Development of the Falls Risk for Older People in the Community (FROP-Com) screening tool*
1 National Ageing Research Institute, Preventive and Public Health Division, Parkville, Melbourne, Victoria 3095, Australia
2 University of Melbourne, Centre for Molecular, Environmental, Genetic and Analytic Epidemiology, Melbourne 3095, Australia
3 Faculty of Health Sciences, LaTrobe University and Northern Hospital, Bundoora, Victoria 3083, Australia
4 Monash University Accident Research Centre, Monash University, Victoria 3800, Australia
Address correspondence to: Melissa A. Russell. Tel: (+61) 3 83872200; Fax: (+61) 3 83872153. Email: m.russell{at}nari.unimelb.edu.au
 |
Abstract |
|---|
 TopNotes Abstract IntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
Background: the aim of this study was to develop a brief screening tool for use in the emergency department (ED), to identify people who require further assessment and management.
Methods: this prospective study included 344 community-dwelling older people presenting to an ED after a fall. After direct discharge participants had a home-based assessment performed that included the Falls Risk for Older People in the Community (FROP-Com), a comprehensive, yet simple, multifactorial falls risk assessment tool. They were then monitored for falls for 12 months. The items from the FROP-Com assessment tool predictive of falls in a multifactorial logistic regression were used to develop the FROP-Com screen.
Results: the items significantly predictive of falls and combined to form the FROP-Com screen were: falls in the previous 12 months, observation of the person's balance and the need for assistance to perform domestic activities of daily living. At the cut-off with the highest Youden index sensitivity was 67.1% (95% CI 59.9–74.3) and specificity was 66.7% (95% CI 59.8–73.6).
Conclusion: the FROP-Com screen has a relatively good capacity to predict falls. It can be used in time-limited situations to classify those at high risk of falls who require more detailed assessment and management.
Keywords: accidental falls, aged, geriatric assessment, risk factors, elderly
|
Introduction |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
An estimated 8% of all people aged 65+ are admitted to a hospital following a fall [1]. In the UK, 647,721 adults aged 60+ present to a hospital annually after a fall, with 443,297 of these people treated and discharged directly from the emergency department (ED) [2]. After an ED presentation older fallers are at risk of further falls, with 52% falling again in the following 12 months in lieu of intervention [3]. The presentation to the ED is an opportunity to identify those at risk of further falls.
Two studies, including the landmark PROFET study, have found intensive multifactorial interventions to be effective in improving outcomes for older people presenting to the ED after a fall [3, 4]. These effective interventions consisted of blanket assessment and management from a medical specialist and allied health professionals. Such intensive programs can prevent falls but questions remain regarding the cost and feasibility (especially in regard to the staffing levels) of such programs. The authors therefore recommended ‘evidence based stratified care pathways for the management of fallers .... to ensure the appropriate use of limited resources’ [4, 5].
Screening for falls risk in the ED provides an approach to executing stratified care pathways and improve efficiency by directing only those at high risk of falls to full evidence-based care. Falls risk screening tools, for older people presenting to the ED after a fall, have not been published with the relevant predictive data for implementing stratified care pathways.
Falls risk screening is particularly pertinent in the ED, where preventative care competes with the demands of injury management and timely discharge from the ED. Current international research indicates that falls prevention measures are not being routinely implemented within the ED [6–8]. A falls risk screening tool can serve as a reminder to clinicians to check for falls risk, in addition to guiding practice.
We have developed a falls risk assessment tool [the Falls Risk for Older People in the Community (FROP-Com) tool] for detailed falls risk assessment in the community setting. The FROP-Com assessment tool covers 13 risk factors in 26 questions with ordinal (0–3) or dichotomous scoring. It has high inter-rater reliability [intraclass correlation coefficient (ICC): 0.81 (95% CI 0.59–0.92)] and moderate predictive validity [sensitivity: 71.3% (95% CI 64.4–78.3), specificity: 56.1% (95% CI 48.9–63.4)] [9]. The full FROP-Com and its guidelines for scoring and further assessment and interventions are available at http://www.mednwh.unimelb.edu.au/research/research_falls_service.htm.
As the first step in the validation process this paper reports a data-driven approach to develop a brief falls risk screening tool for use in the ED—the FROP-Com screen—from the items in the FROP-Com assessment tool. The objectives of this study were to (a) identify the combination of risk factor items derived from the FROP-Com assessment tool that were most predictive of future falls in a sample of older people presenting to the ED after a fall and (b) to evaluate the predictive accuracy and reliability of the abbreviated combination of items identified in (a).
|
Methodology |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
Study design and participants
Development of the FROP-Com screen
The detailed methodology of sampling and assessment has been described previously [9].
The study, undertaken as a prospective cohort study, included participants randomised into the control arm of a randomised controlled trial (RCT). The RCT investigated the effect of a multifactorial falls risk assessment and targeted intervention program on falls and injury. Participants were recruited through seven acute hospital EDs in Melbourne, Australia. Patients were eligible if they were aged 60 years or older, lived in the community; presented to an ED as a result of a fall; were discharged directly home following emergency care and were able to walk independently. All participants in the control group (the sample for this study) received usual care from the ED staff, including potential falls prevention services (falls clinic, physiotherapy or occupational therapy).
The project was approved by the Human Research Ethics Committees at the participating hospitals.
Reliability of the FROP-Com screen
Nested within the RCT, 20 consecutively recruited participants were included in each of the intra-rater and inter-rater reliability sub-studies. Participants were excluded if they did not have functional English or if they had a major medical event occurring between the two assessments. This was defined as any event for which the participant sought emergency medical care at a hospital.
|
Data collection |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
Development of the FROP-Com screen
Baseline measures
After discharge from the ED, participants were visited at home by one member of the research team (a physiotherapist, occupational therapist or medical doctor, allocation depending on time availability). At this home visit the following data were collected and tests administered:
- Demographics
- The injury sustained in fall resulting in the ED presentation
- The services put in place by the ED staff
- The FROP-Com assessment tool.
Follow-up
The binary outcome measure used was the occurrence of falls (no falls versus at least one fall) in the 12-month follow-up period. Falls data were collated using falls diaries and the baseline assessment results were not accessed during the follow-up period. Any participants not completing the follow-up period and not sustaining a fall prior to their early withdrawal were excluded from the analysis. A fall was defined as an event which results in a person coming to rest inadvertently on the ground or lower level [10].
Reliability of the FROP-Com screen
In the intra-rater reliability study a physiotherapist performed all of the baseline assessments and then returned to the participant's house 2 weeks later to repeat all of the items in the FROP-Com assessment tool. In the inter-rater reliability study a physiotherapist or doctor performed the baseline and repeat assessments. In the repeat assessment, conducted 1 week after the baseline assessment, the second clinician was masked to the initial results.
|
Statistical analysis |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
Development of the FROP-Com screen
Power analysis
Using a sample size estimation procedure for a proportion, 350 participants were calculated to be required for this study. This calculation was based on having a sensitivity of 65% [9] and CI range of 5%.
All items from the FROP-Com assessment tool were included in the analysis to determine the items most predictive of falls, except those that were impractical for a short screening test [items: home environment, functional behaviour, walking safely in the community, change in personal activities of daily living status after a fall (ADL), change in domestic ADL status after a fall, change in physical activity status after a fall]. Hence, the remaining 20 items (listed in Table 1) were entered into the analysis.
|
Multiple logistic regression was used to construct the screening tool. Univariate logistic regression was initially performed to identify the factors most strongly associated with falls occurrence, without adjustment for other factors (P = 0.05). Associated factors were entered into the multiple logistic regression according to the strength of the association, assessed by the P-value for the null hypothesis of no association between the predictive factors and outcome. Odds ratios were not used to judge the strength of association due to the differing degrees of freedom between the dichotomous and linear variables.
The significantly associated factors from the multiple logistic model were combined to form the FROP-Com screening tool. A receiver operating characteristic (ROC) curve was plotted for the FROP-Com screen and area under the curve (AUC) was calculated. A ROC curve displays sensitivity versus 1 – specificity for all possible cut-off scores and AUC is a summary measure of test performance over all cut-offs. Sensitivity, specificity, negative and positive predictive values were calculated and the cut-off with the highest Youden index (sensitivity + specificity –1) was found [11]. Sensitivity is the proportion of true positives correctly identified whilst specificity is the proportion of true negatives correctly identified [12]. The positive predictive value is the proportion of test positives that are truly positive and the negative predictive value is the proportion of test negatives that are truly negative [12]. Likelihood ratios for different FROP-Com screen scores were also calculated. The likelihood ratio expresses the odds that a given level of a diagnostic test result would be expected in a patient with (as opposed to one without) the target disorder [13].
The ROC procedure was repeated with a subset of participants who did not receive any falls prevention services from the ED and with the outcome of recurrent (1 or no falls versus 2 or more falls) falls in the 12-month follow-up. The ROC analysis was repeated with recurrent falls as an outcome to further assess the predictive capacity of the FROP-Com screen with this clinically important outcome.
Reliability of the FROP-Com screen
Intra-rater and inter-rater reliability of the newly developed FROP-Com screen were determined using ICCs [14].
|
Results |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
Development of the FROP-Com screen
Participants were enrolled between January 2003 and December 2006. The flow of the participants through the study, participant characteristics and details of index fall are reported elsewhere [9]. To summarise, 1,342 patients were referred by ED staff and 718 consented to take part in the RCT. Three hundred and sixty one participants were randomised into the control arm. However, 16 participants were excluded as they did not complete the 12-month follow-up period and did not have a fall prior to their withdrawal and one was excluded due to incorrect randomisation. The 344 participants in the study had a mean age of 75.9 (95% CI 75.0–76.8), and 69.2% (95% CI 64.3–74.1) were female [9]. As a result of standard care in the control group 17.2% (95% CI 13.3–21.6) of participants received physiotherapy, 10.2% (95% CI 7.2–13.9) received occupational therapy and 3.8% (95% CI 2.0–6.4) were referred to falls and balance clinics.
In the 12-month follow-up period, 100 (29.1%) participants sustained recurrent (
two falls), 64 (18.6%) sustained one fall and 180 (52.3%) sustained no falls.
In the multifactorial logistic regression, number of falls in the past 12 months, observation of the person's balance and the question regarding the need for assistance to perform domestic ADLs were statistically significant predictors of falls (Table 1). These three items were the items chosen for the FROP-Com screening tool.
The mean score for the FROP-Com screen was 3.60 (95% CI 3.39–3.81) from a possible 9 (each of the three items scored 0–3). Higher scores were indicative of higher falls risk. The AUC for the ROC of the three-item FROP-Com screen was 0.73 (95% CI 0.67–0.79). The Youden index was highest (0.34) at a cut-off of 3/4. At this point sensitivity was 67.1% (95% CI 59.9–74.3), specificity was 66.7% (95% CI 59.8–73.6), positive predictive value was 64.7 (95% CI 57.0–71.9) and negative predictive value was 69.0% (95% CI 61.5–75.7). The number of fallers and non-fallers for each score, sensitivity, specificity, positive predictive values and negative predictive values for all cut-offs and likelihood ratios are contained in Table 2.
|
The AUC, for the FROP-Com screen with recurrent falls as the outcome, was 0.70 (95% CI 0.64–0.76). The Youden index (0.29) was highest at the cut-off of 3/4 (Table 3).
|
Of the 263 participants (76.5%) who did not receive any potential falls prevention services, 122 (46.4%) went on to fall. These participants, in comparison to the participants receiving falls prevention services, performed significantly better on the FROP-Com assessment tool and other measures of falls risk [9]. For the participants not receiving potential falls prevention services the AUC for the FROP-Com screen was unchanged from the AUC of the ROC for the full sample of participants (AUC: 0.73; 95% CI 0.67–0.79). The Youden Index was highest (0.36) at a cut-off of 3/4 (Table 3).
Reliability of the FROP-Com screen
The characteristics of the participants in the reliability studies have been previously published [9]. Intra-rater and inter-rater Kappa statistics for individual FROP-Com items are contained in supplementary data for the previously published article [9].
The ICC for intra-rater reliability for the FROP-Com screening tool was 0.87 (95% CI 0.70–0.95) and the ICC for inter-rater reliability was 0.89 (95% CI 0.75–0.96).
|
Discussion |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
We developed a FROP-Com screening tool, using the three items from the FROP-Com assessment tool most strongly predictive of falls: number of falls in the past 12 months, observation of the person's balance and the question regarding the need for assistance to perform domestic ADLs. This combination resulted in a screening tool with a moderate level of predictive ability and good levels of reliability.
The FROP-Com screen is the first falls risk screening tool developed for use in the ED to have predictive accuracy data available. The one previous tool described for use in the ED, the Falls Risk Proforma [15], serves a quite different purpose. It contains items addressing injury assessment and test results, such as bloods and urinalysis and is therefore more of a guide for medical care of the older faller in the ED than a tool for predicting further falls. It does not give an overall score of falls risk so has no attached sensitivity, specificity or likelihood ratios.
The most appropriate cut-off point for the FROP-Com screen should be considered in the context of the setting requirements. With a higher specificity there are fewer people classified incorrectly as fallers who do not go on to fall (false positives) and with higher sensitivity there are fewer people classified incorrectly as non-fallers who go on to fall (false negatives). It is proposed that a positive result on the FROP-Com screening tool would trigger the implementation of multidisciplinary evidence-based care. In this case the risk of false negatives would outweigh the risk of false positives (i.e. it would be worse to miss a person going on to have a fall than to undertake a full assessment and provide interventions that the non-faller may benefit from anyway). Hence a cut-off of 2/3 may be the most appropriate choice, resulting in only 20% of fallers incorrectly classified as low risk and 50% of non-fallers classified as high risk (Table 2). However in the clinical situation, if resources are scarce, the cut-off point could be raised to decrease the numbers of false positives.
The clinically useful level of the Youden index has not been previously documented. Hospital falls risk screening tools with the greatest accuracy have pooled Youden indices of 
0.4 [16]. In this study the highest Youden index was 0.34. However, a caveat on this comparison is the very different populations and period of follow-up time. In comparison to other falls risk screens developed for the community, the FROP-Com screening tool performs favourably. Firstly, the FROP-Com was developed in the gold standard of a 12-month prospective study using falls diaries. Other screening tools have had sensitivity and specificity data published from concurrent studies [17], used retrospective falls as an outcome [18], not included single fallers in the analysis [19] or had high loss to follow-up rates [20]. Compared to the screening tools developed in well-executed prospective studies [21–23] the FROP-Com screen has a slightly higher level of overall predictive power (calculated by the Youden index or AUC). However, a clinical decision to adopt a falls risk screening tool should be based on the study being well designed, the predictive accuracy of the tool, appropriateness of the questions/tests in the clinical setting proposed and the similarity of the clinical population to that studied so that negative and positive predictive values can be applied.
A limitation of this study is that the FROP-Com screen has not been validated in a second sample. This is a necessary next step in its development. The sensitivity and specificity reported here can be extrapolated to other populations of community dwelling older people. However, negative and positive predictive values alter with falls prevalence; therefore, external validity is limited to populations with similar high falls risk profiles.
The FROP-Com screen requires no equipment or specialist knowledge and could be performed in 1–2 min by any health professional in the ED. Minimal staff training would be required using the resources already available: http://www.mednwh.unimelb.edu.au/research/research_falls_service.htm. The FROP-Com could be used as a stand-alone falls risk screen or be easily incorporated into a larger assessment of geriatric problems. It could be used to stratify patients into low or high risk, with those found to be at high risk ideally referred to services providing evidence-based care [3, 4]. An additional benefit of the FROP-Com screen is that two of the risk factors (safety mobilising and lack of independence performing ADLs) can be acted upon directly for the improvement of general health for those found to be at lower risk of falls.
In summary, the newly developed FROP-Com screen consists of the risk factors of number of falls in the past 12 months, observation of balance and needing assistance performing ADLs. It is a quick and easy to apply falls risk screening tool with a moderate capacity to predict falls. The FROP-Com screen could be used to stratify people into low or high risk. Such stratification provides a more feasible approach to implementing evidence-based falls prevention in the ED by capturing those who are likely to most benefit from evidence-based intensive multidisciplinary intervention.
|
Key points |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
- The items from the FROP-Com assessment tool found to be predictive of falls over a 12-month period and chosen to form the FROP-Com screen were number of falls in the previous 12 months, observation of the person's balance and whether the person required assistance to perform domestic ADLs.
- The AUC for the ROC of the newly formed three-item FROP-Com screen was 0.73 (95% CI 0.67–0.79).
- The Youden index was highest (0.34) at a cut-off of 3/4 (from a possible highest risk score of 9). At this point sensitivity was 67% (95% CI 59.9–74.3) and specificity was 67% (95% CI 59.8–73.6).
- Taking a minute or two the FROP-Com screen could be used as a stand-alone tool or be incorporated into a larger assessment of geriatric problems. If a positive result were found, evidence-based falls prevention interventions could be implemented.
|
Funding |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
Australian Government Department of Veterans’ Affairs and the Victorian Department of Human Services. The funders played no role in the production of this research. LD and SD are supported by the Australian National Health and Medical Research Council.
|
Conflict of interest |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
The authors have no conflicts of interest.
|
Acknowledgements |
|---|
The participants in the ‘Falls Aren't Us study’, Melitta Giummarra, Sue Williams, Terry Sullivan, Sally A’Beckett, Leslie Dowson, Mike Dosrevitch, Joe Ibrahim, Jenny Schwartz, Sarah Tarquinio, Marcia Fearn, Fiona Bremner.
|
Notes |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
*This research was undertaken at the Preventive and Public Health Division, National Ageing Research Institute, Poplar Rd, Parkville, Victoria, 3052, Australia.
|
References |
|---|
|
TopNotesAbstractIntroductionMethodologyData collectionStatistical analysisResultsDiscussionKey pointsFundingConflict of interestReferences |
|---|
- Stel V, Smit J, Pluijm S, Lips P. Consequences of falling in older men and women and risk factors for health service use and functional decline. Age Ageing (2004) 33:58–65.
[Abstract/Free Full Text] - Scuffham P, Chaplin S, Legood R. Incidence and costs of unintentional falls in older people in the United Kingdom. J Epidemiol Community Health (2003) 57:740–44.
[Abstract/Free Full Text] - Close J, Ellis M, Hooper R, Glucksman E, Jackson S, Swift C. Prevention of falls in the elderly trial (PROFET): a randomised controlled trial. Lancet (1999) 353:93–7.[CrossRef][Web of Science][Medline]
- Davison J, Bond J, Dawson P, Steen I, Kenny R. Patients with recurrent falls attending Accident and Emergency benefit from multifactorial intervention—a randomised controlled trial. Age Ageing (2005) 34:162–8.
[Abstract/Free Full Text] - Close J. Prevention of falls—a time to translate evidence into practice. Age Ageing (2005) 34:98–100.
[Free Full Text] - Donaldson M, Khan K, Davis J, et al. Emergency department fall-related presentations do not trigger fall risk assessment: a gap in care of high-risk outpatient fallers. Arch Gerontol Geriatr (2005) 41:311–17.[Medline]
- Salter A, Khan K, Donaldson M, et al. Community-dwelling seniors who present to the emergency department with a falls do not receive guideline care and their risk profile worsens significantly: a 6-month prospective study. Osteoporos Int (2006) 17:672–83.[CrossRef][Web of Science][Medline]
- Kalula S, de Villiers L, Ross K, Ferreira M. Management of older patients presenting after a fall—an accident and emergency department audit. S Afr Med J (2006) 96:718–21.[Medline]
- Russell M, Hill K, Day L, Blackberry I, Dharmage S. The reliability and predictive accuracy of the falls risk for older people in the community assessment (FROP-Com) tool. Age Ageing. Advance Access published on June 19, 2008; doi:10. 1093/ageing/afn129.
- Kellog International Working Group on Prevention of Falls by the Elderly. The prevention of falls in later life. Danish Med Bull (1987) 34:1–24.
- Youden W. Index for rating diagnostic tests. Cancer (1950) 3:32–5.
- Kirkwood B, Sterne J. Essential medical statistics: 2nd edition (2003) Malden, USA: Blackwell.
- Sackett D, Haynes R, Guyatt G, Tugwell P. Clinical epidemiology. A basic science for clinical medicine (1991) Boston: Little, Brown and Company. 2nd edition.
- Shrout P, Fleiss J. Intraclass correlations: uses in assessing rater reliability. Psychol Bull (1979) 86:420–8.[CrossRef][Web of Science][Medline]
- Close J, Hooper R, Glucksman E, Jackson S, Swift C. Predictors of falls in a high risk population: results from the prevention of falls in the elderly trial (PROFET). Emerg Med J (2003) 20:421–5.
[Abstract/Free Full Text] - Haines T, Hill K, Walsh W, Osborne R. Design-related bias in hospital fall risk screening tool predictive accuracy evaluations: systematic review and meta-analysis. J Gerontol Biol Sci Med Sci (2007) 62:664–72.
[Abstract/Free Full Text] - Cwikel J, Fried A, Biderman A, Galinsky D. Validation of a falls-risk screening test, the Elderly Fall Screening Test (EFST), for community dwelling elderly. Disabil Rehabil (1998) 20:161–7.[Web of Science][Medline]
- Stalenhoef P, Diederiks J, Knottnerus J, Witte L, Crebolder H. The construction of a patient record-based risk model for recurrent falls among elderly people living in the community. Fam Pract (2000) 17:490–6.
[Abstract/Free Full Text] - Stalenhoef P, Diederiks J, Knottnerus J, Kester A, Crebolder H. A risk model for the prediction of recurrent falls in community-dwelling elderly: A prospective cohort study. J Clin Epidemiol (2002) 55:1088–94.[CrossRef][Web of Science][Medline]
- Nandy S, Parsons S, Cryer C, et al. Development and preliminary examination of the predictive validity of the Falls Risk Assessment Tool (FRAT) for use in primary care. J Public Health (2004) 26:138–43.[CrossRef]
- Covinsky K, Kahana E, Kahana B, Kercher K, Schumacher J, Justice A. History and mobility exam index to identify community-dwelling elderly persons at risk of falling. J Gerontol Med Sci (2001) 56A:M253–9.
[Abstract/Free Full Text] - Pluijm S, Smit J, Tromp A, et al. A risk profile for identifying community-dwelling elderly with a high risk of recurrent falling: results of a 3-year prospective study. Osteoporos Int (2006) 17:417–25.[CrossRef][Web of Science][Medline]
- Tromp A, Pluijm S, Smit J, Deeg D, Bouter L, Lips P. Fall-risk screening test: a prospective study on predictors for falls in community dwelling elderly. J Clin Epidemiol (2001) 54:837–44.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||