Article Text
PDF
Abstract
Objective Although several falls risk assessment tools are available, it is unclear which have been validated and which would be most suitable for primary care practices. This systematic review aims to identify the most suitable falls risk assessment tool for the primary care setting (ie, requires limited time, no expensive equipment and no additional space) and that has good predictive performance in the assessment of falls risk among older people living independently.
Design A systematic review based on prospective studies.
Methods An extensive search was conducted in the following databases: PubMed, Embase, CINAHL, Cochrane and PsycINFO. Tools were excluded if they required expensive and/or advanced software that is not usually available in primary care units and if they had not been validated in at least three different studies. Of 2492 articles published between January 2000 and July 2020, 27 were included.
Results Six falls risk assessment tools were identified: Timed Up and Go (TUG) test, Gait Speed test, Berg Balance Scale, Performance Oriented Mobility Assessment, Functional Reach test and falls history. Most articles reported area under the curve (AUC) values ranging from 0.5 to 0.7 for these tools. Sensitivity and specificity varied substantially across studies (eg, TUG, sensitivity:10%–83.3%, specificity:28.4%–96.6%).
Conclusions Given that none of the falls risk assessment tools had sufficient predictive performance (AUC <0.7), other ways of assessing high falls risk among independently living older people in primary care should be investigated. For now, the most suitable way to assess falls risk in the primary care setting appears to involve asking patients about their falls history. Compared with the other five tools, the falls history requires the least amount of time, no expensive equipment, no training and no spatial adjustments. The clinical judgement of healthcare professionals continues to be most important, as it enables the identification of high falls risk even for patients with no falls history.
Trial registraion number The Netherlands Trial Register, NL7917; Pre-results.
- general medicine (see internal medicine)
- primary care
- preventive medicine
- public health
Data availability statement
Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplemental information. All data relevant to the study are included in the article or uploaded as supplementary information. Additional data are available upon reasonable request from the first author.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
This review is built on extensive literature regarding falls risk assessment tools that are suitable for the primary care setting and representations of their predictive performance.
We endeavoured to reduce bias by only including falls risk assessment tools that have been validated at least three times in different studies and by assessing the risk of bias.
Given that different studies used different cut-off scores, addressed modified versions of the same tools and presented different outcome measures, it was difficult to combine the results and reach a convincing conclusion.
Introduction
Worldwide, falls are the second leading cause of accidental or unintentional injury deaths.1 On average, one of every three people aged 65 years or older falls at least once a year,2 and an estimated 646 000 people die each year due to the consequences of falls.1 These numbers are increasing as society ages.3 The consequences of falls can range from scratches or bruises to hip fractures, brain injuries or even death.4 5 Falls can have a major, long-lasting negative impact on the quality of life and self-management of older people.4–6 The treatment and rehabilitation of falls incidences are correlated with high costs in the healthcare sector.5 7 Therefore, the provision of falls prevention is important for older people.
Society is ageing, and older people are living independently at home for longer.3 The first point of contact for health problems is the general practitioner (GP). The approaches adopted by GPs vary, with some providing no falls prevention care at all, while others are quite active with regard to falls prevention. Given that only 20% of all older patients inform their GPs about their falls, GPs are unaware of the occurrence of 80% of the falls among their patients8 9 and they are thus likely not to know which of their patients are at risk of falls. This situation results in a delay or lack of treatment for falls risk among older people, despite the availability of potentially effective falls prevention interventions.10–14
The early identification of high falls risk among older people is a prerequisite to providing adequate care in time to reduce the risk of falls. Many tools are available for assessing falls risk, including the Timed Up and Go (TUG) test, the Tinetti Balance, the Berg Balance Scale (BBS) and the American Geriatrics Society/British Geriatrics Society guidelines for clinical practice. In a previous review, Gates et al summarise the accuracy of tools for predicting the risk of falling among older adults living in communities. They conclude that there is insufficient evidence to show that any instrument was adequate for predicting falls and they neither report nor consider implications for practice. It thus remains unclear which falls risk assessment tools have good predictive performance and might be suitable for practice.
The high workload associated with primary care places constraints on the time of practitioners.15 16 They also have limited resources for expensive equipment (eg, platforms, sensors), and their practices generally have little space.17–20 A suitable falls risk assessment tool for primary care settings should therefore require limited time, no expensive equipment and no space adjustments. This systematic review aims to identify falls risk assessment tools that are the most suitable for primary care (ie, quick (<5 min), no expensive equipment or specific resources required) and that have demonstrated good predictive performance in assessing the risk of falls among older people living independently. In this study, an assessment tool is understood as a tool that defines the nature of a specific problem: whether a patient does or does not have a high risk of falls.21 No additional assessment is required to identify high or low falls risk. Additional assessment is needed only to explore which intervention is needed to reduce a patient’s risk of falls.
Methods
Study selection
A systematic literature search was conducted in the following databases: PubMed, Embase, CINAHL, Cochrane and PsycINFO, using the search keywords presented in figure 1 (see online supplemental additional file 1). Medical Subject Headings (MeSH) terms were used when possible. Additional articles were included after snowballing. The flowchart for the literature search is displayed in figure 2.
Supplemental material
Search keywords.
Flowchart for the literature search.
Eligibility criteria and study selection
The proportion of older people is increasing, and the current population of older people is ageing differently than was the case 20 years ago (eg, people are becoming older and are more vulnerable to chronic diseases).22 23 Given the importance of validating suitable falls risk assessment tools in the current population of older people, the review included articles published between January 2000 and July 2020 that met the criteria for inclusion (as presented in figure 3).
Eligibility criteria.
This review includes only prospective studies, thus making it possible to summarise the predictive performance of falls risk assessment tools.24 In addition, our final analysis includes only tools that have been assessed in at least three different studies. This was done in order to ensure the validity of the tools that were included, as studies are likely to differ (eg, in terms of the age, sex or frailty of the selected population).
The first round of exclusion based on title was performed by WMAM. All articles from the second round of exclusion based on abstract were reviewed by WMAM. In addition, JCK, CJL and IAMvdG each reviewed 67 articles from a sample of 200 articles from the second round of exclusion. Given the high level of agreement between the reviewers, only the sample of 200 articles was reviewed independently by two reviewers to identify differences in scoring. For the third round of exclusion, WMAM reviewed all full texts, with JCK, CJL and IAMvdG each reviewing one-third of all full texts. Differences between reviewers were discussed until consensus was reached. In total, 26 articles were included in this study.
Quality appraisal
The quality of the included studies was assessed independently by two reviewers (WMAM, together with JCK, CJL or IAMvdG) using the Quality in Prognosis Studies tool.25 26 Articles were classified as being of low quality (*), referring to high potential bias; moderate quality (**), referring to moderate potential bias; or high quality (***), referring to low potential bias. The reviewers resolved differences through discussion until consensus was reached.
Analysis
This review investigates the predictive performance of prognostic tests for predicting the likelihood of experiencing a fall. The predictive performance of a prognostic test is often described similarly to that of diagnostic tests, based on diagnostic accuracy.24 In this review, diagnostic accuracy refers to the ability to discriminate accurately between fallers and non-fallers according to various measures, including sensitivity, specificity and area under the curve (AUC).27 To this end, data regarding sensitivity, specificity and AUC were extracted from the articles and described.
Sensitivity refers to the ability to classify individuals correctly as being at risk of falls, and specificity refers to the ability to classify individuals correctly as not being at risk of falls.28 A diagnostic test has good predictive value if sensitivity and specificity are >70%.29 The AUC is the area under the receiver operating characteristic (ROC) curve, which represents the accuracy of the test. The ROC curve can be used to select the best cut-off score for most optimal sensitivity and specificity, with greater AUC reflecting a better test. The accuracy of a diagnostic test is considered good or excellent if the AUC is >0.7.27 We ranked the outcomes, taking into account the cut-off values for good sensitivity, specificity and AUC.27 29
When analysing the results, we also considered criteria regarding the suitability of falls risks assessment tools for the primary care setting. The time available to primary healthcare providers is limited, due to their high workload.15 16 19 20 They also have limited resources for expensive equipment (eg, platforms, sensors), and their practices generally have little space.17 18 When analysing the results, we therefore considered the following criteria for a suitable tool: limited time, no expensive equipment and no spatial adjustments.
Patient and public involvement
Before conducting the systematic review, an informal focus group was conducted with primary care professionals (four GPs, two practice nurses and three district nurses)—the end-users—to identify their needs and wishes regarding falls risk assessment tools. We used the results of this informal focus group, together with previous literature, to define the suitability criteria used in this study. This ensured that the perspective of primary care professionals was taken into account when analysing the results of the review. No patients were directly involved in this systematic review.
Results
The 27 articles included in this review identify a total of six falls risk assessment tools. Each of these tools is described below and presented in table 1. Further details about the included articles are provided in online supplemental additional file 2.
Supplemental material
Falls risk assessment tools included in this review
Timed Up and Go test
The TUG test takes only a few minutes to complete, and it was described in 14 studies.30–43 In this test, participants are asked to stand up from a chair, walk 3 m, turn, walk 3 m back and sit down again. The time taken to perform this task indicates high or low falls risk. The cut-off scores reported in the articles varied from 10.9 s to 13 s. The AUC is described in 11 studies, ranging from 0.46 to 0.89. In eight studies, sensitivity ranges from 10% to 83.3%, and specificity ranges from 28.4% to 96.6%.
Gait Speed test
The Gait Speed test, based on a distance of 4 m, takes only a few minutes to complete, and it is evaluated in four studies.34 44–46 In this test, participants are asked to walk 4 m at their usual pace. The time taken to complete the task is recorded, and Gait Speed is calculated (m/s). An AUC value of 0.5 is reported by Bongers et al,44 and a value of 0.77 is reported by Tsutsumimoto et al.45 In an investigation of AUC for different follow-up periods and for any or recurrent falls, Kang et al34 report values ranging from 0.54 to 0.68. Sensitivity and specificity were reported in two studies,45 46 ranging from 38.4% to 100% and from 23.9% to 84.7%, respectively, depending on the cut-off scores.
Berg Balance Scale
The BBS evaluates a participant’s balance based on 14 items scored along a 5-point Likert scale and takes 15–20 min to complete. The score for each item ranges from 0 to 4 points, with an overall maximum score of 56 points. Balance is evaluated by asking participants to perform a variety of sitting, transferring and standing positions. In an assessment of which cut-off scores on the BBS best predict the risk of falling, Muir et al47 distinguish between single and multiple falls. They report an AUC of 0.68 for multiple falls with a cut-off score ≤53, and an AUC of 0.59 for a single fall with a cut-off score ≤54. A lower value of 0.47 is reported by Melzer et al.37 Sensitivity and specificity are reported in studies by Muir et al47 (25%–69%) and by Ersoy et al48 (53%–87%).
The Tinetti tests
The Tinetti tests are widely used tests for assessing the risk of falling, but there are many variations. One is the Performance Oriented Mobility Assessment (POMA) total, which consists of two components to assess balance (POMA-B) and gait (POMA-G) and takes about 20 min to complete. For the POMA-B test, which is assessed in four studies,41 46 49 50 participants are asked to perform nine different movements to assess balance. Depending on the cut-off scores, sensitivity ranges from 23% to 89%, with specificity ranging from 47% to 91.3%. An AUC of 0.66 is reported by Bizovska et al,49 but no cut-off scores are specified, and the comparison concerns multiple falls, thus excluding single falls. In the POMA-G, participants are asked to perform six different movements to assess gait. It is recommended to conduct this test in a corridor. The only study to specify the space used for the test is by Bizovska et al49: a well-lit corridor with a length of 30 m. Faber et al50 and Trueblood et al41 report sensitivities ranging from 21% to 64% and specificities ranging from 63% to 95%. Bizovska et al49 do not report any specific results, as they found no significant differences between fallers and non-fallers in relation to the POMA-G.
The Functional Reach test
The Functional Reach (FR) test is validated in three studies.36 40 51 In this test, participants are asked to hold their arms in front of them in an angle of 90 degrees, stretch forward as far as possible and return to the starting position. The distance between the starting position and the stretched position is used as an indicator of the risk of falling. This test takes less than 5 min to complete. The AUC is reported in two studies,36 40 varying from 0.51 to 0.60. Murphy et al51 mention a sensitivity of 73% and a specificity of 88%.
Falls history
Five studies explore the accuracy of falls history (FH),52–56 which takes only a few minutes to assess. These five studies apply different definitions of FH, with the most common being at least one fall in the previous year. Tiedemann et al56 and Nitz et al55 report AUC values ranging from 0.64 to 0.71. Sensitivity and specificity are explored in four studies, with sensitivity ranging from 39% to 69% and specificity ranging from 63% to 82%.
Quality appraisal
The methodological quality of all articles was assessed (see table 1). Three articles were classified as high quality, 21 articles as moderate quality and 3 articles as low quality.
Discussion
This study aimed to identify falls risk assessment tools that are suitable for the primary care setting (ie, they require limited time, no expensive equipment and no additional space) and that have good predictive performance in assessing the risk of falling among older people who are living independently. This systematic review identifies six falls risk assessment tools for the primary care setting. The vast majority of the included studies identify the falls risk among older people over a period of 12 months (mean: 15 months; minimum: 6 months; maximum: 9 years; see online supplemental additional file 2). None of these tools appears to be adequate in discriminating between people who are and are not at high risk of falling, taking into account the thresholds for good diagnostic accuracy (AUC >0.7), as proposed by Šimundić.27 These findings do not change when considering only the articles of moderate and high quality. Four studies report AUC values >0.7 for the TUG test,33 42 Gait Speed test45 and FH,56 thereby indicating good diagnostic accuracy.27 In most of the articles, however, the AUC values range from 0.5 to 0.7, thus indicating insufficient diagnostic accuracy for all of the tools addressed. Furthermore, the sensitivity and specificity of the same tool varied substantially across studies. We are therefore unable to draw convincing conclusions.
The results of this review are corroborated by other studies. For example, even though the TUG test is widely used to assess falls risk, other studies have also reported a lack of predictive ability for this test with regard to falls.57 58 Furthermore, as stated by Gates et al,59 ‘At present, recommending any screening test for routine clinical use is not possible. Despite the number of studies that have been conducted, no strong evidence exists that any screening test is useful for identifying fallers’ (Gates et al, p1113–1114).59 The current systematic review, conducted 13 years later, leads to the same conclusion. The lack of conclusive evidence to identify falls risk assessment tools with adequate predictive performance and accuracy persists to date. It is therefore impossible to select an assessment tool based on predictive performance. Our review nevertheless adds valuable information to the existing body of literature concerning the tool that is currently most suitable for use by primary care providers to identify patients who are at high risk of falls.
Primary healthcare providers have limited time and lack resources for expensive equipment, space and training.15–20 In light of these constraints, the results of this study suggest that the most suitable tool is FH, as it takes only a few minutes to conduct and requires no training, expensive equipment or spatial adjustments. The BBS and the Tinetti tests would not be suitable, as they take 15–20 min to complete and require training to conduct. The TUG and Gait Speed tests are both quick (<5 min), but they require training and space (>4 m) to conduct. Although the FR test is quick (<5 min) and does not require much space, it requires more training than FH and the AUC values reported are lower than those for FH.
Despite the fact that it is insufficient, the diagnostic accuracy of FH is the same or even better than that of most of the other five falls risk assessment tools (see table 1). Based on the clinometric evaluation of four falls risk assessment tools, Barker et al60 also identify FH as a suitable assessment tool, stating that ‘the predictive validity of all tools was found to be low, with no tool offering greater ability to identify residents who would fall than a simple screening question “has the resident fallen in the past 12 months?”’ (Barker et al, p919).60 Patient FH is also used in many multifactorial assessment tools and algorithms, and it appears to be an important factor in the risk of falling (OR: not significant–14.02).48 53 55 61–68 The use of FH nevertheless eliminates the possibility of identifying first-time fallers. Although this is clearly a major disadvantage, older people might be less willing to start and complete falls prevention interventions if they have not previously experienced a fall. They often do not consider themselves at high risk of falling.69 70 The experience of a previous fall might therefore enhance motivation to start and complete a falls prevention intervention.71
According to a study by Nordin et al,72 the assessment of falls risk through the combination of clinical judgement and FH among a population of frail older people was superior to performance-based measures. Meyer et al73 even assert that the use of falls risk assessment tools should be avoided, ‘since it has no clinical consequences other than the waste of scarce nursing resources’ (Meyer et al, p421).73 Due to increasing work pressure15–18 and lack of awareness,74 75 healthcare professionals might not assess a patient’s risk of falling based solely on clinical judgement, as it is not part of any systematic assessment strategy. The systematic assessment of falls risk by combining FH and the expertise of healthcare professionals might therefore be an adequate strategy.
Practice recommendations
In daily practice, GPs can ask their older patients during consultation if they have had a fall during the past 12 months. Even if a patient has not had a fall, the GP might still identify a high falls risk based on clinical judgement (eg, walking or sitting difficulties due to strength and balance problems, dizziness, use of benzodiazepines, visual impairment). If a high falls risk is suspected after such a brief assessment, the GP could investigate the underlying cause of the falls risk by conducting a multifactorial assessment so that adequate care can be provided. It should be noted that, in this study, FH is defined as an assessment tool and not as a screening tool. A falls risk assessment tool defines the nature of the problem, and thus whether a patient is or is not at high risk of falling.21 No additional assessment is required to identify high or low falls risk. Additional assessment (eg, multifactorial assessment) is needed only to determine which intervention is needed in order to reduce a patient’s high falls risk. Screening tools are intended to evaluate the possible presence of specific problems. A screening tool would require additional assessment in order to verify that a patient has a high falls risk.21
Depending on the organisation of the GP practice, the GP could also refer the patient to another healthcare provider (eg, a practice nurse specialised in geriatric care), who might have more time to investigate the underlying cause of the falls risk. A patient’s falls risk could be reduced by conducting a brief falls risk assessment that leads to a comprehensive multifactorial assessment to identify the underlying causes, followed by multifactorial interventions that address any risk factors that have been identified.76–78 The clinical practice guidelines of the American Geriatrics Society/British Geriatrics Society recommend conducting falls risk assessments annually.79
Strengths and limitations
This review was not registered at PROSPERO, the international prospective register of systematic reviews. This could have caused duplication of review topics. Nonetheless, no ongoing reviews were found in the PROSPERO register that specifically focus on suitability of falls risk assessment tools for the primary care setting.
In this review, the initial screening of titles and abstracts was performed by one researcher (WMAM). For the second round of selection, a sample of 200 articles was reviewed independently by a second researcher (JCK, CJL or IAMvdG), based on abstract (>95% consensus). Even though this is an acceptable procedure according to the Cochrane Handbook for Systematic Reviews of Interventions, each screening step should ideally be performed by at least two people working independently.80 Our results might therefore be subject to bias due to our method of study selection.
The results of this review were difficult to combine. Different studies used different cut-off scores, addressed modified versions of the same tests and presented different outcome measures. These differences between studies made it difficult to arrive at a convincing conclusion based on the results.
Given that we have included at least three studies for each tool, it would seem feasible to conduct a meta-analysis for each tool. We did not do this, however, for two reasons. First, the diversity between studies assessing the same tools was quite high. For example, there were substantial differences in cut-off scores, follow-up periods and study populations (eg, in terms of sex, age), as well as in the criteria for inclusion and exclusion and the quality of the studies. These differences rendered a meta-analysis unsuitable for most tools. Second, the results of our study are clear without conducting a meta-analysis: none of the six tools identified in the review appears to be adequate in discriminating between people who are and are not at high risk of falling, taking into account the thresholds for good diagnostic accuracy (AUC >0.7), as proposed by Šimundić.27 Another limitation is related to the possibility of publication bias against studies with worse outcomes, which might have led to an overestimation of the predictive performance of the falls risk assessment tools that were included. All of these limitations support our conclusion that none of the tools addressed has sufficient predictive performance.
Further research
The underlying cause of falls is often multifactorial and complex. This makes it difficult, if not impossible to adequately identify people who are at high risk of falling using only a physical test or brief questionnaire. None of the falls risk assessment tools identified in this review, all of which focus on falls history, balance, gait and/or strength problems, is capable of adequately identifying older people with high falls risk. It is therefore important to investigate other ways of assessing high falls risk in the primary care setting among older people who are living independently. The predictive performance of falls risk assessment tools could potentially be enhanced by developing a multifactorial assessment tool that also takes into account a person’s behaviour and environment.
Taken together, the results of this systematic review indicate that the predictive performance of the six falls risk assessment tools identified in the studies reviewed is insufficient. Overall, FH appears to be the same or even better than the other five tools. In addition, this tool is most suitable for the primary care setting, as it is quick and does not require equipment, space or training. The combination of FH and the clinical judgement of a healthcare professional could be a promising strategy in the primary care setting for identifying older people who are at high risk of falling, such that they can be provided with adequate falls prevention care. This could reduce both falls and fear of falling, thereby maintaining or improving quality of life and prolonging autonomy for older people.
Data availability statement
Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplemental information. All data relevant to the study are included in the article or uploaded as supplementary information. Additional data are available upon reasonable request from the first author.
Ethics statements
Patient consent for publication
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors WMAM conducted the systematic review and wrote the manuscript. JCK, CJL and IAMvdG reviewed articles for inclusion, reviewed the quality check and provided feedback on the manuscript. All authors read and approved the final manuscript.
Funding This work was supported by ZonMw, the Netherlands Organisation for Health Research and Development (Grant Number 531001210).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.