Effectiveness of a multidisciplinary rehabilitation program in real-world patients with chronic back pain: A pilot cohort data analysis

2.1Enrollment process

This was a retrospective cohort study of patients who attend the BIA or pain-CBT at a tertiary pain clinic between January 2017 and November 2019. Inclusion criteria for both programs were patients with any chronic back pain (i.e., neck, upper, and lower back pain). The BIA was designed for those who significantly reduced daily function and needed an intensive outpatient program whereas the pain-CBT was for patients who had relatively better daily function and needed to learn self-management skills. All patients in both groups were referred by pain physicians, but the enrollment process was different between the two programs. All patients in the BIA were evaluated by pain psychologists to determine their appropriateness for the group-based pain psychology program and by physical therapists to determine whether the patient could safely participate in the exercise-based interventions. Tests and measures during the physical therapy evaluation included lumbar ROM, core testing, balance testing, a six-minute walk test (6MWT) with BP/HR measurements pre-and post 6MWT [19]. Then, patients who had an insurance approval attended the BIA program. In contrast, not all patients in the pain-CBT underwent psychology evaluation because pain physicians could make a referral to the pain-CBT. The pain-CBT was periodically offered free of charge and insurance approval was not needed.

2.2Intervention content

Our programs covered the published CBT [20] and ACT contents [21]. Specifically, the pain-CBT included pain psychoeducation and pain self-management techniques, listed in Table 1. The pain-CBT was run by a pain psychologist or 4 supervised postdoctoral psychology fellows. The 2-hour CBT was offered over 8 weeks (total 16 hours). In addition to the pain-CBT, the BIA covered the ACT content: values, mindfulness, committed action, defusion, and self as context [21]. The BIA was run by a pain psychologist and a physical therapist. The BIA consisted of 12 sessions over a 6-week period (total 24 hours of pain psychology and 24 hours of PT sessions). In each BIA session were 2-hour pain psychology and 2-hour PT (Fig. 1).

Figure 1.

Comparison of survey administration and the treatment course for Pain-CBT and BIA programs. Post-Tx: post-treatment.

Our PT program included the evidenced-based educational contents and exercise interventions: pain neuroscience education as well as graded aerobic exercise and supervised/individualized resistance exercise [22, 23](Table 1). Tai Chi and Yoga were also included as they had shown to reduce pain and improve disability for patients with low back pain [24, 25]. The Tai Chi program was led by a Physical Therapist that was certified in Tai Chi for Rehabilitation Sequence [24]. The Tai Chi was taught in seated and standing positions. The yoga was based off the published protocol [25], including supine poses, seated poses and standing poses for 20 minutes each. Patients had a graded aerobic exercise program that increased in duration by 10–20% each week throughout the 6 weeks. The individual exercise program, which was created by the physical therapist at the initial evaluation, focused each individual’s impairments and self-selected goals. The individual program was also upgraded each week throughout the program.

2.3Measurements

Retrospective chart reviews were conducted to assess medical diagnoses. The current study extracted data of the following measures completed within one week of the first and last sessions.

Our primary outcome was the Patient-Reported Outcomes Measurement Information System (PROMIS)®, patient-centered outcome measures [26], which was administered using CAT [27]. The current study included the PROMIS item banks for physical and psychosocial health domains. The physical health domains were upper extremity movement, mobility, pain interference, pain behavior, fatigue, and sleep-related impairment. The psychosocial health domains were depression, anxiety, social role satisfaction, emotional support, and social isolation. All PROMIS scores were converted to T scores, with higher scores indicating worse health status in each domain.

Secondary outcomes were the PROMIS Pain Intensity measure assessing the intensity of pain at worst and average for the past 7 days on a 0–10 scale [28] and the Pain Catastrophizing Scale (PCS) assessing levels of catastrophic thinking about pain [29]. The total PCS scores range from 0 to 52, with higher scores indicating greater pain catastrophizing. Finally, the body map was used to assess the number of painful sites, ranging from 0 to 45 [30].

2.4Statistical analysis

Appropriate (χ2, Mann-Whitney U, and independent t) tests were conducted to examine whether the two groups’ sociodemographic variables and clinical symptoms were significantly different at baseline. Two(time) by Two(group) repeated measures MANOVAs were conducted for the three sets of outcomes: a) PROMIS-physical health measures, b) PROMIS-psychosocial health measures, and c) pain ratings and pain catastrophizing scores, followed by post-hoc univariate analyses. p values of < 0.05 were considered statistically significant. Cohen’s d was computed to calculate effect sizes, and d of 0.2, 0.5, and 0.8 would indicate small, medium, and large effects, respectively [31]. Sensitivity analysis was conducted to examine whether significant difference in treatment outcomes (i.e., significant time by group interaction) would remain significant even after adjusting for the pre-existing difference in sociodemographic variables.

We collected data until our sample size was a total of 40 patients for the current pilot study and intended to have 20 for each group. However, our final sample included 18 patients for pain-CBT and 28 patients for BIA and the current study computed the effect sizes, which can be used for a future study’s power analysis. No significant violation of multicollinearity and homogeneity were observed as all correlation coefficients were < 0.685 and p values of Box’s M tests were > 0.600. There were no missing values except for education and legal claim.

Figure 2.

Comparisons of pre- and post-treatment outcomes. A)–G) are PROMIS T scores of Mobility, Pain Interference, Pain Behavior, Fatigue, Depression, Anxiety, and Social Role Satisfaction, respectively, with higher T scores indicating worse health status in each domain. H) is PCS total score. Errors = SEM.

3.1Baseline patient characteristics

Sociodemographic states were not significantly different between the two groups (p′s > 0.426, Table 2). Patients were predominantly middle-aged, married female, and White/Caucasian with college or higher degree. About two thirds endorsed currently not working. A small percentage of patients in Pain-CBT and BIA programs reported receiving disability benefits (17.9 and 27.8%, p= 0.426), having injury-related pain (38.9 and 28.6%, p= 0.466), and taking a legal action for pain (23.5 and 3.6%). Because less than 5 patients with a legal claim, Fisher’s exact test was conducted and the result indicated legal claim was marginally higher in pain-CBT group (p= 0.060).

The two groups were not significantly different in the number of painful sites (p= 0.295), but pain durations were marginally higher in the BIA group (p= 0.068). Most patients (67.4%) had comorbid diagnosis and patients in the BIA had a trend for more comorbid health conditions (p= 0.095). Different from our expectation, independent t tests indicated that pain, physical, and psychosocial health states were not significantly different between the groups at baseline (p′s > 0.143, Table 2). The median of attended sessions was 7.0 (IQR = 6.0–8.0) for the pain-CBT and 11.5 (IQR = 9.6–12.0) for the BIA.

3.2Health outcomes of the pain-CBT and BIA programs

A 2(time) × 2(groups) repeated measures MANOVA for physical health outcomes (i.e., PROMIS-mobility, pain interference, pain behaviors, fatigue, and sleep impairment T scores) revealed a significant interaction effect, Wilks’ Λ= 0.71, p= 0.014, and a significant main effect of time, Wilks’ Λ= 0.63, p= 0.002. However, we did not find a significant group effect (p= 0.376). Post-hoc tests showed significant interaction effect only in mobility, F(1, 44) = 8.51, p= 0.006 (Fig. 2A), and pain behaviors, F(1, 44) = 4.99, p= 0.031 (Fig. 2C). Mobility and pain behaviors T scores were reduced by 2.32 and 2.25T after the BIA and < 1.0T after pain-CBT (Table 3). Additionally, post-hoc analysis revealed a significant time effect in pain interference, F(1, 44) = 21.01, p< 0.001 (Fig. 2C) and fatigue, F(1, 44) = 9.12, p= 0.004 (Fig. 2D). Pain interference and fatigue scores were reduced by 3.30T and 3.89T, respectively, after the pain-CBT or BIA (Table 3). In summary, mobility and pain behavior were improved only after the BIA whereas pain interference and fatigue were improved after pain-CBT or BIA programs, with medium sized effects (d′s = 0.47–0.73).

We conducted sensitivity analysis for the significant interaction effect. When entering legal claims as a covariate, the previously significant interaction effect in mobility, F(1, 42) = 14.23, p= 0.001, and pain behaviors, F(1, 42) = 14.77, p= 0.035, remained significant. This suggests that significant improvement of mobility and pain behavior only after BIA is unlikely to be driven by less legal claims in the BIA group.

Another 2(time) × 2(group) repeated measures MANOVA for psychosocial health outcomes (i.e., PROMIS-depression, anxiety, emotional support, social isolation, and social role satisfaction T scores) revealed significant interaction, Wilks’ Λ= 0.66, p= 0.005 and time effects, Wilks’ Λ= 0.75, p= 0.032. However, we did not find a significant group effect (p= 0.781; Fig. 2E–G). Post-hoc tests showed no significant time by interactions (ps > 0.099), but a significant time effect only in depression, F(1, 44) = 6.61, p= 0.014, anxiety, F(1, 44) = 4.64, p= 0.037, and social role satisfaction, F(1, 44) = 11.50, p= 0.001, of which T scores were improved by 2.48–3.15T (Table 3). This suggests that depression, anxiety, and social role satisfaction were improved after pain-CBT or BIA programs, with small to medium effects (ds = 0.29–0.73).

Finally, a 2(time) × 2(group) repeated measures MANOVA was conducted to compare PCS scores and the worst and average pain ratings. We found no significant interaction (p= 0.584) and group effect (p= 0.151), but a significant time effect, Wilks’ Λ= 0.75, p= 0.006. Post-hoc tests indicated a time effect in PCS scores, F(1, 44) = 14.06, p= 0.001 (Fig. 2H), but not in worst and average pain ratings (p’s > 0.207). PCS scores were reduced by 4.98 points after pain-CBT or BIA (Table 3), with a moderate effect (d= 0.57).