Stage 1: Identifying potential outcomes.

A comprehensive list of outcomes related to falls prevention for people with MS, PD and stroke was derived from multiple sources. Firstly, outcomes used in existing literature were extracted from an umbrella review of systematic reviews investigating the effectiveness of falls prevention interventions for people with MS, PD and stroke [11]. A systematic literature search using 15 electronic databases, grey literature searches and hand-screening of reference lists was undertaken to identify potentially relevant systematic reviews. Quantitative systematic reviews, mixed-methods systematic reviews (quantitative components only) or pooled analyses investigating the effectiveness of non-surgical or non-pharmacological falls prevention interventions were included in the umbrella review. Further details regarding the search strategy, inclusion and exclusion criteria, and data extraction are reported elsewhere [11, 25]. This umbrella review included 18 systematic reviews, representing 73 unique primary studies. The outcomes reported, their definitions, the outcome measures used and the time points for measurement were recorded for each systematic review. This list was complemented with additional outcomes that are not typically assessed in trials but were considered meaningful and important to people with MS, PD and stroke. These outcomes were identified through a qualitative study undertaken among people living with MS, PD and stroke [26]. A total of 20 individuals participated in online focus groups or semi-structured telephone interviews. Audio-recorded data were transcribed verbatim and analysed through the six phases of reflexive thematic analysis, using a predominantly inductive approach [27–29]. Outcomes that are desirable to participants were identified through this qualitative study and were combined with those reported in the systematic reviews included in the umbrella review. Any outcomes suggested by participants that were not assessed in the systematic reviews were added to the longlist of potential outcomes for the Delphi surveys. The list was further supplemented by discussions with the members of the public and patient involvement (PPI) panel. Finally, respondents to the subsequent Delphi survey had the opportunity to propose additional outcomes at the end of round one.

Stage 2: Development of the Delphi survey.

The lists of outcomes generated from each of the above sources were reviewed and discussed by members of the research team and the PPI panel to identify all distinct outcomes to be included in the Delphi survey rounds for consideration as a core outcome. The online software Qualtrics (Provo, UT) was used to design and administer the three Delphi survey rounds. All outcomes identified through the first stage of this study were presented in alphabetical order in the survey to avoid potential weighting [30]. The survey was developed with input from the PPI panel to promote ease of completion. Following its development, the survey was pilot tested on researchers (n = 2), patients (n = 2) and clinicians (n = 2) to assess for clarity and was modified accordingly prior to formal circulation to participants.

Delphi survey respondents were asked to score each outcome using the Grading of Recommendations Assessment, Development and Evaluations (GRADE) nine-point Likert scale, where a score of 1–3 signifies that an outcome is of limited importance, 4–6 indicates an outcome that is important but not critical, and 7–9 denotes a critically important outcome [15, 31]. This study used the ‘70/15%’ consensus definition, with consensus on inclusion defined as 70% or more of the participants rating it is critically important (7–9) and less than 15% rating it as being of limited importance (1–3), while consensus that an outcome should be excluded from the COS was defined as 70% or more of the participants rating is as being of limited importance (1–3) and less than 15% rating it as critically important (7–9) [15]. Any score distribution outside of those described above demonstrated a lack of consensus with respect to the inclusion or exclusion of the outcome in the COS [15].

Stage 3: Prioritisation of outcomes.

The consensus process involved a three-round online Delphi process. In round one, participants were asked to rate the importance of each outcome using the GRADE scale outlined above and were also encouraged to provide the rationale for each of their scores. At the end of the survey, participants had a free-text option to suggest additional outcomes for inclusion in the next round of the surveys. The outcomes suggested were reviewed by members of the research team to determine if they were unique outcomes [30]. All outcomes from round one were brought forward to round two to allow participants to reflect on the scores and feedback of each participant group before deciding whether or not they wished to change their score.

In round two of the Delphi process, participants were provided with their own score, the overall ratings for the overall group and for each stakeholder group, and a summary of the reasons that participants gave for their scoring of each outcome. Participants were asked to reflect on this information and then rescore all outcomes again using the GRADE scale. Following the completion of round two of the Delphi process, outcomes were separated into three categories: category A (outcomes reaching the criteria for consensus on inclusion), category B (outcomes not reaching consensus) or category C (outcomes reaching the criteria for consensus on exclusion). Outcomes in category A were added to the preliminary COS, outcomes in category B were added to the list of ‘supplementary outcomes’ and were considered for inclusion in a third round of the Delphi process, and outcomes in category C were excluded from any further consideration for the COS.

In line with the predefined criteria described in the study protocol, a third round of the Delphi surveys was completed for any outcomes in category B that were rated as critically important by more than 50% of participants and rated as having limited importance by less than 15% in round two. As per previous rounds, participants were provided with their own score, the overall ratings for the overall group and for each stakeholder group, and a summary of the reasons that participants gave for their scoring of each outcome before being asked to rescore the outcomes. Following round three, the scores for each outcome were analysed to determine if they now met the criteria for category A or if they were to remain in category B.

Following a research team meeting where the very high number of outcomes in category B after three rounds of the surveys was noted and discussed, it was decided by team members that the outcomes in category B should be further subdivided to allow for adequate time to discuss potential outcomes for the final COS during the consensus meeting. Any outcome on this list that >50% of participants had rated as critically important and <15% of participants had rated as being of limited importance were added to category B1 and were carried forward for further discussion at the meeting as they were nearing the predefined threshold for inclusion. Outcomes in category B that <50% of participants had rated as critically important or >15% had rated as being of limited importance were added to category B2 and were excluded from any further consideration for the final COS as it was decided that they had not been sufficiently prioritised by the overall group throughout the Delphi process to warrant inclusion. Outcomes in category B2 remained on the supplementary outcomes list from this point onwards.

Stage 4: Identification of outcome measures and definitions.

This stage of the project was centred around the selection of a single instrument for measuring each core outcome, to ensure consistency and comparability across all future trials. A pragmatic approach was taken to this process with outcome measures and definitions extracted from the umbrella review or else identified through targeted literature searches. Our research team and stakeholder group reviewed the available evidence and provisionally recommended the use of a single outcome measure after consideration and discussion of the following: the frequency with which the measure has been used and its responsiveness in existing studies, the time and resources required to use the measure, and the available data on their psychometric properties as outlined in the COnsensus-based Standards for the selection of health status Measurement INstruments (COSMIN) recommendations (validity, reliability, responsiveness and interpretability) [32, 33].

Stage 5: Consensus meeting.

A virtual face-to-face consensus meeting was held using the Zoom platform on the 5^th of January 2023. This online interactive meeting took place with representatives from each stakeholder group to discuss, vote and gain consensus on the final COS, in addition to the definitions and measures to be used to assess these outcomes. At the end of second Delphi survey, all participants were asked to select if they would like to take part in the consensus meeting. A poll was sent around to those who responded yes to determine their availability to attend at different dates and times. The meeting took place on the date that suited the majority of participants.

The meeting commenced with a presentation describing the preliminary COS and the category B1 outcomes that were near the inclusion criteria. The ratings from the overall group and from each stakeholder group were presented for each outcome on these lists. This was followed by a discussion between attendees regarding the reasons for including or excluding each outcome from the COS and a final vote. The definition for consensus was a minimum of 70% of participants and at least one patient participant voting for the outcome to be included in the COS [34]. Any outcomes not meeting these criteria were added to the supplementary outcomes list accompanying this COS. For each outcome achieving consensus for inclusion in the final COS, recommendations regarding outcome measures and definitions were discussed. Attendees were encouraged to provide feedback on the recommendations and/or to suggest alternatives for discussion. Following this, a final vote on the selection of a single outcome measure and definition, where applicable, took place. Once again, the definition for consensus was a minimum of 70% of participants and at least one patient participant voting for the outcome measure/definition to be recommended alongside the COS.

Recommendation.

Fall incidence should be assessed in all future mixed-diagnosis falls prevention interventions for people with MS, PD and stroke. This outcome should be reported as falls per person per year. Fall events should be prospectively recorded daily with follow-up by team members at least monthly to reduce recall bias. A fall event should be defined as ‘an unexpected event in which the participant comes to rest on the ground, floor, or lower level’ [18].

Key discussion points.

Falls rate and total number of falls were prioritised as core outcomes in the Delphi survey rounds. Due to the degree of overlap between these outcomes and subsequent outcome measures, a recommendation was made to the attendees at the consensus meeting to combine these outcomes into one outcome, fall incidence. No counterarguments were made to this proposal and all attendees voted for fall incidence to be included in the COS. Our research team proposed the use of falls per person per year to report this outcome as it allows for comparison of studies with different reporting periods [37]. Further, given the issue of under-reporting of fall events associated with the retrospective recording of falls [38, 39], we suggested that falls be prospectively recorded daily with follow-up by team members at least monthly to reduce recall bias [37]. Potential challenges with the monthly follow-up by the research team, particularly with respect to large intervention studies with high numbers of participants were noted. A counter argument proposed that follow-up with participants affords the research team the opportunity to clarify what happened and determine whether it meets the definition for a fall, thereby improving the accuracy of findings.

The second part of the discussion focused on the definition of a fall. From our review of the literature, we identified that the most commonly used definition is ‘an unexpected event in which the participant comes to rest on the ground, floor or lower level’ [18]. However, from our qualitative research we identified discrepancies between that definition and what patients consider a fall event, namely that patients consider a fall event a ‘loss of balance’ and believe that instances where an external support or individual prevents them from ending up on the ground should be considered a fall event. In light of these findings, our research team proposed making amendments to the definition of a fall that were reflective of these views. Following an extensive discussion, consensus from attendees was that the original definition should be included with the COS as extra elements opens the definition to interpretation by participants and so could compromise the accuracy of the outcomes. Participants stated that instances where an individual has a voluntary reaction to reach out and grab a support demonstrates a positive reactive falls prevention strategy and so should not be considered a fall event. It was also suggested that amendments could lead to the overreporting of fall events in circumstances where an overprotective person caught an individual to prevent a fall despite the individual having the balance reactions to prevent the fall themselves. Finally, participants felt that the amendments to the definition would describe near-fall events rather than actual fall events.

Recommendation.

Injurious fall incidence should be assessed in all future mixed-diagnosis falls prevention interventions for people with MS, PD and stroke. This outcome should be reported as injurious falls per person per year. Injurious falls should be prospectively recorded daily with follow-up by team members at least monthly to reduce recall bias. The standardised system for categorising and defining fall related injuries proposed by Schwenk et al. (2012) should be used (Table 3) [40]. Each category should be reported even if a study is not powered to detect effects as this will enable future meta-analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Standardised system for categorising and defining fall-related injuries.

https://doi.org/10.1371/journal.pone.0294193.t003

Key discussion points.

Due to degree of overlap between the outcomes and outcomes measures, a recommendation was made to combine number of injurious falls, number of falls resulting in healthcare utilisation and number of fall-related fractures into one outcome, injurious fall incidence. All attendees agreed that injurious fall incidence should be included in the final COS.

A systematic review undertaken by Schwenk et al. (2012) highlighted the substantial variation in the definitions and methods used to measure and document injurious falls [40]. Following this, they developed a standardised system for categorising and defining fall-related injuries [40]. Our research team proposed the use of this standardised system to measure injurious fall incidence as it captures the data of the three outcomes that were combined to generate this outcome. Some concerns over the accuracy of measures to assess injurious falls were expressed, with participants describing whether or not someone will seek medical attention for an injurious fall can be very subjective and system-dependent. A patient participant also described how if an individual who does not typically fall suddenly experiences a fall event, they may seek medical attention not necessarily because they are injured but out of concern that their condition is progressing or that they need to change treatment. Following a discussion, participants ultimately agreed that the use of a standardised system was necessary to compare fall-related injuries and voted to include this measure.

Recommendation.

Quality of life should be assessed in all future mixed-diagnosis falls prevention interventions for people with MS, PD and stroke. This outcome should be measured using the 5-Level EuroQoL-5 (EQ-5D-5L) [41].

Key discussion points.

During the discussion on whether to include quality of life in the COS, it was suggested that only falls outcomes should be included in the final COS since they align with the primary aim of the intervention, the prevention of falls. In response, other participants stated that the prevention of falls should not be the sole aim of an intervention. They described how fall events could be reduced by preventing an individual from completing all activities but did not believe that this should be considered a successful intervention. Rather, they stated that the goal of the intervention should be to improve quality of life by enabling individuals to complete their desired activities in a safer manner. It was also suggested that, given the importance placed upon quality of life by patient participants, individuals would not adhere to an intervention that was not improving this outcome. Following this discussion, all attendees voted to include quality of life in the final COS.

Our research team proposed the use of the EQ-5D-5L to measure quality of life as it has been shown to be valid and reliable across multiple conditions and settings, it is available in more than 130 languages, can be used in cost-effectiveness analysis and requires minimal time to complete [42, 43]. One clinician stated that the measure can be slightly more time consuming in practice to interpret correctly if you are not familiar with the measure. No other arguments for or against its use were made and during the final vote the EQ-5D-5L reached consensus for inclusion.

Recommendation.

Falls self-efficacy should be assessed in all future mixed-diagnosis falls prevention interventions for people with MS, PD and stroke. This outcome should be measured using the Falls Efficacy Scale-International (FES-I) [44].

Key discussion points.

While discussing whether to include falls self-efficacy in the final COS, one participant suggested combining falls self-efficacy, balance confidence and fear of falling into one outcome or else only including one of them in the COS. Other attendees disagreed with this proposal, asserting that the three outcomes are different constructs and as such should be kept separate. Following the final vote, falls self-efficacy was included in the COS.

Our research team proposed the use of the FES-I to assess falls self-efficacy as it has shown to be valid and reliable across multiple conditions, has been translated and tested in multiple languages, is available for free, and requires no equipment and minimal time [45–50]. Participants agreed that this is the best available measure for assessing falls self-efficacy and voted to include it.

Recommendation.

Fear of falling should be assessed in all future mixed-diagnosis falls prevention interventions for people with MS, PD and stroke. This outcome should be assessed using the direct question ‘Are you afraid of falling?’ and five-point Likert scale: not at all, slightly, moderately, very or extremely.

Key discussion points.

No counterarguments were made to the inclusion of fear of falling and during the final vote it reached consensus for inclusion in the COS. In our umbrella review, FES-I was the most commonly reported measure for assessing fear of falling, however, evidence has shown that falls self-efficacy and fear of falling are two distinct constructs [7, 51]. In the absence of a validated outcome measure, our research team proposed the use of a direct question and five-point Likert scale to measure fear of falling. Participants described benefits of the Likert scale including that it requires minimal time and is easy to administer. One clinician proposed changing the wording of the question from ‘are you afraid of falling?’ to ‘are you concerned about falling?’ as the terminology of the question may feed into the fear. However, a counterproposal was put forward that if we are measuring fear directly that it should be in the question to ensure we are capturing it accurately and not crossing over into other constructs such as falls self-efficacy. Following the vote, the direct question and five-point Likert scale reached consensus for inclusion.

Recommendation.

Activity curtailment due to fear of falling should be assessed in all future mixed-diagnosis falls prevention interventions for people with MS, PD and stroke. This outcome should be assessed using the direct question ‘Do you think that fear of falling has made you cut down on any activities that you used to do?’ with the following response options: yes, no, don’t know or refused.

Key discussion points.

All participants voted to include activity curtailment due to fear of falling in the COS. Our research team proposed the use of the Fear of Falling Avoidance Behaviour Questionnaire (FFABQ) as it has been validated for older adults with PD and stroke, no equipment is required, takes minimal time to complete and has high clinical feasibility [52]. A participant suggested the use of a single-item direct question for activity curtailment that is used in falls research for older adults: ‘Do you think that fear of falling has made you cut down on any activities that you used to do?’ [53, 54]. Participants were in favour of this single-item question and voted to include it in the final COS.

Recommendation.

Cost-effectiveness should be assessed in all future mixed-diagnosis falls prevention interventions for people with MS, PD and stroke. Researchers should consult with a health economist to determine the most appropriate way to calculate cost-effectiveness of an intervention.

Key discussion points.

Cost-effectiveness was among the category B1 outcomes that had not met the criteria for inclusion in the preliminary COS but were brought forward for consideration at the consensus meeting as they were nearing the consensus threshold (i.e., >50% voted as being critically important and <15% voted as being of limited importance). When discussion on these category B1 outcomes was invited, participants proposed the inclusion of cost-effectiveness in the final COS despite not reaching consensus for priority in the survey rounds. The group recognised the challenges of doing a cost-effectiveness analysis, especially in terms of international standardisation, but felt it was important to include the outcome in the COS to optimise translation of interventions to practice and to promote adoption for funding by healthcare authorities. Participants also suggested that where similar interventions demonstrated similar effects, a cost-effectiveness analysis could be the factor to determine which to implement. Consequently, participants voted to include cost-effectiveness in the final COS. No recommendation regarding what measure to use was made by the group, instead, they recommended consulting a health economist to inform the decision.

Recommendation.

The anticipated mechanism of impact of an intervention should be considered and articulated when additional outcomes are assessed and reported.

Key discussion points.

While they had been prioritised for inclusion in the COS during the Delphi survey rounds, the remaining five outcomes on the preliminary COS (dynamic balance, balance confidence, objectively assessed mobility, understanding of personal falls risk factors, and falls self-management skills) did not reach consensus for inclusion in the final COS. Following an interactive discussion between attendees, it was agreed that while these outcomes are important, their selection should be dependent on a number of factors such as the proposed mechanism of impact of the intervention, the aim of the intervention, and participant characteristics (e.g., stage of disease or level of mobility). Therefore, participants concluded that as they are only relevant to specific interventions, e.g., dynamic balance is only applicable to interventions aiming to reduce falls through improving balance, they should not be included as a core outcome for all falls prevention interventions.

Recommendation.

Ability to independently perform activities of daily living, falls risk, level of physical activity, lower limb strength, number of fallers, number of falls resulting in a long lie, number of recurrent fallers, objectively assessed ability to perform activities of daily living, self-efficacy, static balance, and time to first post-intervention fall should remain on the supplementary outcomes list.

Key discussion points.

Those in attendance at the consensus meeting agreed that none of the remaining outcomes in category B1 (ability to independently perform activities of daily living, falls risk, level of physical activity, lower limb strength, number of fallers, number of falls resulting in a long lie, number of recurrent fallers, objectively assessed ability to perform activities of daily living, self-efficacy, static balance, and time to first post-intervention fall) were important enough to be included in the final COS. Consequently, these 11 outcomes remained on the supplementary outcomes list.