Long-Term Medicare Costs associated with Opioid Analgesic Therapy vs. Spinal Manipulative Therapy for Chronic Low back Pain in a Cohort of Older Adults

Abstract

Objectives

The purpose of this study was to compare Medicare healthcare expenditures for patients who received long-term treatment of chronic low back pain (cLBP) with either opioid analgesic therapy (OAT) or spinal manipulative therapy (SMT).

Methods

We conducted a retrospective observational study using a cohort design for analysis of Medicare claims data. The study population included Medicare beneficiaries enrolled under Medicare Parts A, B, and D from 2012 through 2016. We assembled cohorts of patients who received long-term management of cLBP with OAT or SMT (such as delivered by chiropractic or osteopathic practitioners) and evaluated the comparative effect of OAT vs SMT upon expenditures, using multivariable regression to control for beneficiary characteristics and measures of health status, and propensity score weighting and binning to account for selection bias.

Results

The study sample totaled 28,160 subjects, of whom 77% initiated long-term care of cLBP with OAT, and 23% initiated care with SMT. For care of low back pain specifically, average long-term costs for patients who initiated care with OAT were 58% lower than those who initiated care with SMT. However, overall long-term healthcare expenditures under Medicare were 1.87 times higher for patients who initiated care via OAT as compared to those initiated care with SMT (95% CI 1.65-2.11; p <0.0001).

Conclusions

Adults aged 65-84 who initiated long-term treatment for cLBP via OAT incurred lower long-term costs for low back pain, but higher long-term total healthcare costs under Medicare as compared to patients who initiated long-term treatment with SMT.

INTRODUCTION

Burden of Low Back Pain Among Older Adults

The crisis of opioid prescribing in the Unites States (US) reflects in part a failure in the management of low back pain (LBP). Low back pain is highly prevalent in the US, and is the most common condition for which opioid analgesics are prescribed.¹ Low back pain is also one of the most common reasons for a physician visit in general,² and for older US adults in particular.³ The point prevalence of nonspecific back pain among older adults is approximately 30%.⁴ Older adults often suffer from recurrent episodes of LBP:⁵ Between 2000 and 2007 the total number of US adults with chronic low back pain (cLBP) increased by 64%.⁶ Spinal pain is the most costly of all medical conditions.⁷ A systematic review of LBP cost of illness studies that included 27 studies published between 1997 and 2007 found that direct costs ranged from $12 to $90 billion annually, and indirect costs ranged from $7 to $28 billion annually.⁸ Because spine care in the US has shown a decline in positive outcomes despite these rising costs,⁹ it is important to assess the comparative value of treatments for cLBP.

Opioid Analgesic Therapy for Low Back Pain

Evidence-based management of LBP for older adults often includes prescription analgesics, including opioids.¹⁰ A recent examination of health claims data found that 31% of patients with LBP received prescriptions for opioid analgesics within the first 6 months of initial diagnosis; within 3 years the percentage increased to 42%.¹¹ However, although opioid prescribing for LBP is often prolonged,¹² 3 systematic reviews concluded that the long-term effectiveness of opioid therapy for cLBP is unknown.^13,14,15 The hazards of opioid analgesics are well known however: the use and misuse of opioid analgesic therapy (OAT) has led to 3.8% of patients developing opioid use disorder,¹⁶ and up to 26% of prescribed patients reporting opioid dependence.¹⁷ In 2018 nearly 70% of US drug overdose deaths involved an opioid, and 46,802 people (128 per day on average) died from opioid-related overdoses.¹⁸ In 2013 the cost of prescription opioid misuse totaled more than $78 billion, and among 6,917 Medicare patients, the additional annual per patient cost associated with diagnosis of opioid abuse or dependence was over $17,000.¹⁹ Furthermore, among patients with a chronic disabling musculoskeletal disorder, higher dosing of opioids correlates with higher pain severity, greater disability, and higher levels of depression.²⁰ Recent studies suggest that OAT should be used sparingly and with caution for patients with cLBP due to the risk of adverse effects.^11,21

Spinal Manipulative Therapy for Low Back Pain

Non-pharmacological management of spinal pain is associated with decreased use of opioids,^22-25 and the National Academy of Medicine and the Joint Commission on Accreditation of Healthcare Organizations have recommended the use of non-pharmacological therapies as effective alternatives to pharmacotherapy for management of pain.^26,27 Recent systematic reviews have found spinal manipulative therapy (SMT) to be an effective treatment for cLBP,^28-30 and current clinical guidelines recommend non-pharmacological therapies, including SMT, as a first-line approach to management of LBP.³¹ Chiropractic management of LBP, which often involves treatment with SMT, may offer a safer alternative to analgesic medication: an observational study of adults with LBP found that the likelihood of adverse drug events was significantly lower for recipients of chiropractic care as compared to non-recipients.²⁰ Among older Medicare beneficiaries with an office visit risk for a neuromusculoskeletal problem, the adjusted risk of injury to the head, neck, or trunk among recipients of SMT was 76% lower as compared to recipients of primary care.³² More recently, a systematic review on the benefits and harms of SMT for treatment of cLBP found limited evidence regarding adverse events.³³ Thus, there is little cause for concern about the safety of SMT for LBP, but for long-term treatment of cLBP, the overall value of SMT remains uncertain. A series of US government reports based upon medical record reviews found that chiropractors, (who provide more than 97% of all SMT services under Medicare,³⁴) frequently provide excessive and unnecessary SMT, particularly as “maintenance care”, which is provided on an ongoing long-term basis, ostensibly to prevent spinal problems from recurring or worsening.^35-37 However, the government reports did not follow standardized review methods and failed to consider the value of chiropractic services.³⁸ It is also uncertain whether SMT offers a cost advantage for management of cLBP. A rigorously designed study that modelled the 1-year cost-effectiveness of adding nonpharmacologic interventions for cLBP to usual care analyzed data from randomized trials of nonpharmacologic therapies, and found that 15 of 17 therapies were cost effective from the payer perspective, but the results for chiropractic care were less favorable for patients with high-impact chronic pain than for a typical mix of patients.³⁹ A study of Medicare claims data found that patients with multiple comorbidities who only used SMT for cLBP had significantly lower costs,⁴⁰ and in a quasi-experimental study of over 84,000 older Medicare beneficiaries, Davis et al. found that increased access to chiropractic spinal manipulation was correlated with reduced spine-related costs.⁴¹ However, the comparative value of SMT for older adults with cLBP remains uncertain.

Cost of Care for Chronic Low Back Pain Under Medicare

Due to the high-cost burden imposed upon the Medicare system by care of patients with cLBP, it is critically important to identify high-value approaches to the treatment of cLBP. Until now, the long-term costs associated with OAT and SMT have not been rigorously compared. Therefore, the purpose of this study was to compare Medicare expenditures for cLBP among recipients of OAT vs. SMT. We hypothesized that among older Medicare beneficiaries receiving long-term care for cLBP, initiation of treatment via OAT would be associated with higher costs, as compared to SMT.

METHODS

Study Design and General Approach

Our approach to testing the hypothesis was to retrospectively examine outcomes in nationally representative samples of Medicare claims data, using a cohort design to evaluate the comparative costs of 2 approaches to management of cLBP. This study was conducted in the context of a multi-aim NIH-funded investigation of the comparative value of OAT vs. SMT for long-term care of Medicare beneficiaries with cLBP.

Study Population, Sampling, and Data Sources

The study population included community dwelling Medicare fee-for-service beneficiaries aged 65-84 years, residing in a US state or the District of Columbia, and continuously enrolled under Medicare Parts A, B, and D from 2012 through 2016, and alive on December 31, 2016. We restricted the population to subjects with an episode of cLBP beginning on a date of service in 2013 (index date) and defined as occurring with the recording of 2 paid claims with primary diagnosis of LBP at least 90 days but less than 180 days apart. Claims for diagnosis of LBP were identified by International Classification of Diseases (ICD-9) diagnosis codes, including 739xx series codes, which are commonly used by chiropractors to identify a manipulable condition of the lower back. Claims were restricted to outpatient office visits with clinicians specializing in General Practice, Family Practice, Internal Medicine, Osteopathic Manipulative Medicine, Physical Medicine and Rehabilitation, Chiropractic, Physical Therapist in Private Practice, or Pain Management. We excluded subjects diagnosed with cancer or in hospice care to avoid confounding by indication for use of opioids. Subjects over age 84 were excluded due to age-related reduction in the utilization of spinal manipulation.

Cohorts

We created 2 cohorts of patients: those who initiated long-term management of cLBP with either SMT or OAT. SMT was identified in claims data by current procedural terminology (CPT) code 98940, 98941 or 98942. Encounters were restricted to services provided in an outpatient setting. We defined long-term management by SMT as ≥12 office visits for spinal manipulation for LBP in any 12-month period, including at least one visit per month.⁴² OAT was identified through outpatient pharmacy prescription fills for opioid analgesics and analgesic formulations including opioids, as identified by drug code.⁴³ We defined long-term management by OAT as 6 or more standard 30-day supply prescription fills in a 12 month period.^14,44 The point in time of cohort accrual (index date) was the date of first office visit for an episode of cLBP. For subjects with more than one episode of cLBP, only the first was counted for purposes of cohort inclusion. Capture of subject data during the 12-month period before the index date (using 2012 and 2013 data) allowed exercise of sample inclusion and exclusion criteria, identification of chronic conditions, and measurement of comorbidity scores. Measurement of outcomes drew from 2013-2016 data, from the index date forward.

Subject Characteristics

We generated descriptive statistics on the demographic characteristics and health status of included subjects. Subject age was categorized as 65-69, 70-74, 75-79, and 80-84. Sex as a biological variable was collected as male or female. To meet data suppression rules, we aggregated race and ethnicity data to 2 categories: “White” and “Other / Unknown”. (The Centers for Medicare and Medicaid Services (CMS), requires suppression of data for cells with n<11 to prevent disclosure of protected health information.) As indicators of lower socioeconomic status, we used eligibility for the Part D low-income subsidy, and dual eligibility for both Medicare and Medicaid. Diagnostic codes for low back pain were categorized as non-specific LBP, radiculopathy, herniated disc, spondylolisthesis, sprain/strain, or spinal stenosis (739xx series codes, which are commonly used by chiropractors to identify a manipulable condition of the lower back were categorized as non-specific LBP). Charlson comorbidity scores were calculated based upon recorded diagnoses. As measures of health status, we calculated Charlson comorbidity scores, and collected data on diagnosis of comorbid chronic conditions (osteoarthritis of the hip or knee, which may confound the indication for opioids, and fibromyalgia and depressive disorder, which may affect prognosis for older adults with cLBP). For subjects who received OAT, we collected class of opioid prescribed at index.

Outcomes Measurement and Statistical Analysis

We sought to compare the impact of OAT vs SMT as initial choice of treatment for long-term care of cLBP. Our rationale for taking this approach was that previous studies on management of LBP have found that initial or early treatment with chiropractic care results in fewer back surgeries,⁴⁵ lower costs,⁴⁶ and less use of prescription opioids.²⁵ We generated descriptive statistics on the demographic characteristics and health status of included subjects. To measure cumulative cost outcomes from index date through 2016, we measured allowed charges (the dollar amount set by Medicare as payment in full, which typically includes patient cost-sharing), and payments (the dollar amount reimbursed by Medicare). We analyzed FFS charges and payment data for outpatient, inpatient, and pharmacy claims. Part D cost measures included both the amount paid by the prescription drug plan and the beneficiary out of pocket payment. We first analyzed for costs of analgesic medications and costs associated with encounters for care of LBP. Next, to evaluate for any potential cost offset effects that might be associated with utilization of SMT as compared to OAT, we measured overall expenditures, in which clinical claims were not restricted to diagnosis of LBP, and pharmacy fills were not restricted to analgesics. We conducted a comparison of overall costs between cohorts OAT and SMT using propensity score methods in combination with multivariable regression (ie, doubly robust). The multivariable model for the dependence of the outcome on group controlled for beneficiary characteristics and measures of health status, including age, sex, race, state of residence, Charlson comorbidity index, LBP diagnostic category, and the presence of specific comorbidities such as hip or knee arthritis, depression, or fibromyalgia. The propensity model for the exposure (group) included the same covariates. In the weighted propensity approach, we estimated the treatment effect using the multivariable outcome model with weighting by the reciprocal of the propensity. In the binned propensity approach, we included deciles of the propensities as a covariate in the multivariable outcome model. All statistical analyses were performed using SAS (SAS Institute, Cary, NC).

RESULTS

Table 1 displays demographic characteristics and measures of health status by cohort. The overall study sample totaled 28,160 subjects, of whom 21,731 (77%) initiated long-term care of cLBP with OAT, and 6,429 (23%) who initially chose SMT. Average subject age ranged between 72.6 and 73.1, and approximately two-thirds of subjects were 74 years of age or younger. Females outnumbered males by nearly 3:1, and those who identified as White greatly outnumbered other racial/ethnic groups. The proportions of subjects with indicators of lower socioeconomic status were higher among patients who received OAT. Comorbidity scores trended higher among patients who received OAT. In both cohorts most patients were diagnosed with non-specific LBP. There were higher proportions of subjects diagnosed with radiculopathy, spondylolisthesis and spinal stenosis among patients who received OAT, as compared to those who received SMT. The frequency of cases of herniated disc was low, and so infrequent in the SMT cohort that data suppression was required. Diagnosis of sprain/strain also occurred infrequently. Higher proportions of patients in the OAT cohort were diagnosed with depressive disorder and osteoarthritis of the hip or knee, as compared to the SMT cohort; fibromyalgia was more frequently diagnosed in the SMT cohort. Most patients (81-82%) who chose OAT were prescribed a schedule 2 opioid; approximately 3% were prescribed a schedule 3 drug, and 15% received a schedule 4 drug.

Table 1.

Patient Characteristics and Health Status

Cohort	SMT		OAT
N	6,429		21,731		P-Value
Age at cohort accrual (years)	Mean	SD	Mean	SD
	73.1	4.8	72.6	4.8	<0.0001
Age Category (years)	n	%	n	%
65-69	2,060	32.00%	8,071	37.10%
70-74	2,216	34.50%	7,070	32.50%
75-79	1,427	22.20%	4,426	20.40%
80-84	726	11.30%	2,164	10.00%	<0.0001
Sex
Male	1,795	27.90%	5,734	26.40%
Female	4,634	72.10%	15,997	73.60%	0.0153
Race/Ethnicity
White	6,127	95.30%	18,327	84.30%
Other / Unknown	302	4.70%	3,404	15.70%	<0.0001
Socioeconomic Status
Low-income subsidy	827	12.90%	10,397	47.80%	<0.0001
Dual eligibility	679	10.60%	8,971	41.30%	<0.0001
Charlson Comorbidity Score
0	3,918	60.90%	9,184	42.30%
1	1,554	24.20%	6,343	29.20%
2	622	9.70%	3,505	16.10%
3	227	3.50%	1,532	7.00%
4+	108	1.70%	1,167	5.40%	<0.0001
Chronic Condition
Osteoarthritis of the Hip	223	3.50%	1,344	6.20%	<0.0001
Osteoarthritis of the Knee	732	11.40%	3,551	16.30%	<0.0001
Fibromyalgia	2,004	31.20%	6,273	28.90%	0.0054
Depressive Disorder	723	11.20%	6,268	28.80%	<0.0001
LBP Diagnosis Category
Non-Specific Low Back Pain	6,134	95.40%	16,376	75.40%	<0.0001
Radiculopathy	192	3.00%	3,307	15.20%	<0.0001
Herniated Disc	*	*	44	0.20%	0.0005
Spondylolisthesis	11	0.20%	276	1.30%	<0.0001
Sprain/Strain	30	0.50%	130	0.60%	0.0485
Spinal Stenosis	59	0.90%	1,598	7.40%	<0.0001

Notes: SMT = subjects who initiated SMT in 2013 for long-term management of cLBP; OAT = subjects who initiated OAT in 2013 for long-term management of cLBP; n = number of subjects; n/a = not applicable

^*data suppressed in accordance with CMS rules

Table 2 displays average costs per beneficiary by cohort. Predictably, costs for analgesics were higher for those who chose long-term treatment by OAT. For encounters specifically associated with primary diagnosis of LBP, average costs for patients who received long-term care via OAT were 58% lower than those who received long-term SMT. However, for overall healthcare costs under Medicare, the cost pattern was reversed. In terms of overall costs, patients who received long-term treatment via OAT, both allowed charges and payments were higher for Parts A, B, and D as compared to SMT. Over the 4-year study period, for management of cLBP specifically, per-beneficiary Medicare payments were $4,693 lower for patients who received OAT versus SMT (a 58% savings); the average annual per patient cost was $1,173 lower. However, regarding total healthcare costs to Medicare for the study period, OAT was associated with payments that were $15,717 higher for Part A, $20,928 for Part B payments, and $7,877 for Part D payments compared to SMT. Patients who initiated care with OAT incurred total adjusted Medicare payments 1.87 times higher than patients who initiated care with SMT (Cost Ratio 1.87; 95% CI 1.65-2.11; SE 0.06; p <0.0001). (Table 3)

Table 2:

Average Costs per Beneficiary ($) 2013-2016

	Total			Annual
Cost Measure	Mean (SD)	Mean (SD)	Difference	Mean		Difference	% Difference
	SMT	OAT	OAT-SMT	SMT	OAT	OAT-SMT	OAT vs. SMT
Low Back Pain
Allowed Charges (1))	11,133 (15,877)	4,627 (8,451)	−6,506	2,783	1,157	−1,626	−58%
Medicare Payment	8,155 (12,171)	3,462 (6,558)	−4,693	2,039	866	−1,173	−58%
Analgesics
Total Charges (2)	394 (1,674)	2,740 (6,780)	2,346	99	685	586	595%
Medicare Payment	244 (1,330)	2,067 (6,210)	1,823	61	517	456	747%
Beneficiary Cost Share	151 (513)	673 (1,244)	522	38	168	130	346%
Overall
Allowed Charges - Part A	46,704 (108,288)	117,827 (199,181)	71,123	11,676	29,457	17,781	152%
Allowed Charges - Part B	47,030 (70,127)	71,243 (81,532)	24,213	11,758	17,811	6,053	51%
Beneficiary Cost Share - Part D	2,494 (2,874)	2,864 (3,540)	370	624	716	92	15%
Payments - Part A	12,337 (26,823)	28,054 (44,959)	15,717	3,084	7,014	3,930	127%
Payments - Part B	36,812 (55,465)	57,740 (65,550)	20,928	9,203	14,435	5,232	57%
Payments - Part D	6,386 (16,419)	14,263 (25,748)	7,877	1,597	3,566	1,969	123%

Notes: SMT = subjects who initiated SMT in 2013 for long-term management of cLBP; OAT = subjects who initiated OAT in 2013 for long-term management of cLBP; SD = standard deviation; vs. = as compared to

Table 3.

Initiation of Care for cLBP with OAT vs. SMT: Comparison of Medicare Costs

	Propensity Scoring Method	Cost Ratio	CI	SE	P
Charges	Binned	2.03	1.96 - 2.10	0.02	<.0001
	Weighted	2.19	1.93 - 2.49	0.07	<.0001
Payments	Binned	1.73	1.68 - 1.79	0.02	<.0001
	Weighted	1.87	1.65 - 2.11	0.06	<.0001

OAT – opioid analgesic therapy; SMT = spinal manipulative therapy; Estimate = exponentiated log; CI = confidence interval; SE = standard error; P = probability; Binned = binned propensity scoring controlling for categorical variable; Weighted = inverse probability of treatment weighting; Charges = Average total Medicare charges per beneficiary; Payments = Average total Medicare payments per beneficiary

DISCUSSION

We hypothesized that among older Medicare beneficiaries receiving long-term care for cLBP, treatment via OAT would be associated with higher costs, as compared to SMT. The results support our hypothesis regarding long-term overall costs to Medicare, but not for long-term costs incurred specifically for clinical care of cLBP. Because per visit costs for SMT are relatively low as compared to many other services provided under Medicare, the higher costs for management of cLBP with SMT may be due at least in part to the higher number of office visits that successful treatment by SMT often requires. However, Weeks et al. found that treatment of Medicare patients with SMT alone resulted in lower per episode costs for cLBP as compared to conventional medical care.⁴⁰ Our findings suggest that the per-episode cost advantage of SMT may have been lost in the population of patients who require long-term ongoing treatment. However, the higher costs of SMT for LBP specifically were evidently offset by lower long-term costs in other areas, resulting in lower costs to Medicare overall. Weeks et al. also reported lower overall costs to Medicare for older adults who received SMT for cLBP.⁴⁰ We speculate that the cost offset effect may be due at least in part to savings resulting from quaternary prevention of complications associated with opioid use. In 2013 the estimated annual per patient cost of managing prescription opioid misuse was $17,052 (95% CI $13,472 - $20,632).^14,19 In this study, among all subjects with cLBP, 77% initiated care with opioids, while only 23% started with spinal manipulation. Thus, there may be significant potential for long-term cost savings under Medicare through increased utilization of SMT as an alternative to OAT for cLBP. Further research is needed to identify the sources of potential cost savings associated with use of SMT. Although we analyzed by initial choice of treatment, relatively few patients crossed over (3% of subjects crossed over from OAT to SMT, and 5% crossed over from SMT to OAT). Thus, costs may have been driven as much or more by the treatment used for long-term care as by the initial choice of treatment.

Limitations

General limitations of using health claims data for observational research include inconsistencies in billing practices and coding of procedures and diagnoses, lack of diagnoses in prescription drug claims, and lack of specific clinical findings. In addition, this study may have been limited by lack of a pain severity indicator other than diagnostic category. The presence of more severe and comorbid conditions in the OAT cohort raises the possibility of confounding by indication, but we controlled for osteoarthritis of the knee and hip that may have prompted opioid prescriptions, and for fibromyalgia and depressive disorder, which can impact prognosis. We also excluded subjects with a diagnosis of cancer or use of hospice care to reduce confounding by indication among recipients of OAT. We employed robust methods intended to minimize the impact of selection bias that may result from a predisposition among recipients of spinal manipulation to avoid the use of prescription drugs and for patients with more severe disease to be treated with opioids. It can be difficult to distinguish between continuous pain and recurrent episodes of acute pain, but our definition of cLBP tried to ensure that the cases were ongoing rather than episodic. Despite these measures, we were unable to consider all confounding variables, the inherent limitations of observational design inhibit causal inference, and explicit assessment of changes in underlying health status were not possible. However, analysis of large multi-year claims datasets allows cost-efficient evaluation of long-term outcomes.

Conclusion

Adults aged 65-84 who received long-term treatment for cLBP via OAT incurred lower long-term costs for LBP, but higher long-term total healthcare costs under Medicare as compared to patients who received long-term treatment with SMT.