THE RELATIONSHIP BETWEEN SOCIOECONOMIC STATUS AND THE OUTCOME OF LUMBAR EPIDURAL STEROID INJECTIONS FOR LUMBAR RADICULOPATHY

Abstract

OBJECTIVE:

Investigate the relationship between socioeconomic status (SES) and pain reduction from epidural steroid injections (ESIs) for lumbar radiculopathy.

METHODS:

The retrospective cohort consisted of patients undergoing ESI for lumbar radiculopathy (n=544). Numeric pain rating scale (NPRS) was measured at baseline and two weeks post-ESI. SES was estimated using median family income in patients’ ZIP code. Linear and mixed models examined demographic and clinical differences in pain pre- and post-injection and whether family income moderated the effect.

RESULTS:

Majority of patients were white (72.4%), female (56.4%), engaged in physical activity (68.2%), and underwent unilateral, transforaminal ESI (86.0%, and 92.1%, respectively). Non-white patients and those who did not engage in physical activity had higher baseline pain (p<0.05)). Lower SES was associated with higher baseline pain (β=0.06 per $10,000p=0.01). Patients with lower SES experienced larger improvement in pain following ESI:-1.56 units for patients in the 10^th percentile of family income versus −0.81for 90^th percentile. Being a current smoker was associated with higher pain (β=0.76 p=0.03) and engaging in structured physical activity less pain (β=-0.07 p<0.01).

CONCLUSIONS:

Lower SES was independently associated with higher pain alleviation after controlling for other potentially influential demographics. Modifiable lifestyle factors maybe a target of potential intervention.

INTRODUCTION

Low back pain is the third most common reason for primary care visits in the United States and the leading global cause of years lost to disability.¹ When conservative management fails, interventional treatments such as epidural steroid injections (ESI) are often considered. It is estimated that 9 million ESIs are performed yearly in the United States for low back pain and radiculopathy.² However, recent research has called their effectiveness into question for certain spine conditions. Most notably, a randomized trial of ESIs for neurogenic claudication and central canal stenosis concluded that injection of glucocorticoid with lidocaine offered minimal or no benefit compared to injection with lidocaine alone.³ Those results contrast with prior studies in individuals with low back pain and radiculopathy which have shown benefit.⁴ In general ESIs have varying degrees of evidence in support (or against) dependent on different pathologies. Controversy over the use of ESIs and variable patient outcome has fueled several studies investigating potential mediators of injection outcomes and identifying patients who are most likely to benefit from specific interventional approaches has become a priority for the low back pain research community.⁵

Recent evidence suggests the application of a biopsychosocial model is required to fully understand the complexity of this public health burden.⁶ Specifically, chronic low back pain has been shown to disproportionately affect people with a lower socioeconomic status (SES).⁷ The reasons for this are unclear and likely multifactorial, but include disparities in environmental exposures, income, health literacy, and access to health care.⁸

The absence of worsening pain with walking, a shorter duration of pre-injection symptoms, a positive femoral stretch test, and the absence of pain with lumbar extension have been suggested to be predictors of favorable ESI outcomes.⁹ Suggested unfavorable psychological mediators of pain have also been identified, such as disturbed mood, negative self-efficacy, high anxiety, and poor mental health.¹⁰ Socioeconomic factors influencing outcomes are not well described, and variables that may be related to SES, such as medical insurance, employment status and family income have not been reliably studied in patients undergoing lumbar ESI.¹¹

Prior orthopedic literature has revealed significant associations between SES and post-surgical improvement in function, reoperation rates, and healthcare resource utilization in procedures such as total knee/hip arthroplasty and surgical correction of lumbar stenosis.¹² Examining this association in lumbar ESIs for radicular pain may help clarify the contribution of SES to the variability in outcome. The primary hypothesis was that SES (using estimated family income) is independently associated with the magnitude of pain reduction at 2-weeks following ESI in patients with radicular pain. Secondary analyses involved evaluating the relationship between SES and baseline pain.

METHODS

The study was approved by the Institutional Review Board and was granted a waiver of informed consent. The setting was an musculoskeletal outpatient clinic within an academic free-standing rehabilitation hospital. All ESI procedures had been performed between December 31^st, 2013 and December January 1^st, 2017.

PATIENT COHORT

This study was a retrospective cohort study of patients who underwent an ESI for lumbar radiculopathy with or without low back pain. Patients aged ≥18 years with symptomatic radicular pain or radiculopathy who underwent lumbar ESI during the study dates were included. Patients with a history of lumbar spine surgery or vertebral fracture or with missing pain score (pre- or post-ESI) documented were excluded. In patients who underwent multiple ESIs, only the first injection performed at the institution was included. All ESIs were performed by one of five board certified spine physiatrists under fluoroscopic guidance using a standardized transforaminal (TFESI) or interlaminar (ILE) epidural approach, with injection of 1–2 ml of 1% lidocaine around the spinal nerve root, followed by 1 ml of dexamethasone (10mg/ml).

PAIN RESPONSE-RELATED VARIABLES

The potential pain response-related variables assessed included demographic (age, sex, BMI, race, marital status), medical/social history (psychotropic medications, smoking, alcohol intake), socioeconomic factors (employment status, insurance status, estimated family income) and physical activity participation (structured included running, strength training or other sporting activity and unstructured included walking only). The type and location of injection (bilateral or unilateral, transforaminal epidural or interlaminar epidural injections, and single or multi-level) was also assessed. The principal measure of SES used was estimated family income. This was estimated by geocoding patients’ ZIP code to the U.S. Census Bureau 2011–2015 American Community Survey 5-Year estimate of median family income. Using zip code as a proxy has been shown to be the most consistent way to monitor SES differences in health outcomes.¹³

PRIMARY OUTCOME MEASURE

The primary outcome measure was radicular pain scores measured at baseline and two-weeks post-injection using the verbal 11-point (0–10) numeric pain rating scale (NPRS). The NPRS is a segmented numeric version of the visual analog scale in which a respondent selects a whole number (0–10) that best reflects the intensity of their pain and is validated for musculoskeletal pain.¹⁴ The NPRS was anchored on the left with the phrase “No Pain” and on the right with the phrase ‘Worst Imaginable Pain.’

STATISTICAL ANALYSIS

After assessing the distribution of NPRS pain scores, it was determined that they were suitable for modeling by linear regression. Demographic characteristics were identified and associated with pre-injection pain scores using univariate models, then included all covariates with a univariate strength of association p ≤ 0.2 in a multivariable model to report adjusted associations. Mixed models were then used with a random effect for person and a fixed effect for time to examine: (1) the change in pain scores from pre-injection to 2 weeks post injection; and (2) whether estimated family income moderated the change in pain scores over time. An interaction was included between estimated family income and time to assess the moderation effect. Demographic characteristics identified above were included as fixed effects. Logistic regression was used to examine the association between dropout and baseline demographic and clinical characteristics. Unless otherwise noted, we used α=0.05 to assess statistical significance. Analyses were conducted using Stata, version 14 (College Station, TX) and R version 3.5.2 (Vienna, Austria), along with StatTag.¹⁵

RESULTS

A total of 1105 ESIs were performed on 780 patients from January 2014 through December 2016. After excluding 325 repeat injections and 236 patients for whom two-week follow-up data were unavailable, we analyzed pain scores among 544 patients. Patients with two week follow-up data were older (58.9 years versus 55.4 years, 95% CI: 0.95, 6.0, p=0.03 years) and more likely to be physically active (68.2% versus 57.6%, 95% CI: 2.8%, 18.3% p<0.01).

Baseline patient characteristics (Table 1)

Table 1:

Patient demographics and characteristics (N=544)

CHARACETERISTIC	VARIABLE	OVERALL
AGE (MEAN (SD))		58.9 (16.4)
BMI (MEAN (SD))		27.8 (5.5)
SEX, n (%)	Female	307 (56.4)
	Male	237 (43.6)
RACE, n (%)	White	394 (72.4)
	Black	39 (7.2)
	Hispanic	17 (3.1)
	Asian	9 (1.7)
	Native American	1 (0.2)
	Other/Unknown	84 (15.4)
MARITAL STATUS, n (%)	Married	312 (57.4)
	Domestic Partner	5 (0.9)
	Single	138 (25.4)
	Widowed	37 (6.8)
	Divorced	26 (4.8)
	Unknown	26 (4.8)
EMPLOYMENT STATUS, n (%)	Full-time	235 (43.2)
	Part-time	16 (2.9)
	Not Employed	45 (8.3)
	Retired	192 (35.3)
	Student	2 (0.4)
	Unknown	54 (9.9)
INSURANCE STATUS n (%)	Commercial	206 (37.9)
	Medicare	184 (33.8)
	Managed Care	101 (18.6)
	Medicaid	40 (7.4)
	Worker’s Compensation	13 (2.4)
MEDIAN FAMILY INCOME [IQR]		118888.50 [67405.25, 150960.00]
SMOKING STATUS n (%)	Current	36 (6.6)
	Never	309 (56.8)
	Quit	198 (36.4)
	Unknown	1 (0.2)
ALCOHOLIC BEVERAGES PER WEEK		1.5 [0.00–4.00]
PHYSICAL ACTIVITY n (%)	Structured	187 (34.4)
	Unstructured	184 (33.8)
	None	172 (31.6)
	Unknown	1 (0.2)
PSYCHOTROPIC MEDICATIONS n (%)	Yes	212 (39.0)
	No	325 (59.7)
	Unknown	7 (1.3)
BASELINE PAIN SCORE, Median [IQR]		6.00 [4.00, 8.00]

Average patient age was 58.9 years (SD16.4 years) and BMI was 27.8 (SD5.5). Just over half of the patients were female (n=307, 56.4%), married (n=312, 57.4%), and never smoked (n=309, 56.8%). The majority of patients were white (n=394, 72.4%) and engaged in some form of physical activity (n=371, 68.2%). Median estimated family income was $118,888.50 (IQR $67,405.25– $150,960.00). The majority of patients had commercial or Medicare insurance (71.7%). The median pain score pre-injection was 6 points (IQR 4–8) with the majority undergoing a unilateral (86.0%), single level (99.1%) and transforaminal ESI (92.1%).

Relationship between patient characteristics and baseline pain (Table 2)

Table 2:

Adjusted associations between pre-injection pain scores and demographic and baseline characteristics (n=544)

VARIABLE	Adjusted Beta	(95% CI)		p-value
		LL	UL
MEDIAN FAMILY INCOME (per 10K)	−0.06	−0.11	−0.02	0.01*
AGE (years)	0.007	−0.01	0.02	0.39
SEX Female	REF.
Male	−0.17	−0.57	0.23	0.41
RACE White	REF.
Non-White	0.66	0.19	1.13	0.01*
MARITAL STATUS Married/Domestic Partner	REF.
Divorced/Single/Widowed/Unknown	−0.14	−0.55	0.26	0.49
INSURANCE STATUS Medicaid	REF.
Commercial	−0.49	−1.31	0.33	0.24
Medicare	−0.15	−0.95	0.66	0.73
Managed Care	−0.51	−1.41	0.38	0.26
Worker’s Compensation	−0.27	−1.71	1.18	0.72
SMOKING STATUS Never	REF.
Current	0.68	−0.11	1.47	0.09
Quit	−0.03	−0.46,	0.39	0.87
ALCOHOLIC BEVERAGES PER WEEK	−0.06	−0.12,	0.00	0.07
PHYSICAL ACTIVITY No Physcial Activity	REF.
Structured Activity	−0.70	−1.20,	−0.19	0.01*
Unstructured Activity	−0.59	−1.07,	−0.10	0.02*
PSYCHOTROPIC MEDICATIONS No	REF.
Yes	0.31	−0.09,	0.71	0.13
Unknown	−0.25	−1.95,	1.45	0.77
CONSTANT	6.78	5.43,	8.13	<0.01

Sex, race, physical activity, marital status, insurance status, smoking, psychotropic medications, and alcoholic beverages consumed per week were all individually associated with pre-injection pain score (p≤0.2). Table 2 shows the adjusted associations between these demographic characteristics and pre-injection pain scores. Patients with lower estimated family income presented with higher pain scores on average (0.06 per every $10,000 less in estimated family income). Non-white patients presented with pain scores over a half point higher on the NPRS (β=0.66, 95%CI: 0.19–1.13) than white patients. Not engaging in physical activity was associated with a baseline pain score that was higher by half a point, compared with patients with structured or unstructured physical activity respectively which were significantly lower (β=-0.70, 95%CI: −1.20, −0.19 p=0.01; β=-0.59, 95%CI: −1.07, −0.10, p=0.02).

Relationship between baseline patient characteristics and response to ESI (Table 3)

Table 3:

Association of pain scores with treatment, demographic and baseline characteristics (n=544)^†

VARIABLE	Adjusted Beta	(95% CI)		p-value
		LL	UL
TREATMENT EFFECT BY MEDIAN FAMILY INCOME ‡
• 10th percentile	−1.56	−1.91	−1.22	<0.01**
• 25th percentile	−1.46	−1.75	−1.17	<0.01**
• 50th percentile	−1.14	−1.35	−0.94	<0.01**
• 75th percentile	−0.92	−1.18	−0.65	<0.01**
• 90th percentile	−0.81	−1.13	−0.49	<0.01**
AGE (YEARS)	0.005	−0.010	0.02	0.51
SEX Female	REF.
Male	−0.23	−0.59	0.13	0.21
RACE White	REF.
Non-White	0.41	−0.01	0.83	0.06
MARITAL STATUS Married/Domestic Partner	REF.
Divorced/Single/Widowed/Unknown	−0.02	−0.38	0.34	0.92
INSURANCE STATUS Medicaid	REF.
Commercial	−0.59	−1.32	0.15	0.12
Medicare	−0.19	−0.92	0.53	0.60
Managed Care	−0.49	−1.29	0.31	0.23
Worker’s Compensation	−0.01	−1.30	1.28	0.99
SMOKING STATUS Never	REF.
Current	0.76	0.06	1.47	0.03*
Quit	−0.14	−0.52	0.24	0.46
ALCOHOLIC BEVERAGES/WEEK	−0.05	−0.11	0.00	0.05
PHYSICAL ACTIVITY None	REF.
Structured	−0.70	−1.15	−0.25	<0.01**
Unstructured	−0.25	−0.69	0.19	0.26
PSYCHOTROPIC MEDICATIONS No	REF.
Yes	0.27	−0.10	0.63	0.15
Unknown	0.12	−1.40	1.65	0.88
ESI LATERALITY Unilateral	REF.
Bilateral	0.41	−0.11	0.93	0.12
ESI PROCEDURE Transforaminal	REF.
Interlaminar	0.03	−0.063	0.68	0.94
CONSTANT	6.96	5.72	8.20	<0.01

^†Estimated from a linear mixed model with pain scores as the dependent variable and a random effect for participant. Adjusted betas indicate the association between demographic and baseline characteristics and pain scores overall. For example, compared with patients who never smoked, current smokers score an average of 0.76 units higher (95% CI 0.06–1.47). Average reduction in pain score from baseline to 2 weeks post injection by median family income. For example, for patients in the 25^th percentile of median family income, pain scores are reduced by an average of 1.46 units (95% CI 1.17–1.75) from baseline to 2 weeks post injection.

^‡Average reduction in pain score from baseline to 2 weeks post injection by median family income. This is expressed as percentile of median family income.

^*p<0.05

Pain scores were significantly reduced over the two-week period, but the amount of reduction depended on estimated family income. Figure 1 illustrates pain scores pre-injection and two weeks later by percentiles of estimated family income. Prior to injection, average pain score for a patient in the 25^th percentile of estimated family income was 6.1 (5.8, 6.4), compared with an average pain score of 5.5 (5.2, 5.7) for a patient in the 75^th percentile of estimated family income. However, two weeks after injection, their pain scores were closer: 4.6 (4.4, 5.0) and 4.5 (4.3, 4.8), respectively. This is further illustrated in Figure 2, which shows that ESIs resulted in a larger reduction in pain scores for patients with lower estimated family income. Table 3 shows estimated reductions in pain scores by percentile of family income. For example, patients in the 10^th percentile for median family income had reduction in pain scores by 1.56 points on the NPRS (95%CI: −1.91, −1.22, p<0.01), whereas those in the 90^th percentile only had a 1.13-point reduction in pain (95%CI: −1.13, −0.49, p<0.01) on the NPRS.

Figure 1:

Average pain scores pre− and post−injection by annual family income

Figure 2:

Average reduction in pain score by annual family income

Even after adjusting for treatment effect moderated by family income, several baseline characteristics remained associated with pain scores (Table 3). Patients who were current smokers had higher pain scores overall (β=0.76, 95%CI: 0.06,1.47, p=0.03), compared with patients who never smoked. In addition, compared with patients who were not physically active, patients who were physically active engaging in structured activity – also had lower pain scores overall (β=-0.70 95%CI: −1.15, 0.25, p<0.01)]. However, after adjusting, pain scores did not differ significantly by sex, race, or age.

DISCUSSION

This study was designed to assess the relationship between SES (measured by estimated median family income) and the outcome of ESIs. In keeping with the hypothesis, the study did find that median family income had an independent association with change in pain score following ESI. Specifically, lower SES was associated with higher baseline pain but a larger response to ESI at 2 weeks. Median family income by zip code was chosen as the measure of SES as it is the most commonly used income measure for the classification of SES and has been described as the most consistent way to monitor SES differences in health outcomes.¹³ To the author’s knowledge there are no prior studies investigating the relationship between SES and spine injection-related outcomes, though there are similar findings in other studies evaluating this relationship in chronic musculoskeletal pain and following orthopedic surgical procedures.¹⁶ For example, multiple studies have shown that family income does have an association with deprivation and patient reported outcomes following total joint arthroplasty.¹⁷ A prior study by Barrack et al¹⁸, showed that individuals who had undergone total knee arthroplasty and had a median family income less than $25,000 per year were more likely to report less satisfaction and functional ability post-surgery. Psychological stressors, race, lower educational background and being a Medicaid recipient have also been significant predictors for severe musculoskeletal pain in an older population.¹⁹ In a study evaluating radiographic knee osteoarthritis and pain, lower individual and community SES was also associated with worse function and pain.²⁰ Similarly Dorner et al²¹ found that individuals in the lowest SES group when compared to the highest were two to three times more likely to feel disabled by pain in multiple body sites.

The finding in this study that lower SES had a better response to ESI suggests the potential benefits of these procedures in this population. Conversely, higher SES was associated with a lesser response to lumbar ESI may in part be due to the patient’s pre-injection status. i.e. individuals with lower SES in general had greater baseline pain. This is supported by a recent study that demonstrated that the most socioeconomically disadvantaged individuals with symptomatic lumbar disc herniations present with the worst functional limitations and pain levels.²² This may in part due to the fact that patients from higher socioeconomically advantaged communities potentially have greater access to and utilization of conservative care such as physical therapy and pain management services leading to less baseline pain. It is important to note that the statistical differences in pain response to injections between percentiles of median household income may not necessarily represent clinical significance.

The present study also found that being of non-white race was associated with higher baseline pain. There is recent evidence that analgesic medication prescription habits and provision of work accommodations by physicians for chronic pain has some potential relationship with the race of the patient.²³ There are other known differences in medical care and access across socioeconomic statuses, which likely impact patient outcomes. For example, insurance status itself has been shown to be associated with pain. A study by Drazin et al²⁴ showed that insurance status does impact post-surgical outcomes for lumbar spinals stenosis. It is important to note that there is significant variability in insurance coverage for rehabilitative services across private insurers, Medicare, and Medicaid.²⁵ For example, Medicaid beneficiaries in some states are entitled to only a single therapy evaluation between physical, occupational, and speech therapy. These restrictions on visits have resulted in outpatient therapy practices not accepting Medicaid patients.¹⁶ This may impact patient orthopedic outcomes, for example Sabesan and colleagues¹⁷ found that patients on Medicaid had a higher risk of complications after treatment of proximal humeral fractures. In the present cohort, insurance status was not found to be associated with ESI outcome, however there was an overall low proportion of individuals on Medicaid insurance which may have impacted this finding. In the present study non-white race was associated with higher baseline pain but was not found to be associated with response to ESI when adjusting for SES. Prior surgical studies however have shown that non-white race does predict worse outcome after ACL reconstruction.²⁶

Secondary analysis revealed other more modifiable lifestyle factors that were associated with higher baseline pain and also response to ESI. In the present study, a current smoker in particular was associated with higher baseline pain. This is in keeping with prior studies that have shown that smoking is associated with the presence of low back pain²⁷ and increased pack-years of smoking has a positive association with the musculoskeletal pain intensity of an individual.²⁸ Physical activity participation also had a relationship to baseline pain and the outcome of the ESIs after accounting for median family income. Those who engaged in structured physical activity in particular, had lower baseline pain. A recent large meta-analysis had similar findings in that physical activity level was associated with a lower prevalence of low back pain.²⁹ The present study findings that after adjusting for median family income, engaging in structured physical activity is associated with outcome of an ESI is of particular importance for a spine proceduralist, as it suggests that encouraging physical activity participation pre-injection may improve outcome. Although not assessed in this study, it is known that physical activity participation does have a relationship with SES. The reasons for this are likely multi-factorial, but it is known that individuals with higher SES have fewer neighborhood safety concerns, for physical activity participation, contributing to overall physical and mental health.³⁰

A major strength of the study is that it has a large sample size and is from a single center, limiting variability in treatment and procedure technique. However, this study is limited by the retrospective observational design, the number of patients with missing chart data, and limited information available on prior treatments the patient’s may have received. We also do not have information on prior or on-going treatments such as physical therapy for lumbar radiculopathy which may have also impacted outcome. It is important to note that the NPRS used in the present study, provides a mono-dimensional evaluation of pain and does not account for the multi-dimensional aspect of the pain experience or function which would be better addressed in a prospective study. Similarly, the use of zip code to estimate family income (which was the primary indices for SES used) may not adequately capture all aspects of SES that influence outcomes, particularly at the individual level. The follow-up evaluation of these patients was also short (2 weeks). This timeline represents the follow-up timeline patients present at our clinic and is classically the optimal time at which patients feel improvement in their symptoms following an ESI. However future studies should evaluate more longer-term pain trajectories. Finally, this single center study was in a population with an estimated median family income which is almost double the national and a prospective study with a wider range of incomes could be more generalizable. Future studies should examine the impact of co-morbidities such as affective disorders, metabolic syndrome, and diabetes on the outcome. Although the present study did not find any relationship between being on a psychotropic medication and pain, a prospective study could assess the psychological state and expectations of an individual pre-injection since this may also impact ESI outcome.

CONCLUSION

The present study investigated the relationship between an individual’s SES, (using estimated family income) and the outcome of ESIs for radicular pain. The pivotal finding in this study is that in a large cohort of patients undergoing lumbar ESI for radicular pain, higher SES was independently associated with less pain alleviation after controlling for other potentially influential demographic factors. Most of the variability in patients’ response to the ESI procedure was not purely explained in the multivariate model by SES, which is consistent with the complex nature of chronic pain symptomatology. Engagement in physical activity and the smoking status appear to be significant factors tied to the severity of radicular pain.

What is Known?

Epidural spinal injections are a common treatment prescribed for individuals with lumbar radicular symptoms, however their ability to improve pain can be variable.

What is New?

This retrospective chart review found that at 2 weeks post-epidural, lower socioeconomic status was independently associated with higher pain alleviation after controlling for other potentially influential demographic and clinical characteristics.