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 of Interventions to Prevent Falls in Older Adults

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

Show details


Key Questions and Analytic Framework

Using the methods of the USPSTF,60 we developed an analytic framework (Figure 1) and four key questions (KQs) to guide our literature search and systematic review. These KQs were designed to evaluate the effectiveness and harms of primary care relevant interventions to prevent falling in older adults. Interventions relevant to primary care include those conducted in primary care, judged feasible to be delivered in primary care, or those easily referred from primary care (Appendix A). We grouped these interventions into five main categories: multifactorial assessment and management, single clinical treatment (with or without screening), clinical education/behavioral counseling, home hazard modification, and exercise/physical therapy. The KQs used for this search were:

Figure 1. Analytic Framework and Key Questions.

Figure 1

Analytic Framework and Key Questions.

KQ 1.

Is there direct evidence that primary care interventions reduce fall-related injury, improve quality of life, reduce disability, or reduce mortality when used alone or in combination to reduce falling in community-dwelling older adults?


Do these interventions reduce injury, improve quality of life, reduce disability, or reduce mortality in older adults specifically identified as high risk for falls?

KQ 2.

Do primary care interventions used alone or in combination in community-dwelling older adults prevent falling?


Do these interventions prevent falling in older adults specifically identified as high risk for falls?


Are there positive outcomes other than reduced falling, and related morbidity and mortality, that result from primary care interventions to prevent falling?

KQ 3.

What are the adverse effects associated with interventions to prevent falling?

KQ 4.

How are high-risk older adults identified for primary care interventions to prevent falling?

Search Strategy and Selection Criteria

We initially searched for relevant existing systematic reviews in the Cochrane Database of Systematic Reviews, the Database of Abstracts of Reviews of Effects, and the Health Technology Assessments and MEDLINE databases, as well as the Institute of Medicine, the Agency for Healthcare Quality and Research (AHRQ), and NICE Web sites. We developed separate literature searches for each KQ based on our review of this literature. We used one good-quality 2003 Cochrane systematic review and meta-analysis, which conducted a comprehensive search and had detailed reporting, as a foundation for our literature search for KQs 1 and 2.54

We conducted the search for KQs 1 and 2 (Appendix B Table 1) in MEDLINE, the Cochrane Central Registry of Controlled Trials, and the Cumulative Index to Nursing and Allied Health Literature (CINAHL) from the end of the Cochrane review54 search date of 2002 through February 2009. We limited the scope of KQ 4 to the high-risk definitions and assessments used in the primary care relevant interventions included for KQs 1 and 2. The search for KQ 3 focused on the harms of interventions included in KQs 1 and 2, which include multifactorial assessment and management, clinical education/behavioral counseling, home hazard modification, exercise/physical therapy, liquid energy and protein dietary supplementation, and hip protectors. We did not systematically search for harms of vitamin D supplementation or harms of vision screening and early treatment as these have both been reviewed in recent AHRQ-funded evidence reports. We searched in MEDLINE and CINAHL beginning in 1992 through February 2009, as 1992 was the earliest publication date of the included trials. For all KQs, we also obtained references from outside experts and through reviewing bibliographies of other relevant articles and systematic reviews.

Two investigators independently reviewed all abstracts and articles against inclusion and exclusion criteria. Discrepancies were resolved by consensus. Inclusion and exclusion criteria were developed for each KQ and are detailed in Appendix B Table 2. Briefly, for KQs 1, 2, and 4, we included only randomized controlled trials conducted among community-dwelling older adults in settings generalizable to the U.S. outpatient and ambulatory primary care populations (see Appendix A for a definition of primary care feasible or referable interventions). We excluded trials that were not designed to assess falls prevention based on assessment of falling or falls as a primary or secondary outcome. For KQ 3 (evaluating harms), we included both trials and observational studies that included primary care relevant interventions to prevent falls conducted in settings generalizable to U.S. primary care populations. Case series and case reports were excluded unless they addressed fatal harms. All trials were limited to English-language articles.

Article Review and Data Abstraction

After dual-reviewing articles for inclusion, two independent investigators critically appraised all included articles using design-specific criteria (Appendix B Table 3). Discrepancies in quality ratings were resolved by consultation with a third investigator. All studies rated as poor quality were excluded from the review. The USPSTF’s Methods Workgroup has defined a three-category quality rating of “good,” “fair,” and “poor” based on specific criteria.60 We reviewed a total of 1,179 abstracts and 425 articles for KQs 1, 2, and 4 and 765 abstracts and 84 articles for KQ 3 (see Appendix B Figure 1 for search results and article flow). A listing of excluded studies and reasons for exclusion can be found in Appendix C Table 6 (KQ 1), Appendix C Table 7 (KQs 2 and 4), and Appendix D Table 3 (KQ 3).

One investigator abstracted data from included studies into evidence tables and a second investigator reviewed these data for accuracy. We abstracted pre-specified study details into evidence tables, including population (age, gender, dwelling, race/ethnicity, socioeconomic status, fall history, inclusion and exclusion criteria); study design, location, and recruitment strategy; number assessed for eligibility, excluded, and randomized; definition and instrument used to identify population at risk, if any; intervention type (multifactorial assessment and management, single clinical treatment, exercise/physical therapy, home hazard assessment, or counseling/education) and description, including key elements, intensity, and duration; length of followup; outcomes; and any recorded adverse effects. Furthermore, we categorized interventions by hours of contact, calculating overall dose in hours; high-intensity interventions had more than 75 hours of contact, moderate-intensity interventions had 26–75 hours, low-intensity interventions had 10–26 hours, and very low-intensity interventions had 0–9 hours. Relevant outcomes for abstraction were determined a priori. Outcomes for KQ 1 included fall-related fractures; quality of life as measured by the SF-12, SF-36, or EuroQol; disability as measured by activities of daily living and instrumental activities of daily living; and mortality (see Table 1 for a description of the outcome measures). For KQ 2, we included number of falls and person-years if provided, number of fallers, and number of frequent fallers. If raw numbers were not available, we also included rate ratio and odds ratio or risk ratio. For KQ 3, we included any adverse effect requiring unexpected medical attention (e.g., fall-related fractures, hospitalization, and mortality), as well as any paradoxical increase in falls or fallers. Complete evidence tables are included in Appendix C Tables 15 (KQs 1 and 2) and Appendix D Tables 1 and 2 (KQ 3).

Table 1. Outcome Measures of Interventions to Prevent Falls.

Table 1

Outcome Measures of Interventions to Prevent Falls.

This review included 39 articles representing 36 unique trials for KQ 1, 51 articles representing 47 trials for KQs 2 and 4, and 49 articles representing 48 trials and one systematic review for KQ 3.

Literature Synthesis

Evidence was synthesized by type of intervention into five main categories:

  1. Multifactorial assessment and management. Multifactorial assessment and management interventions include a clinical assessment of two or more domains of functioning, generally supplemented by assessment of falls-related or general geriatric risk factors and/or conditions, with assessment results used as a basis for remedial management. In this review, multifactorial risk assessments may have been a comprehensive geriatric assessment or a falls-focused assessment, generally including two or more of the following screenings: vision, gait, mobility, strength, medication use, cognitive impairment, orthostatic hypotension, and environmental risks. Management approaches were categorized as comprehensive (treatments and education to comprehensively address risks, conditions, or functional limitations identified through the assessment) or noncomprehensive (less comprehensive interventions that provided only referral or provided treatment of selected risks, conditions, or functional limitations).
  2. Single clinical treatment. Single clinical treatment protocols were defined as those with or without screening to identify persons needing treatment for a single fall-related risk factor, including vision correction, medication optimization/adjustment, assistive device prescription, pharmacological/nutritional interventions, treatment for orthostatic hypotension or urinary incontinence, and hip protectors.
  3. Clinical education or counseling. Education or behavioral counseling included interventions delivered by primary care clinicians and related health care staff to assist patients in adopting, changing, or maintaining behaviors related to fall risk, including exercise, fall risk reduction, and a home hazard checklist.
  4. Home hazard modification. Home visits to identify and remove potential fall hazards, adding grab bars and handrails, or otherwise modifying the environment to improve mobility and safety.
  5. Exercise/physical therapy. Organized programs for individuals or small groups that are part of a health care setting or widely available for referral in most communities, including physical exercise, mobility/gait training, muscle strengthening, balance training, and training for recurrent fallers. Programs may be home-based or occur in a community setting.

We conducted meta-analyses to quantitatively estimate the effect size of falls prevention interventions on fall-related and mortality outcomes. Separate analyses were conducted for each intervention category. For single clinical treatments, the analyses were further stratified by treatment type. In the case of trials with multiple intervention arms,61–63 we calculated estimates for combined intervention arms when the interventions were variations of the same intervention type (i.e., two exercise programs).

For binary outcomes (fallers, fallers with fractures and mortality), a risk ratio and its standard error were calculated using the raw numbers reported from each study and combined using a random effects model.64,65 We used the reported risk ratio66 or odds ratio67 for two studies66,67 that did not report raw numbers. For the latter, the estimate of odds ratio was very close to 1 and provided a good approximate for risk ratio. For mortality, another analysis using the fixed effect model was also performed as a sensitivity analysis since the events were rare. In this case, a fixed effect model could provide a better estimate.68 For the number of falls, the rate ratio and its standard error were obtained from the studies, if reported. If not, they were calculated based on a Poisson distribution if the study reported the number of falls and the corresponding person-time. Rate ratios were also combined using a random effects model.

In the two studies that used clustered randomization69,70 and another study whose design had potential clustering effect,71 we used the reported estimate if the study ate adjusted for clustering effect. Otherwise, we adjusted for clustering effect by multiplying the standard error of the risk or rate ratio by the square root of the design effect. Here, design effect=1+(m−1)ρ, where m is the average cluster size and ρ is the intracluster correlation coefficient. In the main analysis, ρ is assumed to be 0.60 for household clustered studies,71 to be comparable with the reported values,70 and 0.05 for a physician clustered study69 as a conservative estimate. Sensitivity analyses were also performed by assuming a range of plausible values for ρ.

We assessed the presence of statistical heterogeneity among the studies using standard chi-square tests and the magnitude of heterogeneity was estimated using the I2 statistic.72 A series of random effects meta-regression models were used to examine possible sources of heterogeneity and to investigate whether the size of effect measure estimates were associated with various study-level characteristics. In all cases, the outcome was the log of the risk ratio for having a fall. Separate models were run for each predictor, which included mean age, average age of 80 or older (yes vs. no), percent female, percent with a fall during the previous year, presence of several specific components, comprehensiveness or intensity of the intervention, and whether the sample was comprised of high-risk participants. See Appendix F Table 3 for detailed descriptions of how these predictors were defined and which group of trials comprised the samples for each of the predictors. We also conducted sensitivity analyses to determine if selecting the more intense or more comprehensive intervention arm or excluding outliers changed effect size or statistical significance. Test of publication bias on whether the distribution of the effect sizes was symmetric with respect to the precision measure were performed using funnel plots and Egger’s linear regression method73 when the number of studies was about 10 or more.74

All analyses were performed using Stata 10.0 (StataCorp LP, College Station, Texas).

USPSTF Involvement

The authors worked with USPSTF liaisons at key points throughout the review process to develop and refine the analytic framework and KQs and resolve issues around scope and approach. AHRQ funded this research under a contract to support the work of the USPSTF. AHRQ staff provided oversight throughout the project.


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...