The Spine Journal 15 (2015) 1744–1755

Clinical Study

Clinical outcomes of early and later physical therapist services for older adults with back pain Sean D. Rundell, DPT, PhDa,b,*, Alfred C. Gellhorn, MDc, Bryan A. Comstock, MSd, Patrick J. Heagerty, PhDd,e, Janna L. Friedly, MDa,b, Jeffrey G. Jarvik, MD, MPHb,f a Department of Rehabilitation Medicine, University of Washington, 325 Ninth Avenue, Seattle, WA 98104, USA Comparative Effectiveness, Cost, and Outcomes Research Center, University of Washington, 4333 Brooklyn Ave NE, Seattle, WA 98105, USA c Department of Rehabilitation and Regenerative Medicine, Weill Cornell Medical Center, 525 E 68th St, New York, NY 10021, USA d Center for Biomedical Statistics, University of Washington, 4333 Brooklyn Ave NE, Seattle, WA 98105, USA e Department of Biostatistics, University of Washington, 4333 Brooklyn Ave NE, Seattle, WA 98105, USA f Department of Radiology, Neurological Surgery, Health Services, Orthopedics and Sports Medicine and Pharmacy, University of Washington, 325 9th Ave, Seattle, WA 98104, USA b

Received 2 April 2014; revised 5 February 2015; accepted 1 April 2015

Abstract

BACKGROUND CONTEXT: The timing of physical therapy (PT) services and its association with later function and pain are not clear, especially in older adults. PURPOSE: The purpose of this study was to compare clinical outcomes of patients receiving early or later PT services with those not receiving PT among older adults presenting to primary care for a new visit for back pain. STUDY DESIGN/SETTING: Prospective cohort study using the Back Pain Outcomes Using Longitudinal Data registry. PATIENT SAMPLE: A total of 3,705 adults 65 years and older with a new visit for back pain were included. OUTCOME MEASURES: The outcome measures were Roland-Morris Disability Questionnaire (RMDQ), Pain Numerical Rating Scales, and EuroQol-5D. METHODS: We studied two phases of PT utilization: early (0–28 days) and later (3–6 months). At baseline, we selected the participants with complete 12 months of patient-reported outcomes and electronic medical record data. Early PT was defined as initiating PT less than or equal to 28 days from the index visit for back pain. The no early PT group consisted of patients with no PT, no injections, no surgery, and no chiropractic within 28 days. We restricted the later phase analysis to patients with pain greater than 2 of 10 and an RMDQ score greater than 4 to create a subsample of patients with continuing clinically important back pain. We defined later PT as initiating PT between 3 and 6 months after the index visit. The no later PT group consisted of patients without any PT during this time. We used propensity score matching followed by multiple linear regression to estimate the mean difference in outcome. Sensitivity analysis examined clinically important change and dose of PT use among the early PT group.

FDA device/drug status: Not applicable. Author disclosures: SDR: Grant: AHRQ R01 HS019222-01 (I, Paid directly to institution); Fellowship Support: PhRMA Comparative Effectiveness Research Dissertation Award through the Pharmaceutical Outcomes Research and Policy Program at the University of Washington (B). ACG: Grant: AHRQ R01 HS019222-01 (I, Paid directly to institution). BAC: Grant: AHRQ R01 HS019222-01 (I, Paid directly to institution). PJH: Grant: AHRQ R01 HS019222-01 (I, Paid directly to institution). JLF: Grant: AHRQ R01 HS019222-01 (I, Paid directly to institution). JGJ: Grant: AHRQ R01 HS019222-01 (I, Paid directly to institution); Royalties: (B); Stock Ownership: Cofounder and stockholder of PhysioSonics (2% ownership); Consulting: HealthHelp (C); Scientific Advisory http://dx.doi.org/10.1016/j.spinee.2015.04.001 1529-9430/Ó 2015 Elsevier Inc. All rights reserved.

Board/Other Office: Comparative Effectiveness Advisory Board for GE Healthcare through October 2012 (B). The disclosure key can be found on the Table of Contents and at www. TheSpineJournalOnline.com. Back Pain Outcomes Using Longitudinal Data is funded by Agency for Healthcare Research and Quality grant R01 HS019222-01. The authors report no potential conflicts of interest related to this study. * Corresponding author. Comparative Effectiveness, Cost, and Outcomes Research Center, University of Washington, Box 359455, 4333 Brooklyn Ave NE, Seattle, WA 98105, USA. Tel.: (206) 221-2316. E-mail address: [email protected] (S.D. Rundell)

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

1745

RESULTS: The early PT group had better functional status with an adjusted mean RMDQ of 1.1 points less than the no early PT group (95% confidence interval: –2.2, –0.1) and less back pain of –0.5 (–0.9, –0.1) at 12 months. There was no difference between early PT groups at 3 and 6 months. The odds of a 30% improvement in function or pain were not different between these matched groups at 12 months, but the early PT group had increased odds of a 50% improvement in function at 12 months (odds ratio: 1.58, 95% confidence interval: 1.04, 2.40). There was no difference between later groups at 12 months. Greater dose of PT use within the early PT group was associated with better functional status (p5 .01). CONCLUSIONS: We found that among older adults presenting to their primary care providers for a new episode of back pain, early referral to PT resulted in no or minimal differences in pain, function, or health-related quality at 3, 6, or 12 months compared with a matched group that did not receive early PT. Secondary analysis show that patients initiating early PT may be somewhat more likely to experience 50% improvement in function at 12 months. Ó 2015 Elsevier Inc. All rights reserved. Keywords:

Back pain; Older adults; Physical therapy; Comparative effectiveness research; Function; Outcomes

Introduction Back pain is one of the most common reasons for primary care visits in the United States, and it has a considerable burden worldwide [1,2]. Adults with acute back pain typically have a good prognosis [3,4], but after 3 months of symptoms, the prognosis is considerably less favorable as the improvement in pain and disability is small for the patients with ongoing pain and disability [3,5,6]. Little is known about the prognosis for new episodes of back pain for older adults, but initial evidence suggests it is less favorable [7]. A greater proportion of older adults continue to have back pain and limited function at 3 months compared with the studies of younger adults [3,4,7]. Because of the good prognosis for most episodes of acute back pain, clinical guidelines for the diagnosis and treatment of back pain recommend minimal initial interventions: reassurance of a favorable outcome, encouragement to remain active, and strategies for selfcare. Nonsteroidal anti-inflammatories can be used as needed for symptom relief [8]. These guidelines are not specific to older adults, who may have additional comorbidities, greater degenerative conditions, and a different prognosis than younger adults [7,9,10]. Because of these differences, it is possible that an alternative approach to managing new episodes of back pain in older adults may be more appropriate than what is suggested by the current guidelines. The timing of physical therapy (PT) and its association with later function and pain are not clear, especially in older adults. Using Medicare data, Gellhorn et al. [11] found that early PT was associated with less ensuing back-related health-care utilization in patients with back pain compared with those receiving later PT. However, no outcomes data on pain and back-related function are available in Medicare claims databases. Health systems have implemented the use of early PT in new pathways

for managing back pain and reported positive outcomes [12], but more rigorous comparative effectiveness research on the clinical outcomes of this strategy in older adults is currently not available. The purpose of this study was to compare clinical outcomes between patients receiving early or later PT services with those not receiving early or late PT in a cohort of older adults presenting to primary care settings with a new visit for back pain. Secondary objectives are to describe older adults with back pain who receive early or later PT compared with those not receiving PT services during these times.

Methods Design, data source, and subjects We conducted a cohort study using participants from the Back Pain Outcomes Using Longitudinal Data (BOLD) registry. Back Pain Outcomes Using Longitudinal Data is a prospective cohort of adults, 65 years and older, presenting to primary care settings for a new visit of back pain within three integrated health systems: Kaiser Permanente in Northern California, Henry Ford Health System in Detroit, MI, and Harvard Vanguard Medical Associates/ Harvard Pilgrim Health Care in Boston, MA. We collected patient-reported outcomes from participating patients and ascertained health-care utilization from the electronic medical record (EMR) for 12 months after the index visit for back pain. Key clinical exclusion criteria include a health-care encounter for back pain 6 months before the index date, previous lumbar spine surgery, developmental spine deformities, inflammatory spondyloarthropathy, known spinal malignancy or infection, and serious medical comorbid conditions with life expectancy less than 1 year.

1746

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

Context A number of studies have shown that early and effective interventions for a number of spinal conditions may limit long-term disability. The authors sought to investigate the impact of early referral to physical therapy (PT) on outcome in elderly patients with low back pain. Contribution Propensity matching was performed on more than 480 patients who were treated with early PT and a group that received late referral to PT (n599). The authors maintain that early referral to PT resulted in no clinically meaningful differences in outcomes as measured at 3, 6, or 12 months following treatment. The extent of improvement in symptoms, however, may be greater among those referred for early therapy. Implications This was a subset analysis of patients enrolled in the BOLD registry, the primary design of which was not intended to answer this study’s question. Patients included in this study represent a very small percentage of the total number of patients participating in the registry. The authors acknowledge important differences between the two cohorts and propensity matching can only account for those factors considered when executing the matching algorithm. Differences between the patients considered in this study and those encountered in daily clinical practice may also impair the capacity for the results of this work to be translated to a substantial extent. —The Editors

We screened, consented, and enrolled patients at all BOLD study sites after an eligible index visit for back pain. Baseline interviews took place 0 to 21 days after the index visit. Some participants received treatment before they completed the baseline interview, and this resulted in a proportion of the sample with baseline measures occurring after the initiation of PT. Because treatment received before the baseline interview could have affected the baseline questionnaire responses, we excluded participants who initiated PT more than 1 day before the baseline interview from our hypothesis testing analysis. The assembly of this cohort is further described by Jarvik et al [13]. The decision-making and clinical guidelines for newonset back pain are different from decision making for persistent back pain. Consequently, we examined referral to PT during two phases: an early phase from 0 to 28 days and a later phase from 90 to 180 days. For the early phase analysis, we selected BOLD participants who had completed 12 months of patient-reported outcomes data and

had complete EMR data as of November 2013. We restricted the later phase analysis to patients not receiving early PT who had continuing clinically important back pain: a back pain Numerical Rating Scale (NRS) greater than 2 of 10 and a Roland-Morris Disability Questionnaire (RMDQ) greater than 4 at 3 months. This restriction ensured that a clinically meaningful change could occur, and those with resolved back pain were not included in the later phase analysis [14,15] (Figure). Treatment groups We defined the initiation of PT as the day a Current Procedural Terminology (CPT) code for PT evaluation (CPT code: 97001) was documented in the EMR. At one study site (Detroit), specific PT CPT codes were not available in the EMR. However, an indicator of when a PT visit occurred was available. For this site, the PT evaluation was defined as the first PT visit after the index visit. Early PT was defined as having a PT evaluation and no injections, no surgery, and no chiropractic less than or equal to 28 days from the index visit. We excluded 121 participants receiving PT less than or equal to 28 days from the index visit who did not have a PT evaluation code to avoid including patients with ongoing PT for non–back-related care. At the Detroit site, associated International Classification of Diseases, Ninth Revision, codes were available for the PT indicator variable, and we excluded two people receiving ongoing PT for non–back-related care during the early phase. The comparison group for the early phase analysis comprised patients with no PT, no injections, no surgery, and no chiropractic during the first 28 days after the index visit. This comparison group corresponds to guideline concordant care [8]. After restricting the later phase to patients not receiving early PT who had continuing clinically important back pain, we defined later PT as initiating PT from 90 to 180 days of the index visit. The comparison group for the later phase analysis comprised patients with no PT visits during the same time period. Outcomes We collected all patient-reported outcome measures from patients at baseline and 3, 6, and 12 months. The primary outcome measure for BOLD and this study is the RMDQ. The RMDQ is valid, reliable, commonly used, and recommended for back pain research [16,17]. The BOLD registry used a modified version that added ‘‘leg pain/sciatica’’ to each question [13]. The RMDQ is scored from 0 to 24, with 0 signifying no limitation in function and 24 being maximal functional limitation. The minimal clinically important difference has been described as 2 to 3 points or at least a 30% improvement [14,18,19]. Secondary outcome measures are the pain NRS and the EuroQol-5D (EQ-5D). We measured back and leg pain intensities separately using a 0 to 10 NRS for average pain in

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

1747

Figure. Flow diagram of study design

the past 7 days. Self-reported pain intensity using the NRS has good content validity in older adults [20]. The EQ-5D is a preference-based, generic health-related quality-of-life measure. The EQ-5D index has five domains (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression). Each domain is described using three levels of health, 1 (no problems) to 3 (extreme problems), and it is recommended for back pain studies [21] (www.euroquol. org).

back pain as less than 1, 1 to 3, 3 to 6, and 6 to 12 months, 1 to 5 years, or greater than 5 years; the participant’s expectation of recovery using a 0 (no confidence) to 10 (complete confidence) scale [25]; and the Brief Pain Inventory (BPI) that has seven domains (general activity, mood, walking, work, interpersonal relationships, sleep, and enjoyment of life), each scored on a 0 (no interference) to 10 (interferes completely) scale and the composite BPI score that is the average of all domain scores [26].

Covariates

Analysis

We ascertained demographic variables of age, gender, ethnicity (Hispanic/non-Hispanic), race, marital status, and education at the baseline interview. We recorded the back pain International Classification of Diseases, Ninth Revision, code for the index visit from the EMR and categorized them as axial pain, back and leg pain, spinal stenosis, or other. Health status variables measured at baseline included smoking history and a screen for anxiety and depression symptoms using the Patient Health Questionnaire4 (PHQ-4). The PHQ-4 contains two questions about anxiety symptoms and two about depressive symptoms. The cutoff score for a positive screening test for either anxiety or depression is greater than or equal to 3 [22–24]. We collected additional back pain characteristics at baseline: the duration of participants’ current episode of

We used descriptive statistics to characterize the cohort and each treatment group at baseline. We calculated the mean and standard deviation of unadjusted outcome measures at 3, 6, and 12 months for each treatment group. We used descriptive statistics of PT CPT codes for the early PT and no early PT groups to characterize the use and content of PT services for the two sites with these data available (Northern California and Boston). The content of PT was categorized as active (therapeutic exercise, neuromuscular reeducation, gait training therapy, group therapeutic procedures, therapeutic activity, and self-care management training), manual therapy, or passive (all modalities) treatments. We performed two parallel analyses, one for the early phase and one for the later phase. We used a two-step

1748

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

Table 1 Baseline characteristics of patients presenting to primary care for a new back pain visit by timing of PT use Early phase: 0–28 d Variables

Early PT (n5484)

No later PT (n51,903)

Later PT (n599)

All (n53,705)

(6.9) (65.1) (6.0)

73.5 (6.4) 300 (62.0) 34 (7.0)

74.5 (7.1) 1,331 (69.9) 129 (6.8)

73.3 (6.3) 65 (65.7) 4 (4.1)

73.9 (6.9) 2,398 (64.7) 227 (6.2)

(15.0) (75.0) (10.0)

47 (9.8) 363 (75.8) 69 (114.4)

338 (18.0) 1,351 (71.8) 192 (10.2)

15 (15.5) 72 (74.2) 10 (10.3)

524 (14.3) 2,751 (75.1) 388 (10.6)

(31.7) (28.3) (40.0)

123 (25.4) 148 (30.6) 213 (44.0)

715 (37.8) 552 (29.1) 627 (33.1)

28 (28.6) 32 (32.7) 38 (38.8)

1,140 (30.9) 1,057 (28.6) 1,496 (40.5)

(61.2) (16.5)

307 (63.6) 74 (15.3)

1,096 (57.8) 310 (16.4)

67 (68.4) 14 (14.3)

2,273 (61.5) 603 (16.3)

(22.3)

102 (21.1)

490 (25.8)

17 (17.4)

818 (22.1)

(31.5) (18.3) (6.8) (5.9) (15.5) (22.1) (2.8) (2.5) (61.5) (6.5) (2.5)

175 144 42 30 53 40 5.1 5.6 346 9.9 3.6

(3.7)

6.5 (3.2)

4.6 (3.6)

5.1 (3.7)

5.5 (3.7)

(0.17)

0.73 (0.17)

0.71 (0.17)

0.70 (0.18)

0.75 (0.17)

(53.8) (40.3) (5.9) (7.9) (12.1)

268 195 20 44 75

(55.5) (40.4) (4.1) (9.1) (15.5)

1,015 753 128 191 272

(53.5) (39.7) (6.8) (10.0) (14.3)

57 39 3 15 20

(57.6) (39.4) (3.0) (15.2) (20.2)

1,997 1,490 208 298 463

(54.1) (40.3) (5.6) (8.0) (12.5)

(68.3) (19.7) (5.9) (6.1)

266 173 12 33

(55.0) (35.7) (2.5) (6.8)

1,284 369 130 120

(67.5) (19.4) (6.8) (6.3)

59 30 4 6

(59.6) (30.3) (4.0) (6.1)

2,466 807 202 230

(66.6) (21.8) (5.5) (6.2)

(16.5) (59.4) (24.1)

38 (7.9) 405 (83.7) 41 (8.5)

No early PT (n53,221)

Demographics Age (y), mean (SD) 73.9 Sex (F), n (%) 2,098 Ethnicity (Hispanic), n (%) 193 Race, n (%) Black 477 White 2,388 Other 319 Education, n (%) High school graduate or less 1,017 Some college or trade school 909 At least college graduate 1,283 Marital status, n (%) Married or partner 1,966 Separated, divorced, or never 529 married Widowed 716 Back pain history and current episode Symptom duration, n (%) !1 mo 1,014 1–3 mo 587 3–6 mo 218 6–12 mo 190 1–5 y 497 O5 y 710 Back pain intensity, mean NRS (SD) 4.9 Leg pain intensity, mean NRS (SD)y 5.3 Leg pain present, n (%) 1,979 Functional status, mean RMDQ (SD) 9.5 Pain interference with activity, mean 3.3 BPI (SD) Expectation for recovery, mean 5.3 NRS (SD) General health status EQ-5D, mean (SD) 0.76 Other Smoking, n (%) Never 1,729 Quit O1 y ago 1,295 Smoker or quit !1 y ago 188 Positive depression screen, n (%) 254 Positive anxiety screen, n (%) 388 Index Diagnostic Code Category, n (%) Axial pain 2,200 Back and leg pain 634 Lumbar spinal stenosis 190 Other 197 Study site, n (%) Detroit 532 Northern California 1,913 Boston 776

Later phase: 3–6 mo (n52,002)*

(36.2) (29.8) (8.7) (6.2) (11.0) (8.3) (2.8) (2.6) (71.5) (6.1) (2.5)

467 310 129 133 362 500 5.7 5.3 1,304 11.5 3.9

(24.6) (16.3) (6.8) (7.0) (19.0) (26.3) (2.5) (2.5) (68.6) (6.2) (2.4)

372 (19.6) 1,101 (57.9) 430 (22.6)

22 25 18 3 18 13 5.7 5.5 82 12.0 4.3

(22.2) (25.3) (18.2) (3.0) (18.2) (13.1) (2.6) (2.6) (82.8) (5.9) (2.5)

13 (13.1) 81 (81.8) 5 (5.1)

1,189 731 260 220 550 750 5.0 5.3 2,325 9.5 3.3

(32.1) (19.8) (7.0) (6.0) (14.9) (20.3) (2.8) (2.5) (62.8) (6.5) (2.5)

570 (15.4) 2,318 (62.6) 317 (22.1)

BPI, Brief Pain Inventory; EQ-5D, EuroQol-5D; F, female; NRS, Numerical Rating Scale; PT, physical therapy; RMDQ, Roland-Morris Disability Questionnaire; SD, standard deviation. * Participants not receiving early PT and with RMDQ O4 and back pain NRS O2 at 3 months. y Patients with leg pain present only.

process to minimize differences between treatment groups. First, we used a propensity score–matching approach to create balanced treatment groups for each phase [27–29]. We used logistic regression to estimate the predicted

probability of initiating PT during the early phase or during the later phase. We included predictors hypothesized a priori to be associated with PT utilization in each model. These were age, gender, race, education, marital status,

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

1749

Table 2 Characteristics of early phase propensity score–matched groups Chi-square or t test Variables Demographics Age (y), mean (SD) Sex (F), n (%) Ethnicity (Hispanic), n (%) Race, n (%) Black White Other Education, n (%) High school graduate or less Some college or trade school At least college graduate Marital status, n (%) Married or partner Separated, divorced, or never married Widowed Back pain history and current episode Symptom duration (mo), n (%) !3 3–12 $12 Back pain intensity, mean NRS (SD) Leg pain intensity, mean NRS (SD)* Leg pain present, n (%) Functional status, mean RMDQ (SD) Pain interference with activity, mean BPI (SD) Expectation for recovery, mean NRS (SD) Health status EQ-5D, mean (SD) Positive depression screen, n (%) Positive anxiety screen, n (%) Smoker, n (%) Study site, n (%) Detroit Northern California Boston

Control (n5515)

Early PT (n5166)

p Value

Unmatched early PT (n5318)

73.6 (6.7) 335 (65.1) 33 (6.4)

73.2 (6.0) 113 (68.1) 11 (6.6)

.50 .48 .92 .91

73.6 (6.6) 187 (58.8) 23 (7.2)

49 (9.5) 397 (77.1) 69 (13.4)

14 (8.4) 130 (78.3) 22 (13.3)

124 (24.1) 165 (32.0) 226 (43.9)

44 (26.5) 41 (24.7) 81 (48.8)

338 (65.6) 98 (19.0) 79 (15.3)

109 (65.7) 25 (15.1) 32 (19.3)

332 56 127 5.3 5.5 361 10.0 3.7 6.3

(64.5) (10.9) (24.7) (2.9) (2.6) (70.1) (6.5) (2.5) (3.4)

108 26 32 5.4 5.6 118 10.3 3.8 6.4

(65.1) (15.7) (19.3) (2.5) (2.6) (71.1) (6.1) (2.4) (3.3)

0.74 48 59 27

(0.18) (9.3) (11.5) (5.2)

0.73 15 20 7

(0.17) (9.0) (12.1) (4.2)

33 (10.4) 233 (73.3) 47 (14.8) .20 79 (24.8) 107 (33.7) 132 (41.5) .32 198 (62.3) 49 (15.4) 70 (22.0) .14

36 (7.0) 389 (75.5) 90 (17.5)

11 (6.6) 129 (77.7) 26 (15.7)

.71 .65 .81 .67 .58 .62 .42 .91 .84 .60 .84

211 46 61 4.9 4.0 228 9.7 3.5 6.6

(66.4) (14.5) (19.2) (2.9) (3.3) (71.7) (6.1) (2.5) (3.1)

0.73 29 55 13

(0.17) (9.1) (17.3) (4.1)

27 (8.5) 276 (86.8) 15 (4.7)

BPI, Brief Pain Inventory; EQ-5D, EuroQol-5D; F, female; NRS, Numerical Rating Scale; PT, physical therapy; RMDQ, Roland-Morris Disability Questionnaire; SD, standard deviation. * Patients with leg pain present only.

smoking status, baseline RMDQ, baseline back pain NRS and leg pain NRS, baseline BPI, baseline EQ-5D, PHQ-4 anxiety score, PHQ-4 depression score, expectation for recovery, duration of current symptoms, and study site [30–32]. For later PT, we conducted a separate propensity score–matched analysis using the 3-month outcome measures rather than baseline measures. We matched patients receiving early or later PT to as many as four controls based on propensity score using calipers of 0.2 standard deviations of the mean propensity score (0.009 for early PT and 0.010 for later PT) [13]. We initially assessed the balance between groups by comparing the means and standard deviations or proportion of variables, and we also examined the standardized percentage bias of variables between matched groups [33]. We then tested the differences in means using a t test and differences in proportions using a chi-square or Fisher exact

test. The final model for creating the propensity score included age, gender, race, Hispanic ethnicity, education, marital status, smoking (yes/no), RMDQ, back pain, leg pain, EQ-5D, BPI, PHQ-4, expectation for recovery, back pain duration less than or equal to 3 months (yes/no), and study site. For the second step, we applied multiple linear regression models to these matched groups to estimate the mean difference in outcome scores. We performed this step to minimize any residual confounding. We adjusted the model using the propensity score, baseline or 3-month RMDQ, baseline or 3-month back and leg pain, pain expectations, back pain duration, and study site as covariates in these models. We selected these adjustment variables a priori because we considered them important confounders. We generated adjusted point estimates and 95% confidence intervals (CIs) using these models.

1750

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

Table 3 Characteristics of later phase propensity score–matched groups Chi-square, t, or Fisher exact test Variables Demographics Age (y), mean (SD) Sex (F), n (%) Ethnicity (Hispanic), n (%)y Race, n (%) Black White Other Education, n (%) High school graduate or less Some college or trade school At least college graduate Marital status, n (%) Married or partner Separated, divorced, or never married Widowed Back pain history and current episode Symptom duration (mo), n (%) !3 3–12 $12 3-Mo back pain intensity, mean NRS (SD) 3-Mo leg pain intensity, mean NRS (SD)* 3-Mo functional status, mean RMDQ (SD) 3-Mo pain interference with activity, mean BPI (SD) Baseline expectation for recovery, mean NRS (SD) General health status 3-Mo EQ-5D, mean (SD) Positive depression screen, n (%) Positive anxiety screen, n (%) Smoker, n (%)y Study site, n (%) Detroit Northern California Boston

Control (n5215)

Later PT (n571)

74.2 (6.5) 142 (66.1) 4 (1.9)

74.1 (6.5) 25 (64.8) 2 (2.8)

31 (14.4) 170 (79.1) 14 (6.5)

10 (14.1) 55 (77.5) 6 (8.5)

60 (27.9) 62 (28.8) 93 (43.3)

22 (31.0) 20 (28.2) 29 (40.9)

141 (65.6) 28 (13.0) 46 (21.4)

49 (69.0) 10 (14.1) 12 (16.9)

p Value .88 .85 .64 .86

Unmatched later PT (n528) 71.4 (5.3) 19 (67.9) 2 (7.1) 5 (17.9) 17 (60.7) 4 (14.3)

.88 6 (21.4) 12 (42.9) 9 (32.1) .72 18 (64.3) 4 (14.3) 5 (17.9) .94 96 35 84 5.6 5.3 13.9 4.7 4.8

(44.7) (16.3) (39.1) (1.9) (2.3) (4.3) (1.9) (3.5)

31 15 25 5.4 5.3 13.8 4.6 5.2

(43.7) (21.1) (35.2) (1.9) (2.2) (4.5) (2.2) (3.5)

0.66 21 38 11

(0.18) (9.8) (17.7) (5.1)

0.66 9 11 3

(0.18) (12.7) (15.5) (4.2)

28 (13.0) 176 (81.9) 11 (5.1)

10 (14.1) 58 (81.7) 3 (4.2)

.36 .78 .94 .94 .50 .84 .49 .67 1.00 .94

16 6 6 6.5 6.6 15.5 5.7 4.9

(57.1) (21.4) (21.4) (1.9) (2.2) (5.3) (2.6) (4.0)

0.68 6 9 0

(0.22) (21.4) (32.1) (0)

3 (10.7) 23 (82.1) 2 (7.1)

BPI, Brief Pain Inventory; EQ-5D, EuroQol-5D; F, female; NRS, Numerical Rating Scale; PT, physical therapy; RMDQ, Roland-Morris Disability Questionnaire; SD, standard deviation. * Patients with leg pain present only. y Fisher exact test.

Secondary analyses examined the proportion achieving a minimal clinically important difference in function and back pain, tested for a dose-response relationship, and tested for heterogeneity of treatment effects between early PT and 12-month outcomes. A minimal clinically important difference was defined as at least 30% improvement for the RMDQ and back pain NRS [19]. We also examined the proportion achieving a stricter criterion of 50% improvement for each outcome. We then compared the proportion of patients with 30% and 50% improvement in function and back pain between matched groups. Emulating the methods of the primary analysis, we used multiple logistic regressions to estimate the adjusted odds ratio (OR) for a clinically meaningful change by the early or later PT groups versus their respective comparators.

To test a dose-response relationship, we added a PT dose variable to the linear regression models in Step 2. We categorized dose as low (1–2 PT CPT codes), medium (3–9 PT CPT codes), and high ($10 PT CPT codes) based on the estimated number of corresponding visits. We only performed this secondary analysis using the two sites with PT CPT code–level data available (Northern California and Boston). We selected three potential treatment effect modifiers a priori: education, leg pain, and symptom duration at baseline. We included categorical versions of these variables in the linear regression models and tested their interaction with treatment group for heterogeneity of treatment effect. We performed all the analyses with Stata IC, version 12.1 (StataCorp, College Station, TX, USA) and set the significance level at a p value of less than .05 for analyses.

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

Results We analyzed 3,705 of the 5,239 participants from the BOLD registry (Figure). During the first 28 days, 484 (13.1%) initiated PT. The early PT group tended to have fewer African-Americans, more education, shorter duration of back symptoms, greater proportion with leg pain, higher expectations for recovery, and more participants with anxiety symptoms (Table 1). At 3 months, 2,002 (54.0%) were eligible for the later phase analysis. Of these eligible participants, 99 (4.9%) initiated PT during the later phase. Participants initiating later PTappeared to have more education, a greater proportion that were married or living with a partner, shorter duration of symptoms, more leg pain symptoms, fewer smokers, more depression symptoms, and more anxiety symptoms (Table 1). A total of 1,105 (29.8%) participants used PT services at any time. In the early PT group, 371 of the 446 (83.2%) from the 2 sites with PT CPT code–level data available used less than 10 PT CPT codes total over the entire 12 months. Among the participants from the two sites with PT CPT code–level data, the median number of PT CPT codes in the early PT group was 4 (interquartile range: 2, 8), and active treatments accounted for about half of all PT services, median 2 (interquartile range: 1, 4). After propensity score development, the final matched groups for both phases had standardized percentage bias of less than 10.0% for all variables. For the early phase, the group means or proportions of baseline variables were not statistically different from each other (p$.14). For the later phase, there were no differences between matched groups in means or proportions of baseline demographics or 3-month patient-reported outcomes (p$.36) (Tables 2 and 3). We present unadjusted outcomes by unmatched treatment group in Table 4. There appear to be minimal differences in outcomes between the early PT group and no early PT group at 3 months. However, by 12 months, the early PT group had less functional limitation, less back and leg pain, and a higher health-related quality of life. The later PT group had greater functional limitation and lower healthrelated quality of life at 12 months compared with the group not receiving PT in the later phase (Table 4). Adjusted outcomes of the matched groups show no difference in pain or function between early PT and no early PT at 3 months. At 12 months, the early PT group had better functional status, with mean RMDQ of 1.1 points less than the no early PT group (95% CI: –2.2, –0.1) and less back pain of –0.5 (–0.9, –0.1) points different. There was no difference in the odds of a 30% improvement in function or back pain at 3 and 12 months for the early PT group compared with the no early PT group. However, the early PT group had increased odds of a 50% improvement in back pain at 3 months (OR: 1.59; 95% CI: 1.07, 2.36) and in function at 12 months (1.58; 1.04, 2.40). There was no difference in function or pain between the matched later phase groups at 12 months.

1751

The later PT group had decreased odds (OR: 0.35; 95% CI: 0.13, 0.95) of a 30% improvement in function at 12 months and no difference for an important change in pain at 12 months (Tables 5 and 6). Table 7 presents baseline characteristics of the matched early PT users by dose category for the secondary analysis. Among the two sites with PT CPT code–level data available, greater PT utilization within the early PT group was associated with better functional status at 12 months (test for trend, p5.01). Early PT patients who had greater than or equal to 10 PT CPT codes had the most pronounced benefit with a mean RMDQ that was 3.2 (95% CI: –5.2, –1.2) points lower than the no early PT group. There was also a dose-response relationship between amount of early PT and back pain intensity at 12 months (test for trend, p5.036). The early PT group with greater than or equal to 10 CPT codes had mean back pain score of 1.1 (95% CI: –1.9, –0.3) points less than the no early PT group (Table 8). There was no statistically significant treatment effect modification between early PT and education, leg pain, or symptom duration (p5.41, p5.13, and p5.24 respectively). Discussion We found that among older adults who presented to their primary care provider for a new episode of back pain, early referral to PT resulted in minimal improvement in function Table 4 Unadjusted clinical outcomes at 3, 6, and 12 months by treatment group Treatment group

RMDQ

Back pain NRS

Leg pain NRS

EQ-5D

3 Mo Early phase* No early PT Early PT Later phasey No later PT Later PT Early phase No early PT Early PT Later phase No later PT Later PT Early phase No early PT Early PT Later phase No later PT Later PT

9.1 (6.6) 8.9 (6.9)

3.7 (2.7) 3.5 (2.6)

3.0 (2.9) 2.9 (2.9)

0.77 (0.18) 0.77 (0.15)

13.3 (4.7) 14.1 (4.7)

5.6 (1.9) 5.6 (1.9)

4.3 (2.9) 5.0 (2.7) 6 Mo

0.68 (0.16) 0.65 (0.18)

8.7 (6.7) 7.9 (6.9)

3.7 (2.7) 3.1 (2.5)

3.0 (2.9) 2.6 (2.6)

0.77 (0.17) 0.79 (0.16)

11.8 (5.9) 12.8 (5.6)

4.9 (2.5) 4.9 (2.0)

3.9 (2.9) 4.2 (2.5) 12 Mo

0.71 (0.17) 0.68 (0.16)

8.4 (6.8) 7.1 (6.8)

3.6 (2.7) 3.2 (2.6)

2.9 (2.9) 2.4 (2.5)

0.78 (0.18) 0.81 (0.16)

11.4 (6.3) 12.7 (6.2)

4.8 (2.5) 5.0 (2.4)

3.9 (2.9) 4.2 (2.7)

0.72 (0.17) 0.68 (0.18)

EQ-5D, EuroQol-5D; NRS, Numerical Rating Scale; PT, physical therapy; RMDQ, Roland-Morris Disability Questionnaire. Note: Values are mean (standard deviation). * Participants with no injections, no surgery, and no chiropractic. y Participants with no early PT and RMDQ O4 and back pain NRS O2 at 3 months.

1752

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

Table 5 Adjusted clinical outcomes at 3, 6, and 12 months by matched treatment group Treatment group

RMDQ

Back pain NRS

EQ-5D

3 Mo Early phase* No early PT Early PT Difference

9.3 (8.7, 9.9) 9.1 (8.3, 9.9) 0.2 (1.1, 0.7)

No early PT Early PT Difference

9.0 (8.3, 9.6) 8.4 (7.6, 9.1) 0.6 (1.6, 0.4)

No early PT Early PT Difference

8.9 (8.2, 9.7) 7.8 (7.0, 8.6) 1.1 (2.2, 0.1)

Later phasey No later PT Later PT Difference

3.8 (3.5, 4.1) 3.5 (3.1, 3.8) 0.3 (0.7, 0.1) 6 Mo 3.7 (3.4, 4.1) 3.2 (2.8, 3.5) 0.6 (0.9, 0.2) 12 Mo 3.8 (3.5, 4.1) 3.3 (3.0, 3.7) 0.5 (0.9, 0.1) 12 Mo

10.5 (9.3, 11.7) 11.4 (10.1, 12.7) 0.9 (0.6, 2.5)

4.9 (4.4, 5.4) 4.9 (4.3, 5.4) 0.04 (0.7, 0.6)

0.76 (0.74, 0.77) 0.77 (0.75, 0.79) 0.01 (0.01, 0.04) 0.77 (0.76, 0.79) 0.79 (0.77, 0.81) 0.02 (0.01, 0.05) 0.78 (0.76, 0.80) 0.80 (0.78, 0.82) 0.02 (0.01, 0.05)

0.72 (0.68, 0.76) 0.69 (0.65, 0.72) 0.03 (0.09, 0.02)

EQ-5D, EuroQol-5D; NRS, Numerical Rating Scale; PT, physical therapy; RMDQ, Roland-Morris Disability Questionnaire. Note: Values are mean (95% confidence interval). * Participants with no injections, surgery, or chiropractic, and initiated PT after interview. y Participants with no early PT and RMDQ O4 and back pain NRS O2 at 3 months.

and back pain at 12 months compared with a matched group that did not receive early PT. Although the groups appeared to have similar functional outcomes at 3 months, those receiving early PT showed continued small improvements in function between the 3- and 12-month follow-up; the group that did not receive early PT appeared to plateau in terms of function. More participants in the early PT group achieved a clinically meaningful change in function at 12 months using the stricter criteria of 50% improvement.

In contrast, among subjects who had continuing moderate back pain at 3 months and did not receive early PT, initiating later PT did not appear to provide additional benefit. Compared with a matched control group, neither group had a significant decrease in their pain or function at 12-month follow-up. We described the use of PT services in integrated health system settings, which has not previously been reported, to our knowledge. Patterns of PT referral were found to be roughly similar to the previous studies

Table 6 Number, proportion, and adjusted OR for achieving a minimal clinically important difference at 3, 6, and 12 months by matched treatment groups Treatment group

30% Improvement RMDQ

50% Improvement RMDQ

30% Improvement back pain NRS

50% Improvement back pain NRS

246 (43.5) 86 (48.3) 1.14 (0.77, 1.68)

167 (29.6) 72 (40.5) 1.59 (1.07, 2.36)

234 (43.8) 95 (56.6) 1.78 (1.17, 2.71)

194 (36.3) 74 (44.1) 1.35 (0.90, 2.01)

252 (48.9) 84 (50.6) 1.09 (0.68, 1.73)

196 (38.1) 74 (44.6) 1.34 (0.85, 2.12)

72 (33.6) 23 (32.4) 1.17 (0.52, 2.63)

48 (22.4) 13 (18.3) 0.61 (0.20, 1.88)

3 Mo Early phase, n (%)* No early PT (n5565) Early PT (n5178) aOR (95% CI)

184 (32.3) 63 (35.4) 0.96 (0.64, 1.44)

143 (25.1) 45 (25.3) 0.89 (0.57, 1.41)

No early PT (n5534) Early PT (n5168) aOR (95% CI)

183 (34.3) 71 (42.3) 1.31 (0.86, 1.98)

148 (27.7) 55 (32.7) 1.17 (0.76, 1.79)

No early PT (n5515) Early PT (n5166) aOR (95% CI)

197 (38.3) 79 (47.6) 1.39 (0.92, 2.10)

153 (29.7) 67 (40.4) 1.58 (1.04, 2.40)

6 Mo

12 Mo

12 Mo Later phase, n (%)y No later PT (n5215) Later PT (n571) aOR (95% CI)

52 (24.2) 10 (14.1) 0.35 (0.13, 0.95)

32 (14.9) 10 (14.1) 0.77 (0.28, 2.15)

aOR, adjusted odds ratio; CI, confidence interval; NRS, Numerical Rating Scale; OR, odds ratio; PT, physical therapy; RMDQ, Roland-Morris Disability Questionnaire. * Participants with no injections, surgery, or chiropractic, and initiated PT after interview. y Participants with no early PT and RMDQ O4 and back pain NRS O2 at 3 months.

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

1753

Table 7 Baseline characteristics of early PT users by dose category (Northern California and Boston sites only) Variables Demographics Age (y), mean (SD) Sex (F), n (%) Ethnicity (Hispanic), n (%) Race, n (%) Black White Other Education, n (%) High school graduate or less Some college or trade school At least college graduate Marital status, n (%) Married or partner Separated, divorced, or never married Widowed Back pain history and current episode Symptom duration (mo), n (%) !3 3–12 $12 Back pain intensity, mean NRS (SD) Leg pain intensity, mean NRS (SD)* Leg pain present, n (%) Functional status, mean RMDQ (SD) Pain interference with activity, mean BPI (SD) Expectation for recovery, mean NRS (SD) General health status EQ-5D, mean (SD) Positive depression screen, n (%) Positive anxiety screen, n (%) Smoker, n (%) Study site, n (%) Northern California Boston

Low (1–2 PT CPT codes), n546

Medium (3–9 PT CPT codes), n580

High ($10 PT CPT codes), n529

73.6 (5.9) 31 (67.4) 1 (2.2)

73.1 (6.2) 56 (70.0) 7 (8.8)

73.5 (6.9) 18 (62.1) 2 (6.9)

2 (4.4) 36 (78.3) 8 (17.4)

3 (3.8) 70 (87.5) 7 (8.8)

3 (10.3) 19 (65.5) 7 (24.1)

12 (26.1) 14 (30.4) 20 (43.5)

19 (23.8) 18 (22.5) 43 (53.8)

7 (24.1) 5 (17.2) 17 (58.6)

34 (73.9) 2 (4.4) 10 (21.7)

50 (62.5) 17 (21.3) 13 (16.3)

20 (69.0) 4 (13.8) 5 (17.2)

31 9 6 4.6 5.0 33 8.1 3.1 6.7

(67.4) (19.6) (13.0) (2.5) (2.9) (71.7) (4.4) (2.3) (3.4)

55 8 17 5.5 5.6 51 10.0 3.7 6.2

(68.8) (10.0) (21.3) (2.5) (2.6) (63.8) (5.9) (2.3) (3.4)

18 6 5 5.8 5.7 24 12.5 4.6 6.9

(62.1) (20.7) (17.2) (2.3) (1.8) (82.8) (7.2) (2.8) (3.1)

0.79 3 3 2

(0.11) (6.5) (6.5) (4.4)

0.72 6 7 4

(0.16) (7.5) (8.8) (5.0)

0.69 4 8 0

(0.22) (13.8) (27.6) (0)

43 (93.5) 3 (6.5)

67 (83.8) 13 (16.3)

19 (65.5) 10 (34.5)

BPI, Brief Pain Inventory; CPT, Current Procedural Terminology; EQ-5D, EuroQol-5D; F, female; NRS, Numerical Rating Scale; PT, physical therapy; RMDQ, Roland-Morris Disability Questionnaire; SD, standard deviation. * Patients with leg pain present only.

after a new episode of back pain in other outpatient settings. We found that 30% of subjects received PT at some point in the year after a physician visit for a new episode of back pain. This is more than Gellhorn et al. [11] reported in a large cohort of Medicare beneficiaries, 16% within 1 year, but it is similar to studies using data from the National Ambulatory Medical Care Survey, where PT referrals were made in 21% to 27% of cases [34,35].

It is notable that the group receiving early PT was different in potentially important ways from no early PT group in the unmatched cohort. Subjects receiving early PT tended to have a higher level of education and were less likely to be smokers; they reported shorter mean symptom duration, a higher frequency of leg pain, and a higher expectation for recovery. A number of studies have focused on predictors for developing chronic back pain, and most

Table 8 Adjusted clinical outcomes at 12 months by PT dose for matched analysis of early phase (Northern California and Boston sites only) RMDQ

Back pain NRS

Early PT group

Mean (95% CI)

Difference from reference (95% CI)

Mean (95% CI)

Difference from reference (95% CI)

None (n5479) Low PT use (n546) Medium PT use (n580) High PT use (n529)

7.9 7.3 7.1 4.7

Reference 0.6 (2.3, 1.1) 0.9 (2.2, 0.50) 3.2 (5.2, 1.2)

3.8 3.5 3.5 2.6

Reference 0.3 (1.0, 0.3) 0.3 (0.8, 0.2) 1.2 (2.0, 0.4)

(7.2, (5.9, (5.8, (2.9,

8.7) 8.7) 8.4) 6.6)

(3.5, (2.9, (3.0, (1.9,

4.1) 4.1) 4.0) 3.4)

CI, confidence interval; NRS, Numerical Rating Scale; PT, physical therapy; RMDQ, Roland-Morris Disability Questionnaire.

1754

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755

authors agree that psychosocial factors are important determinants. In general, patient expectations for recovery appear to influence prognosis, whereas smoking status, education level, and the presence of leg pain do not appear to be independent prognostic indicators [25,36]. In the group that received early PT, the effect of this early intervention was most marked at the 12-month follow-up evaluation. In contrast with many interventions for back pain, which show short-term benefits for pain and function, we were interested to see the improvements of early PT continuing, and in fact increasing, between 3 and 12 months. These results are consistent with the previous analyses that found decreased back pain recurrence rates at 12 months in subjects receiving PT compared with those receiving medical management [37,38]. The reason for this sustained improvement may relate to instruction on self-management and home exercises, which are frequently a part of a formalized PT program. However, we could not evaluate the specific PT programs in detail, and PT as measured in this analysis likely represents a somewhat heterogeneous treatment intervention for both content and dosage. On average, the content of PT received was essentially half active treatment. This sustained improvement may also relate to the benefit of early activity in the acute stage of back pain, which has been shown to be superior to inactivity in a number of studies [39–41]. The long-term benefits seen in the early PT group were not seen in the group that received PT later. This is consistent with the previous studies that compare early PT with later PT [11,42]. The group receiving PT later likely represents subjects with more complex persistent back pain. In our analysis, the group receiving later PT had a longer duration of symptoms, more leg pain, higher levels of depression and anxiety, and a lower level of education in general compared with the group receiving early PT. The lack of response in this cohort deserves further study and may represent a group of patients who require higher levels and interdisciplinary care. The total amount of PT services received by the subjects in the early PT group was quite variable. Our subgroup analysis evaluating the dose of PT services received found that the vast majority of subjects received either ‘‘low’’ (1–2 CPT codes) or ‘‘medium’’ (3–9 CPT codes) amounts of PT services, with only 15% receiving a ‘‘high’’ ($10 CPT codes) dose. Low and medium doses may correspond to approximately one to four visits, depending on the length of time for visits. There was a significantly lower level of back pain and functional limitation at 12 months in those receiving high dose of PT. Although we detected a trend toward improved outcomes with higher doses of early PT, this was not a prespecified objective and must be considered hypothesis generation. Our analysis has a number of strengths. We were able to report long-term clinical outcomes on a large cohort of older adults, with well-described demographic and functional status. The propensity score matching used in the analysis

allowed us to isolate the effect of early referral to PT by creating highly similar treatment groups. Given the complex influence of psychosocial factors, creating a matched cohort where these factors were similar between those receiving early PT and those not receiving early PT allowed us to examine the role of PT more directly. Finally, we measured and reported patient outcomes directly, rather than inferring such outcomes based on health-care utilization trends. There are some limitations to the study. Although we were able to capture patient expectations of recovery, we did not assess patient treatment preferences directly. These may be important independent predictors of improvement with any given treatment [38]. Additionally, the content and dose of PT were variable after the referral, and we were not able to control these factors directly. Therefore, the treatment as analyzed represents a somewhat heterogeneous treatment. This is similar to the situation in routine clinical practice, where patients may attend a variable number of sessions of PT after a physician referral. Future research should examine the content and dose of PT more directly as we found evidence for a dose-response relationship in our subgroup analysis. Another limitation is that we were not able to account for all cointerventions. Specifically, we did not account for medication usage, which may affect patients’ pain. Additionally, participants in the later phase analysis may have used one or more concomitant interventions that could be associated with later outcomes. In summary, we designed this study to answer a question that clinicians frequently encounter: in an older adult, who presents with a new episode of back pain, with or without radicular features, will he or she benefit from an early course of PT and how much improvement can we expect from this intervention? We found that among older adults presenting to their primary care providers for a new episode of back pain, early referral to PT resulted in no or minimal differences in pain, function, or health-related quality at 3, 6, or 12 months compared with a matched group that did not receive early PT. Secondary analysis show that patients initiating early PT may be somewhat more likely to experience 50% improvement in function at 12 months. Among patients with continuing back pain, initiating later PT was not significantly associated with improved function at 12 months. Additionally, subgroup analysis suggests that patients using more PT services had the largest gains, but additional research is warranted to continue to define which patients will benefit most from formalized PT. References [1] Schappert SM, Rechtsteiner EA. Ambulatory medical care utilization estimates for 2006. Natl Health Stat Report 2008;8:1–29. [2] Vos T, Flaxman AD, Naghavi M, Lozano R, Michaud C, Ezzati 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 2013;380:2163–96.

S.D. Rundell et al. / The Spine Journal 15 (2015) 1744–1755 [3] Pengel LH, Herbert RD, Maher CG, Refshauge KM. Acute low back pain: systematic review of its prognosis. BMJ 2003;327:323. [4] Henschke N, Maher CG, Refshauge KM, Herbert RD, Cumming RG, Bleasel J, et al. Prognosis in patients with recent onset low back pain in Australian primary care: inception cohort study. BMJ 2008;337:a171. [5] Costa Lda C, Maher CG, McAuley JH, Hancock MJ, Herbert RD, Refshauge KM, et al. Prognosis for patients with chronic low back pain: inception cohort study. BMJ 2009;339:b3829. [6] Carey TS, Garrett JM, Jackman AM. Beyond the good prognosis. Examination of an inception cohort of patients wtih chronic low back pain. Spine 2000;25:115–20. [7] Scheele J, Enthoven WT, Bierma-Zeinstra SM, Peul WC, van Tulder MW, Bohnen AM, et al. Course and prognosis of older back pain patients in general practice: a prospective cohort study. Pain 2013;154:951–7. [8] Chou R, Qaseem A, Snow V, Casey D, Cross JT Jr, Shekelle P, et al. Clinical Efficacy Assessment Subcommittee of the American College of Physicians; American College of Physicians; American Pain Society Low Back Pain Guidelines Panel, 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;147:478–91. [9] Hicks G, Morone N, Weiner DK. Degenerative lumbar disc and facet disease in older adults. Spine 2009;34:1301–6. [10] Rudy TE, Weiner DK, Lieber SJ, Slaboda J, Boston JR. The impact of chronic low back pain on older adults: a comparative study of patients and controls. Pain 2007;131:293–301. [11] Gellhorn AC, Chan L, Martin BI, Friedly J. Management patterns in acute low back pain. The role of physical therapy. Spine 2012;37: 775–82. [12] Blackmore CC, Mecklenburg RS, Kaplan GS. At Virginia Mason, collaboration among providers, employers, and health plans to transform care cut costs and improved quality. Health Aff (Millwood) 2011;30:1680–7. [13] Jarvik JG, Comstock BA, Bresnahan BW, Nedeljkovic SS, Nerenz DR, Bauer Z, et al. Study protocol: the back pain outcomes using longitudinal data (BOLD) registry. BMC Musculoskelet Disord 2012;13:64. [14] Stratford PW, Binkley J, Solomon P, Finch E, Gill C, Moreland J. Defining the minimum level of detectable change for the Roland-Morris questionnaire. Phys Ther 1996;76:359–65; discussion 366–368. [15] Childs JD, Piva SR, Fritz JM. Responsiveness of the numeric pain rating scale in patients with low back pain. Spine 2005;30:1331–4. [16] Deyo RA, Andersson G, Bombardier C, Cherkin DC, Keller RB, Lee CK, et al. Outcome measures for studying patients with low back pain. Spine 1994;19(18 Suppl):2032S–6S. [17] Deyo RA, Diehr P, Patrick DL. Reproducibility and responsiveness of health status measures. Statistics and strategies for evaluation. Control Clin Trials 1991;12(4 Suppl):142S–58S. [18] Patrick DL, Deyo RA, Atlas SJ, Singer DE, Chapin A, Keller RB. Assessing health-related quality of life in patients with sciatica. Spine 1995;20:1899–908; discussion 1909. [19] Ostelo RW, Deyo RA, Stratford P, Waddell G, Croft P, Von Korff M, et al. Interpreting change scores for pain and functional status in low back pain. Towards international consensus regarding minimal important change. Spine 2008;33:90–4. [20] Herr KA, Spratt K, Mobily PR, Richardson G. Pain intensity assessment in older adults: use of experimental pain to compare psychometric properties and usability of selected pain scales with younger adults. Clin J Pain 2004;20:207–19. [21] Deyo RA, Battie M, Beurskens AJ, Bombardier C, Croft P, Koes B, et al. Outcome measures for low back pain research. A proposal for standardized use. Spine 1998;23:2003–13.

1755

[22] Kroenke K, Spitzer RL, Williams JBW. The Patient Health Questionnaire-2. Validity of a two-item depression screener. Med Care 2003;41:1284–92. [23] Kroenke K, Spitzer RL, Williams JB, Monahan PO, L€owe B. Anxiety disorders in primary care: prevalence, impairment, comorbidity, and detection. Ann Intern Med 2007;146:317–25. [24] Kroenke K, Spitzer RL, Williams JB, L€owe B. An ultra-brief screening scale for anxiety and depression: the PHQ-4. Psychosomatics 2009;50:613–21. [25] Iles RA, Davidson M, Taylor NF, O’Halloran P. Systematic review of the ability of recovery expectations to predict outcomes in nonchronic non-specific low back pain. J Occup Rehabil 2009;19:25–40. [26] Cleeland CS, Ryan KM. Pain assessment: global use of the brief pain inventory. Ann Acad Med Singapore 1994;23:129–38. [27] Rosenbaum P, Rubin D. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 1985;39:33–8. [28] St€urmer T, Joshi M, Glynn RJ, Avorn J, Rothman KJ, Schneeweiss S. A review of the application of propensity score methods yielded increasing use, advantages in specific settings, but not substantially different estimates compared with conventional multivariable methods. J Clin Epidemiol 2006;59:437–47. [29] Austin PC. The performance of different propensity-score methods for estimating relative risks. J Clin Epidemiol 2008;61:537–45. [30] Mielenz TJ, Carey TS, Dyrek DA, Harris BA, Garrett JM, Darter JD. Physical therapy utilization by patients with acute low back pain. Phys Ther 1997;77:1040–51. [31] Chevan J, Riddle DL. Factors associated with care seeking from physicians, physical therapists, or chiropractors by persons with spinal pain: a population-based study. J Orthop Sports Phys Ther 2011;41:467–76. [32] Freburger JK, Carey TS, Holmes GM. Physical therapy for chronic low back pain in North Carolina: overuse, underuse, or misuse? Phys Ther 2011;91:484–95. [33] Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med 2009;28:3083–107. [34] Hart LG, Deyo RA, Cherkin DC. Physician office visits for low back pain. Frequency, clinical evaluation, and treatment patterns from a U.S. national survey. Spine 1995;20:11–9. [35] Jackson JL, Browning R. Impact of national low back pain guidelines on clinical practice. South Med J 2005;98:139–43. [36] Chou R, Shekelle P. Will this patient develop persistent disabling low back pain? JAMA 2010;303:1295–302. [37] Hides JA, Jull GA, Richardson CA. Long-term effects of specific stabilizing exercises for first-episode low back pain. Spine 2001;26:E243–8. [38] Moffett JK, Torgerson D, Bell-Syer S, Jackson D, LlewlynPhillips H, Farrin A, et al. Randomised controlled trial of exercise for low back pain: clinical outcomes, costs, and preferences. BMJ 1999;319:279–83. [39] Wand BM, Bird C, McAuley JH, Dore CJ, MacDowell M, De Souza LH. Early intervention for the management of acute low back pain: a single-blind randomized controlled trial of biopsychosocial education, manual therapy, and exercise. Spine 2004;29:2350–6. [40] Malmivaara A, H€akkinen U, Aro T, Heinrichs ML, Koskenniemi L, Kuosma E, et al. The treatment of acute low back pain—bed rest, exercises, or ordinary activity? N Engl J Med 1995;332:351–5. [41] Pengel LH, Refshauge KM, Maher CG, Nicholas MK, Herbert RD, McNair P. Physiotherapist-directed exercise, advice, or both for subacute low back pain: a randomized trial. Ann Intern Med 2007;146: 787–96. [42] Fritz JM, Brennan GP, Hunter SJ, Magel JS. Initial management decisions after a new consultation for low back pain: implications of the usage of physical therapy for subsequent health care costs and utilization. Arch Phys Med Rehabil 2013;94:808–16.