Predicting falls in older adults: an umbrella review of instruments assessing gait, balance, and functional mobility

1Over one-third of adults aged 65 years and older fall at least once a year. Increasing age, frailty, comorbidity, impaired gait, muscle weakness, and impaired balance all contribute to the risk of falls. Falls are a major cause of disability and constitute the leading cause of injury-related mortality in people aged above 75 years. The importance of an individualised approach to screening, assessment, and intervention is emphasised across professional guidelines such as the Steadi Algorithm. There is no clear consensus on the specific choice of fall assessment; however, professional guidelines state that adults at high risk should be able to access individually tailored multifactorial measures based on a comprehensive assessment. This should include assessment of gait, balance, and motor function with targeted interventions to address any limitation since these domains are associated with an increased risk of falls. Assessing these limitations could help to identify older adults at risk of falling and allow targeted intervention to reduce this risk.

2Multiple approaches to assess gait, balance, and functional mobility have been developed including the Berg Balance Scale (BBS), the Timed Up and Go (TUG) test, and gait speed testing, such as the dual-task gait test. Although widely used across clinical practice, there appears to be little standardisation and difficulty determining the most appropriate tool. Systematic reviews of individual tools have provided limited and conflicting evidence for a tool’s predictive ability, thus precluding the ability to make clear clinical recommendations. To this end, we performed an umbrella review to synthesize the findings across multiple systematic reviews to help develop recommendations for clinical practice.

3The aim of this umbrella review was to systematically review, critically appraise, and summarize the existing reviews on the use of assessment tools of gait, balance, and functional mobility to predict falls in older adults or distinguish fallers from non-fallers. This review is part of a larger initiative on behalf of the task force on global guidelines for falls in older adults (details available at https://worldfallsguidelines.com/). This paper presents a summary of the umbrella review for Working Group 1, and the findings will be fed into a wider consensus development process to develop key recommendations in the assessment and management of falls for older adults.

4This umbrella review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) and the protocol was previously registered on PROSPERO’s international online register of systematic-, rapid-, and umbrella reviews (PROSPERO CRD42020225101).

5The electronic academic databases MEDLINE, PsychINFO, Embase, and the Cochrane database for Systematic Reviews were searched from inception to November 23rd, 2020. The searches search were then updated on April 20th, 2022. To ensure a broad review of available literature, no restrictions on publication date were applied. A comprehensive search strategy was developed with the support of a research librarian using a combination of medical subject heading (MeSH) terms and key words for the concepts of older adults, gait, balance, and functional mobility assessments, and falls prediction. Only studies in English were included. The full search strategy for MEDLINE is presented in Additional file 1 at the end of this document and this strategy was adapted for each of the included databases. The reference lists of included papers were also reviewed to identify any further relevant reviews for inclusion.

6We included the following types of review studies:Narrative reviews, defined as reviews that may or may not present a systematic synthesis of findings from all individual studies included;Systematic reviews without meta-analysis, defined as having an explicit reproducible methodology including a systematic search that aims to identify all studies that meet pre-specified eligibility criteria followed by a systematic presentation and synthesis of the findings of all included studies;Systematic reviews with meta-analysis, defined as systematic reviews using statistical techniques to combine and summarize the results of multiple studies.

7Narrative reviews, defined as reviews that may or may not present a systematic synthesis of findings from all individual studies included;

8Systematic reviews without meta-analysis, defined as having an explicit reproducible methodology including a systematic search that aims to identify all studies that meet pre-specified eligibility criteria followed by a systematic presentation and synthesis of the findings of all included studies;

9Systematic reviews with meta-analysis, defined as systematic reviews using statistical techniques to combine and summarize the results of multiple studies.

10We excluded the following types of studies: conference abstracts, student theses, books, book chapters, and papers reporting empirical data from a single study rather than reviewing more than one study. Reviews which included technology-based instruments only were excluded, as there is another on-going systematic review on this topic from Working Group 8 of the task force on global guidelines for falls in older adults (PROSPERO CRD42021241177).

11We included reviews of empirical studies in older adults (women and/or men), aged 60 years or older, in any setting. Specifically, we included reviews in all the following settings: the community, and primary and secondary care settings, including long-term care institutions, rehabilitation, and acute hospital settings. We also included reviews that presented data from various age groups in case they presented data on a subgroup of older adults aged 60 years or above separately. Following this, we excluded reviews examining individuals exclusively younger than 60 years of age.

12Reviews that included validated assessments of gait, balance, and functional mobility to predict falls or to distinguish fallers from non-fallers.

13Our primary outcome of interest was the prediction of falls. Secondary outcomes were as follows: reliability, validity including sensitivity, specificity, feasibility, and cost of the assessments.

14Two reviewers (KR, DBJ) independently screened titles and abstracts of all records for eligibility, using the online software package Rayyan (https://www.rayyan.ai/). Disagreements were resolved by the assessment of a third reviewer (GO). Full text articles were retrieved and screened independently by two reviewers (KR, DBJ) with disagreements resolved by the assessment of a third reviewer (GO).

15Three reviewers (KR, DBJ, GO) extracted the data by using a pre-defined data extraction form developed specifically for this review. The following data were extracted:Review details: author(s), year of publication, country of lead author, type of participants, review objective, number of participants, age range of participants, mean age of participants, and proportion of women.Search details: sources searched, type of analysis (narrative review, systematic review without meta-analysis, or systematic review with meta-analysis), number of studies included in the review, design of studies included, and countries in which included studies were conducted.Critical appraisal: date range of included studies, critical appraisal tool(s) used in the review, and critical appraisal score.Gait, balance, and functional mobility tests assessed: fall prediction outcome, measurement of falls, predictive ability, reliability, validity (specificity, and sensitivity).Cost: any cost analysis conducted.

16Review details: author(s), year of publication, country of lead author, type of participants, review objective, number of participants, age range of participants, mean age of participants, and proportion of women.

17Search details: sources searched, type of analysis (narrative review, systematic review without meta-analysis, or systematic review with meta-analysis), number of studies included in the review, design of studies included, and countries in which included studies were conducted.

18Critical appraisal: date range of included studies, critical appraisal tool(s) used in the review, and critical appraisal score.

19Gait, balance, and functional mobility tests assessed: fall prediction outcome, measurement of falls, predictive ability, reliability, validity (specificity, and sensitivity).

20Cost: any cost analysis conducted.

21Three reviewers (KR, DBJ, GO) assessed the risk of bias of the included studies using the Risk of Bias Assessment Tool for Systematic Reviews (ROBIS). ROBIS assesses four domains: 1) study eligibility criteria; 2) identification and selection of studies; 3) data collection and study appraisal; and 4) synthesis and findings.

22To provide key clinical and research recommendations on assessment tools for fall prevention, the findings were synthesised for the most commonly reported gait, balance, and functional mobility assessments. Due to the heterogeneity of the reviews with regards to participant characteristics, settings, and assessment protocols, it was not appropriate to conduct a meta-analysis. A narrative synthesis was conducted for each gait, balance, and functional mobility assessment that was reported by more than two review studies. The narrative synthesis was conducted based on the review type and quality, as well as the number of reviews addressing this assessment and the key findings. For each review, the results were interpreted to indicate whether the findings in relation to the assessment tool’s predictive ability for falls were favourable, not favourable, inconsistent, or unclear (if data could not be extracted). An overall summary for each assessment was then made based on the highest quality available evidence. The synthesis is presented in tabular format; in the tables, the studies are ordered based on their quality.

23The literature search identified a total of 2736 potentially relevant records. Of these, 543 were duplicates. The titles and abstracts of the remaining 2213 records were screened. After excluding 2092 items in the screening, the full texts of 121 articles were assessed for eligibility. After excluding further 90 records (50 were not review papers; 18 did not assess falls; 9 were technology-based instruments only; 7 were duplicate records; 5 were not in older adults; and 1 was not in English), we included 31 records in our analyses. Figure 1 at the end of this document shows the PRISMA flow-chart.

24Table 1 presents a summary of the 31 included review studies. Three were categorised as narrative reviews, 17 as systematic reviews without meta-analysis, and 11 were systematic reviews with meta-analyses. Nine reviews reported on community dwelling older adults only, one reported on long term care settings only, one reported on emergency department settings only, and 13 reported studies across a range of settings including community, supported living, residential care, outpatient and inpatient settings. Four reviews provided no details on settings. Three reviews reported that they included older adults with cognitive impairment. Healthy community-dwelling older people were the primary focus of reviews however older people with neurological disorders were included in one review, older people receiving inpatient stroke rehabilitation were included in one review, and older people being assessed in the emergency department was the focus of one review.

25Of the 31 included reviews, ten were globally deemed at low risk of bias, eight at unclear risk of bias, and 13 at high risk of bias (Table 2). Areas of high or unclear risk of bias primarily related to limiting searches with language restrictions, selection and data extraction not done in duplicate, and a lack of quality appraisal of the individual studies.

26The most frequently reported gait, balance, and functional assessments for falls prediction included the following tests: TUG, BBS, tests of gait speed, dual task assessments, single leg stance, Functional Reach Test (FRT), tandem gait and the chair stand test.

27The TUG consists of a combination of standing from a chair and walking 3 m, turning and returning to sitting. The TUG test was reported in thirteen reviews (Table 3). Three reviews demonstrated favourable findings, four reviews reported unclear or inconsistent findings, five reviews demonstrated not favourable findings, and one review reported no extractable data on TUG’s ability to predict falls. Across all review studies, the evidence was inconsistent on the ability of the TUG to predict falls. There is some evidence, however, from some subgroup analysis that the TUG may have a role in fall prediction for the lower functioning older adult population.

28The BBS is a balance test with a series of 14 balance tasks that assess a person’s ability to safely balance. Tasks include sitting-to-standing, turning 360 degrees and standing on one leg. The BBS was reported in nine review papers (Table 4).

29Three reviews demonstrated favourable findings, one review reported inconsistent findings, and four reviews demonstrated not favourable findings, on the BBS ability to predict falls. One review did not report any results to extract. Across all the review papers, the evidence for using the BBS to predict falls was inconsistent, and based on the best available evidence, the use of the BBS as a balance assessment used in isolation is not recommended to predict falls. There was some evidence from one review that the BBS may have a predictive role in a stroke clinic population.

30Gait speed is the measurement of the time it takes to complete a walk over a given distance in the participant’s preferred or maximum pace and was reported in ten review papers (Table 5). Seven reviews demonstrated positive findings. One reported low sensitivity on the ability of gait speed to predict falls in community dwelling older adults, and one reported that a timed walk was not an independent predictor of falls in long term care settings. One review reported that gait speed did not predict falls in cognitive impaired older adults, however a subgroup analysis showed evidence for gait speed predicting falls. One review reported no data to extract. Different distances were used across the studies including 4, 6, 10, and up to 25 m distances. Two reviews investigated usual gait speed. One review reported mainly preferred walking speed. One review reported that of the eight studies that assessed gait speed, six found slow gait speed under standard conditions to predict falls.

31Details on the gait speed protocol was lacking in three reviews. The best available evidence suggested that gait speed was a useful measure in predicting falls in community dwelling older adults.

32Dual task assessments are the combination of a physical task (such as walking) and either a second physical task (such as holding an object) or a cognitive task (such as counting) and was reported in seven review papers (Table 6). In detail, four reviews demonstrated favourable findings, two review reported unclear findings, and one review demonstrated not favourable findings on the ability of dual task testing to predict falls. Evidence for the ability of dual task testing to predict falls over single balance tests was inconsistent; however, the best available evidence suggested that dual task testing had the ability to predict falls. The optimal type of dual task test is still unclear.

33The single leg stance test is a single leg standing balance test and was reported in five reviews (Table 7). One review reported favourable findings. Three reviews reported unclear findings on its ability to predict falls and one review demonstrated not favourable findings. Overall, the evidence was inconsistent for the ability of the single leg stance to predict falls.

34The Functional Reach Test is a functional balance test and was reported in nine review papers (Table 8). Six review papers demonstrated favourable findings, and three reported not favourable findings on the ability of the Functional Reach Test to predict falls. The evidence across all the reviews was inconsistent for the predictive ability of the Functional Reach Test.

35The Tinetti test and the POMA test are task-oriented balance tests and were reported in eight review papers (Table 9). Two review papers demonstrated positive findings, five review papers reported unclear findings, and one review paper reported not favourable findings on the ability of the Tinetti test or POMA to predict falls. There were inconsistent findings across all the reviews on the predictive ability of the Tinetti and POMA test.

36Tandem gait and stance is a standing balance test and a heel to toe walking test and was reported in eight review papers (Table 10). One review paper demonstrated favourable findings in tandem stand. Two review papers concluded that tandem walk was a significant predictor of falls, and one review demonstrated that only tandem walk had the ability to predict falls. Five review papers reported unclear associations, and one review reported that the test did not predict falls. The findings across the reviews were inconsistent on the ability of the tandem gait to predict falls. However, tandem walk showed promising results in selecting the population in need of a further evaluation [ref].

37The CST measures the ability to get up from chair without using arms, time taken to get up five times, or number of chair stands over 30 seconds, and was reported in five review papers (Table 11). One review paper demonstrated favourable results, 3 papers reported unclear results,, and one review reported inconsistent findings on the ability of the CST to predict falls. Overall, the evidence was inconsistent for the ability of CST to predict falls.

38This umbrella review aimed to systematically and critically appraise the evidence on gait, balance, and functional mobility assessments used to predict falls for older adults. A total of 31 review papers were identified, which were mainly systematic reviews without meta-analysis and of low quality with high risk of bias. There were inconsistencies in the findings across all the review papers. The present umbrella review determined that there is not one single gait, balance, and functional mobility assessment that can be used in isolation to predict falls in community-dwelling older adults. The TUG was the most frequently assessed single test for falls prediction, but the findings were inconsistent in its ability to predict falls. There is, however, favourable evidence to suggest that gait speed can be useful in predicting falls and might be included as part of a comprehensive evaluation for older adults. Some positive results were found in dual task assessment as predictors of falls.

39Clinical practice guidelines recommend multifactorial interventions to prevent falls in community dwelling older adults who are at an increased risk of falls. Such interventions contain an initial assessment of risk factors for falls and subsequent customised interventions for each patient based on risk factors. The inconsistencies reported across the included review papers highlight the importance of making a clinical judgement including risk factors such as previous falls, cognitive impairment, comorbidity, polypharmacy, activities of daily living, psychological factors, vision impairment, cognitive impairment, and footwear.

40Clinicians are encouraged to consider individual needs and contexts when evaluating falls risk in older adults. The inconsistencies reported across the review papers of the present umbrella review may have been influenced by the wide range of settings and clinical characteristics included in the individual studies. It is thus challenging to make recommendations for specific settings using the evidence from this review, in light of the degree of heterogeneity across the evidence available. Based on the evidence from this review, we are unable to recommend using the Timed Up and Go, Berg Balance Scale, Chair Stand Test, One Leg Stand, or Functional Reach, alone as single tests for the prediction of falls in older adults. We acknowledge however that these tests have value in assessing mobility and balance limitations and in identifying appropriate targeted interventions.

41In post stroke patients, one review reported positive results on the BBS for fall prediction, whereas the Functional Reach Test showed positive results in populations with cognitive impairments.

42Gait speed appeared most promising in fall prediction and has also been associated with other important outcomes like survival and functional capacity. Gait speed is a simple measurement, with no need for expensive equipment and can be performed quickly. The favourable findings in this review indicate gait speed is feasible to complete for community-dwelling older people and older outpatients of stroke clinics. However, gait speed should be assessed through a clearly defined protocol, which specifies the distances to walk or the time allocated to walking, and whether participants walk at their usual speed or maximum speed. One review suggested that the assessment at usual pace gait speed over 4 m might represent a highly reliable instrument to be implemented. Given the number of older people who could benefit from fall risk assessment, an inexpensive assessment tool that can be used in different settings is appealing.

43Dual task assessments showed promise in its ability to predict falls, with some evidence suggesting that it was a better predictor of falls than single task assessments. But importantly, differences in testing protocols could have influenced the results, warranting future research with standardised protocols to allow further synthesis of this finding.

44The findings from the present umbrella review demonstrate that it is feasible to complete an assessment of gait speed in older adults across a range of settings including the community, long-term care institutions, and rehabilitation settings. Based on the assessment of the falls risk, it is important that interventions are offered to reduce this risk. Exercise programmes have been demonstrated to reduce the rate of falls, particularly for community-dwelling older adults; the most effective programmes include balance and functional mobility exercises.

45The TUG was the most frequently reported assessment for falls prediction. The TUG is a simple and low-cost test that is easy to administer and has been previously recommended in clinical practice guidelines including the guidelines posited by the American Geriatrics Society/British Geriatrics Society (AGS/BGS). However, this umbrella review demonstrated that its fall predictive ability was inconsistent. This inconsistency may be explained by heterogeneity in the settings and populations studied, the use of different cut-off times, and the mixed quality of the evidence. One review suggested that the TUG may have a role in predicting falls in lower functioning or institutionalized older adults.

46To the authors’ knowledge, this is the first umbrella review examining gait, balance, and functional mobility assessments in the prediction of falls for older adults. Further strengths of this umbrella review included following PRISMA reporting guidelines, conducting a comprehensive search for evidence with the support of a research librarian, and clearly stating the objectives beforehand and ensuring transparency with the published protocol. Furthermore, the selection of the included studies was performed in duplicate with a third reviewer resolving any conflicts. The reviewers assessed the quality of the included studies in pairs, using a quality assessment tool designed specifically to assess the risk of bias in systematic reviews, and differences were discussed and resolved between reviewers.

47However, this umbrella review has some limitations. The included studies were too heterogeneous to allow for direct comparison of results in a united meta-analysis. In addition, studies involved both prospectively and retrospectively reported falls, which might have contributed to some of the heterogeneity. Also, many of the included review studies were considered to have a high or unclear risk of bias with a lack of clear reporting. This limited the data that could be extracted and synthesised. The differences in the included studies were not statistically assessed, following advice from a statistician that it was not possible to do so, due to heterogeneity between studies. We excluded review papers that were not available in English due to the resources available for the review, therefore, it is possible that the language restriction in the selection of the included studies may have affected the results by introducing a risk of selection bias. We chose to exclude grey literature (e.g., papers that are not published in peer-reviewed journals) to ensure a certain level of quality in the included studies. In this umbrella review no differentiation between falls, multiple fallers or injurious falls were made.

48It was not possible from the umbrella review to provide guidance on the critical level of performance in gait speed associated with higher risk of falling. The optimal cut-off in gait speed to predict falls has not been universally defined and accepted, although different cut-offs (eg 1 m/s, 0.8 m/s, 0.6 m/s) have been associated with various adverse health outcomes, including falls. Based on a systematic literature review, an International Academy on Nutrition and Ageing (IANA) expert panel advised to assess GS at usual pace over 4 m and to use the easy-to-remember cut-off point of 0.8 m/s to predict the risk of adverse outcomes.

49The Short Physical Performance Battery (SPPB) was included in the search terms for this review; however, it was only reported in one review, limiting the ability to draw conclusions about its predictive ability in this review. In 2018, the SPPB has been accepted by the European Medical Agency as an assessment tool to assess frailty. In addition, as the SPPB is widely used in geriatric and other medical fields, it is gaining importance. Single studies reported mixed results with regards to fall prediction. Therefore, future research investigating its ability to predict falls and injurious falls is urgently needed.

50Further studies are required to investigate the applicability and validity of fall risk assessment tests in different populations with varying functional levels. Different frailty or intrinsic capacity status may influence the fall predictive ability of these tests. Older frailer adults with cognitive or physical impairment may not be able to perform hazardous tasks or follow complex instructions. Low resource settings may lack the equipment and trained staff to perform the more sophisticated tests, despite their potential effectiveness.

51We acknowledge that assessing gait, balance, and functional mobility may form only one part of an assessment of falls prediction. Falls prediction approaches may need to further account for the multifactorial nature of falls and the extensive list of factors that can contribute to the risk of falling. The development of multifactorial falls prediction models is an area of ongoing research; however, further work is required before their widespread use is advocated.

52Overall, there is not one single gait, balance, and functional mobility assessment alone that can be used in isolation to predict fall risk in community-dwelling older adults. The best available evidence suggests that gait speed is a useful measure in predicting falls and should be considered as part of a comprehensive evaluation of fall risk for older adults. We found that dual task assessments demonstrate some potential to predict falls in older adults, warranting further research in this area.