You Are Here:
A Systematic Evidence Review for the U.S. Preventive Services Task Force
Release Date: December 2010
By Yvonne L. Michael, ScD, SM; Evelyn P. Whitlock, MD, MPH; Jennifer S. Lin, MD, MCR; Rongwei Fu, PhD; Elizabeth A. O'Connor, PhD; and Rachel Gold, PhD, MPH
The information in this article is intended to help clinicians, employers, policymakers, and others make informed decisions about the provision of health care services. This article is intended as a reference and not as a substitute for clinical judgment.
This article may be used, in whole or in part, as the basis for the development of clinical practice guidelines and other quality enhancement tools, or as a basis for reimbursement and coverage policies. AHRQ or U.S. Department of Health and Human Services endorsement of such derivative products may not be stated or implied.
This article was first published in Annals of Internal Medicine on December 21, 2010 (Ann Intern Med 2010;153:815-825; http://www.annals.org).
Background: Falls among older adults are both prevalent and preventable.
Purpose: To describe the benefits and harms of interventions that could be used by primary care practitioners to prevent falling among community-dwelling older adults.
Data Sources: The reviewers evaluated trials from a good-quality systematic review published in 2003 and searched MEDLINE, the Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, and CINAHL from the end of that review's search date to February 2010 to identify additional English-language trials.
Study Selection: Two reviewers independently screened 3423 abstracts and 638 articles to identify randomized, controlled trials (RCTs) of primary care-relevant interventions among community-dwelling older adults that reported falls or fallers as an outcome. Trials were independently critically appraised to include only good- or fair-quality trials; discrepancies were resolved by a third reviewer.
Data Extraction: One reviewer abstracted data from 61 articles into standardized evidence tables that were verified by a second reviewer.
Data Synthesis: Overall, the included evidence was of fair quality. In 16 RCTs evaluating exercise or physical therapy, interventions reduced falling (risk ratio, 0.87 [95% CI, 0.81 to 0.94]). In 9 RCTs of vitamin D supplementation, interventions reduced falling (risk ratio, 0.83 [CI, 0.77 to 0.89]). In 19 trials involving multifactorial assessment and management, interventions with comprehensive management seemed to reduce falling, although overall pooled estimates were not statistically significant (risk ratio, 0.94 [CI, 0.87 to 1.02]). Limited evidence suggested that serious clinical harms were no more common for older adults in intervention groups than for those in control groups.
Limitation: Interventions and methods of fall ascertainment were heterogeneous. Data on potential harms of interventions were scant and often not reported.
Conclusion: Primary care-relevant interventions exist that can reduce falling among community-dwelling older adults.
Primary Funding Source: Agency for Healthcare Research and Quality.
Falls are a serious threat to the lives, health, and independence of older adults. Falls are caused by complex interactions among multiple risk factors, which are characterized as intrinsic (patient related) or extrinsic (external to the patient) (1-3). Between 30% and 40% of community-dwelling persons 65 years or older fall at least once per year (4-5). Falls were the leading cause of fatal and nonfatal injuries among persons 65 years or older (6). The death rate due to falls is 10 per 100,000 persons for those aged 65 to 74 years and 147 per 100,000 persons for those aged 85 years or older (7). The estimated direct medical costs for fatal and nonfatal fall-related injuries among community-dwelling persons 65 years or older in 2000 was $19.2 billion (8), with one study estimating that this cost could reach $43.8 billion by 2020 (9).
Falls among older adults are preventable (8, 10). In 2006, the American Geriatrics Society and the British Geriatrics Society published an updated evidence-based practice guideline recommending that older adults at high risk for falls receive a multifactorial fall-risk assessment and individualized, targeted interventions to address the risks and deficiencies identified in the assessment (8). Physicians face significant barriers to intervening to prevent falls, however, including lack of awareness and appropriate knowledge, competing risks, and difficulty assessing risk (11-13). Therefore, we conducted a systematic review of outpatient interventions available to primary care clinicians to prevent falls in older adults to support the U.S. Preventive Services Task Force (USPSTF) recommendation process.
We followed a standard protocol for this review. The USPSTF gave guidance on formulating 4 key questions (Appendix Figure 1), 2 of which are addressed in this review: Do primary care interventions reduce risk or rate of falls or fallers among community-dwelling older adults? What are the adverse effects associated with interventions to reduce falls or fallers? We grouped these interventions into 5 main categories: multifactorial assessment and management; single clinical treatment (vitamin D, vision correction, and medication management); clinical education or behavioral counseling; home-hazard modification; and exercise or physical therapy. Expanded definitions for the categories are provided in Appendix Table 1, and a more detailed description of our methods is available in our full report (14).
Searches and Selection Process
We searched multiple databases (MEDLINE, Cochrane Database of Systematic Reviews, the Database of Abstracts of Reviews of Effects, and Health Technology Assessments) and Web sites (Institute of Medicine, the Agency for Healthcare Quality and Research [AHRQ], and National Institute for Health and Clinical Excellence) to identify relevant, good-quality systematic reviews published between January 1991 and October 2007. We found 13 existing systematic reviews for interventions to prevent falls. Using guidelines for integrating systematic reviews (15), 2 independent reviewers assessed relevancy (research questions and scope) and quality of identified reviews. We used citations from 1 recent good-quality review of fall interventions (16) and then searched MEDLINE, the Cochrane Central Register of Controlled Trials, and CINAHL from the end of the previous review's search date of January 2002 to February 2010 to identify additional trials. We searched MEDLINE and CINAHL from 1992 (the earliest publication date of the included trials) through February 2010 to locate studies of harms for the included interventions. Our review of harms of vitamin D supplementation and vision screening or early treatment was limited to previously synthesized evidence (17-18). We included randomized, controlled trials (RCTs) of community-dwelling older adults (average age ≥65 years) in settings generalizable to U.S. primary care populations. We included trials if they were designed to assess fall prevention based on assessment of falling or falls as a primary or secondary outcome. We excluded trials not conducted in primary care or other settings with a primary care-comparable population (for example, hospitals, nursing homes, rehabilitation centers, or other long-term care facilities) and trials without a true control group. Inclusion and exclusion criteria and search strategies are available in the full report (14).
Quality Assessment and Data Abstraction
Two investigators independently screened all abstracts for inclusion. We screened 3423 abstracts and 638 full-text articles (Appendix Figure 2). Two investigators independently critically appraised all articles by using the USPSTF quality criteria (19). In brief, the USPSTF quality criteria for RCTs includes consideration of assembly and maintenance of comparable groups; differential loss to follow-up or overall high loss to follow-up; a clear definition of the intervention; equal, reliable, and valid outcome measurement; and intention-to-treat analysis. Good-quality and fair-quality articles meeting the inclusion criteria were retained, and studies with fatal flaws were rated poor quality and not included. Discrepancies were resolved through consultation with a third investigator. One reviewer abstracted relevant information into standardized evidence tables for each included article, and a second reviewer checked the abstraction.
Data Synthesis and Statistical Analysis
We qualitatively synthesized the included trials and summarized the results in tables, stratifying the discussion of evidence by similar intervention groupings (for example, physical activity, vitamin D, vision correction, and multifactorial assessment and management). When possible, we also quantitatively pooled fall-related outcomes to estimate the effect size of these intervention groupings. Falling was assessed in a variety of ways (number of fallers, fall rate, time to first fall, and number of frequent fallers). Number of fallers was the most consistently assessed measure of falling; the other measures were used more selectively. Thus, primary analyses estimated relative risk for falling by using random-effects models (20-21).
We conducted separate analyses for each intervention category. For single clinical treatments, the analyses were further stratified by treatment type. For trials with multiple intervention groups in which the interventions were variations of the same intervention type (for example, 2 exercise programs), we calculated estimates for the more intense groups. We assessed the presence of statistical heterogeneity among the studies by using standard chi-square tests and estimated the magnitude of heterogeneity by using the I2 statistic (22). A series of random-effects meta-regression models were used to examine potential sources of heterogeneity in fall risks; such sources include mean age, average age of 80 years or older, proportion of women, proportion of participants with a history of falling in the previous year, comprehensiveness or intensity of the intervention, and whether the sample comprised high-risk participants. All meta-analyses were done by using Stata software, version 10.1 (StataCorp, College Station, Texas).
Role of the Funding Source
AHRQ funded this work, provided project oversight, and assisted with internal and external review of the draft evidence synthesis. The authors worked with 3 members of the USPSTF to develop the analytic framework and resolve issues involving the scope of the review. The draft systematic review was reviewed by 8 external peer reviewers, then revised for the final version. Agency approval was required before this manuscript could be submitted for publication, but the authors are solely responsible for the content and the decision to submit it for publication.
We included 54 RCTs (with a total of 26,102 participants) that tested primary care interventions to prevent falling. Only 41 of the 111 trials from the Cochrane review (16) were included in our review, owing to differences in inclusion criteria. The most common reasons for exclusion from our review were that a true control group was lacking, the population studied was institutionalized or recruited from inpatient settings, and interventions could not be conducted in primary care or be referred to from primary care. Evidence tables and tables of excluded studies for each key question are available in the full report (14). Details of each included study are summarized in Appendix Tables 2 to 6. The Table shows a summary of evidence by intervention category.
Multifactorial Assessment and Management
We evaluated 19 multifactorial assessment and management trials (7099 participants) with 21 active intervention groups (23-41). All trials except 1 were limited to populations at high risk for falling (36). Most interventions assessed primary risk factors for falls that were identifiable during clinical evaluation: medication use, visual acuity, home environment, and gait and balance. On the basis of evidence that multifactorial assessment interventions that incorporate comprehensive management strategies are more successful in preventing falls (13), interventions were categorized during the abstraction phase by 2 independent investigators as comprehensive (complete and active management of fall risk factors and conditions identified in the multifactorial assessment, including provision of case managers or home nurses) or noncomprehensive (provide only partial or limited management of identified fall-related risk factors). Control groups primarily received usual care, although 2 trials used "attention control groups" (30, 32) and 1 trial provided "a minimal intervention of mainly education" as a control group (42).
The majority of studies were rated as fair quality. Most studies did not report whether treatment allocation was blinded or whether the persons who conducted follow-up assessments were blinded to the treatment condition. Most trials used prospective methods to assess falling; 4 noncomprehensive trials assessed falling retrospectively by using a questionnaire (23, 25, 29, 40). Studies made different assumptions in their analyses, including assumptions about loss to follow-up and falls that may have resulted in outcome misclassification. For example, 1 study counted all lost participants (including those who died) as fallers (24), but another included falls that occurred before loss to follow-up (30). The remaining studies excluded participants who were lost to follow-up or were dead from the analyses or did not report the analyses assumptions.
Pooled estimates involving all 19 multifactorial assessment and management trials suggested a possible reduction in falls with some of these interventions (risk ratio, 0.94 [95% CI, 0.87 to 1.02]), although this finding was not statistically significant. Pooled estimates for the 6 comprehensive trials also suggested a protective effect (risk ratio, 0.89 [CI, 0.76 to 1.03]), although this finding was not statistically significant either (Figure 1). Multifactorial and management interventions were clinically heterogeneous, and the pooled statistical heterogeneity, particularly for the comprehensive trials, was high (I2 = 73.1%). Meta-regression models did not identify any study-level variable, including the comprehensiveness of the intervention, that clearly explained the heterogeneity in the effect estimate. Of note, however, the 4 trials with retrospective assessment of falls more often reported greater reduction in the number of fallers in the intervention group than the control group, compared with trials that used prospective assessment methods. Removing these trials, all of which involved noncomprehensive interventions, further attenuated the observed effects of the noncomprehensive interventions toward the null.
We found no additional studies beyond the trials included in the efficacy section. Overall, there was limited evidence for clinically significant harms from multifactorial assessments. One good-quality trial conducted in New Zealand (312 participants) had a higher proportion of fallers and frequent fallers at 12 months in the intervention group than in the control group (27). Only 5 fair-quality trials explicitly reported additional adverse effects. Adverse effects in these trials were minor and included minor musculoskeletal symptoms or increases in nonurgent outpatient visits to primary care (28, 30, 33, 38, 40).
Exercise or Physical Therapy
We evaluated 18 trials of exercise or physical therapy interventions (3986 participants) that had 21 active intervention groups (43-60). Trials of exercise or physical therapy tested a variety of interventions in 3 major categories: gait, balance, or functional training (exercises designed to develop dynamic strength, flexibility, and agility needed for everyday activities); strength or resistance exercise; and general exercise (including walking, cycling, aerobic activity, and endurance exercise). All but 1 trial (45) included some type of gait, balance, or functional training, and all but 6 trials (50-52, 58-60) used strategies in more than 1 category. Control groups received no intervention or education. Three trials evaluated interventions that lasted 12 months or longer (44, 48, 56). The duration of intervention in the remaining trials ranged from 6 to 26 weeks (median, 12.5 weeks). Treatment intensity (total hours of contact) ranged from 2 to 243 hours (median, 28 hours). Several trials assessed falling retrospectively by using questionnaires (44, 50, 53, 54, 57).
In pooled analysis, exercise or physical therapy interventions reduced the risk for falling by 13% (CI, 6% to 19%), although most trials showed non-statistically significant differences (Figure 2). In meta-regression models, the effect size of the pooled estimate was not influenced by age, sex, history of falling, or risk status of the participants. Meta-regression of the number of hours of physical activity, however, suggested that more intensive physical activity interventions produced a small but statistically significant reduction in the risk for being a faller.
We found no evidence of an increase in falls or fallers in 18 fair-quality or good-quality trials. These trials examined exercise or physical therapy interventions to prevent falls (43-60); in addition, 2 additional fair-quality trials (496 participants) were identified in our harms search (61-62). Three of these trials reported other adverse effects. Two trials reported falls while exercising as instructed, although there was no increase in the number of fallers in the intervention group (56, 61). One fair-quality trial (424 participants) that explicitly evaluated for adverse effects found that persons in the exercise group had more physician visits for abnormal heart rhythm than did persons in the control group, but it reported no statistically significant differences for serious harms, including clinically significant abnormal laboratory or other diagnostic testing, hospitalization, or life-threatening events (62).
Single Clinical Treatments
We evaluated 9 trials of vitamin D supplementation (5809 participants) (63-71). Five of the trials included only women (66-67, 69-71); the proportion of women in the other trials ranged from 51% to 80% (63-65, 68). Five trials were conducted in populations defined as high risk because of recent falls or vitamin D deficiency (64, 67-70). The remaining 4 studies used populations that were unselected except for age 65 years or older (63, 65, 66, 71). All studies were rated as fair quality. Only 3 trials assessed self-reported falls prospectively by using a diary or questionnaire (64, 66, 68). The remaining trials assessed self-reported falls retrospectively with periods of recall ranging from 6 weeks to 12 months (60, 62, 64, 66-68).
The daily oral doses of vitamin D in the intervention ranged from 10 IU (66) to 1000 IU (70) (median, 800 IU). One study provided a single intramuscular injection of 600,000 IU of vitamin D (64). The duration of active intervention ranged from 8 weeks (67) to 3 years (63, 66, 71) (median, 12 months). Two studies evaluated ergocalciferol (vitamin D2) (64, 70), and the remaining studies evaluated cholecalciferol (vitamin D3). Six trials included calcium supplements with vitamin D (63, 67-71). The control groups ranged from no intervention (69, 71) to placebo (63-66) or calcium supplements only (67-68, 70).
Efficacy. Vitamin D with or without calcium was associated with a 17% (CI, 11% to 23%) reduced risk for falling during 6 to 36 months of follow-up (Figure 3). Trials of vitamin D with calcium compared with no treatment or placebo did not support any added benefit of calcium (63, 69, 71). Age, sex distribution, history of falling, or risk status of the participants did not affect the pooled estimate.
Harms. On the basis of the 9 fair-quality trials included in our review, we found no increase in falls, fallers, or other major adverse events. Only 3 trials (926 participants) specifically reported adverse effects—transient and asymptomatic hypercalciuria or hypercalcemia in the intervention group—but no differences in adverse effects or clinically significant harms, such as incident kidney stones, cancer, ischemic heart disease, or stroke (65-66, 70).
Four fair-quality or good-quality trials (1437 participants) evaluated the effect of surgical and nonsurgical vision correction (after screening for visual impairment) on risk for falling (49, 72-74). All of the trials included high-risk populations that were selected for frailty or presence of cataracts. Two trials evaluated expedited cataract surgery compared with routine wait-list controls (73-74). The other 2 studies compared vision screening and referral or treatment with wait-list control or usual care (49, 72).
Efficacy. Vision correction did not reduce the proportion of fallers.
Harms. Although evidence on harms of vision screening or treatment of early impairment of visual acuity in older adults is sparse, 1 study suggested a possible harm (72). In a fair-quality trial conducted in Australia (616 participants), frail older adults received a comprehensive eye examination and approximately 44% received subsequent treatment of identified vision problems (72). Compared with control participants, persons in the intervention group had a higher proportion of fallers (65.0% vs. 49.8%; P < 0.001) and frequent fallers (37.9% vs. 30.6%; P = 0.003) (72). There was also a non-statistically significant increase in fall-related fractures. Trial investigators hypothesized that corrected vision may have increased the level of activity of these frail older adults, thereby increasing their risk for falling (72).
Medication Assessment and Withdrawal
We evaluated 1 small trial (48 participants) of medication assessment and withdrawal in older adults taking psychotropic medications compared with usual care (47). An additional 10 studies with multifactorial assessment and management interventions also included medication assessment and withdrawal (23-25, 31-32, 37-41).
Efficacy. Medication assessment and withdrawal of therapy with medication alone were not associated with a reduced fall rate (47).
Harms. No harms were reported in the trials evaluating withdrawal of therapy with medication.
We evaluated 3 trials of home-hazard modification (2348 participants) (48-49, 75). All 3 trials evaluated an in-home assessment with modification of any identified hazards (for example, adding nonslip tape to rugs and steps) or provision of free safety devices (such as grab bars), compared with usual care or a social control. Two of these interventions also included behavioral counseling of intervention participants (48, 75).
Risk for falling was reduced by 7% to 41%, although only 1 home-hazard modification trial (196 participants) reported a statistically significant beneficial effect on risk for falling compared with control participants (48).
There was no evidence of increased falls or fallers, based on the 3 fair-quality trials that included home-hazard modification interventions. None of these trials reported additional adverse events.
Clinical Education or Behavioral Counseling
We evaluated 1 trial of behavioral counseling (310 participants) in high-risk older adults (76). The intervention group received seven 2-hour weekly group sessions conducted by an occupational therapist and one booster session held 3 months after the final group session. The control group received 2 social visits conducted by an occupational therapist. An additional 13 trials incorporated educational components into a multifactorial clinical assessment, clinical management, or home-hazard modification intervention (23-25, 30-31, 33-38, 46, 75).
Education alone was not associated with a reduction in fall risk.
There was no increase in falls or fallers, and no additional adverse effects were reported.