Interventions to improve outdoor mobility among people living with disabilities: A systematic review

The review in brief

The evidence is very uncertain about most benefits and harms of intervention options aimed to improve outdoor mobility for adults with disabilities, both in the short and long term, and it is currently unclear if any intervention options to improve outdoor mobility are effective and safe.

What is this review about?

Around 15% of the global population live with disabilities and participate in fewer activities outside the home and in the community, and commonly also experience a reduced quality of life. Interventions in this review target outdoor mobility aiming to improve the ability to move, travel and manage orientation outside the home and in the community. These interventions can influence the number of activities outside the home, participation and quality of life but may also lead to harm like falls, injuries and hospitalization.

What is the aim of this review?

The review examines the benefits and harms of interventions aimed to improve outdoor mobility for adults living with different disabilities.

What are the main findings of this review?

What studies are included?

A total of 12,894 studies were reviewed. Twenty-two studies that evaluated interventions aiming to improve outdoor mobility compared to interventions that did not aim to improve outdoor mobility were identified. Twelve ongoing studies with upcoming results were also located. Sixteen studies examined interventions where skills aimed to be used in the outdoor environment were trained. Five studies examined physical training aiming to improve the capacity to move around in outdoor environments, such as treadmill walking. No studies were identified for other types of interventions, such as educational or cognitive behavioural interventions. Most interventions were delivered to people after stroke (13 studies) and older adults living with disabilities (five studies) and were delivered mainly in high income countries.

What are the benefits and harms of skill training?

The evidence is uncertain about the benefits and harms of skill training for all people living with disabilities, both in the shorter and longer term for the possibility to be active outside the home, participation, quality of life and for minor and major harms.

Skill training may improve engagement in everyday life activities in people living with disabilities in the shorter term, but we are uncertain in the longer term. Skill training may improve both activity outside the home and health-related quality of life for people living with cognitive impairments in the shorter term, but we are uncertain in the longer term.

What are the benefits and harms of physical training?

The evidence is uncertain about the benefits and harms of physical training for all people living with disabilities, both in the shorter and longer term for the possibility to be engaged in everyday life activities, participation, quality of life and for minor and major harms. Physical training may improve activity outside the home in people living with disabilities in the shorter term, but we are uncertain in the longer term.

What are the benefits and harms of different treatment options aimed to improve outdoor mobility?

The evidence is uncertain about any difference for benefits and harms of different outdoor mobility treatment options.

What do the findings of this review mean?

Despite identifying 22 studies, the evidence still remains uncertain about most benefits and harms of treatments to improve outdoor mobility, both in the short- and long term. Furthermore, the reported effects of these interventions were considered either small or trivial, and could be of limited relevance to people living with disabilities. The evidence is currently uncertain if there are interventions that can improve outdoor mobility for people living with disabilities and can improve other important outcomes, while avoiding harms.

Future studies that are using more rigorous study designs and include important measurements of improvement for people living with disabilities are needed to reduce this uncertainty to guide policy decisions, treatment decisions of healthcare providers and carers, and decisions to engage in interventions to improve outdoor mobility for people living with disabilities.

How up-to-date is this review?

The evidence is up to date covering publications until January 2023.

Types of studies

Eligible study designs included in the review are RCTs where participants have been randomly allocated to an intervention or control group. To minimize the risk of bias of the effect estimates we excluded cluster randomized trials, cross-over trials and quasi-randomized trials in this review. We also excluded non-randomized studies, observational studies and case reports.

The decision to restrict the study design to only randomized controlled studies of interventions was done because it minimizes the risk of any bias influencing the effect estimates of these interventions. Randomization is the only approach to prevent systematic differences in  baseline characteristics of participants within the intervention and control group in terms of both known, unknown or unmeasured confounders (Cochrane Handbook). We acknowledged the trade-off between our restrictive criteria for study designs—which could lead to including fewer studies within the review—compared with a broader criterion including more studies but subsequently with a higher risk of bias.

Types of participants

Participants eligible for inclusion were all adults, aged 18 or above, living with one or multiple disabilities. All disabilities, including physical, mental, cognitive or sensory impairments, or activity limitations and participation restrictions, were included. The underlying disease, disorder or cause of disability, as described in the included studies, were used to describe the study populations. Additional measures and classifications were used, if available, to further describe each study population.

For example, we planned to include study populations living with:

Diseases of the musculoskeletal system or connective tissue, like osteoarthritis and fractures.

We included participants in all types of settings and did not limit the inclusion to any specific demographics.

We excluded studies where more than 50% of the population were below the age of 18. We also excluded studies where more than 50% of the population did not have disabilities, defined as impairments in a person's body structure or mental functioning, activity limitations or participation restrictions.

Types of interventions

The review included interventions that aimed to improve outdoor mobility for people living with disabilities. It includes, for example, educational interventions, physical training interventions, cognitive or behavioural training interventions or skill training interventions. Interventions could be delivered on their own or in packages of several interventions by a single health care provider or a multidisciplinary team.

The included interventions could be delivered in the community, for example, within healthcare, the workplace, within the home, digitally or by groups in the community. The interventions could be delivered directly related to an injury, but also several years after having received a diagnosis, commonly for people living with a progressive disorder. Interventions aimed solely aimed at health promotion and/or primary prevention were excluded. For example, interventions aiming to prevent a future decline in outdoor mobility for older adults who currently do not have limited mobility outdoors were excluded. The reason was that such health promotive or preventative interventions are substantially different from those aiming to improve already limited mobility, and the benefits and harms would be different, particularly regarding falls. In addition, such interventions could be delivered to populations without disabilities and limited outdoor mobility, which is different from the population targeted in this review.

No restrictions were placed on duration, intensity and way of delivery of the intervention, as it can vary due to different needs of the participants and resources for delivery in the trial context. We extracted and described the duration, intensity and frequency for all included interventions.

The focus of the intervention to improve outdoor mobility had to be clearly described as an aim or purpose of the intervention to be included. Such descriptions were extracted for all included interventions. Interventions without an aim or purpose to improve outdoor mobility were excluded. For example, interventions aiming to only improve physical activity levels or walking speed for the participants. In addition, interventions with a focus on only improving mobility indoors or personal ADL were excluded.

Environmental components could be included as a part of an intervention only if they were part of multi-component interventions that include components aimed at the capacity or performance of the participants as well. Such interventions could include home modifications in combination with skill training aiming to increase the performance to travel outside, or skill training to use buses.

Interventions that solely target alterations in the environment were excluded from this review, because such interventions have fundamentally different mechanisms of action and delivery. Further, alterations in the environment often serve as a mediator for other interventions and it may also be difficult to use randomized controlled trials since blinding may be problematic. This includes interventions targeting infrastructure, changes in organizations, routines or policies, implementation of ramps or lifts, implementation of digital support systems, only providing mobility aids, adjustments in the home or increased access to training sites or parks.

We included two comparisons in the review: 1.

Outdoor mobility intervention compared with a control intervention not expected to improve outdoor mobility. The control interventions included:

A usual care-control group as stated by the trial authors.

Important outcomes

We included any of the six following important outcomes in the review:

1. Activity outside the home

Activity outside the home (defined according to ICF as ‘the execution of a task or action by an individual’) could be measured in continuous outcome measures, for example:

Number of times going outside the home

The outcome measures were extracted and categorized according to at what time after starting the intervention they were collected, defined as short-term (≤6 months) or long-term (≥7 months).

2. Engagement in everyday life activities

Engagement in everyday life activities (defined according to Kielhofner and colleagues (Taylor, 2017) as ‘engagement in work, play, or activities of daily living that are part of one's socio-cultural context and that are desired and/or necessary to one's well-being’) can be measured by dichotomous outcome measures, for example, where study participants answered ‘yes’ or ‘no’ to questions such as:

‘Can you do what you want to do in your life?’

The outcome measures were extracted and categorized according to at what time after starting the intervention they were collected, defined as short-term (≤6 months) or long-term (≥7 months).

3. Participation

Participation (defined according to ICF as ‘involvement in a life situation’) could be measured in continuous outcome measures, for example:

Impact of Participation and Autonomy Questionnaire (IPA) (Cardol, 2001)

The outcome measures were extracted and categorized according to at what time after starting the intervention they were collected, defined as short-term (≤6 months) or long-term (≥7 months).

4. Health-related Quality of Life

Health-related quality of life (defined according to WHO, 2012 as ‘individuals’ perceptions of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards and concerns’) could be measured in continuous outcome measures, for example:

SF36 (Ware, 2007)

The outcome measures were extracted and categorized according to at what time after starting the intervention they were collected, defined as short-term (≤6 months) or long-term (≥7 months).

5. Major adverse events

Major adverse events could be based on dichotomous or continuous outcome measures, and included:

Mortality

The outcome measures were extracted and categorized according to at what time after starting the intervention they were collected, defined as short-term (≤6 months) or long-term (≥7 months).

6. Minor adverse events

Minor adverse events could be based on dichotomous or continuous outcome measures, for example:

Falls

The outcome measures were extracted and categorized according to at what time after starting the intervention they were collected, defined as short-term (≤6 months) or long-term (≥7 months).

Falls leading to hospitalization or injuries requiring medical attention was be classified as Major Adverse Events. Other falls were subsequently be classified as a Minor Adverse event, which has been estimated to be approximately 98.8% of all falls in a hospital setting (Kobayashi, 2017).

Duration of follow-up

We included all duration of follow-up, but they were separated into two different categories, that is, short-term (6 months or less) and long-term (7 months or more) after starting the intervention.

Types of settings

We included interventions delivered in any setting, for example, in hospitals, clinics, the workplace, within the home, digitally or by groups or organizations in the community. We expected that the different settings could have varying impacts on the estimated effects of the intervention and explored this heterogeneity within a subgroup analysis.

Electronic searches

The search strategy for all databases are available in Supporting Information: Appendix 1. The search was conducted in June 2022 and an updated search was conducted in January 2023 for all databases.

Searching other resources

An additional hand search was made in the reference list of systematic reviews identified in our literature search, together with a manual search of the reference lists of the studies included in the review.

Description of methods used in primary research

We expected primary research to use a variety of methods for determining the efficacy of interventions to improve outdoor mobility, and that some of these studies used a randomized controlled design. The interventions differed in their delivery and had different active components. We anticipated that most comparison groups would be categorized as treatment as usual or no intervention. The studied populations with disabilities would vary, but we expected that a large proportion of the populations in the included studies would be people living with disabilities after stroke.

Selection of studies

Two reviewers (MR, BI) independently screened titles and abstracts and excluded reports which did not match our inclusion and exclusion criteria. The Covidence software was used for the screening process. Any disagreements were solved by discussion together with a third reviewer (EML). When insufficient information to assess eligibility was available in the title or abstract the report was assessed in full text.

Two reviewers (MR, EML) independently assessed the full-text versions of the reports in Covidence and disagreements were solved by discussion or together with a third reviewer by consensus (SI). If the eligibility of the report was still uncertain, attempts were made to contact the primary investigators of the report for clarification.

Data extraction and management

Two reviewers (MR, BI) independently extracted data from the included reports. The data extraction form (Supporting Information Appendix 2) was piloted by the reviewers in Covidence and revised before starting data extraction. Disagreements in data extraction were solved by discussion or together with a third reviewer (EML) by consensus. The analyses were conducted in RevMan Web.

Assessment of risk of bias in included studies

Two reviewers (MR, BI) independently assessed the risk of bias in each study outcome using 'Risk of bias in randomized trials (RoB 2)' tool (Sterne, 2019). Disagreements in the risk of bias-evaluations were solved together with a third reviewer (EML) by consensus.

We assessed risk of bias for each study outcome using the 5 domains from RoB2.

Domain 1: Bias arising from the randomization process

For each domain, a series of signalling questions (yes, probably yes, no information, probably no, no) will guide the judgement of risk of bias (low risk, some concerns and high risk). We included text alongside these judgements to provide supporting information for our decisions. The risk of bias judgements is presented alongside the effect estimates in the forest plot of a meta-analysis.

Measures of treatment effect

Statistical analyses were done using Review Manager (RevMan Web) and outcome data was synthesized and summarized in meta-analyses if they were sufficiently homogeneous, both clinically and statistically. If a meta-analysis was not feasible, we present the results narratively and, if applicable, within a forest plot.

Dichotomous data were analysed using relative risk ratio (RR) together with 95% confidence intervals (CI).

Continuous data were converted to standardized mean difference (SMD) together with 95% CIs for analysis. If data in the reports were not presented in means, standard deviations or effect sizes, methods suggested in the Cochrane Handbook was used to calculate the standardized mean difference.

If both adjusted and non-adjusted effect estimates were reported in the same study, we extracted data from the adjusted estimate if sufficient data and assumptions were reported to use this effect estimate in a pooled analysis.

Unit of analysis issues

Studies with multiple outcome measurement time-points were extracted and categorized according to what time after commencing treatment they were collected, that is, short-term (≤6 months) or long-term (≥7 months). If several time points were reported within one of the categories, we used and extracted the time point with the longest follow-up.

When a study with three or more intervention arms fit our inclusion criteria for more than one intervention or comparison we only included the intervention arm that is most closely aligned with the study aim and inclusion criteria. In addition, we performed a sensitivity analysis exploring the potential impact of our choice.

Criteria for determination of independent findings

The review focused on each eligible study and not on each report of a study. Different reports from the same study population was handled as one study.

If several outcome measures were used to measure the same or a similar outcome in an included study, we only used one of the outcome measures as described in the Cochrane Handbook (Cochrane Handbook). All outcomes are transparently reported in the Characteristics of included studies-table and sensitivity analysis was planned to be conducted, when applicable, using all outcomes reported in the outcome category to test the robustness (see Sensitivity Analysis section).

To mitigate any risk of outcome-driven choices when several outcome measures are reported within the same outcome category within a study, we extracted outcomes based on a hierarchy inspired by examples in the Cochrane Handbook. 1.

When possible, we extracted the outcome measure that is the primary outcome in each included study.

In addition, we planned to consult a statistician specialized in psychometrics for guidance if two measurements within the same outcome domain are identified and reported within a study.

A sensitivity analysis was planned when several outcomes are reported within the same domain to explore if our choice of outcome has an impact on the results.

If only a part of a measurement or a subscale were presented (e.g., only the Social Function-domain in SF36) and full data for this outcome measure were not available through contact with study authors, we explored if the part or subscale is overlapping with the construct of the intended outcome in a meaningful way. If not, the reported outcome measure was excluded from the analysis.

Dealing with missing data

Missing outcome-related data were identified in the data extraction process of the included full-text reports. We attempted to contact the primary investigators of the report to obtain missing data. If missing data were not possible to obtain from the primary investigators, or calculated based on guidance from the Cochrane Handbook which prevents calculating a SMD or a RR, the report was excluded from a meta-analysis and the Summary of Findings-table (SoF-table) but were planned to be presented separately in the result section. The missingness of outcome data was handled according to the guidance provided in the Cochrane Handbook for data extraction and Risk of Bias judgements.

Assessment of heterogeneity

To assess the heterogeneity of the effect estimates we used the I ² statistic and the τ ² statistic together with a visual assessment of the forest plots.

We used the thresholds, as reported in the Cochrane Handbook, as a rough guide to the interpretation of the I ² statistic:

0%–40%: might not be important

Heterogeneity was further explored in subgroup analysis and sensitivity analysis.

Assessment of reporting biases

We planned to create and examine a funnel plot to explore the possibility of small-study biases. In interpreting the funnel plot, we planned to examine the different reasons possible for funnel plot asymmetry as outlined in section 10.4 of the Cochrane Handbook, and relate this to the results of the review. If we were able to pool more than 10 trials, we planned to undertake formal statistical tests to investigate funnel plot asymmetry and follow the recommendations in section 10.4 of the Cochrane Handbook. To assess outcome reporting bias, we checked trial protocols against published reports. For studies published after 1 July 2005, we screened the Clinical Trial Register at the International Clinical Trials Registry Platform of the World Health Organization (http://apps.who.int/trialssearch) for the trial protocol.

Data synthesis

Based on guidance provided in the Cochrane Handbook, the decision to conduct a meta-analysis was based on the level of heterogeneity of the interventions included, the included populations and based on the effect estimate in our comparisons and the predefined subgroups. For example, if the clinical heterogeneity is considered too large and a statistical synthesis of effect estimates from very different studies could lead to misleading effect estimates, no meta-analysis was conducted.

We expected a large variation between different interventions included in this review and, due to the heterogeneity, we used a random-effects model. The heterogeneity was further explored within subgroup analyses and sensitivity analyses. If we judged meta-analysis to be inappropriate according to guidance from the Cochrane Handbook, we planned to analyse and interpret individual studies separately in a narrative synthesis.

Subgroup analysis and investigation of heterogeneity

Subgroup analyses was planned to be conducted in three different areas based on: 1.

Population characteristics. Subgroup analysis of population characteristics will be grouped according to diagnose group as defined in the included studies. For example: stroke, dementia, fractures, osteoarthritis and older adults.

Sensitivity analysis

Sensitivity analyses was conducted to explore the potential impact the methods chosen and used in the review could have on our conclusions. Exploration of the potential impact was planned to be done by:

Removing studies using ‘per protocol’ or ‘as treated’ analysis (adherence to intervention) instead of intention to treat (assignment to intervention).

Treatment of qualitative research

We did not include qualitative research.

Summary of findings and assessment of the certainty of the evidence

Two reviewers independently rated the certainty of the body of evidence for the different included outcomes. We used the GRADE system to rank the certainty of the evidence using the GRADEprofiler Guideline Development Tool software (GRADEpro) following the guidelines in the Cochrane Handbook and GRADE Handbook.

The GRADE approach includes five domains: study limitations (risk of bias), unexplained heterogeneity and inconsistency of effect, imprecision, indirectness, and publication bias to assess the certainty of the body of evidence for each outcome.

The GRADE system uses the following criteria for determining the certainty of evidence:

High: we are very confident that the true effect lies close to that of the estimate of the effect

The GRADE system uses study design as a marker of quality. RCTs are considered to be a higher certainty of evidence but can be downgraded for important limitations.

Factors that may decrease the certainty in a body of evidence are:

Serious or very serious study limitations (risk of bias)

We included a total of 4 ‘Summary of findings’ (SoF) tables to present the main findings of the review which will follow the structure as described in the (Cochrane Handbook Chapter 14). We included key information concerning the certainty of evidence, the magnitude and confidence interval of the effect of the interventions examined, and the sum of available data on the outcomes. 1.

Summary of Findings-Tables 1 and 2 summarized the 6 outcomes measured in the short term (≤6 months) for outdoor mobility interventions compared with a control intervention.

Results of the search

The PRISMA flowchart (Figure, 2) illustrates the full search process and the specific search strategies in each database and the number of identified articles is available in Supporting Information: Appendix 1.

OPEN IN VIEWER

Figure 2

PRISMA 2020 flow diagram for new systematic reviews which included searches of databases and registers only.

We identified a total of 22,333 articles from our electronic and supplementary searches and 12,894 articles were left after duplicates were removed. We screened these records and excluded 12,550 irrelevant articles. We assessed 344 articles in full-text, excluded 310 studies not matching our inclusion criteria and finally included 34 studies. Of these, 12 studies reported our prespecified outcomes, 10 studies did not report any of our prespecified outcomes, and 12 studies are still ongoing.

Included studies

Detailed population-, outcome- and study-characteristics are described in Table 1 and detailed intervention characteristics are described in Table 2. The 22 studies included a total of 2675 participants and examined a total of 24 different outdoor mobility interventions. Two studies explored the efficacy of two different outdoor mobility interventions (Ada, 2012; Kim, 2016) to a control intervention and 20 studies compared one outdoor mobility intervention to a control intervention. The included sample sizes ranged from 18 (Mendoza, 2015) to 632 participants (Mänty, 2009). All studies included were individually randomized controlled trials.

Excluded studies

We excluded 310 articles after assessing the full text. All excluded articles are available in the Excluded studies section. Of these, 283 articles were not aimed to evaluate interventions to improve outdoor mobility, 14 articles did not use a randomized design, six articles were conducted in populations without disabilities, five articles evaluated interventions aimed to change the environment or infrastructure and two articles were retracted.

Ongoing studies

We identified 12 ongoing studies.

Seven of the ongoing studies (Dean, 2021; Drks, 2021; Edgren, 2019; Haeger, 2022; Nct, 2015; Nct, 2020a; Salbach, 2019) are exploring outdoor mobility interventions in older adults. Two studies (Isrctn, 2016; Nct, 2020) are exploring outdoor mobility interventions after stroke, one in people with Parkinson's Disease (Nct, 2018a), one in mobility scooter users (Mortenson, 2017) and one in a general population of adults living with long-term disabilities (Nct, 2018). The specific type of outdoor mobility interventions could not be described due to limited descriptions in trial registries and protocols of most of the ongoing studies.

Overall risk of bias

Risk of bias for each study and the reported outcome is presented in Figure 3. Detailed risk of bias assessments and consensus responses to each signalling question is available in Supporting Information Appendix 3.

OPEN IN VIEWER

Figure 3

Detailed risk of bias for each outcome.

Out of the 56 reported outcomes in the included studies, 55 outcomes had an overall high risk of bias, one outcome had some concerns and no outcome had an overall low risk of bias.

Bias arising from the randomization process

We judged 25 outcomes reported in five studies (DePaul, 2015; Fairhall, 2012; Lord, 2008; Magaziner, 2019; Turunen, 2020) to have low risk of bias arising from the randomization process.

We judged 27 outcomes reported in five studies (Ada, 2012; Kim, 2014; Logan, 2004; Logan, 2014; Ullrich, 2021) to have some concern of bias arising from the randomization process, due to unclear reporting of allocation concealment and baseline imbalances in potential prognostic factors for outdoor mobility.

We judged four outcomes reported in three studies (Best, 2016; Miller, 2019; Rantanen, 2015) to have high risk of bias arising from the randomization process due to having no allocation concealment and large baseline imbalances in potential prognostic factors for outdoor mobility.

Bias due to deviations from intended interventions

We judged eight outcomes reported in three studies (DePaul, 2015; Miller, 2019; Ullrich, 2021) to have low risk of bias due to deviations from intended interventions due to using well-designed attention control groups and thus minimizing participant knowledge and preferences about group allocations.

We judged 26 outcomes reported in six studies (Logan, 2004; Logan, 2014; Lord, 2008; Magaziner, 2019; Rantanen, 2015; Turunen, 2020) to have some concern of bias due to deviations from intended interventions, mainly due to using comparison groups with only some attention control, or common practice interventions that do not aim to improve outdoor mobility. This would lead to knowledge about the group allocations, potential participant preference and change in health-related behaviours with the potential to impact outcomes.

We judged 22 outcomes reported in five studies (Ada, 2012; Best, 2016; Fairhall, 2012; Kim, 2014; Turunen, 2020) to have high risk of bias due to deviations from intended interventions, mainly due to using comparison groups with no intervention delivered and with a strong likelihood of participant preference, knowledge about the group allocations and changed health-related behaviours impacting the outcomes.

Bias due to missing outcome data

We judged seven outcomes reported in three studies (Ada, 2012; Best, 2016; Logan, 2014) to have low risk of bias due to missing outcome data, due to having minimal loss of participants at follow-ups and no or minimal differences between groups in reasons for discontinuing and patient characteristics.

We judged 16 outcomes reported in five studies (Fairhall, 2012; Logan, 2014; Miller, 2019; Rantanen, 2015; Turunen, 2020) to have some concern of bias due to missing outcome data, mainly due to moderate loss of participants at follow-ups, but with no, or unclear attempts to correct for any bias and a potential impact on the results.

We judged 33 outcomes reported in nine studies (Ada, 2012; DePaul, 2015; Kim, 2014; Logan, 2004; Logan, 2014; Lord, 2008; Magaziner, 2019; Turunen, 2020; Ullrich, 2021) to have high risk of bias due to missing outcome data, mainly due to substantial loss of participants at follow-ups with a potential of major impact on the results, together with no or unsuccessful attempts to correct for any bias.

Bias in measurement of the outcome

We judged 10 outcomes reported in four studies (DePaul, 2015; Logan, 2014; Miller, 2019; Ullrich, 2021) to have low risk of bias in measurement of the outcome, mainly due to well-designed attention control groups and thus minimizing participant preference and knowledge about the group allocations and limiting the impact on self-reporting outcomes, or using objective outcomes such as mortality measured in registries with limited possibility of introducing bias.

We judged 19 outcomes reported in four studies (Logan, 2004; Logan, 2014; Lord, 2008; Magaziner, 2019) to have some concerns of bias in measurement of the outcome, mainly due to using self-reported measures and using comparison groups with only some attention control, or common practice interventions that do not aim to improve outdoor mobility with a potential to lead to participant preference with a possibility to impact outcome reporting.

We judged 27 outcomes reported in seven studies (Ada, 2012; Best, 2016; Fairhall, 2012; Kim, 2014; Logan, 2014; Rantanen, 2015; Turunen, 2020) to have high risk of bias in measurement of the outcome. This was mainly due to using self-reported outcome measures and using comparison groups with no intervention with a strong likelihood of participant preference and a strong likelihood to impact outcome reporting.

Bias in selection of the reported result

We judged 39 outcomes reported in eight studies (Ada, 2012; Best, 2016; DePaul, 2015; Fairhall, 2012; Logan, 2014; Magaziner, 2019; Miller, 2019; Ullrich, 2021) to have low risk of bias in the selection of the reported results, due to having pre-specified protocols available before any data analyses took place and no deviations on statistical or outcome measures.

We judged eight outcomes reported in three studies (Kim, 2014; Logan, 2004; Rantanen, 2015) to have some concern of bias in the selection of reported results mainly due to having retrospectively registered protocols with some potential deviations related to reporting of subscales, or no protocol with some concerns regarding reporting of only certain subscales of an outcome measure.

We judged nine outcomes reported in two studies (Lord, 2008; Turunen, 2020) to have high risk of bias in selection of reported results due to no protocol available and substantial opportunities to report certain outcome measures and scales and selection of certain analyses, or having substantial deviations from important pre-specified outcomes or selection of new outcome measures without justification.

Comparison 1.1: Skill training interventions versus control intervention not aimed at improving outdoor mobility

Activity outside the home in the shorter term (≤6 months)

Six studies (Best, 2016; DePaul, 2015; Logan, 2004; Logan, 2014; Miller, 2019; Ullrich, 2021) reported this outcome.

Overall, the evidence is very uncertain if skill training interventions improve activity outside the home among people with disabilities in the shorter term (SMD: 0.18; 95% CI: −0.20 to 0.56; I ² = 83%; 925 participants; six studies; very low certainty evidence; Analysis 1.1). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for very serious imprecision and one level for very serious inconsistency.

OPEN IN VIEWER

Figure Analysis 1.1

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 1: Activity outside the home ≤6 months.

Subgroup analysis of skill training interventions in people after stroke did not reduce uncertainty and the evidence is very uncertain if skill training interventions may improve activity outside the home among people after a stroke (SMD: −0.26; 95% CI: −0.36 to 0.88; I ² = 92%; 739 participants; three studies; very low certainty evidence; Analysis 2.1). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for very serious imprecision and one level for very serious inconsistency.

OPEN IN VIEWER

Figure Analysis 2.1

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 1: Skill training—Activity outside the home ≤6 months.

Subgroup analysis of skill training interventions in people using wheelchairs did not reduce uncertainty and the evidence is very uncertain if skill training interventions may improve activity outside the home among wheelchair users (SMD: −0.18; 95% CI: −0.66 to 0.30; I ² = 0%; 68 participants; two studies; very low certainty evidence; Analysis 2.1). The certainty of evidence was downgraded by one level for very serious risk of bias and two level for very serious imprecision.

Subgroup analysis of skill training interventions in people with cognitive impairments slightly reduced uncertainty and may improve activity outside the home among people living with cognitive impairments (SMD: 0.44; 95% CI: 0.07 to 0.81; I ² = NA; 118 participants; one study; low certainty evidence; Analysis 2.1). The certainty of evidence was downgraded by one level for serious risk of bias and one level for serious imprecision.

Activity outside the home in the longer-term (≥7 months)

Two studies (Logan, 2004; Logan, 2014) reported this outcome, and both were conducted among people after a stroke.

Overall the evidence is very uncertain if skill training interventions improve activity outside the home in the longer term (SMD: 0.38; 95% CI: −0.55 to 1.30; I ² = 96%; 672 participants; two studies; very low certainty evidence; Analysis 1.2). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for very serious imprecision, and one level for very serious inconsistency.

OPEN IN VIEWER

Figure Analysis 1.2

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 2: Activity outside the home ≥7 months.

Engagement in everyday life activities in the shorter term (≤6 months)

Three studies (Logan, 2004; Logan, 2014; Lord, 2008) reported this outcome, and all were conducted among people after stroke.

Overall skill training interventions may improve engagement in everyday life activities among people living with disabilities in the shorter term (RR: 1.46; 95% CI: 1.16 to 1.84; I ² = 7%; risk difference [RD]: 0.15; 95% CI: −0.02 to 0.32; I ² = 71%; 692 participants; three studies; low certainty evidence; Analysis 1.3). The certainty of evidence was downgraded by two levels for very serious risk of bias.

OPEN IN VIEWER

Figure Analysis 1.3

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 3: Engagement in everyday life activities ≤6 months.

Engagement in everyday life activities in the longer-term (≥7 months)

Two studies (Logan, 2004; Logan, 2014) reported this outcome, and both were conducted among people after stroke.

Overall the evidence is very uncertain if skill training interventions improve engagement in everyday life activities in the longer term (RR: 1.40; 95% CI: 0.91 to 2.15; I ² = 69%; RD: 0.14; 95% CI: −0.09 to 0.37; 600 participants; two studies; very low certainty evidence; Analysis 1.4). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for very serious imprecision and one level for very serious inconsistency.

OPEN IN VIEWER

Figure Analysis 1.4

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 4: Engagement in everyday life activities ≥7 months.

Participation in the shorter term (≤6 months)

Seven studies (Ada, 2012; DePaul, 2015; Kim, 2014; Logan, 2004; Logan, 2014; Lord, 2008; Miller, 2019) reported this outcome.

Overall, the evidence is very uncertain if skill training interventions improve participation among people living with disabilities in the shorter term (SMD: 0.01; 95% CI: −0.23 to 0.25; I ² = 53%; 886 participants; seven studies; very low certainty evidence; Analysis 1.5). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for very serious imprecision and one level for serious inconsistency. Analysis 2.2, 2.3, 2.4, 2.5.

OPEN IN VIEWER

Figure Analysis 1.5

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 5: Participation ≤6 months.

OPEN IN VIEWER

Figure Analysis 2.2

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 2: Physical training—Activity outside the home ≤6 months.

OPEN IN VIEWER

Figure Analysis 2.3

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 3: Activity outside the home ≥7 months.

OPEN IN VIEWER

Figure Analysis 2.4

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 4: Engagement in everyday life activities ≤6 months.

OPEN IN VIEWER

Figure Analysis 2.5

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 5: Engagement in everyday life activities ≥7 months.

Subgroup analysis of skill training interventions in people after stroke did not reduce uncertainty and the evidence is very uncertain if skill training interventions may improve participation among people after stroke (SMD: −0.06; 95% CI: −0.20 to 0.31; I ² = 55%; 846 participants; six studies; very low certainty evidence; Analysis 2.6). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for very serious imprecision and one level for serious inconsistency.

OPEN IN VIEWER

Figure Analysis 2.6

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 6: Skill training—Participation ≤6 months.

Subgroup analysis of skill training interventions among people using wheelchairs did not reduce uncertainty and the evidence is very uncertain if skill training interventions may improve participation for people using wheelchairs (SMD: −0.43; 95% CI: −1.06 to 0.20; I ² = NA; 40 participants; one study; very low certainty evidence; Analysis 2.6). The certainty of evidence was downgraded by one level for very serious risk of bias, two levels for very serious imprecision.

Participation in the longer-term (≥7 months)

Three studies (Ada, 2012; Logan, 2004; Logan, 2014) reported this outcome, and all were conducted among people after stroke.

Overall, the evidence is very uncertain if skill training interventions improve participation among people living with disabilities in the longer term (SMD: −0.05; 95% CI: −0.31 to 0.20; I ² = 54%; 674 participants; three studies; very low certainty evidence; Analysis 1.6). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for serious imprecision and one level for serious inconsistency.

OPEN IN VIEWER

Figure Analysis 1.6

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 6: Participation ≥7 months.

Health-related quality of life in the shorter term (≤6 months)

Four studies (Ada, 2012; Logan, 2014; Rantanen, 2015; Ullrich, 2021) reported this outcome.

Overall, the evidence is very uncertain if skill training interventions improve health-related quality of life among people living with disabilities in the shorter term (SMD: 0.13; 95% CI: −0.20 to 0.46; I ² = 74%; 779 participants; four studies; very low certainty evidence; Analysis 1.7). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for very serious imprecision and one level for serious inconsistency.

OPEN IN VIEWER

Figure Analysis 1.7

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 7: Health-related quality of life ≤6 months.

Subgroup analysis of skill training interventions among people after stroke did not reduce uncertainty and the evidence is very uncertain if skill training interventions may improve health-related quality of life for people after a stroke (SMD: −0.13; 95% CI: −0.47 to 0.21; I ² = 52%; 552 participants; two studies; very low certainty evidence; Analysis 2.9). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for very serious imprecision and one level for serious inconsistency.

OPEN IN VIEWER

Figure Analysis 2.7

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 7: Physical training—Participation ≤6 months.

OPEN IN VIEWER

Figure Analysis 2.8

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 8: Participation ≥7 months.

OPEN IN VIEWER

Figure Analysis 2.9

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 9: Skill training—Health-related quality of life ≤6 months.

Subgroup analysis of skill training interventions among people with cognitive impairments reduced uncertainty slightly and skill training interventions may improve health-related quality of life for people with cognitive impairments (SMD: 0.49; 95% CI: 0.12 to 0.88; I ² = NA; 118 participants; one study; low certainty evidence; Analysis 2.9). The certainty of evidence was downgraded by one level for serious risk of bias, one level for serious imprecision.

Subgroup analysis of skill training interventions among older adults with disabilities did not reduce uncertainty and the evidence is very uncertain if skill training interventions may improve health-related quality of life for older adults living with disabilities (SMD: 0.38; 95% CI: 0.00 to 0.76; I ² = NA; 109 participants; one study; very low certainty evidence; Analysis 2.9). The certainty of evidence was downgraded by two levels for very serious risk of bias and one level for serious imprecision.

Health-related quality of life in the longer-term (≥7 months)

Two studies reported this outcome (Ada, 2012; Logan, 2014), and both were conducted among people after stroke.

Overall, the evidence is very uncertain if skill training interventions improve health-related quality of life among people living with disabilities in the longer term (SMD: −0.05; 95% CI: −0.23 to 0.13; I ² = 0%; 496 participants; two studies; very low certainty evidence; Analysis 1.8). The certainty of evidence was downgraded by two levels for very serious risk of bias and one level for serious imprecision.

OPEN IN VIEWER

Figure Analysis 1.8

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 8: Health-related quality of life ≥7 months.

Major harms in the shorter term (≤6 months)

Two studies reported this outcome (DePaul, 2015; Lord, 2008) by using: a composite measure of major cardiovascular events, stroke or mortality; and recurrence of stroke as measurements. Both studies were conducted among people after stroke.

Overall, the evidence is very uncertain if skill training interventions lead to major harm among people living with disabilities in the shorter term (RR: 0.83; 95% CI: 0.27 to 2.54; I ² = 0%; RD: −0.01; 95% CI: −0.13 to 0.10; 496 participants; two studies; very low certainty evidence; Analysis 1.9). The certainty of evidence was downgraded by one level for very serious risk of bias and two levels for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 1.9

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 9: Major adverse events ≤6 months.

Major harms in the longer-term (≥7 months)

One study reported this outcome (Logan, 2014) by using mortality as a measurement. The study was conducted among people after stroke.

Overall, the evidence is very uncertain if physical training interventions lead to major harm among people living with disabilities in the longer term (RR: 0.98; 95% CI: 0.45 to 2.14; I ² = NA; RD: 0.00; 95% CI: −0.03 to 0.03; 568 participants; one study; very low certainty evidence; Analysis 1.10). The certainty of evidence was downgraded by two levels for very serious risk of bias and one level for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 1.10

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 10: Major adverse events ≥7 months.

Minor harms in the shorter term (≤6 months)

Two studies (Ada, 2012; DePaul, 2015) reported this outcome by using falls as a measurement. Both studies were conducted among people after stroke.

Overall, the evidence is very uncertain if physical training interventions lead to any minor harm among people living with disabilities in the shorter term (RR: 1.14; 95% CI: 0.63 to 2.04; I ² = 0%; RD: 0.03; 95% CI: −0.12 to 0.18; 124 participants; two studies study; very low certainty evidence; Analysis 1.11). The certainty of evidence was downgraded by two levels for very serious risk of bias and one level for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 1.11

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 11: Minor adverse events ≤6 months.

Minor harms in the longer-term (≥7 months)

Two studies (Ada, 2012; Logan, 2014) reported this outcome by using falls as a measurement. Both studies were conducted among people after stroke.

Overall, the evidence is very uncertain if physical training interventions lead to any minor harm among people living with disabilities in the longer term (RR: 1.00; 95% CI: 0.84 to 1.18; I ² = 0%; RD: 0.00; 95% CI: −0.08 to 0.08; 627 participants; two studies study; very low certainty evidence; Analysis 1.12). The certainty of evidence was downgraded by two levels for very serious risk of bias and one level for serious imprecision.

OPEN IN VIEWER

Figure Analysis 1.12

Comparison 1: Specific outdoor mobility interventions versus control interventions, Outcome 12: Minor adverse events ≥7 months.

Comparison 1.2: Physical training interventions versus control intervention not aimed at improving outdoor mobility

Major harms in the shorter term (≤6 months)

Two studies (Magaziner, 2019; Turunen, 2020) reported this outcome using: a composite score of severe-, life-threatening- or fatal harms; and need of emergency room service.

Overall, the evidence is very uncertain if physical training interventions lead to major harm among people living with disabilities in the shorter term (RR: 0.63; 95% CI: 0.20 to 2.03; I ² = 59%; RD: −0.06; 95% CI: −0.22 to 0.10; 317 participants; two studies; very low certainty evidence; Analysis 1.9). The certainty of evidence was downgraded by one level for very serious risk of bias, one level for serious imprecision and one level for serious inconsistency.

Subgroup analysis of physical training interventions in older adults living with disabilities did not reduce uncertainty and the evidence is very uncertain if physical training interventions lead to major harm among older adults living with disabilities (RR: 0.36; 95% CI: 0.14 to 0.95; I ² = NA; RD: −0.16; 95% CI: −0.29 to −0.02; 109 participants; one study; very low certainty evidence; Analysis 2.12). The certainty of evidence was downgraded by two levels for very serious risk of bias, and one level for serious imprecision.

OPEN IN VIEWER

Figure Analysis 2.10

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 10: Physical training—Health-related quality of life ≤6 months.

OPEN IN VIEWER

Figure Analysis 2.12

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 12: Major adverse events (Dichotomous measures) ≤6 months.

Subgroup analysis of physical training interventions in people with musculoskeletal injuries did not reduce uncertainty and the evidence is very uncertain if physical training interventions lead to major harm among people with musculoskeletal injuries (RR: 1.20; 95% CI: 0.38 to 3.81; I ² = NA; RD: −0.01; 95% CI: −0.05 to 0.07; 210 participants; one study; very low certainty evidence; Analysis 2.12). The certainty of evidence was downgraded by one level for serious risk of bias, and two levels for very serious imprecision.

Comparison 2: Skill training interventions of different length

One study (Ada, 2012) compared two different skill training interventions for people with stroke which differed in length of the intervention period (3 sessions a week over 8 weeks versus 3 sessions a week over 16 weeks) and reported this outcome. Analysis 2.7, 2.8, Analysis 2.8, 2.10, 2.11, 2.13, 2.14, 2.15.

OPEN IN VIEWER

Figure Analysis 2.11

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 11: Health-related quality of life ≥7 months.

OPEN IN VIEWER

Figure Analysis 2.13

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 13: Major adverse events (Dichotomous measures) ≥7 months.

OPEN IN VIEWER

Figure Analysis 2.14

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 14: Minor adverse events (Dichotomous measures) ≤6 months.

OPEN IN VIEWER

Figure Analysis 2.15

Comparison 2: Specific outdoor mobility interventions versus control interventions—by conditions (Subgroup analyses), Outcome 15: Minor adverse events (Dichotomous measures) ≥7 months.

Participation in the shorter term (≤6 months)

Overall, the evidence is very uncertain if skill training interventions of different intensity have an impact on participation in the shorter term (SMD: 0.00; 95% CI: −0.48 to 0.48; I ² = NA; 68 participants; one study; very low certainty evidence; Analysis 3.1). The certainty of evidence was downgraded by one level for serious risk of bias, and two levels for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 3.1

Comparison 3: Different intensity of outdoor mobility interventions, Outcome 1: Participation ≤6 months.

Participation in the longer-term (≥7 months)

Overall, the evidence is very uncertain if skill training interventions of different intensity have an impact on participation in the longer term (SMD: 0.09; 95% CI: −0.39 to 0.57; I ² = NA; 68 participants; one study; very low certainty evidence; Analysis 3.2). The certainty of evidence was downgraded by one level for serious risk of bias, and two levels for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 3.2

Comparison 3: Different intensity of outdoor mobility interventions, Outcome 2: Participation ≥7 months.

Health-related quality of life in the shorter term (≤6 months)

Overall, the evidence is very uncertain if skill training interventions of different intensity have an impact on health-related quality of life in the shorter term (SMD: −0.12; 95% CI: −0.59 to 0.36; I ² = NA; 68 participants; one study; very low certainty evidence; Analysis 3.3). The certainty of evidence was downgraded by one level for serious risk of bias, and two levels for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 3.3

Comparison 3: Different intensity of outdoor mobility interventions, Outcome 3: Health-related quality of life ≤6 months.

Health-related quality of life in the longer-term (≥7 months)

Overall, the evidence is very uncertain if skill training interventions of different intensity have an impact on health-related quality of life in the longer term (SMD: 0.20; 95% CI: −0.27 to 0.68; I ² = NA; 68 participants; one study; very low certainty evidence; Analysis 3.4). The certainty of evidence was downgraded by one level for serious risk of bias, and two levels for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 3.4

Comparison 3: Different intensity of outdoor mobility interventions, Outcome 4: Health-related quality of life ≥7 months.

Minor harms in the shorter term (≤6 months)

Overall, the evidence is very uncertain if skill training interventions of different intensity have an impact on minor adverse events in the shorter term (RR: 5.45; 95% CI: 0.70 to 42.73; I ² = NA; RD: 0.15; 95% CI: 0.00 to 0.29; 68 participants; one study; very low certainty evidence; Analysis 3.5). The certainty of evidence was downgraded by one level for serious risk of bias, and two levels for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 3.5

Comparison 3: Different intensity of outdoor mobility interventions, Outcome 5: Minor adverse events ≤6 months.

Minor harms in the longer-term (≥7 months)

Overall, the evidence is very uncertain if skill training interventions of different intensity have an impact on minor adverse events in the shorter term (RR 0.60; 95% CI 0.32 to 1.13; I² = NA; RD −0.21; 95% CI −0.45 to 0.04; 68 participants; one study; very low certainty evidence; Analysis 3.6). The certainty of evidence was downgraded by one level for serious risk of bias, and two levels for very serious imprecision.

OPEN IN VIEWER

Figure Analysis 3.6

Comparison 3: Different intensity of outdoor mobility interventions, Outcome 6: Minor adverse events ≥7 months.

Subgroup analyses

Due to a limited number of studies conducted for each outcome and type of intervention in combination with a substantial risk of bias, there would be very limited confidence in any results of subgroup analyses when assessed with the ICEMAN tool for the credibility of subgroup effects (Schandelmaier, 2020). Hence, we did not conduct any subgroup analyses for: risk of bias; different control interventions; different intervention areas; different intervention delivery; different intervention duration; different intervention mobility modes; and different intervention settings.

Sensitivity analyses

Removing outcome measures being reported by subscales or parts of measurements: One study, Logan, 2014, measured and presented only subscales of SF36, where Social Functioning was one of the primary outcomes and was included in our primary analyses. The study did not present any formal analyses of full scores or other sub-scores of SF36, but states that ‘The other domains from the SF-36v2 were also analysed at 6 months and showed no significant difference between groups’, which was not different from the initial analysis done and did not change our conclusions.

As either all, or none, of the included studies were excluded from the pre-planned sensitivity analyses, it was not feasible to conduct sensitivity analyses related to: Removing studies using ‘per protocol’ or ‘as treated’ analysis; Removing studies having a high risk of bias due to missing outcome data; Including the reported outcomes for the second active intervention arm; Including the second reported outcome measure in studies reporting multiple outcome measures for the same outcome; Removing studies where environmental interventions are combined with interventions to increase individuals’ capacity and performance, or; Removing studies using balance training as an intervention for the outcome of adverse events, including falls.

Current quality and uncertainties in the evidence base

The vast majority of the studies and interventions had an overall high risk of bias for almost all outcomes due to methodological limitations in several domains. Missing outcome data, possibly due to the longer follow-up periods and difficulties keeping participants in the trial, where people in the control group dropped out to a larger extent, are the most notable reasons. Additionally, half of the interventions used inactive control groups receiving no treatment or unspecified treatment as usual, which often introduced a risk of bias. Additional risks of bias were often introduced in the reporting of the subjectively participant-reported outcome measurements due to knowledge and possible preference of the group allocation.

Most of the included studies, albeit with clearly stated aims to improve outdoor mobility, did not assess if this was achieved or not because of lack of any type of outdoor mobility measure. Instead, surrogate outcomes of unclear predictive markers such as measures of walking speed or measures of strength and endurance of the lower extremities were used. Consequently, it was not possible to assess if increased outdoor mobility was achieved. Future studies should aim to include valid measures of outdoor mobility to evaluate if improvements in surrogate measurements also improve outdoor mobility. To be able to evaluate any potential beneficial effects of increased outdoor mobility, such as engagement in activities, quality of life or participation, such outcome measures should be included in future trials. Turning to another quality aspect, most of the included studies enroled few participants and were not powered to detect important estimates of either benefits or harms, resulting in wide confidence intervals and further uncertainty about potential effects. When several studies and their included interventions reported the same outcome, the reported results often had substantial or considerable heterogeneity in the reported effects, adding to the uncertainty about the actual effectiveness of the outdoor mobility interventions. Due to the many potential reasons for this heterogeneity and few included studies reporting on each outcome we were unable to confidently explore and explain these differences.

Due to the low number of included studies reporting on each outcome, funnel plots were not useful to assess any publication bias, which leads to uncertainties regarding the presence of any selective publication that could impact our conclusions.

Improvements in future studies

To improve the certainty of evidence, future studies should be conducted in the context they are supposed to be delivered in by professionals working in these contexts. For example, in multicenter studies including clinics outside of university hospital settings and in close collaboration with clinical units to enrol relevant participants that reflect the actual target population of interest. Collaborations between researchers, clinics and healthcare professionals are likely needed to create reliable evidence and enrol enough participants to identify important benefits and harms of interventions. But this is often difficult due to limitations in funding and often a lack of large scale coordination within the healthcare systems and the clinicians who meet, enrol and support the heterogeneous group of people living with disabilities.

To reduce the uncertainty related to very imprecise effect estimates, future studies should use strategies to reduce attrition and be powered to detect changes in outdoor mobility as well as other outcomes, including harms, in the long term. To improve enrolment and avoid trials using very small sample sizes, outdoor mobility interventions could be implemented and evaluated as part of standard clinical practice. This could improve the number of enroled participants, and lead to more useful evidence due to similar settings and delivery being used in both research and practice.

Careful planning of the randomization process, including allocation concealment procedures to avoid any potential biases and reporting these strategies is needed to improve assessments of the randomization procedure and possible impact on the outcomes. Strategies to deal with potential baseline imbalances between intervention groups that can happen due to chance in smaller studies, could be improved by, for example, plans for adjustments for specific covariates or sensitivity analyses.

Using more rigorous control interventions is needed to reduce bias due to deviation from intended intervention and would improve our confidence in the evidence. Complete blinding using a placebo is not feasible in outdoor mobility interventions research, but matching intervention length or actual time spent with an interventionist, and avoiding using ‘no treatment’ controls would limit the risk of bias. Of note, Miller, 2019 designed an active control intervention using seminars of only the theoretical skills needed for outdoor mobility in combination with health promotion - but without any practical application of skill training. In addition, studies following up and comparing differences in health behaviours or use of additional care that could impact the outcomes between groups would mitigate the risk of bias.

Creating and openly sharing detailed pre-specified analysis plans would improve our certainty of the evidence and limit any potential risks of bias of selective reporting. Finally, clear reporting of the actual intervention and the delivery is needed to access specific mechanisms of the interventions and compare these across different studies.

Additional implications for research is described in Authors’ conclusions section.