Preventing falls among older people with mental health problems: a systematic review

We searched for published and unpublished English language studies that evaluated an intervention aimed at preventing or reducing falls in older people with a mental health problem, regardless of setting. Studies were identified from searching a range of electronic databases including Pubmed, NHS evidence, Cochrane Library inc, Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, CENTRAL, NHS Economic Evaluation Database and the Health Technology Assessment Database, Cinahl, AMED, BNI, Embase, HMIC, PsychInfo; and by using lateral search techniques such as checking reference lists, and using the ‘Cited by’ option on Web Of Science (WoS) and Google Scholar, and the ‘Related articles’ option on PubMed and WoS. Searches were first conducted in February 2011 and updated in November 2012 with lateral searching ongoing until October 2013; no date restrictions were applied. Further details of the search terms used can be seen in Table 1.

We included both randomised controlled trials and controlled evaluations of fall prevention interventions for people aged 60 and over with any mental health problem including dementia, depression or psychosis. Studies that did not have a specific mental health focus were included as long as a significant percentage (e.g. 50% or more) of participants had a mental health problem, or data was reported separately for those with a mental health condition. As preliminary searches suggested there was a lack of controlled studies we included uncontrolled studies, but only if all participants had a mental health problem. We included single focus or multi-factorial interventions involving environmental, exercise, technological, psychological, educational, and health related components. Our primary interest was in studies delivered in in-patient mental health settings but interventions delivered in other settings were included as long as the study included older people with a mental health problem. The primary outcomes of interest were the number of participants sustaining at least one fall (fallers) and data relating to the rate or number of falls. Secondary outcomes included fall related injuries, hip fractures, service use and patient satisfaction. In addition, we searched for qualitative studies or process evaluations that identified barriers and facilitators to the implementation and uptake of interventions; in particular looking at whether specific guidance is required for this group.

Two reviewers independently screened titles and abstracts of citations identified by electronic searches, applied the selection criteria to potentially relevant papers and extracted data from included studies using a standardised form which was piloted prior to use. Any disagreements were resolved by consensus or by discussion with a third reviewer. We extracted information on the type of study design, study aims, participants, setting, intervention (including details of provider, duration and intensity) and outcomes.

Two reviewers independently assessed the quality of included studies. RCTs and controlled studies were assessed using the Cochrane Collaboration risk of bias tool [20], and uncontrolled studies were assessed on domains adapted from two check lists [21, 22]. Full details of the quality domains can be seen in Table 2.

The interventions in the included studies were classified using the fall prevention classification system developed by the Prevention of Falls Network Europe (ProFANE) [23]. This groups studies by setting (nursing/residential care, hospital, community), by combination (single, multifactorial) and by type of intervention including exercise, social environment, environment/assistive technology, knowledge, and other (multi-sensory stimulation). Populations were not homogenous and the Chi-Square test and I² test [24] revealed significant heterogeneity and thus studies were not pooled in a meta-analysis. Instead we have presented data in the text and in a table with an indication of whether the effect of the intervention was positive, negative or not statistically significant. Where possible we have reported dichotomous outcomes as relative risks or incidence rate ratios and continuous data as mean differences, both with 95% confidence intervals. Where data were not available to allow us to calculate effect sizes we have presented data as reported in the paper (for example P values).

We identified 4614 studies from our electronic and lateral searches of which 27 papers [25‐51] reporting 21 studies met the inclusion criteria. Studies were published between 1997 and 2013, with all but one published in the last ten years. Seventeen studies were randomised, with eight being cluster RCTs [26, 27, 33, 34, 39, 41‐43], and nine with randomisation at the level of the individual [28, 32, 36, 38, 44, 45, 48, 50, 51]. Four studies were uncontrolled [29, 30, 37, 46]. Length of follow up ranged from 3 to 12 months. We found no qualitative studies or process evaluations. A flow chart detailing the identification of studies can be seen in Figure 1.

Nine studies were conducted in Europe, five in Australia, five in North America and two in Japan. Fourteen studies took place in nursing homes or residential care [26‐30, 34, 36, 39, 41‐43, 45, 46], or included participants who mostly resided in nursing or residential care [52], three took place in hospital [33, 38, 48], one in a respite day centre [37], one a geriatric outpatient clinic [32] and two in participants’ own homes [44, 51]. In ten studies participants all (or most) had dementia or cognitive impairment [27‐30, 36‐39, 45, 51, 52], four studies reported sub-group analyses for participants with cognitive impairment [33, 35, 41] or depression [44] and the rest included 48% or over of people with mental health problems (largely dementia).

Ten studies involved multi-factorial interventions [32, 34, 37, 39, 41, 42, 44, 47, 48, 51]; and included components such as staff training, physical activity or training and environmental assessment. Of the single component interventions two focused on knowledge based interventions [26, 33], five on the environment [27, 29, 30, 38, 46] two on physical activity or exercise [28, 43], and two on sensory stimulation, one involving , olfactory stimulation with lavender oil [45], and the other multisensory stimulation through a Snoezelen room [36]. ProFANE classifications can be seen in Table 3 and further details of included studies (including the links between papers) in Table 4.

All 21 included studies reported falls data in some form, although seven did not include a specific definition of a fall [27, 28, 30, 36‐38, 46]. In the others a fall was defined as unintentionally [34, 41‐44, 48, 51], inadvertently [33, 39, 45, 47], or unexpectedly [32] coming to rest on the ground, floor, or a lower level; whether or not an injury was sustained [34, 43, 44, 47]. One study included syncopal falls [48], one both witnessed and non-witnessed falls [29]; one only took account of falls where medical intervention was needed [26], and one categorised injurious falls as those resulting in serious injuries that received medical treatment [42].

Results of the risk of bias assessment can be seen in Figure 2 (RCTs) and Figure 3 (uncontrolled studies). Of the 17 randomised controlled studies sequence generation was considered adequate in thirteen studies, as was allocation concealment. However, only ten studies were judged to have both adequate sequence generation and allocation concealment [27, 33, 36, 38, 41, 44, 45, 47, 48, 51]. Four studies [33, 45, 47, 48] met all quality criteria and six met five or more of the six quality criteria [32, 41, 43, 44, 47, 51]. One RCT [28] met none of the criteria. Of the four uncontrolled studies three met three out of six of the criteria [29, 37, 46] and one met two out of six [30], but all are at high risk of bias because they have no control group.

Results are presented by setting and according to whether the intervention was single or multifactorial. A summary of the data, including effect sizes, can be seen in Table 5. Study quality is reported in the text for those studies judged to be at low or high risk of bias; all quality scores can be seen in Table 5.

Two studies looked at exercise programmes in nursing or residential care. One [28] (high risk of bias) found a significant reduction in numbers of fallers and the other (low risk of bias) [43] no difference in number of fallers but a non-significant reduction in rate of falls and hip fractures. Four studies looked at some form of environmental intervention. One, a study of a wander garden found a significant reduction in the mean number of falls [30] but this was an uncontrolled study and at high risk of bias. Of the other social environmental interventions two studies involving increased supervision [29, 46] reported a significant reduction in number of fallers, but both were at high risk of bias. A study comparing dementia care mapping (DCM), person-centred care (PCC) and usual care found a reduction in falls in the DCM group compared to control but an increase in the PCC group compared to control [27]. One study which involved a knowledge based intervention for staff reported a significant reduction in fallers but no difference in the rate of falls [26]. Two studies evaluated a sensory intervention. One study of lavender patches (low risk of bias) found a non-significant reduction in number of fallers and a significant reduction in incidence of falls [45] and a study of multisensory stimulation in a Snoezelen room found no difference in the number of falls in the intervention group compared to control [36].

Five studies looked at multifactorial interventions in nursing homes or residential care. Common components of these interventions were exercise, medication management, changes to the environment and activities to improve staff knowledge. Two studies report a significant reduction in the rate of falls [40, 41] in participants with cognitive impairment. However, in the latter there was a non-significant increase in the number of hip fractures in the intervention group compared to control. One low risk of bias study found a slight (non-significant) reduction in fallers but a non-significant increase in fall related A&E admissions [52], and one a non-significant reduction in injurious falls and number of recurrent fallers [42]. One study which undertook subgroup analyses for participants with cognitive impairment found that the intervention significantly reduced the number of fallers in those with no cognitive impairment but not in the group with cognitive impairment [35].

Neither of the hospital based studies found a reduction in falls in those with cognitive impairment. A study of multimedia patient education (low risk of bias) compared multi-media education with health professional follow-up, multi-media materials alone and usual care [33]. Although the complete intervention reduced falls amongst participants who were cognitively intact this was not the case for those with cognitive impairment. Amongst those with cognitive impairment there was an increase in the number of fallers, fall incidence and injurious falls in the group that received the complete intervention but no difference in fallers or fall rate in the group that received materials only. The other hospital based RCT evaluated the impact of nurse led individualised advice to staff aiming to reduce agitation and the use of psychotropic medication in patients with confusion and behavioural problems. They found a non-significant increase in falls in the intervention group compared to control [38].

There was one (low risk of bias) multifactorial hospital based intervention [48]. They found a significant reduction in fallers and in fall incidence in people with dementia. There was also a reduction in injuries and serious injuries.

Four studies took place in the community. Only one [44], a multifactorial fall prevention program, found a significant reduction in falls. Although the study did not focus on people with a mental health problem they do report results of a sub-group analysis in which they found a reduction in the incidence of falls in people with depressive symptoms. A pilot RCT found a non-significant reduction in fallers and incidence of falls in older people receiving a home hazard reduction and exercise fall prevention programme [51], an uncontrolled pilot study of a fall prevention programme for people with dementia found no difference in fallers [37] and a multifactorial intervention in a geriatric outpatient clinic reported a non-significant increase in fallers and falls [32].

We found 17 RCTs and four uncontrolled studies that evaluated the impact of interventions to prevent falls in older people with mental health problems, including dementia, delirium and depression. The nature of the interventions varied considerably and involved a variety of components including physical activity and exercise, risk assessment, environmental modification, staff training, increased supervision, patient education, and sensory interventions. Evidence relating to fall reduction was inconsistent. Of the 14 studies that reported the number of fallers, eight found a reduction which was statistically significant in five, and of the 14 that reported rate or number of falls eight found some evidence of a significant reduction. However, four studies found a non-significant increase in falls in the intervention group compared to the control.

Although all the studies included participants with mental health problems only ten studies had a specific mental health focus. Studies predominantly focused on cognitive impairment, dementia and depression and there were no studies including patients with other psychiatric disorders who are likely to be on medications associated with increased risk of falls. The majority of studies were undertaken in nursing and residential homes, and there were none in mental health inpatient settings.

Cochrane reviews looking at fall prevention interventions for older people have found evidence that multifactorial interventions are effective in preventing falls in community dwelling participants and also those in hospitals and nursing care facilities, and that multi-component exercise is effective in community settings [53, 54]. We also found evidence to suggest that multifactorial interventions which included exercise can be effective in older people with cognitive impairment and depression, with four of six studies finding a reduction in falls which was significant in two [40, 48] although there is insufficient evidence to support physical activity or exercise alone with this population.

Only six of the studies, all multifactorial in design, included medication review. This is surprising given that the American Geriatric Society and British Geriatric Society [55] in their Clinical Practice Guideline for fall prevention, argue that the strongest risk associations for falls are with psychotropic medication and polypharmacy and that evidence supports withdrawal or reduction of psychotropic medication.

Although the evidence from this review is not robust it does challenge the American Geriatrics Society and British Geriatrics Society guideline [55] which states that there is insufficient evidence to support any recommendation to reduce fall risk for older people with cognitive impairment. However, the review does support current UK guidance from the National Institute for Health and Care Excellence [56] which recommends that all older people at increased risk of falling are considered for individualised multifactorial interventions including strength and balance training, home hazard assessment, vision assessment and medication review; and that those in in care settings such as care homes receive multifactorial interventions including exercise. In addition, all older people taking psychotropic medication should have this reviewed and where possible discontinued. NICE guidance also recommends that health care professionals dealing with patients at risk of falling should be educated in falls assessment and prevention. A recent audit of UK inpatient NHS settings that included mental health settings [57] found that 35% of nurses and 61% of doctors had not received training in fall prevention in the past year thus highlighting that work is needed in order to meet NICE recommendations.

We used systematic and rigorous methods to synthesize the current evidence on the effectiveness of fall prevention interventions for older people with mental health problems. However, there are a number of methodological issues that could have a bearing on the validity of these results. Due to the scarcity of evidence relating to falls prevention in older people with mental health problems we included four uncontrolled studies which, because they are at increased risk of bias, should be interpreted cautiously. Moreover, only four of the controlled studies met all of the methodological criteria on our checklist and many appeared underpowered to detect a reduction in falls. We included seven studies that included participants with and without mental health problems and which did not report sub-group analyses for those with mental health problems. It is possible that the results from these studies are less generalizable to older people with mental health problems.

The review involved a diverse range of interventions, participants and outcomes and in light of this heterogeneity we judged meta-analysis to be inappropriate. This made direct comparisons between studies more difficult. Nevertheless, despite this, we were still able to make judgements about the strength and consistency of the findings.

The lack of studies we found could reflect the additional methodological and ethical challenges which work in these settings and with these population groups entails. Researchers have highlighted the difficulties of undertaking studies with older patients with mental health problems [41, 52], in particular when there are issues around capacity to give informed consent [58]. However, the UK Mental Capacity Act [59] provides researchers with the means to address this, through using assent, though the process adds time and thus cost to the research. Despite this, researchers need to overcome these difficulties in order to provide practitioners with the evidence of effectiveness they need for practice.

Older people with mental health problems are likely to be older (as incidence of dementia increases with age) and frail with physical and mental health co-morbidities and this needs to be accounted for when designing interventions. The authors of one of the studies included in our review [32] suggest that their intervention failed because it was too complex and specialised and that interventions need to be tailored to the needs of frail older people, particularly when incorporating intensive physical therapy.

Alderson and Roberts [60] argue that reviews such as this one, where the review uncovers very little evidence which practitioners can use, still need to be reported. They argue that uncovering uncertainty can help to improve the evidence base through stimulating more and better quality research.

Staff working with patients with mental health conditions need access to evidence to support their practice with this population group. Staff may be aware of the risks factors for falls, however strategies to prevent falls frequently include choosing management strategies such as restraint and observation [19, 61] rather than interventions which would increase balance and strength. In addition, in-patient mental health settings present unique patient-safety issues [62]. Without evidence of effectiveness of interventions staff will continue to struggle to provide care for those at risk of falls who also have a mental health issue such as dementia [19]. Further work is needed to help us to understand what interventions work in which sub groups of patients. This review therefore adds further weight to the calls for further research to help understand which elements of multifactorial interventions work, in what combination for specific groups of patients, the acceptibility of these to older people, and their cost-effectiveness [41, 52, 55, 56].