Comparison of a Stratified Group Intervention (Start Back) with Usual Group Care in Patients with Low Back Pain: a Nonrandomized Controlled Trial

RANDOMIZED TRIAL

Comparison of a Stratified Group Intervention (STarT Back) With Usual Group Care in Patients With Low Back Pain: A Nonrandomized Controlled Trial

Susan E. Murphy, PhD, Catherine Blake, PhD,y Camillus K. Power, MD,z and Brona M. Fullen, PhDy

outcome over the controls (P ¼ 0.031). The medium-risk strati- Study Design. A nonrandomized controlled trial. fied intervention demonstrated equally good outcomes Objective. This study aims to explore the effectiveness of (P ¼ 0.125), and low-risk stratified patients, despite less interven- group-based stratified care in primary care. Summary of Background Data. Stratified care based on tion, did as well as the historical controls (P ¼ 0.993). Conclusion. Stratified care delivered in a group setting demon- psychosocial screening (STarT Back) has demonstrated greater strated superior outcomes in the high-risk patients, and equally clinical and cost-effectiveness in patients with low back pain. good outcomes for the medium and low-risk groups. This model, However, low back pain interventions are often delivered in embedded in primary care, provides an early and effective groups and evaluating this system of care in a group setting is model of chronic disease management and adds another dimen- important. Methods. Patients were recruited from 60 general practices and sion to the utility of the STarT Back system of care. Key words: low back pain, physiotherapy-led group linked physiotherapy services. A new group stratified intervention intervention, primary care, STarT back tool, stratification. was compared with a historical nonstratified control group. Level of Evidence: 2 Patients stratified as low, medium and high risk were offered risk- Spine 2016;41:645–652 matched group care. Consenting participants completed self-report measures of functional disability (primary outcome measure), pain, psychological distress, and beliefs. The historical control received a generic group intervention. Analysis was by intention to treat. ow back pain (LBP) is a leading cause of disability Results. In total, 251 patients in the new stratified intervention worldwide with an estimated 632 million people 1 and 332 in the historical control were included in the primary affected. It is one of the principle reasons for visiting L 2 analysis at 12 weeks. The mean age of patients was 43 10.98 general practitioners and constitutes a major drain on 3 years. Overall adjusted mean changes in the RMDQ scores were health system resources. Current best practice advocates higher in the stratified intervention than in the control arm at that LBP patients are managed conservatively within a 4 12-week follow-up (P ¼ 0.028). Exploring the risk groups, indivi- biospychosocial framework. However, despite the prolifer- dually the high-risk stratified group, demonstrated better ation of clinical trials, available treatments tend to produce at best, small to moderate effects.5–7 Emerging evidence demonstrates that stratifying patients into more hom- From the BackCare Programme, Orthopaedic Department, University y ogenous groups and offering targeted treatment leads to Hospital, Waterford, Ireland; School of Public Health, Physiotherapy 8–10 and Population Science, University College, Dublin, Ireland; and zPain better patient outcomes. Service, Adelaide and Meath Hospital, Dublin, Ireland. The STarT Back Trial has demonstrated efficacy with Acknowledgment date: May 26, 2015. First revision date: October 8, 2015. respect to prognostic risk stratification on an individual Acceptance date: October 9, 2015. basis within the controlled environment of a randomized The manuscript submitted does not contain information about medical 8 device(s)/drug(s). controlled trial (RCT). The translation of this model of An unrestricted educational grant from Pfizer Healthcare, Ireland, was care to everyday clinical practice is the next phase in the received in support of this work. implementation of this system of care. LBP interventions No relevant financial activities outside the submitted work. can be delivered individually or in groups. The use of Address correspondence and reprint requests to Susan E. Murphy, PhD, group-based programs is well established and is equally Back Care Programme, Orthopaedic Department, University Hospital, efficacious as individual physical therapy and has the added Waterford, Ireland; E-mail: [email protected] advantage of promoting self-management and greater 4,11,12 DOI: 10.1097/BRS.0000000000001305 cost-effectiveness. Spine www.spinejournal.com 645 Copyright © 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. RANDOMIZED TRIAL Stratified Group Intervention (STarT Back) Murphy et al

The STarT Back system, in particular, allows early identi- and though stratified using the STarT Back Tool, received a fication of the more psychologically distressed patients (high 12-week generic group (‘‘one treatment fits all’’) interven- risk). Approximately one-third of primary care patients tion. Physiotherapists delivering the intervention were have a psychosocial dominance8,13 and these are the patients unaware of which risk group the patient belonged to. who pose the greatest burden to health services.8 Identifying Patients attended four 90-minute group exercise/education early effective systems of care for these patients is import- sessions, with 8 to 10 participants over 4 weeks. Sessions ant.14 Low-intensity structured psychological interventions were conducted within a biopsychosocial framework and are not routinely offered in primary care and only advocated included positive evidence-based messages on managing if initial interventions fail.4 To date, the effectiveness of LBP, general ergonomic advice, a stability exercise program, group-stratified care has not been explored in the primary and physical activity promotion.16 Physical activity levels care setting. and adherence to the exercise program were monitored over the duration of the program. MATERIALS AND METHODS In the stratified arm, low-risk patients attended a single The methods are reported in full in the protocol.15 This study 1.5 hour small group education/exercise session promoting compared a new stratified group intervention with a histori- active management of their LBP and outlining positive messages on maintaining a healthy spine, in line with the cal nonstratified control treatment, within a single clinical 16 center, thus removing the risk of differences in management ‘‘The Back Book.’’ Patients were instructed with simple between centers. A pragmatic study, embedded in the clinical back exercises and encouraged to be active in their lifestyle. setting, it particularly aimed to explore outcomes in the more The medium-risk intervention was modeled on the historical group intervention (see above) as that intervention psychologically distressed high-risk group. Patient assess- 12 ment occurred at baseline and at 12 weeks. had demonstrated efficacy. As per the historical control, Our study tested the hypotheses that patients attended four 90-minute group exercise/education sessions over 4 weeks with the same content and structure as (1) Patients stratified to the new ‘‘high-risk’’ group inter- the historical intervention. vention (STarT Back) will have better physical and High-risk patients attended four 120-minute group ses- psychological outcomes than a ‘‘high-risk’’ historical sions. Each group consisted of approximately 4 to 6 patients control group. and an informal problem-solving approach was utilized. (2) Patients stratified to the ‘‘medium’’ and ‘‘low-risk’’ The content was modeled on the ‘‘high-risk’’ intervention 8 group interventions (STarT Back) will have better delivered in the STarT Back Trial. Sessions utilized cogni- physical and psychological outcomes than the tive behavioral (CBT) strategies to promote self-manage- ‘‘medium’’ and ‘‘low’’ risk historical control groups. ment and address unhelpful beliefs and behaviors around LBP. Each session also included an exercise component Ethical approval was granted from University Hospital similar to the medium-risk intervention. Physical therapists Waterford ethics committee (June 2011). delivering the high-risk intervention received additional training to enhance their CBT skills.17 Participants Patients referred by their GP to a primary care physiother- Outcomes Measures apy led spinal triage clinic in the south east of Ireland were Demographic data and clinical outcomes were gathered at invited to participate (February 2012–June 2013). Patients baseline and 12 weeks. The Roland Morris Disability Ques- were included if they were over 18 years old, English speak- tionnaire (RMDQ: scale 0–24; high scores indicate severe ing, had LBP of greater than 3 months duration, with or disability)18 was the primary outcome measure in keeping without associated leg symptoms. Patients with potentially with prior research.8 serious spinal pathology, serious illness, previous spinal Secondary measures included LBP intensity [visual ana- surgery, or who were pregnant were excluded. Physiothera- log scale (VAS)],19 back beliefs [back beliefs questionnaire pists were at senior or clinical specialist level. (BBQ)],20 distress [distress and risk assessment method (DRAM)],21 and a 6-point self-rated scale (worse, Procedure unchanged, <25% better, 25–50% better, 51–75% better, At initial screening, patients were routinely assessed, out- and >75% better). The DRAM measures depression [modi- come measures collated, and patients stratified according to fied Zung depression index (Zung)] and anxiety [modified their level of risk of persisting symptoms: low, medium, or somatic perception questionnaire (MSPQ)]. Combined high risk (STarT Back Tool). Written consent was obtained. scores categorize the patient as normal (not distressed), Both stratified and nonstratified interventions contained at risk of distress, distressed somatic, and distressed education and exercise components. All patients received depressed.21 a patient manual that included all the relevant information and an exercise log. Statistics The historical nonstratified control consisted of patients Data were screened for normality and missing values who attended the clinic (November 2009–October 2011), imputed using the multiple imputation algorithm in the

646 www.spinejournal.com April 2016 Copyright © 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. RANDOMIZED TRIAL Stratified Group Intervention (STarT Back) Murphy et al

statistical package for the social sciences (SPSS).22 The (71%, n ¼ 413) were female. Patients in both study arms intention-to-treat analysis compared treatment response were classified as having moderate levels of disability in primary and secondary outcomes from baseline to (RMDQ) and pain (VAS) (Table 1). 12 weeks, using analysis of covariance (ANCOVA) models.23 Baseline measures were input as a covariate and Baseline Data statistical significance was expressed in a Wald x2 statistic, In both study arms, a significant difference was found with the critical value set at P <0.05. Our first hypothesis between the 3 risk groups with respect to pain (VAS), tested the superiority of stratified treatment over historical disability (RMDQ), back beliefs (BBQ), and distress care for the high-risk subgroup. We aimed to detect a (DRAM), with patients in the lower risk groups having 1.8-point difference in the RMDQ for the new high-risk better scores (Table 1). intervention over controls, with 80% power and alpha (2-tailed) ¼ 0.05. Equivalence of the stratified treatment Between-group Analysis model to standard care for medium and low-risk groups was determined if the lower boundary of the 95% confidence Disability interval (95% CI) around the mean difference RMDQ did not Overall, the new stratified care produced superior results exceed the MCID 2.0 threshold. The total minimum sample over nonstratified controls in RMDQ. There was evidence required was 368 patients, 72 in each of the high-risk groups to support our study hypothesis that the high-risk stratified and 56 in each of the medium and low-risk groups.15 group did better than the nonstratified control (P ¼ 0.031), with a mean difference of 1.9 points in RMDQ (effect RESULTS size 0.34). No difference was found between the new strati- Overall 881 patients participated in the study: stratified arm fied group and nonstratified controls for the medium n ¼ 365 and nonstratified control data n ¼ 516 (Figure 1 (P ¼ 0.125) and low-risk groups (P ¼ 0.993), in keeping illustrates recruitment and patient flow). Follow-up was with our hypothesis (Table 2). 70% (n ¼ 251) in the stratified intervention and 65% (n ¼ 332) in the nonstratified historical control. The STarT Pain Back Screening Tool classified 41 (16%) as low risk, 142 The new stratified care also produced better overall VAS (57%) as medium risk, and 68(27%) as high risk in the outcomes than the nonstratified controls (P ¼ 0.015). The stratified arm, and 59 (18%), 181 (54%), and 92 (28%), medium-risk group demonstrated better than expected respectively, in the nonstratified historical control. The results with a significant improvement in pain (P ¼ 0.007) mean (SD) age was 43 (10.98) years and the majority (Table 2). High-risk stratified patients also did better with

Figure 1. Study flow chart.

Spine www.spinejournal.com 647 Copyright © 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. RANDOMIZED TRIAL Stratified Group Intervention (STarT Back) Murphy et al

TABLE 1. Patient Baseline Characteristics All Patients Low-Risk Patients Medium-Risk Patients High-Risk Patients Stratified Control Stratified Control Stratified Control Stratified Control n ¼ 41 n ¼ 59 n ¼ 142 n ¼ 181 n ¼ 68 n ¼ 92 Patient (n) (n ¼ 251) (n ¼ 332) (16%) (18%) (57%) (54%) (27%) (28%)

Age, mean 43.4 (11.42) 43.13 (10.67) 43.88 (10.24) 43.15 (10.41) 43.54 (11.78) 43.06 (11.13) 42.82 (11.48) 43.25 (10.00) (SD) Sex, female, 170 (68) 243 (73) 28 (68) 48 (81) 97 (68) 133 (73) 45 (66) 62 (67) n (%) Medicolegal, 34 (12) 31 (9) 2 (4.9) 3 (5) 16 (12) 13 (7) 16 (22) 15 (16) n (%) VAS, mean 5.00 (2.31) 4.8 (2.32) 2.48 (1.69) 3.23 (1.85) 5.07 (1.99) 4.73 (2.24) 6.38 (2.04)y,z 5.99 (2.13)y,z (SD) RMDQ, mean 9.98 (5.00) 10.99 (5.04) 5.07 (3.27) 7.07 (3.78) 9.42 (4.17) 10.97 (4.63) 14.10 (4.19)y,z 13.57 (4.48)y,z (SD) BBQ, mean 24.35 (6.88) 23.47 (6.7) 28.20 (6.32) 27.91 (6.23) 25.32 (6.03) 23.78 (6.67) 19.89 (6.76)y,z 20.01 (5.07)y,z (SD) DRAM Zung, mean 22.23 (10.68) 22.31 (11.7) 16.41 (8.51) 15.44 (10.63) 20.38 (9.23) 21.49 (10.01) 29.69 (10.95)y,z 28.21 (12.6)y,z (SD) MSPQ, mean 7.22 (6.22) 7.96 (6.3) 4.35 (3.66) 5.26 (5.79) 6.58 (5.50) 7.66 (5.47) 10.25 (7.58)y,z 10.30 (7.29)y,z (SD) DRAM category n, (%) Normal 82 (33) 111 (34) 22 (55) 37 (64) 51 (36) 57 (33) 9 (14) 17 (19) At risk 110 (44) 126 (38) 17 (43) 16 (28) 65 (46) 75 (43) 28 (42) 35 (39) Distressed 21 (9) 35 (11) 0 (0) 3 (5) 15 (10) 24 (14) 6 (8) 8 (9) somatic Distressed depressed 36 (14) 50 (16) 1 (2) 2 (3) 11 (8) 18 (10) 24 (36) 30 (33) Startback, n (%) 61 (17) 106 (21) 199 (54) 258 (50) 105 (29) 152 (29) BBQ indicates Back Beliefs Questionnaire; DRAM, Distress and Risk Assessment Method; MSPQ, Modified Somatic Perception Questionnaire; RMDQ, Roland and Morris Disability Questionnaire; ROM, range of motion; S1, sacrum 1; T12, thoracic 12; VAS, visual analog scale; ZUNG, Modified Zung Questionnaire. Low risk/medium risk, P < 0.05. yLow risk/high risk, P < 0.05. zMedium risk/high risk, P < 0.05.

respect to pain (P ¼ 0.110), low-risk patients did worse ‘‘improved >75% category’’ (42% stratified care vs. 20% (P ¼ 0.57) but neither were statistically significant. historical) (Table 3).

Patient’s Beliefs DISCUSSION There was no significant difference in patient’s beliefs (BBQ) This study explored the implementation of group-based between stratified and nonstratified care for the total, low, stratified care in a primary care setting. LBP patients man- medium, or high-risk groups (Table 2). aged over an 18-month period in a stratified group model of care demonstrated better patient outcomes than nonstrati- Distress fied care. Of note, patients in the high-risk stratified inter- Mean change in ZUNG depression scale was greater for the vention did particularly well with respect to disability stratified care overall, although not significant (P ¼ 0.065), (RMDQ), the primary outcome compared with the non- and this was also reflected in the low, medium, and high-risk stratified control. The medium-risk stratified patients dem- groups. For the MSPQ, there were no significant differences onstrated equally good outcomes compared with the between stratified and nonstratified care overall nonstratified controls, and despite less intervention, low- (P ¼ 0.847), or for individual risk groups (Table 2). risk patients did as well as the nonstratified controls. The When patients self-rated benefit was reviewed over findings of this implementation study provide further 6 possible change categories, a significant difference in clinical evidence for the effectiveness of the STarT Back proportions was found overall (x2 ¼ 43.6, df 5, P ¼ system of care and add another dimension to the utility of 0.001). These differences were most apparent in the the tool.

648 www.spinejournal.com April 2016 Copyright © 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. RANDOMIZED TRIAL Stratified Group Intervention (STarT Back) Murphy et al

TABLE 2. Primary and Secondary Outcomes; Intention-to-treat Analysis Mean Change Significance: Pre Mean Post Mean 3 Months Mean ANCOVA Baseline Tool Classification Data set (SD) (SD) (SD) Value as a Covariate RMDQ Total Control (n ¼ 332) 10.99 (5.04) 7.57 (5.75) 3.42 (5.12) Wald x2 4.88, df1 Stratified (n ¼ 365) 10.15 (5.14) 6.31 (4.97) 3.84 (5.11) P ¼ 0.028 Low risk Control (n ¼ 59) 7.07 (3.78) 4.59 (4.09) 2.48 (3.92) Wald x2 0.0001, df1 Stratified (n ¼ 61) 4.93 (3.48) 3.62 (3.46) 1.31 (3.94) P ¼ 0.993 Med. risk Control (n ¼ 181) 10.97(4.63) 7.37 (5.42) 3.60 (5.10) Wald x2 2.38, df1 Stratified (n ¼ 199) 9.75 (4.27) 6.00 (4.68) 3.76 (4.99) P ¼ 0.125 High risk Control (n ¼ 92) 13.56 (4.92) 9.88 (6.34) 3.58 (5.78) Wald x2 4.68, df1 Stratified (n ¼ 105) 13.92 (4.48) 8.44 (5.36) 5.48 (5.34) P ¼ 0.031 VAS Total Control (n ¼ 332) 4.80 (2.33) 3.57 (2.57) 1.23 (2.59) Wald x2 5.98, df1 Stratified (n ¼ 365) 5.08 (2.40) 3.31 (2.38) 1.77 (2.72) P ¼ 0.015 Low risk Control (n ¼ 59) 3.23 (1.86) 2.44 (1.97) 0.79 (2.06) Wald x2 3.65, df1 Stratified (n ¼ 61) 2.40 (1.68) 2.35 (2.03) 0.05 (2.12) P ¼ 0.057 Med risk Control (n ¼ 181) 4.74 (2.24) 3.42 (2.46) 1.32 (2.74) Wald x2 7.42, df1 Stratified (n ¼ 199) 5.16 (2.08) 3.07 (2.28) 2.09 (2.56) P ¼ 0.007 High risk Control (n ¼ 92) 5.99 (2.13) 4.54 (2.78) 1.45 (2.58) Wald x2 2.56, df1 Stratified (n ¼ 105) 6.48 (2.01) 4.31 (2.43) 2.17 (2.95) P ¼ 0.110 Mean Change Significance: Pre Mean Post Mean 3 Months Mean ANCOVA Baseline Tool Classification Data Set (SD) (SD) (SD) Value as a Covariate BBQ Total Control (n ¼ 332) 23.56 (6.72) 26.13 (7.45) 2.56 (6.36) Wald x2 2.53, df1 Stratified (n ¼ 365) 23.90 (6.93) 27.28 (7.49) 3.38 (6.59) P ¼ 0.113 Low risk Control (n ¼ 59) 27.89 (6.19) 29.37 (6.89) 1.07 (6.27) Wald x2 1.56, df1 Stratified (n ¼ 61) 27.98 (6.04) 30.51 (6.67) 2.53 (6.44) P ¼ 0.212 Med risk Control (n ¼ 181) 23.78 (6.67) 26.72 (7.17) 2.93 (6.27) Wald x2 0.78, df1 Stratified (n ¼ 199) 24.90 (6.46) 28.49 (6.86) 3.58 (6.56) P ¼ 0.378 High risk Control (n ¼ 92) 19.92 (5.15) 22.68 (7.22) 2.57 (6.50) Wald x2 0.59, df1 Stratified (n ¼ 105) 19.62 (6.12) 23.11 (7.38) 3.49 (6.75) P ¼ 0.440 ZUNG Overall Control (n ¼ 332) 22.31 (11.70) 19.29 (11.23) 3.02 (10.26) Wald x2 3.46, df1 Stratified (n ¼ 365) 22.92 (10.59) 18.31 (10.61) 4.61 (10.26) P ¼ 0.065 Low risk Control (n ¼ 59) 15.44 (10.63) 13.22 (9.86) 2.22 (8.26) Wald x2 0.62, df1 Stratified (n ¼ 61) 16.52 (7.90) 12.64 (8.52) 3.88 (8.70) P ¼ 0.423 Med risk Control (n ¼ 181) 21.49 (10.01) 18.37 (9.93) 3.12 (10.15) Wald x2 1.65, df1 Stratified (n ¼ 199) 21.22 (9.37) 16.63 (9.52) 4.59 (9.61) P ¼ 0.199 High risk Control (n ¼ 92) 28.21 (12.63) 24.98 (11.99) 3.20 (11.59) Wald x2 1.24 df1 Stratified (n ¼ 105) 29.88 (10.55) 24.78 (10.57) 5.10 (10.39) P ¼ 0.27 Mean Change Significance: Pre Mean Post Mean 3 Months Mean ANCOVA Baseline Tool Classification Data Set (SD) (SD) (SD) Value as a Covariate MSPQ Overall Control (n ¼ 332) 7.96 (6.31) 6.41 (5.99) 1.55 (5.11) Wald x2 0.11, df1 Stratified (n ¼ 365) 7.44 (6.22) 6.06 (5.23) 1.38 (5.06) P ¼ 0.847 Low risk Control (n ¼ 59) 5.36 (5.79) 4.04 (4.52) 1.22 (4.22) Wald x2 0.65, df1 Stratified (n ¼ 61) 4.43 (4.26) 3.81 (3.84) 0.62 (4.17) P ¼ 0.707 Med risk Control (n ¼ 181) 7.66 (5.47) 6.16 (5.66) 1.47 (5.32) Wald x2 0.005 df1 Stratified (n ¼ 199) 7.21 (5.85) 5.71 (4.99) 1.50 (5.06) P ¼ 0.607 High risk Control (n ¼ 92) 10.30 (7.29) 8.42 (6.81) 1.88 (5.24) Wald x2 0.06, df1 Stratified (n ¼ 105) 9.60 (7.06) 8.00 (5.71) 1.60 (5.50) P ¼ 0.906 BBQ indicates Back Beliefs Questionnaire; MSPQ, Modified Somatic Perception Questionnaire; RMDQ, Roland and Morris Disability Questionnaire; SD, standard deviation; VAS, visual analog scale; ZUNG, Modified Zung Questionnaire.

Spine www.spinejournal.com 649 Copyright © 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. RANDOMIZED TRIAL Stratified Group Intervention (STarT Back) Murphy et al

TABLE 3. Secondary Outcome Measures Self-rated Benefit Scale <25% 25–50% 51–75% >75% Better Better Better Better Unchanged Disimproved Total Stratified, n (%) 12 (5) 38 (15) 54 (21) 105 (42) 33 (13) 9 (4) 251 Control, n (%) 17 (5) 83 (25) 83 (25) 67 (20) 79 (24) 3 (1) 332 583 DRAM Worse No Better Improved Stratified, n (%) 31 (12.8) 121 (49.8) 91 (37.4) 243 Control, n (%) 59 (18.8) 161 (51.3) 94 (29.9) 314 DRAM indicates Distress and risk assessment method.

Comparing both arms of the study, the proportion of With respect to psychological status (particularly dis- patients in each risk group at baseline is similar. The tress), although improvements were gained in both study majority of patients have more physical obstacles to recov- arms, the new stratified intervention did not yield signifi- ery, which is consistent with other primary care popu- cantly greater benefits. This is disappointing, as the high- lations.8 The clear differences in baseline domains (pain, risk intervention was specifically designed to improve disability, distress, and back beliefs) between the individual patient’s psychological status. However, this is often the risk groups (low vs. medium vs. high) provide further case with low-intensity psychological interventions and support for the clinical utility of the STarT Back Tool as various explanations have been proposed—treatment not a measure of discriminating between patients of varying being sufficiently tailored to individual patients’ psychologi- clinical severity.13 cal characteristics or the diversity of psychological Maximizing function in patients with CLBP is para- approaches utilized leads to inconsistency in results.28,29 mount, particularly in patients who display high levels of In this study, the outcome measures were dictated by functional and psychological impairment (high-risk). The measures already in place for the historical nonstratified changes achieved in function for the stratified high-risk control. These measures may have been a little outdated and patients provide a substantial endorsement for this new lacked breath to evaluate sufficient changes in psychological model of care. The introduction of a psychologically status. Additional and alternative measures incorporating informed intervention that addresses patient’s fears and fear avoidance, self-efficacy, and catastrophization should anxieties around their LBP delivers a greater impact on a be considered.10 patient’s ability to function. A key objective of practice The new stratified intervention yielded greater patient particularly with chronic conditions is to restore function self-rated benefit, which gives an added dimension to the and enable patients to experience a better quality of life.24 findings. Capturing the patient’s own perspective of prog- High-risk patients are burdensome to the health service and ress and improvement gives further support to the study developing a more targeted treatment intervention that findings.30 The more homogenous groupings of patients addresses their needs in the primary care setting is funda- with more tailored treatment may have accounted for this. mental to improving patient outcomes and creating greater Attrition rates were moderately high in the study but efficiency in health care delivery.14 Comparable changes consistent with other back pain trials.11,31 The pragmatic with respect to pain were not achieved for the stratified high- nature of the study may account for this, wherein lifestyle risk intervention. However, this is consistent with other factors may limit ongoing participation.32 Although a level chronic pain populations, wherein a good outcome is deter- of attrition occurred, overall significant patient numbers mined by a significant improvement in function, irrespective completed the interventions, which supports the validity of of changes in pain.25–27 the study findings.33 In addition, missing data were imputed, The stratified medium-risk intervention demonstrated and when this was done, the main study findings were equally good outcomes compared with the nonstratified upheld, further supporting the validity of the study findings. historical control. A greater change was not anticipated, The group model of stratified care gives physiotherapists as the new stratified medium-risk intervention replicated the an alternative stratified management option for their LBP historical nonstratified control. Low-risk stratified patients patients.8 Group interventions offer a cost-effective and self- also compared well with the nonstratified control, which is directed model of care in line with guidelines.4 The relatively particularly relevant as these patients received far less care low level of contact time involved (high risk n ¼ 8 hours, and did just as well. These findings are consistent with both medium risk n ¼ 6 hours, low risk n ¼ 1.5 hours) and being the STarT Back and IMPaCT Back Trials and have particu- delivered by a single discipline (physiotherapy) makes it an lar cost implications for health care providers.8,10 attractive cost option for health care commissioners. In

650 www.spinejournal.com April 2016 Copyright © 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. RANDOMIZED TRIAL Stratified Group Intervention (STarT Back) Murphy et al

addition, the placement of this intervention in primary care 2. Berenguera A, Pujol-Ribera E, Rodriguez-Blanco T, et al. Study protocol of cost-effectiveness and cost-utility of a biopsychosocial ensures that patients are targeted at an earlier point in the multidisciplinary intervention in the evolution of non-specific sub- disease pathway, thereby minimizing chronicity and reduc- acute low back pain in the working population: cluster randomised ing health care utilization in line with government trial. BMC Musculoskelet Disord 2011;12:194. policy.14,24 The results of our study compare well with 3. Dagenais S, Roffey D, Wai E, et al. Can cost utility evaluations 8 inform decision making about interventions for low back pain? the STarT Back Trial, and though the model of care and Spine J 2009;9:944–57. study designs differed, the current study provides further 4. National Institute for Health and Clinical Excellence (NICE). The support for the clinical effectiveness of the STarT Back Acute Management of Chronic (longer than 6 weeks) Non-specific system of care and adds another dimension to its use. Low Back Pain. 2009. Available at: www.nice.org.uk/nicemedia/ pdf/LowBackPainFullGuidelineConsultation.pdf. Accessed June 15, 2012. Study Limitations 5. Hayden JA, van Tulder MW, Malmivaara AV, et al. Meta- This implementation study, the most feasible option in the analysis: exercise therapy for non specific low back pain. Ann available resources, was designed to reflect routine clinical Intern Med 2005;142:765–75. 6. Chou R, Huffman LH. Nonpharmacologic therapies for acute and practice. Some may argue that the use of a historical control chronic low back pain: a review of the evidence for an American leads to differences between the control and intervention Pain Society/American College of Physicians clinical practice groups. However, every effort was made to ensure that both guideline. Ann Intern Med 2007;147:492–504. study arms were as homogenous as possible.15 There was a 7. Balthazard P, de Goumoens P, Rivier G, et al. Manual therapy followed by specific active exercises versus a placebo followed by degree of attrition from the study that may adversely impact specific active exercises on the improvement of functional disabil- on the validity of the study findings. The outcome measures ity in patients with chronic non-specific low back pain: a random- were constrained by the historical measures already in place ized controlled trial. BMC Musculoskelet Disord 2012;28:162. 8. Hill JC, Whitehurst DG, Lewis M, et al. Comparison of stratified thereby limiting the scope of outcome. In addition, more primary care management for low back pain with current best longterm follow-up is required to assess the sustainability of practice (STarT Back): a randomised controlled trial. Lancet this model of care. 2011;378:1560–71. 9. Fersum K, Dankaerts W, O’Sullivan P, et al. Integration of sub- classification strategies in randomised controlled trials evaluating CONCLUSION manual therapy treatment and exercise therapy for a non specific This group model of stratified care (STarT Back), embedded chronic low back pain: a systematic review. Br J Sports Med in the primary care setting, provides an early and effective 2010;44:1054–62. model of chronic disease management for LBP patients and 10. Foster N, Mullis R, Hill J, et al. Effect of stratified care for low back pain in family practice (IMPaCT Back): a prospective popu- is recommended to maximize patient outcomes. lation-based sequential comparison. Ann Fam Med 2014;12: 102–11. 11. Lamb SE, Hansen Z, Lall R, et al. Group cognitive behavioural Key Points treatment for low-back pain in primary care: a randomised controlled trial and cost-effectiveness analysis. Lancet 2012;375: Stratified care (STarT Back ) delivered in a group 916–23. setting demonstrated superior outcomes over a 12. Murphy S, Blake C, Power CK, et al. Outcomes of a group education /exercise intervention in a population of patients with nonstratified group control. non-specific low back pain: a 3-year review. Ir J Med Sci High-risk patients did particularly well with 2014;183:341–50. respect to disability, thereby providing an 13. Hill JC, Dunn KM, Lewis M, et al. A primary care back pain effective early system of care for the more screening tool: identifying patient subgroups for initial treatment. Arthritis Rheum 2008;59:632–41. psychologically distressed patients and mini- 14. National Health Service (NHS) Spinal Task Force Commissioning mizing potential chronicity and health care Spinal Services. Getting the Service Back on Track. A Guide for utilization. Commissioners of Spinal Services. Available at: http://www.na- Medium-risk patients did equally well as the tionalspinaltaskforce.co.uk/pdfs/NHSSpinalReport_- nonstratified control and low-risk patients, vis7%2030.01.13.pdf. Accessed February 1, 2014. 15. Murphy SE, Blake C, Power CK, et al. The effectiveness of a despite less intervention did as well as the stratified group intervention using the STarTBack screening tool in nonstratified control. patients with LBP-a non randomised controlled trial: study protocol. BMC Musculoskelet Disord 2013;14:342. 16. Royal College of General Practitioners and NHS Executive. The Back Book. London: 2002; The Stationery Office. Acknowledgments 17. Main CJ, Sowden G, Hill JC, et al. Integrating physical and psychological approaches to treatment in low back pain: the The authors wish to acknowledge Pfizer Healthcare Ireland development and content of the STarT Back trial’s ‘high-risk’ and the Irish Pain Society for unrestricted educational grants. intervention (StarT Back; ISRCTN 37113406). Physiotherapy 2012;98:110–6. References 18. Roland M, Morris R. A study of the natural history of back pain. 1. Lim SS, Vos T, Flaxman AD. A comparative risk assessment Part 1: Development of a reliable and sensitive measure of dis- of burden of disease and injury attributable to 67 risk factors ability in low back pain. Spine 1983;8:141–4. and risk factors clusters in 21 regions, 1990–2010: a systematic 19. Hagg O, Fritzell P, Nordwall A. The clinical importance of changes analysis for the global burden of disease study. Lancet 2012; in outcome scores after treatment for chronic low back pain. Eur 380:2224–60. Spine J 2003;12:12–20.

Spine www.spinejournal.com 651 Copyright © 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. RANDOMIZED TRIAL Stratified Group Intervention (STarT Back) Murphy et al

20. Symonds TL, Burton AK, Tillotson KM, et al. Do attitudes and 27. Fullen BM, Blake C, Horan S, et al. Ulysses: the effectiveness beliefs influence work loss due to low back trouble? Occup Med of a multidisciplinary cognitive behavioural pain manage- 1996;46:25–32. ment programme-an 8-year review. Ir J Med Sci 2014;183: 21. Main CJ, Wood PL, Hollis S, et al. The Distress and Risk Assess- 265–75. ment Method. A simple patient classification to identify distress 28. Pincus T, McCracken L. Psychological factors and treatment and evaluate the risk of poor outcome. Spine 1992;17:42–52. opportunities in low back pain. Best Pract Res Clin Rheumatol 22. SPSS Inc. IBM Statistical Package for the Social Sciences (SPSS) 2013;7:625–35. Version 20. Chicago, IL: SPSS Inc; 2012. 29. Kerns RD, Sellinger J, Goodin BR. Psychological treatment of 23. Brom GF, Fransen J, Lennons WA. A simple sample size formula chronic pain. Annu Rev Clin Psychol 2011;411–34. for analysis of covariance in randomized control trials. J Clin 30. Morley S. Efficacy and effectiveness of cognitive behaviour therapy Epidemiol 2007;60:1234–8. for chronic pain: progress and some challenges. Pain 2011; 24. Department of Health and children. ‘‘Tackling Chronic Disease. A 152:S99–106. Policy Framework for the Management of Chronic Diseases.’’ 31. Cairns MC, Foster NE, Wright C. Randomized controlled Available at: http://www.dohc.ie/publications/tackling_chronic_ trial of specific spinal stabilization exercises and conventional diease.html. Accessed August 8, 2014. physiotherapy for recurrent low back pain. Spine 2006;31: 25. McCracken L, Eccleston C. A prospective study of acceptance of 676–81. pain and patient functioning with chronic pain. Pain 32. MacPherson H. Pragmatic clinical trials. Complement Ther Med 2005;118:164–9. 2004;12:136–40. 26. Loeser J. The multidisciplinary pain center: treating co-morbid- 33. Fersum K, O’Sullivan P, Skouen J, et al. ‘‘Efficacy of classification- ities. Giamberardino M, Jenson T, editors. Chapter 19. Pain based cognitive functional therapy in patients with non-specific Comorbidities. Understanding and Treating the Complex Patient. chronic low back pain: a randomized controlled trial’’. Eur J Pain International Association for the Study of Pain, IASP press; 2012. 2013;17:916–28.