Bridging gaps across levels of care in rehabilitation of patients with rheumatic and musculoskeletal diseases: Results from a stepped-wedge cluster randomized controlled trial

Objective To compare the effectiveness of a structured goal-setting and tailored follow-up rehabilitation intervention with existing rehabilitation in patients with rheumatic and musculoskeletal diseases. Design A pragmatic stepped-wedge cluster randomized trial. Setting Eight rehabilitation centers in secondary healthcare, Norway. Participants A total of 374 adults with rheumatic and musculoskeletal diseases were included in either the experimental (168) or the control group (206). Interventions A new rehabilitation intervention which comprised structured goal setting, action planning, motivational interviewing, digital self-monitoring of goal progress, and individual follow-up support after discharge according to patients’ needs and available resources in primary healthcare (the BRIDGE-intervention), was compared to usual care. Main measures Patient-reported outcomes were collected electronically on admission and discharge from rehabilitation, and after 2, 7, and 12 months. The primary outcome was patients’ goal attainment measured by the Patient Specific Functional Scale (0–10, 10 best) at 7 months. Secondary outcome measures included physical function (30-s Sit-To-Stand test), health-related quality of life (EQ-5D-5L-index), and self-assessed health (EQ-VAS). The main statistical analyses were performed on an intention-to-treat basis using linear mixed models. Results No significant treatment effects of the BRIDGE-intervention were found for either primary (Patient Specific Functional Scale mean difference 0.1 [95% CI: −0.5, 0.8], p = 0.70), or secondary outcomes 7 months after rehabilitation. Conclusion The BRIDGE-intervention was not shown to be more effective than existing rehabilitation for patients with rheumatic and musculoskeletal diseases. There is still a need for more knowledge about factors that can improve the quality, continuity, and long-term health effects of rehabilitation for this patient group.


Introduction
Rheumatic and musculoskeletal diseases are among the largest contributors to disability worldwide. 1 The diseases affect individuals by reducing physical and psychosocial health and constitute a substantial societal burden, which is predicted to rise markedly. 2 Multidisciplinary rehabilitation is often required 3 in form of coordinated health interventions applied to optimize function and minimize disability, with the overall purpose to enable patients to live their lives in line with personal preferences, needs, and goals. 4,5 There is solid evidence that multidisciplinary rehabilitation provides beneficial health effects by improving function and health-related quality of life, [6][7][8] but the effects tend to be small and decline quickly. 9,10 One explanation may be that patients need continued support over a longer period of time to be able to reach their rehabilitation goals and implement new, healthy habits in their daily life. 11,12 A lack of efficient and systematic follow-up after rehabilitation discharge may be one factor contributing to the diminishing effects. 13 Further, without good communication, coordination, and continuity of care and support, patients may experience fragmented and poorly integrated health services from multiple providers, which again may result in suboptimal outcomes. 14,15 The need for improved quality, continuity, and care coordination within and across healthcare settings is acknowledged as a challenge to be prioritized 14 both internationally 16 and in Norway. 17 This will involve creating relationships and conditions to support informed and coherent interactions between the patient and multiple rehabilitation providers, within and across different locations, and over prolonged periods of time. 14,15 Further, patient involvement and shared decision-making in the planning and tailoring of interventions are considered crucial to improve patient motivation, quality of care, adherence to treatment, self-care, and health outcomes. [16][17][18][19] To strengthen the quality and continuity of rehabilitation, we developed the BRIDGE-intervention, designed to act as a bridge between healthcare levels. Underpinned by theories of behavioral change and empirical evidence of clinical effects, the BRIDGE-intervention comprised five main components: structured goal setting, 20,21 action-and coping planning, 21 digital self-monitoring of progress, 22 tailored telephone follow-up support, 13,23 and use of motivational interviewing 24,25 in goal setting and follow-up conversations. We hypothesized that the intervention would help patients achieve their rehabilitation goals and improve or prolong beneficial health outcomes. Accordingly, the aim of this study was to evaluate if the BRIDGE-intervention was more effective than usual care in improving goal attainment and health outcomes in patients with rheumatic and musculoskeletal diseases.

Study design and settings
The trial was developed in line with the Medical Research Council recommendations for the design and evaluation of complex interventions, 26 and built on several foregoing studies in which knowledge and tools for improving quality and continuity in rehabilitation were produced. 9,[11][12][13][27][28][29][30][31][32] Following a pilot feasibility cohort study, 11 we performed a pragmatic stepped-wedge cluster randomized trial 33 between August 2017 and August 2019. The steppedwedge design was selected to facilitate patient recruitment and the practical implementation of the intervention at multiple sites, and to protect against potential between-group contamination and disappointment effects commonly associated with a parallel roll-out while maintaining a robust methodology for scientific evaluations.
Eight Norwegian rehabilitation centers (clusters) in secondary care participated, comprising five rehabilitation institutions and three hospital rheumatology departments. The centers started simultaneously to include patients in the control condition (delivering current rehabilitation) before they one after another introduced the BRIDGE-intervention in a predefined, randomized order. Accordingly, the intervention was rolled out in sequences with an initial period in which all centers provided usual care (control), and a closing period in which all centers provided the BRIDGE-intervention to patients (Box 1). The study was registered in ClinicalTrials.gov (NCT03102814) prior to launch.

Participants
Patients were eligible for the study if aged ≥ 18 years; admitted to rehabilitation due to one of the following verified diagnoses: inflammatory rheumatic diseases, systemic connective tissue disorders, osteoarthritis, osteoporosis, fibromyalgia or chronic widespread pain, or non-specific low back, neck, or shoulder pain (persistent for > 3 months); able to read, understand and complete questionnaires in Norwegian; had access to a computer or equivalent device for digital data collection, and possessing a personal electronic credential for secure identification online. Exclusion criteria were cognitive impairment, severe psychiatric disorder(s), and fracture(s). Patients were screened for eligibility and recruited on admission to rehabilitation stay by local project coordinators at each participating center.

Control and experimental interventions
Medical treatment as usual was provided to both groups. The control intervention consisted of the rehabilitation provided at the participating centers at the study start, which varied slightly in structure and content. Six centers offered inpatient stays of 3-4 weeks duration, while two hospital departments offered shorter stays of 2 weeks, either as an inpatient (center 7) or outpatient (center 4) rehabilitation. 1 Multidisciplinary rehabilitation was provided through teams consisting of at least four health professions. The content of the rehabilitation provided included combinations of group sessions, individual sessions, and self-led activities. Details of the control intervention have been previously described 34 (Additional file 1, available at https://doi.org/10.1186/s12913-021-06164-2).
The BRIDGE-intervention was added to already existing rehabilitation and tailored to individual patients. The intervention was directed at improving the routines and communication skills of health professionals in secondary healthcare as regard goal setting, action planning, and psychological support, and at engaging and motivating patients to take an active role in pursuing their rehabilitation goals over time. Further, the intervention included telephone follow-up and help with connecting patients with required health-and social services, or other types of support, in their municipality of residence after rehabilitation discharge. Further, digital graphics based on self-reported outcomes were made available to the BRIDGE-intervention group (but not the control group), providing patient feedback on individual progression.
To facilitate the implementation of the BRIDGEintervention, two booklets were designed to support health professionals and patients, respectively, with suggestions and templates for structured communication about, and registration of goals, action plans, and follow-up, along with information, instructions, communication tools, and illustrative examples. A preparatory reflection task ("The shoe") and a YouTube tutorial on rehabilitation goals were made available to the patients before goal-setting meetings with healthcare personnel. The study materials are available online at https://diakonhjemmetsykehus.no/nkrr/ prosjekter/bridge-studien.
An overview of the BRIDGE-trial, with a description of how and when the BRIDGE-intervention was implemented in relation to the control condition and the patients' rehabilitation course, is provided in Box 1. Details on the content of the intervention components and the behavioral change techniques applied, with connections to the presumed mechanisms of action 35 are provided in Appendix Table 1.

Training of the intervention providers and fidelity monitoring
Training of the intervention providers was organized via outreach visits to each center by the research team (ALSS, IK, TND) who gave lectures, arranged workshops, and delivered the study materials. To minimize potential contamination of the control condition, the visits were arranged approximately one week before each center shifted to the intervention condition. During the trial, the centers received supervisory support from the chief study coordinator (ALSS), and regular project meetings were held that addressed study logistics and implementation.
To ensure and monitor that the BRIDGE-intervention was delivered according to the study protocol, a fidelity check list was used by the health professionals at each rehabilitation center (Appendix Table 2). Additionally, the patients reported their needs for follow-up after rehabilitation discharge, whether plans were made to meet these needs, and whether they had received follow-up, or not, in their municipality of residence. A quality indicator set for the rehabilitation of patients with rheumatic and musculoskeletal diseases 30 was used to monitor the quality of the rehabilitation provided throughout the trial.

Data collection
Patient-reported data were collected by means of an electronical portal, using a system delivered by Checkware (www.checkware.com). Information on socio-demographics, disease, medication, and lifestyle factors was collected at inclusion. Patient motivation for pursuing their rehabilitation goals was self-recorded at discharge from rehabilitation. All patients completed a core set of instruments for measuring outcomes of rehabilitation in rheumatic and musculoskeletal diseases 29 on five occasions: on admission and discharge from rehabilitation, and after 2, 7, and 12 months in their home settings.

Outcome measures
The applied rehabilitation core set includes patientreported outcome measures that cover different aspects of health and function. 29 The primary outcome in the present trial was patients' goal attainment 7 months after rehabilitation measured by the Patient Specific Functional Scale. 36 Baseline for this particular outcome was set at discharge because goals may change during a rehabilitation stay. 32 In this instrument, the patients identified 1-5 goals in terms of "important activities that you have difficulty performing, and that you think are relevant rehabilitation goals for you to proceed with after discharge." The patients rated their current status regarding each goal on 11-point numeric rating scales ranging from 0 (unable to perform) to 10 (able to perform without problems).
Secondary outcome measures were physical function, health-related quality of life, and self-assessed health status. Physical function was measured by the 30-s Sit-To-Stand Test. 37 This is a clinical field test in which a chair of standard height is used and the maximum number of full stands, with arms crossed across the chest, during 30 s is recorded. An instructional video that demonstrated the correct performance of the test accompanied the written patient instructions. Health-related quality of life was measured by the generic questionnaire EQ-5D-5L (EuroQol 5 Dimensions, www.euroqol.org), 38 which covers five domains: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Each domain is scored on a five-level scale ranging from 1 (no problems) to 5 (unable to/extreme problems). Patient responses were transformed into summary index scores by applying societal preference weights based on normative reference material from a UK population. 39 EQ-5D-5L-index scores range from below zero to 1, where 1 denotes the maximum health-related quality of life. Patients' self-assessed overall health status was recorded on the vertical EuroQol-visual analog scale, numbered from 0 to 100, with the lower endpoint labeled "The worst health you can imagine" and the upper labeled "The best health you can imagine." 38 The remaining seven outcome measures of the core set were used as tertiary outcomes: Coping was measured by the Effective Musculoskeletal Consumer Scale (0-100, 100 is best), a 17-item questionnaire assessing knowledge, attitudes, and behaviors about self-management skills. 29 Functioning in daily activities was measured by the 12-item Hannover Functional Ability Questionnaire (0-24, 0 is best). 29 Social participation was measured by one question from the COOP/ WONCA questionnaire addressing whether "your physical or mental health has limited your social activities or contact with others," rated on a scale from 1 to 5, where 1 is best. 29 Mental health was measured by the Hopkins Symptom Checklist-5 (0-4, 0 is best), a five-item questionnaire assessing mainly anxiety and depression, or degree of psychological distress. 29 Pain and fatigue were measured on 11-point numeric rating scales ranging from 0 to 10, where 0 is best (no pain, no fatigue). 29 The patients' motivation to work purposefully to achieve their stated goals was also measured on an equivalent scale; 0-10, where 10 is best (maximum motivation). 29 Sample size calculations Sample size calculations were performed based on the primary outcome using information from a prior study testing the rehabilitation core set. 29 It was estimated that a sample size of 140 participants (70 in each group) was needed to detect a mean difference of 1 point in the Patient Specific Functional Scale scores between the two groups, with a 0.05 significance level and a power of 80%, assuming a standard deviation of 1.84 for the scale, an intra-cluster correlation ρ (rho) < 0.005, approximately equal cluster sizes, and a dropout rate of 25%.

Randomization and blinding
The cluster-level randomization was performed using a computer-generated list of random numbers. The eight clusters were given a number between 1 and 8 before they were randomly allocated to one of eight sequences (Box 1). The sequential intervention rollout plan was presented to the clusters at the study start, and accordingly, the intervention providers were not blinded to group allocation.
For patients, group allocation was determined by their admission dates, and whether the pertaining clusters were in the control-or intervention phase on that date. All patients were given identical study information on admission before consenting to participate, but to avoid unblinding 40 disparate verbal information was given, by leaving out the particulars of the experimental intervention to potential participants in the control group. Hence, the patients knew they were trial participants and which treatment they were expected to receive, but they may have been blinded to which study group they belonged. Consequently, the patient-reported outcome assessments may also have been performed blinded to allocation status. Week numbers in which the clusters crossed over from the control condition to the intervention condition are provided in the initial dark grey cell for each cluster. The length of each step was 5 weeks (with an extension of 1 week for periods including national holidays). Patients recruited during the control condition stayed in the control group, and patients recruited in the intervention condition stayed in the intervention group. Individual patient-reported data were collected at 5 time points: admission and discharge, and 2, 7, and 12 months after admission to rehabilitation. The follow-up period continued throughout 2018 until the end of August 2019, implying a total trial duration of 106 weeks.
The BRIDGE-intervention comprised five interacting components, which were partly added to, and partly provided to modify existing practice, supported by training by the study team and two guidance booklets, designed for patients and healthcare personnel, respectively. The intervention was designed to improve the quality and continuity of rehabilitation processes, with a focus on a high degree of patient involvement, follow-up support, and coordination of health services across levels of care.

Statistical analyses
Descriptive statistics were used to summarize and compare the baseline characteristics of the patients in the control-and intervention groups. The main statistical analysis for estimating treatment effects was performed with an intention-to-treat approach using all available data and a 3-level linear mixed model with random intercepts for the cluster (center) and patient (repeated measures), and with treatment group, baseline values, seasonal and calendar time treated as covariates. This method provided the estimated mean outcome level at each measurement time point, and the results of the group comparisons were reported as mean differences [intervention minus control] with 95% confidence intervals (CIs). Dropout analyses were performed to determine the response rate and attrition, and to explore the characteristics of completers and dropouts. The significance level was set at 0.05. Analyses were conducted in Stata IC16.
Several sensitivity analyses were conducted to investigate the robustness of our findings and the influence of missing data. These included within-cluster comparisons of the control-and intervention group, unadjusted intention-to-treat-analyses of the total sample using all available data, and rigorous per-protocol analyses. The perprotocol population was defined based on information from the fidelity checklist combined with patient-reported data on whether follow-up was needed and planned at discharge, and received within 7 months after rehabilitation (Appendix Table 2). To be included in the per-protocol analyses, a patient was required to fulfill 16 out of 19 protocol components, in addition to having contributed with data on the primary outcome.
Additionally, descriptive statistics were used to examine which components of the BRIDGE-intervention were received or not (yes/no) in the sub-group of patients who did not receive the BRIDGE-intervention as intended (the non-per protocol population). Further, we compared the control-and intervention group (total sample) with respect to patient-reported received follow-up (yes/no) in primary healthcare at 2 and 7 months after rehabilitation by using Pearson's Chi-square test.

Patient and public involvement
Two patient research partners who participated as members of the project steering committee provided feedback on the project materials and the study's design-, conduct, and development. The steering committee further comprised one healthcare professional from each participating rehabilitation center who actively contributed with advice to better adapt the study activities to local conditions 26 and to interpret the findings. Furthermore, two study-independent representatives from primary healthcare also provided advice, together with a representative from a tertiary healthcare rehabilitation unit, and an international expert in rehabilitation based in the Netherlands.

Ethics
The study was approved by the Norwegian Regional Committee for Medical Research Ethics (REK South-East, 2017/665), and conducted in accordance with the Helsinki Declaration and the ICMJ Recommendations for the Protection of Research Participants. All study participants provided written, informed consent before enrolment.

Participant flow
Out of 769 eligible patients invited to participate, 421 consented. Of these, 46 withdrew or were excluded before baseline. A total of 374 patients (206/168 in the control/intervention group) completed the baseline assessments and were included in the analyses. Figure 1 depicts the flow of participants throughout the course of the study. The various stated reasons for dropout and exclusion are described for each group in Appendix Table 3.
Dropout analyses comparing baseline characteristics of completers (n = 245) and non-completers (n = 129) of the primary outcome at 7 months follow-up showed that non-completers were more likely to be from the intervention group, were significantly younger, smokers had shorter disease duration, and more of them (27% vs. 16%) were diagnosed with fibromyalgia syndrome, compared to completers. At 12 months follow-up, this pattern was repeated, but with slightly smaller differences (data not shown).

Baseline characteristics
The baseline characteristics of the included patients are shown in Table 1. The majority were middle-aged, married, women, with longstanding musculoskeletal disease, generally active, slightly overweight, and recipients of social security benefits, with relatively low levels of education. More than 70% used analgesics regularly, and approximately 20% used biological medicines. The treatment groups were mostly well-balanced with respect to baseline characteristics, however, some differences in the distribution of diagnoses were seen. The intervention group included a greater proportion of patients with fibromyalgia syndrome, while more patients with the inflammatory rheumatic disease were included in the control group. Further, the patients in the intervention group were significantly younger, with shorter disease duration compared to the control group.

Effect outcomes
The results from the main intention-to-treat analyses are presented in Table 2. No significant treatment effects were found for primary or secondary outcomes. For the primary outcome; goal attainment, a negligible between-group difference of 0.1 [95% CI −0.5, 0.8], p = 0.70 was found, with a confidence interval indicating at most modest effects. Similar results were found for the secondary outcomes; physical function (mean difference 0.9 [95% CI: -0.4, 2.2], p = 0.18), health-related quality of life (mean difference 0.0 [95% CI: −0.0, 0.0], p = 0.99), and self-assessed health (mean difference −0.1 [95% CI: −4.1, 3.9], p = 0.98) at 7 months follow-up. However, for the tertiary outcome functioning in daily activities, a significant result in favor of the intervention group was found (mean difference −1.7 [95% CI −2.7, −0.7], p = 0.001), which remained significant after a Bonferroni Correction. However, the effect was not maintained at 12 months. For both groups, the mean outcome scores remained fairly stable throughout all measurement time points from discharge to 12 months follow-up. Figure 2 shows the development over time in the primary and secondary outcome scores from baseline to the 12 months follow-up, by treatment arm.
The per-protocol analyses produced similar results to the intention-to-treat-analyses (adjusted/unadjusted) with respect to group differences and p-values. However, only 31% (n = 52/168) of the patients in the intervention group were included in the perprotocol analyses. Of those excluded (n = 116) from the per-protocol analyses, 14 lacked checklist information. Of the remaining (n = 102) nearly all received the BRIDGE-intervention components delivered in secondary healthcare, but most (n = 99) reported not to have received required follow-up (health-and social services) in primary care after discharge.
The results from the within-cluster comparisons of the control-and intervention groups (data not shown) also produced similar results with group differences near zero, with one exception (center 2) demonstrating a moderate positive effect of the BRIDGE-intervention, compared to the control condition (mean difference 2.7 [95% CI 0.3, 5.1]).

Follow-up received
Explorative comparisons of the control-and intervention group with respect to patient-reported received follow-up in primary healthcare showed few significant differences, except that the control group received more follow-up from nurses at medical doctors' offices within 2 months after rehabilitation, and that the intervention group received more follow-up from the social security office within 7 months after rehabilitation. As concerns the healthcare services most frequently used by the patients, the groups were relatively similar (Table 3).

Missing data
A high proportion of missing data was exposed. Missing data occurred both as consequence of  Table 3 for details on reasons for refusal, withdrawal, and exclusion. attrition and incomplete completion of questionnaires. For the primary outcome, the proportions of missing data were 29% in the control-and 41% in the intervention group at 7 months follow-up ( Table 2, numbers analyzed provided in the subtext). Similar proportions of missing data were also found for the other outcome measures at 7 months. Attrition was 19% (40/206) in the control group and 27% (45/168) in the intervention group at 7 months (Figure 1), which means that 10% and 14% of the missing data in the controland intervention group, respectively, occurred due to inadequate completion of questionnaires by patients who stayed in the study at this point.

Discussion
This study aimed to evaluate the effectiveness of providing a multicomponent health behavior change intervention (the BRIDGE-intervention) as an adjunct to usual care in the rehabilitation of patients with rheumatic and musculoskeletal diseases. The results showed no significant treatment effects of the BRIDGE-intervention on either goal attainment, physical function, health-related quality of life, or self-assessed health. For both groups, the mean primary and secondary outcome scores remained fairly stable from discharge to 12 months follow-up, indicating small, but lasting effects of the rehabilitation provided to both groups. Our findings differ from the findings of a recent systematic review of reviews summarizing the evidence base for psychological interventions in patients with rheumatoid arthritis, 41 concluding that intervention components such as goal setting, action planning, motivational interviewing, and supportive counseling, result in small to moderate improvements in biopsychosocial patient outcomes in addition to those achieved by the standard care, and that longer lasting interventions with a maintenance component appear particularly effective. They also deviate from those of two other systematic reviews 42,43 reporting that combining many health behavior change techniques yield higher effects and increase the possibility of achieving health behavior change in patients with chronic musculoskeletal conditions. One possible reason for the lack of effect in our study may be related to implementation difficulties. The BRIDGE-intervention was highly complex, 26 comprising numerous interacting components, that were delivered by a variety of healthcare personnel, within a range of different real-life settings, thereby providing considerable room for variation in the delivery of the intervention. To optimize implementation, the BRIDGE-intervention was highly protocolized with detailed manuals to guide both health professionals and patients in adhering to the intervention procedures throughout the study. However, this may unintentionally have added to the complexity of both delivering and receiving the intervention. This hypothesis is supported by the fidelity estimates, which showed that only 31% of the patients in the intervention group received the BRIDGEintervention as intended, with follow-up after rehabilitation discharge as the most frequently lacking factor. Our findings thereby confirm other reports stating that care coordination across service levels still are among the weakest elements in the rehabilitation process. 11,14,44 As the BRIDGE-intervention specifically targeted improvement of the transitions between levels of care, we expected that patients in the intervention group would receive more follow-up than the control group. However, we found that the groups received an approximately equivalent amount of follow-up, reported as a use of health-and social services after rehabilitation discharge. Variations in cluster-level delivery of the BRIDGE-intervention may potentially   explain the observed variation in the primary outcome between clusters. The way institutions and individuals adopt and integrate new trial-related work into existing routines, is influenced by the degree to which the new work is perceived to be relevant and legitimate. 45 Further, provider-level factors, such as experience, engagement, knowledge, and skills may influence the fidelity of delivery. 46 Ultimately, the responses and actions of the individuals who received the BRIDGE-intervention must have influenced its effectiveness, 45 as it depended on these individuals changing behavior in relation to set goals.  A strength of this study is the robust research design, which may have kept the included patients blinded to group affiliation and thus protected against disappointment effects and contaminant factors during the intervention roll-out, which strengthens the internal validity of the study. Further, the pragmatic orientation enabled evaluations in real-world settings, strengthening the external validity. Other methodological strengths are the use of the rehabilitation core set of outcomes for patients with rheumatic and musculoskeletal diseases, 29 reflecting the patient-reported desired effects of rehabilitation on everyday functioning across several life domains, as well as the rigorous statistical analyses accounting for calendar time and data clustering.
The study has several limitations. The cluster randomized study design is prone to postrandomization selection bias, 40 and there is a risk of performance bias as it was not possible to blind the intervention providers. Moreover, several reports stress that there might be a risk of both type I and type II errors associated with few clusters in stepped-wedge cluster randomized trials. 47,48 Although the exact lower limit for a number of clusters in such trials is currently not known, the BRIDGE trial would probably have benefited from including more than eight clusters.
A substantial limitation is a large amount of missing data, which reduces the statistical power and the representativeness of the sample, and impacts on the validity and generalizability of the results. Another limitation is the lack of longitudinal data on patient motivation and self-managed goal-directed activity after rehabilitation discharge, which could have provided valuable information in the interpretation of the findings. Finally, the fact that missing data and attrition were highest in the experimental group compared to the control group, may also indicate that receiving in the BRIDGE-intervention was burdensome to some patients.
This study demonstrates that shortcomings still exist in rehabilitation transitions between levels of care. The multicomponent BRIDGE-intervention was not shown to be more effective than usual care in terms of improving goal attainment, physical function, health-related quality of life, and self-assessed health in patients with rheumatic and musculoskeletal diseases. However, the implementation of the intervention was not optimal, as the majority of patients did not receive the required follow-up in their municipality of residence after rehabilitation discharge. This finding will be further explored in an upcoming study. Further research should focus on factors that can raise treatment fidelity among health care providers, bridge gaps across levels of care, and thus improve the quality, continuity, and long-term health effects of rehabilitation for this patient group.

Clinical messages
• The BRIDGE-intervention did not lead to better goal attainment and clinical outcomes in the rehabilitation of patients with rheumatic and musculoskeletal diseases, compared with usual care. • Further research is required to improve the continuity of rehabilitation across levels of care and the long-term health effects of rehabilitation for this patient group.

Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Research Council of Norway (grant no. 2017/260661/H10 and grant no. 328657).

ORCID iDs
Gunnhild Berdal https://orcid.org/0000-0003-4138-9465 Anita Dyb Linge https://orcid.org/0000-0001-5638-5497 Note 1. These two centres correspond to clusters 4 and 7 in Figure 1 in the present paper, but to centres 2 and 6, respectively, in the link provided (Additional file 1). 34 In collaboration with the multidisciplinary team, patients discussed and set from one to maximum of five rehabilitation goals during goal-setting meetings organized at admission and discharge from rehabilitation. The goal-setting process was structured according to the G-AP approach 1 and included four stages: (i) goal negotiation and goal setting, (ii) action planning, coping planning and measurement of confidence to complete plan, (iii) action (carry out plan), and (iv) evaluation, feedback, and reorientation. The process was guided by detailed instructions in two booklets designed for patients and healthcare professionals, respectively. 2 Plans for self-management and follow-up: Strategies for achieving individual rehabilitation goals and maintaining self-management and health-promoting behavior were discussed with the rehabilitation team and specified by the patient in written plans using predefined templates, including if-then plans by identifying potential risk situations, difficulties, and barriers, as well as sources of support.

KPMG. Evaluation of the escalation plan for habilitation and rehabilitation (2017-2019). Prepared on behalf of the
Specific examples of such plans were available, along with a scale (ranging from 0 to 10) on which the patients could rate their confidence to complete their plan. Before discharge, follow-up plans were developed by the multidisciplinary team in collaboration with the patients. These plans were tailored to the patients' individual needs and available resources in their municipality of residence.

CL1
I have/the team has collaborated with P on the development of rehabilitation goals and P has written 3-5 goals in his/her workbook.

CL2
I have introduced P to the electronic portal and P has logged in with his/her secure credential.

CL3
P has recorded his/her goals in the electronic portal (in the Patient Specific Functional Scale) and also completed the other questionnaires in the portal on admission to rehabilitation.

Rehabilitation plan and implementation phase
For each goal, P has in collaboration with me/the team developed a rehabilitation plan in his workbook. The plan includes strategies for resolving potential risk situations and implementation barriers.

CL5
I have/the team has given P positive feedback on what has worked well in the implementation of planned measures.

CL6
P and I/the team have over time adjusted the specific measures and sub-goals according to the patient's needs, in order to achieve a belief in mastery (self-efficacy) and agreed on expectations of the result.

CL7
Rehabilitation goals after discharge and the individually tailored follow-up plan P has written 3-5 goals in the workbook for the time after discharge and P has transferred these goals to the electronic portal (Patient Specific Functional Scale at discharge).

CL8
P has completed the rehabilitation plans for each of the goals, including strategies for resolving potential risk situations and barriers. P has put this in writing in the last part of his/her workbook.

CL9
P has a distinct plan for his/her own efforts and has transferred it as "plans independent of health personnel" in the electronic portal.

CL10
I have /the team has collaborated with P about relevant follow-up interventions from other services, and P has entered the result of this collaboration in the electronic portal as "follow-up in the municipal health service after the rehabilitation discharge."

CL11
P and I have agreed on a date and time for a follow-up phone call concerning rehabilitation goal progression and goal-directed activities at home, approximately one month after discharge.

CL12
Use of the electronic portal and the digital feedback report after discharge I am assured that P feels informed and capable to log in to the electronic portal at home.

CL13
I have provided P with information on how the digital feedback report (regarding P's scores on the rehabilitation core set) can be used after discharge, alone or together with others, as a support to achieve the goals and maintain the initiated activities and lifestyle changes.

CL14
Follow-up after discharge I have conducted the first mandatory phone call guidance with P, as agreed.

CL15
If applicable: several additional phone call follow-ups have been implemented (2, or 3, or up to maximum of 4 calls).

CL16
Guidance on goal-directed activities and goal attainment have been topics in the telephone conversation(s), including an evaluation of potential needs for necessary follow-up in the municipality or local community.

CL17
If needed: I have helped P to contact relevant healthcare services after discharge.

CL18
CL sum (counts of yes) ≥ 16 was required to be included in the PPP (Continued) The per-protocol population in the BRIDGE trial was based on information from two sources: A fidelity checklist for the delivery of 18 intervention sub-components, completed by the patient's primary healthcare contact at the rehabilitation centers during the intervention delivery (Table 1, items 1-18) (yes/no). An electronic portal in which the patients in the intervention group recorded from their home settings whether needed and planned follow-up was received, or not, after rehabilitation discharge ( To be included in the PP population, the patients had to: Either A) have reported a need for follow-up after rehabilitation discharge, B) planned follow-up according to needs before discharge, and C) received at least one type of follow-up in his/her residential municipality according to stated needs, within 7 months after rehabilitation discharge, or reported no need for follow-up after discharge ( Table 2).
The rationale behind the construction of the composite variable 19 (yes/no) is shown in Table 2, and a final "yes" was regarded as mandatory to be included in the per-protocol population. In addition, the patients had to have contributed with data on the primary outcome at the 7 months measurement time point. Furthermore, 16 out of the 18 checklists (CL) items had to be verified delivered (ticked yes) by the patient's primary contact (health professional) at the rehabilitation center. Table A3. Listed reasons for patient rejection of study participation, exclusion, and drop out.
T0 reasons for declining participation (total n = 348) 1 in the control phase (n = 166) in the intervention phase (n = 182) Did not fulfill inclusion criteria (n = 46) Concerns regarding electronic data security (n = 1) Lack of computer skills or not fond of electronic tasks (n = 7) Not interested or no stated reason (n = 63) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 24) Reluctance towards longstanding commitment or a follow-up period decided by others (n = 13) Other illness (n = 2) Other reasons (n = 10) Did not fulfill inclusion criteria (n = 24) Lack of computer skills or not fond of electronic tasks (n = 5) Not interested or no stated reason (n = 125) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 16) Reluctance towards longstanding commitment or a follow-up period decided by others (n = 10) Other illness (n = 2) T1 reasons for exclusion or withdrawal before baseline assessments (total n = 46) 2 in the control phase (n = 26) in the intervention phase (n = 20) Incomplete baseline data (n = 4) Did not fulfill inclusion criteria (n = 6) Technical login problems (n = 7) Concerns regarding electronic data security (n = 1) Not interested or no stated reason (n = 3) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 3) Other illness (n = 2) Incomplete baseline data (n = 3) Did not fulfill inclusion criteria (n = 2) Technical login problems (n = 3) Lack of computer skills or not fond of electronic tasks (n = 1) Not interested or no stated reason (n = 4) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 7) T2 reasons for exclusion or withdrawal before discharge (total n = 23) 3 in the control phase (n = 11) in the intervention phase (n = 12) Not interested or no stated reason (n = 5) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 2) Other illness (n = 4) Interrupted rehabilitation before completing discharge assessments (n = 1) Concerns regarding electronic data security (n = 1) Not interested or no stated reason (n = 3) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 5) Other illness (n = 2) T3 reasons for exclusion or withdrawal before 2 months follow-up (total n = 32) 4 in the control phase (n = 16) in the intervention phase (n = 16) Not interested or no stated reason (n = 13) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 2) Other illness (n = 1) Technical login problems (n = 1) Not interested or no stated reason (n = 11) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 2) Deterioration of health status (RMDs) (n = 2) T4 reasons for exclusion or withdrawal before 7 months follow-up (total n = 30) 5 in the control phase (n = 13) in the intervention phase (n = 17) Lack of computer skills (n = 1) Not interested or no stated reason (n = 9) Lack of energy or questionnaires and/or intervention perceived as too comprehensive (n = 1) Deterioration of health status (RMDs) (n = 1) Other illness (n = 1) Technical login problems (n = 1) Not interested or no stated reason (n = 15) Other illness (n = 1) T5 reasons for exclusion or withdrawal before 12 months follow-up (total n = 32) 6 in the control phase (n = 20) in the intervention phase (n = 12) Technical login problems (n = 2) Not interested or no stated reason (n = 18) Technical login problems (n = 2) Not interested or no stated reason (n = 9) Other reasons (n = 1) 1−6 T0 = request for participation, T1 = baseline, T2 = discharge, T3 = two months follow-up, T4 = seven months follow-up, T5 = twelve months follow-up.