A systematic review has demonstrated that resistance training effectively improves muscle strength in adults over 50 with sarcopenia, a condition closely linked to reduced autonomy in activities of daily living. ²⁴ Furthermore, finding from the ongoing SHADE LSR indicate that, among all types of physical exercise, resistance training is currently one of the most effective interventions for preventing HAD in older adults, consistently yielding positive results across all analyses. ²⁵

Several studies included in the SHADE living systematic review based their resistance training interventions on existing multicomponent exercise programmes, such as the Vivifrail physical exercise programme, ²⁶ High Intensity Functional Exercise (HIFE) Training, ²⁷ and the Otago exercise programme. ^{26– 28} These programmes served as a foundation for selecting specific exercises. Researchers then used these selections to develop their own tailored training programmes. ^{29– 31} Some studies incorporated power training, also referred to as high-velocity training, which involves performing the concentric phase of resistance exercises at a higher speed. ^{29, 31, 32} This type of training has been shown to be more effective than moderate-velocity resistance exercises in improving the physical function of healthy older adults. ^{33– 35} However, there is no evidence that high-velocity training has an impact on ADL for hospitalised older people. In the studies included in the LSR, as of October 2025, patients were typically hospitalised for a short duration, with a mean length of stay of 8.05 days at discharge. Observing meaningful effects of resistance training within such a limited timeframe is challenging, particularly when advanced modalities such as velocity-based training are considered.

In addition, the studies did not describe how to instruct patients in these modalities. They also lacked guidance on which patient groups might benefit most from the intervention. To describe resistance training modalities precisely, it has been recommended to report the exercise selection, training frequency and length, intensity, number of repetitions and set endpoint, number of exercises, progression, the fractional and temporal distribution of the contraction modes per repetition, time under tension and its range of motion. ^{36, 37} None of the included studies in the LSR contain all these parameters.

Multicomponent exercise programmes are well established in geriatric care. Such programmes have been shown to have a positive influence on strength and balance, thereby reducing fear of falling. This fear is a factor that correlates with both the risk of falls and limitations in ADL. ³⁸ Furthermore, these multicomponent exercise programmes have also been shown to reduce HAD and adverse events associated with hospitalisation. ²³

The ongoing SHADE LSR, which included 84 studies in October 2025, highlights that balance, resistance, and gait training are the key components within multicomponent programmes that most effectively prevent hospital-associated disability (HAD) in older adults during acute hospitalisation. ²⁵

The multicomponent exercise programmes included in the SHADE LSR comprised balance, resistance, and gait training, and showed variation in design and implementation. Balance exercises typically included standing balance training on stable or unstable surfaces, as well as seated balance exercises tailored for patients with reduced physical capacity. ^{30, 39, 40} Resistance training was performed as mentioned for resistance training above. Gait training mostly involved walking, either with or without supervision or walking aid, in the hospital corridors. ^{29, 40, 41} Other walking exercises described in the LSR included stepping practice and navigating small obstacles. ^{32, 42} The multicomponent exercise programmes reviewed did not specify how the proportion of each component was determined. Nor did they explain how the exercises and their load was adapted for different patient types or which safety measures were implemented.

For our research, we will use a modified Delphi study design. ⁴³ This project comprises two distinct Delphi studies, each dedicated to one exercise programme: the first focusing on resistance training and the second on a multicomponent exercise programme combining balance, resistance and gait training. This approach aims to achieve consensus on two exercise programmes designed to prevent HAD in patients within an acute hospital setting. This Delphi design is recognised as valuable for evidence-based practice, as it facilitates the incorporation of existing knowledge into practical guidelines. ⁴⁴ Our methodology is modified from the typical Delphi study in that not all questions in the first round will be open-ended. This will enable the use of the knowledge acquired in the LSR. ¹

Participants for both Delphi studies (i.e. resistance training only and multicomponent programme with a combination of balance, resistance and gait training) will be recruited using a non-probability purposive sampling strategy.

We consider the sample of participants to be homogeneous, since they will all be clinicians or researchers working with older people. A sample size of 10 to 15 participants is considered adequate for a homogeneous group. ⁴⁵ We plan to include 30 participants per type of training programme (i.e. resistance training and a multicomponent exercise programme combining balance, resistance and gait training). A sample size of approximately 30 participants per group should ensure the generalisability of the results, even with a dropout rate of up to 20%. ^{43, 46}

The participants on the expert panel will be clinicians, mostly, from various regions of Switzerland, including physiotherapists and sports scientists, who actively instruct physical exercise to older people. Snowball recruitment will be conducted. To recruit physiotherapists and sport scientists working in hospitals and rehabilitation centres, we will contact the heads of the medical and therapeutic services at hospitals and geriatric rehabilitation centres in Switzerland, asking them to forward our email to their employees. To recruit physiotherapists in private practice, we will ask the two Swiss physiotherapy associations, Physioswiss and the Swiss Association of Independent Physiotherapists (ASPI), to circulate an email to their members. We will also contact different physiotherapy rehabilitation community of interest, such as plateforme-Reha and Interessengemeinschaft Physiotherapie Rehabilitation Akutspital (IGPTR A) to find potential experts in different fields. We will proceed in the same way with the Swiss Associations of Professionals in Adapted Sports Activities (ASP-APA and Sportwissenschaft.ch). To ensure that we contact a wide range of people working in these fields, we will also contact Swiss associations active in Health Sport and Sports Therapy, such as Rheumaliga, Le Réseau Santé & Sport (LRSS), Allianz Bewegung, Sport und Gesundheit, Pro Senectute, Public Health Schweiz, Swissheart and Swisscardiotherapists. To approach Swiss researchers in this field, we will reach out to alumni and collaborators from universities working in this area and identify researchers through their published articles. Across all categories, we will also leverage our professional network, combining direct personal outreach with targeted promotion on LinkedIn. The recruitment will take place between October and November 2025.

These two modified Delphi studies will be open to clinicians who instruct physical exercise to older people (over 65 years old) in facilities such as hospitals, rehabilitation centres, and private practices such as physiotherapists and sports scientists. Participants must have worked with this population for at least five years. ⁴⁷ Researchers specialising in geriatrics and physical exercise with at least three publications in this field will also be eligible. ⁴⁷ As the questions will be asked in English, participants should be able to understand basic written English to participate.

Participants will be included in the study only after providing written consent immediately before Round 1 survey. They must also agree to the publication of their anonymised responses.

In each modified Delphi study (i.e. resistance training only and multicomponent exercise programme with a combination of balance, resistance and gait training), we will conduct three rounds, as conducting between two and three rounds is recommended to avoid participant fatigue and ensure the relevance of the results. ⁴³ To minimise attrition, we will keep the interval between the closure of one round and the start of the next as short as possible. ⁴³ Two participation reminders will be sent during each round: one shortly before the initial deadline and another immediately after, granting participants an additional week to respond. The investigator team will use the browser-based research electronic data capture (REDCap) software ⁴⁸ to develop the questionnaires for the three rounds, focusing on developing questions designed to achieve the objectives of each round. This software will also be used to gather and store the data. The questionnaires will be the same for all participants of one Delphi study. The study will start in November 2025 and should end before July 2026.

5.6.1. Round 1

The primary aim of the first modified Delphi round is to gather participants’ perspectives on the appropriateness of the exercise interventions identified in the ongoing LSR. ¹ A secondary objective is to assess the levels of patient functional ability for which these interventions are suitable and to propose progression strategies and safety measures for each level. Additionally, this round will explore alternative exercise options and seek consensus on feasible implementation modalities. Finally, it will identify facilitators and barriers to implementing such training in a hospital setting from both clinician and patient reported perspectives.

Participants will first be provided with an overview of the scientific evidence gathered in the LSR regarding physical exercise interventions for preventing hospital-associated disability (HAD) in hospitalised older adults. Following this, their demographic and professional data will be collected. Next, participants will evaluate the exercises identified in the studies included in our LSR for their suitability in a Swiss hospital setting for older adults with varying levels of functional ability. For these exercises, participants will indicate how they would adapt them for patients with different ability levels and specify the safety parameters to be included. They will also be invited to propose additional exercises appropriate for the respective training concept (i.e., resistance or multicomponent programme). Each exercise will be described in terms of its name, starting position, movement pattern, and instructions. During this first round, participants will also define the appropriate load for each ability level in hospitalised patients aged over 65 years. Similarly, expert consensus will be sought on training frequency, duration, exercise volume, contraction timing and proportion, as well as exercise type to determine their appropriateness for different patient profiles. Finally, clinicians and researchers will identify facilitators and barriers to implementing such exercise programmes for the prevention of HAD in older adults within a hospital setting.

Due to the wide range of patients’ needs and capacities, we will use a predetermined classification system for patients based on the three-level dependency scale of the commonly used Barthel Index. This scale categorises patients as “severely dependent patient”, “moderately dependent patient” or “slightly dependent or independent patient”. ^{49, 50} To facilitate responses, these different ability levels categories will be defined descriptively rather than by specific point thresholds on the Barthel Index: -

Severely dependent: patient is unable to perform any or most ADLs without substantial help

In the first round of the present modified Delphi study, exercises consensus questions will be presented using Likert items with a six-point ordinal scale to minimise questionnaire length and reduce the risk of participants dropping out. Additionally, open-ended questions will be included to clarify sources of disagreement, identify missing exercises and explore possibilities regarding exercise modalities. Based on the results obtained, we will: i) compile a pool of potentially relevant exercises, ii) define a set of modalities for a preliminary exercise programme and iii) compile a pool of potential facilitators and barriers to implementing such exercise programmes for the prevention of HAD in older adults in a hospital setting.

5.6.2. Round 2

The second rounds of the modified Delphi study will aim to collect the participants’ opinion on the developed physical exercise programmes. To this end, we will present the results of the first round to the participants. For the consensus questions, the results will be summarised using the median and interquartile range (IQR) of the participants’ ratings. The key themes and concepts identified through the open-ended responses will be organised systematically and presented in a structured manner.

Participants will be asked to indicate their level of agreement with the proposed exercises, their progression and safety measures using Likert items with a six-point ordinal scale. Participants will then be asked to do the same with the modality descriptions and the facilitators and barriers to implementing such exercise programmes for the prevention of HAD in older adults in a hospital setting. In case of disagreement, volunteering clinicians will be invited to suggest modifications or propose alternative approaches. The same question types as in round 1 (i.e. agreement questions using Likert items with a six-point ordinal scale and open text questions will be used to clarify disagreement).

5.6.3. Round 3

The aim of the third modified Delphi study round is to gather participants’ opinions on the revised physical exercise programme, modalities, facilitators and barriers to implementing such exercise programmes for the prevention of HAD in older adults in a hospital setting. The results of the consensus questions from round two will therefore be presented with summary statistics such as medians and IQR to the participants. Any elements that fail to reach consensus will be adapted and re-evaluated using the same procedure as in round two.

The electronic data collection system REDCap ⁴⁸ will be used to collect and manage study’s data. REDCap surveys will first be tested by team members and colleagues not involved in developing the questions. This will provide feedback on completion time and comprehensibility. The surveys will then be adapted and sent to the participants via email. Participants’ identities will be coded, and access to these data will be restricted to the three main investigators and two bachelor students. All participants’ contact information will be permanently deleted upon completion of the study.

The data from the modified Delphi studies will be reported in accordance with the reporting guidelines for Delphi techniques in health sciences, as proposed by Spranger et al. ⁵¹

We will use Likert items with a six-point ordinal scale, as this lies between four and seven categories that have been considered as optimal. ⁴³ Given that participants are expected to constitute a homogeneous group with sufficient expertise to address all questions, a neutral “don’t know” response option will not be included. This approach is intended to encourage participants to take a definitive position on each item, thereby supporting the consensus-building process. ⁵²

5.8.1. Quantitative analysis of the modified Delphi study rounds

The characteristics of the participants, including their work setting, primary training, years of experience, age, and gender, will be presented in a table using mean values and their standard deviation. The number of participants in each round of the modified Delphi studies will be presented using a participant flow diagram. This diagram will visually represent participant retention and attrition across the different modified Delphi rounds.

5.8.1.1. Level of agreement

In this study, consensus will be defined as the level of agreement, which will be measured as follows: Likert items with a six-point ordinal scale will be used to assess participants’ agreement or disagreement with each question in the questionnaire. Participants will be asked to indicate their level of agreement on a scale from 1 to 6 (1 = Very strongly agree, 2 = Strongly agree, 3 = Agree, 4 = Disagree, 5 = Strongly disagree, and 6 = Very strongly disagree). During the analysis phase, the Likert items with a six-point ordinal scale will be dichotomised as suggested by Limotai et al. (2020) as follows: i)

responses 1 to 3 will be considered positive

Our a priori criterion will be as follows: if more or equal to 80% of participants agree with a statement, we will assume that they are in favour of it, and it will be included in the final version of the training programme. Statements (e.g. type of exercise, duration, frequency, safety measures) will be included if approved by 80% of participants (Likert items: 1–3) and excluded if disapproved (Likert items: 4–6) by 50% or more of participants after each round of the modified Delphi studies. ⁵³ All intermediate values will be included in the next round. We will also describe the distribution by providing the median and IQR for each task.

Items for which no agreement was reached by the end of the three rounds will be presented separately, using the results from the third round.

5.8.1.2. Stability of response/internal reliability

The stability of responses (i.e. referring to the level of agreement between rounds in the modified Delphi study and serving as a measure of reliability) will be reported using the median and interquartile range (IQR) across rounds. The stability of results will be determined using a Wilcoxon matched-pairs signed-rank test. ⁴³

5.8.2. Qualitative analysis of the different modified Delphi study rounds

For the open-ended questions, we will primarily apply an adapted version of the argument-based qualitative analysis strategy (AQUA), complemented by qualitative content analysis for those sections where AQUA cannot be applied due to the nature of the questions. ⁵⁴ First, all responses will be read thoroughly to ensure familiarity with the content. Next, each response will be segmented into distinct arguments, translated into British English if originally written in German or French, and harmonised for language style. ⁵⁴ Similar suggestions will then be grouped into categories using both inductive and deductive approaches, as outlined in qualitative content analysis, and ranked by frequency. ⁵⁵ The arguments within each category will be condensed and summarised to identify core statements and establish links between similar ideas. ⁵⁴ Two investigators will carry out this step separately and their results will be compared. If there is persistent conflict, a third person will intervene and make the final decision. The research group will then meet to decide which elements to include or exclude in the next round.