Predictors of Persistent Disability and Back Pain in Older Adults with a New Episode of Care for Back Pain

Abstract

Objective. To identify predictors of persistent disability and back pain in older adults.

Design. Prospective cohort study.

Setting. Back pain outcomes using longitudinal data registry.

Subjects. Five thousand two hundred twenty adults age 65 years and older with a new primary care visit for back pain.

Methods. Baseline measurements included: demographics, health, and back pain characteristics. We abstracted imaging findings from 348 radiology reports. The primary outcomes were the Roland-Morris Disability Questionnaire (RMDQ) and back pain intensity. We defined persistent disability as RMDQ of 4/24 or higher at both six and 12 months and persistent back pain as pain 3/10 or higher at both six and 12 months.

Results. There were 2,498 of 4,143 (60.3%) participants with persistent disability, and 2,099 of 4,144 (50.7%) had persistent back pain. Adjusted analyses showed the following characteristics most strongly predictive of persistent disability and persistent back pain: sex, race, worse baseline clinical characteristics of back pain, leg pain, back-related disability and duration of symptoms, smoking, anxiety symptoms, depressive symptoms, a history of falls, greater number of comorbidities, knee osteoarthritis, wide-spread pain syndromes, and an index diagnosis of lumbar spinal stenosis. Within the imaging data subset, central spinal stenosis was not associated with disability or pain.

Conclusion. We found that many predictors in older adults were similar to those for younger populations.

 Chronic Back Pain, Disability, Predictors, Prognostic Factors, Older Adults

Introduction

The risk for persistent pain and disability is understudied in older adults with back pain. Persistent back pain is an important problem for older adults as it contributes to a lower quality of life. Older adults with persistent low back pain have greater functional limitations and more depressive symptoms compared with those without persistent back pain [1,2]. Persistent back pain in older adults has a large economic impact as well. Health care utilization and charges for back pain in Medicare beneficiaries are increasing [3,4]. In order to better develop prevention and management strategies for this major and important health issue among older adults, we need a better understanding of the predictors of persistent disability and back pain. This information is also vital to inform clinical decision-making regarding whom to target with interventions and what modifiable risk factors may be suitable targets for intervention.

Few prognostic studies of back pain involving older adults have prospectively followed participants from either a new visit or new onset of back pain. Most well-designed prognostic studies of back pain have included adults of all ages [5–9,10]. Collectively, these studies have found psychological variables (depression, anxiety, and loneliness), non-back-related joint pain, older age, and higher baseline pain or functional limitation to be predictors of later persistent back pain. Women also tended to have a greater prevalence of persistent back pain [5,7,11]. Yet, these studies do not help us fully understand how prognostic factors for older adults with back pain may vary from those for younger adults. In addition to the host of physiological changes due to aging, comorbidities are common in older adults [12]; they have more disabling pain [13], and they have a higher prevalence of degenerative processes [14] compared with younger adults, which may impact the importance of risk factors.

Few prospective studies have focused on the transition to persistent back pain specifically in older adults [9,11]. The only inception cohort study found that longer duration of symptoms, higher pain intensity, previous back pain, lack of pain radiating below the knee, greater comorbidities, lower expectation of recovery, and a longer duration of the timed ‘Up and Go’ test were associated with non-recovery at three months in a Dutch cohort [9]. To our knowledge, no inception cohort study with long-term follow-up of older adults beyond three months has been published. Consequently, there is a need for additional research that prospectively examines risk factors for developing persistent back-related disability and back pain after a new episode of care for back pain in older adults.

Our primary objective was to investigate predictors of persistent back-related disability and back pain over one year in older adults with a new primary care visit for back pain. Based on results of earlier studies, we hypothesized that female sex, arthritis in nonlumbar spine joints, depressive symptoms, anxiety symptoms, higher baseline pain, longer duration of symptoms, and lower expectation regarding recovery would be associated with having persistent disability and back pain. We investigated several additional potential predictors without having a specific hypothesis, and these should be considered exploratory analyses. Our secondary objectives were to investigate predictors of persistent back-related disability and back pain in the subgroup with acute back pain and examine prevalent lumbar imaging findings associated with disability and back pain 12 months after the index visit. We hypothesized that the presence of central spinal stenosis diagnosed by imaging would be associated with greater disability and pain at 12 months.

Methods

StudyDesign, Setting, andParticipants

We investigated predictors of persistent disability and back pain in older adults using data from the Back pain Outcomes using Longitudinal Data (BOLD) prospective cohort study. The formation of this cohort, its measures [15], and the clinic course of participants’ outcomes over 12 months [16] are described elsewhere. Briefly, BOLD is a cohort study of 5,239 older adults (≥65 years old) presenting to primary care settings for new visits for back pain during the years 2011–2013. They were allowed to have any duration of back symptoms. We recruited the cohort from three integrated health systems: Kaiser Permanente in Northern California, Henry Ford Health System in Detroit, and Harvard Vanguard Medical Associates in Boston. The main exclusion criteria were: a health care visit with a back pain diagnosis code or lumbar procedure code less than six months before the index visit, prior lumbar spine surgery, developmental spine deformities, inflammatory rheumatic spondyloarthropathy, known spinal malignancy or infection, history of cancer in the past five years, history of human immunodeficiency virus infection, serious medical comorbid condition with life expectancy of less than one year, no telephone, planning to move within a year, unable to understand English, or severe cognitive impairment that would interfere with answering questions. We contacted participants at three, six, and 12 months to collect patient-reported outcomes either by telephone or mailed questionnaire, depending on the study site.

For this analysis, we selected participants from the BOLD registry who had reached their 12-month follow-up and had electronic medical record (EMR) data available for 12 months before and after the index visit as of October 2014. The institutional review boards at the University of Washington, Kaiser Permanente Northern California, Henry Ford Health System, and Harvard Vanguard Medical Associates all approved the BOLD protocol.

Variables/Measurement

We collected baseline demographic data (including age, gender, ethnicity [Hispanic vs non-Hispanic], race, marital status, education level, employment status, and study site) by interview within three weeks of an eligible index visit for back pain. We categorized the back pain–related International Classification of Diseases, 9th revision (ICD9) code at index as: axial pain, back and leg pain, lumbar spinal stenosis, or other (see the Appendix). We asked participants about their expectation for recovery on an 11-point numerical rating scale (0 = no confidence in recovery, 10 = complete confidence in recovery), smoking status, and the duration of current back pain in months. We classified participants as having acute back pain if the duration of their current back pain was three months or less. We measured depressive and anxiety symptoms using the Patient Health Questionnaire-4 Depression and Anxiety Screen (PHQ-4) [17,18], and we measured general health status with the EQ-5D index score [19]. Pain interference with activity was assessed using the Brief Pain Inventory (BPI). The BPI score is an average of seven domains, scored from 0 (no interference) to 10 (interferes completely) [20]. We asked about number of falls in the prior three weeks using the Behavioral Risk Factor Surveillance System falls screening questions [21].

Using electronic medical records, we looked for the following comorbidities (as defined by at least one ICD9 code during the 12 months prior to the index visit): knee osteoarthritis, hip osteoarthritis, osteoporosis, cervical pain, and chronic widespread pain/fibromyalgia. We selected these comorbidities because they were hypothesized to be associated with persistent LBP a priori (see the Appendix). We also calculated Quan’s Comorbidity Index using the electronic medical record data from the 12 months prior to index [22].

We created an analytic subsample of randomly selected imaging records across all three study sites. Out of the 2,703 imaging reports with a qualifying lumbar imaging Current Procedural Terminology, 4th edition (CPT) code (CPT Codes: 72100, 72110, 72114, 72131, 72132, 72133, 72148, 72149, and 72158), we randomly selected 400 imaging reports, stratified on site and imaging modality (radiographs, magnetic resonance imaging [MRI], or computed tomography [CT] scans). The size of this subsample was based on practical considerations for the time and effort of manual abstraction. We excluded reports with incomplete or duplicate data (N = 25) and nonlumbar images (N = 14). We also found that 13 participants had multiple unique reports for the same imaging modality. For these participants, we selected the report created closest to the index date and excluded the other report, giving us a final total of 348 imaging reports for the analysis.

We abstracted imaging findings from these radiology reports. As degenerative changes in the lumbar spine are relatively stable over time and new imaging findings are uncommon [23,24], we abstracted lumbar imaging procedures that occurred within seven days prior to index and up to the 12-month follow-up. Two investigators (SDR and NJD) independently abstracted each radiology report using a standardized data collection form. Any discordant findings were adjudicated by a third investigator (JGJ), a neuroradiologist with over 20 years’ experience in spine imaging.

We measured patient-reported outcomes for back pain at baseline and three, six, and 12 months. We used a modified version of the Roland-Morris Disability Questionnaire (RMDQ) to measure back-related disability. The RMDQ is a 24-item questionnaire (0 = none, 24 = maximum), and we modified it by adding “or leg (sciatica)” to the words “back pain” for each item [15,25,26]. We recorded average back and leg pain intensity during the prior seven days using an 11-point pain numerical rating scale (NRS) (0 = no pain, 10 = worst pain) that has been validated for older adults [27]. We defined the presence of persistent disability as RMDQ of 4/24 or higher at both six and 12 months, and we defined persistent back pain as back pain NRS of 3/10 or higher at both six and 12 months [28,29].

MissingData

We accounted for missing outcomes using inverse probability weights. Missing 12-month outcomes were associated with: older age, nonwhite race, lower education, higher baseline RMDQ, cervical pain, baseline anxiety symptoms, and baseline depressive symptoms at P ≤ 0.1. We included these variables along with study site in a logistic model to estimate the probability of having a competed 12-month RMDQ. This method reduces potential selection bias by creating a “pseudo population” with a similar probability of having a complete 12-month RMDQ [30]. Most participants with missing six-month outcomes also had missing 12-month outcomes, and most with missing RMDQ scores also had missing back pain NRS scores. Consequently, we applied the inverse probability weights for a completed 12-month RMDQ in all regression models as it essentially accounted for the missing data of either outcome at both six and 12 months.

Statistical Analysis

We used means and standard deviations (SDs) or frequencies and proportions of demographic variables, baseline health characteristics, and baseline patient-reported outcomes to describe the sample. We report the prevalence of imaging findings using frequencies and proportions within the subsample with abstracted imaging records. We described unadjusted six- and 12-month outcomes using mean and SD of back pain NRS and RMDQ. We also present the number and proportion of participants meeting our definition of persistent disability and persistent back pain at six and 12 months.

We examined the associations between each predictor variable and persistent disability or persistent back pain using logistic regression models. We reweighted all models using the inverse probability weights for complete outcomes. We ran separate models for each predictor, adjusting for age, gender, race, marital status, education level, employment status, and study site. We present odds ratios (ORs) and their 95% confidence intervals (CIs) to describe each predictor’s association with a patient having persistent disability or persistent back pain after a new episode of care for back pain. We performed these analyses on all participants meeting the inclusion and exclusion criteria to inform our primary objective and a subset restricted to those with back pain duration three months or less for our secondary analysis.

We tested the association between imaging findings and 12-month back-related disability or back pain using the subset of abstracted radiology reports. We used multiple linear regressions to test the association between each radiographic feature and back-related disability or back pain at 12 months. Due to the smaller sample size for imaging findings, we used the continuous back-related disability or back pain scores to enhance statistical power. We applied the inverse probability weights and adjusted for age, gender, race, and study site. All statistical analysis was performed using Stata IC, version 12.1 (College Station, TX, USA). We used a standard two-sided significance value of P < 0.05.

Results

We analyzed data from 5,220 participants from the BOLD registry (Figure 1). The mean age was 73.8 years (SD = 6.9), 3,375 (64.7%) were women, 1,386 (26.6%) were nonwhite, 2,176 (41.7%) were at least college graduates, and 3,177 (60.9%) were married or living with a partner. About half of the participants (N = 2,756, 52.8%) had back pain for three months or less. Almost two-thirds (N = 3,291, 63.1%) had leg pain (Table 1).

Figure 1

Flow diagram of study enrollment.

Table 1

Baseline characteristic of older adults with a new visit for back pain

	All participants with 12-mo follow-up (N = 5,220)
Variable	Number responding	Mean or number	SD or %
Demographics
Age (mean, SD), y	5,220	73.8	6.9
Sex (N, % male)	5,220	1,845	35.3
Ethnicity, (N, % Hispanic)	5,195	305	5.9
Race (N, %)	5,220
Black		788	15.1
Native American/Alaskan/Hawaiian or Pacific Islander		44	0.8
Asian		197	3.8
White		3,834	73.5
Other		299	5.7
NA		58	1.1
Education (N, %)	5,220
Less than high school graduate		313	6.0
High school graduate/GED or vocational/tech/trade school		1,444	27.7
Some college		1,271	24.4
Four year college graduate		1,285	24.6
Professional or graduate degree		891	17.1
NA		16	0.3
Marital status (N, %)	5,220
Married or partner		3,177	60.9
Separated or divorced		600	11.5
Never married and single		255	4.9
Widowed		1,173	22.5
NA		15	0.3
Employment (N, %)	5,220
Working full-time/part-time		601	11.5
Retired (not due to ill health)		4,233	81.1
Retired or disabled because of ill health		145	2.8
Other		220	4.2
NA		21	0.4
Study site (N, %)	5,220
Detroit		953	18.3
Northern California		3,164	60.6
Boston		1,103	21.1
Symptom duration (N, %)	5,220
<1 mo		1,746	33.5
1–3 mo		1,010	19.4
3–6 mo		342	6.6
6–12 mo		313	6.0
1–5 y		772	14.8
>5 y		1,032	19.8
NA		5	0.1
Back pain intensity (mean, SD NRS*)	5,220	5.0	2.8
Leg pain present (N, % yes)	5,218	3,291	63.1
Leg pain intensity (mean, SD NRS)†	3,291	5.5	2.5
Functional status (mean, SD RMDQ‡)	5,220	9.6	6.4
Pain interference with activity (mean, SD)	5,220	3.3	2.5
Health characteristics
EQ5D§ (mean, SD)	5,205	0.76	0.2
Smoking (N, %)	5,220
Never		2,888	55.3
Quit > 1 y ago		1,996	38.2
Smoker or quit < 1 y ago		321	6.2
NA		15	0.3
Expectation for recovery (mean, SD)	5,206	5.5	3.7
Positive anxiety screen (N, % yes)	5,208	642	12.3
Positive depression screen (N, % yes)	5,206	431	8.3
Osteoporosis (N, % yes)	5,220	277	5.3
Knee osteoarthritis (N, % yes)	5,220	372	7.1
Hip osteoarthritis (N, % yes)	5,220	95	1.8
Cervical pain (N, % yes)	5,220	262	5.0
Widespread pain syndromes (N, % yes)	5,220	278	5.3
Falls in prior 3 wks	5,220
0		4,834	92.6
1		291	5.6
2+		91	1.7
NA		4	0.1
Quan Comorbidity Index Score	5,220	0.85	1.4
Index Diagnostic Code Category (N, %)	5,220
Axial pain		3,542	67.9
Back and leg pain		1,093	20.9
Lumbar spinal stenosis		286	5.5
Other		299	5.7

NA = not answered; NRS = numerical rating score; RMDQ = Roland-Morris Disability Questionnaire.

*

Scored from 0 to 10 on the numerical rating scale.

†

Participants with leg pain only.

‡

Scored from 0 to 24 on the Roland-Morris Disability Questionnaire.

§

Scored from 0 to 1 on the EQ-5D-5L index.

The final subsample with imaging findings included 348 imaging reports from 324 individuals: 155 radiographs, 149 MRIs, and 44 CTs (Figure 1). We describe the prevalence of imaging findings in the lumbar spine by imaging modality in Table 2. The most common findings were disc space narrowing (radiographs: N = 119 [76.8%], MRI: N = 61 [40.9%], and CT: N = 24 [54.6%]), moderate-to-severe foraminal or lateral recess stenosis (MRI: N = 79 [53.0%], and CT: N = 31 [70.5%]), and mild listhesis (radiographs: N = 55 [35.5%], MRI: N = 65 [43.6%], CT: N = 23 [52.3%]) (Table 2).

Table 2

Prevalence of anatomical features and their association with 12-mo disability or back pain scores (adjusted for age, sex, race, study site and missingness)

Anatomic feature			Back-related disability (RMDQ)				Back pain (pain NRS)
Plain films (N = 155)	N	(%)	Beta	95% confidence interval		P values	Beta	95% confidence interval		P values
Any decreased disc space	119	76.8	1.02	−1.43	3.47	0.41	−0.35	−1.33	0.63	0.48
Moderate to severe central stenosis	1	0.7
Any lateral recess/foraminal stenosis	0	0.0
Moderate to severe facet arthropathy	31	20.0	1.26	−1.52	4.05	0.37	0.68	−0.62	1.98	0.30
Grade 1 listhesis	55	35.5	1.23	−1.23	3.68	0.33	0.00	−1.09	1.09	1.00
Severe scoliosis	0	0.0
Severe kyphosis	3	1.9
Compression fracture	17	11.0	−2.41	−5.99	1.18	0.19	−0.78	−2.42	0.87	0.35
MRI (N = 149)
Moderate to severe disc dessication	53	35.6	−0.55	−2.95	1.86	0.65	−0.38	−1.43	0.67	0.48
Any decreased disc space	61	40.9	1.86	−0.49	4.22	0.12	0.29	−0.78	1.36	0.59
Medium to large disc bulge	39	26.2	−1.11	−4.00	1.78	0.45	−0.44	−1.81	0.93	0.53
Medium to large disc protrusion	16	10.7	0.02	−4.76	4.80	0.99	−0.08	−2.01	1.85	0.93
Any disc extrusion	6	4.0	4.09	−3.16	11.34	0.27	−0.01	−2.94	2.91	0.99
Any disc herniation	15	10.1	−2.92	−6.10	0.25	0.07	−0.70	−2.13	0.73	0.33
Type 1 modic changes*	11	7.4	5.75	1.03	10.47	0.02	2.12	0.58	3.67	0.01
Moderate to severe central stenosis	37	24.8	−1.61	−4.42	1.19	0.26	−0.96	−2.19	0.28	0.13
Any lateral recess or foraminal stenosis	79	53.0	−2.61	−5.40	0.19	0.07	−1.05	−2.23	0.13	0.08
Nerve root compression	46	30.9	1.59	−1.35	4.54	0.29	0.36	−0.94	1.65	0.59
Moderate to severe facet arthropathy	42	28.2	−0.24	−3.14	2.66	0.87	0.07	−1.23	1.38	0.91
Grade 1 listhesis	65	43.6	−1.85	−4.28	0.57	0.13	−0.46	−1.54	0.62	0.40
Compression fracture	10	6.7

Anatomic feature			Back-related disability (RMDQ)				Back pain (pain NRS)
CT (N = 44)	N	(%)	Beta	95% confidence interval		P values	Beta	95% confidence interval		P values
Decreased disc space	24	54.6	1.50	−4.16	7.16	0.59	0.88	−1.57	3.34	0.47
Medium to large disc bulge	12	27.3	3.75	−1.47	8.96	0.15	2.28	−0.74	5.29	0.13
Medium to large disc protrusion	4	9.1	6.40	1.93	10.87	0.01	3.71	1.60	5.83	0.00
Any disc extrusion	1	2.3	4.48	−0.26	9.22	0.06	−0.05	−2.28	2.19	0.97
Any disc herniation	4	9.1	−0.12	−13.71	13.46	0.99	−0.29	−5.84	5.27	0.92
Moderate to severe central stenosis	19	43.2	1.84	−3.94	7.62	0.52	1.50	−1.07	4.08	0.24
Any lateral recess or foraminal stenosis	31	70.5	3.95	−2.74	10.65	0.24	3.48	1.64	5.31	0.00
Nerve root compression	12	27.3	2.43	−3.86	8.72	0.44	1.34	−1.23	3.91	0.30
Moderate to severe facet arthropathy	20	45.5	4.93	0.39	9.47	0.03	2.18	0.38	3.99	0.02
Grade 1 listhesis	23	52.3	3.78	−2.75	10.30	0.25	0.52	−2.18	3.21	0.70
Severe scoliosis	2	4.6
Compression fracture	1	2.3
MRI and CT combined (N = 188)
Any decreased disc space	85	45.2	1.81	−0.29	3.92	0.09	0.30	−0.63	1.24	0.52
Medium to large disc bulge	51	27.1	−0.09	−2.56	2.38	0.94	0.18	−1.00	1.36	0.76
Medium to large disc protrusion	19	10.1	1.08	−3.31	5.47	0.63	0.59	−1.24	2.41	0.53
Any disc extrusion	7	3.7	3.94	−2.08	9.95	0.20	−0.09	−2.53	2.35	0.94
Any disc herniation	19	10.1	−1.60	−5.05	1.85	0.36	−0.39	−1.86	1.08	0.60
Moderate to severe central stenosis	54	28.7	−1.42	−3.78	0.95	0.24	−0.55	−1.59	0.50	0.30
Any lateral recess or foraminal stenosis	107	56.9	−1.39	−3.85	1.07	0.27	−0.30	−1.40	0.79	0.58
Nerve root compression	58	30.9	1.59	−0.95	4.13	0.22	0.54	−0.57	1.64	0.34
Moderate to severe facet arthropathy	61	32.5	0.77	−1.73	3.26	0.55	0.61	−0.44	1.67	0.26
Grade 1 listhesis	83	44.2	−0.72	−2.85	1.41	0.51	−0.27	−1.21	0.66	0.56

*

Endplate edema. CT = computed tomography; MRI = magnetic resonance imaging; NRS = numerical rating scale; RMDQ = Roland-Morris Disability Questionnaire.

Outcomes

At 12 months, the mean back pain intensity decreased from 5.0 (SD = 2.8) to 3.6 (SD = 2.7), and the mean RMDQ decreased from 9.6 (SD = 6.4) to 8.4 (SD = 6.8). After a year of follow-up, 60.3% (2,498 of 4,143) of participants had persistent disability and 50.7% (2,099 of 4,144) had persistent back pain.

Unadjusted Associations

Most predictors were associated with both persistent disability and persistent back pain in the unadjusted analyses (see the Appendix). However, a comorbid diagnosis of osteoporosis was not associated with either persistent disability or persistent back pain. Additionally, a prior diagnosis of hip osteoarthritis was not significantly associated with persistent disability or persistent back pain.

Within the acute back pain subgroup (duration of symptoms ≤ 3 months), most predictors were also associated with persistent disability and persistent back pain for the unadjusted analysis, Similarly, neither comorbid osteoporosis nor a prior diagnosis of hip osteoarthritis was significantly associated with either outcome. Within this subgroup, the index diagnosis code categories were not consistently associated with the outcomes as well (see the Appendix).

Adjusted Associations

Estimates adjusted for demographic variables, study site, and missing outcomes found that for every 10-year increase in age there was a modest increase in persistent disability (OR = 1.28, 95% CI =  1.14–1.43). Men had less persistent disability (OR =  0.67, 95% CI =  0.57–0.78) and persistent back pain (OR =  0.66, 95% CI =  0.57–0.76) compared with women. Compared with whites, blacks were more likely to have persistent disability (OR =  1.36, 95% CI =  1.08–1.72) and persistent pain (OR =  1.52, 95% CI =  1.22–1.90). High educational attainment was associated with lower odds of persistent disability and pain (Table 3).

Table 3

Adjusted association with persistent disability or persistent back pain (adjusted for demographics, site, and missingness)

	Persistent disability at both 6 and 12 mo (RMDQ ≥ 4)				Persistent back pain at both 6 and 12 mo (NRS ≥ 3/10)
Variable	Odds ratio	95% confidence interval		P values	Odds ratio	95% confidence interval		P values
Age	1.02	1.01	1.04	<0.001	1.01	1.00	1.02	0.162
Sex (male)	0.67	0.57	0.78	<0.001	0.66	0.57	0.76	<0.001
Race
White	Reference				Reference
Black	1.36	1.08	1.72	0.009	1.52	1.22	1.90	<0.001
Other	1.10	0.87	1.39	0.438	1.07	0.85	1.33	0.584
Education
Less than high school graduate	Reference				Reference
High school graduate/GED or trade school	1.02	0.69	1.51	0.906	1.19	0.84	1.67	0.332
Some college	0.96	0.65	1.42	0.835	1.13	0.79	1.60	0.507
Four-year college graduate	0.63	0.42	0.93	0.021	0.80	0.56	1.14	0.218
Professional or graduate degree	0.53	0.36	0.80	0.002	0.66	0.46	0.95	0.025
Marital status
Married or partner	Reference				Reference
Separated or divorced	0.93	0.74	1.16	0.513	0.99	0.80	1.23	0.924
Never married and single	1.28	0.92	1.77	0.149	0.93	0.67	1.29	0.664
Widowed	1.07	0.88	1.30	0.472	1.03	0.85	1.24	0.776
Employment
Working full-time/part-time	Reference				Reference
Retired (not due to ill health)	1.21	0.97	1.50	0.090	1.08	0.87	1.35	0.470
Retired or disabled due to ill health	2.23	1.31	3.79	0.003	2.62	1.61	4.25	<0.001
Other	1.60	1.09	2.36	0.017	1.25	0.86	1.83	0.246
Symptom duration
<1 mo	Reference				Reference
1–3 mo	1.37	1.14	1.64	0.001	1.35	1.12	1.63	0.001
3–6 mo	1.55	1.18	2.05	0.002	2.17	1.65	2.86	<0.001
6–12 mo	2.74	2.01	3.73	<0.001	2.75	2.07	3.67	<0.001
1–5 y	3.40	2.71	4.25	<0.001	3.85	3.10	4.78	<0.001
>5 y	4.10	3.32	5.07	<0.001	4.66	3.82	5.68	<0.001
Functional status (RMDQ)	1.19	1.17	1.21	<0.001	1.13	1.11	1.14	<0.001
Back pain intensity (NRS)	1.23	1.20	1.26	<0.001	1.26	1.23	1.30	<0.001
Leg pain intensity (NRS)*	1.15	1.11	1.20	<0.001	1.13	1.09	1.17	<0.001
Leg pain present (Yes)	2.28	1.98	2.62	<0.001	1.87	1.63	2.14	<0.001
Pain interference (BPI)	1.39	1.35	1.44	<0.001	1.29	1.25	1.33	<0.001
EQ-5D	0.01	0.00	0.01	<0.001	0.03	0.02	0.04	<0.001
Smoking
Never	Reference				Reference
Quit > 1 y ago	1.20	1.04	1.38	0.012	1.06	0.92	1.21	0.425
Smoker or quit < 1 y ago	1.64	1.19	2.24	0.002	1.55	1.16	2.08	0.003
Expectation for recovery	0.85	0.83	0.86	<0.001	0.85	0.84	0.87	<0.001
Positive anxiety screen	2.03	1.60	2.57	<0.001	1.65	1.33	2.03	<0.001
Positive depression screen	2.91	2.10	4.02	<0.001	2.10	1.60	2.75	<0.001
Osteoporosis	0.75	0.56	1.01	0.055	0.83	0.62	1.10	0.192
Knee osteoarthritis	1.69	1.29	2.22	<0.001	1.40	1.10	1.79	0.007
Hip osteoarthritis	1.63	0.96	2.76	0.070	1.62	0.99	2.65	0.053
Cervical pain	1.27	0.94	1.70	0.117	1.31	0.99	1.73	0.056
Widespread pain syndromes	2.05	1.50	2.81	<0.001	2.03	1.52	2.72	<0.001
Falls in prior 3 wks
0	Reference				Reference
1	0.86	0.65	1.14	0.291	0.94	0.71	1.24	0.661
2+	6.27	2.83	13.88	<0.001	1.84	1.09	3.10	0.022
Quan Comorbidity Index	1.20	1.14	1.27	<0.001	1.16	1.10	1.22	<0.001
Index Diagnostic Code Category
Axial pain	Reference				Reference
Back and leg pain	1.29	1.09	1.52	0.003	1.05	0.89	1.23	0.550
Lumbar spinal stenosis	1.83	1.36	2.47	<0.001	1.69	1.27	2.25	<0.001
Other	1.35	1.00	1.83	0.050	1.34	1.00	1.78	0.047

BPI = Brief Pain Inventory; NRS = numerical rating scale; RMDQ = Roland-Morris Disability Questionnaire.

*

Participants with leg pain only.

We found worse baseline clinical characteristics (i.e., higher levels of back pain, leg pain, back-related disability, longer duration of symptoms, and worse health status) were associated with both persistent disability and persistent back pain (Table 3). Compared with people who never smoked, those who were current or recent smokers were more likely to have persistent disability and persistent back pain (OR =  1.64, 95% CI =  1.19–2.24, and OR =  1.55, 95% CI =  1.16–2.08, respectively). Increased expectations for recovery were associated with better outcomes, such that for every one-point increase in expectation of recovery, the odds of persistent disability and persistent back pain decreased by 15% (OR =  0.85, 95% CI =  0.83–0.86, and OR =  0.85, 95% CI =  0.84–0.87, respectively). Comorbid anxiety symptoms, depressive symptoms, knee osteoarthritis, and widespread pain syndromes were associated with both persistent disability and persistent back pain. Participants who reported two or more falls within the three weeks prior to baseline were more likely to have persistent disability and persistent pain. The odds of persistent disability increased 20% for every one-point increase on Quan’s Comorbidy Index (OR =  1.20, 95% CI =  1.14–1.27), and the odds of persistent pain increased 16% (OR =  1.16, 95% CI = 1.10–1.22). Compared to participants with a diagnosis code in the category of axial back pain at their index visit, participants who had a diagnosis code for lumbar spinal stenosis were more likely to have persistent disability and persistent pain (Table 3).

Predictors of persistent disability and persistent back pain within the acute back pain subgroup were very similar to those for the entire cohort. Notable differences include an index diagnosis code for spinal stenosis and being retired/disabled due to ill health no longer being associated with persistent disability and persistent pain in this subgroup (Table 4).

Table 4

Acute back pain subgroup, adjusted association with persistent disability or persistent back pain (adjusted for demographics, site, and missingness)

	Persistent disability at both 6 and 12 mo (RMDQ ≥ 4)				Persistent back pain at both 6 and 12 mo (NRS ≥ 3/10)
Variable	Odds ratio	95% confidence interval		P values	Odds ratio	95% confidence interval		P values
Age	1.03	1.02	1.05	<0.001	1.02	1.00	1.03	0.038
Sex (male)	0.60	0.50	0.73	<0.001	0.61	0.50	0.74	<0.001
Race
White	Reference				Reference
Black	1.48	1.10	2.00	0.01	1.51	1.12	2.04	0.007
Other	1.12	0.83	1.51	0.47	1.04	0.77	1.40	0.819
Education
Less than high school graduate	Reference				Reference
High school graduate/GED or vocational/tech/trade school	0.95	0.56	1.62	0.862	0.92	0.56	1.52	0.746
Some college	1.00	0.59	1.71	0.999	0.86	0.52	1.43	0.566
Four-year college graduate	0.71	0.41	1.22	0.215	0.71	0.42	1.19	0.197
Professional or graduate degree	0.59	0.34	1.02	0.060	0.53	0.32	0.90	0.019
Marital status
Married or partner	Reference				Reference
Separated or divorced	0.82	0.61	1.10	0.175	0.94	0.70	1.27	0.696
Never married and single	0.99	0.66	1.50	0.976	0.97	0.65	1.45	0.881
Widowed	0.96	0.75	1.23	0.740	1.00	0.78	1.28	0.993
Employment
Working full-time/part-time	Reference				Reference
Retired (not due to ill health)	1.20	0.91	1.58	0.197	1.09	0.82	1.46	0.540
Retired or disabled because of ill health	1.47	0.75	2.88	0.257	1.97	1.04	3.73	0.037
Other	1.61	1.02	2.56	0.043	1.40	0.88	2.23	0.150
Functional status (RMDQ)	1.16	1.14	1.18	<0.001	1.10	1.09	1.12	<0.001
Back pain intensity (NRS)	1.19	1.15	1.23	<0.001	1.22	1.18	1.26	<0.001
Leg pain intensity (NRS)*	1.12	1.06	1.17	<0.001	1.07	1.02	1.12	0.003
Leg pain present (yes)	2.35	1.96	2.82	<0.001	1.95	1.62	2.35	<0.001
Pain interference (BPI)	1.35	1.29	1.40	<0.001	1.26	1.21	1.31	<0.001
EQ-5D	0.01	0.01	0.03	<0.001	0.05	0.03	0.09	<0.001
Smoking
Never	Reference				Reference
Quit > 1 y ago	1.20	1.00	1.45	0.056	1.07	0.88	1.30	0.480
Smoker or quit < 1 y ago	1.55	1.04	2.29	0.030	1.61	1.09	2.37	0.016
Expectation for recovery	0.86	0.83	0.88	<0.001	0.88	0.85	0.90	<0.001
Positive anxiety screen	1.77	1.32	2.38	<0.001	1.48	1.11	1.96	0.007
Positive depression screen	2.55	1.72	3.79	<0.001	1.84	1.28	2.65	0.001
Osteoporosis	0.65	0.43	0.96	0.032	0.64	0.42	0.96	0.031
Knee osteoarthritis	1.88	1.30	2.73	0.001	1.54	1.09	2.17	0.015
Hip osteoarthritis	1.68	0.78	3.65	0.188	2.43	1.14	5.18	0.021
Cervical pain	1.29	0.87	1.93	0.206	1.53	1.03	2.26	0.034
Widespread pain syndromes	1.67	1.10	2.55	0.017	1.81	1.19	2.75	0.005
Falls in prior 3 wks
0	Reference				Reference
1	0.86	0.60	1.23	0.410	1.03	0.72	1.48	0.858
2+	5.98	2.27	15.77	<0.001	1.71	0.85	3.43	0.134
Quan Comorbidity Index	1.19	1.11	1.28	<0.001	1.13	1.06	1.21	<0.001
Index Diagnostic Code Category
Axial pain	Reference				Reference
Back and leg pain	1.49	1.20	1.84	<0.001	1.18	0.95	1.46	0.133
Lumbar spinal stenosis	1.04	0.56	1.95	0.892	1.48	0.79	2.77	0.218
Other	1.89	1.20	2.97	0.006	1.90	1.25	2.89	0.003

BPI = Brief Pain Inventory; NRS = numerical rating scale; RMDQ = Roland-Morris Disability Questionnaire.

*

Participants with leg pain only.

No findings from the radiograph imaging reports were associated with 12-month outcomes (Table 2). However, a few findings on MRI and CT were associated with 12-month outcomes. Type one Modic changes (endplate edema) reported on MRI had some of the strongest associations with 12-month back-related disability (mean 5.8 points greater vs those without type one Modic changes, 95% CI =  1.0–10.5) and 12-month back pain (mean 2.1 points higher, 95% CI =  0.6–3.7). On CT, compared with participants with mild or no facet joint arthropathy, those with moderate or severe facet joint arthropathy had worse back-related function and back pain scores at 12 months, and participants with a moderate-to-severe disc protrusion had worse back-related disability and back pain intensity at 12 months compared with those with mild or no disc protrusions. These findings were no longer associated with 12-month outcomes when the common findings from the CT results were combined with the MRI result (Table 2).

Discussion

Among older adults with a new primary care visit for back pain, we found that female sex, comorbid pain conditions, depressive symptoms, anxiety symptoms, higher baseline pain, longer duration of symptoms, and lower expectation regarding recovery were associated with persistent disability and back pain. We also found older age, race, education level, worse baseline back-related function, worse baseline health status, a history of falls, comorbidity burden, a lumbar spinal stenosis diagnosis, and smoking were additional exploratory risk factors for persistent disability and persistent back pain. Predictors were similar for older adults with acute back pain compared with those with any duration of back pain, suggesting future research on predictors and screening tools can be investigated for new back pain visits regardless of symptom duration.

Our results are consistent with much of the recent research examining predictors of poor outcome in adults of all ages with back pain from primary care settings. Several studies have found worse back pain outcomes to be associated with psychosocial factors and clinical characteristics such as pain intensity, disability, and leg pain [31–35]. The commonalities between our results and these prior studies suggest that many predictors of poor outcome for back pain may be similar for adults of all ages, and clinicians may be able collect and use similar information for clinical decision-making with adults of all ages with back pain. Using the strongest known risk factors for poor outcome, the STarT Back Screening Tool was developed and validated in primary care settings to screen patients with back pain for risk of a poor outcome [10,36]. Four constructs from the STarT Back Screening Tool (i.e., leg pain, comorbid pain, disability, and expectation for improvement) were also risk factors that we found associated with persistent back pain and persistent disability in older adults, but we did not collect baseline data on the other constructs included in this tool. This similarity indicates that further work investigating the clinical utility of the STarT Back, or other similar screening tools [37], in older adults living in the United States may be productive.

Many of our findings are also consistent with another study focused only on older adults with back pain. In a cohort of Dutch older adults with new episodes of low back pain, Scheele et al. investigated predictors for nonrecovery of back pain at three months. Common predictive variables at baseline between our study and theirs were: longer duration of back pain, greater severity of back pain, greater number of comorbidities, and expectation for recovery. The presence of leg pain was one risk factor with contradictory results between the studies. We found the presence of leg pain was greater in those with persistent disability and pain. Scheele et al. reported that older adults with pain below the knee have a lower risk of nonrecovery [9].

Our study examined longer-term outcomes (six and 12 months) than Scheele et al. [9]. Regardless of the length of follow-up, many predictors of poor outcome appear similar between the two studies. This is consistent with our prior research showing that pain and disability outcomes improve little after three months [16]. Additionally, a review by Chou et al., which involved patients of all ages, found predictors of chronic disabling low back pain were similar for three, six, or 12 month outcomes [31].

Given the morbidity and incidence of falls in older adults [38], our exploratory finding that a history of falls was one of the strongest predictors of persistent disability and persistent pain is noteworthy. Moreover, the presence of musculoskeletal pain is an important risk factor for falls and recurrent falls among community-dwelling older adults [39,40]. Taking this knowledge into consideration along with our results implies that assessing falls history and falls-related risk factors in older adults with back pain may be important for determining risk for multiple important outcomes for older adults: falls, chronic disability, and chronic pain. Future studies of back pain in older adults should include a history of falls and associated falls risk factors in order to see whether our exploratory finding is robust.

The primary strength of our cohort study of older adults is its large size and one-year follow-up. By including only patients identified during a new primary care visit for back pain, our large inception cohort adds to the literature by providing some of the best estimates of risk factors for persistent back pain and persistent disability in older adults with a new episode of care for back pain. Another strength of this study is the excellent long-term follow-up, with roughly 85% of participants completing 12-month outcome surveys. Additionally, our use of inverse probability weights further minimizes the potential selection bias due to attrition.

An important limitation of this study is that some potentially important predictors of persistent back pain were not measured in the BOLD cohort. Prior studies suggest self-efficacy, passive coping, catastrophizing, and illness perception beliefs may often predict the development of persistent pain [33–35]. Additionally, potentially important predictors that are common in older adults such as frailty, gait speed, polypharmacy, and cognitive performance were not included in this study. These measures were not included in the baseline assessment for the BOLD cohort, and other studies will be needed to determine their association with persistent back pain in older adults. A second limitation is that not all participants underwent lumbar imaging. The substantially smaller sample size for the imaging analysis reduced the precision of our estimates, and for several imaging findings, especially those that were uncommon, we found some inconsistencies in outcomes based on imaging modality used (MRI vs CT for disc protrusion and facet arthropathy). Additionally, the subsample receiving lumbar imaging may be different from those who did not have these diagnostic procedures. We caution that these results may not be generalizable to all older adults with back pain as selection bias may exist.

A third limitation is relying on the use of ICD-9 codes for exact medical comorbidities. Absence of a diagnosis by ICD-9 coding does not preclude the possibility that a patient had a certain back pain or medical diagnosis, so misclassification of comorbid health conditions may be present. A fourth limitation is that residual selection bias may be present if unmeasured variables are associated with missing data and persistent disability or back pain. Another limitation is that we did not have centralized interpretations of the imaging—leading to unknown variation in the accuracy of reporting of imaging findings. Lastly, persistent disability and persistent back pain were common outcomes, and readers should be aware that odds ratios do not approximate relative risk with common outcomes.

We found that many predictors of persistent back pain and disability in older adults with a new visit for back pain are similar to those for younger populations, and these predictors were similar for a subgroup with acute back pain. Clinicians and researchers can now be more confident in the generalizability of these predictors to many older adults with new episodes of care for back pain. Most imaging findings were not associated with persistent disability and pain. As lumbar spine imaging likely provide little useful prognostic information, clinicians should be judicious regarding their use of imaging for older adults with back pain. Future research will need to further investigate novel prognostic factors that may be unique to older adults, whether there are statistical interactions with predictors such as age or gender, and the validity and clinical utility of screening tools for persistent back pain in older adults.

Appendix

Table A1

ICD-9 codes for determining the presence of selected comorbidities

Knee osteoarthritis	715.16	715.26	715.36	715.96	716.66
Hip osteoarthritis	715.15	715.25	715.35	715.95	716.65
Osteoporosis	733.00	733.02	733.03	733.09
Cervical pain	721.0	721.1	722.0	722.4	722.71	722.81	722.91	723.x	847.0
Fibromyalgia/chronic pain syndrome	338.2x	338.4	729.1

Table A2

ICD-9 codes for classification of Index Diagnostic Code Category

Axial pain	307.89, 721.3, 721.5, 721.9, 722.5, 722.51, 722.52, 722.6, 724.2, 724.5, 724.6, 739.3, 739.4, 846, 846.1, 846.8, 847.2, 847.3, 847.9
Back and leg pain	722.1, 722.2, 724.3, 724.4
Lumbar spinal stenosis	724, 724.01, 724.02, 724.03, 724.09
Other	733.1, 733.10, 733.13

Author notes

Funding sources: This work was supported by grants 1R01HS01922201 and 1R01HS022972-01 from the Agency for Healthcare Research and Quality.

Sean Rundell’s doctoral dissertation work was partially supported by a PhRMA Comparative Effectiveness Research Dissertation Award through the Pharmaceutical Outcomes Research and Policy Program at the University of Washington.

Disclosure and conflicts of interest: Dr. Jarvik has the following potential conflicts of interest; although they do not relate directly to the subject of this manuscript, he lists them in the spirit of full disclosure. He has served on the Comparative Effectiveness Advisory Board for GE Healthcare through October 2012. He is a cofounder and stockholder of PhysioSonics, a high intensity focused ultrasound company, and receives royalties for intellectual property. He serves as a consulting medical editor for Google as well as a consultant for HealthHelp, a radiology benefits management company. Finally, he is a co-Editor of Evidence-based Neuroradiology, published by Springer Publishing. All other authors have no conflicts of interest to declare.