Abstract
Background
Acupuncture is used globally as an alternative treatment for patients with low back pain (LBP), effectively reducing pain and improving physical activity. However, the use and profile of acupuncture for LBP in middle-aged and older adults remains understudied. The aim of this study was to identify key factors associated with the use of acupuncture treatment in this population.
Methods
Using the 2018 China Health and Retirement Longitudinal Study (CHARLS) dataset, a cross-sectional analysis of 7,929 respondents aged 45 years and older with LBP was conducted. A two-way stepwise regression model was used to identify significant correlates of demographics, health status, and healthcare choice factors with acupuncture use, and multiple theory-driven interaction terms were introduced into the stepwise model and assessed for their enhancement of model fitting and predictive performance by likelihood ratio tests with AUC comparisons. In addition, stratified analyses by gender, age (45–59, 60–74, and ≥ 75 years), and residence were conducted to verify the stability of the model findings.
Results
Among 7,929 respondents with LBP from the CHARLS 2018 dataset, 1,097 (13.8%) comprised the acupuncture group while 6,832 (86.2%) formed the non-acupuncture group. Significant correlates included females (OR = 1.28, 95% CI: 1.09–1.49), pain distress level (OR = 1.54, 95% CI: 1.32–1.78), massage (OR = 8.81, 95% CI: 7.53–10.32), herbal medicine (OR = 2.65, 95% CI. 95% CI: 2.28–3.09), western medications (OR = 1.92, 95% CI: 1.65–2.24).
Conclusion
LBP patients with certain characteristics tend to use complementary and alternative interventions, including acupuncture. In the future, it is necessary to combine the evaluation of efficacy and mechanism studies to clarify the optimal combination of multimodal treatment and the applicable population, in order to promote the use of acupuncture in the treatment of LBP.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12906-025-05086-4.
Keywords: Low back pain, Acupuncture, CHARLS, Cross-sectional study, Logistic regression, Healthcare utilization
Introduction
Low back pain (LBP) encompasses a range of often overlapping pain types, such as nociceptive, neuropathic, and nonspecific pain. Anatomically, it refers to pain or discomfort in the area below the 12th rib level and above the gluteal folds, which may or may not be accompanied by leg pain [1]. LBP is an extremely common symptom experienced by people of all ages and is the result of the dynamic interaction between social, psychological, and biological factors [2].
The burden of disease reports from China and its provinces ranked LBP as the leading cause of years lived with disability from 2005 to 2016 [3]. The annual prevalence of LBP among Chinese adults ranges from 20.88–29.88% [4], with higher rates observed in females compared to males [3]. The loss of productivity and medical expenses due to LBP is substantial, making it a pressing global public health issue.
Acupuncture is globally used as an alternative therapy for patients with LBP, effectively reducing pain and improving daily functional impairments, thus enhancing patients’ quality of life and mental health [5]. Guidelines from the American College of Physicians and the American Pain Society recommend acupuncture for the treatment of LBP [6, 7]. Acupuncture has demonstrated clinical efficacy in the management of both acute and chronic LBP, with existing evidence indicating an absence of significant adverse effects [5, 8, 9]. Furthermore, comparative clinical studies have shown this therapeutic modality to exhibit superior efficacy and a more favorable safety profile compared to conventional oral pharmacotherapy [10]. It is often combined with other non-invasive treatments, such as chiropractic and physical therapy, to enhance its effectiveness. In the treatment of chronic LBP, acupuncture can improve pain and functional impairment in the short term, whether used alone or as an adjunct to conventional therapy [11].
Previously, a prospective cohort study in Germany reported that female patients, those with chronic pain, low functional capacity, and older individuals with LBP were more likely to receive acupuncture treatment [12]. In the UK, LBP is the most common reason among patients opting for acupuncture [13]. Additionally, regardless of the extent of acupuncture use and its perceived benefits within the healthcare system, there is a general trend towards complementary medicine [14]. A nationwide retrospective cohort study in Korea involving 130,089 participants demonstrated that acupuncture treatment was associated with a reduced rate of lumbar surgery among LBP patients [15]. However, the factors influencing patients’ choice to use acupuncture have not been fully explored. Given the prevalence and significance of acupuncture treatment among LBP patients in China, it is necessary to identify the factors associated with acupuncture use in middle-aged and older Chinese patients with LBP. To fill this gap, this study aims to analyze the prevalence of acupuncture use and the factors influencing the choice of acupuncture treatment among middle-aged and older patients with LBP.
Materials and methods
Study population
The China Health and Retirement Longitudinal Study (CHARLS) is a comprehensive interdisciplinary survey project led by the National School of Development at Peking University and executed by the China Social Science Survey Center. Aiming to analyze population aging issues and promote interdisciplinary research on aging in China, CHARLS developed a high-quality dataset containing the social, economic, and health status of Chinese residents aged 45 years and above and their family members through interviews. CHARLS follows a baseline sample tracking research model, which involves a nationwide baseline survey in 2011 and subsequent biennial follow-ups, which included 17,706 Chinese middle-aged and elderly respondents in 28 provinces (autonomous regions and municipalities directly under the central government), 150 counties, 450 communities (villages), and 10,257 households. Details of CHARLS have been published in previous literature [16].
This study is a secondary analysis of 2018 data from the CHARLS. Respondents aged 45 years and older who reported LBP were selected, as detailed in Fig. 1. The 121 excluded observations were not retained during the initial data processing. However, a sensitivity analysis based on simulated data suggested that their exclusion had no substantial impact on the final model results. The CHARLS protocol had received permission from the Biomedical Ethics Committee of Peking University, and all respondents provided informed consent at the time of participation (IRB00001052-11015).
Fig. 1.
Flow chart for study population selection. “Respondents” are defined as Chinese residents aged 45 and over and their spouses who participated in the CHARLS survey
Data collection
With reference to previous studies [17–19]the following data were collected for this study:(i) Sociodemographic characteristics: age, gender, marital status, residence, education level, smoking status, alcohol status, work status, insurance status. (ii) Health status: general health status, suffering from chronic disease, depressive symptoms, pain distress level, location of pain, difficulty in carrying weights over 5 kg, difficulty in stooping, kneeling or crouching, night sleep time, social activity. (iii) Health services choices: Whether the patient sought outpatient care in the past month (General hospital, Specialized hospital, Chinese medicine hospital, Primary care institutions). Whether the patient sought inpatient care in the last year(General hospital, Specialized hospital, Chinese medicine hospital, Primary care institutions). Take the following measures to reduce the pain: Herbal medicine, Western medications, Acupuncture, Massage or Other. Take OTC drugs during the past month. Respondents with self-reported back, waist, and buttock pain were defined as LBP patients according to the anatomical definition.
Outcome measures
Respondents were asked to report whether they were currently taking acupuncture treatment to reduce the pain, with a yes or no response.
Handling of missing data
Before the data analysis, we conducted a missing value analysis on the dataset. Figure S1 illustrates the distribution of variables with missing data in our study. Table S1 shows the variables with missing values and their proportions. To preserve the data structure and relationships between variables as much as possible, we applied multiple imputation by chained equations (MICE) using the mice package in R. The imputation model incorporated all analysis variables (demographics, health behaviors, pain-related outcomes) as predictors, and with the exception of the ‘general health status’ variable, which had a high missing rate (5.91%), the missing rates for the other variables were low, mostly less than 1%, and therefore it was reasonable to assume missing at random (MAR) assumption. Binary variables were imputed via logistic regression. Ordinal variables used proportional odds models (polr). Continuous variables employed predictive mean matching to retain distributional properties. Generate 5 interpolated datasets with 50 iterations per chain. Due to the high rate of missingness of the variable “general health status” (5.91%), a missing not at random (MNAR) sensitivity analysis was conducted to assess the impact of different missingness mechanisms on the model estimation (Supplementary materials 1). The remaining variables did not require additional MNAR sensitivity analyses due to the very small percentage of missing at random and were treated with multiple interpolation.
Statistical analysis
The study used acupuncture (Yes/No) as a binary dependent variable. The independent variables included: Continuous variables: age; Binary variables: gender, marital status, residence area, work status, insurance status, suffering from chronic disease, social activity, depressive symptoms, and multiple medical providers and treatments; Ordinal variables: smoking, drinking, health status, night sleep time, difficulty in carrying weights over 5 kg, difficulty in stooping, kneeling, or crouching, and pain distress level. Nominal Variables: location of pain. To compare the sociodemographic, health status, and health service use characteristics between the acupuncture treatment group and the non-acupuncture treatment group, we employed the following statistical methods: Chi-square test and Fisher’s exact test for categorical variables where appropriate, and independent samples t-test for comparing mean differences in continuous variables.
Given the study involves 7,929 observations and 32 variables, a bidirectional stepwise logistic regression analysis was conducted to explore factors influencing acupuncture use, considering interactions between variables and balancing model complexity and explanatory power. Ordered categorical variables (e.g., pain distress level) were tested for linear, quadratic, and cubic trends in the analyses by orthogonal polynomial comparisons, and the age variable was standardised with a Z-score for interpretation. To test for a nonlinear relationship between age and the probability of acupuncture use, quadratic and cubic terms were added to the model for age, respectively (Supplementary materials 2). The model retained all trend terms to fully assess potential nonlinear relationships. Multicollinearity tests showed all variables had variance inflation factor (VIF) values very close to 1 (Table S2), which indicating minimal multicollinearity among these variables in the model. All variables with bivariate P-values < 0.2 were included to identify the most parsimonious model for predicting acupuncture use among middle-aged and older patients with LBP. Model fit was assessed using the Hosmer-Lemeshow test.
Interaction analysis
Based on the multivariate logistic regression model, we further considered the interactions between key variables and explored whether there were effect modifications of different predictors in specific populations. Specifically, we introduced a number of potential interaction terms, such as “western medications × massage”, “pain distress level × general health status”, into the final model after stepwise regression screening to assess whether there were interactive moderating effects of treatment modality, somatic functional status, and other factors on the choice of acupuncture treatment.
All interaction terms were selected based on the principles that (1) there was a possible association in medical or behavioural mechanisms; (2) the main effect was included in the regression model and was statistically significant; (3) Sufficient sample support at the data level. After the interaction terms were included, the goodness of fit between the models with and without the interaction terms was compared using the likelihood ratio test (LRT), and changes in predictive performance were assessed by calculating the AIC value and the area under the ROC curve (AUC) for each model. To avoid the problem of multicollinearity due to the introduction of interaction terms, we used the GVIF for diagnostics and ensured that all GVIF^(1/(2×df)) < 2 (Table S3). The final model retained the interaction terms that improved in both statistical significance and model performance (Table S4).
Stratified analyses
To evaluate the stability of the regression model results across different populations, we stratified the analysis based on gender and age (dividing age into three groups: 45–59, 60–74, and ≥ 75), and residence. In the data preprocessing stage, we first used the cut function to generate age stratification variables. However, it should be noted that in each stratified subset, some factor variables (such as “gender” and “age group” in the original data set) only retained one level in a single subgroup, so the data cleaning strategy (including eliminating factors with only a single level) was adopted before model construction. This preprocessing method ensures that the factor variables are correctly coded in modeling, but may also lead to different independent variables included in each subset. To ensure the robustness of the hierarchical model, we used two-way stepwise regression in each subset to further screen out the key factors affecting the selection of acupuncture.
All statistical analyses were performed using R software (version 4.4.0), with a significance level set at α = 0.05.
Results
Among the 7,929 middle-aged and older patients with LBP, 1,097 (13.8%) used acupuncture for treatment. Table 1 presents a binary comparison of the sociodemographic characteristics between middle-aged and older patients with LBP who used acupuncture and those who did not. Acupuncture users were predominantly female, lived in non-urban areas, did not smoke, had affected work status, had insurance, and were relatively younger (P < 0.05).
Table 1.
Associations between acupuncture use and demographic characteristics among Chinese middle-aged and older adults with LBP, using chi-square tests for categorical variables and t-test for continuous variables
| Acupuncture treatment use | P-value | ||
|---|---|---|---|
| No(n = 6832), n(%) | Yes(n = 1097), n(%) | ||
| Gender | <0.001 | ||
| Female | 4136 (60.5) | 758 (69.1) | |
| Male | 2696 (39.5) | 339 (30.9) | |
| Marital status | 0.669 | ||
| Married/partnered | 5688 (83.4) | 921 (84.0) | |
| Married/spabs/separated/divorced/widowed/never married | 1135 (16.6) | 176 (16.0) | |
| Residence | 0.023 | ||
| Urban | 2395 (35.1) | 424 (38.7) | |
| Non-urban | 4437 (64.9) | 673 (61.3) | |
| Education level | 0.819 | ||
| No formal education | 1957 (28.6) | 316 (28.8) | |
| Elementary school | 3125 (45.7) | 498 (45.4) | |
| Middle/high/vocational school | 1673 (24.5) | 267 (24.3) | |
| College degree or higher | 77 (1.1) | 16 (1.5) | |
| Smoking status | < 0.001 | ||
| No | 4258 (62.3) | 756 (68.9) | |
| Abstinence | 890 (13.0) | 124 (11.3) | |
| Yes | 1684 (24.6) | 217 (19.8) | |
| Alcohol status | 0.128 | ||
| No | 4752 (69.6) | 784 (71.5) | |
| Rarely | 460 (6.7) | 82 (7.5) | |
| Often | 1620 (23.7) | 231 (21.1) | |
| Work status | |||
| Affected* | 3316 (48.6) | 633 (57.8) | < 0.001 |
| Not affected | 3506 (51.4) | 462 (42.2) | |
| Insurance status | 0.007 | ||
| Yes | 6622 (96.9) | 1080 (98.5) | |
| No | 210 (3.1) | 17 (1.5) | |
| Age, Mean(SD) | 62.2(10.2) | 61.2(9.6) | 0.004 |
SD Standard deviation
*Affected: Inability to work or significant productivity loss due to LBP or other health conditions
As shown in Table 2, the acupuncture group reported slightly higher proportions of “very poor” and “poor” health status, and slightly lower proportions of “good” and “very good” health status. The proportion of patients with at least one chronic disease was slightly higher in the acupuncture group (58.5% vs. 52.4%), with a significant difference (P < 0.001). The proportion of patients without depressive symptoms was lower in the acupuncture group (55.2% vs. 62.0%), with a significant difference (P < 0.001). There was a significant difference in the level of pain distress between the two groups (P < 0.001). The acupuncture group reported a higher proportion of being “quite a bit/very”. The proportion of patients reporting pain in multiple areas including “Back-Waist-Buttocks” was higher in the acupuncture group (20.1% vs. 12.6%). There was a significant difference in the difficulty of lifting objects heavier than 5 kg (P < 0.001), with a slightly higher proportion of the acupuncture group reporting “Can not do” (18.3% vs. 13.4%). The difficulty level of bending, kneeling, or squatting showed significant differences (P < 0.001), and the acupuncture group reported a higher proportion of difficulties or inability to complete these tasks. The proportion of patients participating in social activities was significantly lower in the acupuncture treatment group compared to the non-treatment group (P < 0.001).
Table 2.
Associations between acupuncture use and health status, by Chinese middle-aged and older adults with LBP
| Acupuncture treatment use | ||
|---|---|---|
| No(n = 6832), n(%) | Yes(n = 1097), n(%) | |
| General health status | ||
| Very poor | 635 (9.3) | 163 (14.9) |
| Poor | 2061 (30.2) | 390 (35.6) |
| Fair | 3306 (48.4) | 463 (42.2) |
| Good | 505 (7.4) | 51 (4.6) |
| Very good | 325 (4.8) | 30 (2.7) |
| Suffering from chronic disease* | ||
| Yes | 3583(52.4) | 642 (58.5) |
| No | 3249(47.6) | 455(41.5) |
| Depressive symptoms | ||
| Yes | 2595(38.0) | 492(44.8) |
| No | 4237(62.0) | 605(55.2) |
| Pain distress level | ||
| A little | 3146(46.0) | 347 (31.6) |
| Somewhat | 1403(20.5) | 207 (18.9) |
| Quite a bit | 1271(18.6) | 266 (24.2) |
| Very | 1012 (14.8) | 277 (25.3) |
| Location of pain | ||
| Back | 441 (6.5) | 55 (5.0) |
| Waist | 3155(46.2) | 404 (36.8) |
| Buttocks | 129 (1.9) | 12 (1.1) |
| Back-Waist | 1770(25.9) | 308 (28.1) |
| Back-Buttocks | 52 (0.8) | 7 (0.6) |
| Waist-Buttocks | 422 (6.2) | 90 (8.2) |
| Back-Waist-Buttocks | 863 (12.6) | 221 (20.1) |
| Difficulty in carrying weights over 5Kg | ||
| No | 5114(74.9) | 745(67.9) |
| Can still do | 650(9.5) | 115(10.5) |
| Need help | 150(2.2) | 36(3.3) |
| Can not do | 918(13.4) | 201(18.3) |
| Difficulty in stooping, kneeling or crouching | ||
| No | 3281(48.0) | 404(36.8) |
| Can still do | 2266(33.2) | 415(37.8) |
| Need help | 208 (3.0) | 44(4.0) |
| Can not do | 1077(15.8) | 234(21.3) |
| Night sleep time | ||
| < 7 | 4352 (63.7) | 735 (67.0) |
| 7 ~ 9 | 2183 (32.0) | 323 (29.4) |
| 9~ | 297 (4.3) | 39 (3.6) |
| Social activity | ||
| No | 4622 (67.6) | 670 (61.3) |
| Yes | 2210 (32.4) | 427 (38.7) |
*Suffering from at least one of the 14 chronic diseases: hypertension, dyslipidemia, high blood sugar, malignancy, lung disease, liver disease, heart disease, Stroke, Kidney disease, emotional and mental problems, memory-related disorders, and asthma
Table 3 outlines the associations between healthcare service choices and acupuncture use. In terms of seeking outpatient and inpatient treatment, patients’ choice of general hospital, specialized hospital, Chinese medicine hospital, and Primary care institutions was significantly related to choosing acupuncture treatment (P < 0.001). Among the treatment modalities, the acupuncture group might prefer to combine OTC medication/other treatment for pain relief (75.48% vs. 66.96%, 23.79% vs. 19.39%). The acupuncture group used herbal medicine (56.06% vs. 17.52%), massage (53.24% vs. 8.24%) were both significantly higher than the non-acupuncture group. The proportion of western medications was significantly higher in the non-acupuncture group (45.78% vs. 31.54%) than in the acupuncture group.
Table 3.
Associations between acupuncture use and health services choices, by Chinese middle-aged and older adults with LBP
| Acupuncture treatment use | P-value | |||
|---|---|---|---|---|
| No(n = 6832), n(%) | Yes(n = 1097), n(%) | |||
| Outpatient Care | ||||
| General hospital | Yes | 482(7.06) | 119(10.84) | < 0.001 |
| No | 6350(92.94) | 978(89.16) | ||
| Specialized hospital | Yes | 47(0.69) | 21(1.91) | < 0.001 |
| No | 6785(99.31) | 1076(98.09) | ||
| Chinese medicine hospital | Yes | 86(1.26) | 35(3.19) | < 0.001 |
| No | 6746(98.74) | 1062(96.81) | ||
| Primary care institutions* | Yes | 731(10.7) | 193(17.59) | < 0.001 |
| No | 6101(89.3) | 904(82.41) | ||
| Inpatient Care | ||||
| General hospital | Yes | 239(3.5) | 70(6.38) | < 0.001 |
| No | 6593(96.5) | 1027(93.62) | ||
| Specialized hospital | Yes | 27(0.4) | 11(1) | 0.014 |
| No | 6805(99.6) | 1086(99) | ||
| Chinese medicine hospital | Yes | 33(0.48) | 17(1.55) | < 0.001 |
| No | 6799(99.52) | 1080(98.45) | ||
| Primary care institutions * | Yes | 94(1.38) | 32(2.92) | < 0.001 |
| No | 6738(98.62) | 1065(97.08) | ||
| OTC drugs | Yes | 4575(66.96) | 828(75.48) | < 0.001 |
| No | 2257(33.04) | 269(24.52) | ||
| Herbal medicine | Yes | 1197(17.52) | 615(56.06) | < 0.001 |
| No | 5635(82.48) | 482(43.94) | ||
| Western medications | Yes | 3128(45.78) | 346(31.54) | < 0.001 |
| No | 3704(54.22) | 751(68.46) | ||
| Massage | Yes | 563(8.24) | 584(53.24) | < 0.001 |
| No | 6269(91.76) | 513(46.76) | ||
| Other treatments | Yes | 1325(19.39) | 261(23.79) | 0.033 |
| No | 5507(80.61) | 836(76.21) | ||
*Primary care institutions: Community healthcare center, Township hospital, Health care post, Nursing home, Other. Outpatient Care also included Village clinic/Private clinic
Table 4 presents the results of a logistic regression analysis identifying the statistically significant predictors of acupuncture treatment use. The Hosmer and Lemeshow goodness-of-fit statistic for this regression model indicated that the model was appropriate (P = 0.701).
Table 4.
Logistic regression identifying the statistically significant predictors of acupuncture use
| Predictors of Acupuncture Treatment Use | Odds ratio | 95%C.I | P-value |
|---|---|---|---|
| Gender | |||
| Male | 1 | 0.002 | |
| Female | 1.28 | 1.09, 1.49 | |
| Insurance status | |||
| No | 1 | 0.037 | |
| Yes | 1.76 | 1.07, 3.11 | |
| Social activity | |||
| No | 1 | 0.024 | |
| Yes | 1.19 | 1.02, 1.38 | |
| Outpatient Care | |||
| General hospital | |||
| No | 1 | 0.074 | |
| Yes | 1.26 | 0.98, 1.61 | |
| Specialized hospital | |||
| No | 1 | 0.046 | |
| Yes | 1.88 | 0.99, 3.43 | |
| Chinese medicine hospital | |||
| No | 1 | 0.011 | |
| Yes | 1.84 | 1.14, 2.92 | |
| Primary care institutions | |||
| No | 1 | < 0.001 | |
| Yes | 1.51 | 1.23, 1.85 | |
| Herbal medicine | |||
| No | 1 | < 0.001 | |
| Yes | 2.65 | 2.27, 3.09 | |
| Western medications | |||
| No | 1 | < 0.001 | |
| Yes | 1.92 | 1.65, 2.24 | |
| Massage | |||
| No | 1 | < 0.001 | |
| Yes | 8.81 | 7.53, 10.32 | |
| Difficulty in stooping, kneeling or crouching | |||
| Linear | 1.34 | 1.12, 1.58 | < 0.001 |
| Quadratic | 0.89 | 0.72, 1.11 | 0.291 |
| Cubic | 0.97 | 0.75, 1.27 | 0.823 |
| Pain distress level | |||
| Linear | 1.54 | 1.32, 1.78 | < 0.001 |
| Quadratic | 1.05 | 0.91, 1.22 | 0.509 |
| Cubic | 0.96 | 0.82, 1.21 | 0.627 |
| Age | 0.89 | 0.82, 0.96 | 0.004 |
Among demographic and social factors, females demonstrated 28% higher odds of acupuncture utilization compared to males (OR = 1.28, 95% CI: 1.09–1.49; P = 0.002). Insured individuals had 1.76 times higher odds of chosing acupuncture than uninsured individuals (OR = 1.76, 95% CI: 1.07–3.11; P = 0.037). Participation in social activities was positively associated with acupuncture use (OR = 1.19, 95% CI: 1.02–1.38; P = 0.024). Older age marginally reduced the likelihood of acupuncture use (OR = 0.89, 95% CI: 0.82–0.96; P = 0.004).
In terms of healthcare utilisation patterns, mainly focusing on outpatient care, patient visits to Chinese medicine hospital (OR = 1.84, 95% CI: 1.14–2.92; P = 0.011) or Primary care institutions (OR = 1.51, 95% CI: 1.23–1.85; P < 0.001) were associated with a significantly associated with increased acupuncture use. General hospital (OR = 1.26, P = 0.074) had no significant effect, and specialized hospital (OR = 1.88, P = 0.046) were at the borderline of significance. Concurrent use of herbal medicine (OR = 2.65, 95% CI: 2.27–3.09; P < 0.001) or western medications (OR = 1.92, 95% CI: 1.65–2.24; P < 0.001) was strongly predictive of acupuncture utilization. The strongest predictor was massage therapy, with users demonstrating 8.81-fold higher odds of selecting acupuncture (OR = 8.81, 95% CI: 7.53–10.32; P < 0.001).
For health status and physical function limitations, difficulty in stooping, kneeling, or crouching showed a linear association with acupuncture use (OR = 1.34, 95% CI: 1.12–1.58; P < 0.001), though quadratic and cubic terms were non-significant (P > 0.05). Higher levels of physical pain distress were also linearly associated with increased acupuncture adoption (OR = 1.54, 95% CI: 1.32–1.78; P < 0.001), with no evidence of non-linear trends. The results suggest that as the level of functional limitation and pain distress increases, the tendency of patients to choose acupuncture treatment may also rise.
Revised model with interaction terms
The inclusion of interaction terms one by one revealed two notable findings (Table S4). First, a synergistic effect was observed between western medications and massage therapy: individuals adopting both modalities had 61% higher odds of selecting acupuncture compared to those using either therapy alone (OR = 1.61, 95% CI: 1.17–2.23; P = 0.004). Second, among patients with severe functional limitations (inability to stoop and difficulty carrying 5 kg items), acupuncture utilization was significantly elevated (OR = 1.61, 95% CI: 1.15–2.24; P ≈ 0.005). However, the overall model improvement was marginal (Figure S2, ΔAUC = + 0.002), and most interaction terms showed non-significant or inconsistently directional effects (all P > 0.05 unless specified).
Comparison of the Stepwise model with the stratified model
The key variables (e.g., massage, herbal medicine, western medications, etc.) obtained by stepwise regression screening in the original model all showed significant positive associations in the full sample. Stratified analyses showed that the direction and magnitude of effects of these core variables were generally consistent across subsets of different age groups, genders, and places of residence.
The OR for massage was approximately 10.21 in the 45–59 year old group, and the OR for massage remained high in the 60–74 and ≥ 75 year old groups, respectively (Table 5); in the gender stratification, in both the female and male groups, herbal medicine and western medications showed positive associations consistent with the stepwisel model-consistent positive effects (Table 6). Although the stratified models differed somewhat in the combination of independent variables included due to differences in sample size and distribution of factor variables within subgroups, the overall stability of the effects of the main predictors was good. The direction of the effects of the core predictors (pain distress level, massage, herbal medicine, and western medications) remained consistent across the gender, age, and residence strata (Tables 5, 6 and 7), supporting the robustness of the regression model: there was a strong tendency to use acupuncture among patients who chose complementary therapies (e.g., massage, herbal medicine) and who reported more severe pain or dysfunction. Demographic differences, particularly gender and insurance coverage, further shaped the utilisation patterns.
Table 5.
Logistic regression results by age group
| 45–59 (n = 3307) | 60–74 (n = 3574) | >=75 (n = 985) | ||||
|---|---|---|---|---|---|---|
| OR (95% CI) | P-value | OR (95% CI) | P-value | OR (95% CI) | P-value | |
| Gender | ||||||
| Male | 1 | 0.045 | 1 | 0.020 | - | - |
| Female | 1.28 (1.01, 1.64) | 1.35 (1.05, 1.75) | - | |||
| Marital status | ||||||
| Married/partnered | 1 | 0.135 | - | - | - | - |
| Married, spabs/separated/divorced/widowed/never married | 1.38 (0.89, 2.08) | - | - | |||
| Work status | ||||||
| Affected | 1 | 0.067 | - | - | - | - |
| Not affected | 1.26 (0.98, 1.60) | - | - | |||
| Social activity | ||||||
| No | 1 | 0.006 | - | - | 1 | 0.143 |
| Yes | 1.39 (1.10, 1.75) | - | 1.42 (0.88, 2.27) | |||
| Difficulty in carrying weights over 5Kg | ||||||
| Linear | 1.62 (1.15, 2.26) | 0.005 | - | - | - | - |
| Quadratic | 0.94 (0.62, 1.47) | 0.795 | - | - | - | - |
| Cubic | 0.79 (0.47, 1.34) | 0.362 | - | - | - | - |
| Pain distress level | ||||||
| Linear | 1.60 (1.25, 2.05) | < 0.001 | 1.59 (1.28, 1.97) | < 0.001 | - | - |
| Quadratic | 0.96 (0.76, 1.22) | 0.741 | 1.19 (0.96, 1.49) | 0.119 | - | - |
| Cubic | 1.16 (0.91, 1.47) | 0.233 | 0.81 (0.63, 1.02) | 0.079 | - | - |
| Outpatient - General hospital | ||||||
| No | 1 | 0.067 | - | - | - | - |
| Yes | 1.39 (0.97, 1.98) | - | - | |||
| Outpatient - Chinese medicine hospital | ||||||
| No | 1 | 0.036 | - | - | 1 | 0.05 |
| Yes | 2.10 (1.03, 4.12) | - | 3.67 (0.93, 13.08) | |||
| Outpatient - Primary care institutions | ||||||
| No | 1 | 0.038 | 1 | 0.002 | - | - |
| Yes | 1.42 (1.01, 1.96) | 1.57 (1.17, 2.09) | - | |||
| Inpatient - Chinese medicine hospital | ||||||
| No | 1 | 0.102 | - | - | - | - |
| Yes | 3.68 (0.77, 18.39) | - | - | |||
| Herbal medicine | ||||||
| No | 1 | < 0.001 | < 0.001 | 1 | < 0.001 | |
| Yes | 2.76 (2.17, 3.51) | 2.56 (2.04, 3.20) | 3.09 (1.92, 4.96) | |||
| Western medications | ||||||
| No | 1 | < 0.001 | < 0.001 | 1 | 0.104 | |
| Yes | 2.02 (1.59, 2.56) | 2.05 (1.63, 2.59) | 1.49 (0.93, 2.41) | |||
| Massage | ||||||
| No | 1 | 1 | < 0.001 | |||
| Yes | 10.21 (8.02, 13.04) | < 0.001 | 8.61 (6.80, 10.92) | < 0.001 | 10.81 (6.43, 18.32) | |
| Other treatments | ||||||
| No | - | - | 1 | 0.039 | 1 | 0.069 |
| Yes | - | 1.31 (1.01, 1.68) | 0.57 (0.30, 1.02) | |||
| Alcohol status | ||||||
| Linear | - | - | 1.23 (0.99, 1.51) | 0.06 | 0.66 (0.39, 1.07) | 0.109 |
| Quadratic | - | - | 0.77 (0.55, 1.09) | 0.126 | 0.44 (0.22, 0.94) | 0.025 |
| Insurance status | ||||||
| No | - | - | 1 | 0.073 | - | - |
| Yes | - | 2.17 (1.00, 5.52) | - | |||
| Difficulty in stooping, kneeling or crouching | ||||||
| Linear | - | - | 1.46 (1.16, 1.85) | 0.002 | - | - |
| Quadratic | - | - | 0.76 (0.57, 1.02) | 0.064 | - | - |
| Cubic | - | - | 0.85 (0.60, 1.22) | 0.368 | - | - |
| Location of pain | ||||||
| Waist | 1 | |||||
| Back | - | - | - | - | 0.53 (0.16, 1.44) | 0.252 |
| Buttocks | - | - | - | - | 0.00 (NA) | 0.982 |
| Back-Waist | - | - | - | - | 0.92 (0.52, 1.59) | 0.762 |
| Back-Buttocks | - | - | - | - | 0.00 (NA) | 0.992 |
| Waist-Buttocks | - | - | - | - | 0.40 (0.13, 1.04) | 0.078 |
| Back-Waist-Buttocks | - | - | - | - | 0.94 (0.49, 1.74) | 0.848 |
Table 6.
Logistic regression results by gender
| Female (n = 4894) | Male (n = 3035) | |||
|---|---|---|---|---|
| OR (95% CI) | P-value | OR (95% CI) | P-value | |
| Age | 0.99 (0.98-1.00) | 0.018 | 0.99 (0.97-1.00) | 0.057 |
| Insurance status | ||||
| No | 1 | 0.034 | - | - |
| Yes | 2.06 (1.10–4.23) | - | ||
| Social activity | ||||
| No | 1 | 0.004 | - | - |
| Yes | 1.31 (1.09–1.58) | - | ||
| Difficulty in carrying weights over 5Kg | ||||
| Linear | 1.37 (1.11–1.68) | 0.003 | - | - |
| Quadratic | 0.96 (0.72–1.28) | 0.758 | - | - |
| Cubic | 0.78 (0.55–1.13) | 0.182 | - | - |
| Pain distress level | ||||
| Linear | 1.55 (1.29–1.86) | < 0.001 | 1.45 (1.11–1.88) | 0.006 |
| Quadratic | 1.13 (0.94–1.36) | 0.181 | 0.94 (0.72–1.22) | 0.627 |
| Cubic | 0.91 (0.75–1.10) | 0.336 | 1.05 (0.80–1.37) | 0.720 |
| Outpatient - Chinese medicine hospital | ||||
| No | 1 | 0.011 | - | 0.022 |
| Yes | 2.12 (1.17–3.74) | 1.52 (1.05–2.17) | ||
| Outpatient - Primary care institutions | ||||
| No | 1 | 0.007 | - | < 0.001 |
| Yes | 1.40 (1.09–1.78) | 2.52 (1.94–3.28) | ||
| Herbal medicine | ||||
| No | 1 | < 0.001 | 1 | < 0.001 |
| Yes | 2.74 (2.27–3.32) | 2.52 (1.94–3.28) | ||
| Western medications | ||||
| No | 1 | < 0.001 | 1 | < 0.001 |
| Yes | 1.98 (1.63–2.41) | 1.86 (1.44–2.42) | ||
| Massage | ||||
| No | 1 | < 0.001 | 1 | < 0.001 |
| Yes | 10.29 (8.46–12.54) | 7.45 (5.69–9.76) | ||
| Depressive symptoms | ||||
| No | - | - | 1 | 0.075 |
| Yes | - | 1.27 (0.97–1.65) | ||
| Location of pain | ||||
| Waist | 1 | |||
| Back | 0.98 (0.64–1.47) | 0.925 | - | - |
| Buttocks | 0.49 (0.17–1.20) | 0.148 | - | - |
| Back-Waist | 1.26 (1.01–1.58) | 0.042 | - | - |
| Back-Buttocks | 1.28 (0.42–3.26) | 0.636 | - | - |
| Waist-Buttocks | 1.44 (1.00-2.04) | 0.047 | - | - |
| Back-Waist-Buttocks | 1.44 (1.11–1.87) | 0.006 | - | - |
Table 7.
Logistic regression results by residence
| Non-urban (n = 5110) | Urban (n = 2819) | |||
|---|---|---|---|---|
| OR (95% CI) | P-value | OR (95% CI) | P-value | |
| Age | 0.98 (0.97, 0.99) | 0.001 | - | - |
| Gender | ||||
| Male | 1 | 0.017 | 1 | 0.008 |
| Female | 1.32 (1.05, 1.64) | 1.41 (1.10, 1.82) | ||
| Alcohol status | ||||
| Linear | 1.21 (1.01, 1.45) | 0.034 | - | - |
| Quadratic | 0.88 (0.65, 1.21) | 0.405 | - | - |
| Insurance status | ||||
| No | 1 | 0.076 | - | - |
| Yes | 1.81 (0.98, 3.69) | - | ||
| Work status | ||||
| Not affected | 1 | 0.145 | - | - |
| Affected | 1.17 (0.95, 1.44) | - | ||
| Social activity | ||||
| No | 1 | 0.104 | 1 | 0.109 |
| Yes | 1.18 (0.97, 1.43) | 1.22 (0.96, 1.55) | ||
| Difficulty in stooping, kneeling or crouching | ||||
| Linear | 1.40 (1.13, 1.75) | 0.003 | 1.14 (0.86, 1.50) | 0.368 |
| Quadratic | 0.99 (0.76, 1.31) | 0.963 | 0.77 (0.54, 1.11) | 0.145 |
| Cubic | 0.93 (0.67, 1.30) | 0.647 | 1.04 (0.68, 1.63) | 0.864 |
| Pain distress level | ||||
| Linear | 1.54 (1.27, 1.87) | < 0.001 | 1.52 (1.18, 1.94) | 0.001 |
| Quadratic | 1.01 (0.84, 1.22) | 0.902 | 1.13 (0.88, 1.44) | 0.350 |
| Cubic | 0.93 (0.76, 1.13) | 0.469 | 1.02 (0.79, 1.32) | 0.869 |
| Outpatient - Specialized hospital | ||||
| No | 1 | 0.032 | - | - |
| Yes | 2.51 (1.05, 5.67) | - | ||
| Outpatient - Primary care institutions | ||||
| No | 1 | 0.008 | 1 | 0.015 |
| Yes | 1.40 (1.09, 1.78) | 1.56 (1.08, 2.22) | ||
| Herbal medicine | ||||
| No | 1 | < 0.001 | 1 | < 0.001 |
| Yes | 2.71 (2.23, 3.29) | 2.66 (2.07, 3.43) | ||
| Western medications | ||||
| No | 1 | < 0.001 | 1 | < 0.001 |
| Yes | 2.02 (1.66, 2.48) | 1.81 (1.42, 2.32) | ||
| Massage | ||||
| No | 1 | < 0.001 | 1 | < 0.001 |
| Yes | 10.21 (8.26, 12.63) | 7.40 (5.82, 9.43) | ||
| Other treatments | ||||
| No | 1 | 0.144 | - | - |
| Yes | 1.18 (0.94, 1.49) | - | ||
During the stratified analyses, we found that some subsets had a situation where only one level of the factor variable was retained, leading to coding errors in the construction of the design matrix. In addition, due to some differences in the distribution of missing values across subsets, untreated may cause inconsistent sample sizes in model estimation.To address these issues, we performed rigorous data cleaning on each subset prior to modelling, eliminating variables with only a single level and performing complete case rejection of missing values. Although this approach may have reduced the sample size of some subsets, it ensured the validity and stability of the model estimates.
Age-Stratified results
45–59 years: Females remained a higher propensity compared to males (OR = 1.28, P = 0.045), social activity (OR = 1.39, P = 0.006), difficulty in carrying weights over 5Kg (linear trend OR = 1.62, P = 0.005), pain distress level (linear trend OR = 1.60, P < 0.001), herbal medicine (OR = 2.76, P < 0.001), western medications (OR = 2.02, P < 0.001) and massage (OR = 10.21, P < 0.001) were significant. Chinese medicine hospital (OR = 2.10, P = 0.036) and primary care institutions (OR = 1.42, P = 0.038) outpatient care showed borderline significance.
60–74 years: Females (OR = 1.35, P = 0.020) and difficulty in stooping, kneeling or crouching (linear trend OR = 1.46, P = 0.002) were significant. Herbal medicine (OR = 2.56, P < 0.001), western medications (OR = 2.05, P < 0.001) and massage (OR = 8.61, P < 0.001) remained strongly predictive.
≥ 75 years: Only herbal medicine (OR = 3.09, P < 0.001) and massage (OR = 10.81, P < 0.001) retained significance. Social activity and western medications lost significance (P > 0.05), while alcohol consumption showed a quadratic trend (OR = 0.44, P = 0.025).
Gender-Stratified results
Age showed a negative correlation effect that was significant among females (OR = 0.99, P = 0.018) and borderline significant among males (OR = 0.99, P = 0.057). Among males, the results involving “Back-Waist” (OR = 1.26, P = 0.042), “Waist-Buttocks” (OR = 1.44, P = 0.047), and “Back-Waist-Buttocks” (OR = 1.44, P = 0.006) regions were significant, whereas no pain sites reached significance in males.The linear trend for pain distress was significant in both males (OR = 1.45, P = 0.006) and females (OR = 1.55, P < 0.001). Insurance (OR = 2.06, P = 0.034), social activity (OR = 1.31, P = 0.004), difficulty in carrying weights over 5Kg (linear trend OR = 1.37, P = 0.003) were retained in females. Herbal medicine, western medications, and massage showed significant associations in both sexes, with the female group showing a stronger association (OR = 2.74 vs. 2.52; OR = 1.98 vs. 1.86; OR = 10.29 vs. 7.45).
Residence-Stratified results
Age was only significant in non-urban areas (OR = 0.98, P = 0.001), but the effect was weak. The higher propensity for females to choose acupuncture persisted in both non-urban (OR = 1.32, P = 0.017) and urban areas (OR = 1.41, P = 0.008). Alcohol (linear trend OR = 1.21, P = 0.034) was a risk factor only in non-urban areas, which may reflect the higher prevalence of heavy drinking or alcohol-related comorbidities in rural populations. The linear trend for Difficulty in stooping, kneeling or crouching to be significant was only in non-urban areas (OR = 1.40, P = 0.003). Pain distress level was significant in both areas (non-urban OR = 1.54, urban OR = 1.52), suggesting a generalised role for pain in driving outcomes, independent of geography. Outpatient - Specialized hospital was significant only in non-urban areas (OR = 2.51, P = 0.032), and Outpatient - Primary care institutions was significant in both strata, slightly stronger in urban areas (OR = 1.56 vs. 1.40). Herbal medicine, western medications and massage effects were consistent in both areas (OR ≈ 2.7 for herbal medicine, OR ≈ 1.8 ~ 2.0 for western medications), and massage was stronger in non-urban areas (OR = 10.21 vs. 7.40).
Discussion
In this nationally representative cross-sectional study, the utilization rate of acupuncture among middle-aged and older adults with LBP in China was 13.8% (1097/7929). This result is similar to the 13.97% (1058/7576) reported in a previous study [17]. In the United Kingdom, 19% of the patient population choosing acupuncture treatment visited the clinic for LBP [13]. The rate of acupuncture use in the European chronic pain population (including LBP) was 13% [20]. In Korea, 32.2% of elderly patients with LBP chose acupuncture treatment [21]. In Dr. Vitaly’s 2003 study [20]LBP patients accounted for 11.2% of acupuncture outpatients in a Beijing hospital, whereas national surveys conducted in Japan [21]Australia [22]and the United States [23] showed overall acupuncture utilization rates of 6% (with 50.9% of these patients suffering from LBP), 9.2% (with 20.7% of these patients suffering from LBP), and 4.1% (with 34.0% of these patients suffering from LBP), respectively. Although these data demonstrate a certain degree of prevalence of acupuncture in the treatment of LBP, direct comparisons remain challenging due to differences in study contexts, sample sizes, and surveyed populations. In a related study [15] in Korea showed that the rate of lumbar spine surgery in patients with LBP in the acupuncture group (0.54%) was significantly lower than that in the non-acupuncture group (0.85%) at follow-up after two years, and the risk rate of lumbar spine surgery tended to be further reduced in the elderly group after acupuncture treatment. Another meta-analysis [24] showed that fusion surgery was not superior to non-surgical treatments in terms of pain and disability outcomes in short- or long-term follow-up. There are serious gaps in the evidence for the safety and efficacy of treatments for elderly patients with LBP, and the rapidly rising rate of imaging of incidental pathology may put the elderly at risk of inappropriate invasive treatments as well as persistent pain.28 Although we have not yet obtained definitive data related to the rate of surgery in Chinese patients with LBP, the comparison of the 19% selection rate in the UK and 32.2% in Korea suggests that there may be scope for further promotion of acupuncture therapy in China to reduce unnecessary surgery and invasive treatments in the middle-aged and elderly population with LBP.
The integration of age, gender, and residence stratified analyses with the stepwise model reveals both shared and distinct determinants of the outcome. The stepwise model identified massage, herbal medicine and western medications, pain distress level, and primary care institutions(outpatient) as robust predictors across all strata (with differences in the ≥ 75 years stratum). With only 12.5% of the sample in the ≥ 75 years stratum, most variables were not included in the model due to sparse data and lack of statistical validity (e.g. gender, pain distress level). Massage (OR = 10.81, P < 0.001), herbal medicine (OR = 3.09, P < 0.001) still maintained significance, limited by the sample size it is not yet possible to infer the association of key variables such as gender with the results. These predictors consistently showed strong associations, underscoring their universal relevance in the studied population.
The choice of western medications, herbal medicine, and massage shows an increasing positive correlation with choosing acupuncture for LBP. Previous studies have shown that massage and spinal manipulation are the most commonly offered forms of complementary and alternative medicine (CAM), and the use of CAM is associated with LBP. The frequent use of CAM for LBP suggests its popularity among both patients and physicians [25]. Among older adults with LBP, those receiving spinal manipulation reported higher satisfaction and better health-related quality of life compared to those receiving prescription medication [26]. Another study [27] indicated that patients with musculoskeletal pain are more likely to choose multiple treatments compared to other patients. These results, combined with our study, suggest that integrative conservative treatment plans, including massage and acupuncture, may be more suitable for older adults with LBP. Overall, the results suggest a trend where patients seeking diverse medical opinions or treatments are more inclined to choose acupuncture. The diverse treatment behavior patterns (Fig. 2) also show that patients using acupuncture are more likely to use other forms of complementary therapies. The combination of diverse complementary therapies reflect the trend towards integration of traditional and modern medicine in the practice of LBP management, while the fear of surgical risks, concerns about postoperative rehabilitation [28, 29]and recognition of the safety and efficacy of non-invasive treatments [7, 30] in the group of middle-aged and elderly LBP sufferers may contribute to a preference for non-invasive treatments. These results provide insights into the relationship between different healthcare service choices and the use of acupuncture. The strong association between acupuncture and alternative medicine warrants further investigation to understand how these practices complement each other and the motivations behind their concurrent use.
Fig. 2.
Distribution of different treatment combinations in the population treated with acupuncture
Pain distress level exhibited a linear dose-response relationship, aligning with global evidence on pain as a driver of healthcare utilization and disability, which was reflected in this study by the indicator Difficulty in stooping, kneeling or crouching in the stepwise model as well as in the stratified results. Research [30] indicates that patients using complementary health approaches, including acupuncture, experience lower disability and pain severity related to LBP, which may reflect perceptions of the potential benefits of acupuncture in relieving pain and improving physical function.
The strong linkage with primary care institutions’ outpatient clinics and Chinese medicine hospital’ outpatient clinics aligns with their role as primary providers of traditional therapies, likely driven by both service availability and patient predisposition. Notably, the stabilising effect of primary care institutions, predominantly community healthcare centers and township hospitals, may reflect the policy emphasis on the integration of traditional medicine into primary care in China in recent years [31]. In contrast, the non-significant association with general hospital’ outpatient clinics (OR = 1.26, P = 0.074) underscores the dominance of Western medicine in routine acute care, where acupuncture remains peripheral. The borderline significance of specialized hospital’ outpatient clinics (OR = 1.88, P = 0.046) warrants cautious interpretation, the broader confidence interval (0.99–3.43) suggests potential heterogeneity in patient profiles or service patterns across specialties. Future studies should employ causal inference models to disentangle whether facility type directly drives acupuncture adoption or merely correlates with preexisting patient preferences.
Our study also highlights a pronounced gender disparity in acupuncture utilization among middle-aged and elderly LBP patients, with females exhibiting significantly higher odds of choosing acupuncture compared to males. This pattern persists across both urban and non-urban settings, suggesting systemic rather than contextual drivers. The observed gender gap aligns with global trends where females are more likely to adopt complementary therapies for chronic pain managemen [17, 19, 32–34]and the mechanisms behind this need to be deconstructed from a multi-dimensional physiological-psychosocial perspective. The prevalence of LBP is higher in females than in males [35]. The prevalence of LBP tends to increase with age [36]especially in females, which may be related to the fact that decreasing levels of oestrogen weaken its protective effects on the skeletal, muscular, and nervous systems, which exacerbates the perception of pain and prompts females to be more proactive in seeking complementary therapies, such as acupuncture. Females are more likely to use any form of healthcare compared to males, who may be more inclined to tolerate or choose monotherapy [17]. In addition, in a study of gender differences in neural responses after acupuncture [37]it was found that females showed more pronounced brain activation and differences in activation areas in the relevant brain regions compared to males, and that this difference in neural activity may influence the subjective experience of and preference for acupuncture treatments in both sexes. Socio-cultural gender role stereotypes suggest that females may be more inclined to express pain and seek medical support, but may also be labelled as ‘emotional’ due to social pressures, whereas males are encouraged to show strength and patience [38]. This preference is a complex phenomenon shaped by a combination of hormonal fluctuations, neurological response differences and social role expectations.
Increasing age was significantly associated with a lower probability of acupuncture selection in middle-aged and elderly patients with LBP. In the gender-stratified analyses, although the age effect reached statistical significance in females (P = 0.018) and was not significant in males (P = 0.057), the effect was weak, and the gender : age interaction term was not statistically significant in the interaction model (P > 0.05), and the difference in the significance of the age in the main effect may have resulted from the difference in the sample sizes (the sample sizes were larger in women, and the detection of a small effect in the greater ability to detect small effects), it is uncertain whether this difference is real.
The contribution of health insurance to acupuncture choice among middle-aged and elderly LBP patients in this study showed an overall significant (OR = 1.76, P = 0.037), supporting the role of insurance coverage in promoting the use of complementary therapies. In the age stratification, a borderline significant association was observed only in the 60–74 year old group (OR = 2.17, P = 0.073), suggesting that the driving effect of insurance coverage on acupuncture choice may be concentrated in this group. The weak association of OR = 1.81 (P = 0.076) in non-urban areas may reflect the low actual reimbursement of traditional Chinese medicine services by primary care insuranc [39, 40]e, and the probability of acupuncture use was significantly higher among females than among the uninsured, whereas there was no significant association among males, possibly due to the fact that there were greater differences between male participants and the uninsured in terms of confounders such as pain severity and type of occupation resulting in the variables being eliminated by stepwise regression.
Limitations
The data on health status, medical behaviours and health service use covered in this study relied on self-reporting by respondents, and the findings may introduce recall bias, while the results of this study may not be able to be extrapolated to other patient groups of other cultural backgrounds or age groups. While controlling for health status, insurance factors, etc., respondents’ subjective treatment beliefs about the efficacy of the combination of treatments, respondents’ occupational type, income level, and other possible residual confounders were not measured. The cross-sectional data did not allow us to infer whether acupuncture “drove” other treatment choices or whether patients also chose acupuncture due to their preference for the combined regimen. In addition, due to data availability, we were not able to further differentiate between specific subtypes of LBP and details such as type of acupuncture, treatment acupoints, and frequency of treatment, and longitudinal studies are needed to validate the temporal relationship, efficacy, and so on. Differences in stratified results may be affected by residual confounding, small sample sizes in some subgroups lead to unstable estimates, while sparse data due to uneven sample sizes may mask specific patterns in subgroups such as the elderly. In the future, age-specific mechanisms need to be validated by larger cohorts of older adults and mechanistic studies.
Conclusion
This study demonstrated that LBP patients with certain characteristics tend to use complementary and alternative interventions, including acupuncture. Although the statistical association suggests a synergistic trend, there is a need to combine the evaluation of efficacy with mechanistic studies in the future to clarify the optimal combination of multimodal treatments and the applicable populations, in order to improve the clinical practice and the public health policy to promote the use of acupuncture in the treatment of LBP.
Supplementary Information
Acknowledgements
Not applicable.
Clinical trial number
Not applicable.
Authors’ contributions
YL is first author. ZJ and CB designed the study. YL, XX, and XH participated in data cleaning and data analysis. YL drafted the manuscript. YL, XX contributed to the interpretation of the results, and ZJ, CB, XM made important revisions to the manuscript for important intellectual content and approved the final version of the manuscript.All authors have approved the final article.
Funding
This research was supported by grants from the High-level TCM Key Discipline of National Administration of Traditional Chinese Medicine (NO.zyyzdxk-2023068) and the Shanghai Clinical Medical Research on Acupuncture and Moxibustion (NO. 20MC192050) and the Outstanding Leader Plan of Shanghai (No. 060) in the public sector.
Data availability
The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.” and in the submission system “The data that support the findings of this study are available on the website of the China Health and Retirement Longitudinal Study (CHARLS) at https://charls.charlsdata.com/pages/data/111/zh-cn.html. To access and use this survey data for research purpose, an approval should be obtained from the CHARLS team at Peking University. The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Declarations
Ethics approval and consent to participate
The CHARLS study was approved by the Ethics Review Committee of Peking University (approval number: IRBO0001052-11,015). All methods were carried out in accordance with relevant guidelines and regulations, and all participants signed informed consent forms when participating.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Chunling Bao, Email: chunlingb@163.com.
Xiaopeng Ma, Email: pengpengma@163.com.
References
- 1.Murthy V, Sibbritt D, Adams J, Broom A, Kirby E, Refshauge KM. Self-prescribed complementary and alternative medicine use for back pain amongst a range of care options: results from a nationally representative sample of 1310 women aged 60–65 years. Complement Ther Med. 2014;22(1):133–40. 10.1016/j.ctim.2013.11.013. [DOI] [PubMed] [Google Scholar]
- 2.Hartvigsen J, Hancock MJ, Kongsted A, et al. What low back pain is and why we need to pay attention. Lancet. 2018;391(10137):2356–67. 10.1016/S0140-6736(18)30480-X. [DOI] [PubMed] [Google Scholar]
- 3.Jia N, Zhang M, Zhang H, et al. Prevalence and risk factors analysis for low back pain among occupational groups in key industries of China. BMC Public Health. 2022;22(1):1493. 10.1186/s12889-022-13730-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Li Q, Peng L, Wang Y, Yang Y, Wang Z. Risk factors for low back pain in the Chinese population: a systematic review and meta-analysis. BMC Public Health. 2024;24(1):1181. 10.1186/s12889-024-18510-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Mu J, Furlan AD, Lam WY, Hsu MY, Ning Z, Lao L. Acupuncture for chronic nonspecific low back pain. Cochrane back and neck group. Ed Cochrane Database Syst Reviews. 2020;2020(12). 10.1002/14651858.CD013814. [DOI] [PMC free article] [PubMed]
- 6.Rubin R. Medicare proposes coverage of acupuncture for lower back pain. JAMA. 2019;322(8):716. 10.1001/jama.2019.11573. [DOI] [PubMed] [Google Scholar]
- 7.Qaseem A, Wilt TJ, McLean RM, et al. Noninvasive treatments for acute, subacute, and chronic low back pain: a clinical practice guideline from the American college of physicians. Ann Intern Med. 2017;166(7):514–30. 10.7326/M16-2367. [DOI] [PubMed] [Google Scholar]
- 8.Greville-Harris M, Hughes J, Lewith G, et al. Assessing knowledge about acupuncture: a survey of people with back pain in the UK. Complement Ther Med. 2016;29:164–8. 10.1016/j.ctim.2016.10.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Vickers AJ, Vertosick EA, Lewith G, et al. Acupuncture for chronic pain: update of an individual patient data meta-analysis. J Pain. 2018;19(5):455–74. 10.1016/j.jpain.2017.11.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lin H, Wang X, Feng Y, et al. Acupuncture versus oral medications for acute/subacute non-specific low back pain: a systematic review and meta-analysis. Curr Pain Headache Rep. 2024;28(6):489–500. 10.1007/s11916-023-01201-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Liu L, Skinner M, McDonough S, Mabire L, Baxter GD. Acupuncture for low back pain: an overview of systematic reviews. Evid Based Complement Alternat Med. 2015;2015:328196. 10.1155/2015/328196. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Chenot JF, Becker A, Leonhardt C, et al. Determinants for receiving acupuncture for LBP and associated treatments: a prospective cohort study. BMC Health Serv Res. 2006;6:149. 10.1186/1472-6963-6-149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hopton AK, Curnoe S, Kanaan M, Macpherson H. Acupuncture in practice: mapping the providers, the patients and the settings in a national cross-sectional survey. BMJ Open. 2012;2(1):e000456. 10.1136/bmjopen-2011-000456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Zhang Y, Lao L, Chen H, Ceballos R. Acupuncture use among American adults: what acupuncture practitioners can learn from National health interview survey 2007? Evid Based Complement Alternat Med. 2012;2012:710750. 10.1155/2012/710750. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Koh W, Kang K, Lee YJ, et al. Impact of acupuncture treatment on the lumbar surgery rate for low back pain in Korea: a nationwide matched retrospective cohort study. PLoS One. 2018;13(6):e0199042. 10.1371/journal.pone.0199042. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China health and retirement longitudinal study (CHARLS). Int J Epidemiol. 2014;43(1):61–8. 10.1093/ije/dys203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mei F, Dong S, Li J, Xing D, Lin J. Preference of musculoskeletal pain treatment in middle-aged and elderly Chinese people: a machine learning analysis of the China health and retirement longitudinal study. BMC Musculoskelet Disord. 2023;24(1):528. 10.1186/s12891-023-06665-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Zhou Y, Wushouer H, Vuillermin D, Ni B, Guan X, Shi L. Medical insurance and healthcare utilization among the middle-aged and elderly in China: evidence from the China health and retirement longitudinal study 2011, 2013 and 2015. BMC Health Serv Res. 2020;20:654. 10.1186/s12913-020-05522-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Knezevic NN, Candido KD, Vlaeyen JWS, Van Zundert J, Cohen SP. Low back pain. Lancet. 2021;398(10294):78–92. 10.1016/S0140-6736(21)00733-9. [DOI] [PubMed] [Google Scholar]
- 20.Napadow V, Kaptchuk TJ. Patient characteristics for outpatient acupuncture in Beijing, China. J Altern Complement Med. 2004;10(3):565–72. 10.1089/1075553041323849. [DOI] [PubMed] [Google Scholar]
- 21.Ishizaki N, Yano T, Kawakita K. Public status and prevalence of acupuncture in Japan. Evid Based Complement Alternat Med. 2010;7(4):493–500. 10.1093/ecam/nen037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Xue CCL, Zhang AL, Lin V, Myers R, Polus B, Story DF. Acupuncture, chiropractic and osteopathy use in Australia: a national population survey. BMC Public Health. 2008;8:105. 10.1186/1471-2458-8-105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Burke A, Upchurch DM, Dye C, Chyu L. Acupuncture use in the united states: findings from the national health interview survey. J Altern Complement Med. 2006;12(7):639–48. 10.1089/acm.2006.12.639. [DOI] [PubMed] [Google Scholar]
- 24.Xu W, Ran B, Luo W, Li Z, Gu R. Is lumbar fusion necessary for chronic low back pain associated with degenerative disk disease?? A meta-analysis. World Neurosurg. 2021;146:298–306. 10.1016/j.wneu.2020.11.121. [DOI] [PubMed] [Google Scholar]
- 25.Jf C. Use of complementary alternative medicine for low back pain consulting in general practice: a cohort study. BMC Complement Altern Med. 2007;7. 10.1186/1472-6882-7-42. [DOI] [PMC free article] [PubMed]
- 26.Kizhakkeveettil A, Bezdjian S, Hurwitz EL, et al. Spinal manipulation vs prescription drug therapy for chronic low back pain: beliefs, satisfaction with care, and qualify of life among older medicare beneficiaries. J Manipulative Physiol Ther. 2021;44(8):663–73. 10.1016/j.jmpt.2021.12.007. [DOI] [PubMed] [Google Scholar]
- 27.Cai Y, Boyd DL, Coeytaux RR, Østbye T, Wu B, Mao Z. Treatment of chronic conditions with traditional Chinese medicine: findings from traditional Chinese medicine hospitals in Hubei, China. J Altern Complement Med. 2015;21(1):40–5. 10.1089/acm.2014.0125. [DOI] [PubMed] [Google Scholar]
- 28.Wu Q, Cui X, Guan LC, et al. Chronic pain after spine surgery: insights into pathogenesis, new treatment, and preventive therapy. J Orthop Transl. 2023;42:147–59. 10.1016/j.jot.2023.07.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Huang J, Li P, Wang H, Lv C, Han J, Lu X. Exploring elderly patients’ experiences and concerns about early mobilization implemented in postoperative care following lumbar spinal surgery: a qualitative study. BMC Nurs. 2023;22(1):355. 10.1186/s12912-023-01510-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Urits I, Wang JK, Yancey K, et al. Acupuncture for the management of low back pain. Curr Pain Headache Rep. 2021;25(1):2. 10.1007/s11916-020-00919-y. [DOI] [PubMed] [Google Scholar]
- 31.Fang L, Zhu RR, Sang Z, et al. World health organization survey on the level of integration of traditional Chinese medicine in Chinese health system rehabilitation services. Integr Med Res. 2023;12(2):100945. 10.1016/j.imr.2023.100945. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Wu MY, Lee YC, Lin CL, Huang MC, Sun MF, Yen HR. Trends in use of acupuncture among adults in Taiwan from 2002 to 2011: a nationwide population-based study. PLoS One. 2018;13(4):e0195490. 10.1371/journal.pone.0195490. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Cui J, Wang S, Ren J, Zhang J, Jing J. Use of acupuncture in the USA: changes over a decade (2002–2012). Acupunct Med. 2017;35(3):200–7. 10.1136/acupmed-2016-011106. [DOI] [PubMed] [Google Scholar]
- 34.Wardle JL, Adams J, Sibbritt DW. Acupuncture in Australian general practice: trends in reimbursed acupuncture services from 1995 to 2011. Acupunct Med. 2013;31(1):45–50. 10.1136/acupmed-2012-010244. [DOI] [PubMed] [Google Scholar]
- 35.Xu S, Qi J, Liu C, et al. Evaluation of three decades of the burden of low back pain in China before COVID-19: estimates from the global burden of disease database 2019. J Glob Health. 2024;14:04006. 10.7189/jogh.14.04006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Wáng YXJ, Wáng JQ, Káplár Z. Increased low back pain prevalence in females than in males after menopause age: evidences based on synthetic literature review. Quant Imaging Med Surg. 2016;6(2):19906–19206. 10.21037/qims.2016.04.06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Oliver MI, Pearson N, Coe N, Gunnell D. Help-seeking behaviour in men and women with common mental health problems: cross-sectional study. Br J Psychiatry. 2005;186:297–301. 10.1192/bjp.186.4.297. [DOI] [PubMed] [Google Scholar]
- 38.Samulowitz A, Gremyr I, Eriksson E, Hensing G. Brave men and emotional women: A Theory-Guided literature review on gender bias in health care and gendered norms towards patients with chronic pain. Pain Res Manag. 2018;2018:6358624. 10.1155/2018/6358624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Zeng Z, Luo Y, Tao W, et al. Improving access to primary health care through financial innovation in rural China: a quasi-experimental synthetic difference-in-differences approach. BMC Prim Care. 2024;25:195. 10.1186/s12875-024-02450-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Liu S, Meng W, Yu Q, et al. Evaluation and countermeasures of contracted services of Chinese family Doctors from demanders’ point of view — a case study of a City. BMC Health Serv Res. 2022;22(1):1534. 10.1186/s12913-022-08891-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.” and in the submission system “The data that support the findings of this study are available on the website of the China Health and Retirement Longitudinal Study (CHARLS) at https://charls.charlsdata.com/pages/data/111/zh-cn.html. To access and use this survey data for research purpose, an approval should be obtained from the CHARLS team at Peking University. The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.