Effect of home-based exercise programs with e-devices on falls among community-dwelling older adults: a meta-analysis

Eight online databases were searched for published works which were published until May 2022, namely: PubMed, Cochrane, CINAHL, Scopus, Ovid MEDLINE, ScienceDirect, Web of Science and Embase. In addition, ongoing and unpublished trials were searched in the Cochrane Controlled Register of Trials. The complete search strategy is summarized in Supplementary Table 1. Later, a manual search of relevant randomized controlled trials (RCTs) from reference lists was carried out, producing three additional relevant records. All the mentioned articles were written in English.

The eligibility criteria were strictly established according to the PICOS (population, intervention, comparison, outcome and study design) framework (Table 1) [50]. The target population was older adults (aged at least 60 years) who could participate in home-based exercises after professional evaluation, living independently in communities [51]. Because of the differences in life expectancy, the definition of older adults varies from country to country. According to WHO, in most regions of the world the rates of fall-related fatalities are highest among adults over the age of 60 years [52]. Therefore we used community-dwelling adults aged 60 years or over as the target population. Interventions included HEPE alone or in combination with other multifactorial interventions. The content of the HEPE should be acceptable, feasible and safe for the elderly (including but not limited to the following: simple balance training, strength training with moderate resistance or modified comprehensive training). Meanwhile, taking the variable physical level of the elderly into consideration, the exercise programs should also be personalized or under remote supervision by professionals. Comparison included standard physiotherapy or traditional home-based exercise (including group-based exercise classes, telephone call-based exercises, home visit-based exercises or booklet-based exercises), usual medical care or no intervention. Eligible studies should contain at least one of the following outcomes: the risk or rate of falls, balance ability, lower extremity strength, cognition function, severity of falls and fear of falling. Study design should meet the standard of a RCT [53].

In this SR, the selection process was guided by the PRISMA flow diagram [50]. All articles identified from the searches were exported to EndNote v.20, and duplicates were removed automatically. The remaining articles were screened manually using the titles and abstracts. The verification of eligibility of the selected articles was done using the study titles and abstracts. Next, full texts were downloaded. Studies with inappropriate populations, interventions, comparisons and outcomes were excluded. Finally, the included studies were assessed on their suitability for meta-analysis. The number of studies included and excluded at each step and the reasons for exclusion are listed in the PRISMA diagram (Figure 1).

The quality assessment was carried out by two raters based on the Physiotherapy Evidence Database (PEDro) scale [54,55]. This is a rating scale to assess the methodological quality of clinical trials which consists of 11 items encompassing external validity. A total score is reached by adding the ratings of items 2 to 11 for a final score of 0–10. Its authors have suggested that a PEDro score of <4 should be considered ‘poor’, 4 or 5 should be considered ‘fair’, 6–8 should be considered ‘good’ and 9 or 10 can be considered ‘excellent’ [56]. A third rater then resolved any potential disagreement.

The following data from each study were extracted: basic information and publication details; participants’ characteristics; intervention characteristics; comparison characteristics; and outcome measurements. When there was a missing or inconsistent outcome because of the heterogeneity of the study designs, a narrative synthesis approach was performed to examine the results, rather than a meta-analysis. Dichotomous data (fall rate during the follow-up period) and quantitative data (including the result of timed up and go [TUG], five-times sit-to-stand test [FTSST] and physiological profile assessment [PPA]) extracted were then entered into Review Manager (v. 5.3; the Cochrane Collaboration, Oxford, UK) for meta-analysis.

All statistical analyses were performed with Review Manager v. 5.3. Heterogeneity was initially explored through visual inspection of the forest plots, then a test for statistical heterogeneity was carried out using the I² and χ² test. A p-value of < 0.05 was considered to be statistically significant. According to the percentage of I², values greater than 50% are considered to represent substantial heterogeneity. When this occurred, we attempted to explain the cause of the variation. If the value was less than 30%, the overall estimate was produced by a fixed-effects model. Otherwise, a random-effects model was used, for which the CIs are broader than those of a fixed-effects model [57]. Stability and publication bias were tested by a one-study leave-out sensitivity analysis.

A total of 4198 records were gathered from various databases, and three additional articles were found by hand-searching. After automatically removing the duplicates, 2231 records were kept. Then a screening according to the title and abstract was carried out by three independent authors, from which 2002 records were excluded. Finally, 229 records with corresponding full texts were further screened judging by the eligibility criteria; 12 studies (participants n = 1648) met all of the inclusion criteria and were included in the meta-analysis. The main reasons for exclusion are listed in the PRISMA diagram (Figure 1). The characteristics of the included studies are summarized in Table 2. The average age of the participants ranged from 70.22 to 85.2 years and all of them were community-dwelling older adults. The HEPEs were all personalized according to each participant’s fitness level by the professionals. The interventions required 2–7 days per week (or 60–120 min per week) and lasted from 6 weeks to 24 months. According to most of the exercise programs, 30 min per day is an average intervention intensity.

The e-devices used in the studies included mobile phones [44], tablets [63–65], personal computers [59,62], a combination of all [31,66,67] or a customized system [58,60,61].

The overall quality of the included studies could be considered as good (PEDro score: 6.83; range: 5–9; Table 3). None of the studies showed poor methodological quality. Only one study got 5 points, which should be considered as ‘fair’; besides that, ten studies had good quality and one study had excellent quality.

Among the 12 included studies, six reported the proportion of fallers, while the remaining six did not mention any information about the proportion of fallers or rate of falls in the end point measurements. Therefore, a total of 1242 participants were included in the meta-analysis (Figure 2). The proportion of fallers is defined as the ratio of fallers to the total number of participants. The data were generally collected during the interventions or the follow-up period through a self-report method. A lower risk ratio indicates a reduced rate of falls. The pooled effect of HEPEs in reducing the proportion of fallers is statistically significant (risk ratio [RR]: 0.82; 95% CI: 0.72–0.95; Z = 2.11; p = 0.006), favoring HEPE groups to control groups, as determined by the fixed-effects method. Meanwhile, the heterogeneity tests indicated that nonheterogeneity was found across the six studies (I² = 0%; χ² = 2.20; p = 0.82).

Seven studies utilized the TUG test to estimate the balance of the participants, which were pooled in meta-analysis by a random-effects model.

TUG is a simple and widely used test of overall functional mobility, assessing the individual’s ability to transfer, walk and change direction [68]. According to the results from TUG, a shorter time taken to complete the whole test indicates better balance and functional performance [69]; therefore we took TUG as the index of balance in the meta-analysis with 865 participants included (Figure 3A). The HEPE group showed an improvement trend in balance, but without statistical significance (MD: -0.71; 95% CI: -1.47 to 0.05; Z = 1.83; p = 0.07).

Three studies (Rosie et al. [58], Boongird et al. [62] and Hong et al. [63]) used the Berg balance scale (BBS) to measure balance in the elderly. The scale consists of 14 items, scored from 0 to 4, which are added to make a total score between 0 and 56 [70]. Evidence has shown that total scores below 45 are associated with a higher risk of falls; an individual with a history of falls and a total score below 51 is highly predictive of falls; a score of less than 40 is associated with a fall risk of almost 100% [71]. Rosie et al. reported that the HEPE group had a significant within-group improvement in BBS score (mean improvement 1.67 ± 2.64 points; p = 0.001) compared with the control group [58]. Hong et al. further proved a significant interaction effect between group and time regarding the scores of BBS (p = 0.03) [63]. According to Boongird et al., the result from BBS did not show a significant improvement for the HEPE group, but maintained a fair level compared with the control group (p = 0.258) [62]. Haines et al. utilized the Balance Outcome Measure for Elder Rehabilitation to test the balance, but no statistical significance was found in their study [59]. Shake et al. [64] used the Short Physical Performance Battery in the study, which is a well-established tool to assess lower extremity muscular strength, agility and dynamic balance [72]; the composite scores of the entire battery showed only a main effect of time (F_1,83 = 13.86; p < 0.001) [64].

Of note, two studies (Schoene et al. [60] and Li et al. [66]) conducted a dual-task TUG test as well; this is a modified TUG test combined with cognitive-related dual tasks. Previous research has shown that dual-task TUG may have an added value for fall prediction over TUG [73,74]. Schoene et al. found that the HEPE group improved significantly in the dual-task TUG test (F_31,1 = 4.226; p = 0.049) [60]. A similar result was also shown in another study (mean difference [MD]: -2.35; p < 0.001) [66].

Meanwhile, there was high heterogeneity across the seven studies included in the meta-analysis (I² = 80%; χ² = 30.50; p < 0.001). The heterogeneity could be caused by different ages of the participants or the intervention duration. Thus a subgroup analysis was conducted to explain the heterogeneity.

Six studies reported the changes of lower extremity strength by FTSST. Previous studies have proved the excellent reliability and validity of FTSST to assess sitting and standing performance and lower limb strength [75] in both healthy adults and those with pathologies [76,77]. The improvement of HEPE on lower extremity strength is significant (MD: -0.94; 95% CI: -1.71 to -0.47; Z = 3.93; p < 0.001) as determined by a random-effects model. The MDs ranged from -1.34 [64] to 0.40 [60]. A heterogeneity test suggested a fair heterogeneity across the studies (I² = 42%; χ² = 8.60; p = 0.13) (Figure 4).

Physiological fall risk was estimated using the PPA in four studies. PPA involves a series of simple tests of vision, peripheral sensation, muscle force, reaction time and postural sway, and is a useful tool in quantifying fall risks [78,79]. According to the meta-analysis results, there was no statistical significance (MD: -0.22; 95% CI: -0.46 to 0.01; Z = 1.90; p = 0.06) across the included studies with 731 participants involved, but the overall trend favored the HEPE groups. Meanwhile, a fair heterogeneity existed across the studies (I² = 48%; χ² = 5.73; p = 0.13) (Figure 5).

The fall risk could also be measured by fall efficacy, which is related to the level of fear of falling. According to previous studies, various kinds of fall efficacy scales have been developed and proved to be accurate in predicting physiological fall risks [80]. Rosie et al. reported the fall efficacy by the modified falls self-efficacy scale, from which no statistical difference was found either in-group (p = 0.393) or between-group (p = 0.818) [58]. Haines et al. measured the fall efficacy by the activities-specific balance confidence scale [59]. Similarly, a negative result was found, according to a linear regression analysis (regression coefficient = -0.30; 95% CI: -1.49 to 0.89; p = 0.61) [59]. Boongird et al. applied the Thai fall efficacy scale to measure the fall efficacy in their trial and found that fall efficacy improved significantly in the HEPE group at the 6-month follow-up (p = 0.003) [62]. According to Hong et al., there were no changes in falls efficacy in either activities of daily living (p = 0.55) or social activities (p = 0.51) by means of the Korean falls efficacy scale [63]. However, the additional measurement of fear of falling, assessed through a fear of falling questionnaire, was carried out in their study and revealed a significant interaction effect between group and time (p = 0.009). The HEPE group showed a better fall efficacy than the control group (p = 0.008).

The fall risk in the HEPE group was significantly decreased, by 18% (RR: 0.82; 95% CI: 0.72–0.95) compared with the controls according to the changes in the proportion of fallers. Notably, a previous review indicated that balance and functional exercises reduced the rate of falls by 24% (RR: 0.76; 95% CI: 0.70–0.81), while multiple types of exercise (most commonly balance and functional exercises plus resistance exercises) reduced the rate by 34% (RR: 0.66; 95% CI: 0.50–0.88) [51]. This result suggests that HEPEs may be more effective in reducing the fall rate if multiple types of exercise are applied.

Our findings suggested that the overall effect of HEPEs was not significant in improving balance as measured by TUG. A previous review indicated that elderly adults showed an age-related decline in balance and gait with increased gait variability and an associated increased risk of falls [81]. Considering that age is one of the factors that may affect the balance and intervention effect, a subgroup analysis was then carried out; we found that balance was significantly improved when the average age of involved participants was over 75. Of note, the traditional home-based interventions seemed to have limited impact on balance improvement in older adults (aged over 75). For example, a RCT enrolled 64 subjects whose average age was 75.9 years; after 15 weeks of home-based exercise, no significant improvement was found in balance as measured by TUG (p = 0.90) [82]. Therefore this is an exciting finding, because HEPE could be a more suitable intervention for older adults compared with the traditional home-based programs. On the one hand, the HEPEs took advantage of the better adherence brought by e-devices. As previous studies have shown, 67% of participants using the digital program continued to exercise regularly compared with 35% for the paper booklet (p = 0.036) [83]. Similarly, a RCT comparing adherence to laboratory-based trials using a supervised e-device versus traditional home-based trials yielded similar results [39]. Therefore a digital program was able to facilitate long-term maintenance of regular exercise so as to bring about a better curative effect on balance. On the other hand, as all the HEPEs involved were individually tailored, the exercise programs fitted the functional level of the elderly well. Evidence has shown that tailored exercise programs cause significant improvements in balance compared with fixed ones [84]. For example, in the study of Rosie et al., the results showed a significant within-group improvement in balance in the HEPE group (p = 0.001), while no change was revealed in the control group performing knee extension (p = 0.258) [58]. This aligned with another SR which found that individualized home-based exercise programs promoted autonomy and were therefore seen as superior to generic fall-prevention programs among participants at an average age of 80.1 years [85]. The mentioned two reasons might explain why the HEPEs could significantly improve balance in people over 75 years of age.

The result of another subgroup analysis based on different intervention durations suggested that >16 weeks was an appropriate intervention duration, as a significant improvement in balance was found compared with the subgroup training for less than 16 weeks. A SR aimed at dose–response relationships in balance training in older adults illustrated that balance training lasting for less than 11 weeks resulted in lower effects on balance performance [86]. Another previous review which examined the efficacy of exercise to reduce falls recommended at least 2 h of training per week for a training period of more than 6 months to reduce falls [87]. In our SR, the average amount of exercise was 121.25 min per week, which was in accordance with the analysis by Sherrington et al. [87]. These findings might indicate that an intervention period for HEPE of more than 16 weeks could be more effective in improving overall balance performance.

Muscle strength (especially in the lower extremities) is one of the factors that are closely related to falls in older adults [88]. The result found that lower extremity strength was significantly improved in HEPE groups as measured by FTSST. The positive finding might be because most of the HEPEs consisted of strength or functional training. Almost all of the studies delivered individually tailored muscle strength training, including stepping or stretching exercises that would increase in difficulty over time, as recommended by Gschwind et al. and Shake et al. [31,60,61,64,65]. A more comprehensive program including vestibular rehabilitation, core and limb strengthening exercises, joint motion range exercises and balance exercises was used in the study of Yi et al. [44]. Therefore, in order to achieve better results in improving lower limb strength, multimodal exercise programs were recommended in HEPEs.

Conversely, the effect of HEPEs on fall risk as measured by PPA was not significant. The between-trial heterogeneity (I² = 48%) could be explained by a single study from Gschwind et al. [31]. The heterogeneity reduced to 28% (χ² = 2.79; p = 0.25) when this trial was removed from the meta-analysis, and the overall effect of HEPEs on fall risks was then significant (MD: -0.32: 95% CI: -0.58 to -0.07; Z = 2.46; p = 0.01). According to Gschwind et al., a low adherence rate may explain the limited improvement of PPA. There was a significant three-way interaction for fall risk assessed by the PPA in a subgroup analysis between the high-adherence (>90 min of exercise per week), low-adherence (<90 min of exercise per week) and control groups (p = 0.044) [31]. In addition, two-way analysis comparing the high-adherence group with the control group revealed a significantly larger effect in favor of the high-adherence group for fall risk (p = 0.031) [31]. This result was in accordance with our finding that only a certain amount of exercise (at least 120 min per week for over 16 weeks) can achieve significant improvement in physical behavior. Our finding was better supported by the following studies which validated the dose–response relationship between fall risk and fall effectiveness. Haines et al. found a negative result of fall efficacy as measured by the activity-specific balance confidence scale after only 8 weeks of intervention (regression coefficient: -0.30; 95% CI: -1.49 to 0.89; p = 0.61) [59]. Contrary to this, another RCT suggested that a 24-week home-based tele-rehabilitation program could significantly reduce the risk of falls (RR: 0.60; 95% CI: 0.44–0.83; p < 0.001) [89].

Meanwhile, it was important to know whether exercise-induced overall physical improvement was stable over time. Evidence indicated that the fall prevention program with exercise intervention improved functional performance at 3 months in community-dwelling older adults with risk of falls, but did not reduce falls at 1-year follow-up [90]. Arkkukangas et al. conducted a trial of 12 weeks of fall prevention exercise for community-dwelling older adults, in which a significant within-group improvement was found in physical performance (p = 0.04), fall efficacy (p = 0.02) and physical activity level (p = 0.02) according to a short-term follow-up [91]. However, the positive result was not reproduced in their 2-year follow-up study [92]. Likewise, Naseri et al. found a similar time–efficacy relationship at 6-month follow-up [93]. With reference to the mentioned studies, we suggest conducting HEPEs on a permanent basis to counteract time-related declines in overall physical performance. In addition to applying different kinds of e-devices, the following improvements were adopted by the included studies to enable better long-term adherence: the exercise programs were modified and supplemented to suit the older adults in seven studies [31,44,58,61,63,64,66]; two studies provided a video of the exercises [62,67]; and three studies provided remote guidance for any possible problems [59,60,65].

Of note, the positive changes in dual-task TUG were also enlightening. As dual-task TUG is a cognitive-related test [74], this might mean that HEPEs could improve cognitive function in older adults. In addition to the studies of Schoene et al. and Li et al. [60,66], Shake et al. also paid attention to cognitive changes; these results suggested a significant main effect of group exercise (p < 0.05) and an interaction effect (p = 0.02) [64]. However, a review suggested that physical activity or exercise alone was able to improve the attentional control so as to help improve cognition in older at-risk individuals [94]. Likewise, a meta-analysis among older adults also showed that exercise training had a significant benefit for cognitive function (p < 0.001) [95]. Therefore the improvements associated with dual-task TUG or cognitive tasks were mostly attributed to the exercise itself. Whether the intervention of e-devices could improve cognitive function needed further experimental proof. So any evidence of potential benefits from HEPEs should be interpreted cautiously, given the limited evidence of changes to cognitive function.

This review, to our knowledge, is the first SR and meta-analysis examining the effectiveness of HEPEs on fall prevention among community-dwelling older adults. Especially in the background of COVID-19, this cost-friendly and contactless intervention is worthy of being studied and popularized. Meanwhile, intervention using e-devices has solved the problem of poor adherence of home-based exercises to a great extent. A previous review proved that telehealth combined with e-devices was able to reduce fall risk and fall efficacy significantly [40], and evidence also showed that individualized home-based exercise programs can lead to significant improvements in physical activity, balance, mobility and muscle strength for older people [85]. In the review we focused on home-based exercise programs delivered by e-devices; we not only proved the effectiveness of HEPEs for fall prevention, but also reached a quantitative conclusion on the optimum exercise volume when using a HEPE. The specific volume of exercise can help healthcare providers or physiotherapist to make an informed decision for community-dwelling older adults in relation to falls. Furthermore, in addition to qualitative and quantitative perspectives, we compared HEPEs with traditional home-based exercise programs and face-to-face exercise programs in terms of fall rates. Last but not least, our findings were based on 12 studies that can be considered, on average, of good quality (PEDro score = 6.83).

This review is not without limitations. First of all, the small number of available studies and the lack of a consistent set of tests make our conclusions preliminary in essence. In fact, the sample sizes of the included studies were uneven, ranging from 30 to 503. As such, the results had smaller effect sizes and may not be representative of larger populations. Secondly, the review excluded older adults with a diagnosis of chronic diseases such as multiple sclerosis, cancer, Parkinson’s disease or Alzheimer’s disease, because these patients may have a different performance in terms of balance [96], strength [97], mobility [98] or cognitive ability [99], which increases the risk of falls and restricts the effectiveness of HEPEs. Therefore future studies should consider investigating the effects of HEPEs on older adults with different chronic diseases. Thirdly, most of the included studies were conducted in Asia and Australia; the result might be different in other regions, as many factors have been proved to influence fall risk, including ethnicity and health status [100]. Finally, almost all the studies were from developed countries, so the results might not be replicable in nondeveloped countries or remote areas.

In terms of implications for practice, the intervention pattern of a HEPE should be personalized and contain multiple types of exercise (including balance, strength training and functional training) to achieve a better curative effect. Next, the duration of intervention should be more than 16 weeks. Besides that, a permanent exercise plan is recommended in order to counteract time-related declines in overall physical performance. The frequency of HEPE-based exercise should be at least 120 min per week. Last but not least, step-by-step instructions and important advice regarding the method of application of the e-devices are necessary to increase the acceptance of new technologies among the elderly [101].

In terms of implications for research, as we did not find significant positive results from HEPEs on cognitive performance, future studies could pay attention to the special effects of HEPEs on cognition. Furthermore, multiple test methods should be applied in any future study in order to reduce the deviation of results caused by single tests and reinforce the credibility of the results.