A Systematic Review of the Guidelines and Delphi Study for the Multifactorial Fall Risk Assessment of Community-Dwelling Elderly

As falls are among the most common causes of injury for the elderly, the prevention and early intervention are necessary. Fall assessment tools that include a variety of factors are recommended for preventing falls, but there is a lack of such tools. This study developed a multifactorial fall risk assessment tool based on current guidelines and validated it from the perspective of professionals. We followed the Meta-Analysis of Observational Studies in Epidemiology’s guidelines in this systematic review. We used eight international and five Korean databases to search for appropriate guidelines. Based on the review results, we conducted the Delphi survey in three rounds; one open round and two scoring rounds. About nine experts in five professional areas participated in the Delphi study. We included nine guidelines. After conducting the Delphi study, the final version of the “Multifactorial Fall Risk Assessment tool for Community-Dwelling Older People” (MFA-C) has 36 items in six factors; general characteristics, behavior factors, disease history, medication history, physical function, and environmental factors. The validity of the MFA-C tool was largely supported by various academic fields. It is expected to be beneficial to the elderly in the community when it comes to tailored interventions to prevent falls.


Introduction
Approximately one-third of all people over 65 years of age experience at least one fall, and 15% fall at least twice in their lifetime. [1]. Falls are among the most common causes of injury to the elderly, and they can lead to physical disability, including fractures that result in long-term disability, and reduced exercise capacity; they can even be fatal [2]. The mortality rate for fall-related injuries was 61.6 per 100,000 United States residents aged ≥ 65 years in 2016 [3]. Falls associated with the elderly are also related to the financial burden, not only for the suffering patients but also the increased costs for elderly medical expenses in the health care system. In 2015, costs for falls to Medicare alone totaled over US$ 31 billion in the United States [4]. As falls affect physical, mental, and economic conditions, prevention and early intervention are necessary.
Although there is an increase worldwide in the falls associated with the elderly in the community, the integrated multi-factor assessment tools based on evidence are limited. The limitations of previous fall assessment tools involve the independent identification of physical, psychological, or environmental factors. There were several "physical function" instruments used in the assessment of the risk of falling, which were the Berg Balance Scale, the Timed Up and Go Test, and the Tinetti Balance elderly residing in the community. The type of outcome was factors and/or items of multifactorial fall risk assessment, and the type of study involved guidelines. The search terms are reported in Table S1.
First, two researchers (LWS and KJE) independently reviewed the titles and abstracts of the searched articles. Second, we reviewed the full manuscripts of eligible studies and recorded the reasons for exclusion for each study. The inclusion criteria were as follows: (a) studies in which research subjects were community-dwelling elderly defined as aged 65 and over, (b) studies in which research interventions had a multifactorial fall risk assessment, and (c) studies in which the evidence was based on guidelines only. Exclusion criteria were as follows: (a) studies in which research subjects were in facilities (e.g., hospitals or nursing homes), (b) studies in which research subjects had a specific disease (e.g., community-dwelling elderly with Parkinson's disease), (c) studies in which the guidelines had interventions but no assessment components, (d) studies not published in English or Korean; (e) studies that did not contain guidelines, and (f) studies for abstract or conference proceedings only.

Delphi Study
We conducted a Delphi study to facilitate consensus among Korean experts. Prior studies on the Delphi research method state that about 10 panelists were needed to minimize errors and maximize reliability or judged that 8-12 people were appropriate [21]. If the number of experts is too small, it is difficult to agree on an adequate number of topics, and if they are too many, it is a time-consuming process. We recruited eleven experts for the Delphi panel. However, nine experts agreed to participate, and two experts refused. All experts who participated in the study were informed about the aims of the study and provided informed consent.
To prepare for the first round, the research team developed indicators for each element of the multifactorial fall risk assessment tool among community-dwelling elderly that originated from the reviewed guidelines. When planning a Delphi study, we set the criteria for the end of the rounds as a completed round for the expert's consensus, and not as the number of specific rounds [21]. The first round was open. The first Delphi meeting with a multidisciplinary expert panel was held from October 13 to 26, 2016, by e-mail. Experts reviewed opinions about the appropriateness of classification; the necessity to add, correct, delete, and integrate the determinants identified in the systematic review; and the need to change their order. The validity of the Delphi technique was increased using qualified experts [22]. The expert group consisted of a total of nine Ph.D. experts, three geriatric medicine professors, two medical doctors, two nursing professors, one nurse, one police science professor, and all of them had previous fall-related research or practical experience for over five years.
We included scoring beginning with the second round. The second Delphi meeting with the same expert panel was held from 22 December, 2016, to 19 January, 2017, by e-mail. The mean, standard deviation, median, and interquartile range of experts' opinions about the necessity and applicability dimensions were presented in the questionnaires that followed each round. An expectation of the Delphi process was for the expert group to reach a consensus; this study reached a consensus among experts in the third round. During the three rounds of the Delphi questionnaires, data were collected by e-mail. The experts reviewed opinions and decided the appropriateness of the items. They considered reasons to add, correct, delete, and integrate the items from determinants, as well as changes to the order. In addition, the expert panel was asked to evaluate each item on a 5-point Likert scale (strongly disagree to strongly agree) along the two dimensions of necessity and applicability to the community-dwelling elderly. Data from each round were analyzed, and experts received feedback that presented information, including the written opinions and anonymous results of the ratings.
To select the components of the final questionnaires for the tool, we analyzed additional opinions from the panel of experts. The criteria chosen for scoring the survey were as follows: content validity ratio (CVR) ≥ 0.78 (minimum value for nine panelists), degree of consensus (DoCs) ≥0.75, and degree of convergence (DoCv) ≤0.50. Cronbach's alpha test was used to determine internal consistency when the criteria were scored higher than 0.7. Furthermore, to evaluate stability, only items with coefficients of variation (CV) of 0.80 or more were deleted [22]. Self-assessment of the research design was conducted to ensure the quality, all of which met its standards. The questions were, "What criteria will be used to determine which items to drop?" and "What criteria will be used to determine to stop the Delphi process?" [21] 3. Results Figure 1 shows an updated flow chart of the search results, and the previous chart is reported in Figure S1. After updating the search for guidelines, one guideline was added [23]. Of the 2072 articles retrieved by our database search, 92 were selected based on the titles and abstracts. We included a total of nine articles describing guidelines for multifactorial fall risk assessment among community-dwelling elderly [24][25][26][27][28][29][30][31]. The included guidelines are described in Table 1. The nine guidelines are classified by country: two were from Canada [29,30], one from Australia [24], one from Ireland [27], one from the United States of America [23], and the other four guidelines were not restricted by country [25,26,28,31]. Likewise, the participants' ages in nine of the guidelines were over 65 years. There were no gender restrictions in any of the guidelines. All nine articles were classified by the person who performed the assessment tool: one by the health care provider [28], one by the physical therapist [25], two by health professionals [24,26], one by community health workers [30], one by the primary health care teams [31], one by clinicians [23], and two were not identified [27,29]. The number of factors for each guideline was two to four.

Systematic Review and an Initial List of Potential Standards
The results of the quality assessment of guidelines, using the Appraisal of Guidelines for Research and Evaluation II (AGREE II), indicated that they ranged from 66.7 to 100.0% ( Table 2). The Australian Commission on Safety and Quality in Health Care guidelines scored highest on the overall assessment (100.0%), while all the other guidelines scored 66.7%. The six domain scores of the AGREE II were evaluated separately. The highest scored domain was the "Scope and Purpose" (83.0%), and the lowest scored domain was "Applicability" (36.5%). We discussed the results of the quality assessment and concluded that no guidelines would be excluded when conducting the Delphi study.
The initial factors and items that resulted from our systematic review and the discussion by the researchers are listed in Table 3. We excluded ethnicity (Race), thyroid dysfunction, hearing, risk-taking behavior, and weather and climate from the list of items through the systematic review, because they did not fit due to ambiguity. Altogether, eight items were selected for behavioral factors, 17 for biological factors, three for environmental factors, and two for general factors. Since the factors and items for fall risk in updated guidelines have not been newly added, the Delphi has not been implemented again.

Open Round
For the four factors and 30 items chosen, we performed the open round with a panel of experts (nine experts from five fields), providing their thoughts on the suitability of the Multifactorial Fall Risk Assessment Tool for Community-Dwelling Older People (MFA-C) in narrative form. The typical answers related to factors and items needed to be modified, added, reordered, integrated, or moved to other factors. As a result, four factors (behavior, biological, environmental, and general) were reclassified into seven factors (general characteristics, behavior factors, disease history, medication history, physical function, cognitive function, and environment factors), and the existing 30 items were reorganized according to these new factors. At this time, the disease history item was moved to the factor level, and 10 items were added and included in that factor (Table 4).

Consensus in Scoring Rounds
Nine experts from five fields participated in the scoring round. Through the first round (the open round), 39 items under six factors were suggested. The scoring round was conducted twice, and a total of three rounds (one open round and two scoring rounds) were completed in nine months.
In the second round (the first scoring round), expert panelists agreed on 33 out of 39 items (84.6%) ( Table 4). The scoring round comprised segments for the necessity and applicability of the scale to community-dwelling elderly. In the necessity segment, the expert panel agreed on CVR, DoCs, DoCv, and CV. In the applicability segment, the CVR value of the medication side effect in the medication history factor was less than 0.79. The low-income item of the general characteristics factor, vitamin D deficiency of the behavior factor, incontinence of the disease history factor, the medication side effect of the medication history factor, the cardiac function of the physical function factor, and the cognitive capacity of the cognitive function were all less than 0.75 for DoCs or higher than 0.50 for DoCv.
Of these six items that corresponded with the exclusion criteria, three items (low income, incontinence, and cardiac function) were re-included based on the expert panel's judgment. Additionally, all of the CVs were less than 0.80. However, another three items (mediation side effect, vitamin D deficiency, and cognitive capacity) were excluded from this round after reaching an expert consensus. The experts concluded that medication side effects and cognitive capacity were duplicated with the newly added items of the disease history factor. In addition to identifying vitamin D deficiency, a blood test had to be performed. However, the expert panel determined that it would be inappropriate for community workers to assess the risk of falls and that this would place an economic burden on the elderly. In the third round (the second scoring round), the panels reached 100.0% agreement (36 of 36), thereby concluding the scoring round. Therefore, the final version of MFA-C had 36 items in six factors (Table 5).    Notes: ADL = activities of daily living; TCAs = tricyclic antidepressants, SSRIs = selective serotonin reuptake inhibitors; TUG = time up and go test; ROM = range of motion; EKG = electrocardiogram, BP = blood pressure; HR = heart rates; CVA = cerebrovascular accident.

Discussion
We systematically reviewed previously distributed individual fall risk factors, thereby facilitating the potential prevention of and early intervention in falls through the development of a multifactorial assessment tool that can be applied practically in the community. To our knowledge, this is the first study to develop a fall risk assessment tool through the Delphi study in various fields based on systematic review results that include multiple fall risk factors in the guidelines published. Previous studies have shown that there are differences in the items for developing a fall risk assessment tool based on the varied experiences of nurses or physicians [32]. Representatively, the tool by the National Health Service (NHS) in Bristol comprises 13 items: history of falls, medications, postural hypotension, alcohol intake, nutrition and osteoporosis, vision, hearing, walking/gait, transfers, function, continence, environmental hazards, and cognition [18]. Compared with the tool provided by the NHS, our tool was developed with more comprehensive and detailed assessment items related to the risk of falling. For a more accurate and in-depth verification of effectiveness using our fall risk assessment tool, systematic reviews of guidelines and confirmation of various expert opinions were necessary.

Items Excluded from this Multifactorial Assessment Instrument
Among the final items presented in this study, we excluded a lack of vitamin D, medication side effects, and cognitive capacity, all of which were considered fall risk items in the existing eight guidelines. Several studies reported that vitamin D reduced the risk of falls, and one meta-analysis estimated a 20% reduction in fall risk through vitamin D supplementation in the elderly [33]. These studies posited that the correlation between low serum 25-hydroxyvitamin D (25(OH)D) and increased falls was due to the lack of 25(OH)D, which leads to muscle weakness and poor balance [34]. As a result, this could lead to decreased physical performance and aging [34]. However, it also indicates that vitamin D deficiency does not have a direct effect on falls, but somewhat weakens the musculoskeletal system, resulting in falls. In this study, the final fall risk assessment tool includes the musculoskeletal function item of the physical function factor. Therefore, the Delphi panelists excluded vitamin D from the risk assessment tool because it was a duplication. In addition, recent studies have shown that supplemental vitamin D did not prevent falls [35], nor did it have a significant correlation with falls [36]. Furthermore, the National Institute for Health and Care Excellence (2013) does not recommend the use of vitamin D for fall prevention because there is a lack of robust evidence regarding the required dosage or method of administration [16]. For this reason, the expert panelists determined that invasive and costly vitamin D testing to assess fall risk was inappropriate for the elderly.
Furthermore, two items (medication side effect and cognitive capacity) in the Delphi phase were excluded because they were considered to overlap with other items of the disease history factor. In particular, the medication side effect item in the existing guidelines did not list specific disease names; therefore, the use of the item to perform a fall risk assessment could reduce the reliability of the evaluation because the results would vary according to the person performing the evaluation.
In this study, only the "fear of falling" was identified as an item related to psychological characteristics. Recent studies have reported that fall-related psychological concerns directly affected falling and its complications [7]. Therefore, it is suggested that psychological characteristics related to falls be summarized and organized for future study.

Additional Items in This Multifactorial Assessment Instrument
Most previous guidelines were developed to describe the past disease history, name of the drugs, and environmental risk of falls in an open ended form question. We tried to organize the list of items correctly to increase the concordance rate of the data analysis even though the person who assesses fall risks varies. This is significant in improving the reliability of this tool compared to other tools.
First, after reflecting on the opinions of experts in various academic fields, new items were added under the disease history factor that had not appeared in previous guidelines. The Delphi panelists thoroughly reviewed the specific factors and items and gave specific opinions on each. Disease history includes these items: stroke, dementia, Parkinson's, cardiovascular disease, respiratory disease, peripheral neuropathy, diabetes, chronic pain, arthritis, and osteoporosis. Therefore, our study differs from a guideline that includes only a few medical history items such as osteoporosis, depression, and cardiac disease [25]. We identified diseases that affect falls based on evidence and expert opinions and added them to our multifactorial assessment tool.
Regarding the relationship between falls and disease, neurological diseases such as strokes, dementia, Parkinson's, and peripheral neuropathy are traditionally associated with aging. These conditions might share common cognitive dysfunctions that affect the control of gait and balance [37]. They can limit complex and goal-oriented activities requiring the constant awareness of body movements [38]. Second, some studies identified that cardiovascular diseases in the elderly also increased the risk of falls [39] because the elderly are generally frail with noticeable cognitive decline and multi-morbidity [39]. Similarly, diabetes, arthritis, osteoporosis, and chronic pain are diseases or symptoms with high correlations with the types of fractures that are the most common outcomes of falls [40][41][42].
Moreover, hypoglycemia is the most significant cause of fall episodes [42]. A recent study reported that the adjusted odds of fall-related fractures among patients with hypoglycemic events were 70% higher than in patients without it [43]. These studies consider one explanation to be certain diabetes medications that may increase the risk of fracture and thereby worsen fall-related outcomes [44].
Additionally, arthritis and osteoporosis can lower vitamin D and bone mineral density. Both have been frequently suggested as factors that heighten the risk of bone fracture and falling [40]. Additionally, recent literature reported that elders with multisite pain had a 51% higher chance of fall risk [41]. Research has suggested that those with pain have excessive psychological concerns regarding low balance confidence, reduced self-efficacy of falling, and have mobility limitations such as slower gait pattern and difficulties in activities of daily living (ADL) [45].
Second, we specifically evaluated the use of a wider range of drugs than those included in the existing guidelines-particularly, psychoactive and cardiovascular drugs. Our study included a separate process of sorting and merging related medicines based on the Delphi expert panels. As a result, health care providers received a more comprehensive review of the drugs that affect falls in the elderly. We added those medication names to the medication history factor.
Finally, in our study, experts who participated in Delphi also considered the assessment items related to the residential environment. Based on their recommendations, we added concrete environmental items such as light, carpet, and height of the bed to the residential environment factor.

Limitations and Strengths
Publication limitations may have been present due to the inclusion of English and Korean-only published guidelines. Additionally, our study has a limitation related to validity. Among the methods to confirm the validity of the tool, only expert validity was used. Face validity was not applied. To overcome this problem, we collected the opinions of various fields related to falls and verified validity in various ways by calculating DoCv, and DoCs as well as CV and CVR. This is demonstrated clearly in various factors affecting the falls of the elderly based on worldwide guidelines. Most of the fall risk screening instruments found in the literature tend to focus on one single risk factor [6,46].
Additionally, evidence-based guidelines are developed to assist the practitioner, community residents, and policymaker to make informed clinical decisions [19,47]. Guidelines are valuable resources that play an integral role in improving the intervention and management of various health conditions. We clarified why we extracted each fall risk item based on evidence and expert opinions.
This research gathered all existing factors and filled in missing factors related to falls by collecting various expert opinions. This study increased its validity by adding expert opinions gathered through Delphi studies, in addition to a systematic review method. In this study, the strength of our research was the breadth of expertise within our multidisciplinary panel. These experts thoroughly reviewed the selected guidelines and provided professional opinions on all specific factors and items. Our multifactorial fall risk assessment tool will help to determine proper fall prevention interventions for the elderly in communities.
We clarified why we extracted each fall risk item based on evidence and expert opinions. Conversely, most tools did not describe the criteria for classifying the fall risk items as factors [46,48]. Therefore, the items affecting fall risk that were included in other guidelines were different for each tool. This tool was developed by a thorough, evidence-based approach through the Delphi study and built upon existing guidelines, and so it can be used universally in any country.
All the included guidelines can be internationally used because they did not reflect the situation of a specific country. Therefore, it is necessary to confirm the generalizability of using the tool by identifying whether each multifactorial fall assessment tool has been translated into the language of each country and verifying its validity.

Conclusions
Health care providers can use comprehensive falls risk screening tools to identify the elderly who are at risk of falling. We developed a multifactorial fall risk assessment tool based on evidence, assessing general characteristics, behavior factors, disease history, medication history, physical function, and environmental factors that reflect the characteristics of the elderly in a community. Although there were existing guidelines, the multifactorial risk factors for falls suggested by each guideline were inconsistent. Therefore, this study attempted to reach a consensus. This study increased the validity of our tool by adding expert opinions gathered through Delphi studies in addition to a systematic review method. This multifactorial fall risk assessment tool, created through this systematic methodology, is expected to be beneficial to the elderly in the community when designing comes to tailored interventions to prevent falls.