Non-inferiority of hybrid outpatient telerehabilitation for patients with back pain: 3-month follow-up of a randomized controlled trial

Abstract

BACKGROUND

International studies identified comparable or better effects for telerehabilitation compared with face-to-face rehabilitation or no rehabilitation in people with back pain. In German rehabilitation centers, a standardized back school for patients with back pain is provided usually face-to-face as part of a multimodal rehabilitation program.

AIM

To examine the non-inferiority of a three-week, digitally assisted, multimodal rehabilitation that applies a digital version of a standardized back school (intervention group [IG]) against the same rehabilitation program applying the back school face-to-face (control group [CG]).

DESIGN

Our study was a non-blinded multicenter randomized controlled trial. Recruitment was conducted from 2022 to 2023. We analyzed outcomes at the end of rehabilitation and 3 months later.

SETTING

Implementation of the study and enrollment of participants was conducted in 8 German outpatient rehabilitation centers.

POPULATION

Rehabilitants aged 18-65 years with back pain were included.

METHODS

284 patients with back pain were randomized into the IG or CG using computer-generated block randomization. We excluded 14 patients as they withdrew their consent and requested removal of their data. We finally included 270 patients (IG: N.=127, CG: N.=143). The primary outcome was self-reported pain self-efficacy (10-60 points). Secondary outcomes were, amongst others, current health status and pain.

RESULTS

Our primary adjusted intention-to-treat analysis demonstrated that hybrid digitally assisted rehabilitation was non-inferior to face-to-face rehabilitation at the end of rehabilitation (b=-0.55; 95% CI=-2.75 to ∞) and at the 3-month follow-up (b=0.24; 95% CI=-2.86 to ∞). These results were in line with a non-adjusted intention-to-treat analysis, an adjusted complete case analysis, and an adjusted per-protocol analysis. Secondary outcomes were tested for superiority. Our primary adjusted intention-to-treat analysis found no significant group differences in the secondary outcomes.

CONCLUSIONS

This study provides evidence that hybrid digitally assisted rehabilitation in patients with back pain is a sound alternative to face-to-face rehabilitation in an outpatient rehabilitation setting.

CLINICAL REHABILITATION IMPACT

Hybrid digitally assisted rehabilitation can improve flexibility and access to rehabilitation. Further studies should examine which components and which time frame of rehabilitation can be digitized without any loss of effectiveness.

Low back pain is a leading cause of work disability and was responsible for approximately 64 million years lived with disability worldwide in 2019.¹ In Germany, about 15.6% of adults have chronic back pain (18.5% of women and 12.4% of men).² Back pain is associated with high costs due to sick leave, use of health services and disability pensions.^{3, 4} However, rehabilitation programs with a multidisciplinary biopsychosocial approach improve pain and disability in individuals with chronic low back pain,⁵ and are strongly recommended in guidelines for the treatment of back pain around the world.^{6, 7}

An overview of systematic reviews suggests that physiotherapeutic telerehabilitation in diseases such as arthrosis, low back pain, and hip and knee replacements can achieve as comparable or even better outcomes relative to personal rehabilitation or non-rehabilitation.⁸ A scoping review found that remote physiotherapy has similar effectiveness as face-to-face delivery, but it is less costly for patients and health care providers.⁹ Furthermore, patients chose remote telerehabilitation to support face-to-face therapy sessions rather than to replace them. Consequently, a hybrid approach that incorporates both may be the most efficient and acceptable.⁹

Our randomized controlled trial compared a hybrid rehabilitation program with a multimodal back school delivered digitally to a rehabilitation program that delivered the same back school conventionally in face-to-face sessions. We intended to demonstrate that patients completing the hybrid rehabilitation program achieve as comparable pain self-efficacy as patients completing the back school within a conventional face-to-face rehabilitation program. We also aimed to investigate the usage, satisfaction and usability of the digital application.

Materials and methods

Ethical approval and consent to participate

This investigation conformed to the principles outlined in the Declaration of Helsinki. The Ethics Committee of the University of Lübeck (21-462) approved of the study on March 25, 2022, before its start. Participation in the study was voluntary. Written and informed consent was necessary to participate in the study. Model consent forms and patient information forms given to the participants can be found in the supplement of the study protocol.¹⁰ The study has been registered in the German Clinical Trials Register (DRKS00028770).

Trial design

We conducted a randomized controlled trial to test the non-inferiority of the hybrid rehabilitation program.¹⁰ We recruited participants in eight outpatient rehabilitation centers. In this paper we report our findings at the end of rehabilitation and the 3-month follow-up. These short-term effects give us insight into the immediate outcomes of the intervention.

We did not apply stopping guidelines, and we did not expect adverse health events. We prepared this article according to the Consolidated Standards of Reporting Trials (CONSORT) and the extension to non-inferiority and equivalence randomized trials.^{11, 12}

Eight centers for outpatient rehabilitation were involved in recruitment, data collection, and implementation of the rehabilitation program. The centers were located in seven different German cities: Berlin, Bielefeld, Frankfurt am Main, Jena, München, Paderborn, and Regensburg. The intervention and control groups were recruited in the same rehabilitation center, but the digital back school of the intervention group was used remotely and was provided by Caspar Health, an e-health company that offers digital rehabilitation and aftercare.

Participants

Rehabilitants aged 18-65 years and with back pain (ICD-10 M50-M54, post-acute and acute rehabilitation) were included. Patients without stable internet, an appropriate electronic device for using the internet or executing app-related videos, and a suitable camera for communication were excluded. Patients who did not speak German were also excluded.

Intervention group

The intervention group received a hybrid rehabilitation program, during which a standardized back school was delivered digitally. The standardized back school developed from the Federal German Pension Insurance consists of seven modules and is based on the health action process approach and the fear-avoidance beliefs model.^13-15 It represents the standard back school in all participating outpatient rehabilitation centers. The effectiveness of this back school has been shown in a randomized controlled trial within face-to-face inpatient rehabilitation.¹⁴ Overall, the rehabilitation consists of a 3-week individual rehabilitation program at the outpatient rehabilitation center, with all treatments and therapies following the therapy standards of the Federal German Pension Insurance for patients with chronic back pain.¹⁶ The back school curriculum represents only a part of this 3-week rehabilitation. The Caspar application was used for the digital implementation of the back school and its remote delivery. Supplementary Digital Material 1, Supplementary Table I shows an overview of the back school in the intervention group, based on the Template for Intervention Description and Replication (TIDieR) checklist¹⁷ and the TIDieR-Telehealth checklist.¹⁸ A more detailed description of the intervention is published in our study protocol.¹⁰

Control group

The control group received a conventional rehabilitation program, during which the standardized back school was provided in face-to-face meetings. The overall rehabilitation consists of a 3-week individual rehabilitation program at the outpatient rehabilitation center, with all treatments and therapies following the therapy standards of the Federal German Pension Insurance for patients with chronic back pain.¹⁶ The back school curriculum represents only a part of this 3-week rehabilitation. Supplementary Digital Material 2, Supplementary Table II shows an overview of the back school in the control group, based on the Template for Intervention Description and Replication (TIDieR) checklist.¹⁷ A more detailed description of the intervention is published in our study protocol.¹⁰

Both groups

In addition to the back school, both groups received additional treatments during the three-week individual rehabilitation program at the outpatient rehabilitation center. All centers provided a multimodal interprofessional program consisting of exercise therapy, physiotherapy, massage and other physical therapies, social and psychological counseling, patient education, pain management, and relaxation training. Treatments followed the therapy standards of the Federal German Pension Insurance for patients with chronic back pain.¹⁶ The daily dose was around four and a half hours. The programs aimed to improve the patient’s functional limitations to enable participation in work and everyday life. Most extensive part of the three-week program is exercise which accounts for around half of all therapies. Patients have to request rehabilitation at the German Pension Insurance. The request is assessed by the German Pension Insurance and requires approval. Rehabilitation is approved to improve or restore work ability and to avoid disability pensions.

Outcomes

Primary outcome

The primary outcome was self-reported pain self-efficacy, which was measured by the German adaption of the Pain Self-Efficacy Questionnaire (PSEQ; German: Fragebogen zur Erfassung der schmerzspezifischen Selbstwirksamkeit [FESS]).^{19, 20} The score ranges from 10-60. Higher scores indicate better pain self-efficacy.¹⁹

Secondary outcomes

The secondary outcomes were general health (one item; Copenhagen Psychosocial Questionnaire [COPSOQ]; 0-10 points),²¹ mental health, functional capacity, and pain (IRES-24 questionnaire; 0-10 points);²² and cognitive and behavioral pain management,²³ including action-oriented coping, cognitive restructuring, subjective coping competence, mental distraction, counter-activities and relaxation as subscales for cognitive and behavioral pain management (Fragebogen zur Erfassung der Schmerzverarbeitung, [FESV]; 4-24 points).²³

Additional secondary outcomes were motivational self-efficacy, determined by the ratings for three statements based on Schwarzer et al. (3-12 points),²⁴ disorder and treatment knowledge on back pain (0-50 points), self-efficacy in practicing gained knowledge (0-20 points) and electronic health literacy (E Health Literacy Scale, [eHEALS]; 8-40 points).²⁵ Three self-developed questions on self-informing behavior, exercise adherence after the end of rehabilitation, and knowledge adherence after the end of rehabilitation were each rated categorically and converted into a binary variable (at least once per week vs. less than once per week).

Additional work-related secondary outcomes comprised work ability assessed by three questions of the Work Ability Index.^{26, 27} The first and second questions asked about the current ability to do physical work and mental work (2-10 points). The third question asked participants to rate their work ability from 0-10 points.^{28, 29} Furthermore, current sickness absence, the duration of sickness absence (in weeks), and the current employment status (yes or no) were assessed.

Patient satisfaction was assessed by the client satisfaction questionnaire (8-32 points).^{30, 31}

Moreover, the digital intervention was rated by the intervention group only. We assessed the system usability of the Caspar application (0-100 points with at least 70 points being rated as acceptable),^{32, 33} an overall rating of the Caspar application,³⁴ and asked two additional self-developed questions on the frequency of use of the Caspar application and the electronic device used. Scores of the system usability scale were aggregated into five categories (0-<60, >60-69, >70-79, >80-89, >90-100).³³

Sample size

We determined that a difference of four points for pain self-efficacy¹⁹ is the minimum relevant difference between the intervention and control group.³⁵ The minimum required sample size for our non-inferiority analysis was 242 participants (one-sided error: 5%; power: 90%) to perform an intention-to-treat analysis after multiple imputation. We planned to recruit 320 participants from eight outpatient rehabilitation centers (intervention group: N.=160; control group: N.=160).

Randomization and blinding

Randomized allocation was done in a one-to-one ratio. The principal investigator at the University of Lübeck generated all randomization sequences using Stata 16.1 (StataCorp, College Station, TX, USA). For each outpatient rehabilitation center, 40 assignments were randomly combined in blocks of four and eight, and the University of Lübeck provided the rehabilitation centers with 40 identical, non-transparent, sealed envelopes numbered from 1-40. The envelopes contained the information about the group allocation. Prior to opening the envelope, its content was unknown to everyone, excluding the principal investigator who was not involved in recruitment. The envelopes were handed out consecutively to the participants after they had given informed consent. A study assistant in each outpatient rehabilitation center enrolled the participants and registered the group allocation in the study list (Microsoft Excel) using the identification number. Due to the nature of the intervention, participants and treatment staff in the study were not blinded. As we assessed data by self-report using questionnaires, the assessors were also not blind.

Statistical analysis

We calculated descriptive statistics to determine the sample characteristics, the dosage of the delivered treatment, and usage of the Caspar application. We determined baseline group differences with a two-sample t-test for continuous variables and the Pearson Chi-square Test for categorical variables. Moreover, we calculated the system usability of the Caspar application and its overall assessment.

To test the primary hypothesis of non-inferiority, we applied the confidence interval (CI) method.^{11, 12} We determined -4 points as the non-inferiority margin. Four points are slightly below the smallest clinically important difference (5.5-8.5 points) reported in a recent systematic review on pain self-efficacy in patients with low back pain.³⁵ We assumed non-inferiority of the hybrid rehabilitation if the lower boundary of the one-sided 95% CI exceeded -4 points. We adjusted the estimate of the treatment effect for the baseline score of the dependent variable, as well as for the treatment center (outpatient rehabilitation center). We used linear regression models to calculated the estimates, and we tested non-inferiority in separate models for the end of rehabilitation and at the 3-month follow-up. To justify conducting linear regression for our main analyses, we inspected the assumptions. The results can be found in Supplementary Digital Material 3 (Supplementary Figures 1-4). We considered age, sex, level of education, electronic health literacy, and motivational self-efficacy as moderators of the treatment effect on our primary outcome. We assessed the interaction effects through linear regression models and the corresponding P values.

To estimate group differences for secondary continuous and binary outcome variables, we performed linear or logistic regression. We determined regression coefficients or odds ratios, their 95% CI, and P values. We tested all secondary outcome variables based on superiority of the intervention group. We tested superiority to identify potential group differences that, in the case of non-inferiority, could be used to formulate a preference for one of the two treatments.³⁶

To perform an intention-to-treat analysis, we imputed missing values by multiple imputation using 20 independent data sets.^{37, 38} We included parameters without missing values (group allocation, the location of the outpatient rehabilitation center, sex, age and native German language skills) as covariates in the imputation model. We combined the parameter estimates according to Rubin’s rules.³⁹

In sensitivity analyses, we calculated an unadjusted intention-to-treat estimate, an estimate for a complete case analysis adjusted for the baseline score of the dependent variable and for the treatment center, and an adjusted per-protocol analysis estimate. The latter analysis included only individuals in both groups who participated in six of the seven back school modules.

We considered P values <0.05 to be statistically significant. All hypothesis tests, except for the primary outcome, were two-tailed. Withdrawal of consent led to deletion of the participant’s data. We used Stata 16.1 for the analyses (StataCorp, College Station, TX, USA).

Results

Participants flow and recruitment

From April 5, 2022, to January 31, 2023, study assistants in the outpatient rehabilitation centers assessed a total of 685 patients for enrollment, of which 284 were randomized into the intervention group (N.=138) and the control group (N.=146). Recruitment ended when we reached a sufficient sample size. We excluded 14 patients as they withdrew their consent and requested removal of their data. At the end of rehabilitation 232 participants (80.8%) completed the questionnaire (intervention group: N.=102; control group: N.=130). Three months after the end of rehabilitation, 161 participants (56.2%) completed the questionnaire (intervention group: N.=68; control group: N.=93). We included 270 participants in our intention-to-treat analyses using multiple imputation of missing values. The flow of participants is shown in Figure 1, based on CONSORT.⁴⁰

Figure 1.

—Flow of participants. CG: control group; IG: intervention group.

Sample characteristics

The participants were on average 46.8 years old; 52.6% were women and 55.8% reported at least an intermediate level of education (at least level 3 of the ISCED-11). Table I shows the sample characteristics separately for the intervention and control groups. Both groups had comparable levels of health and functional impairment. No significant baseline differences between the groups were found.

Table I. —Sample characteristics of the participants.

	IG (N.=127)		CG (N.=143)
	N.	M (SD) or %	N.	M (SD) or %	P*
Age in years	127	46.5 (10.4)	143	47.0 (10.4)	0.667
Sex					0.662
Female	65	51.2%	77	53.8%
Male	62	48.8%	66	46.2%
Native German speaker					0.721
Yes	111	87.4%	127	88.8%
No	16	12.6%	16	11.2%
Partnership					0.809
Yes	94	78.3%	109	79.6%
No	26	21.7%	28	20.4%
Employment					0.629
Yes	113	90.4%	124	88.6%
No	12	9.6%	16	11.4%
Employment contractb					0.295
Permanent employment contract	107	96.4%	112	93.3%
Fixed-term contract	4	3.6%	8	6.7%
Off work due to sickness absence					0.537
Yes	65	52.0%	77	55.8%
No	60	48.0%	61	44.2%
Sickness absence (in weeks)	126	9.7 (9.1)	138	9.3 (9.0)	0.705
Level of education					0.480
Low	12	9.6%	8	5.7%
Medium	69	55.2%	79	56.4%
High	44	35.2%	53	37.9%
Primary outcome
Pain self-efficacy (10-60)	120	38.6 (11.4)	136	37.3 (10.0)	0.313
Secondary outcomes
Current health status (0-10)	125	4.9 (1.9)	142	4.8 (1.8)	0.722
Mental health (0-10)	125	4.9 (2.3)	141	5.2 (2.3)	0.385
Functional capacity (0-10)	125	4.4 (2.1)	138	4.1 (2.0)	0.230
Pain (0-10)a	125	3.2 (1.8)	141	2.8 (1.6)	0.073
Action-oriented coping (4-24)	123	16.3 (4.8)	135	16.0 (4.8)	0.654
Cognitive restructuring (4-24)	122	13.3 (4.7)	133	12.5 (4.3)	0.186
Subjective coping competence (4-24)	124	16.7 (4.0)	136	16.3 (4.2)	0.415
Mental distraction (4-24)	122	11.0 (4.4)	138	11.3 (4.8)	0.584
Counter-activities (4-24)	124	13.3 (4.6)	135	12.6 (4.6)	0.229
Relaxation (4-24)	122	11.4 (4.7)	137	10.8 (4.9)	0.273
Motivational self-efficacy (3-12)	124	10.0 (2.1)	138	10.2 (1.7)	0.402
Disorder and treatment knowledge (0-50)	121	23.7 (12.0)	135	24.7 (11.9)	0.500
Self-efficacy in practicing knowledge (0-20)	124	10.3 (4.4)	134	10.3 (4.6)	0.879
Electronic health literacy (8-40)	124	29.2 (6.6)	135	29.4 (5.9)	0.883
Self-informing behavior					0.807
At least once a week	41	32.8%	43	31.4%
Less than once a week	84	67.2%	94	68.6%
Work ability in relation to work demands (2-10)	122	5.9 (1.7)	140	5.9 (2.0)	0.975
Self-rated work ability (0-10)	120	4.4 (2.3)	142	4.5 (2.6)	0.612

Deviations in the number of cases in the rows are due to missing values or non-applicable information. ^aHigh values represent low pain; ^bemployed participants only; *two-sample t-test or Pearson’s Chi-Square Test. CG: control group; IG: intervention group; M: mean; SD: standard deviation.

Delivered treatment doses

A total of 234 (86.7%) patients gave consent to use their medical discharge reports, including 109 (85.8%) in the intervention group and 125 (87.4%) in the control group. Overall, we found a difference of 2.2 h in the treatment dose between the intervention group (69.9 h) and the control group (67.7 h). A detailed overview of the delivered treatment dose is shown in Supplementary Digital Material 4, Supplementary Table III. Furthermore, in the intervention group 20.2% completed 0-3 modules of the back school, 4.6% completed 4-5 modules and 75.2% completed 6-7 modules. In the control group, however, 6.4% completed 0-3 modules, 19.2% completed 4-5 modules and 74.4% completed 6-7 modules. An overview of therapy adherence is presented in the Supplementary Digital Material 5 (Supplementary Table IV).

Usability and use of the Caspar application (intervention group only)

Assessment of the Caspar application at the end of rehabilitation revealed acceptable system usability (mean=75.7, standard deviation=14.4). About 76% of the participants in the intervention group rated the Caspar application as at least acceptable (>70 points). Most patients used the application several times per week and mainly with a mobile phone. We summarized all educational and exercise videos watched per person as the total number of videos watched and divided them into three categories of usage intensity based on percentiles. Half of the patients watched 25-122 videos. Table II presents the usability and use of the digital application.

Table II. —Usability ratings and use of the digital Caspar application.

	N.	M (SD), median (IQR) or %
System usability (0-100)	99	75.7 (14.4)
System usability	99
≥90-100	22	22.2%
≥80-89	29	29.3%
≥70-79	24	24.3%
≥60-69	10	10.1%
<60-0	14	14.1%
Overall assessment of the Caspar application	99
Good or better	80	80.8%
Satisfying or worse	19	19.2%
Frequency of Caspar application use	99
Several times per week or more	80	80.8%
Once per week or less	19	19.2%
Type of electronic device (multiple answers possible)	95
Laptop	18	18.9%
Computer	4	4.2%
Mobile phone	77	81.0%
TV	8	8.4%
Tablet	20	21.0%
Total Caspar application use (number of watched videos)	127	78 (25-122)
Intensity of Caspar application use	127
Low use (0-25 videos)	32	25.2%
Moderate use (>25-122 videos)	64	50.4%
High use (>122 videos)	31	24.4%

IQR: inter quartile range; M: mean; SD: standard deviation.

Primary outcome

In our adjusted intention-to-treat analysis of the primary outcome pain self-efficacy, we tested the non-inferiority of the intervention group. At both the end of rehabilitation and the 3-month follow-up, the lower boundary of the 95% CI for the intervention group exceeded the non-inferiority margin of -4 points and indicated non-inferiority. Findings were similarly in the non-adjusted intention-to-treat analysis, the adjusted complete case analysis, and the adjusted per-protocol analysis (Table III, Figure 2).

Table III. —Pain self-efficacy at the end of rehabilitation and the 3-month follow-up.

	IG			CG			b	One-sided 95% CI
	N.	Predicted values	SE	N.	Predicted values	SE
End of rehabilitation (adjusted ITTA)	127	43.35	0.82	143	43.90	0.73	-0.55	-2.75 to ∞
End of rehabilitation (non-adjusted ITTA)	127	43.87	1.02	143	43.43	0.94	0.44	-2.30 to ∞
End of rehabilitation (adjusted CCA)	94	43.23	0.80	119	44.74	0.71	-1.51	-3.64 to ∞
End of rehabilitation (adjusted PPA)	82	43.72	1.03	93	43.70	0.90	0.02	-2.84 to ∞
3-month follow-up (adjusted ITTA)	127	45.56	1.21	143	45.32	0.97	0.24	-2.86 to ∞
3-month follow-up (non-adjusted ITTA)	127	45.88	1.26	143	45.03	1.02	0.86	-2.35 to ∞
3-month follow-up (adjusted CCA)	63	46.49	1.16	88	45.82	0.98	0.67	-2.36 to ∞
3-month follow-up (adjusted PPA)	82	45.98	1.37	93	45.14	1.17	0.84	-2.66 to ∞

b: regression coefficient; CCA: complete case analysis; CG: control group; CI: confidence interval; IG: intervention group; ITTA: intention-to-treat analysis; PPA: per-protocol analysis; SE: standard error.

Figure 2.

—Non-inferiority test of the hybrid rehabilitation using one-sided 95% confidence intervals at the end of rehabilitation and the 3-month follow-up.

Secondary outcomes

There were no significant differences in our adjusted intention-to-treat analysis (Supplementary Digital Material 6: Supplementary Table V, Supplementary Table VI) and non-adjusted intention-to-treat analysis (Supplementary Digital Material 7: Supplementary Table VII, Supplementary Table VIII) at the end of rehabilitation and the 3-month follow-up. However, the adjusted complete case analysis (b=1.83; 95% CI=0.53 to 3.14; P=0.006) as well as the adjusted per-protocol analysis (b=1.75; 95% CI=0.18 to 3.23; P=0.029) found significantly better counter-activities in favor of the intervention group at the 3-month follow-up (Supplementary Digital Material 8: Supplementary Table IX, Supplementary Table X, Supplementary Table XI, Supplementary Table XII).

Moderator analysis

There were no significant interactions between the potential moderators defined in advance and treatment group at the end of rehabilitation and the 3-month follow-up (age, sex, the level of education, electronic health literacy, and motivational self-efficacy) (P>0.05).

Discussion

In this randomized controlled trial, we compared hybrid rehabilitation that implemented a multimodal digital back school with rehabilitation that provided the same back school conventionally in face-to-face sessions in patients with back pain. We found that hybrid rehabilitation was non-inferior for our primary outcome pain self-efficacy. The results on our primary outcome were robust across all analyses (adjusted and non-adjusted intention-to-treat, adjusted complete case, and adjusted per-protocol analyses). The results for our secondary outcomes indicated better counter-activities in favor of the intervention group in our adjusted complete case and per-protocol analyses at the 3-month follow-up.

We identified three current single-arm studies and six randomized controlled trials to which we could compare our results. The three single-arm observational studies, each of which examined a digitally delivered multimodal treatment, reported significant improvements in several parameters, including pain,^41-43 findings that are consistent with the results of our study. Nonetheless, the single-arm studies provided no evidence regarding the effectiveness of the interventions compared with face-to-face treatments. The six randomized controlled trials also investigated digitally delivered treatment for patients with back pain and found that digitally delivered treatment was either as effective as usual care or even more effective for several outcomes such as pain and disability.^44-49 However, three of these studies did not examine multimodal treatments, but rather exercise programs only.^46-48 Moreover, in one study only the intervention group received a multimodal treatment,⁴⁴ and in another study the intervention group had access to the intervention for three months while the control group had access for six weeks.⁴⁵ Similarly, to our study, Di Cui et al.⁴⁹ performed a randomized controlled trial and investigated a digitally delivered multimodal rehabilitation program that comprised an exercise program, educational components, and interaction with a physiotherapist. The control group received evidence-based face-to-face physiotherapy, comprising an exercise program, educational components, and additional individual treatments (e.g., manual therapy). In both groups, the intervention lasted eight weeks. Despite assessing different outcomes than in our study, their findings are comparable with ours. Both groups made significant improvements in several outcomes, for example, pain and mental health. As in our study there were no significant between group differences at the end of the study.

Strengths and limitations of the study

The results must be interpreted in light of the following limitations. First, the lack of blinding of the participants and some of the involved health experts may have introduced a bias. Second, the high dropout rate in both groups, but higher in the intervention group, increases the likelihood of attrition bias and thus threatens the internal validity of our results. Consequently, we conducted a non-response analysis in which we compared responders with non-responders in total and in both study arms, respectively, at both follow-up dates (at the end of rehabilitation and the 3-months follow-up) (Supplementary Digital Material 9: Supplementary Table XIII, Supplementary Table XIV). We found a few significant baseline differences between the responders and non-responders, such as higher current sickness absence in non-responders, lower pain self-efficacy among non-responders, and a lower level of education among non-responders. However, further analyses suggested that both the intervention and control group were affected equally. Most importantly, the significant differences between responders and non-responders indicate that data are missing at random and support the validity of our multiple imputation. Third, the treatment adherence of the intervention group was slightly lower than that of the control group. Perhaps some participants needed more help in approaching the digital intervention at the beginning of the back school. Additionally, the weekly interactive online meetings in the intervention group were scheduled in the evening and could have been perceived as an additional burden. Fourth, our findings were generated in an outpatient rehabilitation setting and within the German welfare system. Therefore, the back school was not delivered as a stand-alone intervention. Replication of our study in other settings (e.g., a different welfare system) is worthwhile and strongly recommended. Fifth, telerehabilitation has been linked to cost reduction for health care providers as well as patients.⁹ Accordingly, our study would have benefitted from a cost-effectiveness analysis.

Besides its limitations, the study has several strengths. First, we conducted our study in eight rehabilitation centers, supporting the external validity of our findings. Second, we used an appropriate control group. The back school was developed by Meng et al.¹³ and its effectiveness was shown in a randomized controlled trial in an inpatient setting in 2011.¹⁴ When comparing our control group change scores with those of the intervention group in the study by Meng et al.,¹⁴ we found that our control group even achieved higher change scores. We therefore are confident that our face-to-face back school improved outcomes. Third, the digital back school is based on an evidence-based back school, ensuring the feasibility, acceptability, stringency, and relevance of its implementation. Fourth, we have clearly described both treatment arms so that our study could be replicated in different settings.

Our results support telerehabilitation assistance as part of a multimodal rehabilitation program for patients with chronic back pain in an outpatient setting. Further studies should show which components and which time frame of rehabilitation can be digitized without any loss of effectiveness. This could make rehabilitation more flexible. Additional studies should also investigate whether models of hybrid rehabilitation can be effectively implemented in other patient groups (e.g. patients with cardiovascular disorders) and in rehabilitation aftercare.

Conclusions

The findings of this study provide evidence that hybrid digitally assisted rehabilitation in patients with back pain is a sound alternative to face-to-face rehabilitation in an outpatient rehabilitation setting.

Supplementary Digital Material 1

Supplementary Table I

Supplementary Table I.—Description of the intervention group based on the Template for Intervention Description and Replication (TIDieR) checklist and the TIDieR-Telehealth checklist.

Supplementary Digital Material 2

Supplementary Table II

Description of the control group based on the Template for Intervention Description and Replication (TIDieR) checklist.

Supplementary Digital Material 3

Supplementary Figure 1

Variance of the residuals for the independent variables in the main model at the end of rehabilitation.

Supplementary Digital Material 4

Supplementary Table III

Overall delivered treatment dose during the rehabilitation program according to the classification of therapeutic services.

Supplementary Digital Material 5

Supplementary Table IV

Therapy adherence to the back school only.

Supplementary Digital Material 6

Supplementary Table V

Secondary continuous outcomes: adjusted intention-to-treat analysis.

Supplementary Table VI

Secondary binary outcomes: adjusted intention-to-treat analysis.

Supplementary Digital Material 7

Supplementary Table VII

Secondary continuous outcomes: non-adjusted intention-to-treat analysis.

Supplementary Table VIII

Secondary binary outcomes: non-adjusted intention-to-treat analysis.

Supplementary Digital Material 8

Supplementary Table IX

Secondary continuous outcomes: adjusted complete case analysis.

Supplementary Table X

Secondary binary outcomes: adjusted complete case analysis.

Supplementary Table XI

Secondary continuous outcomes: adjusted per-protocol analysis.

Supplementary Table XII

Secondary binary outcomes: adjusted per-protocol analysis.

Supplementary Digital Material 9

Supplementary Table XIII

Sample characteristics of the non-responders versus responders at the end of rehabilitation.

Supplementary Table XIV

Sample characteristics of the non-responders versus responders 3 months after the end of rehabilitation.

Supplementary Table XIV

Sample characteristics of the non-responders versus responders 3 months after the end of rehabilitation.