Abstract
Background
Complementary and Integrative Health (CIH) services are a national priority for the Department of Veterans Affairs (VA) healthcare system and can be effective in reducing chronic pain. Eligible VA patients can receive their CIH care through a VA clinic, or through community care (CC) funded by the VA. The present study compares the effectiveness of 3 CIH services (acupuncture, chiropractic, and medical massage therapy) delivered in direct care by VA vs CC providers at improving veterans’ chronic pain.
Methods
Data were analyzed from the Complementary and Integrative Health Therapy Patient Experience Survey, a longitudinal, self-administered survey of CIH use and health outcomes. Mixed models were used to evaluate the relationship of higher CIH therapy visits delivered by the VA vs CC with pain interference and pain severity, using both raw counts of visits and clinically meaningful groupings of visits.
Results
Among veterans with chronic pain who engaged in CIH services, more CIH visits were associated with lower levels of pain severity and pain interference. VA acupuncture and chiropractic had a stronger beneficial relationship with pain than CC acupuncture and chiropractic, while CC medical massage therapy had a stronger beneficial relationship than VA medical massage.
Conclusions
CIH therapies delivered through the VA and through CC both offer potentially effective means of reducing chronic pain. Some therapies may be more effective when delivered through the VA vs CC, or may indicate lack of full implementation, but the limitations of observational data preclude any causal statements.
Introduction
Chronic pain, or pain on most days or every day, is a prevalent and severe health issue affecting approximately 21% of U.S. adults. 1 U.S. veterans have a greater risk of experiencing chronic pain in part due to combat and training-related injuries, and their pain tends to be more severe than the general population.2,3 The biopsychosocial framework for understanding chronic pain also suggests that injuries, psychological stress, and trauma experienced during active military duty may elevate risk for and mediate progression of chronic pain. 4 Also, chronic pain elevates the risk of several comorbid health problems, including insomnia, physical mobility limitations, and depression.5,6
Many complementary and integrative health (CIH) therapies are evidence-based approaches to address chronic pain and are recommended by the American College of Physicians, the American Academy of Family Physicians, the Centers for Disease Control and Prevention, the Department of Veterans Affairs (VA), and the Department of Defense (DoD).7-10 Eight evidence-based complementary and integrative health therapies (acupuncture, biofeedback, clinical hypnosis, guided imagery, medical massage therapy, meditation, tai chi/qigong, yoga) are included in the veterans medical benefits package under VA Directive 1137 – Provision of Complementary and Integrative Health. These may be made available when clinically appropriate as part of a treatment plan through a VA medical facility (in person or via telehealth) or by Community Care Network providers.
The quality of care provided by the VA, and accessibility of care for veterans, are important areas of on-going improvement in VA, including in the delivery of chronic pain treatment. The VA Maintaining Internal Systems and Strengthening Integrated Outside Networks (MISSION) Act was passed in 2018 to make VA-purchased care in the community, dubbed community care (CC), available to veterans, with 45% of VA enrollees utilizing CC in 2022. 11 However, there remains a need for greater understanding of the effectiveness and utilization of CIH therapies provided in the VA and through CC. One systematic review comparing quality of care broadly between the VA and non-VA healthcare systems found, across 69 articles, that VA care performed better than other systems in effectiveness and safety from medical care-related illness or injury. 12 In another study, effectiveness of acupuncture and chiropractic treatment for chronic pain assessed for the VA and for CC showed no notable differences in physical health. 13 Given that this study only examined differences in pain based on whether or not participants had any VA care, there remains the question of how different amounts of care (eg, number of therapy visits) is related to pain, and whether the same number of visits between VA and CC have different relationships to pain.
The present study explores the associations between 3 CIH therapies and veterans’ chronic pain, comparing the acupuncture, chiropractic, and medical massage therapy delivered in VA with that delivered in CC. These results can inform veterans and VA staffing choices in selecting where to seek out chronic pain treatment and may inform future studies about minimally effective dosages of practitioner-delivered CIH therapies.
Methods
Using a cohort study design, we conducted a longitudinal analysis of the associations between veterans’ use of 3 clinician-delivered CIH therapies (acupuncture, chiropractic, medical message therapy) and chronic pain, comparing therapies delivered by the VA and CC over an 8-month period. We used data collected from 2 sources: a survey distributed from March 2021 to April 2023, and electronic health record (EHR) data from the VA. Acupuncture, chiropractic, and medical massage therapy were selected as the focus of this study because they must be delivered in-person by a practitioner, are more feasibly captured within the VA electronic health records than therapies that can be conducted independent of a practitioner or alone (eg, meditation, yoga), and are the 3 most commonly-delivered therapies. Of the 8 CIH approaches that are part of the medical benefits package acupuncture and massage are the most highly utilized in direct care and in the community. Similarly, chiropractic care is highly utilized in both settings, but is not classified in VHA as a complementary integrative approach. The Greater Los Angeles VA Healthcare System Institutional Review Board determined that this project is a quality improvement effort and not a research study.
Data Sources
CIH Therapy Patient Experience Survey
Measures of pain outcomes came from the CIH Therapy Patient Experience Survey, a longitudinal survey of veterans from 2021 to 2023. The survey sampling frame included 15 324 veterans with chronic musculoskeletal pain who had newly begun CIH therapies. An initial survey was sent via both email and a hard copy mailed version. Follow-up surveys were sent to participants who completed the initial survey, at 1 month, 3 months, and 6 months. Surveys could be completed online, over the telephone, or on paper and mailed back using a prepaid return envelope. Additional details regarding development of the survey are published elsewhere. 14
VA Electronic Health Records
Data on CIH therapy visits and sociodemographic characteristics were obtained from the VA EHR and the VA’s CC data set starting in 2021. The VA uses current procedural terminology (CPT) codes for acupuncture, chiropractic, and medical massage therapies, and the VHA offers incentives to medical centers for recording the delivery of CIH therapies. Prior work has demonstrated that VA EHR records capture the vast majority of CIH therapy visits that are VA-funded, which includes CC. 15
Measures
CIH Utilization
Acupuncture, chiropractic, and medical massage therapy data were extracted from VA EHR for all veterans enrolled in the CIH Therapy Patient Experience Survey. The EHR includes identification for whether a given therapy visit occurred at a VA healthcare facility or in CC.
EHR records about CIH visit dates were supplemented by additional information reported by participants who completed surveys. Each therapy visit was aggregated in 2 ways: (1) over the 8 weeks preceding a survey return date (initial survey, 3-month, and 6-month); and (2) over the 8 months preceding the 6-month follow-up survey return date. Each therapy visit was aggregated by person, type of therapy, and whether the visit was at the VA or CC. For visit counts at initial survey, 3-month, and 6-month follow-up, EHR therapy count measures were then harmonized with survey therapy count measures to avoid double counting in the following way: if a count of therapy visits for a given respondent was higher from their survey than what was recorded in the EHR, the survey value was used but capped at an increase of 3 visits above the EHR value; if the survey visit value was lower than the recorded EHR visit count, then the EHR visit count was used. For CIH therapy visits measures over a full 6 months, only EHR counts were used (as survey measures did not cover the full 8-month window).
To test the association between the 3 therapies and pain using pre-determined thresholds for the number of therapy visits, we re-parameterized therapy counts into five-level categorical variables with levels set at visit count intervals, based on guidance from our Subject Matter Expert (JO). These intervals were 0 visits, 1-4, 5-8, 9-12, and 13 or more.
Outcome Variables
We measured pain interference and pain intensity using subscales of the Brief Pain Inventory (BPI), administered at initial survey date, 3-month and 6-month follow-up. 16 For both BPI subscales (range, 0-10; higher scores = worse function or pain intensity), a 1-point improvement was considered clinically important.
Covariates
Factors that may confound the association between our exposure variables (VA and CC acupuncture, chiropractic, and medical massage therapy visits) and outcome variables (pain interference and pain severity) were included as covariates and described below. We included several sociodemographic participant characteristics from the VA EHR (age, sex, distance to the nearest primary care facility, rural/urban designation, and U.S. region), a covariate from the CIH Therapy Patient Experience Survey (Elixhauser Comorbidity Index score), and a model-specific indicator for whether a participant was engaging in CIH therapies other than the therapy of interest. Both age and sex are associated with the propensity to engage in CIH therapy use among veterans, with those below 65 years of age and women have a higher odds of CIH use. 17 A participant’s distance to the nearest primary care facility was a four-level categorical variable (less than or equal to 2 miles, more than 2 miles and less than or equal to 15 miles, more than 15 miles, and an unknown number of miles). 18 Indicators for urban and rural residence were generated using VHA designations pulled from the EHR. The VA classification of urbanicity and rurality incorporates a population density measure that defines urban as any U.S. Census Bureau-defined urbanized area, rural as an area not defined as urban, and highly rural as a rural territory with a population density of less than 7 civilians per square mile. 19 U.S. region of residence was a five-level categorical variable corresponding to groupings of VA veterans Integrated Services Networks (VISNs). The Elixhauser Comorbidity Index score was a unweighted sum of 11 common health comorbidities (eg, hypertension, alcohol abuse). 20
Analysis
We first used descriptive statistics to determine the distribution of VA and CC CIH therapy visits across the 3 therapy types (acupuncture, chiropractic, and medical massage therapy) among all individuals invited to the survey at the first timepoint (the initial survey). For survey responders these descriptive patterns include harmonized EHR and survey-reported utilization. For non-responders, these patterns represent only data available in the EHR and VHA’s community care billing data. We calculated therapy visit frequencies using the five-level categorical variables of count intervals.
To examine linear associations between the number of therapy visits and pain outcomes among survey responders, we fit 6 multivariable hierarchical linear regression models: 1 for each therapy for pain interference and pain severity. For each therapy, the model included the therapy visit count of both the VA and for community care (eg, the model examining acupuncture included a count variable for acupuncture visits leading up to each survey timepoint done at the VA and a separate variable for acupuncture visits designated as community care). Models incorporated random intercepts for each participant and were adjusted for all sociodemographic variables described earlier. A quadratic term for time was also included to account for temporal time trends.
For therapy visits binned into intervals, we also used multivariable linear models with categorical therapy visit variable as the exposure of interest and either pain interference or pain severity at 6-month follow-up as the outcome. Covariates were all sociodemographic variables as well as the count of therapy visits for the same therapy but the opposing source of care (eg, for the association between intervals of VA-delivered acupuncture and pain interference, the model included a covariate for the initial survey count of CC-delivered acupuncture visits). Models were fit for each permutation of therapy type and source of care, and for each of the 2 outcomes.
Participant observations with missing data for variables included in regression models were excluded from analysis (ie, complete case analysis). Of the variables included in the above models, only the outcomes and age had missing values. Because age only had 2 missing values, the coefficients of these models are nearly equivalent to what would be produced using multiple imputation methods to account for missing at random (MAR) data. All analyses were conducted using Stata version 18.
Results
A total of 15 324 veterans identified as initiating CIH were invited to participate in the CIH Experience Survey and 7064 responded. The 3-month follow-up survey was completed by 1685, and the 6-month follow-up by 3726 participants. EHR data were identified and incorporated for all 15 324 veterans. The sample size for analyses varied, as there were more missing data for the pain interference measure than the pain severity measure. A total of 6452 respondents were used in regression models examining pain interference, and 6469 respondents contributed information and were incorporated into models examining pain severity; in each case, these samples had varying amounts of completed surveys among the 3 timepoints (eg, some participants returned only the initial survey, while others returned some or all of the follow-up surveys as well).
Characteristics of Full Survey Cohort Prior to Baseline (Using EHR Data)
Change in Pain Interference and Pain Severity With Each Additional CIH Therapy Visit Across VA and CC Providers
Bold = P < 0.05.
In developing the community care system, the VA developed Standard Episodes of Care (SEOC) for treatment in the community. Because 12 visits is the number of treatments permitted in the initial episode of care for chiropractic and acupuncture, we chose this number of visits to represent a standard episode of care. When extrapolating regression estimates compared to no therapy visits, predicted pain interference scores related to VA acupuncture visits dropped from 6.3 to 5.2, and for VA chiropractic visits decreased from 6.3 to 5.6 (not shown in tables). This pattern was similar for pain severity (VA acupuncture: 5.9 to 5.0, VA chiropractic: 5.9 to 5.1). Conversely, a standard episode of care for CC acupuncture saw a decrease in pain interference from 6.3 to 6.1, and in pain severity from 5.9 to 5.4. In CC chiropractic we found a predicted decrease in pain interference from 6.3 to 5.9 and pain severity from 5.9 to 5.2.
Difference in Pain Interference and Pain Severity Score (Reference Group: 0 Visits), by VA and CC Across 8 Month Study Period
Bold = P < 0.05.
Discussion
This study compared pain interference and severity for veterans following acupuncture, chiropractic, and medical massage therapy visits to VA and CC providers. Overall, we found that VA acupuncture and chiropractic had a stronger beneficial relationship with pain than CC acupuncture and chiropractic, while conversely CC medical massage therapy had a stronger beneficial relationship than VA medical massage therapy. Extrapolating to 12 visits of acupuncture therapy at the VA resulted in clinically meaningful improvements in pain interference.
While findings may demonstrate that VA does some things better than CC, and vice-versa, this almost certainly depends on who is going to which facility, the comparability of people who tend towards the VA and CC, and the quality of care across clinics. For example, prior studies have shown that VA outpatient clinicians provide similar or better patient experiences than CC. 13 O’Hanlon also points to some studies showing that outcomes were similar for patients receiving care from VA and CC, and they saw no differences in outcomes particularly when CC utilization was high (ie, they did not adjust for differences in utilization); we tried to address this by adjusting for utilization for 1 (VA or CC) while looking at the other. Because the VA uses a whole health approach to care rather than focusing just on a single chief complaint or health problem, patients may be connected with other services and receive help with physical, psychological, social, and economic factors that impact wellbeing. 21 In addition to a particular CIH therapy to treat pain, veterans have access to many other services through the VA and may simultaneously be engaged in other modalities of care that can also improve their pain. While we attempted to account for this by adjusting for whether or not a participant was engaged in other CIH therapies, this does not capture the full spectrum of relevant factors at play.
When examining preset cut points for therapy visits, we found that 5 to 12 visits of acupuncture at VA sites appeared to give the greatest improvements in pain interference and pain severity. Veterans who did not receive more than 4 visits may not have utilized enough acupuncture for there to be a perceptible improvement in their pain. It is also possible that some veterans may have started their care late in the 8-month evaluation period (eg, within days of the 6-month survey follow-up period), before one could see effects on pain interference or severity. These estimates also could indicate that those who are receiving more than 13 acupuncture visits may be a subgroup of chronic pain patients for whom acupuncture is not an effective therapy or need a high dose of acupuncture for benefit, however there are many alternative explanations – including several limitations to the data and study design described below. Because the association of acupuncture with pain interference and severity was not seen from CC until 13 or more visits, there may be other facets of VA care besides the acupuncture itself that are also affecting pain outcomes. For chiropractic we found that veterans with 5-8 VA visits had statistically significantly lower pain interference, similar to those with 5-8 CC visits (−0.42 and −0.38, respectively).
Findings may indicate that a higher number of acupuncture visits were needed to see an association with pain interference. Chiropractic cases in VA are for musculoskeletal reasons only. Acupuncture referrals may be for musculoskeletal diagnoses, internal medicine or mental health diagnoses. It is possible that the diagnosis for which a person is referred to these different therapies could account for this variation. In addition, patients are often referred to acupuncture at a later point in the care continuum. These patients tend to be more complex and have a greater number of chronic health conditions at the point of acupuncture initiation.
VA medical massage therapy had the strongest magnitude of association of the 3 therapies on pain interference, with more visits trending towards more benefit. Medical massage therapy delivered through the VA may differ from what is offered through community care, as coding used for medical massage therapy can indicate delivery by other providers with soft tissue mobilization in scope such as chiropractors, physical therapists, and other providers; the type of medical provider can impact the type of massage therapy, which in turn may moderate the effect of the therapy on pain. We did not have the data to differentiate medical massage therapy visits across different provider types, but future studies may consider this as a potentially fruitful line of inquiry. In addition, medical massage therapy was the least used of the 3 therapies, and the CIH approach which has the fewest staff providers within VA delivering the service, making estimates for these effects less precise.
Our findings should be viewed in light of several limitations. Because the study did not randomly assign CIH therapy, estimates are likely confounded by imbalances in important factors between participants that influence both therapy use and chronic pain. In combination with a sparse amount of data for higher numbers of CIH therapy visits, we run the risk of positivity violations that make it difficult to address confounding even with some of our observed covariates. This may be particularly true for medical massage therapy, given that the VA had only developed a qualification standard for medical massage therapists in 2019 and the COVID-19 pandemic slowed the hiring of massage therapists during the study period. This study took place during the COVID-19 pandemic (March 2021 through March 2023), which may have impacted the results in unknown ways. For example, we do not know how clinician recommendations for self-care and CIH therapies may have changed as a result of the pandemic, and whether these changes differed between the VA and community care clinics. However, during most of the survey period, VA facilities had identified pathways and protocols for patients to receive care that required in-person contact. Moreover, data from the VA electronic health record shows that rates of acupuncture, chiropractic care and massage therapy increased during this period relative to pre-COVID due to expansion of these services. The response rate for follow-up surveys was low (24% at 6 months). The majority of our sample had missing data for 1 or both of the outcomes, and 2 participants were missing data on their age. Because nearly all missing data was in the patient-reported outcome measures, many of the modern approaches to address this (eg, multiple imputation, full information maximum likelihood) would not substantially affect the estimate values from regression models if data were missing at random; however, estimates would still be biased if the probability of having missing outcome data is related to the missing outcome data itself. These results also do not inform whether CIH therapy visits preceded pain measurements or vice-versa; those with a higher number of therapy visits, particularly at the highest count (of 13 or more), may be receiving more visits to address more severe pain or worse health problems overall. The use of other CIH therapies concurrently may impact pain; while we tried to account for this by adjusting for whether someone was using a therapy other than the 1 of interest, we did not differentiate those therapies nor quantify how many different other therapies they were using. While this study tested pre-determined cut points for doses of CIH therapies, there remains a question of optimal dose to achieve clinically meaningful improvement in chronic pain. Future studies should consider dose-ranging trial designs to determine minimally effective dose, with consideration that optimal dosages will likely vary by the type of CIH therapy, the conditions to treat, the chosen ways of measuring conditions, and patient characteristics.
We did not investigate whether the effects of these and other CIH therapies may differ for women veterans, who experience a higher proportion of chronic pain and pain-related health problems compared to men 22 and are more likely to use CIH services at the VA. At the same time, women veterans may worry about experiencing sexual harassment at VA clinics which may have implications for their experiences undertaking these therapies in the VA. Future research should investigate potential moderation of the use and effects of these therapies for women veterans in the VA and in community care. In addition, we did not examine potential moderation between urban and rural residents. Given the MISSION Act was intended, in part, to increase access to care for veterans living far away from VA clinics (and rural veterans tend to live further away than urban veterans), this remains an important question to gauge effectiveness of the policy for increasing access to care, including to CIH therapies for pain.
CIH therapies delivered through the VA and through community care both offer potentially effective means of reducing chronic pain. The VA and CC provide different avenues for veterans to receive treatment for chronic pain, and their effectiveness may vary depending on the type of CIH therapy being received.
Footnotes
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This evaluation was funded as a quality improvement project by the Office of Patient Centered Care and Cultural Transformation and VA QUERI program (PEC 16-354), 
. Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the U.S. Department of Veterans Affairs, or United States Government.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.