shaw WS, Linton SJ, Pransky G. Reducing sickness absence from work due to low back pain: how well do intervention strategies match modifiable risk factors? J Occup Rehabil. 2006;16:591–605.
3. Iles RA, Davidson M, Taylor NF. Psychosocial predictors of failure to return to work in non-chronic non-specific low back pain: a systematic review. Occup Environ Med. 2008;65:507–517.
4. Shaw WS, van der Windt DA, Main CJ, et al. Early patient screening and intervention to address individual-level occupational factors (“blue flags”) in back disability. J Occup Rehabil. 2009;19:64–80.
5. Heitz CAM, Hilfiker R, Bachmann LM, et al. Comparison of risk factors predicting return to work between patients with subacute and chronic non-specific low back pain: systematic review. Eur Spine J. 2009;18:1829–1835.
6. 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;24:205–217.
7. Koes BW, van Tulder M, Lin CC, Macedo LG, McAuley J, Maher C. An updated overview of clinical guidelines for the management of non-specific low back pain in primary care. Eur Spine J. 2010;19:2075–2094.
8. Chou R, Shekelle P. Will this patient develop persistent disabling low back pain? JAMA. 2010;303:1295–1302.
9. Linton SJ, Shaw WS. Impact of psychological factors in the experience of pain. Phys Ther. 2011;91:700–711.
10. Nicholas MK, Linton SJ, Watson PJ, et al. Early identification and management of psychological risk factors (“yellow flags”) in patients with low back pain: a reappraisal. Phys Ther. 2011;91:737–753.
11. Melloh M, Elfering A, Salathé CR, et al. Predictors of sickness absence in patients with a new episode of low back pain in primary care. Ind Health. 2012;50:288–298.
12. Chou R, Qaseem A, Snow V, et al. Diagnosis and treatment of low back pain: a joint clinical practice guideline from the American College of Physicians and the American Pain Society. Ann Intern Med. 2007;147:478–491.
13. American College of Occupational and Environmental Medicine. Occupational Medicine Practice Guidelines: Evaluation and Management of Common Health Problems and Functional Recovery in Workers. 2nd ed. Elk Grove Village, IL: American College of Occupational and Environmental Medicine; 2008:653–781.
14. Van Tulder M, Becker A, Bekkering T, et al. Chapter 3. European guidelines for the management of acute nonspecific low back pain in primary care. Eur Spine J. 2006;15(suppl 2):S169–S191.
15. Bergman S. Management of musculoskeletal pain. Best Pract Res Clin Rheumatol. 2007;21:153–166.
16. Nguyen TH, Randolph DC. Nonspecific low back pain and return to work. Am Fam Physician. 2007;76:1497–1502.
17. Moore JE. Chronic low back pain and psychosocial issues. Phys Med Rehabil Clin N Am. 2010;21:801–815.
18. Pransky G, Buchbinder R, Hayden J. Contemporary low back pain research—and implications for practice. Best Pract Clin Rheumatol. 2010;24:291–298.
19. Dupeyron A, Ribinik P, Gélis A, et al. Education in the management of low back pain: literature review and recall of key recommendations for practice. Ann Phys Rehabil Med. 2011;54:319–335.
20. Foster NE, Delitto A. Embedding psychosocial perspectives within clinical management of low back pain: integration of psychosocially informed management principles into physical therapist practice—challenges and opportunities. Phys Ther. 2011;91:790–803.
21. Shaw WS, Main CJ, Johnston V. Addressing occupational factors in the management of low back pain: implications for physical therapist practice. Phys Ther. 2011;91:777–789.
22. Weiner BK. Spine update: the biopsychosocial model and spine care. Spine. 2008;33:219–223.
23. Shaw WS, Pransky G, Winters T, Tveito TH, Larson SM, Roter DL. Does the presence of psychosocial “yellow flags” alter patient–provider communication for work-related, acute low back pain? J Occup Environ Med. 2009;51:1032–1040.
24. Rainville J, Smeets RJ, Bendix T, Tveito TH, Poiraudeau S, Indahl AJ. Fear-avoidance beliefs and pain avoidance in low back pain—translating research into clinical practice. Spine J. 2011;11:895–903.
25. Hayden JA, Dunn KM, van der Windt DA, Shaw WS. What is the prognosis of back pain? Best Pract Res Clin Rheumatol. 2010;24:167–179.
26. Melloh M, Elfering A, Egli-Presland C, et al. Identification of prognostic factors for chronicity in patients with low back pain. Int Orthop. 2009;33:301–313.
27. Gray H, Adefolarin AT, Howe TE. A systematic review of instruments for the assessment of work-related psychosocial factors (blue flags) in individuals with non-specific low back pain. Man Ther. 2011;16:531–543.
28. Hilfiker R, Bachmann LM, Heitz CAM, Lorenz T, Joronen H, Klipstein A. Value of predictive instruments to determine persisting restriction of function in patients with subacute non-specific low back pain. Systematic review. Eur Spine J. 2007;16:1755–1775.
29. Hill JC, Dunn KM, Lewis M, et al. A primary care back pain screening tool: identifying patient subgroups for initial treatment. Arthritis Rheum. 2008;59:632–641.
30. 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;378:1560–1571.
31. Shaw WS, Pransky G, Patterson W, Linton SJ, Winters T. Patient clusters in acute, work-related low back pain based on patterns of disability risk factors. J Occup Environ Med. 2007;49:185–193.
32. Reme SE, Shaw WS, Steenstra IA, Woiszwillo MJ, Pransky G, Linton SJ. Distressed, immobilized, or lacking employer support? A sub-classification of acute work-related low back pain. J Occup Rehabil. 2012;22:541–552.
33. Steenstra IA, Ibrahim SA, Franche RL, Hogg-Johnson S, Shaw WS, Pransky GS. Validation of a risk factor–based intervention strategy model using data from the readiness for return to work cohort study. J Occup Rehabil. 2010;20:394–405.
34. Westman AE, Boersma K, Leppert J, Linton SJ. Fear-avoidance beliefs, catastrophizing, and distress: a longitudinal subgroup analysis on patients with musculoskeletal pain. Clin J Pain. 2011;27:567–577.
35. McCarthy CJ, Roberts C, Gittins M, Oldham JA. A process of subgroup identification in non-specific low back pain using a standard clinical examination and cluster analysis. Physiother Res Int. 2012;17:92–100.
36. Verra ML, Angst F, Staal JB, et al. Differences in pain, function and coping in Multidimensional Pain Inventory subgroups of chronic back pain: a one-group pretest–posttest study. BMC Musculoskelet Disord. 2011;12:145.
37. Kopec JA, Esdaile JM, Abrahamowicz M, et al. The Quebec Back Pain Disability Scale. Measurement properties. Spine. 1995;20:341–352.
38. Childs JD, Piva SR, Fritz JM. Responsiveness of the Numeric Pain Rating Scale in patients with low back pain. Spine. 2005;30:1331–1334.
39. Radloff LS. The CES-D scale. Appl Psychol Meas. 1977;1:385–401.
40. Sullivan MJL, Bishop SR, Pivik J. The Pain Catastrophizing Scale: development and validation. Psychol Assess. 1995;7:524–532.
41. Woby SR, Roach NK, Urmston M, Watson PJ. Psychometric properties of the TSK-11: a shortened version of the Tampa Scale for Kinesiophobia. Pain. 2005;117:137–144.
42. Eisenberger R, Huntington R, Hutchison S, Sowa D. Perceived organizational support. J Appl Psychol. 1986;71:500–507.
43. Coole C, Drummond A, Watson PJ, Radford K. What concerns workers with low back pain? Findings of a qualitative study of patients referred for rehabilitation. J Occup Rehabil. 2010;20:472–480.
44. May S. Patients' attitudes and beliefs about back pain and its management after physiotherapy for low back pain. Physiother Res Int. 2007;12:126–135.
45. Corbett M, Foster NE, Ong BN. Living with low back pain–stories of hope and despair. Soc Sci Med. 2007;65:1584–1594.
46. Hallegraef JM, Krijnen WP, van der Schans CP, de Greef MH. Expectations about recovery from acute non-specific low back pain predict absence from usual work due to chronic low back pain: a systematic review. J Physiother. 2012;58:165–172.
47. Iles RA, Taylor NF, Davidson M, O'Halloran PD. Patient recovery expectations in non-chronic non-specific low back pain: a qualitative investigation. J Rehabil Med. 2012;44:781–787.
48. Shaw WS, Pransky G, Winters T. The Back Disability Risk Questionnaire for work-related, acute back pain: prediction of unresolved problems at 3-month follow-up. J Occup Environ Med. 2009;51:185–194.
49. Steinley D. K-means clustering: a half-century synthesis. Br J Math Stat Psychol. 2006;59(Pt 1):1–34.
50. Cormack RM. A review of classification. J R Stat Soc. 1971;134:46.
51. Steinley D, Brusco MJ. Choosing the number of clusters in Kappa-means clustering. Psychol Methods. 2011;16:285–297.
52. Tibshirani R, Walther G, Hastie T. Estimating the number of clusters in a data set via the gap statistic. J R Stat Soc. 2001;63(Part 2):411–423.
53. da Cunha Menezes Costa L, Koes BW, Pransky G, Borkan J, Maher CG, Smeets RJ. Primary care research priorities in low back pain: an update. Spine. 2013;38:148–156.
54. Vroman K, Warner R, Chamberlain K. Now let me tell you in my own words: narratives of acute and chronic low back pain. Disabil Rehabil. 2009;31:976–987.
55. Gatchel RJ, Bernstein D, Stowell AW, Pransky G. Psychosocial differences between high-risk acute vs. chronic low back pain patients. Pain Pract. 2008;8:91–97.
56. Metz CE. Basic principles of ROC analysis. Semin Nucl Med. 1978;8:283–298.
57. van Abbema R, Lakke SE, Reneman MF, et al. Factors associated with functional capacity test results in patients with non-specific chronic low back pain: a systematic review. J Occup Rehabil. 2011;21:455–473.
58. Hill JC, Fritz JM. Psychosocial influences on low back pain, disability, and response to treatment. Phys Ther. 2011;91:712–721.
59. Lambeek LC, van Mechelen W, Knol DL, Loisel P, Anema JR. Randomised controlled trial of integrated care to reduce disability from chronic low back pain in working and private life. BMJ. 2010;340:c1035.
60. Jensen LD, Maribo T, Schiøttz-Christensen B, et al. Counselling low-back-pain patients in secondary healthcare: a randomized trial addressing experienced workplace barriers and physical activity. Occup Environ Med. 2012;69:21–28.
61. Linton SJ, Andersson T. Can chronic disability be prevented? A randomized trial of a cognitive–behavior intervention and two forms of information for patients with spinal pain. Spine. 2000;25:2825–2831.
62. Lamb SE, Lall R, Hansen Z, Castelnuovo E, et al. A multicentred randomised controlled trial of a primary care-based cognitive behavioural programme for low back pain. The Back Skills Training (BeST) trial. Health Technol Assess. 2010;14:1–253.
63. Hay EM, Mullis R, Lewis M, et al. Comparison of physical treatments versus a brief pain-management programme for back pain in primary care: a randomised clinical trial in physiotherapy practice. Lancet. 2005;365:2024–2030.
64. Karjalainen K, Malmivaara A, Pohjolainen T, et al. Mini-intervention for subacute low back pain: a randomized controlled trial. Spine. 2003;28:533–540.
65. Indahl A, Velund L, Reikeraas O. Good prognosis for low back pain when left untampered. A randomized clinical trial. Spine. 1995;20:473–477.
66. Staal JB, Hlobil H, Twisk JW, Smid T, Koke AJ, van Mechelen W. Graded activity for low back pain in occupational health care: a randomized, controlled trial. Ann Intern Med. 2004;140:77–84.
67. Shaw WS, Pransky G, Patterson W, Winters T. Early disability risk factors for low back pain assessed at outpatient occupational health clinics. Spine. 2005;30:572–580.
Copyright © 2013 by the American College of Occupational and Environmental Medicine