A transdiagnostic group exercise intervention for mental health outpatients in Germany (ImPuls): results of a pragmatic, multi-site, block-randomised, phase 3 controlled trial

Description of research question and justification for undertaking the trial

Epidemiological data from 2019 suggests that 15.6 % of the German population suffered from any mental disorder in 2019 (point prevalence).¹ Mental disorders result in a considerable burden of disease, for example accounting for 6.4 % of overall disability-adjusted life years (DALYS) assessed in the 2019 epidemiological survey.² From 2008 to 2018, the proportion of mental disorders among all causes of death increased from 2.2 % to 6.1 %.³ The most prevalent disorders in Germany are anxiety disorders and trauma- and stress-related disorders (point prevalence: 7.1 %), major depressive disorders (point prevalence: 4.3 %), as well as insomnia (point prevalence: 4 %).^2,4 Notably, these disorders often occur comorbidly and share common underlying aetiological and even maintenance mechanisms, such as the experience of stressful life events and high perceived stress, low self-efficacy, sleep disturbances, elevated anxiety sensitivity, or repetitive negative thinking.^4–14

Worldwide, 27.2 % of the DALYS attributable to mental disorders can be explained by major depressive disorders and 16.3 % by anxiety disorders.¹⁵ In 2015, health care costs in Germany caused by mental disorders amounted to 44.4 billion euros.¹⁶ Of this, 8.7 billion euros can be attributed to major depressive disorders, 1.7 billion euros to phobic and other anxiety disorders, and 1 billion euros to insomnia. Mental disorders account for 13.1 % of total costs and represent the second highest cost group after cardiovascular disorders (46.4 billion euros, 13.7 % of total costs). Major depressive (RR = 2.63) and anxiety disorders (RR = 1.41) have also been shown to increase the risk of cardiovascular disease.¹⁷ Besides direct costs (e.g., treatment costs), mental disorders cause indirect costs on the German job market. With 14.4 billion euros overall, mental disorders caused the second-highest lost production costs of all diagnosis groups in 2019.¹⁸ They further caused 117.2 million days (16.5% of all days) of incapacity to work, which is the longest absences per sick leave of all disorders.¹⁹ Amongst mental disorders, major depressive disorders accounted for the most days of incapacity to work (33.9 million days), followed by trauma- and stress-related disorders (21.6 million days). Anxiety disorders accounted for 7.6 million days and insomnia for 0.5 million days.

Despite the high prevalence and severe negative impact of mental disorders, it is estimated that in Germany only 10 % of all affected individuals receive evidence-based treatment and only 2.5% receive psychological treatment.²⁰ In addition, even those receiving psychological treatment often have to wait before treatment initiation; for example, 40 % of outpatients waited three to nine months to start psychotherapeutic treatment in German health care settings.²¹ Longer waiting times are associated with worsening and chronicity of symptoms and the development of comorbid conditions.²² The high prevalence and severe burden of mental disorders in combination with the large gap between people in need for treatment and those actually receiving it illustrates the need to develop alternative efficacious, effective and efficient treatments.²³

Exercise, defined as physical activity that is planned, structured and repeated, with the primary aim to improve or maintain physical fitness, has revealed positive therapeutic effects for diverse mental disorders.^24,25 Most of the studied exercise interventions include aerobic activities (i.e. running) or a combination of aerobic exercise with strengthening activities. Recent meta-analyses on major depressive disorders and insomnia have shown large effects for exercise that were comparable to those of psychological treatment and psychopharmacological treatment.^26,27 A recent meta-analysis on PTSD found small to moderate effects; however, in two of the four studies included, the intervention comprised yoga.²⁸ Looking at more recent evidence from RCTs focusing on interventions including aerobic exercise, large treatment effects were found.^29,30 For panic disorder (with and without agoraphobia), RCTs have revealed large effects on symptomatology with both, acute exercise and structured multi-week aerobic exercise programs.^31,32 In addition, moderate to large effects have been reported for exercise as an augmentation to TAU for major depressive disorders, panic disorder and PTSD.^30,33,34 Exercise also seems to be efficacious in a transdiagnostic way as it targets the aforementioned underlying aetiological and maintaining factors that are present across depressive disorders, anxiety disorders and insomnia such as low self-efficacy, sleep disturbances or higher stress perception and adaptive stress coping.^35,36 Key components of exercise interventions that have shown optimal therapeutic efficacy among patients with major depressive disorders, insomnia, panic disorder with or without agoraphobia and PTSD include aerobic exercise at a minimum of moderate intensity (MVAE) either or a combination of MVAE with resistance training, conducted two to three times per week, for 10 weeks with a session duration of at least 30 minutes, partially supervised or non-supervised.^27,35,37,38

Exercise might not only be a promising efficacious (transdiagnostic) treatment but also carry the advantage of being highly efficient, since it can be delivered in group settings with relatively short durations, can be offered to patients with heterogeneous and burdensome mental disorders, can be expected to show a low likelihood of adverse effects, comes at a relatively low cost, and is suited to reduce the risk for cardiovascular diseases that frequently occur comorbid with mental disorders.^27,39,40 Furthermore, exercise can be performed and continued independently without professional supervision or only remote supervision. However, individuals suffering from mental disorders often have difficulties to initiate and maintain a physically active lifestyle, which may be related in part to deficiencies in motivation and exercise-related self-regulatory skills in this population.^41,42 Reassuringly, there is evidence showing that exercise adoption and maintenance are mediated by motivational and volitional aspects, such as intention strength, action planning and barrier management.⁴³ A recent meta-analysis shows that especially self-efficacy in building intentions and action planning is crucial for sustained exercise behavior change.⁴⁴ One possible way to promote such motivational and volitional aspects are the application of behavior change techniques (BCTs).^45,46

The combination of behavior change techniques and exercise as a structured intervention appears highly promising in terms of initiating a sustainable exercise behavior change. Since outpatients have less supervision and contact to their therapists compared to patients in inpatient or rehabilitative mental health care settings, structured exercise interventions in combination with behavior change techniques to overcome general and disorder-specific barriers might be especially important within the outpatient mental health care setting. Indeed solely exercise on prescription (or on referral) for outpatients shows drop-out rates of nearly 80%, whereas structured exercise interventions in combination with BCTs for outpatients show lower dropout rates and stronger effects on mental health.27,47

Therefore, ImPuls was developed and evaluated as a complex exercise program with respect to the Medical Research Council (MRC) framework, specifically designed to scrutinize an additional treatment option in the outpatient mental health care system in Germany.48 It has been successfully evaluated in terms of efficacy and acceptability in a feasibility study for outpatients waiting for psychotherapeutic treatment.49,50 A broader and more comprehensive pragmatic trial was needed to explore the extent to which the intervention also achieves its effect in a real-world setting (e.g., with exercise therapists working in the outpatient setting as intervention deliverers alongside their daily business; as an add-on to treatment as usual; with a realistic referral system).51 Therefore, it is now conducted in a pragmatic multi-site randomized controlled trial to investigate efficacy and cost-effectiveness within the real-world outpatient setting.52

The complexity of ImPuls (i.e., several interacting factors [e.g., BCTs and Exercise], involvement of different actors [exercise therapists, managers of outpatient rehabilitative and medical care facilities, patients] etc.) and the future need to implement the intervention into a comprehensive health service provision prompts the necessity of research beyond a pure evaluation of efficacy, namely process evaluation. Thus, the ImPuls study is accompanied by a comprehensive process evaluation based on the MRC framework and its complement, the former of which provides comprehensive and detailed guidance.53-55 Using a mixed-methods approach may be particularly helpful to understand multiple perspectives, multiple types of causal processes and multiple types of outcomes which in turn are common aspects of implementation research.56 Process evaluation in studies evaluating exercise interventions has been slowly emerging during the last decade.57-59 However, process evaluations of exercise interventions offered to patients with mental disorders is rarely conducted. For example, one study focused solely on selected aspects like adherence rate of the participants, acceptability and feasibility, which means that only specific subcomponents of the Medical Research Council (MRC) Framework were taken into account.55,60 Other studies exclusively conducted qualitative interviews, which ideally should be complemented by quantitative methods to provide an encompassing insight into the processes relevant for implementation.56,61,62 Another study heeds the aforementioned deficiencies, yet apparently seems to omit investigation of interactions (e.g., between participants and the intervention / -deliverers) with regard to the MRC framework key component mechanisms of impact.63 Given the lack of comprehensive process evaluations accompanying exercise programs for patients with heterogeneous mental disorders, the respective evidence for implementation conditions is weak. Consequentially, further research in this area is needed.

Explanation for choice of comparators

The TAU condition will be modelled to represent the typical treatment patients receive in the German outpatient health care system. Therefore, patients will not be actively provided with any treatment but patients are allowed to receive any intervention that is available to them. Any evidence-based treatment provided by the outpatient mental health care system will be recorded, i.e., any psychiatric/pharmacological or psychological/psychotherapeutic intervention. Interventions, delivery or dosage of the intervention can be changed and adapted during the course of the study.

Specific objectives and hypotheses

Despite the promising evidence for MVAE as an intervention for patients with mental disorders, exercise programs or professional exercise therapy are currently not provided as regular health services within the outpatient mental health care system in Germany. With the aim of combining the current evidence on the efficacy of MVAE and sustained exercise behavior change with specific demands of a real-world outpatient health care setting, ImPuls was developed as a manualized group exercise intervention for physically inactive outpatients suffering from major depressive disorders, insomnia, panic disorder with or without agoraphobia and PTSD.⁴⁸ A feasibility study found moderate long-term effects in patients waiting for psychotherapeutic treatment who completed ImPuls compared to a passive control group.^49,50 ImPuls integrates recent findings about the optimal modalities of exercise for therapeutic efficacy, such as optimal frequency, intensity, time/duration and type of exercise (FITT criteria) for the targeted disorders and evidence regarding sustainable behavior change by integrating behavior change techniques (BCTs). The components of this intervention are further tailored towards the specific needs of outpatients with mental disorders in the current German mental health care setting. Specific features are 1) the inclusion of a broad range of heterogenous diagnoses for which prior research has demonstrated therapeutic transdiagnostic efficacy, 2) intervention delivery in group format, and 3) short duration (i.e., only 4 weeks of supervised MVAE sessions carried out inhouse in each study site). The aim of the current pragmatic trial is to investigate the efficacy and cost-effectiveness of implementing ImPuls within the outpatient mental health care setting in Baden-Württemberg, a representative state in South-West Germany. The following hypotheses will be tested:

• 1. Participants in the intervention condition, who have received ImPuls in addition to TAU, will show lower global symptom severity at post-treatment and follow-up assessments compared to a control condition with TAU only.

• 2. Overall costs in the intervention condition will represent a significant saving for the public health system compared to the control condition at post-treatment and follow-up assessments.

• 3a. The intervention will lead to significantly higher levels of MVAE at post-treatment and follow-up assessments compared to the control condition.

• 3b. The effect of condition on the reduction of the primary outcome global symptom severity will be mediated by an increase in MVAE.

• 4. Participants in the intervention condition will show lower disorder-specific symptoms (major depressive disorder, insomnia, panic disorder with or without agoraphobia and PTSD) compared to participants in the control condition at post-treatment and follow-up assessments

We will further assess, if participants in the intervention condition show more instances of clinically significant change compared to participants in the control condition at post-treatment and follow-up assessments (Additional analysis to Hypothesis 1 and 4).

The main objectives of our process evaluation are a) to support the findings of the ongoing pragmatic randomized controlled trial by confirming that its efficacy is truly attributable to the ImPuls intervention and b) to discover further crucial factors for the implementation of ImPuls into real-world outpatient mental health care settings.

The main research questions of the process evaluation are:

1. Implementation:

   1. To what extent do our actions empower exercise therapists (i.e., competence, acceptance) to deliver the intervention?

   2. To what extent do exercise therapists implement intervention components as intended (treatment fidelity) and what are reasons for its potential variance?

   3. Which strategies recruit the most patients and how valid are the referrals in terms of acquisition/inclusion?

   4. How do referring healthcare professionals rate the ImPuls intervention in terms of acceptability, appropriateness and feasibility?

   5. To what extent are all ImPuls sessions offered as planned and all telephone contacts made as scheduled?

2. Context:

   1. What barriers and facilitators do exercise therapists and managers experience concerning the implementation of the ImPuls intervention?

3. Mechanisms of Impact:

   1. To what extent do attitudes of exercise therapists towards the ImPuls intervention (e.g., acceptability, appropriateness) moderate the treatment effects?

   2. To what extent do patients’ integration of core components of the ImPuls intervention (e.g., amount of exercise, barrier management, goal-setting) as well as changes in respective individual behavioral determinants (e.g., action and coping plans; physical activity-related health competencies) mediate the treatment effects?

   3. To what extent do patients’ integration of motivational/volitional core components of the ImPuls intervention (e.g., barrier management, goal-setting, phone contacts) as well as changes in respective individual behavioral determinants (e.g., action and coping plans; physical activity-related health competencies) mediate its effect on their exercise adherence?

   4. To what extent do changes in patients’ transdiagnostic psychological processes (e.g., emotional regulation, repetitive negative thinking or perceived stress) mediate the treatment effects?.

Description of trial design

• Download figure
• Open in new tab

Figure P1:

Research design and measurement points of the entire trial including process evaluation.

Note: follow-up = 12 months after randomization; Inter-1 = once at the end of the supervised ImPuls phase (week 1-4); inter-2 = once at the end of the partially supervised ImPuls phase 1 (week 5-12); inter-3 = once at the end of the partially supervised ImPuls phase 2 (week 13-24); post = after completion of the intervention (12 months after randomization, week 24-26); post-d = end of study period in the outpatient facility; pre = prior to randomization; prep-1 = following the first training; prep-2 = following the second training; rando = after randomization, prior to intervention start in the outpatient facility.

The study will be led by researchers based at the University of Tuebingen in Germany, and will be conducted in ten different study sites across Baden-Württemberg, a region in South-West Germany. Study sites are local outpatient rehabilitative and medical care facilities which were selected to cover the different regions in Baden-Württemberg. The entire project will be conducted between September 2020 and June 2024. The study has been registered at the German Clinical Trial Register (ID: DRKS00024152, 05/02/2021) and has been approved by the local ethics committee for medical research at the University of Tuebingen (ID: 888/2020B01, 02/11/2020). A pragmatic multi-site block-randomized controlled trial with two treatment arms (ImPuls + TAU vs. TAU) and three points of assessment (pre, post, follow-up) will be conducted (see fig. 1). All outcomes will be included at all assessments. Study completion and reporting will be carried out in accordance with the Consolidated Standards of Reporting Trials (CONSORT), the Template for Intervention Description and Replication (TIDieR) and the Consensus on Exercise Reporting Template (CERT).^64,65

Description of study settings

Data will be collected in Baden-Württemberg, a region south-west of Germany. Study sites are mostly big centers of psychosomatic, orthopedic or cardiological rehabilitation and one outpatient unit of physiotherapy. A list of all locations/study sites can be obtained in the registration: https://drks.de/search/en/trial/DRKS00024152

Eligibility criteria

ImPuls Intervention

The exercise intervention “ImPuls” ⁴⁸ will be delivered to groups consisting of 6 patients and will be divided into a supervised and partially-supervised period. BCTs, such as goal setting, self-monitoring, formation of concrete exercise plans and coping planning, will be integrated to promote sustained exercise behaviour change.^45,46,66 The intervention structure and contents are displayed in Figure P2 and Table P1. Participants will receive ImPuls in addition to TAU.

Criteria for discontinuing or modifying allocated interventions

Participants who miss more than four consecutive sessions during the four weeks (two consecutive weeks) of the supervised phase (Weeks 1-4) of the intervention (≥40 %, due to any reason) but continue the study assessments are defined as treatment dropout, as they did not receive the minimum intended dose of the intervention. They can still participate in the remaining sessions of the intervention and will be asked to complete all remaining assessments. A treatment dropout is only counted in the time before an official study dropout (e.g., if a patient completely drops out of the study in Week 1, the patient is indicated as a study dropout and not as treatment dropout). If patients never attended treatment sessions but still take part in the assessments (are not defined as study drop-out), they are solely defined as treatment dropouts.

Participants who intentionally drop out of the entire study, are no longer accessible, or never/no longer participate in assessments are defined as study dropouts. It is always necessary to ask for the exact time when the patient decided to drop out of the study. Participants who intentionally discontinue the treatment or did not receive the minimum intended dose of the intervention (defined as treatment dropouts) and discontinue assessments will be also defined as study dropout. All participants agree to be asked to specify their reasons for study/treatment-dropout on a voluntary basis (once, after treatment/study dropout). If available, reasons for discontinuation will be reported. If more than 50 % of participants discontinue their participation in the intervention group during the first five weeks (Weeks 0-4) of the delivery, discontinuation of the group due to lack of economic efficiency for the study sites is possible. If only one participant of the intervention group remains, he or she can no longer receive the intervention, as delivery of the intervention to less than two participants no longer qualifies as “group-based intervention”.

All participants, including dropouts, will be asked to complete all of the following study assessments on a voluntary basis.

Strategies to improve adherence to intervention protocols, and any procedures for monitoring adherence

Relevant concomitant care and interventions that are permitted or prohibited during the trial

ImPuls will be delivered in addition to standard care/treatment as usual (TAU). Any other standard treatment (psychological or pharmacological treatment) covered by German health insurances are permitted.

Primary, secondary, and other outcomes

Exercise behavior/MVAE

The assessment of self-reported exercise duration and frequency and the assessment of accelerometer -based moderate to vigorous physical activity will serve as two proxies for MVAE. Exercise in minutes per week will be assessed using the self-report Exercise Activity Index of the Physical Activity, Exercise, and Sport Questionnaire [BSA questionnaire].⁹⁴ Participants specify type, duration, and frequency of exercise in the last four weeks. A weekly average of more than 200 minutes of exercise will be checked again to see if it makes sense and in the event of a mistake (i.e., repeated entry of an activity already stated in the leisure time index) it will be defined as missing; in the event of an unrealistic entry, it will be defined as missing as well.

Moderate to vigorous physical activity (MVPA) will be assessed via accelerometer-based sensors (Move 4, movisens GmbH). The sensor assesses physical activity of a person based on kinematic data in three dimensions and atmospheric air pressure. This allows to estimate the amount of physical activities of different intensities for a specified time period based on validated algorithms.⁹⁵ Patients will wear the sensors on the hip for seven consecutive days. A sensor record is considered valid and used for analysis if the sensor was worn for eight hours on at least four out of seven days.⁹⁶ In addition, the participants complete a protocol, in which non-wear time or other problems will be indicated. In case patients indicate medical problems that resulted in sedentary behavior, data will be set manually to N/A. Raw data will be checked, if non-wear time indicated by the participants is identified by the sensors. The data of valid days (at least eight hours wearing time) are added up and divided by the number of valid days and multiplied by 7.⁹⁷ If a person has less than four days with eight hours of sensor data each, then the data for that whole week is set to N/A and will be imputed. Volume of moderate to vigorous physical activity will be indicated in minutes/week and calculated by the daily physical activity of at least moderate intensity (≥ 3 MET)/minutes.

Sociodemographic data

The following sociodemographic data will be assessed during the pre-assessment: gender, age, marital status, living situation, having children, nationality status, mother tongue, highest level of education, current work situation, duration of current mental health problems, use of online therapy programs or therapy applications, use of programs to increase physical activity, current psychiatric treatment, use of psychiatric medication (current and in the past), previous physical and psychological disorders, outpatient psychological treatment (current and in the past) including number and time period, previous inpatient treatment because of psychiatric problems including number and time period, previous day-care treatment including number and time period. Due to technical reasons the variable current psychological treatment at pre-assessment needed to be reassessed by several patients at post-assessment. Therefore, we have several missing answers to this question at pre-assessment.

Assessments to ensure internal validity

Estimated number of participants / Sample

The sample size was determined a priori using power analysis (G*Power, version 3.1.9.2).^131,132 Power analysis was conservatively based on the lowest symptom-related post-treatment effect of exercise (vs. TAU/waiting list) on all included clinical disorders, namely the effect size of d = -0.348 (g = - 0.347) for symptoms of post-traumatic stress disorder.²⁸ A two-sided t-test, alpha level of 0.05, a test power of 80%, an equal cell population, and a dropout rate of 30% were assumed. This calculation resulted in N = 375, which is conservative enough to detect the lowest expected treatment effect at the post-treatment phase. However, we regarded this sample size as the minimum and targeted a total sample of up to N = 600, in order to have enough statistical power for further analysis, e.g., looking at moderators of treatment effects. In April 2022 we decided to recruit the minimum sample size. Due to various measures to contain the COVID pandemic, recruitment was delayed and made more difficult, which is why the original target of 600 participants could not be reached.

Recruitment / strategies for achieving adequate participant enrolment to reach target sample size

Patients will be recruited mainly via inpatient psychiatric departments, family practices, general practitioners and psychiatric and psychotherapeutic outpatient units. The project will be conducted in collaboration with two health insurances, the AOK Baden-Württemberg (AOK) and the Techniker Krankenkasse (TK), who will support the recruitment with a targeted approach of general practitioners, psychiatrists and psychotherapists. All hospitals, clinics and practices will receive information material, such as flyers and posters, to inform their patients about the project. In addition, AOK and TK will publish articles in media distributed for their members during the course of the project. The TK further will inform eligible patients directly via phone calls. Recruitment will be additionally performed via social media posts (Instagram and Facebook), newsletters of professional associations, email distribution lists of universities, local influencers and regional newspapers and television.

Interested patients will first attend a preliminary telephone screening of eligibility criteria and will receive general information about the project (see also fig. 2). Patients will be screened for somatic contraindications for exercise via the Physical Activity Readiness Questionnaire (PARQ) and will be informed that they have to provide a physician referral for ImPuls before pre-assessment.¹³³ In case of any suspicion of somatic contraindications, that might oppose participation (e.g., heart diseases or orthopedic problems), patients will be asked to provide an additional medical consult from their general practitioner or a medical specialist. During this screening, the recruitment strategy is assessed and recorded in the telephone lists in order to assess recruitment success per referrer. Institutions (i.e. clinic, psychotherapy practice, doctor’s practice) are evaluated before pure information material (distributed flyers, posters, table posters) if participants state that they were made aware of the ImPuls program by employees of these institutions. If participants state that they only received the flyer (without a doctor’s or psychotherapist’s consultation), the recruitment strategy is assigned to flyers as well as magazines, newspapers, radio reports, etc., respectively.

Eligible participants will be invited for a first inhouse meeting taking place in a study site close to their residence. Within the meeting, they will provide informed consent for study participation, receive information about the study site, create an individual pseudonym and will be screened initially for symptomatology related to the exclusion criteria to prepare for the structural diagnostic interview. Only scheduled telephone contacts (this excludes miscalls or information-only calls) and only completed interviews at study sites will be counted as telephone screenings and interview sessions, respectively. Discontinued or canceled interviews due to any reasons will be counted as exclusion after telephone screening. The telephone screening and initial interview will be performed by trained research assistants. Following this first inhouse meeting, psychologists with a M.Sc. degree undergoing a training in cognitive-behavior therapy, who will be trained by an external expert for structured clinical interviews, will conduct the structured clinical interview for DSM-5 (SKID-5-CV) to confirm eligibility.¹³⁴ The interviewers were all trained by an external lecturer specialized in assessment using the structured clinical interview for DSM 5 (SCID-5). The entire training consisted of 2 x 3 days with a main focus on hands-on role play-based training for conducting the SCID. Teams of three were formed. The first person was the patient, the second the interviewer, and the third person acted as an external rater who also rated the items of the SICD. If the items did not match, critical items were discussed until there was 100 % agreement between raters. During the assessment phase, weekly supervision sessions with a licensed professional supervisor were held to discuss critical cases. In addition, the supervisor could be contacted at any time during or after a structured interview to discuss ambiguities in the inclusion and exclusion criteria.

Once six patients at the same site will be found to be eligible for participation, they will receive online questionnaires via the web-based data management system REDCap and a accelerometer-based Physical Activity sensors (MOVE 4; movisens GmbH) which will be worn for seven consecutive days (pre-assessment).^103,104 The method used to collect sex/gender data is based on self-reports (female, male and other). Online questionnaires and the assessment of physical activity will be carried out within a period of 14 days. On Day 15, the six patients will be randomized as a group to either the intervention or control condition. In case of an assignment into the intervention condition, study sites will have to start the intervention within 14 days. Before the start of the intervention, global symptom as well as disorder-specific symptom severity will be assessed again to check whether participants meet the cut-off criteria for a mental disorder. In each study site, 60 patients are planned to be recruited and randomized, resulting in 10 allocations per site. The intervention group will complete the exercise intervention in addition to TAU, while the control group will receive TAU within the real-world outpatient mental health care setting in Germany. The TAU condition is intended to represent the typical treatment patients receive in the German outpatient health care system. Thus, patients will not actively be provided with any treatment but they will be allowed to receive any intervention that is available to them. The procedure of the pre-assessment phase will be repeated six months (Post) and 12 months (Follow-up) after randomization. After the completion of all assessments, patients of the control group will receive 450 € as reimbursement for their time.

Allocation

Six eligible patients are randomized as a group to either the intervention or control condition. The randomization sequence is generated independently of the study coordinator and the research team responsible for data collection and management. The sequence is generated using a varying-size permuted block design, stratified by study site. This procedure ensures an appropriate balance in the number of treatment and control groups per study site. Randomization codes are generated digitally and concealed on a secure system. The group-allocation sequence is concealed from the research team responsible for data collection and management until the assignment. Allocation is performed through RedCap “Randomization Module” on the day of intervention assignment with no prior access for anyone but the data manager.

Masking/Blinding

Patients are not blind to allocation. No assessors need to be blinded since primary and secondary outcomes are not based on clinician/assessor ratings. Accelerometer-based data is prepared and processed by trained student assistants according to predefined rules and specifications. Raw data is stored in REDCap and monitored by an external unblinded data manager. The sponsor and his research team responsible for data collection and management and any personal in contact with patients is blinded regarding the randomization procedure. The data analyst of the main analysis regarding efficacy (Hypotheses 1, 3, 4) is blinded regarding the allocation. He receives the final dataset, which is masked for the treatment condition (the “condition” variable only informs Condition A or B but no real labels of the treatment and control conditions). The data analyst will be unblinded after completion of the main analysis regarding efficacy. Analyses regarding health economics (Hypothesis 2) and further analysis regarding implementation/Process evaluation will be conducted after unblinding. An unblinded data manager handles the raw data when exporting the data from REDCap. No one other than the external data manager has access to the longitudinal data before the statistical analysis is completed.

Plans to promote participant retention and complete follow-up

In accordance with the Evaluation Concept and Record of Processing Activities, all data for evaluation runs through REDCap and network drives on a secured LMU internal server providing a stable and traceable coding, storage and security structure as well as data quality measures such as automatic survey reminders, alerts to researchers for telephone reminders to participants, and data type and range checks where appropriate. Missing values on surveys have to be confirmed by participants to be intentional before being able to proceed to the next questionnaire.

The process evaluation data collected via app are stored pseudonymously on a server of the Central Office of the University of Tuebingen and transmitted pseudonymously to the central data management site LMU Munich. Likewise, videos of the group sessions will be recorded by the University of Tuebingen as part of the process evaluation. These videos will be stored on a hard-drive at the Central Office of the University of Tuebingen and transferred to central data management site LMU Munich via secured network drive connection. The routine data of the health insurances (AOK/TK) include outpatient, inpatient, as well as medicines, remedies and aids six months prior and up until six months after completion of the intervention period and are transmitted to the trust center of the Technical University Munich on the basis of a pseudonym assigned by the Tue-CO, which is not known to the patients, and from there double pseudonymized to the Technical University Munich. The data stored at the central data management site LMU Munich (primary and secondary endpoints, data on progression diagnostics and process evaluation) are transmitted to Technical University via a secure connection. From there, they are transmitted to the trust center with the routine data and will be double pseudonymized.

Data management

Statistical Methods

To test Hypotheses 1, 2, 3a and 4 we will use multilevel modeling to establish the treatment effects on the primary outcome (global symptom severity) and the secondary outcomes (major depressive disorders, insomnia, panic disorder, agoraphobia, PTSD, QALYs, exercise). In these analyses, each outcome will be predicted by the group (ImPuls + TAU vs. TAU), time (pre vs. post vs. follow-up), and their interaction. Given that the randomization is stratified by study site, we will account for the effects of study site in all analyses. All analyses will be performed on intention-to-treatment basis, using multiple imputations. Restricted maximum likelihood estimation will be used, which is implemented by R (Version 4) lme4 package. Each model will account for the three-level nested structure of the data: the within-person, between-person, and between-site levels. Random intercepts will be assumed to model the between-person and between-site, and random slopes (on time) the between-person variation. The significance level will be set to be alpha = 0·05.

To test Hypothesis 3b, mediation analyses will be performed to test the indirect effects of the treatment on the primary outcome that are mediated by the changes in the putative mediators (self-reported exercise in minutes per week, accelerometry-based moderate to vigorous physical activity (MVPA); from pre to post, pre to follow-up and post to follow-up). We will compute standardized change scores for the outcome and mediators.¹³⁵ In a path model (specified in the framework of structural equation modeling), the group variable predicts the changes in the mediators, and these mediating factors further predict the change score in the outcome. The indirect effects are defined by the products of the group-mediator and mediator-outcome effects. In case of saturated models, we will compare model fits with and without the indirect paths in terms of AIC and BIC. As exploratory analyses, we will also examine other forms of mediations proposed by Goldsmith et al., encompassing cross-lagged effects and latent change score models.¹³⁶ As recommended by Usami, Murayama and Hamaker, we will first test whether these models fit the data well (and which model fits the data best), and then investigate the indirect effects of the treatment on the outcome.¹³⁷

To further analyze clinical significant changes (additional analysis for Hypothesis 4 and 5), the Reliable Change Index (RCI) will be computed for each individual between pre-, post and follow-up assessments.^138-140 The reliable change index will be calculated each for Post and follow-up GSI scores on the non-imputed dataset. Jacobson-Truax formulation will be used, namely RCI = [GSI (post or follow-up) – GSI (pre)] / S_diff. S_diff was given by sqrt(2)*SD(GSI pre)*sqrt(1 – reliability). The cut-off for a clinically significant change was defined as 12 or smaller. The Mann-Whitney U test will be performed to examine whether the ordinal scores (recovered, improved, unchanged, and deteriorated) are significantly different between the two groups. Sensitivity analyses will be conducted with different analytic approaches. First, analyses on the primary and secondary outcomes will be repeated on the completer sample. Completers are operationalized as those who received at least two complete weeks within the supervised intervention period. Second, complete data analyses will be performed to check if multiple imputations influenced the results.

Concerning process evaluation, we use descriptive statistics to analyze objectives a), b), c), d), e) and f). Moderation analysis is planned to gain more insight concerning objective f) and g) (e. g., with multiple regression analysis) in terms of attendance, dropouts and efficacy. We will also use mediation analysis for objective h) (e.g., structural equation modeling).

Empowerment of exercise therapists (research question 1a) will be analyzed descriptively (mean, standard deviation) with respect to their self-reports on the modified training evaluation scale, the (occupational-) self-efficacy scale and concerning the frequency of supervision.^107,109

We descriptively present the adherence score (mean, standard deviation) as well as the corresponding percentage to check whether the desired high level of adherence (<90%) in the context of the efficacy trial is achieved (research question 1b).

We descriptively present the amount of all patients acquired by each strategy as well as the amount of all patients included in the study by each strategy. We then check which recruitment strategy has the highest inclusion rate and present the corresponding percentage (research question 1c).

Referring healthcare professional’ opinion on the new treatment option is presented descriptively (mean, standard deviation) (research question 1d). To provide information about the dose delivered (research question 1e), we present documentation data of exercise therapists and the web-based ImPuls interface descriptively.

To assess barriers and facilitators (research question 2a), we use quantitative data from questionnaires (barriers and facilitators assessment instrument, satisfaction scale, as well as basic recommendations for outpatient facilities in Germany.^{107,113,118-121} Results are presented descriptively (mean, standard deviation) to provide an overview over contextual characteristics. This data is further complemented with qualitative data from the interviews.

We evaluate the interviews in a deductive-inductive process following the steps of a contentstructuring qualitative content analysis.106 First, two researchers will elaborate a preliminary coding frame for coding based on the interview guideline. Subsequently, 15 % of all interviews are independently coded by those researchers. Inter-coder-reliability analysis is then performed to ensure that the elaborated coding frame is applicable. Potential discrepancies are discussed to refine the coding frame in an iterative process. This process is continued until both researchers agree that the categories are distinct and no new categories need to be added to the coding frame. Afterwards, the remaining interviews are coded. Coding and analysis are done by using the software MAXQDA 2022. In a final step, the statements from all interviews are summarized category by category and used to supplement the quantitative data.

We will conduct mediation and moderation analyses as well as subgroup analyses to gain deeper insight into the impact of the program (mechanisms of change). We check whether exercise therapist’ attitude towards mental disorders (Opening minds scale for health-care providers OMS-HC), evidence-based practice (evidence-based practice attitude scale – German version EBPAS-36D) and the program (measurement of therapy expectation and therapy evaluation of patients (PATHEV) affects treatment effects (research question 3a).^100,111,112 In addition, patient’ application data as well as changes in respective individual behavioral determinants e.g., action and coping plans; Physical activity-related health competencies (PAHCO) will be used to determine the extent to which core components of the intervention have been used and how this affects treatment effects (research question 3b).¹²⁸ Further analysis is done to determine the extent to which motivational/volitional core components of the ImPuls intervention (application data [barrier management, goal-setting], documentation data [phone contacts]) as well as changes in respective individual behavioral determinants (e.g., action and coping plans; [PAHCO]) affect patient’ exercise adherence (research question 3c). Finally, we want to explore whether psychological processes such as emotional intelligence (Trait emotional intelligence questionnaire (TEIQue), emotional regulation (difficulties in emotion regulation scale (DERS)), repetitive negative thinking (perseverative thinking questionnaire PTQ)) or perceived stress (perceived stress scale (PSS)) mediate the treatment effect on global symptom severity (research question 3d).^{124,125,127,129}

Missing data will be handled by multiple imputations.

Data Monitoring

Auditing

Data monitoring is ensured through a combination of automatic electronic validation through REDCap and manual checks, performed by the evaluating site (LMU). A complete audit trail including any data entry, access, modification, timepoint of randomization, user rights and exports is maintained by REDCapAfter data collection, pseudomized REDCap data and logs are exported, personal information (E-mail) is cleaned from the logs and logs are stored with the codelist on a secured drive. pseudomized data is stored for the codelist data retention period wiA site initiation procedure for each site was conducted by the LMU before data collection.

Ethics

Ethics approval has been obtained (ID: 888/2020B01, 02/11/2020).

Informed consent

Interested patients are invited to a first meeting in their respective study center. In this meeting a staff member of the central office from Tuebingen informs patients about the study. All questions are clarified. Afterwards patients sign informed consent. The informed consent files are stored in lockable filing cabinets separated from all other pseudonymized study data. Only selected staff members have access to these cabinets.

Confidentiality

In accordance with the Record of processing activities and data protection concept, participants personal information is collected by recruiting staff at Tuebingen, who input pseudonymized records and researcher-collected data into REDCap and maintain a code list. The only exception is email address, which is also input into REDCap for automatic survey invitation and reminder e-mails, but which is not accessible to LMU data analysis personnel. At the same time, survey responses are not accessible to recruiting staff. Email addresses will be deleted from the dataset after data collection is completed, and the code list will be deleted after 10 years. Final published datasets will be anonymized, i.e., published without pseudonym. Personal patient file data (informed consent, protocols of structural clinical interviews) in written form will be stored in-house of the university of Tübingen in securely lockable filing cabinets. Files will be shredded after 10 years. Other personal data (phone lists) will be anonymized after completion of data collection.

Group session videos for process evaluation are stored separately in order to be rated regarding manual adherence. Individual participants are not identified or rated and videos are stored safely at the conducting institution and will be deleted from the network drive after rating.

Personal information of the diagnostics is stored safely and pseudonymized in files at the Tue-CO.

Declaration of interest

There are no financial or competing interests to declare.

Access to data

The principal investigator (Sebastian Wolf) and the member of the evaluation team (Keisuke Takano, Eva Herzog, Thomas Ehring) will have access to the complete final data. In addition, all other investigators involved in the project will have access to the final data for pre-specified analyses.

Ancillary and Post-Trial Care

No provisions and post-trial care are provided.

Dissemination policy

A publication committee that consists of representatives of all partners develops a dissemination policy.

The study PI proposes potential papers. PhD students or other researcher propose a topic/paper and write an abstract to the committee. The committee finally decides about acceptance and authorships.