Associations Between Early Chiropractic Care and Physical Therapy on Subsequent Opioid Use Among Persons With Low Back Pain in Arkansas

Abstract

Objective

The objective of this study was to estimate the association between early use of physical therapy (PT) or chiropractic care and incident opioid use and long-term opioid use in individuals with a low back pain (LBP) diagnosis.

Methods

A retrospective cohort study was conducted using data from Arkansas All Payers’ Claims Database. Adults with incident LBP diagnosed in primary care or emergency departments between July 1, 2013, and June 30, 2017, were identified. Participants were required to be opioid naïve in the 6-month baseline period and without cancer, cauda equina syndrome, osteomyelitis, lumbar fracture, and paraplegia/quadriplegia in the entire study period. PT and chiropractic treatment were documented over the ensuing 30 days starting on the date of LBP. Any opioid use and long-term opioid use (LTOU) in 1-year follow-up were assessed. Multivariable logistic regressions controlling for covariates were estimated.

Results

A total of 40 929 individuals were included in the final sample, with an average age of 41 years and 65% being women. Only 5% and 6% received PT and chiropractic service, respectively, within the first 30 days. Sixty-four percent had incident opioid use, and 4% had LTOU in the follow-up period. PT was not associated with incident opioid use (odds ratio [OR], 1.07; 95% confidence interval [CI], 0.98-1.18) or LTOU (OR, 1.19; 95% CI, 0.97-1.45). Chiropractic care decreased the odds of opioid use (OR, 0.88; 95% CI, 0.80-0.97) and LTOU (OR, 0.56; 95% CI, 0.40-0.77).

Conclusion

In this study we found that receipt of chiropractic care, though not PT, may have disrupted the need for opioids and, in particular, LTOU in newly diagnosed LBP.

Introduction

Based on data from the 2016 National Health Interview Survey, approximately 50 million U.S. adults report having chronic pain and 19.6 million additional U.S. adults report severe chronic pain that significantly interferes with daily life activities.¹ As a result of the growing popularity in opioid analgesics for pain management, opioid prescriptions have increased drastically in the previous 2 decades. Despite a recent downward trend in opioid prescribing, rates in 2015 were still 3 times higher than prescribed in 1999.^2,3 Misuse, abuse, and overdose increased proportionately with the rise in opioid prescribing, reaching levels now considered “epidemic.”⁴ More than 47 000 opioid-related deaths were reported in 2018 alone.⁵

Low back pain (LBP) is the most common noncancer pain complaint for which opioids are prescribed.⁶ About 80% of adults have at least 1 LBP episode in their lifetime and LBP is among the most common reasons for a physician visit in the United States.^{7, 8, 9} Furthermore, the point prevalence of chronic LBP in adults between the ages of 20 and 69 years was estimated at 13.1%.¹⁰ Despite the lack of compelling evidence,^11,12 between a quarter to a half of patients newly consulting a primary care provider for LBP receive an opioid prescription.^{13, 14, 15}

According to data collected from the 2013 and 2014 National Health and Nutrition Survey, nearly 80% of opioid users take their pain medication long-term.¹⁶ Long-term opioid use (LTOU) often begins with a prescription to a previously opioid-naïve individual for an acute pain condition such as LBP,¹⁷ and several studies have demonstrated how opioids that are prescribed over an extended period of time influences risk of misuse and overdose.^18,19 Compared with short-term use, LTOU increases the risk of opioid use disorders 5- to 40-fold depending on dose, prompting greater attention on guidelines and standards for initial prescribing.¹⁹ Several studies have focused on prescribing patterns surrounding opioid initiation, finding the type of opioid prescribed (short- or long-acting), the days’ supply, the cumulative dose, the number of physician visits and the number of fills in the first month of initiation predict LTOU.^{20, 21, 22, 23, 24, 25} However, little research has focused on the use of nonpharmacological treatment services in pain management as a strategy to reduce LTOU risk, and preliminary studies aimed at examining care by physical therapists (PTs) and chiropractors consistently show a moderate to strong association between early care visits and lower LTOU risk.^{26, 27, 28} However, there is a lack of research that examines both of these services when studying early care decisions in the reduction of opioid initiation and LTOU.

Chiropractors and PTs are both commonly used to manage nonspecific low back pain. Physical therapists have historically been oriented to restoring normal movement through manipulation, exercise, and stretching, while chiropractors have historically focused on manipulation to improve function of the spine to alleviate back pain. Over the past several decades, the practice choices and attitudes of PT and chiropractic care providers have converged toward a multimodal care approach that follows current clinical practice guidelines for musculoskeletal pain treatment.^{29, 30, 31} For instance, providers in both professions tend to follow similar diagnostic criteria and use the same evidence-based treatment modalities for in-office and home exercise, soft-tissue therapy, patient education, and manual therapy.^29,30,32,33 Unsurprisingly, most studies find PTs and chiropractors to be similarly effective for the treatment of LBP.^34,35 However, PT and chiropractic practices are not identical in the United States, and few studies have evaluated the subtle differences in the standards of practice (eg, utilization rates of radiographic imaging, prioritization of treatment modalities) and patient perceptions of care that can potentially affect patient outcomes.^29,33,34,36 For instance, 1 study evaluating the sociodemographic differences between patients who either seek chiropractic care or PT care found that chiropractic care is associated with a higher self-rating of health status and less reported days of disability (missed work, school, or bedridden owing to condition).³⁷ Yet, it is unknown how physical therapy and chiropractic care may influence the use of opioids in patients with LBP.

Therefore, the objective of this study was to identify the influence of early treatment with either PT or chiropractic care on opioid initiation and LTOU among patients with LBP. We hypothesized that the use of either PT or chiropractic care in the early treatment of LBP would result in a decrease in the rate of both opioid initiation and LTOU. Research focused on early care decisions involving nonpharmacological treatment options may aid practitioners by providing further insight into the comparative efficacy of different physical treatment modalities in reducing opioid use.

Materials and Methods

Data Source

The Arkansas All Payers’ Claims Database from January 2013 through June 2018 was used for this study. The Arkansas All Payers’ Claims Database is a deidentified state-level database of medical, pharmacy, and dental claims from both public and private health plans.³⁸ The database also contains information on enrollment status, type of health plan, and sociodemographic characteristics of insured individuals. The data were acquired through the Arkansas Biosciences Institute at the University of Arkansas for Medical Sciences and this study was determined to be exempt from human subject research by the University of Arkansas for Medical Sciences institutional review board.

Study Design and Cohort Selection

A retrospective cohort study design was used. We identified enrollees in Medicaid or commercial health plans with an incident LBP diagnosis in the primary care or emergency department (ED) between July 1, 2013 and June 30, 2017. LBP was identified examining both primary and secondary ICD-9-CM and ICD-10-CM diagnosis codes indicative of LBP (see supplemental file). The first visit with LBP diagnosis was considered the index date. To focus the analysis on adults who will not age up to Medicare and are cancer free and opioid naïve before their LBP diagnosis, the following inclusion and exclusion criteria were applied. Individuals had to be between 18 and 63 years of age at the index date continuously enrolled with both medical and pharmacy benefits for 6 months before and 12 months after the index date. The 6-month period before the index date was considered the baseline period, and the 1 year after the index date was the follow-up period. Individuals were excluded if they had any opioid prescription in the baseline period or had a cancer, osteomyelitis, cauda equina, lumbar/pelvic fracture, or quadriplegia/paraplegia diagnosis in the baseline or follow-up period. Individuals whose first opioid prescription in the follow-up was suboxone (buprenorphine-naloxone) also were excluded. Finally, individuals for whom sex was missing in the Arkansas All Payers’ Claims Database also were excluded.

Study Measures

The main variables of interest were the receipt of PT or chiropractic care on, or within 30 days after, the index LBP diagnosis. Receipt of PT or chiropractic care was assessed individually as well as in a composite measure of receiving either type of care by examining the CPT-4 codes (see supplemental file) in inpatient or outpatient claims with a primary or secondary diagnosis for LBP. The use of lumbar radiographs, lumbar advanced imaging and lumbar injections, and ED and physician office visits where a LBP diagnosis was recorded also were assessed in the 30-day window after the index date using the CPT-4 codes described in the supplemental file. Demographics (age and sex), health system (payer type), year of index diagnosis, baseline mental health comorbidities (mood disorders, schizophrenia, anxiety, personality disorders, and acute reactions to stress), substance use disorders (tobacco, opioid, and other) and a general measure of comorbidity assessed by the Charlson score were created by searching claims for diagnosis codes in the 6-month baseline period.³⁹ Additional covariates describing baseline use of nonsteroidal anti-inflammatory drugs, benzodiazepines, skeletal muscle relaxants, and oral steroids in the baseline were recorded by searching the prescription claims files. The main outcome measures were any opioid use and LTOU. Any opioid use was characterized in the 14-day follow-up period and the 1-year follow-up period after the index LBP diagnosis by searching for any oral opioid analgesic formulations. LTOU was identified in the 1-year follow-up period. LTOU was defined as receiving a total supply of 90 or more days of opioids in any 180-day period without a gap of more than 30 days. An alternative definition used in sensitivity analyses to define LTOU is receiving a ≥120-day supply of opioids, or receiving a ≥90 supply with 10 or more separate prescription dispensings.

Statistical Analysis

Bivariate analyses were conducted between PT and chiropractic care and the opioid use outcome measures using χ² tests. Separate multivariable logistic regression analyses among PT, chiropractic care, and the composite PT-chiropractic care as exposures with any opioid use and LTOU were estimated adjusting for demographic and clinical variables. Odds ratio (OR) and 95% confidence intervals (CI) are reported from the multivariable regressions. The variables included in the model were based on clinical knowledge and past evidence, and no variable selection techniques were used. A correlation matrix between all covariates and variance inflation factors were estimated to detect multicollinearity. SAS version 9.4 (SAS Institute, Cary, North Carolina) was used for all analyses.

Results

There were 253 388 unique individuals with at least 1 LBP diagnosis in the primary care or ED setting between July 2013 and June 2017 (Table 1). After applying the inclusion and exclusion criteria, there were 40 929 cancer-free patients with LBP without prior opioid use who had continuous enrollment for at least 18 months. The average age of the sample was 41 years (standard deviation = 12.61), 65% were female and 80% had commercial insurance (Table 2). The most common mental health conditions were mood disorders (26%) and anxiety disorders (26%), and 21% had nicotine dependence diagnosed. Nineteen percent had nonsteroidal anti-inflammatory prescriptions in the baseline period, and 14% filled at least 1 benzodiazepine prescription in the baseline period. A little more than a third of participants had some lumbar imaging within 30 days of the initial LBP diagnosis, with 34% having a lumbar radiograph and 0.5% having a lumbar MRI or CT.

Table 1.

Cohort Selection Flow Chart for Patients With Low Back Pain

Criteria	Number of Individuals	Number of Individuals Lost
Individuals with low-back pain as primary or secondary diagnosis between July 1, 2013, and June 30, 2017, in a visit with primary care physicians, urgent care, or emergency department (index date)	253 388	Not applicable
Individuals with no opioid prescription in 6 mo before the index date (baseline)	86 903	166 485
Individuals with no low-back pain diagnosis in baseline (incident diagnosis)	77 476	9427
Individuals with continuous enrollment in baseline	70 762	6714
Individuals with continuous enrollment in the 12 mo post–index date (follow-up period)	68 242	2520
Individuals ≥18 y of age on index date	58 590	9652
Individuals ≤63 y of age on index date	50 454	8136
Individuals with both medical and pharmacy benefits and non-Medicare insurance in the study period	47 864	2590
Individuals without cancer in the baseline	46 759	1105
Individuals without osteomyelitis, cauda equina, lumbar/pelvic fracture, and quadriplegia/paraplegia in the baseline	46 599	160
Individuals without cancer in the follow-up period	44 890	1709
Individuals without osteomyelitis, cauda equina, lumbar/pelvic fracture, and quadriplegia/paraplegia in the follow-up period	44 105	785
Individuals without buprenorphine-naloxone as the first opioid prescription after the index date	44 084	21
Individuals without missing/unknown sex	40 929	3155

Table 2.

Demographic and Clinical Characteristics of the Final Cohort (N = 40 929)

Characteristics	Frequency	%
Age, mean (standard deviation)	41.45 (12.61)	NA
Age categories
18-24 y	5047	12.33
25-29 y	4195	10.25
30-39 y	9659	23.60
40-49 y	9484	23.17
50-59 y	9834	24.03
60-63 y	2710	6.62
Sex
Male	14 531	35.50
Female	26 398	64.50
Payer type
Commercial	32 720	79.94
Medicaid	8152	19.92
Both commercial and Medicaid	57	0.14
Mental health disorders
Mood disorders	10 706	26.16
Schizophrenia	1123	2.74
Anxiety disorders	10 816	26.43
Personality disorders	608	1.49
Acute reaction to stress	361	0.88
Adjustment disorders	1362	3.33
Substance use disorder
Tobacco use disorder	8542	20.87
Opioid misuse disorder	428	1.05
Other substance use disorder	2123	5.19
Obesity	8565	20.93
Diabetes	6712	16.40
Charlson Comorbidity Index
0	29 778	72.76
1	7606	18.58
2	1866	4.56
≥3	1679	4.10
Year of index diagnosis
2013	4582	11.19
2014	13 022	31.82
2015	10 453	25.54
2016	8721	21.31
2017	4151	10.14
Opioid use within the first 14 d of index date
No opioid use	28 747	70.24
Any opioid use	12 182	29.76
Opioid use in the 1-y follow-up
No opioid use	18 686	45.65
Any opioid use	22 243	54.35
Long-term opioid usea	1805	4.41
Long-term opioid useb	1951	4.77
Days’ supply among opioid users in the 1-y follow-up period
1-2 d	1863	8.38
3-4 d	5382	24.20
5-7 d	6963	31.30
8-10 d	3610	16.23
11-14 d	479	2.15
15-21 d	1658	7.45
≥22 d	2288	10.29
Type of opioid among opioid users in the 1-y follow-up period
Schedule II long acting	51	0.23
Other Schedule II short acting	218	0.98
Oxycodone short acting	1766	7.94
Hydrocodone short acting	10 894	48.98
Schedule III-IV and nalbuphine	2853	12.83
Tramadol	6461	29.05
Pharmacologic use in the baseline period
NSAIDs	7619	18.62
Benzodiazepines	5529	13.51
Skeletal muscle relaxant	3585	8.76
Oral steroids	4994	12.20
Lumbar procedures (within the first 30 d of index date)
Lumbar radiographs	14 085	34.41
Lumbar advanced imaging (MRI or CT)	3902	0.53
Lumbar injections	301	0.74
Physician office visits with low back pain diagnosis (within the first 30 d of index date)	36 041	88.06
ED visit with low back pain diagnosis (within the first 30 d of index date)	4719	11.53

CT, computed tomography; ED, emergency department; MRI, magnetic resonance imaging; NA, not applicable; NSAIDs, nonsteroidal anti-inflammatory drugs.

^aPrimary long-term opioid use definition.

^bSensitivity long-term opioid use definition.

The use or PT and chiropractic care for LBP was relatively low. Only 11% of participants used either PT or chiropractic care within 30 days of receiving a LBP diagnosis; 5% of the sample used PT and 6% used chiropractic care. Thirty percent of the sample had at least 1 opioid prescription within 14 days of index date and 54% ultimately had 1 or more opioid prescriptions in the 1-year follow-up. Among all opioid users, 64% had their first opioid prescription of less than or equal to 7 days supplied, and 10% had first opioid script with greater than 21 days’ supply. 1805 individuals (4%) had LTOU, using the main definition. With the sensitivity definition, 1951 individuals (5%) had LTOU.

The use of PT in first 30 days was not associated with any opioid use (PT Yes = 55.87%, PT No = 54.26%; P = .139); however, those who had PT were more likely to transition to LTOU (PT Yes = 5.64%, PT No = 4.34%; P = .004) in unadjusted bivariate analyses (Table 3). Chiropractic care, however, was associated with a lower percentage of persons using any opioids (Chiropractic Yes = 39.28%, Chiropractic No = 55.29%; P < .0001) and LTOU (Chiropractic Yes = 1.66%, Chiropractic No = 4.58%; P < .0001) in unadjusted bivariate analyses. In multivariable adjusted analyses, similar relationships were observed. PT was not associated with either any opioid use (OR, 1.07; 95% CI, 0.98-1.18) or LTOU (OR, 1.19; 95% CI, 0.97-1.45), and chiropractic care was associated with a reduction in any opioid use (OR, 0.88; 95% CI, 0.80-0.97) and a more profound reduction in LTOU (OR, 0.56; 95% CI, 0.40-0.77) (Tables 4 and 5). These relationships were preserved when PT and chiropractic care were both entered in the multivariable model. The composite PT-chiropractic care was not associated with any opioid use or LTOU (Table 5). Multicollinearity between covariates was not detected, as none of the variance inflation factors estimates exceeded 2.0.

Table 3.

Bivariate Associations Between Opioid Use Measures and Physical Therapy/Chiropractic Care

Opioid Use	Physical Therapy With Low Back Pain Diagnosis in 0-30 d After Index Date		Chiropractic Care With Low Back Pain Diagnosis in 0-30 d After Index Date		Total
Any opioid use	No n (col%)	Yes n (col%)	No n (col%)	Yes n (col%)
No	17 708 (45.74)	978 (44.13)	17 222 (44.71)	1464 (60.72)	18 686
Yes	21 005 (54.26)	1238 (55.87)	21 296 (55.29)	947 (39.28)	22 243
χ2 = 2.18; P = .139			χ2 = 234.41; P ≤ .001
Long-term opioid use
No	37 033 (95.66)	2091 (94.36)	36 753 (95.42)	2371 (98.34)	39 124
Yes	1680 (4.34)	125 (5.64)	1765 (4.58)	40 (1.66)	1805
χ2 = 8.42; P = .004			χ2 = 45.99; P ≤ .001
Total	38 713	2216	38,518	2411	40 929

col%, column percentage.

Table 4.

Odds Ratios and 95% Confidence Intervals for Physical Therapy and Chiropractic Care as Exposures and Any Opioid Use Within 1 Year of Follow-up as the Outcome

Exposure	Odds Ratio (95% CI)	P Value
Physical therapya	1.07 (0.98; 1.18)	.142
Chiropractic carea	0.88 (0.80; 0.97)	.008
Physical therapy–chiropractic care composite variablea	0.97 (0.90; 1.04)	.405
Physical therapyb	1.07 (0.97; 1.17)	.177
Chiropractic careb	0.88 (0.80; 0.97)	.010

^aEstimates are obtained from separate multivariable logistic regressions that adjust for age, sex, payer type, mood disorders, schizophrenia, anxiety disorders, personality disorders, acute reaction to stress, adjustment disorders, tobacco use disorder, opioid use disorder, other substances use disorder, obesity, diabetes, Charlson Comorbidity Index, year of index diagnosis, lumbar radiographs, lumbar advanced imaging, lumbar injections, ED visit with low back pain diagnosis, physician office visits with low back pain diagnosis, NSAID use in baseline and within 14 days of index date, skeletal muscle relaxants use in baseline and within 14 days of index date, corticosteroids use in baseline and within 14 days of index date, and benzodiazepines use in baseline and within 14 days of index date.

^bEstimates are obtained from multivariable logistic regression in which both physical therapy and chiropractic care were used as 2 separate exposure variables.

Table 5.

Odds Ratios and 95% Confidence Intervals for Physical Therapy and Chiropractic Care as Exposures and Long-Term Opioid Use in 1 Year of Follow-up as the Outcome

Exposure	Odds Ratio (95% CI)	P Value
Physical therapya	1.19 (0.97; 1.45)	.092
Chiropractic carea	0.56 (0.40; 0.77)	.001
Physical therapy–chiropractic care composite variablea	0.93 (0.78; 1.11)	.439
Physical therapyb	1.18 (0.97; 1.44)	.104
Chiropractic careb	0.56 (0.40; 0.78)	.001

^aEstimates are obtained from separate multivariable logistic regressions that adjust for age, sex, payer type, mood disorders, schizophrenia, anxiety disorders, personality disorders, acute reaction to stress, adjustment disorders, tobacco use disorder, opioid use disorder, other substances use disorder, obesity, diabetes, Charlson Comorbidity Index, year of index diagnosis, lumbar radiographs, lumbar advanced imaging, lumbar injections, ED visit with low back pain diagnosis, physician office visits with low back pain diagnosis, NSAID use in baseline and within 14 days of index date, skeletal muscle relaxants use in baseline and within 14 days of index date, corticosteroids use in baseline and within 14 days of index date, and benzodiazepines use in baseline and within 14 days of index date.

^bEstimates are obtained from multivariable logistic regression in which both physical therapy and chiropractic care were used as 2 separate exposure variables.

Discussion

The use of PT and chiropractic care was low in Arkansas, with only 11% using either service within 30 days of LBP diagnosis. Only 5% used PT in this period. Compared with Western states (Alaska, Washington, Oregon, Idaho, Wyoming, and Montana) from 2009 through 2013, almost twice as many (9%) sought PT care the day of the LBP diagnosis and an additional 12% used at some time in the follow-up.⁴⁰ Three of those states had provisions for unrestricted direct access to physical therapists during the study period, meaning that a physician referral was not required to visit a physical therapist.^41,42 As of July 2019, 5 of those 6 states had unrestricted direct access.⁴¹ Arkansas, on the other hand, has direct access with provisions—the provisions could vary by plan and may include physician referral and visit limits, among others.^41,43 This difference in access to PT across states may explain its low rates of utilization after LBP diagnosis in our data. Another study that used national commercial insurance data from 2007 to 2015 reported that 29% of individuals with musculoskeletal pain used PT within 90 days of diagnosis.²⁶ This suggests that PT may be significantly underutilized in Arkansas relative to other states.

Rates of chiropractic care use were also lower than have been reported in other populations.^{44, 45, 46} A survey of Veterans with chronic noncancer pain reported that 56% of respondents reported having received chiropractic care, and 75% intended to use chiropractic care.⁴⁷ A systematic review of cross-sectional and longitudinal survey studies reported that 16% to 74% of patients with back pain received chiropractic care.⁴⁸ Several factors may explain the low rates of chiropractic care in Arkansas. First, Arkansas has the lowest density of chiropractors of approximately 2 chiropractors per 1,000 beneficiaries, which is among the lowest in the country based on Medicare claims data.⁴⁹ Another factor are the reimbursement policies for chiropractic care. Arkansas Medicaid pays for a maximum of 12 visits per year for enrollees at least 21 years of age and requires physician referral.⁴³ It only pays for manual manipulation of spine and does not cover other types of chiropractic services.⁴³ The coverage of chiropractic care in commercial insurance is plan-specific and may not be dramatically different in Arkansas than in other states.⁵⁰

Half of the individuals with LBP had at least 1 opioid prescription in the 1-year follow-up period, and 4% to 5% had LTOU. The use of chiropractic care within 30 days of LBP diagnosis was associated with diminished use of opioids in the short term and, in particular, the long term, in which the risk of LTOU was almost cut in half. Surprisingly, the use of PT did not mitigate the risk for acute or LTOU, which is in contrast to previous studies that have found reductions in opioid use with the use of PT.^{26, 27, 28,51} For example, a study using MarketScan commercial data reported a reduction of opioid use (measured in morphine milligrams equivalent) in patients with LBP by 5% with the use of PT within 90 days of diagnosis.²⁶ Another study used Optum claims data, comprising commercial and Medicare Advantage enrollees, documented a 73% reduction in the odds of LTOU with the use of PT.²⁷ A 53% reduction in the odds of any opioid prescription was observed in Medicaid enrollees with early participation in PT.⁵¹ The definition for PT exposure varies across studies including ours and could partially explain discrepancies in the findings. Two of the studies that report the largest reductions in opioid use required either the first visit to be a PT visit, or a PT visit on the same day as diagnosis to be considered in the PT group,^27,49 whereas our study requires the initial encounter to be in the physician office or the ED. Examining persons with LBP who self-refer to a PT excludes LBP initially managed in the ED or physician office setting where opioids can be prescribed. Although existing observational studies are consistent in finding a reduction in opioid use with PT care, several randomized trials have failed to find differences in opioid use within 1 year between usual care and PT care groups for LBP.⁵²

Existing studies on chiropractic care have reported reductions in opioid use.^46,53,54 A retrospective study on Veterans reported a 4-percentage point decrease in any opioid prescription within 90 days of a chiropractic visit.⁵⁵ A 50% reduction in likelihood of opioid use has been reported in retrospective claims-based studies.^53,54 A clinical trial conducted in active-duty Veterans with LBP documented a 27% reduction in the odds of pain medication use.⁵⁶ A survey-based study reported that more than 80% of patients with chronic pain mentioned they derived benefit from chiropractic treatment.⁵⁷ Our results are in line with these studies and suggest, along with others, that chiropractic may help patients engage in nonpharmacologic self-care for pain despite focusing on passive interventions.⁵⁸ It may also be the case that individuals with LBP who desire to manage pain without undue reliance on medication are more likely to self-select chiropractic care. The retrospective nature of our study cannot determine if this finding represents a selection bias or treatment effect.

Limitations

Selection bias or residual confounding cannot be ruled out despite controlling for a wide range of demographics, health systems, and comorbidities. Variables such as health plan benefits design, rural-urban residence, and pain intensity could have confounded the results. For example, individuals seeking chiropractic care may have been more likely to be enrolled in health plans with more strict opioid prescribing or dispensing regulations. The low risk of LTOU in chiropractic care users could be a result of the plan-level difference instead of a causal effect of chiropractic care on back pain.

We did not assess type and duration of PT and chiropractic care used. The CPT codes used in the claims data for identifying PT and chiropractic care are not sufficiently granular to tease apart their different types. We did not evaluate outcomes such as self-reported pain or health-related quality of life, which would have provided a more complete understanding of the effects of PT and chiropractic care.

The ICD-9/10-CM codes used to identify back pain in our study may be incomplete, in particular when considering the newer ICD-10-CM codes such as M99.01-M99.04, which we did not include to identify back pain. This would have resulted in fewer persons identified for the study and may have under-assessed the frequency of PT or chiropractic care, which required a practitioner-specific CPT-4 code in addition to a back pain diagnosis. This omission is likely minor, as the most frequent ICD-10-CM codes (M54.2 and M54.5) were included and these codes are the most frequently used codes by practitioners for low back pain.⁵⁹ The study data were limited to a single state, Arkansas, which has low rates of chiropractic and PT use and high rates of opioid prescribing, and thus may not generalize to other states or nationally.

Conclusion

The use of PT and chiropractic care was low among Arkansans with LBP and with Medicaid or commercial insurance. In this sample of the population, use of chiropractic care was associated with substantial reduction in likelihood of any opioid use and LTOU, and no association was observed with PT.

Funding Sources and Conflicts of Interest

This study was supported from funds of the UAMS Translational Research Institute grant UL1 TR003107 through the National Center for Advancing Translational Sciences of the National Institutes of Health. B.C.M. received royalties in the past 3 years from TrestleTree, LLC for the development of an opioid risk prediction tool and has received consulting fees from eMAX Health for unrelated studies.

Contributorship Information

Concept development (provided idea for the research): J.M.F., B.C.M.

Design (planned the methods to generate the results): J.M.F., B.C.M., M.A., D.C.

Supervision (provided oversight, responsible for organization and implementation, writing of the manuscript): B.C.M.

Data collection/processing (responsible for experiments, patient management, organization, or reporting data): M.A., D.C.

Analysis/interpretation (responsible for statistical analysis, evaluation, and presentation of the results): M.A., D.C.

Literature search (performed the literature search): A.M.S.

Writing (responsible for writing a substantive part of the manuscript): A.M.S., M.A., D.C.

Critical review (revised manuscript for intellectual content, this does not relate to spelling and grammar checking): A.M.S., B.C.M., D.C., M.A., J.M.F.

Practical Applications.

• •Utilization of chiropractic care was associated with substantial reduction in likelihood of any opioid use and long-term opioid use.
• •The utilization of physical therapy and chiropractic care was low among Arkansans with low back pain and with Medicaid or commercial insurance.
• •Physical therapy and chiropractic care may be significantly underutilized in Arkansas compared with other states.
• •A factor that may affect the utilization of chiropractic care and physical therapy are specialist care reimbursement policies.

Alt-text: Unlabelled box