U.S. flag

An official website of the United States government

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Michael YL, Lin JS, Whitlock EP, et al. Interventions to Prevent Falls in Older Adults: An Updated Systematic Review [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2010 Dec. (Evidence Syntheses, No. 80.)

Cover
  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Interventions to Prevent Falls in Older Adults

Interventions to Prevent Falls in Older Adults: An Updated Systematic Review [Internet].

Show details

4Discussion

Summary of Review Findings

We evaluated 47 randomized controlled trials (n=23,980) testing primary care interventions to prevent falling among community-dwelling older adults against minimal or no-treatment control groups. This represents a substantial body of research on interventions to prevent or reduce falls published since the 1996 USPSTF recommendation. Furthermore, this remains a very active area of international research. We did not include comparative effectiveness trials in our review since they do not provide data on the absolute effectiveness of interventions to prevent falling compared with not intervening. To allow synthesized consideration of the evidence for potential types of interventions, we have organized the Discussion by intervention type rather than by key question. A summary of the overall evidence is provided in Table 15.

Table 15. Summary of Evidence By Key Question.

Table 15

Summary of Evidence By Key Question.

Multifactorial Assessment and Management

We evaluated 14 multifactorial assessment and management trials (n=5,570) with 16 intervention arms conducted in community-dwelling older adults aged 73 to 81 years on average. We found evidence that the most comprehensive interventions that provided medical and social care based on assessment results were more consistently associated with a significant benefit (random effects RR, 0.75 [95% CI, 0.58 to 0.99]) (I2=86.4%). After excluding one outlier, the statistical heterogeneity was moderate (I2=44.4%) and the relative risk was attenuated by 11% but remained marginally significant. Limited evidence suggests that fall-related fractures and disability were reduced. The trials would need more subjects to detect a statistically significant difference in rare outcomes, such as fracture risk, if one existed. These trials also provided limited evidence of other positive outcomes, such as maintenance of falls efficacy and improvements in functional limitations. Firm conclusions are difficult to draw since these outcomes were not consistently reported in this literature. The evidence is adequate that there are not serious harms associated with multifactorial assessment with comprehensive management of identified risks. Minor harms, including paradoxical increased falls and musculoskeletal symptoms, were identified for multifactorial assessment and management programs.

The challenges to providing these comprehensive programs as part of primary care are substantial, and include barriers for both clinicians and payers.50,51 Barriers for clinicians include patient compliance, care fragmentation and lack of coordination, and lack of knowledge and skills.50 Additional barriers to fee-for-service Medicare coverage include cost of services, concern about fraud, legislative limitations, and complex financing structure.51

Single Clinical Treatment

Vitamin D. We evaluated eight trials (n=5,216) of vitamin D supplementation conducted in community-dwelling older adults aged 71 to 77 years on average. While our report found no evidence that vitamin D supplementation (with or without calcium) affects fall-related fractures, we found evidence that vitamin D can effectively reduce the risk for falling (RR, 0.83 [95% CI, 0.75 to 0.91]). One study also reported a statistically lower rate of falls per year associated with vitamin D supplementation.86 The substantial range across trials in types and delivery of vitamin D results limits our ability to compare dosages and determine any threshold dose effect. Since almost all studies showed some effect consistent with benefit, we conclude that we cannot specify a threshold more informative than those coming from dietary reference intakes. There do not appear to be significant clinical harms associated with vitamin D supplementation.

Vision correction. We evaluated four trials (n=1,437) of vision correction conducted in community-dwelling older adults aged 78 to 83 years on average. Our report found no evidence that vision correction can effectively reduce fall-related fractures or risk for falling in populations selected for risk for falling. A single trial reported a significant reduction in fall rate associated with expedited first surgery to correct eye cataracts.87 Falls efficacy was improved (i.e., fear of falling was reduced) as a result of these interventions. Harms associated with vision correction interventions may include a paradoxical increased fall risk.

Medication withdrawal. We evaluated one trial (n=48) of medication withdrawal conducted in community-dwelling older adults with an average age of 75 years who are taking psychotropic medications. In addition, we examined three additional multifactorial assessment and management trials that assessed medication use and provided appropriate intervention. The evidence that medication withdrawal alone reduced the rate of falling was inconclusive. We found no evidence to suggest serious harms of medication withdrawal.

Protein supplementation. We evaluated one trial (n=50) of protein supplementation among frail community-dwelling older adults with an average age of 79 years. Our report found inconclusive evidence that protein supplementation reduced risk for falling. There was no evidence to suggest serious harms of protein supplementation.

Clinical Education/Behavioral Counseling

We evaluated one trial (n=310) of high-intensity behavioral counseling conducted in high-risk community-dwelling older adults. An additional nine trials incorporated low- to high-intensity educational components into a multifactorial assessment and management, single clinical treatment, or home hazard modification intervention.62,69–71,77,78,80,83,112 Our report found no evidence that fall prevention interventions that only included educational and counseling strategies resulted in a reduced risk for falling. There was no evidence to suggest serious harms of clinical education and counseling.

Physical Activity

We evaluated 17 trials (n=3,985) of exercise or physical therapy interventions conducted in community-dwelling older adults. While our report found limited evidence of other health benefits associated with exercise or physical therapy interventions, we found evidence that these interventions reduced risk for falling. The pooled relative risk for exercise or physical therapy interventions was 0.86 (95% CI, 0.80 to 0.92), with little statistical heterogeneity (I2=5.4%). When we stratified the control groups by rate of falling (>35% vs. <35%, the general community rate of falling), we found that interventions appeared to be primarily effective in those at increased risk for falls. No other differences in benefit were observed based on fall-risk status. Limited evidence suggests that functional limitations improved as a result of these interventions. Firm conclusions are difficult to draw since these outcomes were not consistently reported in this literature, with no more than three trials reporting any other positive outcome. No serious harms were identified for exercise or physical therapy programs.

Home Hazard Modification

We evaluated three home hazard modification trials (n=2,348) including community-dwelling older adults. An additional five trials69,75–77,79 (n=1,643) included home-hazard assessment and modification as part of a multifactorial assessment and management intervention. We found limited evidence that home-hazard modification reduced the risk for falling among community-dwelling populations selected based on fall risk factors. There was no evidence to suggest serious harms associated with home hazard modification. One home-hazard modification trial was excluded because it compared the effectiveness of home-hazard modification plus behavioral counseling with home hazard modification alone.123 This large study (n=3,182) recruited community-dwelling adults aged 65 years or older, unselected for fall risk, from a managed care organization in the northwest. Participants randomized to the intervention group were provided with assistance to modify the home hazards identified in the home assessment and also attended a moderate-intensity falls-prevention education program (90-minute classes for 4 weeks). The control group participants received the home hazard assessment with minimal followup. The intervention participants had a significantly reduced risk for falling during the 23 months of followup (39%) compared with the control group (44%). The control treatment in this comparative effective trial was similar to the interventions provided in the two included home hazard modification trials that reported no significant reductions in fall risk.61,70 Additionally, similar to the nonsignificant home hazard modification trials,61,70 the population included in the comparative effectiveness trial was unselected, in contrast to the higher-risk population selected for the successful home hazard modification.63 Further research on these more-intensive home hazard modifications is needed.

Comparison With Other Reviews of Interventions to Prevent Falls

While our results are similar to previous systematic evidence reviews and meta-analyses,53–55 they do differ in some details. Other relevant recent systematic evidence reviews evaluating specific types of interventions (e.g., hip protectors,56 multifactorial assessment,57 and exercise58) also included institutionalized populations. Unlike our review, prior reviews (except one53) included institutional and hospitalized populations. The specific purpose of the current review was to evaluate primary care-based clinical approaches to fall prevention, a narrower focus than any of these earlier reviews.

Given the difference in scope, we will focus on the comparison of the current review with the Cochrane review of interventions for preventing falls in older people living in the community.53 Unlike the current review, Cochrane included comparative effectiveness trials. We included 41 of the 111 trials reviewed in the Cochrane review. The most common reasons for exclusion of studies reported in the prior reviews were quality, study design (generally comparative effectiveness trials), or population (not comparable with primary care). These reasons are detailed in Appendix B Table 2. We included five trials not included in the 2009 Cochrane review—two studies of hip protectors,66,85 two studies of vitamin D,89,99 and one study of Tai Chi exercise114—most of which were published after the Cochrane review’s search period ended.

Similar to the 2009 Cochrane review and meta-analyses, we found no overall reduction in fall risk when all of the 14 multifactorial assessment and management trials were pooled. Unlike the 2009 Cochrane review, however, the comprehensiveness of these interventions was a significant predictor of success. We found a significantly reduced risk for falling when the analysis was limited to the most comprehensive multifactorial assessment and management interventions. One recent systematic review and meta-analysis of multifactorial clinical assessment programs also reported an absence of an overall benefit,57 while another reported a significant reduction in fall risk associated with these interventions.52 The absence of an overall benefit may result from combining multifactorial assessment and management strategies that provide direct intervention with those studies that primarily provide referral.124,125 However, the characteristics of a comprehensive multifactorial assessment and management intervention have not been clearly defined, and different approaches to classification may also lead to different results. For example, while we agreed with the 2009 Cochrane review in the majority of studies that we classified as comprehensive, one of the studies that we classified as comprehensive was classified as noncomprehensive in the 2009 Cochrane review.82 Clarifying the components of a comprehensive multifactorial assessment and management intervention is an important topic of future research.

Similar to the 2009 Cochrane review and meta-analyses, we conclude that exercise programs are effective overall, as did Chang and colleagues55 and Sherrington.58 We found that exercise/physical therapy interventions for community-dwelling older adults may be particularly effective in participants at higher risk for falls based on fall risk among the control group during followup. As in the current review, the 2009 Cochrane review evaluated fall risk at baseline based on history of falling or one or more risk factor for falls at enrollment and found no difference in pooled estimates; it did not evaluate fall risk based on the control group during followup. Sherrington and colleagues reported the opposite finding (more effective among participants at lower risk for falls), but this is likely explained by the inclusion of more frail institutionalized populations in their review.58 We did not sort exercise interventions by components and location, as was done in the 2009 Cochrane review. This review concluded that community-based group exercise interventions, individualized home-based exercise programs with multiple components, and Tai Chi were effective.53

Unlike the 2009 Cochrane review and meta-analyses, we found that vitamin D supplementation was consistent with a significantly reduced risk for falling. We include data from an additional two trials that were not included in the Cochrane review and are generally protective. Also, we did not find a benefit of medication withdrawal outside of comprehensive multifactorial assessment and management. We agree with the Cochrane review that home-hazard assessment and modification interventions did not reduce fall risk.

Harms

Overall, there do not appear to be significant clinical harms associated with effective interventions to prevent falls in older adults—multifactorial assessment and management including direct provision of medical and social care, vitamin D supplementation, and exercise and physical therapy. For interventions without evidence for effectiveness, it appears that harms are small for vision correction in frail older adults.

We did not systematically review the evidence on the harms of vitamin D, vision screening, or early vision correction in older adults because of the availability of recent AHRQ-funded evidence reports. The effectiveness and safety of vitamin D have been recently reviewed by the University of Ottawa Evidence-based Practice Center.126 This review of 19 vitamin D trials in adults found that there was limited evidence that vitamin D intake above current dietary reference intakes is harmful. However, most trials of higher doses of vitamin D were not adequately designed to assess long-term adverse effects. Daily doses ranged from 400 to 4000 IU of vitamin D3 or 5000 to 10,000 IU of vitamin D2. In most trials, reports of hypercalcemia and hypercalciuria were not associated with clinically relevant events. The Women’s Health Initiative reported a 17% increased risk for kidney stones in women aged 50 to 79 years whose daily vitamin D3 intake was 400 IU combined with 1000 mg calcium. Details are available in the full evidence report126 or the original research report.127 In addition, there are currently two ongoing calcium and vitamin D reviews evaluating the harms of vitamin D supplementation. One review is funded by AHRQ and is projected to be completed in June 2009.128 The other review is funded by the Institute of Medicine and seeks to redefine dietary reference intakes; it is projected to be completed in May 2010.129 Finally, the harms of vision screening and early vision correction in older adults have been recently addressed by a separate USPSTF report.130 This report found very sparse evidence for harms of vision screening or early treatment of visual impairment in older adults. In this review, none of the screening studies in primary care settings evaluated potential harms. Harms associated with eyeglasses were limited to a single small observational study showing an association between multifocal lens use and an increased risk for falls (adjusted OR, 2.09 [95% CI, 1.06 to 4.92]). Harms associated with other treatments for uncorrected refractive error were also limited, but included a low incidence of clinically significant harms, such as infectious keratitis, corneal ectasia, and a long-term complication of cataract surgery, posterior capsule opacification. Details are available in the full evidence report.130 Of note, the cataract surgery intervention trials included in this review provide limited evidence for harms since they compared expedited surgery to usual care, in which both groups received cataract surgery.87,91

Contextual Issues

Identification of Persons for Evidence-Based Interventions to Prevent Falls

A practical question facing primary care clinicians is how to feasibly and effectively identify the community-dwelling older adults who are appropriate for falls interventions. Epidemiologic studies demonstrate that fall risk increases dramatically as the number of risk factors increase.122,131 However, it is challenging to translate these findings into a strategy for primary care clinicians to reliably identify persons at risk for falling. The literature we reviewed does not provide clear direction as to how to proceed.

Among the 41 intervention trials we reviewed, few (12%) enrolled unselected older persons. While some (20%) selected persons only on the basis of age (70–80 years or older), the majority of studies (68%) enrolled participants pre-selected for increased risk factors for falls, including history of falls, gait and balance impairment, clinical history (such as stroke, Parkinson’s disease, recent hospitalization, or medication usage), clinical exam findings (e.g., visual defects), or were selected to be in need of the tested intervention (e.g., vitamin D deficiency in supplementation trials, hip fracture risk in hip protector trials). These intervention studies generally used noncomparable sets of self-reported or measured risk factors across a broad range, including more than 15 different domains to identify those at risk for falling (Table 14).

Among the included trials, falls history was the most common risk factor assessed other than age. The definition of falls history varied, with a history of at least one fall during the previous 2 to 12 months required for participation in three studies79,82,102,103 and a history of more serious falls required for four studies.66,77,80,84 Falls history was one of several risk factors assessed to qualify participants for four other interventions,71,89,93,100 although fallers did not make up even half of the selected participants in two of these studies.71,93 Use of falls history identified individuals along a spectrum of risk (as represented by the proportion of fallers in the control group in the subsequent year), even when supplemented by other risk factors (Appendix E Table 1).

Although effective interventions primarily addressed selected, higher-risk participants (or the benefits appeared to be primarily in this group), the methods for identifying higher-risk participants for these interventions also varied widely. When we examined a surrogate measure of actual falls risk among the selected study participants (as represented by the risk for falling in the control groups), we found that control-group fall risk in effective interventions were mostly 50% or greater, but at least exceeded the average community fall risk of 33% to 35% for the comprehensive assessment and exercise/physical therapy intervention types. For vitamin D interventions, benefits were seen even when the control-group fall rates were lower than “community” levels. Perhaps selecting participants for vitamin D supplementation should concentrate on older age (≥70 years) and vitamin D deficiency rather than falls history. Research demonstrates that myopathy associated with vitamin D insufficiency contributes to gait instability, increased body sway, and falls.132

To address the dilemmas raised by the diversity of falls risk assessment approaches in the literature, others have proposed relatively consistent, evidence- and expert opinion-based algorithmic approaches to identifying higher-risk participants for falls interventions.28,31,45,109 These authors all propose an approach that regularly assesses the frequency, context, and sequelae of falls during the previous year among older adults28 or beginning at age 6545,109 or 7031 years. One group suggests that, among those that have not fallen during the previous year, clinicians should ask about gait, balance, or mobility problems with either a positive history of falls or problems with gait, balance, or mobility determining elevated falls risk status.109 Two groups28,45 suggest that, after screening for a history of falls, those reporting a single fall28,45 or those considered to be at risk for falling45 should be observed or tested for balance and gait deficiencies in order to detect elevated falls risk status. Another variation suggests that all participants be observed for gait and balance difficulties as well as having their falls history elicited, but essentially identifies the same group of participants as at elevated risk (i.e., those with two or more falls or with balance or gait difficulties).31 These approaches all essentially agree that those selected as having an elevated risk for future falls by one of these brief screenings should undergo a more in-depth, multifactorial falls risk assessment, as should those presenting to the health care system for falls-related injuries or recurrent falls.28,45 The multifactorial falls risk assessments recommended by various groups for those at elevated risk were fairly consistent across a range of falls risk factors, including circumstances of previous falls,28,45 medical comorbidities,28 cardiovascular and neurological assessment,28,45 lower extremity joints and weakness,28,45 medication use,28,45,109 orthostatic hypotension,45,109 visual impairment,28,45,109 gait,28,45,109 balance28,45,109 and mobility concerns,45,109 impaired functional activities,28,45,109 environmental hazards,28,45,109 cognitive impairment,45,109 fear of falling,45 and urinary incontinence.45 This staged approach limits the receipt of the resource-intensive multifactorial falls risk assessment to those with the greatest risk. Among this group, the comprehensive assessment allows the clinician to define individual risk more precisely and to tailor interventions to the most important modifiable risk factors.

Recent systematic reviews have addressed the issue of risk factor assessment, but many questions remain for clinicians. One systematic review used multivariate analyses in prospective cohort studies to establish a clinically meaningful set of risk factors from among the large list of reported falls risk factors.45 While 24 studies reported multivariate analyses, design and reporting differences limited the review to reporting only the proportion of studies with statistically significant or insignificant findings for each risk factor. This approach did not effectively reduce the number of risk factors or prioritize the falls risk factors that clinicians should consider in community-dwelling adults (e.g., falls history, gait deficit, balance deficit, mobility impairment, fear, visual impairment, cognitive impairment, urinary incontinence, home hazards).

A recent systematic review attempting to overcome the challenges of identifying participants for evidence-based falls interventions explored an approach based on identifying a subset of individuals whose absolute falls risk would theoretically exceed 50%.109 These reviewers identified falls risk factors commonly used to identify participants for effective falls intervention trials (age, falls history, gait and balance impairment, orthostatic hypotension, medication usage, cognitive impairment, visual defects, limitations in basic or instrumental activities of daily living) and examined their prognostic value in 18 medium to large cohorts with prospective ascertainment of any or recurrent falls over 6 to 12 months. The review examined the independent contribution of risk factors after adjustment for other risk factors through multiple regression analyses. The most consistently studied risk factors in multivariate analyses were gait and balance impairment (15 studies), age (11 studies), history of falls (11 studies), medication use (11 studies), visual impairment (11 studies), limitations in functional activities (10 studies), cognitive impairment (11 studies), and orthostatic hypotension (four studies). Among these eight risk factors, only three (history of falls, certain medication use, and gait and balance impairment) provided independent prognostic value in at least half of the applicable multivariate studies. All 11 studies that evaluated a history of falls found that falls during the previous year predicted falls during the following year. Specific medications, such as benzodiazepines or other psychoactive medications, were associated with increased falls risk after multivariate adjustment in about half (six of 11) of prognostic studies. In contrast, cognitive impairment and limitations in activities of daily living were not associated with increased falls risk after adjusting for other falls risk factors in most analyses (two of eight and three of 10, respectively). Among the four remaining falls risk factors (orthostatic hypotension, visual impairment, age, gait and balance impairment), only gait and balance impairment were related to future falls risk in the majority of multivariate studies (10 of 15 applicable studies). Using the likelihood ratios generated from the univariate relationship between the risk factor and subsequent falls, the authors pointed out that in a population with a pre-test probability of falls of 19% to 36% (the “community” rate), any risk factor with a likelihood ratio of at least 2 would increase the post-test probability of falling to 50%. Using this approach, one to three risk factors would be important (history of falls, gait or balance impairment, and psychotropic medication/use of more than four medications). The unadjusted likelihood ratio for falls in the next year in those with previous falls ranged from 2.8 to 3.8. For those with gait or balance impairment, the unadjusted likelihood ratio was 1.7 to 2.4. For medication use, a likelihood ratio of 1.7 to 1.9 was generally associated with psychotropic medications or use of four or more medications.

Considering our review and others’, we can find no simple, validated way to identify participants most likely to benefit from evidence-based falls interventions. Clinicians may follow expert advice to screen based on falls history and simple gait and balance assessment. However, falls are the quintessential example of a clinical problem in which multiple small risks interact, and a problem for which different individuals will have different component risks as part of their risk profile.133 Thus, most current approaches attempt to apply population risk factors to risk-stratify groups of individuals for a clinical problem for which there are markedly different component risks for individuals. The most fruitful approach may be to individualize absolute risk, as has been done using the Framingham risk profile for coronary artery disease.134 Some researchers have attempted to construct risk indices for clinical prediction using multiple regression models,135–138 but these have rarely identified the same set of predictors due in part to differences in cohorts, types of falls outcomes predicted (e.g., any vs. recurrent falls over 1 year or more), and the range of falls risk factors considered. Furthermore, many of these studies do not provide the sensitivity and specificity or discriminant abilities of their risk prediction models. For tools that have determined a clinical index with reasonable sensitivity and specificity, the indices have generally not been validated using another population. Creating good risk-prediction models and tools that are applicable to primary care could be an important step forward in reducing falls among community-dwelling older adults, but its realization will require a series of coordinated research efforts. For example, a recent systematic review of fall risk assessment tools in community settings examined validity and reliability studies for 23 different tools as reported in 14 studies.139 Only three tools (Berg balance scale, functional reach test, and timed Get Up and Go test) were examined in more than one study. Of these, only the timed Get Up and Go test and functional reach test would be clearly feasible for primary care practitioners.

Cost-Effectiveness

There are very few studies examining the cost-effectiveness of interventions to prevent falls in older adults. Only four studies included in this report addressed cost-effectiveness, including two evaluating a comprehensive multifactorial assessment followed by direct provision of care (Yale Frailty and Injuries: Cooperative Studies of Intervention Techniques and VIP trials),63,140 one evaluating a community-based exercise program,141 and one evaluating cataract surgery.142 We also found two additional studies evaluating a nurse-delivered home exercise program in older adults that were not included in our report because they were based on a nonrandomized controlled trial.95,143 One of four cost-effectiveness studies was based on a trial conducted in the United States in the early 1990s,140 while the other three were based on trials conducted in New Zealand and the United Kingdom. Thus, cost-effectiveness estimates are not easily applied to the current U.S. health care system.

Overall, the costs per fall prevented varied widely, with lowest cost (per fall prevented) in community-based exercise interventions141 and highest cost (per fall prevented) in professionally-led in-home programs or comprehensive multifactorial assessment and management.95,140,143 From two cost-effectiveness analyses that allowed calculation of costs per fall and per serious fall, it appears that the costs to prevent a serious fall resulting in injury are approximately twice the costs of preventing any fall.95,141,143 It is difficult to compare these costs given the differences in cost valuation, country setting, and types of interventions evaluated. These analyses were generally well conducted and costs were based on costs incurred in the actual trials. However, all the analyses for these trials were based on a single trial that included moderate to small numbers of participants (approximately 300). Additionally, the cost analyses were limited to the time frames of the trials, which were at the most 2 years. Also, most of the analyses limited the costs incurred to the health care system, even those that stated that they used a societal perspective. We found only one cost-utility analysis using quality-adjusted life years (QALYs). This study evaluated the cost-effectiveness of a first eye cataract surgery in the United Kingdom from a societal perspective.142 Unlike the other cost-effectiveness analyses, this study modeled longer-term costs beyond the trial’s duration. The analyses found that the first eye cataract surgery was not cost-effective over the trial period (incremental cost per QALY, £35,704), owing to an increase in health care utilization during the 3 months post-surgery, but likely cost-effective over the participants’ remaining lifetime (incremental cost per QALY, £13,172). The applicability of this analysis given the difference in health care costs in the United Kingdom and the United States is also unclear. If cost information is important to weighing the evidence, original cost-effectiveness analyses are needed for the interventions deemed effective.

Limitations

Limitations in the body of evidence. Overall, the research on preventing falls in older adults is of fair quality. Concerns about this research include the impracticality of double blinding, failure to blind outcome assessors, significant attrition, less than ideal outcome measures, and heterogeneous treatment approaches. A major limitation of the existing evidence is the lack of data on important outcomes beside falls. Only 28 studies included any health outcomes, and only 13 reported fall-related fractures. The prevention of fractures, injuries, and other serious sequelae is the key reason for intervening to prevent falling.

Falls outcome measurement has improved, as recent trials are more likely to measure falls prospectively using diaries or calendars to minimize recall bias.144 Many trials, such as those that evaluated vitamin D supplementation or exercise interventions, were not sufficiently powered to observe a significant reduction in risk for falling without pooling. Thus, despite some trial evidence, some interventions (e.g., protein supplementation or medication withdrawal) have insufficient evidence. Current research incompletely reports how the trials affected other important outcomes (both harms and benefits, such as disability or functional limitations). Recently the Prevention of Falls Network Europe published a consensus document describing a common data set for fall prevention interventions; the routine use of these assessment instruments and procedures will enhance the quality and comparability of future trials as well as expand the available data on health outcomes and other positive outcomes.145 Although the consensus document does not address harms reporting, this is a critical need, particularly since harms were not systematically evaluated in the majority of fall prevention interventions in this review. Likewise, as the overall body of evidence is large, heterogeneity in the intervention approaches precludes the usefulness of one combined meta-analysis. Within these intervention types, specific limitations may also apply. For example, in the context of medication management by physicians, while use of psychoactive medications or a large number of medications is a clear risk factor for falls, there is not a clear model for clinicians to analyze overall medication use and to balance the benefits and harms of individual medications.146 The absence of such tools limits the effectiveness of interventions evaluating medication management.

While identification of those groups of older adults who are most likely to benefit would help to target labor-intensive interventions,147 inconsistency in the approaches used to identify populations at higher risk for falling makes it impossible to evaluate whether any single approach to identifying high-risk older adults is successful.31,109

While the older adults included in the interventions were heterogeneous with regard to age, fall risk, and overall health, they were homogeneous with regard to race/ethnicity and possibly socioeconomic status. The validity of these findings for nonwhite and lower socioeconomic status populations is an area for future research.

Limitations in our approach. Our review did not include questions examining specific components of the falls prevention programs that influenced the effectiveness of the programs. We limited the falls prevention interventions that we evaluated for the overarching evidence to trials that assessed falls; thus, it is possible that single clinical treatment trials with relevant health outcomes were not included if they did not also assess data on falls. Another limitation of our review is the narrow scope of the other positive outcomes of included fall-prevention interventions. It is possible that the included interventions resulted in other benefits that were not captured in our review. We included trials with varying lengths of followup (between 6 and 24 months) in our meta-analysis. However, the majority of trials assessed outcomes at 12 to 18 months. We used control-group fall risk in the subsequent year as a proxy for actual falls risk. While this measure is not a perfect proxy for what would happen without any contact, it allowed us to illustrate the apparent range of fall risks identified by those selecting based on falls history or other risk factors. However, use of control-group risk for stratifying results should be viewed as suggestive.148

Emerging Issues/Next Steps

Studies addressing the effectiveness of all available clinically-feasible instruments to identify populations at high risk for falling were beyond the scope of this review, but are an important area of research for clinicians.

Ongoing research identifying common modifiable risk factors for falling, such as vestibular dysfunction,149 should be incorporated in future reviews.

Future Research

While the number of studies on the effectiveness of interventions to prevent falls in older adults has dramatically increased since 1960,147 many research questions related to fall prevention among older persons identified in 1994131 remain active research questions today.

One outstanding question, relevant to most of the interventions studied in this review, is how to identify persons at high risk for falling. Currently there is no simple validated way to clinically identify community-dwelling older adults and subgroups of community-dwelling older adults most likely to benefit from evidence-based falls interventions. The development and validation of a standardized assessment of absolute fall risk would allow researchers to quantify individuals’ fall risk and then target persons at different levels of risk with appropriate interventions. Creating good risk-prediction models and tools applicable to primary care would be an important step forward in reducing falls among community-dwelling older adults, but its realization will require a series of coordinated research efforts. Few studies of effective falls prevention interventions also reported the impact of these interventions on fall-related fractures, injuries, utilization, quality of life, disability, and mortality. Thus, these results may reflect selective reporting, and further research is needed.

The effectiveness of certain types of falls prevention interventions remains unclear. Research is needed on the effectiveness of home-hazard modifications for noninstitutionalized populations, and the impact of increased intensity of these interventions. Research is needed to develop a clear clinical model for analyzing and reducing medication use, and the benefits versus harms of such medication withdrawal. Additional research is needed to clarify the specific elements of successful interventions. For example, clarifying the components of a comprehensive multifactorial assessment and management intervention is an important topic of future research. Similarly, the effectiveness of differing intensity levels of exercise/physical therapy interventions, and the most effective components of such programs, remains unclear and deserves further study. In addition, questions remain as to the harms of effective falls prevention interventions and whether there are certain subgroups in whom the harms of such interventions outweigh the benefits. Research is also needed on the cost-effectiveness associated with falls prevention interventions. Finally, further research is needed on the effectiveness of falls prevention interventions in nonwhite populations, and in populations of diverse socioeconomic status.

Conclusions

Falls prevention has been an area of active research since 1996. Falls are an important public health and clinical problem that will only increase as the U.S. population ages. Primary care relevant interventions have demonstrated beneficial effects on falling compared with no or minimal treatment. Specifically, comprehensive multifactorial assessment with direct provision of care, exercise and physical therapy, and vitamin D supplementation were associated with small to moderate reductions in fall risk, with estimates ranging from a 12% to 24% reduced risk. Only minor harms were identified for these interventions. Some evidence supports more robust effects on risk for falling when the interventions are targeted to those at high risk. Since evidence suggests that clinical interventions should target high-risk populations, further research on valid, reliable, clinically feasible tools to identify these populations is imperative. Despite reductions in fall risk, limited evidence supports a beneficial effect on health outcomes, including fall-related fractures, disability, and quality of life. Additional studies sufficiently powered to address these outcomes would be very beneficial. Tested interventions with no clear benefit in community-dwelling participants include visual deficit correction. Very limited evidence is available for protein supplementation or home hazard modification and clinical education or counseling alone.