Exercise frequency in the studies reviewed predominantly ranged from once per week to daily exercise. Exercising once per week appears to be inadequate for falls prevention, with only one pilot study reporting a short-term reduction in falls incidence at that frequency , although it may be sufficient to produce some balance and functional improvements . The minimum effective frequency reported was twice per week , although this effect was noted in the short term only and was not supported by studies of longer duration [10, 22, 35]. Exercising three times per week was the most commonly adopted and most consistently effective approach across the studies reviewed [4, 8, 9, 27, 33, 37, 43, 44, 47, 53, 56, 64]. Where effects were non-significant, some instances of trends towards reductions in falls rates were observed [52, 63]. In other cases, additional factors such as exercise type and intensity may have been influential [29, 32, 42].
Mixed evidence was observed for programmes involving more frequent exercise. Studies aiming for daily exercise reduced falls rates by approximately 21–47 % [2, 17, 28, 30, 62], although adherence to the desired frequency was poor. After 1 year, Barnett et al.  found that only 13 % of participants were exercising daily at home, with the vast majority (91 %) exercising just once per week outside of group exercise classes. Similarly, the participants of Day et al.  reported exercising approximately twice per week outside of group sessions. This suggests that very high exercise frequencies may not be acceptable to older adults, but further supports the finding that exercising three times per week is effective. These findings confirm those of Costello and Edelstein  by indicating that the optimum frequency of exercise for falls prevention is three times per week, since it allows significant benefits to be attained while remaining acceptable to this population.
Clear guidelines exist describing appropriate strength and endurance training intensities for older adults . Many studies stated that their programmes met these guidelines in relation to one or both types of training [4, 30, 32, 35, 47], but most did not provide sufficient detail of exercise intensity. All studies in which guideline intensities were explicitly not met did not significantly reduce falls incidence [28, 29, 42, 55]. This highlights that strengthening and endurance exercise must be of an appropriate intensity to achieve training effects and the ensuing clinical benefits.
Defining an optimal intensity for balance training is problematic, since there is currently no standard measure by which to express the balance training intensity. Thus, its reporting in the studies reviewed was vague and inconsistent, with descriptions including ‘demanding’ , ‘challenging’  and of ‘appropriate and increasing levels of difficulty’  being used. Sherrington et al.  categorised balance training intensity according to the presence or absence of certain components: moving the body’s centre of mass, reducing the base of support and minimising upper limb support. Based on these criteria, the majority of reviewed studies which included balance training may be classified as ‘highly challenging’.
Progression of intensity is vital for continuing training gains to be obtained, but maintaining safety while providing adequate challenge is essential. Costello and Edelstein  proposed that balance training should therefore be conducted at the highest possible level of difficulty without falling or near-falling to ensure sufficient training intensity, and that each exercise should be mastered before progressing to ensure safety. Structured programmes such as the Otago Exercise Programme (OEP) [8, 9, 43, 44] provide a useful framework for clinicians to prescribe and progress balance training, although these principles have been adopted effectively in less rigidly structured programmes without adverse events [17, 19, 37, 61].
Time spent exercising may be considered in numerous ways: the duration of each exercise bout, the duration of the intervention or the total exercise volume, i.e. the cumulative time spent exercising throughout an intervention. The duration of a single bout of exercise varied from approximately 15 to 120 min in the studies reviewed. No consistent relationship between bout duration and the effectiveness of an intervention was observed, although the majority of effective interventions included some bouts of at least 60 min or more. Bouts of longer durations, i.e. 90–120 min in length, were all conducted in group settings—including time for explanation, demonstration, etc.—rather than as single, continuous bouts of exercise.
The total duration of the exercise programmes varied greatly. Effective interventions lasted from 5 weeks  to 2 years . The most commonly observed programme durations were from 15 weeks to 12 months. Most interventions of approximately 4 months duration were reported to reduce falls [17, 19, 28, 62], while 12-month interventions reduced—or showed trends towards reducing—falls incidence [2, 9, 33, 43, 44, 52, 64, 65], with some exceptions [35, 42, 63]. The length of follow-up must be considered as a potential source of bias in short-term studies. Reporting the immediate effects of short-term interventions may show reductions in falls incidence, but falls may occur infrequently, thus participants must be observed over an extended time period to detect true changes. Some evidence suggests that the effects of shorter interventions may be carried over to long-term follow-up [56, 61], while the results of Yamada et al. negate this . Further research investigating participants’ continuing exercise behaviours upon completion of short-term interventions is required.
Total exercise volume also varied; even among interventions of equal duration, e.g. tai chi group of Wolf et al.  spent 64 h exercising compared to approximately 11 h for the balance training group, despite both interventions lasting 15 weeks. To overcome the challenge this presents in comparing interventions, it may be helpful to consider the exercise volume per week for each intervention (Table 2). When considered in this manner, all effective interventions were found to involve at least 1 h of exercise per week. Ineffective interventions that exceeded this volume were noted to have insufficient exercise frequency [10, 22, 35], questionable exercise intensity [20, 42, 55, 63] or involve exercise of an inappropriate type [29, 32], with the exception of the intervention of Shumway-Cook et al.  which displayed a trend towards significant falls reduction. Overall, effective interventions comprised of at least 40 h of exercise over the course of the intervention, slightly less than the cutoff of 50 h suggested previously [50, 51]. Interventions with high exercise volumes [33, 52] were reasonably successful, but cited poor adherence as a barrier to further success. The OEP involves approximately half the volume of exercise of that prescribed by Shumway-Cook et al.  and is successful in both preventing falls and gaining adherence . Thus, while no definitive total exercise volume can be recommended, it is apparent that exercise programmes must meet certain minimum requirements while remaining acceptable to participants to be effective.
Guidelines recommend a comprehensive programme of balance, strength, endurance and flexibility training for all adults aged 65 and over . A number of the studies reviewed implemented such a programme with the effect of reducing falls incidence [2, 20, 30, 65], including the OEP which reduced fall rates and fall-related injuries by approximately 32 % . Similarly, the 36-week FaME programme  delivered a home programme based on the OEP combined with a supervised exercise class which included more challenging balance exercises. This intervention produced a similar immediate reduction in falls rates, increasing to 54 % after 1 year, suggesting additional benefits from group classes and more challenging balance training.
Freiberger et al.  investigated the optimum amount of emphasis to place on each component of a comprehensive programme. Although it did not reach significance, their fitness group—in which strength/flexibility, balance and endurance training each comprised 33 % of the programme—experienced a trend towards a reduction in falls incidence, while the psychomotor group—in which strength and balance combined comprised just 40 % of the programme—did not. Similarly, Day et al.  observed a reduction in falls incidence and improvements in balance with a comprehensive programme which is comprised of 30–35 % balance training. These results suggest that balance training should constitute at least one third of the total programme content and be given at least equal emphasis compared to other components for optimum fall prevention.
The nature of balance training undertaken varied between studies, although training principles remained similar. Exercises which encouraged reducing or leaning beyond the base of support, shifting the body’s centre of mass, minimising upper limb support, coordinating single or dual-task movements, altering sensory feedback and functional activities were commonly used in balance training. Some effective novel approaches included computerised balance training , group games [39, 47] and obstacle courses [47, 61]. Tai chi was also consistently successful in preventing falls and appears to be an effective method of integrating many principles of balance training into one accessible programme [27, 33, 56, 62–64].
Certain characteristics were common to ineffective programmes: firstly, the lack of a balance training component [32, 63]. Secondly, programmes which lacked functional relevance were ineffective [29, 32]. Finally, programmes which lacked exercise progression were also ineffective in preventing falls .
The inclusion of walking in falls prevention exercise programmes has been identified as a contentious issue, with suggestions that walking may increase fall-risk exposure and reduce the emphasis on vital balance training . In the studies reviewed which included a walking component [2, 8, 9, 20, 30, 43, 44, 47, 65], walking was not consistently associated with the effectiveness of the intervention. However, Rubenstein et al.  showed some association between exercise—which included walking—and increased falls risk. Clinicians may identify individuals who are at high risk of falling and would be unsafe undertaking a walking programme independently and chose not to recommend walking in such cases, an approach adopted in the OEP and supported by Sherrington et al. . Other lower risk forms of endurance training, e.g. stationary cycling, swimming, etc., may be substituted if desired.
Delivery and adherence
Shumway-Cook et al.  demonstrated the importance of adherence to the success of a falls prevention programme, with those who attended more than 75 % of exercise classes having 41 % fewer falls than those who attended less than 33 % of classes. How best to ensure adherence remains unclear, although methods of exercise programme delivery may be influential.
Although a uniform population was targeted, various settings were observed for the exercise interventions reviewed, most commonly community centres and participants’ homes. Exercise in either location was effective, as were combined centre- and home-based programmes. A convenient location and accessibility via transport links are vital, as these were cited by participants as major contributing factors to dropping out of a programme . However, Day et al.  provided transport for their participants to attend exercise classes, yet 26 % of participants never attended a class, and only 61 % attended more than half of their sessions. In addition, participants’ adherence to the recommended daily home exercise programme was poor despite being able to complete this in their own homes. It is clear, therefore, that although location is undoubtedly influential, other factors strongly influence exercise uptake and adherence.
Supervision and format (i.e. individual/group sessions) may be two such factors. Supervised group exercise is thought to facilitate uptake and adherence due to the leadership, social support and social outlet provided . However, empowering individuals and encouraging self-regulated behaviour change are recommended to gain long-term motivation and exercise participation [3, 45]. In the studies reviewed, a combined approach with supervised group exercise supplemented by an individual home exercise programme appeared to be a common and effective choice of intervention, potentially providing a beneficial mix of both approaches. Providing individualised home exercise programmes with limited one-on-one supervision, as in the OEP, may also allow similar benefits to be attained, but the cost-effectiveness of this approach does not compare well to group programmes . A novel approach that aimed to achieve adherence and long-term behaviour change by embedding training activities into everyday tasks was also investigated in one pilot study with promising results . Larger trials of this approach are required to determine its acceptability to older adults.
Programme characteristics can also influence adherence. A strong negative correlation between exercise bout duration and adherence has recently been demonstrated in older women  and must be considered when designing an exercise programme for older adults. Low exercise frequency and intensity have been cited as facilitators to programme participation , although—as discussed—minimum requirements must be attained for a programme to be effective. Interventions of 5–12 weeks duration achieved excellent uptake and adherence, while longer interventions were associated with poorer uptake and adherence. However, as already stated, no studies examined long-term exercise behaviours of older adults following completion of short-term falls prevention programmes; thus, it is not possible to ascertain whether shorter interventions can bring about long-term positive changes in exercise behaviour.
Since this review focused on identifying optimum exercise programme characteristics, only studies which examined exercise as a single intervention were included. Thus, the effect of exercise as part of a multifactorial intervention has not been considered. However, a factorial study  demonstrated that exercise was effective singly and in combination with home hazard modification and vision correction. In fact, combining all three interventions was found to bring about the greatest reduction in falls rates. This indicates that—while effective alone—a comprehensive exercise programme is also a vital component of a successful multifactorial intervention.
In terms of outcome measurement, not all studies reviewed utilised prospective falls diaries or calendars to record falls incidence, despite this being the current gold standard in falls data collection . This may lead to inaccuracy in reports of falls incidence rates where retrospective methods are used. It may also render comparison with studies using prospective methods invalid. Statistical methods used to report changes in falls incidence also varied, making direct comparison between studies challenging.