Introduction

Spine-related musculoskeletal pain is the leading cause of disability worldwide and one of the most common reasons for missed work [1]. In the United States (U.S.), healthcare costs for low back and neck pain are rising and as of 2016 were the highest for any condition, with an estimated $134.5 billion for care related to spinal pain [2].

There are many options for treatment of acute or chronic spine-related pain. These range from conservative therapies, such as manual or behavioral therapies, to medications, injection procedures and surgery [3, 4]. Approaches to management of spine-related musculoskeletal pain differ by type of provider, such as chiropractors, physical therapists, primary care medical physicians and medical specialists such as orthopedists and neurologists [5]. In the U.S., chiropractic care is one of the most commonly utilized approaches to treatment of spine-related musculoskeletal pain [6]. Chiropractic care guidelines are concordant with the American College of Physicians’ recommendations for initial management of low back pain (LBP) using non-pharmaceutical treatment [7, 8].

In the midst of rising healthcare costs, it is important to examine not only clinical outcomes but also the cost of intervention strategies for spine-related pain. Although most cases of spine-related musculoskeletal pain can be effectively managed with conservative guideline-concordant non-pharmacological and non-invasive approaches, frequently a patient’s course of care is unnecessarily escalated by use of more invasive, hazardous, and/or costly procedures [9]. The escalation of care for spine-related musculoskeletal pain may include emergency department visits, medical specialist visits, diagnostic imaging, hospitalization, surgery, interventional pain medicine techniques, prescription of drugs with high risk for addiction or abuse, and encounters for complications of spine care (e.g., adverse drug events) [9]. The escalation of spine-related musculoskeletal pain management is closely associated with increased downstream costs.

Gold et al. defined “downstream” costs as those that “may have changed, intentionally or unintentionally, as a result of the implementation strategy and intervention.”[10]p.3 Downstream costs may include those associated with healthcare utilization, patient and caregiver costs, productivity costs and costs to other sectors. For spine-related musculoskeletal pain, most often LBP, an emerging body of evidence suggests that downstream costs are significantly affected by the specialty of the initiating care provider [5]. Such costs typically include diagnostic tests, particularly advanced imaging [11], surgery, specialist care and medication use [12].

The opioid epidemic. For patients with spine-related musculoskeletal disorders, among the most important escalations of care associated with downstream human and societal costs that are receiving recent attention are opioid use, abuse and overdose. In 2017, the U.S. government declared the opioid epidemic to be a Public Health Emergency [13]. This epidemic is still on the rise, with drug overdose deaths increased by 31% in a single year, 2019–2020 [14].

It is not certain which combination of provider and/or therapy offers the most cost-effective approach to managing spine-related musculoskeletal pain. A 2015 systematic review compared the costs of chiropractic care to those of other types of health care [15]. The costs were generally lower when musculoskeletal spine care was managed by chiropractors, though the included studies contained methodological limitations [15]. The purpose of this review was to update, summarize, and evaluate the evidence for the cost of chiropractic care compared to conventional medical care for management of spine-related musculoskeletal pain [15].

Methods

Our team followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol to conduct the review and registered it with PROSPERO in December 2022 prior to data abstraction (CRD42022383145). We elected a priori not to pursue meta-analysis due to heterogeneity of the included studies. Most of the included studies are cohort studies which by their nature can only show associations, cannot prove causation, and are of a lower level of evidence than RCTs, which are the study design usually included in meta-analyses. We conducted the searches and quality assessments from July through December 2022 and data abstraction from January through March 2023. The primary aim of our systematic review was to address the research question: Is chiropractic management of spine-related musculoskeletal pain in U.S. adults associated with lower overall healthcare costs as compared to medical care?

To answer the research question, we formulated PICO elements (Population, Intervention, Comparison, Outcome) as follows:

P: U.S. adults with spine-related musculoskeletal pain

I: Chiropractic management

C: Medical care

O: Healthcare costs and use of procedures estimated to increase downstream costs involved in escalation of care

Costs in a controlled setting are not often comparable to usual and customary costs in a clinical setting [16]. Therefore, in addition to randomized controlled trials, we also included economic and cohort studies that collected data specifically on cost, whether or not treatment outcomes were considered.

Eligibility criteria

Inclusion criteria

  • Published in peer-reviewed journal and available in Medline from inception through 10/31/2022

  • English language

  • Study population comprised of US adults treated for spine-related musculoskeletal pain

  • Compared chiropractic management to medical care

  • Cost data for treatment of spine-related musculoskeletal pain were provided

  • Designs were randomized controlled trial, cohort study or economic evaluation.

Exclusion criteria

  • Reviews, commentaries, abstracts from conference proceedings, theses, cross-sectional descriptive surveys and gray literature.

  • Systematic reviews were not used as part of quality assessment or data abstraction. They were retrieved only to identify eligible studies which were not found in the literature search.

  • Studies with clinical effectiveness outcomes only and no inclusion of cost or utilization data

Literature search

We developed a search strategy based on the PICO terms, with a health sciences librarian working with the other investigators. We made several “trial runs” to refine the strategy to be sure it was as inclusive as possible while screening out obviously non-relevant citations. Our search was conducted exclusively in Medline, as relevant high-quality articles were more likely to be found in journals indexed there. We developed a search strategy with keyword clusters based on our PICO. Most publications about spine-related pain study adults (our P) and “adult” was not helpful as a limiter. Intervention (I) cluster terms were selected from a previously published search string of complementary and alternative medicine terms [17]. The Outcome (O) cluster started with terms used in a prior cost-effectiveness study [18], with the subsequent addition of indexing terms found during early search trial runs. The MeSH heading Cost-Effectiveness Analysis was not yet available at the time of our search. The complete search strategy is shown in Additional File 1.

Retrieved citations were downloaded into an EndNote library (v. 20). Using Rayyan https://rayyan.ai/, [19] online systematic review software, at least two investigators screened titles and abstracts for eligibility and resolved disagreements by discussion. At least two investigators checked the references included in all relevant systematic reviews found in the literature search and added any eligible citations not identified in our literature search to the library. At least two investigators did full-text screening of the titles remaining after title/abstract screening was completed, and disagreements were again resolved by discussion. All authors contributed during the process in review of eligible citations.

Evaluation of risk of bias

We evaluated randomized controlled trials (RCTs) using a checklist modeled after those of the Scottish Intercollegiate Guideline Network (SIGN) [20], which we have used elsewhere [3]. An article was rated as “high quality, low risk of bias,” “acceptable quality, moderate risk of bias,” “low quality, high risk of bias,” or “unacceptable” quality. For studies analyzing treatment costs (e.g., economic studies), we developed a checklist with similar format to those of SIGN checklists [20].

For cohort studies, it was difficult to identify a single appropriate checklist because most seemed designed to be more appropriate to assess prospective cohort studies, and the most recent relevant studies related to our topic are retrospective cohort studies using large datasets. We therefore developed a checklist for prospective cohort studies after reviewing other existing checklists [20]. For retrospective or cross-sectional cohort studies, we developed a checklist combining some features of the SIGN checklist for cohort studies [20] and the NIH tool for observational cohort and cross-sectional studies [21]. These checklists included items assessing comparability of the included cohort groups, as part of the risk of bias assessment. Three investigators (RF, CH and JW), one of whom is an author of a number of cohort studies, piloted and then refined this form with a sample of studies.

Two or more investigators rated each article. Disagreements were resolved by including additional reviewers and discussing differences in ratings until they reached agreement.

Because of the large number of cohort studies, which are considered to have an inherent risk of bias due to their design, we only included studies which the investigators agreed were at minimum “acceptable quality, moderate risk of bias” using the SIGN quality criteria. We excluded any studies that the investigators found to be “low quality, high risk of bias” or “unacceptable quality.”

Data extraction

Because it has been found that data extraction errors are frequent in systematic reviews, we followed the recommendations on data extraction in a review of data extraction guidelines and methods [22]. Before starting the process, we listed all the items we thought were necessary for answering our research question. Then we drafted a data extraction form with these items and two investigators (RF and CH) piloted it on a sample of studies. We then provided brief, online training on use of the forms with the 3 investigators who did the data extraction (RF, CH, DT). This included instructions on how disagreements would be resolved, which was to recheck the source paper and provide it to the other reviewer(s). Two investigators (RF and CH) did independent parallel extraction for all studies, and DT then reviewed the drafted tables; thus the data extraction was triple-checked. We did not attempt to subcategorize patient populations from the included studies.

Results

We concluded the search in November 2022 and retrieved 2247 citations. Figure 1 shows the PRISMA flow chart, showing reasons for exclusions. Nine articles were identified by reference tracking and expert consultation to make the total number of articles for screening 2256. Title and abstract screening of these resulted in 93 articles for full-text screening, with 49 excluded (see Additional File 2 for citations) and 44 remaining for quality assessment and data extraction. Please refer to Table 5 for a summary of findings including a quick-view color coded identification format related to each accepted paper. For ease of viewing, we divided the papers using two headings: (1) types of costs and (2) factors affecting costs.

Fig. 1
figure 1

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram of literature search. *Excluded studies are listed in Supplementary materials

Full size image

Final inclusions and quality assessment

Table 1 lists the study design and quality rating for the 44 included studies. All were rated high or acceptable quality (see Additional File 3 for details for the quality assessment) [20]. There were 4 prospective cohort studies [23,24,25,26], 22 retrospective or cross-sectional cohort studies [9, 12, 27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46], 17 cost studies [5, 47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62] and one randomized controlled trial [63], although 2 of the cost studies used data from RCTs.

Table 1 Included studies, by study design and first author
Full size table

Data extraction and summary

Because of the large number of studies, we separated the data extraction into two tables, using 2017, the year the U.S. declared the opioid epidemic [13], as the dividing line. Table 2 displays data extracted from each included study published between 2018 and 2022. Table 3 displays data extracted from each included study published between 1991 and 2017.

Table 2 Summary of included studies 2018–2022
Full size table
Table 3 Summary of included studies 1991–2017
Full size table

There were 17 included articles published in the 5 years from 2018 to 2022 (approximately 3 articles per year). There were 27 included articles published in the 26 years from 1991 to 2017 (approximately 1 article per year). From 2018 to 2022, most [15] were retrospective/cross-sectional cohort studies, with 1 prospective cohort study and 1 economic/cost study. From 1991 to 2017, most [16] were economic/cost studies, with 7 retrospective/cross-sectional cohort studies, 3 prospective cohort studies and 1 randomized controlled trial.

Summary of cost factors

Table 4 summarizes the findings of all included studies, by year of publication. Below we have grouped these by type of cost and factors affecting cost. Table 5 depicts chiropractic services versus comparisons in terms of higher, lower or no difference in association for each of the identified types of costs and downstream utilization of factors affecting costs. All of the included studies newer than 2009 demonstrated associations that favored chiropractic services in regard to lower costs and lower utilization of services.

Table 4 Summary of findings for chiropractic management vs medical management, by year of publication
Full size table
Table 5 Association of chiropractic care with factors affecting costs, by study
Full size table

Type of costs

  • Total costs Ten studies found that Doctor of Chiropractic (DC) care had lower overall costs [5, 12, 38, 41, 42, 51, 52, 55, 58, 62]. No studies found that DC care had higher overall costs.

  • Costs per episode of care Six studies found that DC care had lower costs per episode [35, 38, 41, 49, 59, 62], and two found that it had higher cost per episode [23, 56].

  • Insurance/compensation costs Four studies found DC care had lower insurance and compensation payment costs [49, 53, 55, 57]. No studies found higher costs.

  • Long-term healthcare costs Two studies found lower long-term healthcare costs associated with DC care [36, 42]. No studies found higher costs.

  • Office visit costs One study found reduced costs for DC office visits [12]; four studies found that DC care had higher costs for office visits [47,48,49, 61]. Two of these studies noted that chiropractic office costs were higher only when referral costs were not included in the calculation. When referral costs were included, chiropractic office visit costs did not differ significantly from medical care costs [48, 61].

Factors affecting costs

  • Diagnostic imaging Fifteen studies found that diagnostic imaging, particularly advanced imaging like MRI, was used less with DC care; six studies were published 2018–2022 [5, 9, 27,28,29, 31] and nine studies from 1991 to 2017 [25, 32, 33, 47, 51,52,53,54, 61].

  • Opioids Eleven studies found that fewer opioid prescriptions were dispensed or filled with DC care. Ten of these were published 2018–2022 [5, 12, 29, 36, 37, 39, 43,44,45,46], and only one between 1991 and 2017 [40].

  • Surgery Eight studies found fewer surgeries with DC care; four published 2018–2022 [27,28,29,30] and 4 published 1991–2017 [26, 51, 54, 61].

  • Hospitalization Seven studies found fewer hospitalizations with DC care; two studies were published from 2018 through 2022 [9, 29] and five from 1991 through 2017 [51, 54, 55, 59, 60].

  • DC as 1st provider Six studies analyzed cost factors related to having a DC as the 1.st care provider. Generally, this was associated with lower downstream costs. Four studies were published 2018–2022 [5, 27, 36, 37] and two published from 1991 through 2017 [26, 38].

  • Injection procedures Five studies found decreased use of injection procedures with DC care; all were published from 2018 through 2022 [9, 27,28,29, 36].

  • Specialist visits (including surgeon referral visits) Five studies found fewer referrals for specialist visits with DC care; three were published from 2018 through 2022 [9, 29, 31] and two published 1991–2017 [32, 33]. Three studies in the 1991–2017 group stated that their analyses had excluded all referral costs [47, 48, 61].

  • Emergency department (ED) visits Two studies, both published from 2018 through 2022, found that fewer ED visits were associated with DC care [9, 28].

Discussion

The purpose of this systematic review was to address our primary research question: Is chiropractic management of spine-related musculoskeletal pain in U.S. adults associated with lower overall healthcare costs as compared to medical care? This is the first systematic review of this type performed since 2015. In that review, Dagenais et al. found that health care costs were generally lower among patients whose spine pain was managed with chiropractic care. Due to the heterogeneity of patient populations and sample sizes each paper was evaluated by three separate reviewers using the checklists previously described in the Methods Sect. [15] As the literature review progressed, we found that in studies published within the past few years, an important aspect of cost began to emerge that went beyond the immediate per episode cost: the type of initial provider was strongly associated with lower downstream costs.

Downstream costs are often incurred after the initial provider has completed the episode of care. Downstream financial costs include expensive and invasive procedures such as hospitalization, surgery, injection procedures and advanced imaging. There are additional financial and non-financial downstream costs associated with the long-term consequences of addiction to opioid analgesics, including work absenteeism, decreased quality of life, psychological distress, and death due to drug overdose.

Bise et al. continued pursuing this concept in a 2023 cohort study, finding an association between the first choice of provider and future healthcare utilization [64]. His team concluded that both chiropractors and physical therapists provide nonpharmacologic and nonsurgical interventions, and that their early use appears to be associated with a decrease in immediate and long-term utilization of healthcare resources. This study adds further confidence in the emerging body of evidence on provider-related cost differentials and provides a compelling case for the influence of conservative care providers as the first provider managing for spine-related musculoskeletal pain. It follows logically that if downstream interventions are reduced, lower healthcare system costs will follow.

nonpharmacologic and nonsurgical interventions, and that their early use appears to be associated with a decrease in immediate and long-term utilization of healthcare resources. This study adds further confidence in the emerging body of evidence on provider-related cost differentials and provides a compelling case for the influence of conservative care providers as the first provider managing for spine-related musculoskeletal pain. It follows logically that if downstream interventions are reduced, lower healthcare system costs will follow.

The potential human and societal cost savings of avoiding overuse of opioid analgesics, with the possibility of overdoses and addiction, is another important emerging concept in the literature. We found that 10 studies published since the U.S. government declared the opioid epidemic in 2017 found reduced dispensing of opioid prescriptions when DCs were the first provider [5, 12, 29, 36, 37, 39, 43,44,45,46]. Only one study published in the earlier time period (1991–2017) included opioid prescribing as a comparison [41].

Overall, as summarized in Table 4, diagnostic imaging, opioid utilization, surgery, hospitalizations, injection procedures, specialist visits and emergency department visits were all reduced where chiropractors were involved early in the case. We did not subcategorize the patient populations (e.g., general population, Medicare, insurance claims) within any of tables.

Limitations of the study

First, most of the included studies were retrospective cohort studies using large databases. Observational studies can only show associations, not prove causation, so definitive conclusions cannot be made about costs. However, their findings were so consistent that they warrant further scrutiny using higher-level study designs. Second, most of the included studies were retrospective cohort studies and therefore not the highest level of evidence. Third, we did not use any single validated checklist to assess study quality. We evaluated several checklists (e.g., SIGN, CHESS, MMAT) before determining that modification of validated checklists was necessary. Existing checklists seemed better-suited to prospective cohort designs and not as well-suited to the included retrospective cohort designs. The included studies were so variable in design and patient populations that it was not possible to pool the results for meta-analysis. Fourth, some large established datasets contained limited cost outcome variables. This made important factors such as pharmaceutical use and costs unavailable if they were not included in the dataset. Fifth, some observational studies using claims data and electronic health records do not provide enough detailed clinical information to determine whether opioid prescriptions were filled, or if filled, were actually used by the patient. Lastly, we searched only the MEDLINE database, and it is possible we missed other relevant articles not indexed there.

Strengths

Although there are few randomized controlled trials available on this topic, there were many well-conducted cohort studies that provided analyses of large datasets with cost and care data identified by provider type.

While there are certainly other factors affecting cost of care, this paper included the most common cost escalators associated with typical care for LBP, including opioids, injection procedures, surgery, specialist visits and emergency department visits.

Based on the substantial body of evidence published since 1991, a trend is developing in US healthcare systems to include chiropractors as an integral part of the medical/healthcare team, as exemplified by the Veterans Administration (VA) chiropractic clinics across the country [65, 66].

Recommendations. When considering this evidence, it may be in society’s best interest for U.S. healthcare organizations and governmental agencies to consider modifying benefit designs to reduce barriers to access to chiropractic providers. Modifying or eliminating pre-authorization requirements, medical doctor gatekeepers, arbitrary visit limits, co-pays and deductibles may all be considered. Eliminating these barriers would allow easier access to chiropractic services, which based on currently available evidence consistently demonstrate reduced downstream services and associated costs.

Conclusion

Patients with spine-related musculoskeletal pain who consulted a chiropractor as their initial provider incurred substantially decreased downstream healthcare services and associated costs, resulting in lower overall healthcare costs compared with medical management. A primary limitation was related to the heterogeneity and sample sizes of the populations and retrospective data sets. While observational studies cannot prove causation, the recurrent theme of the data seems to support the utilization of chiropractors as the initial provider for an episode of spine-related musculoskeletal pain. Future studies using randomized designs will be helpful in clarifying and validating this trend.