A Scoping Review of Fall-Risk Screening Tools in the Emergency Department for Future Falls in Older Adults

The World Health Organization has defined a fall as an event that results in a person coming to rest inadvertently on the ground or floor or lower level [1]. Approximately one-third of adults aged 65 and over fall each year, with half resulting in injury [2]. Falls were the second most common cause of death associated with injury in 2017 [3]. The majority of falls are reported by older adults living independently in the community and often result in presentation to the emergency department (ED) [4]. Falls in older adults are often life-changing [5] and a recent longitudinal analysis determined that approximately 36–44% of patients who came to the ED after experiencing a fall suffered subsequent adverse events, including recurrent falls, repeat ED visits, and death within 1 year [4]. The increased risk of adverse events has emphasized the importance of falls prevention to improve patient outcomes. Considering the heightened risk of ED re-attendances following falls, use of fall-prevention pathways from the ED demonstrates an optimization of health system resources, especially considering that inpatient admission of older adults can herald numerous complications [6], and evidence suggests that the ED is an appropriate environment for screening older adults at risk of occult disorders [6].

The effect of falls on health systems has prompted the establishment of evidence-based guidelines to proactively reduce fall risk [5, 7]. These guidelines recommend the use of screening tools to guide further assessment of appropriate patients and onward referral to relevant preventive measures [8]. Despite this recommendation, fall-risk screening tools’ uptake by clinicians remains poor [9, 10]. Competing demands on ED staff amid human resource constraints necessitates triaging of ED patients by acuity and immediate seriousness [11], and existing literature shows that screening for fall risk requires labour time [10]. The prospect of flagging patients for more comprehensive review has demonstrated both patient benefits [12] and system benefits for ED workflow [13].

The ability of fall-risk screening tools in the ED to estimate risk of future fall remains unclear [2]. Whilst various initiatives resulting from the screening of patients in the ED have demonstrated improved workflow [13, 14], they have not addressed the value of screening tools to estimate risk of future falls, thus initiating comprehensive assessment or preventive measures. A small number of ED-validated fall-risk screening tools exist [15, 16] and several other non-ED-validated fall-risk screening tools exist [10, 17, 18]. One reason for the existence of multiple tools is the multidimensional nature of falls, such that different tools are appropriate for different subgroups of patients [18]. Within the geriatric ED guidelines [7], the boundary between “assessment” and “screening” has become less defined. Some physical tests [19, 20], traditionally regarded as assessments stemming from screening [21], are now recommended as part of a screening examination in certain contexts [22] to guide further assessment. Considering this ambiguity, a scoping review is an appropriate way to clarify key concepts.

Moreover, a key challenge exists in determining who, and when, should be the subject of screening. A retrospective chart review [23] of 350 patients demonstrated that guideline adherence was very low but that the four key characteristics associated with guideline adherence were older age, higher comorbidity, aged care residency, and hospital admission. Commentary by two authors [6, 24] advises the importance of capturing lower-risk patients in the ED. Evidence suggests that the 6-month aftermath of such presentations is a critical period [25] wherein the true benefit of screening in the early and pre-symptomatic stages of a disorder [26] materializes with referral onwards to appropriate preventive services.

The aim of this scoping review was to examine the clinical utility of fall-risk screening tools in the ED as determined by their association with future falls. Review methods and reporting align with the Arksey and O’Malley [27] framework for scoping reviews. The Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews checklist (PRISMA-ScR) [27] is provided in online supplementary File 1 (for all online suppl. material, see https://doi.org/10.1159/000541238).

Our research question was “what is the current evidence base for screening tools in the ED to estimate fall risk and their association with future falls?”. “Screening tools” refer to items used under the broad definition of “risk screen” as distinct from “assessment.” We also chose the term “association with future falls” as we were not seeking to make assumptions related to causality.

We searched PubMed, Embase, and CINAHL for peer-reviewed journal articles published since 2012 that evaluated one or more screening tools to determine patient-level fall risk. This search was last conducted on June 23, 2024. A previous systematic review of fall-risk screening undertaken in 2014 [2] identified only one observational study published before 2012 [28]. For this reason, 2012 was selected as the starting year for this review.

The search strings were based on the three key concepts of “diagnosis and prediction,” “ED,” and “falls.” Numerous synonyms were identified through discussion between authors. Where necessary, wildcard operators using regular expressions were used. Major subject headings pertinent to each database were used where applicable. Search builder strategies were used to link key theoretical constructs. Database-specific search strategies are provided in online supplementary File 2.

English language studies published since 2012 were included. Observational studies were included employing either prospective or retrospective designs. We limited the selected studies to those with subjects 50 years and over who were discharged directly home from the ED, or home from hospital after admission via the ED, including patients discharged to residential aged care facilities. All eligible studies evaluated future fall outcomes following screening. Future falls post-screening included self-reported falls and re-admissions to the ED due to a fall irrespective of injury. We adopted the clinical definition of a fall as an event that results in a person coming to rest inadvertently on the ground, floor, or other lower level [1]. Eligible studies could report falls as a primary or secondary outcome. We included only publications that examined screening tools that were administered in the ED setting or could be completed using only information collected during the ED presentation.

Papers were excluded wherein screening relied on biomechanical or gait parameter measurement requiring specialized measurement equipment or diagnostic imaging. Studies examining screening tool acceptability and staff compliance as a primary objective were excluded. Conference abstracts, letters, and systematic reviews were excluded.

Search results were exported to EndNote 20 [29] for de-duplication. Further de-duplication and blinded abstract screening were performed using Rayyan [30] by two authors (D.W., J.R.). Screening decision conflicts were resolved by discussion with a third author (N.M.W.). A data extraction sheet was developed and approved by all authors. Data extracted after full-text review included study location, study design, sample size information, inclusion and exclusion criteria, details of the screening tool(s) examined, outcome definitions, and details of statistical methods used for analysis.

Key covariates from the selected studies were summarized by narrative synthesis. Methods of analysis ranging from descriptive statistics through to prognostic performance of tools were summarized and compared where applicable.

The full search strategy returned 3,127 records, which decreased to 2,308 records after de-duplication (Fig. 1). Abstract screening identified 75 records for full-text review. A further 60 articles were excluded following full-text review. The main reasons for exclusion were based on the outcomes evaluated (n = 30), study setting (n = 12), and publication type (n = 11).

Sixteen studies were included (Table 1). Five studies originated from the USA [32‒36], four studies were from Australia [15, 37‒39], two studies were from Canada [40, 41], one study was from the Netherlands [42], and one study was from Brazil [16]. One study [10] collected data from the USA and Thailand. One study, from Singapore, was retrieved by reference searching [43].

Fourteen unique screening tools were found. Screening tools varied based on their intended application and factors for estimating fall risk. Eight tools were fall-risk screening tools [10, 15, 16, 32, 34, 35, 37‒39, 44], two tools were rapid physical tests [33, 40, 43, 44], one tool was a functional screening tool [35], one tool was a frailty-screening tool [42], one tool was a trauma triage tool [36], and one tool was a component of a health-related quality-of-life measure [44]. Seven studies evaluated risk-screening tools that were specifically designed for falls. Two screening tools had already been developed and validated for use in the ED [15, 16]. Only one tool was integrated into the electronic medical record (EMR) [32] and all tools required focused time on behalf of participating clinicians.

Studies that evaluated screening tools not incorporating physical tests varied based on recruited study participants. Three studies recruited both fall and non-fall ED presentations [32, 34, 39]. Four studies recruited only patients presenting to the ED with a fall or already identified as “at risk” by other methods [15, 35, 39, 41]. One study [10] recruited patients who had been risk-defined by the “Stopping Elderly Accidents, Deaths, and Injuries” (STEADI) guidelines [21], and another included any patients not requiring hospital admission for a fall [16].

One of the physical test papers only recruited patients who had sustained trauma in the previous 2 weeks [40]. A second physical test paper [33] recruited both fall and non-fall ED presentations but required that patients personally identify risk factors for falling from the STEADI toolkit [21], while another, combining a physical test, fall screen, and quality-of-life measure [44], recruited both fall and non-fall ED patients. The studies embedding fallscreening tools within composite geriatric assessments [37, 38, 43] and the study of the frailty screen [42] recruited both fall and non-fall ED presentations. The trauma triage tool article [36] only recruited patients with fractured hips but did not specify whether they were fall related.

More than half of the studies defined their target population based on the outcome of ED presentation. Six studies were restricted to patients who were discharged home immediately after ED presentation [16, 33, 39‒41, 44]. Seven studies included patients who were either discharged home directly from the ED or were admitted to hospital [10, 15, 34, 37, 38, 42, 43]. One study excluded those discharged to other areas of the hospital [10] before screening could be completed, and one study specifically described the exclusion of patients who were unable to consent [39]. The trauma triage tool study only included those admitted for hip arthroplasty [36] using only information available at the time of presentation.

All sixteen studies collected data under retrospective, prospective, or longitudinal designs. Two studies employed a retrospective, cross-sectional design [32, 36]. Two studies employed a longitudinal design that involved a fall follow-up, reissue of measures from baseline, and issue of a Short-Form 12 (SF-12) [45] for quality-of-life measures based on physical and mental component scores [37, 38].

Future falls were evaluated in all included studies, as the primary [10, 15, 16, 32‒36, 39‒41, 43, 44] or secondary [37, 38, 42] outcome. Eleven studies relied exclusively on patient reporting of falls [16, 33‒35, 37‒43]. Two studies relied on ED presentation clinical codes and manual screening of notes for follow-up fall outcomes [32, 36]. Three studies contacted patients and used patient notes as a secondary data source, although neither of these studies indicated whether clinical codes were used [10, 43, 44]. Five studies defined patient follow-up as 6 months or less [10, 15, 32, 33, 39]. One study did not define a specific follow-up period for evaluation [36]. All other studies comprised varied or multiple follow-up times of 1, 3, 6, 9, and 12 months [16, 35, 37, 38, 40‒44].

A range of statistical methods were used to evaluate associations between patients having risk factors and future falls (Table 2). Reported approaches were descriptive statistics only [38], univariable or multivariable analyses [10, 36, 37, 40], and prognostic accuracy assessment, including sensitivity and specificity, and area under the curve statistics [33‒35, 39, 42]. Three studies assessed both univariable or multivariable associations and prognostic accuracy [16, 32, 44]. One study also constructed a multivariable hazard function for 180-day risk of falls and death adjusted by the Charlson Comorbidity Index [16].

Most tools performed modestly for their assessment of risk of future falls. Analysis of prognostic accuracy of tools revealed a pattern of higher sensitivity (true positive rate) than specificity (true negative rate) for two of the studies on the same tool [15, 34]. One study reported an improvement in the area under the curve from 0.64 to 0.75 after inclusion of additional variables from patients’ EMRs [32]. Only three of the reviewed studies performed sensitivity analyses or attempted to account for missing data [32, 41, 44].

Most analyses were adjusted for covariates to account for known patient risk factors (Fig. 2). Previous falls and high-risk medications had stronger associations with future falls than age and gender. A detailed discussion of covariates and their findings can be found in online supplementary File 3.

Previous falls were a consistently strong predictor of future falls in most of the fall-specific screening studies [10, 15, 16, 32, 34, 39, 41] with the exception of one study [44]. Two of the studies of the two-item screening tool [15, 34] reported very strong associations between the occurrence of two or more falls in the previous 12 months and future falls. Both assessed patients at 6 months follow-up. One of the physical test studies included self-reported previous falls as an explicit predictor variable of self-reported future falls [33].

High-risk medications and six or more medications were consistently associated with self-reported falls at follow-up. All studies reporting on medications had relevant medication questions embedded within their respective tools [10, 14, 15, 32, 34, 37, 39, 41, 44, 46]. One scale [15] gauged risk by six or more medications, whereas the STEADI toolkit [10] targeted medications according to the 2015 Beers Criteria for high-risk medications [47]. Only medications causing light-headedness were associated with the future falls at 6 months in the latter study, though in another study this was not associated with future falls when self-reported [44].

This scoping review aimed to determine the current evidence base for fall-risk screening tools used in the ED with respect to their association with future falls. The studies in this review demonstrated variable prognostic utility with respect to future falls prediction. Almost all tools consistently demonstrated associations between previous falls and high-risk medications with future falls. Evaluation methods varied in terms of statistical methods employed, outcome measures, and timeframes for follow-up.

In keeping with a systematic review of the subject performed a decade ago [2], previous falls and high-risk medications were consistently associated with future falls. Despite the publication of 13 relevant papers since 2014, comparison of screening tools remains difficult because of different sampling methods, inclusion criteria and follow-up periods. This could warrant a call to standardize outcomes and study methods enabling comparison across studies. The discordance of practice with the geriatric ED guidelines [23] demonstrates that screening tools have a role in the ED as a very quickly administered inventory guiding onward primary care referral.

Unlike the recent geriatric emergency care scoping review [48], our scoping review sought to examine only very quick screening tools rather than assessment tests. Acknowledging discordance of guidelines with practice [23] and the numerous reasons for discordance, including time constraints, it is likely that ED fall screening will continue to fall short of geriatric ED guideline recommendations. This was identified in a secondary study of a prospective trial [24], which revealed consistently low rates of completion of numerous guideline items, which are known to be associated with falls, e.g., visual acuity, orthostatic hypotension, high-risk medication, and peripheral neuropathy.

With previous falls and high-risk medications consistently identified as risk factors across almost all included studies in this review, the concept of a quick screen as a risk stratification strategy guiding referral on to a more detailed assessment needs to be considered as the role of screening tools. Whilst the geriatric ED guidelines recommend detailed assessment and intervention in a primary care setting in follow-up, they do not differentiate between a quick screen for risk stratification and a more detailed assessment. Audits of practice [6, 23, 24] have shown greater guideline adherence associated with the four patient-level characteristics of older age, greater comorbidity, residence in assisted-living accommodation, and admission to inpatient or observation wards. It is the lower-risk cohort, often those residing in the community and discharged directly from the ED, who should be the target of quick-screening tools flagging previous falls and high-risk medications.

The higher sensitivity relative to lower specificity reported with many of the ED-specific fall-risk screening tools and physical tests is common in tools designed for use in the ED, where strong negative predictive value is required. Such results are typical of tests delivered as part of an initial survey when ruling out sinister or dangerous pathology in the ED, such as deep vein thrombosis [49] and cervical spine fracture [50].

Existing evidence strongly associates previous falls with future falls [2, 17, 51]. This conclusion weighs heavily on the methodology of selecting only patients who had fallen in many of the studies found by this review. Most studies of risk screening in this review compared the test score or items within the test score with the outcome of a future fall. An important factor in prognostic studies is that the study sample should be representative of the target population [52], otherwise heightening the risk of selection bias [53].

The risk of recall bias when relying on participants reporting the outcome of interest is also a major consideration [52]. Considering the association between cognitive decline and falls [54, 55], this introduces a potentially large source of systematic error in studies relying on self-reporting of falls. Moreover, it has previously been reported that patients may under-report falls due to embarrassment, social stigma, or cultural factors [56]. The lack of sensitivity analysis in almost all of the studies further highlights the risk of bias evident in many of the studies, a risk well documented in existing statistical literature [52].

With the varied methods of participant selection, follow-up, and statistical analysis encountered in the studies reviewed, comparison and subsequent synthesis of results in pooled analysis was problematic. While meta-analysis can offer sophisticated means of enquiry and pooled estimates of screening accuracy, it is unsuitable for this collection of studies with considerable heterogeneity of selection criteria, screening items, and scoring thresholds [57]. The differing methods and inclusion criteria of studies included in this review may warrant a call to standardize outcomes and study methods, so we can compare across studies while noting that this may be challenging due to geographical variations in practice. Included studies evaluated the clinical utility of existing screening tools that have been developed using different statistical methods. We did not appraise the methodological quality of screening tools evaluated.

This study was conducted as a scoping review for the purpose of clarifying key concepts and knowledge gaps within the literature. A discussion such as the nuanced differences between screening, assessments, physical tests and the overlap that occurs between them, as well as the embedding of some screening tools within more comprehensive geriatric assessments, demands a broad focus better served by a scoping review than a systematic review [58].

All tools evaluated require the clinician to deviate from workflow and invest time in dedicated analysis in the context of a busy ED. Considering the typically high acuity caseload of ED, this may not be always possible to accommodate in clinical practice. Many EDs require the input of allied health staff as secondary rather than primary contact practitioners; thus, referrals are generated by emergency physicians and registrars. Even in cases of primary contact nursing and allied health roles, clinician time is often spent involved in detailed assessment of a variety of patient types. This means that multiple simultaneous arrivals, unless automatically screened, are likely to result in fall-risk screening being missed.

Targeted falls prevention, e.g., balance classes and combined geriatric care, has returned mixed results [8, 59, 60]. A shortcoming of many fall-prevention programmes is the provision of care exclusively for those who have already fallen, given their remedial context. While the fall is often the sentinel event, the process of functional, cognitive, and physical decline is long established.

Secondary use of EMR data was identified in one of the studies as worthwhile for augmenting screening of at-risk patients without adding to existing workflows [32]. Clinical decision support systems have already been deployed across various clinical domains as early warning systems and automated alerts, although they have faced significant hurdles [61, 62].

Risk screening tools for future falls in the ED have shown promise; however, sample selection biasing those who have already fallen makes generalization about risk prediction difficult. Whilst it is unlikely to be possible for clinicians to manually screen everyone for fall risk in the ED, digitized systems capturing data from the patient’s EMR and flagging markers associated with heightened falls risk warrants further exploration and evaluation.

Studies of fall-risk screening in the ED aimed at identifying future fallers have already revealed several common risk factors associated with future falls. Current approaches to fall-risk screening in the ED may not be well suited to the time-sensitive nature of ED workflows. Missed fall-risk screening may compromise identification of patients at high risk of falls and contributes to missed opportunities to direct these patients to appropriate care. Further research to examine the role of computerized clinical decision support systems, potentially leveraging automated processing from routinely recorded EMR information, may assist in overcoming some of the workflow challenges likely associated with fall-risk screening approaches identified in this review.

The authors declare no conflict of interest.

No funding was received for the preparation of this manuscript and there was no sponsor’s role.

This study was conceptualized by D. Wickins, S.M. McPhail, and N.M. White. This study was designed and synthesis of data was conducted by D. Wickins and N.M. White. Extraction and analysis of data was conducted by D. Wickins, J. Roberts, and N.M. White. Technical support for figure design was provided by J. Roberts. Critical review and editing of this manuscript were conducted by D. Wickins, J. Roberts, S.M. McPhail, and N.M. White.