Summary of Results

Our examination of the evidence of multifactorial interventions to prevent falls consists of a heterogeneous set of 26 trials. These trials recruited community-dwelling older adults with varying fall risk. The vast majority of the trials recruited participants determined to be at high risk for falls; history of falls was the most common risk factor used for trial recruitment. Each trial examined uniquely designed interventions, such as direct interventions and referrals customized to participants based on an initial risk assessment. Pooled results from 17 RCTs (n=9,737) show that multifactorial interventions reduce the incidence rate of falls at the longest followup time (6–12 months) compared to the control group but also that there is substantial heterogeneity and a lack of precision in the effect size (incidence rate ratio [IRR], 0.79 [95% CI, 0.68 to 0.91]; p=0.001; I²=87.2%). Pooled results from 24 trials (n=12,490) show that there is no statistically significant effect on people experiencing a fall (RR, 0.95 [95% CI, 0.89 to 1.01]; p=0.002; I²=56.4%). While most studies consistently show no or minimal benefit with nonstatistically significant point estimates near 0.9, the effect is imprecise. Multifactorial interventions appear to have no statistically significant effect on mortality (k=23; n=9,721; RR, 0.96 [95% CI, 0.79 to 1.17]; p=0.659; I²=0%). Narrative analysis demonstrates no effect on hospitalization, institutionalization, ADL, IADL, or QOL, but these outcomes are reported in just a small proportion of multifactorial studies.

There is sparse evidence on the harms of multifactorial interventions; conclusions are limited to a few studies (k=4; n=1,466) with incomplete reporting of adverse events. The harms, which are generally minor, include musculoskeletal complaints related to the exercise component of the multifactorial intervention.

Characteristics of Included Studies

Of the included 26 multifactorial trials (31 articles71-101), 15 were from the previous review^{71, 72, 75, 77, 80, 81, 83, 85, 87, 91-96} and 11 were new.^{73, 74, 76, 78, 79, 82, 84, 86, 88-90} Three trials from the previous review were excluded for population,¹⁰² lack of feasibility for primary care to refer to the intervention,¹⁰³ or poor quality.¹⁰⁴

Study Characteristics

We identified seven good-quality^{74, 78, 84, 85, 91, 93, 95} and 19 fair-quality RCTs^{71-73, 75-77, 79-83, 85-90, 92, 94, 96} (n=15,506) with a primary or secondary aim of examining the effectiveness of multifactorial interventions on falls and/or fall-related injuries (Table 2). The majority of the trials were conducted outside of the United States. Six trials were conducted in the United Kingdom,^{72, 74, 75, 83, 84, 92} four in Australia,^{78, 85, 87, 90} three in the United States,^{73, 93, 96}three in the Netherlands,^{76,80, 94} two in Canada,^{71, 81} two in Spain,^79,89 two in Finland,^88,91 and one each in Denmark,⁹⁵ Switzerland,⁸² Sweden,⁸⁶ and New Zealand.⁷⁷ The size of trials (intervention plus control groups randomized for our analysis) ranged from 100 to 5,310 participants. The fall-related outcomes reported in the trials that we included for analysis were falls (k=17; n=9,737), people experiencing a fall (k=24; n=12,490, injurious falls (k=9; n=4,306), fractures (k=5, n=3,236), people experiencing an injurious fall (k=16; n=9,445), people experiencing a fracture (k=5, n=1,937), mortality (k=23; n=9,721), people hospitalized (k=4; n=2,134), people transitioning to institutional care (k=7; n=2,143), QOL (k=4; n=1,104), ADL (k=7; n=2,106), and harms (k=4; n=1,466).

Table 2

Study Characteristics, for Multifactorial Interventions, by Author.

Population Characteristics

Most recruited participants were 65, 70, or 75 years of age or older (Table 3). The recruitment age thresholds were ≥55 years,⁷¹ ≥60 years,^{84, 90} ≥65 years,^{72, 75,76, 80, 81, 83, 86, 91, 92, 95, 96} ≥70 years,^{74, 78, 88, 93, 94} ≥75 years,^{73, 77, 85, 87, 89} ≥80 years,⁸² and ≥85 years.⁷⁹

Table 3

Population Characteristics, for Multifactorial Interventions, by Author.

Mean age ranged from 71.9 years⁷¹ to 85 years.^{79, 82} The percentage of women in the studies ranged from 53.2 percent⁸⁹ to 94 percent.⁷¹ Race and ethnicity were not reported in any of the studies except Wagner et al,⁹⁶ a U.S. study in which 94 percent of participants were white. Measures of socioeconomic status, such as education or income level, varied widely.

Nearly all studies solely recruited older adults living in the community; four RCTs included at least 94 percent community-dwelling older adult^{75, 76, 84, 91} Trials recruited participants from clinics,^{74, 77-79, 87,89, 92, 94} emergency departments, hospitals, or ambulances following a fall,^{72, 75, 80, 83, 84, 90, 95} or multiple settings (e.g., a combination of clinic, emergency department, hospital, self-referral, and/or community).^{71, 76,81, 82, 86, 88, 91} Two U.S. trials recruited patients from health maintenance organizations^{93, 96}; the third U.S. trial recruited patients from a health insurance database.⁷³ An additional RCT recruited patients in Australia from health insurance member databases.⁸⁵

A total of 21 of 26 studies excluded patients with cognitive impairment or inability to provide consent/understand instructions. Half of the included studies specifically excluded participants with cognitive impairment based on the results of various tests (e.g., Mini-Mental State Examination [MMSE], mental status questionnaire, abbreviated mental test, clinical dementia rating scale, or clinical diagnosis documented in medical record).^{72, 75-78, 80, 82, 85, 86, 91-93, 95} An additional eight studies excluded those who could not understand instructions or provide their own informed consent.^{71, 74, 77, 82-84, 88, 89}

Seven trials recruited patients at average risk of falling where the only risk factor for falls was age.^{73, 79, 82, 85, 87, 89, 96} Nineteen trials solely recruited patients at high risk for falls according to various definitions.^{71, 72, 74-78, 80, 81, 83, 84, 86, 88, 90-95} Nearly half of the studies (12 of 26) defined high risk as having a history of falling based on either historical recall of one or more falls in the previous 3 months⁷⁷ or 12 months,^{81, 91, 92} or seeking medical attention in an emergency department, hospital, ambulance, or clinic following a fall.^{72, 76,80, 83, 84, 90, 95} The remainder of the trials recruited participants who fulfilled one or more risk factor criteria from a list of possible risk factors. The most common risk factors included a history of one or more falls;^{74, 88, 94} gait, balance, mobility, high risk medication usage, or high health care utilization;^{74, 86, 88, 93, 94} ADL impairment;^{86, 93} osteoporosis or a history of osteoporotic fracture;^{71, 88} TUG score of >14 seconds;⁷¹ or frailty (3+ Cardiovascular Health Study frailty criteria: slow gait, weak grip, exhaustion, low energy expenditure, or weight loss).⁷⁸

The baseline risk factor measures reported in the studies varied. Measures included a history of falls, fear of falling, comorbidities, number of medications, self-rated health status, ADL/IADL measurements, the MMSE, and/or measures of mobility/balance. In the trials reporting the percentage of participants with falls in the previous year, the range was 31 percent⁸⁹ to 100 percent.^{81, 91, 92} Trials reporting some measure of ADL function generally reported fairly independent ADLs,^{72, 73, 76,79, 81, 83, 87, 92, 95} although there were a few exceptions, such as the trial by Moller et al,⁸⁶ which required participants to need assistance with at least two ADLs, and the trial by Russell et al,⁹⁰ in which a third of participants needed some assistance with ADLs.

Intervention Details

The 26 multifactorial RCT publications described to various degrees the complex assessment and intervention components for the treatment group (Table 4). A trial by Newbury et al⁸⁷ sent only initial risk assessment results to primary care providers (PCPs); there was no further treatment intervention or referrals by the research team. For the remainder of the trials, the intervention groups underwent both an initial assessment and treatment intervention that was largely individualized and based on the risk factors identified in the initial assessment. Treatment and referrals were generally managed by the research team.

Table 4

Intervention Details, for Multifactorial Interventions, by Author.

Intervention Groups

Initial assessment. All 26 RCTs used an initial assessment of modifiable fall risk factors to customize the intervention for each participant. This initial assessment could include a multidisciplinary comprehensive geriatric assessment or an assessment of the risk of falls with any number of the following components: balance, gait, vision, cardiovascular health (e.g., postural blood pressure or pulse, carotid sinus stimulation), medication, environment (e.g., home hazards or personal needs), cognition, and psychological health. The initial assessment occurred in a clinical setting and/or the participant’s home. Nursing professionals nearly always performed the initial assessment, with or without additional professionals (e.g., physical therapists, exercise instructors, occupational therapists, medical doctors, dieticians or nutritionists).

Treatment interventions. Treatment interventions varied substantially across the studies. They generally included multiple _targeted intervention components, such as exercise (unsupervised or supervised, group or individual), psychological (cognitive behavioral therapy), nutrition therapy, knowledge (e.g., via DVDs, lectures, pamphlets), medication management, urinary incontinence management, environment (e.g., assistive technology or dwelling recommendations), and referral to physical or occupational therapy, social or community services, and specialists (e.g., ophthalmologist, neurologist, cardiologist).

In the vast majority of trials (22 of 26 trials), treatment interventions were implemented through a combination of direct treatment administered by the research team as well as specialty referrals generated by the research team. In a minority of trials, the research team administered all of the interventions^{82, 93} or treatment recommendations were solely communicated with the PCP and/or the participant for further action.^{73, 87} In more than half of the trials (k=14), in addition to making direct referrals or directly implementing the recommended interventions, the research team communicated with PCPs (generally to communicate specific or comprehensive risk assessment results);^{71, 72, 76,77, 79-81, 84, 88,8991-93, 96} in two of these trials, PCPs were contacted only to discuss medication changes.^{72, 93} In the majority of trials, the protocols included one or more home visits for initial assessment, environment interventions, and/or physical therapy/exercises. The vast majority of interventions in the trials, however, occurred in the outpatient setting.

We could not quantify the intensity of the multifactorial interventions because total contact time was rarely reported. Most trials reported that they directly offered or referred participants to an exercise or physical therapy intervention. This offer was available either to all participants for one session or serially, or was _targeted to those with balance or gait issues identified in the risk assessment.

Control Groups

Control groups in the trials received usual care^{71-76,78-80, 82-87, 92, 94-96} or usual care plus minimal control (pamphlet, social visit, brief fall-risk advice, letter).^{77, 81, 88-91, 93}

Study Quality

Nearly all of the 26 trials reported randomization methods that were likely to be adequate. (Cohen et al⁷³ modified the probability of assignment to a particular group to ensure reaching _target analysis numbers to balance groups for attrition.) Three studies were cluster RCTs by physician⁹³ or clinical practice site.^{89, 92} Fifteen studies clearly reported adequate outcome assessor blinding,^{72, 75, 77-81, 84, 87, 90, 91, 93-96} and nine reported having unblinded outcome assessors (e.g., those administering interviews or abstracting medical records were not blinded).^{71, 73, 74, 76,82, 86, 88, 89, 92} For the remaining studies, reporting about assessor blinding was unclear.^{83, 85} The multifaceted and customized nature of these multifactorial interventions precluded analysis of intervention adherence rates. With two exceptions,^{73, 83} nearly all trials used intention-to-treat analysis where all randomized participants were included in the analysis regardless of their participation at the conclusion of the study. To ascertain falls, most studies had the patient prospectively record falls on a calendar or in a diary, with or without additional confirmation (e.g., medical records or 3–6 recall phone calls).^71-89 Six studies recorded falls by patient’s recall every 1 to6 months,^90-95 and one study recorded falls as recorded by the patient’s recall at >6 months and in hospital discharge summaries.⁹⁶ Completion rates ranged from 68.6 to 100 percent (all respondents had at least one data point, allowing them to be included in analysis). Eight trials were designed to have enough power to detect differences in the rate of falls,^{72, 74, 83, 84, 88, 89, 91, 95} and nine trials were powered for people experiencing a fall.^{75-77, 80, 81, 88, 89, 92, 94, 96}

Falls

Meta-analysis of 17 multifactorial RCTs^{72-75, 77, 79, 81, 83-86, 88, 90, 91, 93, 95, 99} (n=9,737) demonstrated a lower rate of falls at the longest followup (6–12 months) in the multifactorial group than in the control group with substantial heterogeneity (IRR, 0.79 [95% CI, 0.68 to 0.91]; I²=87.2%) (Figure 2 Table 5). In the control group, the rate of falls per person-year ranged from 0.38 to 7.7 events per person-year at the longest followup. Individual RCTs reported substantial variation in effect size, with wide and overlapping CIs and IRR point estimates ranging from 0.42 to 1.12. Two RCTs particularly notable for much greater beneficial effect sizes were conducted by Close et al⁷² and Logan et al,⁸⁴ with IRR point estimates of 0.42 and 0.45, respectively. Those studies recruited participants from the emergency department⁷² or an ambulance⁸⁴ following a fall; one had specific intervention protocols outlined in the publication,⁷² and the other included intensive interventions (6 physical therapy sessions, 12 group sessions of supervised exercise or education on preventing falls, and up to 12 home sessions⁸⁴). We explored heterogeneity by examining the number of falls by country, date of publication, recruitment setting, fall rate of the control group, recruitment inclusion criteria of average or high risk of falls, mean age, followup period, and study quality. We were unable to explain the high heterogeneity by any single variable except recruitment setting (Appendix C Figure 1). In meta-regression, both clinical and multiple recruitment settings were statistically different than that of the emergency department (p=0.030 and p=0.023, respectively). Caution should be used in interpreting this post hoc subanalysis because heterogeneity was high and formal a priori subgroup credibility ratings were not performed.⁶³ Visual examination of the funnel plot (not included in this report) for the 17 pooled studies did not suggest a risk of small study bias and the result of the Egger test was not statistically significant (p=0.678).

Figure 2

Pooled Analysis of Multifactorial Intervention Randomized, Controlled Trials for Falls at Longest Followup (6–12 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; IRR = incidence rate ratio

Table 5

Falls, for Multifactorial Interventions, by Author.

Injurious Falls

Approximately one-third (k=9; n=4,306) of the 26 multifactorial trials reported injurious falls at 6 to 36 months (Figure 3; Table 6).^{77, 78, 83, 86, 88, 90, 91, 93, 95} Injuries were largely defined as severe injuries (e.g., dislocation, severe sprain, fracture, head injury)^{91, 93} or those requiring medical care,^{83, 95} although some trials also included soft tissue bruises and contusions.^{77, 88} Three trials did not define the types of injuries sustained.^{78, 86, 90} Only one study, by Palvanen et al,⁸⁸ reported a statistically significant reduction in injurious falls in the multifactorial group compared to the control group (IRR, 0.74 [95% CI, 0.61 to 0.89]). The remaining eight trials showed no notable differences: half of these studies showed point estimates near 1.

Figure 3

Forest Plot of Randomized, Controlled Trials for Injurious Falls at Longest Followup (6–36 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; IRR = incidence rate ratio

Table 6

Injurious Falls, for Multifactorial Interventions, by Author.

Fracture

Five trials (n=3,236) reported falls resulting in any fall-related fracture or hip fracture^{78, 88, 90, 91, 95} at 12 to 36 months of followup. The results showed no statistically significant difference between the multifactorial and control groups. Individual IRR point estimates ranged from 0.55 to 1.09 at longest followup (Figure 4; Table 7).

Figure 4

Forest Plot of Randomized, Controlled Trials for Fractures at Longest Followup (12–60 Months) Fractures. Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; IRR = incidence rate ratio

Table 7

Fractures, for Multifactorial Interventions, by Author.

People Experiencing a Fall

Meta-analysis of the 24 multifactorial RCTs^{71-81, 83-86, 88-96} (n=12,490) reporting the number of people experiencing a fall demonstrated no difference at the longest followup (6–18 months) in the multifactorial group compared to the control group (RR, 0.95 [95% CI, 0.89 to 1.01]; I²=56.4%) (Figure 5; Table 8). Four RCTs^{72, 84, 88, 96} reported a statistically significant modest reduction in people experiencing a fall (range of RR, 0.62 to 0.84). The remaining 20 RCTs^{71, 73-81, 83, 85, 86, 89-95} showed no statistically significant difference in the number of people experiencing a fall, with one outlier study (Ciaschini et al⁷¹) reporting a point estimate of 1.51. The Ciaschini study had the youngest population (mean age 71.9 years), which was recruited from the emergency department after a fall.⁷¹ While one cluster RCT by Tinetti et al⁹³ reported an RR of 0.76 (95% CI, 0.58 to 0.98), after adjustment for clustering, this result was no longer statistically significant. Almost half of the studies (9 of 22) showed point estimates greater than 1.0, indicating that the control group had fewer people experiencing a fall, although none of the point estimates was statistically significant. The percentage of people experiencing a fall in the control group ranged from 17.0 to 94.1 percent. We examined variables as listed in the falls results above and were unable to identify any individual factors to explain the heterogeneity.

Figure 5

Pooled Analysis of Multifactorial Intervention Randomized, Controlled Trials for People Experiencing a Fall at Longest Followup (6–12 Months). Abbreviations: CG= control group; CI = confidence interval; IG = intervention group; RR = relative risk (more...)

Table 8

People Experiencing a Fall, for Multifactorial Interventions, by Author.

People Experiencing an Injurious Fall

Meta-analysis of 16 trials^{72-75, 80-82, 85, 86, 90-96} (n=9,445) showed no difference in people experiencing an injurious fall in the multifactorial group compared to the control group at the longest followup (9–36 months) (RR, 0.94 [95% CI, 0.85 to 1.03]; I²=34.3%) (Figure 6; Table 9).

Figure 6

Pooled Analysis of Multifactorial Intervention Randomized, Controlled Trials for People Experiencing an Injurious Fall at Longest Followup (12–36 Months). Abbreviations: CG= control group; CI - confidence interval; IG= intervention group; RR = (more...)

Table 9

People Experiencing an Injurious Fall, for Multifactorial Interventions, by Author.

People Experiencing a Fracture

Only five trials (n=1,937) reported on people with fragility,⁷¹ hip,⁷⁵ fall-related peripheral,⁹⁰ or any fall-related fracture^{89, 92} at 6 to 12 months of followup (control group prevalence range, 0.7% to 6.0%). There was no statistically significant difference between the multifactorial and control groups (Figure 7; Table 10). RR point estimates ranged from 0.17 to 1.02 in the studies with relatively few events, which made estimates unstable.

Figure 7

Forest Plot of Randomized, Controlled Trials for People Experiencing a Fracture at Longest Followup (6–36 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; RR = relative risk

Table 10

People Experiencing a Fracture, for Multifactorial Interventions, by Author.

Mortality

Pooled analysis of 23 RCTs^{71, 72, 74, 75, 77, 79-94, 96} (n=9,721) showed no difference in all-cause mortality at 6 to 36 months in the multifactorial group compared to the control group (RR, 0.96 [95% CI, 0.79 to 1.17]; I²=0%) (Figure 8). Individual study results varied widely, with RRs ranging from 0.20 to 5.03 and wide CIs reflecting a relatively uncommon outcome with few events in most studies (Table 11); not surprisingly, none was statistically significant as the trials were not intended nor powered to affect mortality outcomes.

Figure 8

Pooled Analysis of Multifactorial Intervention Randomized, Controlled Trials for Mortality at Longest Followup (6–36 Months). Abbreviations: CG= control group; CI - confidence interval; IG - intervention group; RR - relative risk

Table 11

Mortality, for Multifactorial Interventions, by Author.

People Transitioning to Institutional Care

Seven RCTs^{72, 74, 79, 81, 82, 87, 92} (n=2,143) reported mixed results on institutionalization (Figure 9; Table 12). The prevalence of institutionalization in the control groups varied substantially, from 0.6 to 20.1 percent. The RR from individual trials ranged from 0.43 to 3.07 with wide confidence intervals, which reflected the relatively few institutionalization events in most trials.

Figure 9

Forest Plot of Randomized, Controlled Trials for People Transitioning to Institutional Care at Longest Followup (6–12 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; RR = relative risk

Table 12

People Transitioning to Institutional Care, for Multifactorial Interventions, by Author.

People Hospitalized

Four RCTs^{72, 84, 93, 96} (n=2,134) reported the outcome of people hospitalized, which revealed no difference in the prevalence of hospitalization in the multifactorial versus control group (Figure 10; Table 13). RR and OR point estimates ranged from 0.57 to 0.98. There was wide variation in the prevalence of hospitalization in the control group (1–53%), indicating that the studies had heterogeneous populations with different baseline risks.

Figure 10

Forest Plot of Randomized, Controlled Trials for People Hospitalized at Longest Followup (12 Months). Abbreviations: CG= control group; CI = confidence interval; IG - intervention group; RR = relative risk Note: One multifactorial intervention study reported (more...)

Table 13

People Hospitalized, for Multifactorial Interventions, by Author.

QOL

Four RCTs (n=1,104) reported QOL outcomes as measured using 12-Item Short Form Survey (SF-12),⁷⁶ 36-Item Short Form Survey (SF-36),^{77, 95} EuroQol EQ-5D,⁷⁸ or EuroQoL EQ-5D Visual Analog Scale (VAS).⁷⁸ Overall, there was no difference between the intervention group and the control group at 12 months of followup (Table 14). Only the study by Vind et al^{95, 101} showed a statistically significant difference in changes in the SF-36 physical health component score from baseline in the intervention group compared to the control group at 12 months of followup, but these changes are unlikely to be clinically meaningful. SF-36 mental health-component mean changes from baseline to 12 months of followup were similar between the intervention group and the control group. In the studies by Elley et al,⁷⁷ Fairfall et al,^{78, 98} and deVries et al,⁷⁶ the changes in QOL scale scores (SF-12, SF-36, EQ5D VAS, EuroQol EQ 5D) from baseline to 12 months were similar between the intervention group and the control group.

Table 14

ADL, IADL, and QOL, for Multifactorial Interventions, by Author.

ADL and IADL

Seven studies (n=2,106) reported ADL outcomes as measured by the Barthel Index (Table 14).^{72, 76,78, 83, 84, 92, 95} These studies compared baseline to the longest followup (6 months⁸³ or 12 months^{72, 76,78, 84, 92, 95}) and showed no statistically significant difference in six of seven studies. The one exception was the trial by Logan et al,⁸⁴ with a statistically significant difference between the intervention and control groups in the proportion of participants with scores above or below the median value of 15 (OR, 2.91 [95% CI, 1.18 to 7.20]); however, the clinical meaningfulness of this difference is uncertain.

Four RCTs (n=1,102) used the Frenchay Activities Index (FAI)^{80, 94, 95} or the Lawton and Brody scale⁷⁶ to report IADL, which showed mixed results (Table 14). The RCT conducted by van Haastregt et al⁹⁴ showed a statistically significant difference in adjusted mean difference in the FAI (1.6 [95% CI, 0.6 to 2.7]) at 12 months that did not persist at 18 months (adjusted mean difference, 1.0 [95% CI, -0.2 to 2.2]). The other three studies showed no statistically significant difference in IADL changes between the intervention and control groups.^{76,80, 95}

Three studies (n=637) reported a combination of ADL and IADL using the Nottingham Extended Activities of Daily Living scale^{77, 84} or Sonn and Asberg⁸⁶ to compare mean differences in scores over 12 months; the results were mixed (Table 14). An RCT by Logan et al⁸⁴ reported that the Nottingham Extended ADL scale was statistically significantly different between the intervention and control groups (mean difference, 3.47 [2.13 to 4.81]).⁸⁴ Two studies reported no statistically significant difference in ADL and IADL combination score changes between the intervention and control groups.^{77, 86}

Fall Risk Status

As mentioned above, most multifactorial trials (19 of 26) recruited participants defined as at high risk for falls using a host of different risk-factor criteria. The most common risk factor used for recruitment was history of falls. For any given outcome, the vast majority of the pooled trials recruited high-risk participants (falls: 14/17; people experiencing falls 19/22), and our exploratory analyses did not suggest differential treatment effect based on whether the trials recruited high-risk individuals, nor did we see a linear association between control-group fall rate and treatment effect. Nonetheless, trials recruiting from emergency settings following a fall demonstrated a greater statistically significantly treatment benefit on falls reduction and in general, these trials recruiting from emergency settings had higher control group fall rates compared to the trials recruiting from clinic settings or a combination of clinic and emergency settings. This exploratory analysis suggests that there may be an association between fall risk and treatment effectiveness.

Harms

Four RCTs (n=1,466) reported any harm in the intervention group.^{78, 80, 91, 93} Only one of these RCTs reported harms (back pain) in the control group for comparison, which showed no difference between the intervention and control groups (2/120 vs. 0/121).⁷⁸ One RCT reported three falls without injury during the exercise component of the intervention.⁹¹ In general, adverse events were rare, minor, and associated with the exercise component of the multifactorial intervention.

Critical Appraisal

Populations and intervention design, components, intensity, and personnel are heterogeneous in these 26 trials. This heterogeneity makes it challenging to understand why the statistical results of the rate of falls varied so widely. Our attempts to explore the heterogeneity yielded one apparently consistent finding: studies recruiting participants from emergency settings exhibited greater benefit related to the rate of falls than did participants in other settings. Otherwise, we were unable to identify any trial or participant characteristic that was more likely to be associated with the rate of falls. Our examination of trials with statistically significant reductions in falls or the number of people experiencing a fall did not reveal an association even after we considered the adequacy of the power of the trial for these outcomes.

Harms were inconsistently reported in a small proportion of the total studies (k=4). The harms reported were generally minor and related to rare musculoskeletal complaints associated with the exercise component of the multifactorial intervention. The lack of collection of adverse events from the control groups further limits conclusions regarding harms.

Summary of Results

The evidence examining exercise interventions to prevent falls consists of a heterogeneous set of 21 trials (n=7,297). About half of the trials recruited participants determined to be at high risk for falls; physical function/mobility limitation was the most common risk factor (alone or in combination) used for trial recruitment. The trials included single or multiple exercise components, and the exercise interventions were primarily conducted in a group setting. Pooled results from 14 RCTs (n=4,663) showed that exercise interventions may reduce the rate of incident falls at 6 to 24 months of followup compared to the control group (IRR, 0.87 [95% CI, 0.75 to 1.00]; p=0.052; I²=57.3%). Pooled results from 10 RCTs (n=4,622) showed a 19 percent reduction in injurious falls (IRR, 0.81 [95% CI, 0.73 to 0.90]; I²=0.0%]). Pooled analysis of 15 RCTs (n=4,926) demonstrated that exercise interventions may reduce the number of people experiencing a fall at 6–24 months compared to the control group (IRR, 0.89 [95% CI, 0.81 to 0.97]; I²=43.9%). Pooled results from 11 RCTs (n=4,263) suggested that exercise interventions have no statistically significant effect on mortality (RR, 0.93 [95% CI, 0.71 to 1.22]; I²=0.0%). There is limited evidence of the effect of exercise on fracture, people hospitalized, people transitioning to institutional care, QOL, or IADLs based on one to three trials for each of these outcomes.

Conclusions about the harms of exercise interventions are limited by a small number of studies (k=8; n=4,107) and incomplete reporting of adverse events. Harms, including musculoskeletal complaints associated with exercise, were generally minor.

Characteristics of Included Studies

Of the 21 included exercise trials^105-125 (35 articles^105-139), nine were from the previous review^{105-112, 122} and 12 were new.^{113-121, 123-125} Eight trials from the previous review were excluded because the population was not representative of general primary care populations,^140-144 they had less than 6 months of followup,^{145, 146} or they were of poor quality.^141-147

Study Characteristics

We identified 16 fair-quality^{105-111, 114-118, 121-124} and 5 good-quality^{112, 113, 119, 120, 125} RCTs (n=7,297) with a primary or secondary aim of examining the effectiveness of exercise on reducing falls and/or fall-related injuries at 6 to 60 months of followup (Table 15). Three trials were conducted in the United States,^{105, 111, 125} eight in Europe (not including the United Kingdom),^{107, 108, 113-115, 119-121} seven in Australia or New Zealand,^{106, 109, 110, 112, 116,117, 122} two in Asia,^{123, 124} and one in the United Kingdom.¹¹⁸ Trial sizes ranged from 55¹⁰⁵ to 1,635¹²⁵ participants randomized to exercise and control arms (for our analysis).

Table 15

Study Characteristics, for Exercise Interventions, by Author.

The outcomes reported in the trials we included for analysis were falls (k=14; n=4,663), people experiencing a fall (k=15; n=4,926), injurious falls (k=10; n=4,622), fractures (k=3, n=2050), people experiencing an injurious fall (k=5; n=2,776), mortality (k=11; n=4,263), IADL (k=3; n=363), QOL (k=3; n=1,179), people hospitalized (k=1; N=98), and people transitioning to institutional care (k=2; n=206). No studies reported ADLs.

Two studies (n=2,480) reported adverse events for intervention and control groups.^{118, 125, 131} An additional six studies (n=1,627) reported adverse events for participants in intervention groups only.^{113, 115, 117, 120, 121, 123}

Population Characteristics

The most common _target population was aged 65 years or older (7 of 20 studies^{105, 109, 116, 118, 121, 123, 124}) (Table 16). Other age _target thresholds were 60 years or older,^{111, 112, 117, 119} 70 years or older,^{108, 113, 114, 120, 122, 125} 75 years or older,^{106, 115} 80 years or older,¹¹⁰ and at least 85 years.¹⁰⁷ The mean age ranged from 68 years¹¹⁹ to 88 years.¹⁰⁷ Six of the studies were conducted exclusively with women,^{110, 113-115, 119, 124} while in one study less than half of the participants were female (42%).¹²⁰ The majority of participants in the remaining studies were women. Three studies reported race/ethnicity; in those studies, participants were almost exclusively white.^{105, 118, 125} Ten studies reported a measure of socioeconomic status (primarily education).^{105, 108, 112, 115, 116, 118, 120, 121, 123, 125} Each study reported socioeconomic status differently, making it difficult to summarize this measure, but most participants who reported socioeconomic status had a high socioeconomic status (e.g., higher level of education).

Table 16

Population Characteristics, for Exercise Interventions, by Author.

Twelve of the 21 studies recruited from a community setting (e.g., population-based registries)^{107, 112-116, 119, 121-125}; the remaining studies recruited from a clinic or hospital^{106, 108-111, 117, 118} or from insurance rolls.^{105, 120} All studies recruited community-dwelling adults.

Nine trials did not specify any fall-risk criteria as a condition for inclusion of participants.^{105, 106, 110, 112, 114, 116, 119, 122, 124} The remaining 12 trials specified fall-risk criteria as a condition for inclusion of participants; the criteria varied widely.^{107-109, 111, 113, 115, 117, 118, 120, 121, 123, 125} Two identified participants with history of falls only,^{113, 120} one identified participants upon discharge from the hospital,¹¹⁷ and one used balance or gait.¹¹⁵ Seven studies defined fall risk using multiple risk factors (e.g., functional test, self-reported limitation in mobility, health status, history of falls, health care use) or used a risk assessment tool to determine risk status.^{107-109, 111, 118, 121, 123, 125} Overall, the most common fall-risk criterion was physical function/mobility limitation alone or in combination with other factors, self-reported or assessed using performance measures.^{107-109, 115, 121, 123, 125}

The baseline measures of health or functional status reported in the studies varied. The measures included living alone, experiencing a fall in the past year, physical function, ADL or IADL baseline score, QOL or self-reported health rating, number of medications, and other factors. Overall, about half of the participants included in these RCTs lived alone and forty percent reported falling in the past year.

Intervention Details

The 21 exercise interventions varied along several dimensions: supervision, individual versus group exercise, duration, frequency, and exercise components (Table 17).

Table 17

Intervention Details, for Exercise Interventions, by Author.

Intervention Groups

Supervision. With one exception,¹¹⁶ participants in all 21 exercise RCTs were supervised during the intervention. Supervision was conducted by a specialized exercise instructor,^{108, 109, 112, 114, 115, 118, 120, 121} physical or occupational therapist,^{107, 110, 111, 113, 117, 119, 122, 124} health professional,^{106, 123} or unspecified supervisor.^{105, 125} Participants in the unsupervised trial were self-directed using a written program.¹¹⁶

Individual versus group exercise. Fifteen of the interventions included group exercise alone^{105, 111, 112, 114, 119-121} or in combination with home-based exercise.^{107-109, 115, 118, 122, 123, 125} The remaining five interventions included only individual-based exercise.^{106, 110, 116, 117, 124}

Duration. Exercise programs ranged from 2 months¹¹¹ to 42 months.¹²⁵ The most common duration was 12 months in five trials.^{106, 109, 110, 114, 117}

Frequency. Exercise sessions were scheduled once per week^{112, 115, 121, 122} up to six times a week.^{117, 125} The most common number of sessions was three per week in seven trials.^{105, 106, 110, 111, 114, 118, 124}

Exercise component. The ProFaNE taxonomy³⁵ defines the following exercise components: gait, balance, and functional training; strength and resistance; flexibility; tai chi/3-D training; general physical activity; and endurance (Table 18). Eight of the RCTs employed a single exercise component.^{108, 111, 112, 115-117, 121} The maximum number of components was five.^{109, 118} The most common type of exercise component was gait, balance, and functional training; 17 of the 21 RCTs employed this component alone^{111, 115, 117, 121, 122} or in combination with another type of exercise.^{106, 107, 109, 110, 113, 114, 118-120, 123-125}

Table 18

Components of Exercise Interventions, as Defined by ProFaNE, by Author.

Thirteen of the RCTs employed resistance training in combination with another exercise component,^{105, 106, 109, 110, 113, 114, 118-120, 122-125} eight included flexibility in combination with another exercise component,^{107, 109, 110, 118, 119, 122, 124, 125} and five included endurance training in combination with another exercise component.^{105, 108, 110, 118, 120} Three of the RCTs included 3-D training, specifically tai chi alone^{108, 112} or in combination with another exercise component.^{107, 109, 118} Five of the RCTs employed general physical activity alone¹¹⁶ or in combination with another exercise component.^{106, 107, 119, 125}

Control Groups

Control groups in the trials were instructed to maintain usual activity levels^{105-108, 111-114, 118-124} and/or received minimal written information about health or preventing falls, or a social visit.^{109, 110, 115-117} One study assigned control group participants to receive a health education workshop that included weekly in-person sessions and 5- to 10-minute instructor-led stretching.¹²⁵

Study Quality

Two studies reported the use of unblinded outcome assessors (e.g., those administering an interview who were not blinded or those abstracting medical records unblinded),^{105, 116} and an additional three studies^{114, 121, 124} did not provide enough information to determine whether the outcome assessors were blinded. To ascertain falls, most studies used prospectively recorded falls diaries with or without additional confirmation (i.e., medical records)^{105, 106, 108, 110-122}; five studies recorded fall outcomes by recall every 1 to 6 months.^{107, 109, 123-125} Adherence to the intervention was reported by 11 of the RCTs as the average percentage of exercise sessions attended, which ranged from 57 percent¹¹⁷ to 95 percent.¹⁰⁹ Eight of the RCTs reported adherence to the intervention as the percentage of participants classified as “adherent.”^{108-110, 115, 117, 118, 120, 122} “Adherent” was defined by the investigators and ranged from any sessions to all sessions. In some cases, adherence was defined as still participating in exercise after 12 months. The percentage of adherent participants reported for the studies ranged from 17 percent¹¹⁸ to 82 percent.¹²⁰ Several studies did not adjust results for differences in baseline characteristics.^{105, 106, 109, 110, 114, 115, 117, 125} In addition, several studies reported poor retention of participants during followup^{108, 111, 118} or minor to moderate differences between intervention and control groups in attrition.^{105, 108, 116, 120, 122, 124}

Falls

Meta-analysis of the 14 exercise RCTs (n=4,663) reporting the outcome of falls demonstrated a significant reduction in the rate of incident falls at the longest followup (6–24 months) in the exercise group compared to the control group, with substantial heterogeneity (IRR, 0.87 [95% CI, 0.75 to 1.00]; p=0.052; I²=57.3%) (Figure 11; Table 19). The baseline fall rate of control groups varied widely, from 0.04 per person-year¹²⁴ to 1.6 per person-year.¹²¹ Individual RCTs reported substantial variation in effect and effect size, with IRR point estimates ranging from 0.47¹¹⁰ to 1.43,¹¹⁷ and wide and overlapping CIs. One RCT was notable for the higher fall rate in the exercise group (IRR, 1.43 [95% CI, 1.07 to 1.93]).¹¹⁷ That study recruited patients (mean age of 81 years) discharged from geriatric care, rehabilitation, and orthopedic wards at any of four public hospitals in Sydney, Australia. The intervention was an individually based program designed to improve balance; participants received instruction from a physical therapist and were asked to do the exercises at home six times a week. At 12 months, 57 percent were participating in exercise sessions and 29 percent were no longer exercising.¹¹⁷

Figure 11

Pooled Analysis of Exercise Intervention Randomized, Controlled Trials for Falls at Longest Followup (6–24 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; IRR = incidence rate ratio NOTES: For studies (more...)

Table 19

Falls, for Exercise Interventions, by Author.

We were unable to explain the high heterogeneity by any single variable, despite examining the outcome of falls by country, date of publication, rate of falls in the control group, inclusion criteria for average or high risk of falls, mean age, followup period, and intervention hours. We also evaluated variability based on the intervention exercise components (e.g., gait, balance, and functional training; strength or resistance; flexibility; tai chi/3-D training; general physical activity; endurance), the number of components in each intervention, and whether the intervention was group only, individual only, or group and individual. None of these characteristics was significant in explaining the variability. Visual examination of the funnel plot for the 14 pooled studies (not included in this report) did not suggest a small study bias, and the Egger test was not statistically significant (p=0.472).

Injurious Falls

Meta-analysis of the 10 exercise RCTs (n=4,622) reporting the outcome of injurious falls demonstrated a significant reduction in the rate of injurious falls at the longest followup (12–60 months) in the exercise group compared to the control (IRR, 0.81 [95% CI, 0.73 to 0.90]; I²=0.0%) (Figure 12; Table 20).^{106, 109, 113-115, 117, 118, 120, 122, 125} Specific definitions for fall-related injury outcomes were reported as injurious falls,^{106, 109, 113-115, 120, 122} fall-induced serious injuries,¹²⁵ or falls with injuries resulting in health care.¹¹⁷ Nine of the ten RCTs reported an effect consistent with a benefit of exercise intervention.^{106, 109, 113-115, 118, 120, 122, 125} Three of the RCTs reported statistically significantly reduced rates of fall-related injuries,^113-115 with significant effects ranging from 0.46¹¹³ to 0.80.¹¹⁵ One RCT reported a nonstatistically significant increase in the rate of injuries among the intervention group compared to the control group (IRR, 1.14 [95% CI, 0.76 to 1.73]).¹¹⁷

Figure 12

Pooled Analysis of Exercise Intervention Randomized, Controlled Trials for Injurious Falls at Longest Followup (6–60 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; IRR = incidence rate ratio NOTE: (more...)

Table 20

Injurious Falls, for Exercise Interventions, by Author.

Fractures

Three RCTs (n=2,047) reporting fall-related fractures show mixed results. One small RCT (n=72) reported a large reduction in the number of fall-related fractures at 60 months followup (IRR, 0.26 [95% CI, 0.07 to 0.97]).¹¹⁴ Two RCTs reported no statistically significant reduction in the rate of fractures in the exercise group compared to the control group (IRR, 0.92 [95% CI, 0.45 to 1.91]¹¹⁷ and IRR, 0.87 [95% CI, 0.63, 1.19]¹²⁵) (Figure 4; Table 21). One study additionally reported a nonsignificant reduction in risk of hip fracture (IRR, 0.76 [0.37 to 1.57]) ¹²⁵

Table 21

Fractures, for Exercise Interventions, by Author.

People Experiencing a Fall

Meta-analysis of the 15 exercise RCTs (n=4,926) reporting the number of people experiencing a fall demonstrated a reduced risk of falling at the longest followup (6-24 months) in the exercise group compared to the control group (RR, 0.89 [95% CI, 0.81 to 0.97]; I²=43.9%) (Figure 13; Table 22).^{105, 107-112, 115-117, 119, 121-123} Individual RCTs reported substantial variation in effect and effect size, with wide and overlapping CIs and RR point estimates ranging from 0.40¹¹⁹ to 1.38.¹¹⁷ The results from three RCTs were consistent with a statistically significant reduction in the number of people experiencing a fall in the exercise group compared to the control group (14-60 percent reduced risk).^{115, 119, 122} One RCT reported a statistically significant increase in the number of people experiencing a fall in the exercise group compared with the control group (RR, 1.38 [95% CI, 1.11 to 1.73]).¹¹⁷ The remaining 11 RCTs showed a nonsignificant benefit in terms of fewer people experiencing a fall in the exercise group than in the control group.^{105, 107-112,116, 118, 121, 123} We explored heterogeneity by examining country, date of publication, prevalence of people experiencing a fall in the control group, average or high fall risk, mean age, followup period, and intervention hours, yet were unable to explain the high heterogeneity by any single variable. When trials with specific exercise components were examined separately for the number of people experiencing a fall, none of these characteristics was significant in explaining the variability.

Figure 13

Pooled Analysis of Exercise Intervention Randomized, Controlled Trials for People Experiencing a Fall at Longest Followup (6–24 Months). Abbreviations: CG= control group; CI = confidence interval; IG = intervention group; RR = relative risk

Table 22

People Experiencing a Fall, for Exercise Interventions, by Author.

Six RCTs^{109, 110, 112, 116, 121, 122} (n=1,890) reported results for the number of people experiencing two or more falls (Table 22). The results of these trials were consistent with a reduced risk of falling for those in the exercise group compared to the control group. Three of these studies reported a statistically significant reduced risk, which ranged from 0.21¹²¹ to 0.54.¹¹² Two studies reported small or moderate nonsignificant reductions of risk of falling at least twice.^{110, 122} One study reported a risk estimate of 1.0.¹¹⁶

People Experiencing an Injurious Fall

Five RCTs (n=2,776) reported at least one outcome related to the number of people experiencing an injurious fall, with longest followup ranging from 12 to 60 months.^{109, 110, 114, 115, 125} All five trials reported results consistent with a reduced risk of injury, although none of the results was statistically significant. Reductions in risk among the exercise group ranged from 0.61¹¹⁴ to 0.90¹¹⁵ (Figure 14; Table 23).

Figure 14

Forest Plot of Exercise Intervention Randomized, Controlled Trials for People Experiencing an Injurious Fall at Longest Followup (12–60 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; RR = relative (more...)

Table 23

People Experiencing an Injurious Fall or Fracture, for Exercise Interventions, by Author.

People Experiencing a Fracture

One trial (n=1,635) showed a nonsignificant, decreased risk for people with fall-related fractures (RR, 0.87 [95% CI, 0.63 to 1.19]) and people with hip fractures (RR, 0.87 [95% CI, 0.41 to 1.81]) at 31 months in the exercise group compared to the control group (Table 23).¹²⁵

Mortality

Pooled analysis of 11 RCTs (n=4,263) showed no significant association with all-cause mortality at longest followup (12–60 months) in the exercise group compared to the control group (RR, 0.93 [95% CI, 0.71 to 1.22]; I² =0.0%) (Figure 15, Table 24).^{106, 107, 109, 110, 113-115, 117, 121, 123} Individual study results showed no statistically significant differences, and RRs varied widely (0.16¹⁰⁶ to 1.13¹²⁵), with wide CIs. The few deaths in most trials made estimates unstable.

Figure 15

Pooled Analysis of Exercise Intervention Randomized, Controlled Trials for Mortality at Longest Followup (12–60 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; RR = relative risk NOTE: For studies (more...)

Table 24

Mortality, for Exercise Interventions, by Author.

People Transitioning to Institutional Care

Two RCTs (n=206) reported no statistically significant difference in the number of people transitioning to institutional care between the exercise and control groups at longest followup (6–12 months). The wide confidence intervals reflect the rare event rate; the prevalence of institutionalization in the control groups varied from 2.8 percent over 6 months¹¹⁹ to 1.5 percent over 12 months¹²¹ (Figure 9, Table 25).

Table 25

Institutionalization, for Exercise Interventions, by Author.

People Hospitalized

One RCT (n=98) reported the outcome of people hospitalized, which showed no statistically significant difference in hospitalization in the exercise group compared to the control group at 12 months¹²³ (Figure 10, Table 26).

Table 26

Hospitalization, for Exercise Interventions, by Author.

QOL

Three RCTs (n=1,179) reporting QOL outcomes as measured by SF-12,¹¹⁷ EuroQol EQ-5D,¹¹⁸ and the Australian Quality of Life scale¹¹⁶ at longest followup (6–12 months) showed no statistically significant differences between the exercise and control groups (Table 27). In general, mean changes in exercise group from baseline to followup were small and not clinically meaningful.

Table 27

ADL, IADL, and QOL, for Exercise Interventions, by Author.

IADL

Three studies (n=363) reporting IADL outcomes as measured with the Lawton and Brody scale (original or modified) at longest followup (6–12 months)^{105, 110, 123} showed no statistically significant differences between the exercise and control groups (Table 27). In general, mean changes in exercise group from baseline to followup were small and not clinically meaningful.

Fall Risk Status

A little over half of the exercise trials (12/21) recruited participants defined as at high risk for falls using a variety of risk factor criteria. The most common risk factor used for recruitment alone or in combination with other risk factors was physical function/mobility limitation (self-reported or measured objectively). For any given outcome, more than half of the pooled trials recruited high-risk participants (falls, 8/14; people experiencing falls, 9/15; injurious falls, 7/10), and our exploratory analyses did not suggest differential treatment effect based on whether or not the trials recruited high-risk individuals, nor did we see an association between control group fall rate and treatment effect.

Harms

Eight RCTs (n=4,107) reported on harms^{113, 114, 117, 118, 120, 121, 123, 125, 131} in the intervention group. Two of these trials also reported harms in the control group for comparison and reported no difference in the rate of serious injuries between the intervention and control groups.^{118, 125, 131} Several studies reported that intervention participants experienced pain, bruising, or both, related to the exercise.^{115, 117, 118, 123, 131} One RCT¹¹⁵ reported a wrist fracture in a participant in the intervention group. Another RCT reported three serious fall injuries during 114,100 physical activity sessions (for a rate of 2.6 per 100,000 sessions).¹²⁵ Several of the studies reported no severe or significant adverse effects or injuries.^{113, 120, 121} In general, adverse events reported for these exercise interventions were minor.

Critical Appraisal

The recruitment strategies in these exercise interventions generally represent community-dwelling older adults and primary care populations at risk of falling. Most of the studies were small, with eight RCTs enrolling fewer than 200 participants and only four including more than 500 participants. In general, interventions were heterogeneous in terms of duration, frequency, and exercise components. With one exception, interventions included exercise supervised by a professional, and in the majority of interventions participants exercised as part of a group rather than exercising alone. We were unable to determine any trial or participant characteristic associated with a reduced risk of falling or the number of falls. The magnitude of effects was generally small or moderate, and the CIs were generally wide and overlapping. Harms were inconsistently reported in a small proportion of the total studies, and those that were reported were generally minor or rare musculoskeletal complaints. The lack of collection of adverse events from the control groups limits further conclusions regarding the potential harms.

Summary of Results

Evidence from seven heterogeneous trials^{113, 148-153} (11 articles^{113, 135, 138, 148-155}) (n=7,531) of different vitamin D3 formulations and dosing schedules and varying baseline fall risk in community-dwelling older adults shows mixed results. The single trial of annual high-dose cholecalciferol (500,000 International Units [IU]) showed an increase in falls, people experiencing a fall, and injurious falls;¹⁵³ the trial of calcitriol showed a reduction in falls and fallers,¹⁴⁹ and the remaining studies showed no statistical difference in falls, people experiencing a fall, or injurious falls. Pooled results showed no statistically significant difference in falls (k=5; n=3,529; IRR, 0.97 [95% CI, 0.79 to 1.20]; I²=75.8%), people experiencing a fall (k=6; n=6,519; RR, 0.97 [95% CI, 0.88 to 1.08]; I²= 60.3%) or mortality (k=6; n=7,084; RR, 1.08 [95% CI, 0.83 to 1.40]; I²=0%) One study showed no difference between vitamin D and control groups in QOL changes at 12 months. The number of people hospitalized or transitioned to institutional care, ADL, and IADL were not reported in any of these studies. Harms were reported variably in four studies and, with one exception, adverse events were no different between the vitamin D and placebo groups.^{148, 149, 152, 153} The exception was the single, annual high-dose cholecalciferol (500,000 IU) trial, which revealed a higher rate of falls, people experiencing a fall, and fall-related injuries among the intervention group.¹⁵³

Characteristics of Included Studies

Of the seven included vitamin D trials, four were included in the previous review,^148-151 and three were new.^{113, 152, 153} We excluded the four trials that did not recruit a population generalizable to a primary care population^156-159 (e.g., those with vitamin D insufficiency or deficiency) and one poor-quality trial.¹⁶⁰

Study Characteristics

We identified three fair-quality^148-150 and four good-quality^{113, 151-153} RCTs (n= 7,531) with a primary or secondary aim of examining the effectiveness of vitamin D on falls, fall-related injuries, or both at 9 to 60 months of followup (Table 28). Two trials were conducted in the United States,^{149, 151} two in Australia,^{152, 153} and one each in the United Kingdom,¹⁵⁰ Switzerland,¹⁴⁸ and Finland.¹¹³ Trial size ranged from 204 participants¹¹³ to 3,314 participants^{150, 153} randomized to the vitamin D and matched control arms.

Table 28

Study Characteristics, for Vitamin D Interventions, by Author.

The outcomes reported in the trials that we included for analysis were falls (k=5; n=3,496), people experiencing a fall (k=6; n=6,519), injurious falls (2 RCTs; n=2,460), mortality (k=6; n=7,084), and QOL (k=1; n=3,314). No studies reported people experiencing an injurious fall, hospitalization, institutionalization, ADL, or IADL. Five studies (n= 3,955) reported harms.^{148, 149, 151-153}

Population Characteristics

Five studies recruited participants aged 70 years and older,^{113, 148, 150, 152, 153} and two RCTs recruited participants aged 65 years and older (Table 29).^{149, 151} Mean age ranged from 71 years^{149, 151} to 76.8 years.¹⁵⁰ Five of the seven studies were conducted exclusively with females,^{113, 149, 150, 152, 153} and two recruited about the same number of males and females.^{148, 151} In the three studies that reported race or ethnicity, the participants were almost exclusively white.^{149, 151, 152} Measures of socioeconomic status were not reported.

Table 29

Population Characteristics, for Vitamin D Interventions, by Author.

Five of the seven studies recruited from the community setting.^{113, 148, 149, 151, 153} One study recruited participants from general practices,¹⁵⁰ and one RCT recruited from the community and clinics.¹⁵² All seven studies recruited community-dwelling adults.

Four trials recruited patients at average risk of falls where the only risk factor was age.^{148, 149, 151, 152} Three trials recruited only patients at high risk (based on varying definitions) for falls.^{113, 150, 153} One study defined high risk as a history of falls in the previous 12 months,¹¹³ and the two remaining studies^{150, 153} defined high risk as the presence of one or more risk factors, including maternal or family history of hip fracture,^{150, 153} self-reported fall,¹⁵³ previous fracture,^{150, 153} low body weight (<58 kg),¹⁵⁰ or self-reported health that was fair or poor.¹⁵⁰

Baseline mean serum 25-hydroxyvitamin D levels ranged from 26.4 ng/ml¹⁵² to 31.8 ng/ml.¹⁴⁹ These values were reflective of mean vitamin D levels in adults aged 60 and older based on 2009–2010 data of the National Health and Nutrition Examination Survey (mean 25-hydroxyvitamin D level, 29 ng/ml).⁶²

Baseline measures of health or functional status varied. They included history of falls, comorbid conditions, baseline ADL or IADL score, QOL, and self-reported health rating.

Intervention Details

Vitamin D3 was administered orally in all studies with various formulations, including cholecalciferol,^{113, 150-153} 1-hydroxycholecalciferol,¹⁴⁸ and calcitriol (Table 30).¹⁴⁹ The dosing schedules varied. The cholecalciferol trials used a dose of 700 IU daily,¹⁵¹ 800 IU daily,^{113, 150} 150,000 IU every 3 months,¹⁵² or 500,000 IU annually.¹⁵³ The other two RCTs administered 1 μg of 1-hydroxycholecalciferol daily ¹⁴⁸ and 0.25 μg of calcitriol twice daily.¹⁴⁹ In two studies, the intervention group received calcium (500 mg/day of calcium citrate malate¹⁵¹ or 1,000 mg daily of calcium carbonate¹⁵⁰) in addition to vitamin D.

Table 30

Intervention Details, for Vitamin D Interventions, by Author.

Vitamin D was administered for 9 months up to 5 years; for two trials, the treatment duration and outcomes followup was 3 years^{149, 151} One trial continued therapy for 3 to 5 years and reported outcomes for up to 12 months after treatment was completed.¹⁵³ The control groups received a matched placebo in six of the seven trials.^{113, 148, 149, 151-153} One open-label study gave participants an educational pamphlet on fall prevention and adequate consumption of calcium and vitamin D.¹⁵⁰

Falls

Meta-analysis of five RCTs (n=3,529) showed no statistically significant effect on falls at longest followup (9–36 months) (IRR, 0.97 [95% CI, 0.79 to 1.20]; I²=75.8%) (Figure 16; Table 31).^{113, 148, 149, 151, 153} This group of trials showed mixed results between the vitamin D and control groups. The RCT (n=213) that used calcitriol showed a statistically significant 37 percent reduction in fall rate in the vitamin D group compared to the control group at 3 years (annual fall rate, 0.27 vs. 0.43; IRR, 0.63 [95% CI, 0.47 to 0.84]).¹⁴⁹ One trial of high-dose annual cholecalciferol treatment (n=2,256) showed a 16 percent increase in falls at 3 years (IRR, 1.16 [95% CI, 1.03 to 1.31]).¹⁵³ The three remaining RCTs reporting this outcome showed no statistically significant difference in falls between the vitamin D and control groups (IRR point estimates 0.87 to 1.12).^{113, 148, 151} The fall rate of the control group varied widely, from 0.37^{148, 151} to 1.18 falls per person-year.¹¹³ Sensitivity analysis removing the high dose annual vitamin D trial showed no statistically significant difference in falls (IRR 0.91 [95% CI, 0.68 to 1.22]).

Figure 16

Pooled Analysis of Vitamin D Intervention Randomized, Controlled Trials for Falls at Longest Followup (9–36 Months). Abbreviations: b.i.d. = twice a day; CG = control group; CI = confidence interval; IG = intervention group; IU = international (more...)

Table 31

Falls, for Vitamin D Interventions, by Author.

Injurious Falls

Two trials (n=2,460) reported falls with soft-tissue injury or injurious falls with mixed results (Figure 3; Table 32).^{113, 153} The annual high-dose (500,000 IU) cholecalciferol study by Sanders et al¹⁵³ showed an increase in injurious falls in the vitamin D group (IRR, 1.15 [95% CI, 1.02 to 1.29]), whereas the study by Uusi-Rasi et al¹¹³ showed no difference at 24 months (IRR, 0.84 [95% CI, 0.45 to 1.57]).

Table 32

Injurious Falls and Fracture, for Vitamin D Interventions, by Author.

Fractures

One trial (n=2,256) administering high-dose (500,000 IU) vitamin D reported fractures. There was no statistically significant difference in fractures between the vitamin D and control groups at 36 months, although the point estimate was above 1 (IRR, 1.25 [95% CI, 0.97 to 1.61]) (Figure 4; Table 32).¹⁵³

People Experiencing a Fall

Meta-analysis of six RCTs (n=6,519) showed no statistically significant difference in people experiencing a fall between the vitamin D and control groups at 9–36 months (RR, 0.97 [95% CI, 0.88 to 1.08]; I²= 60.3%) (Figure 17; Table 33).^{148-150, 152, 153, 155} Similar to the mixed results seen for the falls outcome, the calcitriol study (n=213) showed a statistically significant 23 percent reduction in people experiencing a fall in the vitamin D group compared to the control group (RR, 0.77 [95% CI, 0.61 to 0.98])^{149, 155} and the annual high-dose (500,000 IU) cholecalciferol study showed an 8 percent increase in people experiencing a fall in the vitamin D group (RR, 1.08 [95% CI, 1.03 to 1.14]).¹⁵³ The remaining four RCTs showed no difference in people experiencing a fall between the vitamin D and control groups (RR, 0.84¹⁴⁸ to 1.08¹⁵²). A sensitivity analysis removing the high-dose annual vitamin D trial showed no statistically significant difference in people experiencing a fall (RR 0.94 [95% CI, 0.84 to 1.05]). An additional sensitivity analysis adding trials exclusively recruiting participants with vitamin D insufficiency or deficiency showed a nonstatistically significant reduction in people experiencing a fall with the upper confidence interval including 1 (RR 0.88 [95% CI, 0.78 to 1.00]; I²=83.2%).

Figure 17

Pooled Analysis of Vitamin D Intervention Randomized, Controlled Trials for People Experiencing a Fall at Longest Followup (9–36 Months). Abbreviations: b.i.d. = twice a day; CG = control group; CI - confidence interval; IG - intervention group; (more...)

Table 33

People Experiencing a Fall, for Vitamin D Interventions, by Author.

People Experiencing an Injurious Fall

No trials reported this outcome.

People Experiencing a Fracture

Four trials (n=5,436) reported mixed results for people experiencing a fracture (Figure 7; Table 34).^150-153 Two trials^{150, 151} showed a reduction in people experiencing a fracture; Porthouse et al.¹⁵³ showed a nonstatistically significant reduction in people with physician-confirmed hip fractures at 25 months of followup (RR, 0.39 [95% CI, 0.11 to 1.34]) with wide CIs, and Bischoff-Ferrari et al.^{151, 154} showed a statistically significant reduction in people with nonvertebral fractures at 36 months (RR, 0.46 [95% CI, 0.23 to 0.90]). The high-dose (500,000 IU annually) cholecalciferol trial showed a nonstatistically significant increase in people with nonvertebral fractures at 36 months (RR, 1.22 [95% CI, 0.95 to 1.57]),¹⁵³ and Glendenning et al.¹⁵² reported similar proportions of people experiencing a fracture in both groups (RR, 0.94 [95% CI, 0.40, 2.24]).

Table 34

Person Experiencing a Fracture, for Vitamin D Interventions, by Author.

Mortality

Six RCTs (n=7,084) reporting mortality at 9 to 36 months of followup showed mixed results.^{113, 148-150, 152, 153} Meta-analysis showed no statistically significant difference in deaths between the vitamin D and control groups (RR, 1.08 [95% CI, 0.83 to 1.40); I²=0%). No individual study reached statistical significance (Figure 18; Table 35). Wide CIs were particularly notable in trials with up to two events.

Figure 18

Pooled Analysis of Vitamin D Intervention Randomized, Controlled Trials for Mortality at Longest Followup (9–36 Months). Abbreviations: b.i.d. = twice a day; CG = control group; CI = confidence interval; IG = intervention group; IU = international (more...)

Table 35

Mortality, for Vitamin D Interventions, by Author.

People Transitioning to Institutional Care

No trials reported this outcome.

People Hospitalized

No trials reported this outcome.

QOL

In one study (n=3,314), QOL exhibited similar mean differences in SF-12 mental and physical component scores between the vitamin D and control groups at 12 months compared to baseline (SF-12 mental component adjusted mean difference, 0.03 [95% CI, -0.04 to 0.97] and SF-12 physical component adjusted mean difference, -0.152 [95% CI, -0.10 to 0.7]) (Table 36).¹⁵⁰

Table 36

ADL, IADL, and QOL, for Vitamin D Interventions, By Author.

ADL

No trials reported this outcome.

Fall Risk Status

Three of the seven trials recruited participants at high risk for falls where the most common definition of increased fall risk was based on history of falls as a single factor or as one of many risk factors. Given the few trials and heterogeneity of dosages and formulations, no conclusions can be made about associations between baseline fall risk and treatment effectiveness.

Harms

Five RCTs (n=3,955) reported harms associated with vitamin D and showed no difference in the frequency of adverse events attributable to treatment (Table 37).^{148, 149, 151-154} As noted above in the sections on falls and persons experiencing a fall, one trial reported an increase in falls, people experiencing a fall, and fall-related injuries associated with the annual high dose (500,000 IU) of cholecalciferol. The event rates for several of the reported events that did occur were rare (e.g., kidney stones, diabetes). Transient hypercalcemia was reported in two trials^{148, 149} and described as mild or clinically asymptomatic; a single case of hypercalciuria was reported in the treatment group in one trial.^{151, 154} Most of the conditions are unlikely to be attributable to vitamin D.

Table 37

Harms, for Vitamin D Interventions, by Author.

Critical Appraisal

These seven vitamin D trials tested various formulations and dosing schedules for community-dwelling older adults at average or high risk of falls who were not specifically selected for vitamin D insufficiency or deficiency. Although most trials had followup lasting more than 12 months, only three of the seven trials were designed to have adequate power to detect differences in falls or fractures in these trials. Given the high statistical heterogeneity in the meta-analyses for falls, people experiencing falls, and mortality, we have limited confidence in interpreting these mixed results.

Summary of Results

We identified three trials^{122, 161, 162} (four articles^{122, 129, 161, 162}) that examined the effect of an environment intervention (i.e., a single home visit to reduce home hazards) on falls in older adults at varying fall risk (n=2,175). The results were mixed: one trial¹⁶² showed a 46 percent reduction in falls, while two trials showed no effect. None of the trials reporting people experiencing a fall or injurious falls showed differences between the environment intervention group and the control group. One trial reporting changes in QOL and ADL showed no difference between these groups. No trials reported the outcomes of mortality, people hospitalized, people transitioning to institutional care, or harms. All studies were conducted outside of the United States, and the overall conclusions were limited by few studies that were underpowered for fall outcomes.

Characteristics of Included Studies

Of the three included environment trials, one was included in the previous review¹⁶¹ and two were newly identified.^{122, 162} One study from the previous review was excluded because the population was not representative of the general primary care population.¹⁴¹

Study Characteristics

We found one good-quality RCT¹⁶² and two fair-quality RCTs^{122, 161} that examined the effect of environment interventions on falls (k=3, n=2,175) (Table 38). Two of the trials had a primary aim to reduce falls,^{122, 161} and the other trial had a secondary aim of falls prevention.¹⁶² Two trials took place in Australia^{122, 161} and one took place in the United Kingdom.¹⁶² Trial size ranged from 165 to 1,879 randomly assigned participants. All three trials reported on falls and people experiencing a fall; one of these also reported injurious falls.¹²²

Table 38

Study Characteristics, for Environment Interventions, by Author.

Population Characteristics

All three studies recruited community-dwelling adults aged 70 years and older (Table 39).^{122, 161, 162} One required that all participants had fallen in the year preceding the study¹⁶²; the other two studies reported that 27 percent of their participants had fallen in the previous year¹⁶¹ or 6 percent within the past month.¹²² Mean age ranged from 76 to 79 years. The percentage of females ranged from 52 to 69 percent. The two studies that reported ADLs indicated that participants were still performing most ADLs without limitations (mean score, 18 out of 20¹⁶² or 5.3 out of 6¹²²).

Table 39

Population Characteristics, for Environment Interventions, by Author.

Intervention Details

A nurse,¹⁶¹ home-maintenance staff member,¹²² or occupational therapist¹⁶² conducted a one-time assessment of the participant’s home to identify environment hazards within the home that could contribute to a fall (Table 40). Each study provided some kind of modification to reduce the hazard of falls. Modifications were made during the assessment (if possible)^{161, 162} or through a city program (which paid up to $100).¹²² For hazards not modified during the assessor’s visit or through a city program, the participant was responsible for making changes or hiring someone to complete the work. Only one of the trials had a systematic approach to identifying home hazards and also provided followup phone calls.¹⁶²

Table 40

Intervention Details, for Environment Interventions, by Author.

Participants in the control group in two of the trials were not given any intervention but could continue to receive usual care.^{161, 162} The third trial offered a delayed intervention.¹²²

Falls

Only one trial of high-risk participants, Pighills et al,¹⁶² reported a statistically significant reduction in the number of falls in the intervention group compared to the control group (IRR, 0.54 [95% CI, 0.36 to 0.83]) (Figure 19; Table 41). This trial had a prescriptive protocol for the home assessment and provided followup phone calls as well.¹⁶² The other two trials showed no effect on the rate of falls (IRR, 0.98 [95% CI, 0.81 to 1.19]¹²² and IRR, 1.02 [95% CI, 0.83 to 1.27]¹⁶¹).

Figure 19

Forest Plot of Other Interventions Randomized, Controlled Trials for Falls at Longest Followup (12–24 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; IRR = incidence rate ratio

Table 41

Falls, for Remaining Interventions, by Author.

Injurious Falls

Only one trial captured injurious falls (i.e., a fall that resulted in a cut, scrape, gash, bruise, fracture, head injury, or hospitalization). There were no differences between the intervention and control groups (IRR, 0.97 [95% CI, 0.75 to 1.26])¹²² (Figure 3; Table 42). The same trial found that the intervention group had a nonstatistically significant higher rate of falls with injuries resulting in health care (IRR, 1.47 [95% CI, 0.81 to 2.67]).

Table 42

Injurious Falls for Other Interventions, by Author.

People Experiencing a Fall

All three studies reported that the intervention group had fewer people experiencing a fall than the control group, but the differences were not statistically significant (Figure 20; Table 43). RR or point estimates ranged from 0.83 to 0.93. One study also looked at people experiencing recurrent falls and found similar nonstatistically significant results (people experiencing two or more falls: RR, 0.94 [95% CI, 0.66 to 1.33]; people experiencing three or more falls: RR, 0.73, [95% CI, 0.42 to 1.27]).¹²²

Figure 20

Forest Plot of Other Intervention Randomized, Controlled Trials for People Experiencing a Fall at Longest Followup (12–18 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; RR = relative risk NOTE: Under (more...)

Table 43

People Experiencing a Fall, for Other Interventions, by Author.

People Experiencing Injurious Falls

None of the studies reported people experiencing injurious falls.

Mortality

No studies reported mortality outcomes.

People Transitioning to Institutional Care

No trials reported this outcome.

People Hospitalized

No studies reported hospitalization.

QOL

Using the EuroQol and SF-12, one study reported on the QOL of participants in the intervention and control groups (Table 44). There was no difference from baseline to 12 months between the two groups.¹⁶²

Table 44

ADL, IADL, and QOL, for Other Interventions, by Author.

ADL

One study reported ADLs, measured with the Barthel index (Table 44). There was no difference from baseline to 12 months between the intervention and control groups.¹⁶²

Fall Risk Status

Only one of these three trials recruited participants at high risk for falls based on history of prior falls, so it is not possible to make conclusions about associations between baseline fall risk and treatment effectiveness.

Harms

No studies reported harms.

Critical Appraisal

These three environment trials of average or high-risk, community-dwelling older adults showed mixed results. The largest of the three trials (more than 1,500 participants) had no effect on fall outcomes,¹⁶¹ but the smallest trial, which was of good quality and had a designated protocol and provided telephone followup, resulted in a large reduction in falls.¹⁶² These mixed results suggest that larger studies replicating the trial by Pighills et al¹⁶² are needed.

Summary of Results

Evidence is limited to two underpowered RCTs^{163, 164} (3 articles^163-165) in high-risk older adults, which showed no difference in falls, injurious falls, people experiencing a fall, or mortality with an intervention involving medication management. There were no studies reporting people hospitalized, people transitioning to institutional care, QOL, ADL, or harms. Evidence of the effectiveness and harms of these interventions to reduce falls and fall-related injuries is too limited for conclusions.

Characteristics of Included Studies

Both trials were newly identified studies.

Study Characteristics

We identified two fair-quality RCTs (n=266) conducted in the United States with a primary aim of examining the effectiveness of medication management on recurrent falls^{163, 164} and fall-related injuries (Table 45).¹⁶³ The fall-related outcomes reported in the trials that we included for analysis were falls, people experiencing a fall, fall-related injuries, and mortality at 6 months¹⁶⁴ and 1 year¹⁶³ of followup.

Table 45

Study Characteristics for Medication Management Interventions.

Population Characteristics

Both RCTs recruited participants aged 65 and older (mean age, 75 years) (Table 46). The majority of participants were women (73%), were white (92%), and had an education beyond high school (74%).

Table 46

Population Characteristics, Medication Management Interventions.

One trial recruited participants from a central electronic database from a community pharmacy chain,¹⁶³ while the second trial recruited participants from a fall-prevention workshop.¹⁶⁴ Both trials recruited participants at high risk for falls (i.e., fell at least once in the prior year,^{163, 164} had a fear of falling¹⁶⁴ or took four or more long-term prescription medications,¹⁶³ of which one or more was a central nervous system medication (e.g., benzodiazepines, antidepressants, anticonvulsants, sedative-hypnotics, narcotic analgesics, antipsychotics, skeletal muscle relaxants).

All participants were community-dwelling and had no significant cognitive impairment (fewer than three errors on a six-item MMSE-derived screening test¹⁶³ or able to provide their own consent¹⁶⁴). One study reported that 43 percent of participants used a cane or walker and 49 percent had fallen two or more times in the year prior to randomization.¹⁶³ The second trial reported that 40 percent had fallen in the prior 6 months; most of the participants assessed their health to be good or better.¹⁶⁴

Intervention Details

In both trials, the intervention and control groups received an educational fall-prevention brochure (Table 47).

Table 47

Intervention Details, for Medication Management Interventions.

The medication management intervention included an algorithm-driven consultation with a pharmacist¹⁶⁴ or pharmacy resident.¹⁶³ One trial included a single, 45-minute, face-to-face medication review consultation,¹⁶³ while the second trial included one 60-minute, face-to-face medication review consultation and a followup telephone call at 3 months.¹⁶⁴ The pharmacist either contacted the prescriber to approve the medication changes¹⁶³ or developed an action plan with prescriber communication only when deemed necessary.¹⁶⁴

Falls

One study¹⁶³ reported no difference in the rate of falls between the medication management group and the control group (2.2 vs. 2.1 falls per person year; IRR, 1.01 [95% CI, 0.81 to 1.26]) (Figure 19; Table 41).

Injurious Falls

The same trial¹⁶³ reported no difference in injurious falls between the intervention and control groups at 12 months (IRR, 0.87 [95% CI, 0.62 to 1.24]) (Figure 3; Table 42).

Fractures

No trials report this outcome.

People Experiencing a Fall

Both trials reported no difference in people experiencing one or more falls between the medication management group and the control group (RR, 1.02 [95% CI, 0.79 to 1.31] for Blalock et al.,¹⁶³ RR, 1.16 [95% CI, 0.55 to 2.41] for Mott et al¹⁶⁴) (Figure 20; Table 43).

People Experiencing an Injurious Fall

No trials reported this outcome.

People Experiencing a Fracture

No trials reported this outcome.

Mortality

There was no difference in mortality between the medication management and control groups in one study.¹⁶³ However, the events were rare (3 vs. 2 deaths; RR, 1.50 [95% CI, 0.26 to 8.77]) (Figure 21; Table 48).

Figure 21

Forest Plot of Other Intervention Randomized, Controlled Trials for Mortality at Longest Followup (12–24 Months). Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; RR = relative risk

Table 48

Mortality, for Other Interventions, by Author.

People Transitioning to Institutional Care

The trials did not report this outcome.

People Hospitalized

The trials did not report this outcome.

QOL

The trials did not report this outcome.

ADL

The trials did not report this outcome.

Fall Risk Status

Both trials recruited participants at high risk for falls based on history of prior falls, fear of falling and/or high-risk medication usage, so it is not possible to make conclusions about associations between baseline fall risk and treatment effectiveness.

Harms

Neither study on medication management reported harms.

Critical Appraisal

The evidence on medication management is limited to two underpowered studies that showed no statistically significant effect on the outcome of falls. The literature is too limited to make any conclusions about the effectiveness or harms of medication management interventions.

Summary of Results

Two trials^{166, 167} (five articles^166-170) of cognitive behavioral interventions _targeted community-dwelling older adults at high risk of falling. An 8-week group-based intervention and 16-week individual-based intervention trial showed nonstatistically significant reductions in falls and people experiencing a fall, at 12 and 14 month followup. One trial reported a statistically significant reduction in people experiencing two or more falls at 14 months of followup. Trial results on injurious falls were mixed. Larger trials adequately powered for fall-related outcomes are needed. Neither study reported fracture, people experiencing an injurious fall, people experiencing a fracture, people hospitalized, or people transitioning to institutional care.

Characteristics of Included Studies

We identified two new studies that used cognitive behavioral interventions for reducing fear of falling.^{166, 167}

Study Characteristics

These two fair-quality RCTs (n=929 randomized), conducted in the Netherlands, were aimed to reduce fear of falling and activity avoidance ¹⁶⁶ or to address concerns about falls¹⁶⁷ in community-dwelling older adults (Table 49). Both trials measured falls, injurious falls, people experiencing a fall, mortality, and IADL changes at 12–14 months of followup.

Table 49

Study Characteristics, for Psychological Interventions.

Population Characteristics

Zijlstra et al¹⁶⁶ recruited community-dwelling older adults with a fear of falling or those who avoided activities due to a fear of falling (Table 50). Similarly, Dorresteijn et al recruited community-dwelling older adults who perceived their general health as fair or poor and who had concerns about falling and associated activity avoidance.¹⁶⁷ The mean age was 78 years. Seventy-one percent of participants were female. Approximately half (55%) of the participants lived alone, and more than one-third had fallen more than once in the 6 months before the trial was started in the study by Zijlstra et al.¹⁶⁶ The participants in the study by Dorresteijn et al¹⁶⁷ were more frail, with 61 percent falling in the 6 months prior to the start of the study, 13 percent reporting their general health as poor, 26 percent often or very often concerned about falls, and 23 percent often or very often avoiding activities due to concerns about falls.

Table 50

Population Characteristics, for Psychological Interventions.

Intervention Details

Both studies used a cognitive behavioral intervention designed to reduce fear of falling. A nursing professional facilitated the intervention in both studies (Table 51). Zijlstra et al¹⁶⁶ used a group-based approach with eight weekly 2-hour sessions with the purpose of addressing misconceptions, setting realistic goals for safe activity, reducing home hazards, and promoting physical exercise to increase strength and balance. Six of the eight sessions included 15 minutes of low-intensity physical exercise. Every weekly session assigned homework that included physical exercise. A booster session was provided 6 months after the eight weekly sessions were completed. Dorresteijn et al¹⁶⁷ used an individual approach, with seven total sessions (3 home visits, 4 telephone contacts) over 16 weeks aimed to address concerns about falls, thoughts about falling, physical exercise, asserting oneself, overcoming personal barriers; safe behavior, and managing concerns about falls. The session length varied, with in-person visits ranging from 60-75 minutes and telephone contacts of 35 minutes. The control group for both studies received usual care.

Table 51

Intervention Details, for Psychological Interventions.

Falls

In both trials, the intervention group had nearly identical results with nonstatistically significant lower rates of falls compared with the control group at 12 to 14 months of followup (adjusted IRR, 0.86 [95% CI, 0.65 to 1.14]¹⁶⁶ and adjusted IRR, 0.86 [95% CI, 0.65, 1.13]¹⁶⁷) (Figure 19; Table 41).

Injurious Falls

The trial results for injurious falls were mixed. At 12–14 months, one trial showed a nonstatistically significant reduction in fall-related injuries resulting in medical care in the intervention group compared to the control (adjusted IRR, 0.78 [95% CI, 0.45 to 1.34]),¹⁶⁶ but the other trial showed the opposite effect (adjusted IRR, 1.42 [95% CI, 0.96, 2.10])¹⁶⁷ (Figure 3; Table 42).

Fracture

Both trials did not report this outcome.

People Experiencing a Fall

There was a statistically significant 28 percent reduction in people experiencing a fall at 14 months in the intervention compared to the control group in one trial (RR, 0.72 [95% CI, 0.58 to 0.90]) however after adjustment these results were not statistically significant (adjusted OR, 0.50 [95% CI, 0.23 to 1.08]).¹⁶⁶ There was a statistically significant reduction in people who fell at least twice in the intervention group compared with the control group (RR, 0.59 [95% CI, 0.43 to 0.81]) and these results remained statistically significant even after adjustment for confounders (adjusted OR, 0.38 [95% CI, 0.17 to 0.84]). The study by Dorresteijn et al¹⁶⁷ reported similar results at 12 months, although none was statistically significant (Figure 20; Table 43).

People Experiencing an Injurious Fall

Both trials did not report this outcome.

People Experiencing a Fracture

Both trials did not report this outcome.

Mortality

There was no difference in the number of participants who died between the intervention and control groups in either trial (RR, 0.93 [95% CI, 0.30 to 2.84]¹⁶⁶; RR, 1.01 [95% CI, 0.36 to 2.81]¹⁶⁷) at 12–14 months of followup (Figure 21, Table 48). The study by Zijlstra et al^{166, 170} also reported mortality at 7 years and found no difference between the intervention and control groups (RR, 0.98 [95% CI, 0.77 to 1.25]) (Table 48).

People Transitioning to Institutional Care

Both trials did not report this outcome.

People Hospitalized

Both trials did not report this outcome.

QOL

Both trials did not report this outcome.

ADL and IADL

The study by Dorresteijn et al¹⁶⁷ reported statistically significant improvements in ADL (measured with the Groningen Activity Restriction Scale) for the intervention group versus control group at 12 months followup, but the difference between groups was small (adjusted mean difference, -0.83) (Table 44).

The study by Zijlstra et al¹⁶⁶ examined IADL using the Frenchay Activities Index and found no differences between the intervention and control group over 14 months of followup. The study by Dorresteijn et al¹⁶⁷ reported statistically significant improvements in ADL (measured with the Groningen Activity Restriction Scale) for the intervention group versus control group at 12 months followup, but the difference between groups was small (adjusted mean difference, -1.01) (Table 44).

The study by Dorresteijn et al¹⁶⁷ reported statistically significant improvements in ADL/IADL (measured with the Groningen Activity Restriction Scale) for the intervention group versus control group at 12 months followup, but the difference between groups was small (adjusted mean difference, -1.81) (Table 44).

Fall Risk Status

Both psychological trials recruited participants at high risk for falls based on fear of falling; evidence is too limited to make conclusions about associations between baseline fall risk and treatment effectiveness.

Harms

One trial reported no adverse events or side effects;¹⁶⁶ the other did not report adverse events.¹⁶⁷

Critical Appraisal

These two trials of a primary care referable group or individual, home-based cognitive behavioral intervention for community-dwelling older adults at high risk of falling were fairly well designed. While results showed a nonstatistically significant reduction in falls and number of people experiencing a fall, both trials’ primary aims were to reduce participants’ fear of falling, and in one trial, fall outcomes were collected for the purpose of monitoring safety rather than for assessing primary outcomes.¹⁶⁶ These trials were underpowered to detect differences in fall outcomes, so larger trials are necessary to determine if cognitive behavioral intervention has any effect on fall-related outcomes.

Summary of Results

Six fair- to good-quality RCTs^{113, 120, 122, 171-173} (10 articles^{113, 120, 122, 129, 135, 138, 171-174}) examined the effectiveness of multiple interventions with one to two trials testing each of the following combinations of interventions compared to a control group: exercise+environment, exercise+psychological, exercise+knowledge+fall risk assessment, exercise+vitamin D, and knowledge+environment in older adults of varying risk. There is limited evidence based on three trials designed to have adequate power for falls that knowledge+environment (n=310), exercise+environment+vision (n=272), and exercise+psychological (n=378) interventions reduce falls and/or fallers by 20 to 46 percent. A single underpowered trial (n=453) of exercise+knowledge+falls risk assessment compared to control revealed a nonstatistically significant reduction in falls (IRR, 0.75 [95% CI, 0.52 to 1.09]), but no difference was seen in people experiencing a fall. Two trials of exercise+psychological (n=153) and exercise+vitamin D (n=204) showed no effect on falls (IRRs, 0.94 and 0.99); however, the exercise+vitamin D trial showed a large statistically significant reduction in injurious falls (IRR, 0.38 [95% CI, 0.17 to 0.81]). The evidence on QOL, ADL, IADL, or mortality outcome was limited. No trials reported the outcomes of fracture, people experiencing fracture, people hospitalized, or people transitioning to institutional care.

Characteristics of Included Studies

Of the six included multiple intervention studies, two were included in the previous review,^{71, 172} and four were newly identified.^{113, 120, 122, 173}

Study Characteristics

We identified three fair-quality^{122, 171} and three good-quality^{113, 120, 172} RCTs (n=1,770) with a primary or secondary aim of examining the effectiveness of multiple interventions on falls, fall-related injuries, or both (Table 52). One trial was conducted in the United States,¹⁷² two in Australia,^{122, 171} two in Germany,^{120, 173} and one in Finland.¹¹³ Three RCTs had factorial designs^{113, 120, 122} and one was a cluster-randomized trial of general practices.¹⁷³ Trial sizes ranged from 153¹²⁰ to 453¹⁷² participants (randomly allocated to multiple interventions or matched control groups). Followup time ranged from 12 to 24 months. The outcomes reported in the trials that we included for analysis were falls (k=6; n=1,770), people experiencing a fall (k=4; n=1,413), injurious falls (k=5, n=1,460), people experiencing a fall-related injury (k=1, n=378), mortality (k=3; n=1,035), QOL (k=1; n=258), and harms (k=3, n=810). No studies reported on hospitalization, institutionalization, ADL, or IADL.

Table 52

Study Characteristics, for Multiple Interventions, by Author.

Population Characteristics

Four of the six studies recruited participants aged 70 years and older;^{113, 120, 122, 171} two of these studies had upper age limits (80 years¹¹³ or 90 years¹²⁰) (Table 53). Two RCTs recruited participants aged 65 and older.^{172, 173} Mean age ranged from 74.0 years¹¹³ to 78.4 years.¹⁷¹ One study recruited women exclusively,¹¹³ while in the other five studies women comprised nearly half ¹²⁰ to up to three-fourths of the participants.^171-173 Only the U.S. study reported on race or ethnicity; nearly all were white (95%).¹⁷² Measures of socioeconomic status were reported in only the study by Freiberger et al¹²⁰ (35.4 percent with low educational attainment and 25.7 percent with low income). All but one study¹⁷³ excluded patients with cognitive impairment.

Table 53

Population Characteristics, for Multiple Interventions, by Author.

All six studies recruited community-dwelling adults. Four of the studies recruited a general population of participants from a community^{171, 172} or used population-based registries.^{113, 122} The remaining studies recruited participants from a health insurance company database¹²⁰ or clinic.¹⁷³

Two trials recruited participants at average risk of falling where the only risk factor for falls was the participant’s age.^{122, 172} Four trials recruited patients at high risk for falls, as defined as either a history of falls (in the previous 6 or 12 months),^{113, 120, 171, 173} a fear of falling,^{120, 171, 173}, low physical function (TUG test or Chair Stand Test >10 seconds) or balance deficits.¹⁷³

The measures of health or functional status at baseline in the studies varied. The measures included a history of falls, comorbid conditions, number of medications, baseline ADL or IADL score, quality of life, living alone, and self-reported health rating.

Intervention Details

The trial by Clemson et al¹⁷¹ (knowledge+environment assessment) offered educational sessions (2-hour sessions each week for 7 weeks), an educational session reviewing safety hazards in the home, a home visit and 1.5 hour booster session 3 months after the last session (Table 54). The trial by Fitzharris et al¹²² (exercise+environment+vision intervention) included supervised group exercise classes (1-hour weekly session for 15 weeks), a home visit to modify hazards, and vision screening with appropriate referral. The trial by Freiberger et al¹²⁰ (exercise+psychological) included progressive supervised group exercises or a multicomponent cognitive behavioral program that addressed the thoughts and concerns of elderly people about falls and the hazards of falls (1-hour sessions twice per week for 16 weeks). This trial also provided cognitive training by using exercises to improve participants’ ability to concentrate, process information faster, and improve short-term memory. The second exercise+psychological trial was conducted by Siegrist et al¹⁷³ and included a progressive supervised group exercise program with strength, power, balance, and gait training, a self-management program with perception and functional training, and a cognitive behavioral program aimed to increase self-efficacy (1-hour weekly sessions for 16 weeks) and a 12-week home exercise program. The trial by Shumway-Cook et al¹⁷² (exercise+knowledge+falls risk assessment) involved a total of 156 progressive group exercise sessions (1-hour sessions 3 times per week), nurse-led educational classes (1-hour monthly classes for 6 months), and an assessment of the participant’s risk of falling that was mailed to his or her PCP. The factorial design trial by Uusi-Rasi et al¹¹³ (exercse+vitamin D) offered 800 IU daily of cholecalciferol plus 78 weekly or twice weekly group exercise classes offered over 104 weeks.

Table 54

Intervention Details, for Multiple Interventions, by Author.

For four of the RCTs that used multiple exercise components,^{113, 120, 122, 172} group exercise sessions were supervised by exercise instructors, physical therapists, or fall-prevention instructors. The interventions in all five trials lasted between 7 weeks¹⁷¹ and 104 weeks.¹¹³ All five RCTs included some exercise for gait and balance as well as for strength and resistance. One study specifically mentioned flexibility training¹⁷²; however, stretching was a part of other study protocols as well.^{113, 120} The trial by Clemson et al¹⁷¹ had the least intensive exercise component (educational sessions with brief exercise practice and review).

The intensity and duration of the supervised group exercise classes varied, as detailed above. The total number of exercise sessions offered in the trials ranged from 7¹⁷¹ to 156¹⁷² sessions over 15¹²² to 104 weeks.¹¹³ Four RCTs also made recommendations for home training programs.^{113, 120, 122, 173}

Control groups received usual care.^{113, 120, 122, 171-173} One study provided brochures on preventing falls.¹⁷² In another study the control group received two social visits.¹⁷¹ In the single study with the vitamin D intervention, the control group received placebo in addition to usual care.¹¹³

Falls

Six multiple intervention trials (n=1,770) reported a falls outcome.^{113, 120, 122, 171-173} The rate of falls in the control groups ranged from 0.51 to 2.4 per person-year in the individual studies. The three RCTs designed to have adequate power for falls (exercise+environment+vision,¹²² knowledge+environment,¹⁷¹ and exercise+psychological¹⁷³) showed statistically significant reductions in falls in the multiple intervention group (IRR, 0.80 [95% CI, 0.65 to 0.98]¹²²; IRR, 0.68 [95% CI, 0.57 to 0.83]¹⁷¹; IRR, 0.54 [95% CI, 0.35 to 0.84]¹⁷³). Unlike the other studies in this group, the trial by Clemson et al¹⁷¹ did not have a distinct dedicated supervised exercise program; instead, it provided eight weekly 2-hour educational sessions in which some exercise education and practice was a part of three of the sessions. One underpowered exercise+knowledge+falls risk assessment RCT¹⁷² showed a nonstatistically significant trend that favored the intervention group with an IRR in the same range as the studies by Fitzharris et al¹²² and Clemson et al,¹⁷¹ with an IRR of 0.75 (95% CI, 0.52 to 1.09).¹⁷² The one RCT of exercise+vitamin D, which had the most intensive and longest duration of the exercise intervention, showed no difference in falls (IRR, 0.99 [95% CI, 0.72 to 1.39])¹¹³ (Figure 19; Table 41).

Injurious Falls

Five studies (n=1,460) reported injurious falls at 12 to 24 months.^{113, 120, 122, 172, 173} One study of exercise+vitamin D showed a 62 percent reduction in injurious falls (i.e., falls resulting in medical care) at 24 months (RR, 0.38 [95% CI, 0.17 to 0.81])¹¹³; interestingly, this exercise+vitamin D trial showed no effect on overall falls (IRR, 0.99 [95% CI, 0.72 to 1.39]). Three studies (exercise+knowledge+falls risk assessment,¹⁵⁷ exercise+environment+vision,¹⁰⁵ and exercise+psychological¹⁷³ showed similar nonstatis tic ally significant reductions in injurious falls but with different definitions (i.e., minor or major injuries or those requiring hospitalization,¹²² falls resulting in medical care,¹⁷² or not defined¹⁷³) (IRR, 0.72, ¹⁴⁷ 0.79,¹⁷³ and 0.80¹⁰⁵ respectively). The remaining trial of exercise+psychological intervention showed no effect on injurious falls.¹²⁰ (Figure 3; Table 42).

Fractures

No trials reported this outcome.

People Experiencing a Fall

Four multiple intervention RCTs (n=1,413) demonstrated mixed results regarding the number of people experiencing a fall.^{122, 171-173} The prevalence of people in the control group who fell was similar in the three studies (49.4%,¹⁷³58.2%,¹⁴⁶ 63.5%,¹⁰⁵ 57.3%¹⁴⁷). The trial by Fitzharris et al¹²² (n=272) (exercise+environment+vision) reported a 33 percent statistically significant reduction in people experiencing a fall at 18 months of followup (RR, 0.67 [95% CI, 0.51 to 0.88]). The other three RCTs showed no difference at 12 to 14 months of followup (n=310; RR, 0.90 [95% CI, 0.73 to 1.10], ¹⁴⁶ n=453; adjusted RR, 0.96 [95% CI, 0.82 to 1.13], ¹⁴⁷ and n=378; RR 0.85 [95% CI, 0.68, 1.06]). The two studies, which reported the number of people who had had two or more falls, revealed a beneficial trend favoring the multiple intervention group (RR, 0.74¹⁷¹ and 0.70¹²²), but they were not statistically significant (Figure 20; Table 43).

People Experiencing an Injurious Fall

One exercise+psychological trial (n=378) reported a reduction of people experiencing an injurious fall in the intervention group compared to the control group (RR, 0.75 [95% CI, 0.56 to 1.00])¹⁷³ (Table 55).

Table 55

People Experiencing an Injurious Fall, for Multiple Interventions, by Author.

People Experiencing a Fracture

No trials reporting this outcome.

Mortality

Three studies (n=1,035) reported mortality outcomes at 12 to 24 months (Figure 21; Table 48).^{113, 172} The individual studies reported RRs of 0.67 (95% CI, 0.11 to 3.97) for the exercise+knowledge+fall-risk assessment trial¹⁴⁷; 0.25 (95% CI, 0.01 to 5.48) for the exercise+vitamin D trial, ¹¹⁵ and 0.56 (95% CI, 0.23 to 1.39) for the exercise+psychological trial.¹⁷³ The wide CIs reflected a relatively uncommon outcome with 10 or fewer events in each group.

People Transitioning to Institutional Care

No trials reported this outcome.

People Hospitalized

No trials reported this outcome.

QOL

Only one study (n=258) reported on SF-36 physical and mental health components.¹⁷¹ No statistically significant mean differences between the multiple intervention group and the control group were found (SF-36 physical component mean difference, 0.70 [95% CI, -2.94 to 1.88]; SF-36 mental health component mean difference 0.53 [95% CI, -2.95 to 1.88]) (Table 44).¹⁷¹

ADL

No trials reported this outcome.

Fall Risk Status

Most of these multiple trials (4 of 6) recruited participants at high risk for falls based on history of prior falls, fear of falling, or functional tests of balance/gait; evidence is too limited to make conclusions about associations between baseline fall risk and treatment effectiveness.

Harms

Three trials (n=810) reported either no adverse events^{120, 172} or no severe adverse events¹¹³ associated with the intervention. However, it is unclear how the adverse events were collected or measured.^{113, 120, 172}

Critical Appraisal

Due to the various combinations of intervention types, these fair- to good-quality trials of community-dwelling older adults do not provide a coherent body of evidence; thus, pooling the results is inappropriate. Three trials showed statistically significant reductions in the rate of falls (and one of these showed a reduction in people experiencing a fall as well) despite the differences in the intervention combinations (knowledge+environment, exercise+environment+vision, exercise+psychological).^{122, 171, 173}