Effectiveness of community-based rehabilitation interventions incorporating outdoor mobility on ambulatory ability and falls-related self-efficacy after hip fracture: a systematic review and meta-analysis

This review is reported in adherence to the Preferred Reporting Items for Systematic Review and Meta-analysis statement [17]. The protocol is registered on the International Register of Systematic Reviews (PROSPERO CRD42021236541).

We included randomised controlled trials (RCT) of community-based rehabilitation interventions which incorporated outdoor mobility for persons after hip fracture. Rehabilitation was defined as ‘a set of interventions designed to optimize functioning and reduce disability in individuals with health conditions in interaction with their environment’ [18]. Rehabilitation interventions for participants after hip fracture are often complex incorporating several interacting components. We employed a broad definition of ‘outdoor mobility’ to determine the extent to which outdoor mobility was captured by these components. This definition included components which targeted going outdoors for structured/unstructured exercise/activity to those which targeted going outdoors for participation, e.g. taking public transport. We included RCTs which planned to incorporate supervised outdoor mobility, unsupervised outdoor mobility, and/or encouragement of outdoor mobility irrespective of whether this was completed by all participants within the RCT. We included RCTs irrespective of comparator group, outcomes measured, length of follow-up, and publication year. We excluded protocols, pilot/feasibility studies, secondary analyses of RCTs, and nonrandomised studies. We excluded RCTs not published in English, due to lack of resources for expert translation.

We searched MEDLINE, Embase and PsychInfo (OVID), CINAHL (EBSCOhost), PEDro, and OpenGrey for published and unpublished RCTs from database inception to 13 January 2021.

We used a published search strategy based on population, intervention, and study design (hip fracture, rehabilitation, and randomised controlled trials) limited to human and English language (Supplementary File 1) [19].

Three reviewers screened title and abstracts (R1, R2, R3), and two reviewers screened full texts (R1, R3) of potentially eligible RCTs against eligibility criteria. Conflicts were resolved by consensus. Cohen’s Kappa was estimated at k = 0.7 (moderate agreement) for inter-rater reliability prior to consensus work of full-text screening [20].

Two reviewers (R1, R2) piloted data extraction onto a template adapted from the taxonomy to classify and describe fall-prevention interventions [21]. We sought data for the following data items: author, year, location, sample size intervention group, sample size control group; approach — aim, inclusion criteria, exclusion by dementia/cognitive impairment, other exclusion; base — recruitment, site (s) of delivery, assessment delivered by, intervention delivered by; components — assessment as part of intervention, combination of interventions and description; descriptor intervention — supervised/unsupervised (type, duration, frequency, intensity, individual/group), psychological (cognitive behavioural therapy, other, individual/group), environment, assistive technology, knowledge, post-intervention follow-up (period, type, completeness) and strategies to improve uptake/adherence; descriptor control — routine care/no specific intervention, supervised exercises, medication, knowledge, social environment and other; and outcome — primary, secondary, effect for primary outcome at intervention end and intervention follow-up. Following the pilot, an additional data item specifically related to outdoor mobility was added. One reviewer (R2) extracted the remaining data onto the template. Final extraction was checked for accuracy by a second reviewer (R1). We extracted data from the earliest publication where multiple publications referred to one RCT.

Two reviewers independently assessed risk of bias at the study level using the Cochrane Risk of Bias Tool (R1, R2) [22]. Conflicts were resolved by consensus.

For our first objective, we reported the extent to which outdoor mobility was incorporated into rehabilitation interventions in a narrative synthesis. For our second objective, we completed an inverse variance random effects meta-analysis to estimate standardised mean difference (for continuous outcomes) or risk difference (for binary outcomes) and their 95% confidence intervals. We interpreted a standardised mean/risk difference of <0.2 as null, 0.2–0.49 as small, 0.5–0.79 as medium and ≥0.8 as large [23]. Statistical heterogeneity was evaluated using the inconsistency-value (I²). Results of meta-analysis were presented in tables and forest plots. Meta-analyses were completed in RevMan Version 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2011).

Small-study publication bias was evaluated through interpretation of funnel plots for each outcome.

We identified 5681 articles after de-duplication. We excluded 5569 on abstract screening. We excluded 99 on full-text screening for nonrandomised study design (n = 31), population (n = 10), intervention (n = 55), language (n = 2), no response from author for additional data related to eligibility (n = 2) and leaving 12 papers reporting 11 RCTs (Fig. 1).

Most RCTs were at low risk of bias for random sequence generation (n = 10), blinding of outcome assessor (n = 8), incomplete outcome data (n = 8) or selective reporting (n = 11) (Fig. 2). There was insufficient information to assess allocation concealment for 7 RCTs. Lack of blinding of personnel and participants was the most common reason for high bias assignment (n = 5) [24‐29]. In addition, one RCT did not blind outcome assessors [24, 25].

Detailed characteristics for the 11 RCTs are available in Table 1. RCTs were completed in Australia (n = 1) [29], Finland (n = 1) [30], Germany (n = 2) [26, 31], the Netherlands (n = 1) [32], Sweden (n = 2) [24, 25, 28] and the USA (n = 4) [27, 33‐35]. Sample size ranged from 28 [31] to 240 participants [32]. Participants were older adults (eligible age range from 60 years plus [26, 30, 33] to 75 years plus [31]) admitted with hip fracture and treated surgically. Nine RCTs excluded potential participants based on their cognitive function [24‐27, 29‐32, 34, 35]. Karlsson et al. [28] explicitly stated inclusion of participant irrespective of cognitive status, whilst Magaziner et al. excluded participants with ‘low potential to benefit’ or ‘practical impediments to participation’ [33]. Participants were recruited from acute hospital [24, 25, 28, 29, 34, 35], inpatient rehabilitation [26, 31], clinic/health centres [27, 33], nursing and community care facilities [32] or the community [30]. Outcome assessments were completed by physiotherapists [27, 30, 33], occupational therapist [24, 25, 32], gerontologist and psychologist [26], researchers [28] or were not specified [29, 31, 34].

Seven RCTs compared interventions to routine care. This routine care was described as inpatient services, pathways and discharge planning [29]; inpatient rehabilitation for 2–4 weeks [26, 34]; inpatient rehabilitation based on functional needs and a single home therapy evaluation [35]; or interdisciplinary inpatient rehabilitation, discharge planning, referral to ongoing outpatient rehabilitation [24, 25, 28, 32] including handover to physiotherapists/occupational therapists at residential care facilities [28]. Two RCTs provided written materials (home exercise programme [30], non-exercise related written materials [27]) with no further follow-up. Two interventions were compared to sham active controls including seated activities [31], or seated activities and transcutaneous electrical stimulation [33]. Detailed descriptions for each intervention are available in Table 2.

All 11 RCTs included in this review included outdoor mobility in their intervention. This was explicitly stated by 6 RCTs [24‐28, 32, 33] and confirmed with authors for the remaining 5 RCTs [29‐31, 34, 35]. Outdoor mobility was supervised [27‐29, 31, 33, 35], unsupervised [26, 32] or both supervised and unsupervised [24, 25, 30, 34]. The target duration/distance, frequency, type (e.g. using transport) and/or intensity of outdoor mobility (independent of indoor mobility) was not described for any RCT included in this review. Two authors provided data on the extent to which outdoor mobility was achieved by participants in their RCT [27, 33]. Mangione indicated 83% of participants performed outdoor mobility during the intervention [27]. Cross-sectional data presented at the American Physical Therapy Association, Combined Sections Meeting in 2019 by Mangione et al. [36], indicated the proportion of participants in the larger trial by Magaziner et al. [33] of home exercise after hip fracture who performed outdoor mobility during the home-delivered physical therapy intervention was as follows: visit 3, 44% outdoor walking; visit 8, 57% outdoor walking; visit 16, 62% outdoor walking; visit 24, 63% outdoor walking; and visit 32, 56% outdoor walking (these are for different sample sizes and different seasons). The remaining RCTs did not detail the extent to which outdoor mobility as an intervention component was achieved by participants.

There was no evidence of publication bias for any of the meta-analyses.

Three RCTs selected falls-related self-efficacy (Falls Self-Efficacy Scale (Swedish version) [24, 25], Short Falls Self-Efficacy Scale [26], self-efficacy for walking [35]) as their primary study outcome. One RCT included falls-related self-efficacy as a secondary study outcome noting a between-group difference at 4-month follow-up favouring the intervention group (median (25th and 75th percentiles): intervention 90.5 (80.5, 98.0) comparison: 79.5 (40.0, 92.5)) [29]. Rehabilitation interventions which incorporated outdoor mobility were suggestive (confidence interval crosses null) of a small increase in falls-related self-efficacy at 1–3-month follow-up compared with routine care (standardised mean difference 0.25; 95% CI: − 0.29, 0.78) (Fig. 5). There was substantial heterogeneity in the analysis I² = 87%. On removal of the RCT by Ziden et al. [24], there was no between-group difference in falls-related self-efficacy at 1–3-month follow-up (standardised mean difference − 0.03; 95% CI: − 0.24, 0.18) and I² = 0%. At 12-month follow-up, there were no between-group differences in falls-related self-efficacy for the study by Resnick et al. [35]. Differences in the Falls Self-Efficacy Scale (Swedish version) observed by Ziden et al. persisted at 6 and 12-month follow-up [25].

We identified 12 papers for 11 RCTs which included outdoor mobility in their rehabilitation intervention for participants after hip fracture. There were methodological concerns related to unblinded participants, personnel, and outcome assessors, and a lack of precision in estimates across included RCTs. Our meta-analyses suggest interventions which include outdoor mobility may be beneficial in terms of outdoor walking and falls-related self-efficacy and not beneficial for walking endurance. However, the RCTs did not provide sufficient detail to replicate the intended outdoor mobility component. Furthermore, most RCTs did not provide detail on the extent to which the outdoor mobility component was actually achieved. Coupled with methodological concerns, we cannot determine the extent to which any potential benefit observed across RCTs may be attributed to an outdoor mobility intervention component.

The current review identified 7 additional RCTs not included in a previous review of home-based rehabilitation after hip fracture by Wu et al. [13]. We identified the same RCTs by Ziden et al. [24, 25] and Karlsson et al. [28] investigating the effectiveness of interventions on outdoor mobility. Wu et al. proposed no effect on walking outdoors based on these two studies [13]. We adopted a more conservative interpretation of the meta-analysis highlighting the conflicting evidence for effectiveness between the two studies. We also add to the findings of this earlier review by providing results from analysis of both walking endurance and falls-related self-efficacy.

A previous review by Heldmann and colleagues indicated outcome measure selection should be highly specific to the intervention components to reveal benefits attributable to rehabilitation in older patients [11]. For the current review, ambulatory ability and/or falls-related self-efficacy were selected as a primary outcome for 6 of the 11 RCTs identified suggesting outdoor mobility was a peripheral treatment component for half of included RCTs. Indeed, interventions included multiple treatment components many of which do not have a plausible mechanism for changing ambulatory ability or falls-related self-efficacy, e.g. wound care, mediation and nutrition [28]. In addition, most interventions targeted changes in body function/structures through, e.g. resistance training or flexibility (often supervised), as well as changes in activities or participation through (often unsupervised) indoor and outdoor mobility [26, 28, 30, 31, 33, 35]. The peripheral nature of outdoor mobility to these interventions may explain the lack of reported effectiveness on ambulatory ability and falls-related self-efficacy.

Potential benefits in falls-related self-efficacy and/or ambulatory ability were observed for interventions where outdoor mobility was a more central treatment component. The intervention by Ziden et al. focused explicitly on increasing outdoor mobility through physical activity, cognitive behavioural interventions and engagement of family in discharge planning [24, 25]. The alignment between intervention components and outcomes may explain the positive effect (in terms of earlier recovery of outdoor mobility and increased falls-related self-efficacy) observed compared with routine care [24, 25]. The aerobic training arm of the RCT by Mangione et al. was the only intervention to achieve a clinically meaningful (but not statistically significant) between-group difference for the 6-min walk test at the end of the intervention [27, 37]. The observed difference may be attributed to the relevance of the 6-min walk test to an intervention which focused on 20 min of indoor and outdoor walking (83% of participants performed outdoor mobility) at 65 to 75% of age-predicted maximal heart rate [27]. Whilst promising, these interventions were not without methodological concerns. Ziden et al. failed to blind outcome assessors to group allocation which may have led to overestimation of effectiveness [24, 25]. The RCT by Mangione et al. was small with 12 participants in the intervention group and 10 in the control group leading to a lack of precision in outcome estimates. It is therefore not possible to determine whether an intervention with outdoor mobility as a central component leads to benefits in ambulatory ability or falls-related self-efficacy after hip fracture.

Half of RCTs included in this review incorporated a psychological treatment component (goal setting and/or motivation) [24, 25, 29, 30, 34, 35]. Evidence from stroke and primary prevention supports a key role of psychological components in interventions targeting outdoor mobility. For patients post-stroke, a large UK multicentre trial, the ‘Getting Out of The House Study,’ saw a neutral effect of repeated practice of outdoor mobility on outcomes apart from potentially increasing the number of outdoor journeys (secondary study outcome) [38]. The authors noted that the benefit observed was dependent on the treating therapist — indicating a role of motivation and feedback [38]. This is in keeping with an implementation intervention in Australia which reported a beneficial effect of targeting the behaviour of community rehabilitation teams to deliver more outdoor journeys for people post-stroke on the proportion of people achieving outdoor mobility after the intervention [39]. An umbrella review of primary prevention interventions pointed to feedback as a core behaviour change treatment component for increasing physical activity among older adults [40]. For the current review, only one study incorporated objective feedback with the use of sensor output for unsupervised indoor mobility to inform coaching during the intervention [32]. This objective feedback was not extended to outdoor mobility and may be targeted in a future intervention study [32].

There is uncertainty over the external validity of many of the studies included in this review to the underlying population of patients with hip fracture. Most excluded potential participants with cognitive impairment (170 of 1868 (9%) potential participants, where reported) [24, 27, 29, 32, 34, 35], reflecting up to 30% of the underlying population [41]. Only one RCT included participants’ resident in nursing homes [28] where the incidence of hip fracture is high [42]. Moreover, the structure of community-based rehabilitation varies widely regionally, nationally, and internationally. Therefore, it cannot be certain whether results from Australia, Finland, Germany, the Netherlands, Sweden, and the USA may be generalizable to other contexts both within and across countries.