Abstract
Dementia affects approximately 6% of persons over 65 years of age, increasing to 24% of those aged 85 years and over [1]. Difficult behaviours such as wandering, resistiveness, aggression and noisiness emerge in a substantial proportion of cases. In a British community survey, aggression was reported by the carers for 4% of persons with mild dementia, 14% of those with moderate dementia and 42% of those with severe dementia. Noisiness was described in 8%, 20% and 42% of mildly, moderately and severely affected persons, respectively [2]. The number and severity of disturbed behaviours correlated strongly with carer burden [2,3] and greatly increased the likelihood of admission to residential care [4].
As a result, challenging behaviours are found most commonly in residential facilities. Prevalence rates in nursing homes vary widely from one study to another because of differences in sampling methods [5], definitions and behavioural measures [6]. In a survey of a representative sample of more than 10 000 residents of Australian hostels and nursing homes, staff rated 32% of nursing home residents as having mild behavioural disturbances, 22% as moderate and 14% as severe [7]. In one North American nursing home, 37% of residents were verbally abusive, 24% hit, kicked or threw things, and 11% shouted or screamed at least once each day [8].
Psychotropic medications are used very commonly to reduce the frequency and severity of these behaviours. In the Australian study mentioned previously, 43% of mainstream nursing home residents with moderately severe dementia had taken hypnotics or other sedative medications in the previous fortnight, 42% had taken neuroleptic medications and 18% had taken antidepressants [7]. Evidence in support of their usefulness is limited, however, and side effects are frequent and often hazardous [5].
Non-pharmacological strategies such as behaviour modification and music therapy are promoted as safe, humane and at least as effective as medications, but the evidence for this view is sometimes lacking. This deficiency should be rectified, however, to ensure that treatment strategies are firmly based on evidence acquired through rigorously conducted scientific trials. The purpose of this paper is to provide a systematic review of relevant research findings from 1989 to 1998 inclusive concerning non-pharmacological approaches to behavioural disturbances in people with dementia. The paper is intended to highlight areas that warrant further attention from clinicians and academics and to provide guidelines for future research.
The Cochrane Collaboration has set the gold standard for systematic reviews of the efficacy of medical treatments [9]. The Collaboration, which was established to prepare, maintain and disseminate such reviews, has developed an exhaustive methodology to reduce bias in the selection of scientific reports and in the methods used to summarise their findings. We have adopted the Cochrane approach with the proviso that few of the studies mentioned here were randomised controlled trials. However, clear criteria were used to choose, evaluate and describe appropriate publications. Reality orientation and reminiscence therapy are not covered as the Cochrane Collaboration has recently published material concerning them in its Database of Systematic Reviews [10,11].
Method
Data collection strategies included: (i) manual searches of selected journals, namely Age and Ageing, Alzheimer's Disease and Associated Disorders, American Journal of Alzheimer's Care and Related Disorders, Australian and New Zealand Journal of Psychiatry, Australian Journal on Ageing, British Journal of Psychiatry, Dementia, Geriatrics, Geriatric Nursing, Gerontological Nursing, International Journal of Geriatric Psychiatry, International Psychogeriatrics, Journal of Gerontology, Journal of the American Geriatrics Society and Gerontologist; (ii) searches of references listed in previous reviews; and (iii) searches of Medline, CINAHL, Psyclnfo and Cochrane computer databases for material published in English in the period 1989–1998 inclusive. Computer searches used the key words ‘dementia’ or ‘cognitive impairment’ and ‘behaviour’ (a list of specific search commands is available on request from the authors).
Criteria for the inclusion of original reports in this review were that: (i) dementia had been diagnosed using specified, acceptable criteria; (ii) subjects had one or more overtly difficult behaviours, for example aggression or verbal disruption; (iii) there were comparisons of the frequency or severity of behaviour before and after the introduction of an intervention; (iv) the intervention was not exclusively pharmacological in nature; and (v) the study design was rated as being strong, moderate or weak but not poor (see following for details). Studies that focused on incontinence, social interaction or mood were excluded, as were studies of reality orientation and reminiscence therapy.
Articles that appeared to meet these criteria were selected for further consideration. Decisions were reviewed by all of the authors and disagreements were settled by consensus. Articles that met our criteria were appraised in the following domains: design, sampling technique, setting, behaviours) studied, measurement tools, sources of information, type of intervention, feasibility and results. Feasibility was rated on a scale of easy, moderate and difficult. An intervention was rated as difficult if it required implementation over a long period of time, was expensive, involved extensive staff training or required specific expertise in design and implementation. By contrast, an easy intervention was simple to implement, involved low costs and used readily available materials and skills.
The strength of study design was rated using a tool devised by Forbes [12]. Studies were assessed in six areas: design and allocation to intervention, inclusion criteria, attrition levels, control of confounders, data collection methods and statistical analysis. Papers were rated as pass, borderline or fail in each area. Global ratings of validity were then assigned as follows. ‘Strong’ validity was assigned to papers with at least four passes and no fails; ‘moderate’ validity to those with fewer than four passes but no fails, and ‘weak’ validity for one to two fails. More than two fails led to a rating of ‘poor’ validity. As noted already, papers rated as poor were not considered further. Randomised controlled trials were not automatically rated as strong unless they scored highly in the areas of inclusion criteria, attrition levels, data collection methods and statistical analysis. Since validity levels were based solely on published material, studies may have been under-rated if reports failed to include all required details.
Results
Forty-three papers met the criteria for inclusion listed above (Table 1). The majority of studies were conducted in the USA (30), United Kingdom (five) and Australia (four). Studies from Canada and the Netherlands accounted for the remainder. Twenty-eight were conducted in residential facilities (labelled variously as nursing homes, dementia-specific units, long-term care units, skilled nursing facilities or geriatric centres), seven in psychogeriatric wards of hospitals or psychiatric institutions, seven in the community and one in a mixed community and residential setting.
Summary of review data including author, number of subjects, target behaviour, intervention, feasibility, validity rating, area of weakness, effect on behaviour and statistical test results
Validity ratings of articles were as follows: one strong, 15 moderate and 27 weak. There were five randomised controlled trials. In most cases subjects acted as their own controls giving a moderate rather than strong level of control over confounders. High attrition levels are sometimes inevitable in work with frail, elderly people as demonstrated by the five reports with attrition rates of 20% or higher. Full details of subjects' age, sex and other basic details were lacking in 10 papers. There were gaps in the description of data collection tools and associated reliability and validity in 18 and in the provision of detailed statistical test results in 12.
The articles have been grouped into the following intervention types for further discussion (some studies included more than one type): changes to the physical environment [13,19]; activity programs [20,26]; exposure to music, voice and language [20,27,35]; behaviour therapy [6,36,39]; massage and aromatherapy [40,42]; light therapy [43,46], multidisciplinary teams [47,48] and carer education [25,49,54]. Most attention will be given to papers rated as having moderate or strong validity.
Physical environment
If elements of the physical environment exacerbate the confusion and disorientation experienced by many people with dementia, then adapting, simplifying or enriching their physical surroundings might prove helpful. Such changes have initial set-up costs but need not impinge further on staff time. While five of the studies in this category showed beneficial effects, only three were rated as having moderate validity (Table 1).
Chafetz [13] questioned whether the visual agnosia that often accompanies dementia could be channeled to useful effect. Strips of contrasting coloured tape were laid on the floor near doors to create a visual illusion and thus reduce exiting but without success. Namazi etal. [17] conducted a more complex study to assess the effects of seven types of visual barrier on exiting behaviour. Coloured strips of tape placed on the floor actually increased the mean number of exits from 2.1 to 2.4 per day. By contrast, exiting was eliminated completely when the doorknob was hidden by means of a cloth stretched from one side of the door to the other. The colour and pattern of the cloth seemed not to matter. Painting the doorknob the same colour as the door reduced exits to a mean of 0.6 per day and attaching a secure cover to the knob reduced them further to 0.4 per day. These results were not subjected to statistical analysis. The authors noted that, since conditions were applied in series, learning might have played some part.
In a similar vein, seven residents of a dementia unit triggered the security alarm of an exit door on 115 occasions per week. Contacts fell by a statistically significant margin to 64 per week when the door was obscured using a blind of the same colour and to five per week using an identically coloured cotton sheet [15].
Mayer and Darby [16] found that, when a full-length mirror was placed in front of the door of a psychogeriatric ward, approaches to the door resulted in physical contact with it in 36% of instances compared with a baseline level of 76%. This statistically significant change could not be attributed solely to physical obstruction of the doorway since contact rates were higher at 51% when the mirror was reversed. Five of the nine patients involved in the study watched their reflections for lengthy periods, one added the mirror to her collection of objects to dust and another seemed disturbed by his reflection.
One moderately robust study assessed the impact of two simulated environments on levels of agitation, pacing, trespassing and exiting in nursing home residents [14]. Corridors were chosen as the best setting because prior research had shown that wanderers spent much of their time there. A nature scene was created using murals, posters, artificial plants, tape-recorded birdsong, forest smells and benches. A ‘home and people’ scene included family style photographs, an armchair, coffee table, traditional music and a citrus aroma. Residents' behaviour was measured using direct observation, photoelectric counters and personal activity monitors. The scenarios were inexpensive to set up and led to residents spending significantly longer periods of time in the enhanced environments than previously. There were also non-significant reductions of trespassing into other persons' rooms and attempts to exit the premises. Residents were less agitated (but not significantly so) and appeared to enjoy the enriched settings. All staff members and 78% of relatives were satisfied with the changes.
Whall et al. [19] hypothesised that the agitation and aggression associated with bathing would be reduced in a soothing, pleasant environment that tapped memories of nature. Residents in three control nursing homes were bathed in the usual manner. Those in two ‘treatment’ homes were bathed in rooms featuring tape recordings of birdsong, flowing water and small animals. Large bright pictures were shown in time with the sounds; residents were offered pudding and soft drinks, and nursing aides were trained to focus residents' attention on these various stimuli. There was a significant decline in scores for both agitation and aggression in residents bathed in the enriched setting.
To reduce the aggression that arose when residents of a Veteran's Administration dementia ward were transferred by elevator to a dining room, Negley and Manley [18] provided meals in day rooms to make the transfer unnecessary. As a result, assaults by resident on resident fell from 40 to 21 per month and assaults by resident on staff fell from seven to six. No statistical analysis was reported. Anecdotally, staff felt less anxious, spent more time assisting patients to eat and believed that patients were also less anxious.
Activity programs
Activity programs are offered in most residential facilities in the hope that diversion, stimulation and physical effort will reduce the anxiety and agitation that underpin many disturbed behaviours. While six of the seven studies in this category showed beneficial effects, only three were rated as having moderate validity. Two were randomised controlled trials (Table 1).
Cohen-Mansfield and Werner [20] argued that, if verbal disruption stems from sensory and social isolation, it might be reduced using audiotapes of residents' preferred music, videotapes prepared by family members, conversation, sensory stimulation and simple exercises. Shouting was reduced by 34% through use of music, 50% by videotape and 66% by social interaction. Similarly, repeated requests for attention were reduced by 33% through use of music, 46% by videotape and 94% by social interaction. These changes were statistically significant.
Another study tested the hypothesis that wandering is the result of inactivity, boredom, emotional distress and an inability to communicate by devising a walking program to counter these factors [21]. Trained volunteers walked outside with 11 markedly confused nursing home residents for 90 min on weekday evenings. Talking, singing and holding hands were all encouraged. The mean number of reported aggressive incidents fell to a statistically significant degree from 1.36 per day following non-activity days to 0.84 following activity days.
Confirmation of the value of physical activity came from a study that offered daily 40-minute programs of gentle stretches and exercises to 11 confused elderly residents [23]. Episodes of agitation fell significantly from a total of 495 in the month prior to implementation to 207 in the month of the study. By contrast, episodes increased from 526 to 594 for controls who were offered sessions of reading and poetry.
Evidence in support of activity programs comes from three other studies [22,25,26]. Not all of the programs considered here proved successful though. When sensory integration was taught three times weekly to residents of long-term geriatric facilities, non-significant improvements in levels of agitation were noted in both the experimental and control groups [24]. The authors questioned whether this apparent lack of efficacy might actually have been due to the small sample size, infrequency of sessions and use of insensitive measures.
Music, voice and language
The calming effects of music and the spoken word have been harnessed as treatments for distress and agitation. All 10 studies on this topic were conducted in residential facilities and reported positive effects. Three were rated as having moderate validity (Table 1).
Levels of aggression were reduced to a statistically significant degree when audiotapes of residents' preferred music (as reported by family members) were played during bath times [29]. In this trial, in which 18 participants were randomly assigned to music or no music conditions for a 2-week period and then switched, the mean number of aggressive behaviours per person totalled 122 without music and 66 with it. There were also reductions in particular behaviours, namely yelling, abusive language, verbal resistance and physical resistance which failed to reach statistical significance.
In a complex study of exposure to gentle, monotonous sounds, Burgio et al. [27] tested the responses of 13 nursing home residents to proprietary audio-tapes of either a babbling brook or ocean waves. Nine of the 13 subjects who showed less verbal agitation when exposed to sound proceeded to a second stage in which residents' preferred tapes were played at intervals during afternoons and evenings. The proportion of monitored time occupied by noise-making fell from approximately 50% when tapes were off to 35% when tapes were on, a statistically significant difference. However, nurse aides helped residents to access tapes on only 51% of observed occasions.
In an intriguing study, family members were asked to prepare brief audiotapes recounting family anecdotes, favourite poems or prayers and other cherished memories [35]. Tapes were then played by nursing staff twice a day, once in the morning and once in the afternoon. Over a 2-month period, problem behaviours reduced significantly in frequency while tapes were played in 91% of observed instances. Nursing staff reported that residents often smiled, laughed and sang while tapes were running. Residents accepted the tapes repeatedly over extended periods. No statistical analyses were listed in the paper.
The remaining studies were rated as having weak validity [20,28,30,32]. In two, verbally disruptive behaviour was significantly reduced using tapes of classical music or residents' preferred music [20,28]. When Denney [30] played background music during meal times, levels of agitation dropped by 56% after 1 week and increased again when the music was stopped. Two studies by Gerdner [32,33] recorded a lag effect with a decrease in agitation after, but not during, the intervention period.
Validation therapy, which entails much attention to verbal and non-verbal language, was examined in only one of the studies uncovered by our search [31]. When care staff were trained over a 3-week period in the use of listening, distraction, eye contact, emotional expression and other validating techniques, there was a reduction of reported incidents attributable to difficult behaviours from 3% of the total in the month prior to training to zero per cent later. The report made no reference to statistical analyses.
In a study of communication style per se, Hart and Wells [34] sought to determine the contribution to agitation and other disturbed behaviours of the failure by patients to understand caregivers' language. All of the 15 persons concerned could understand a simple one-phase command, for example, ‘Touch your nose’ but none could follow a three-stage command such as, ‘Point to my face, raise your arms and then clap your hands’. When they were read a series of 60 three-stage commands while being videotaped, they showed an average of 29 manifestations of agitation. By contrast, when read a series of 60 one-stage commands, they showed an average of only three such manifestations. This difference was highly statistically significant. The authors suggested that assessing patients' level of comprehension and purposely formulating sentences to match this level would reduce levels of agitation.
Behaviour therapy
Behaviour therapy has not been trialled extensively in dementia because of doubts that new learning is possible. However, the studies considered here, all of which were rated as having weak validity and described work with very small numbers of patients, suggest that learning can occur even in those with advanced degrees of confusion (Table 1).
Doyle et al. [6] found that contingent reinforcement (rewarding quiet behaviour with individually tailored rewards such as favourite foods), ignoring noisy outbursts and applying carefully selected stimuli such as music, conversation, touch or visual aids successfully reduced noisiness in three out of seven residents with severe dementia. A meta-analysis of the results revealed an overall statistically significant effect.
Bird et al. [38] provided detailed accounts of the use of cued recall in five severely confused persons. As an example, a woman who repeatedly entered other residents' bedrooms and took their possessions was taught in a single 2-hour session to associate a cue, a large red stop sign, with stopping and walking away. The sign was placed at eye height on doorframes. Intervals between trials were gradually lengthened to reinforce new learning and staff were asked to repeat cueing each time the behaviour recurred. Further staff involvement was rarely necessary as inappropriate daily entries dropped from a mean of 44 to two. These measures were not subjected to statistical analysis.
Massage and aroma
Touch and smell are soothing, hence the possible usefulness of massage and aromatherapy in relieving distress and agitation. Two of the three studies on this topic reported positive effects and two were rated as having moderate validity (Table 1).
When four long-stay residents of a psychogeriatric ward were offered a series of 30-min treatments with hand massage or vaporised lavender oil, only one improved and two showed consistently higher levels of disturbance [40]. Snyder et al. [41] reported that hand massage twice a day prior to nursing care led to a statistically significant reduction in the frequency and intensity of some but not all behaviours during morning but not afternoon care. There were also interaction effects between treatment and gender: agitation decreased significantly for women but not for men. In another study, the same authors offered 10-min treatments of hand massage and therapeutic touch in which a hand was placed on each subject's back to ‘transmit calming energy’ [42]. Both treatments produced a statistically significant reduction in subjects' pulse rates and ratings of relaxation of the forehead, eyes and seven other body areas but not in ratings of agitated behaviour.
Light therapy
The possibility that evening agitation (‘sundowning’) and nocturnal disorientation might be reversed by daytime exposure to bright light has been examined in four studies, two of moderate validity. Three of the four reported beneficial effects (Table 1).
Lovell et al. [44] placed six nursing home residents in front of a 2500 Ix light box containing white fluorescent bulbs for 2 h each morning. Treatment took place wherever subjects felt most comfortable, usually in a chair near their bed. The light box, which measured 1 × 2 feet, was placed approximately 1 m away from subjects at a height within their visual fields. Subjects were encouraged to watch television or participate in other normal activities during the treatment. A member of staff remained present throughout. Ratings of agitation were made once every 15 min between 16.00 h and 20.00 h during 3 days of baseline and 5 days post treatment. Mean scores on the scale fell from 20 to 10.5 out of a maximum of 50, a statistically significant change.
Bright lights were applied in the evening in a study of 10 dementia unit residents known to become restless, agitated or noisy later in the day [46]. When residents were exposed to fluorescent bulbs emitting 1500–2000 Ix between 19.00 h and 21.00 h each evening for a week, mean ratings of nocturnal restlessness were halved and the proportion of total daily activity occurring between 23.00 h and 07.00 h fell from 18% at baseline to 12% during treatment. This proportion rose again to 17% when treatment was stopped. There were no statistically significant changes in the use of restraints and as-needed medications.
In a report that failed to demonstrate any effect, seven community-resident subjects were exposed to 2 h of 2000 Ix light each morning [43]. Light sources were worn on the head, 3–4 cm from the eye, to allow participants to move freely. This might have reduced efficacy but, equally, outcomes in this study were measured with uncommon precision using a mechanical monitor and were possibly less subject to bias. The investigators, who were principally interested in shifts in circadian rhythm, suggested that treatment be given when core body temperature is at its lowest, approximately 2 h prior to waking, and that treatment periods be extended from 2 to 4 h. It is unlikely, however, that these suggestions could be implemented easily in daily life.
Multi-disciplinary teams
It is now customary in aged mental health services for staff to work in multidisciplinary teams. Patients benefit from the contributions of doctors, nurses and allied health professionals but few studies have attempted to measure the changes in patients' behaviour that follow from teams' involvement. Only two such studies were uncovered in our search, both by the same researchers. Validity ratings were low for one and moderate for the other (Table 1). The earlier study utilised psychotropic medications and psychological interventions but had high attrition rates resulting in only two out of four subjects completing the exercise [47].
The later study used a stronger design and a larger clinical team comprised of a psychiatrist, clinical psychologist, nurse, social worker and occupational therapist [48]. Members of Group One received assistance immediately. Members of Group Two, who acted as controls, received assistance 16 weeks later. Interventions included a diverse range of individually tailored treatments including psychotropic medications, activities, behavioural strategies and carer education. Fifteen of the 20 members of Group One showed a significant improvement in behaviour compared with only two of 13 in Group Two. The authors speculated that delays in providing assistance might result in problems becoming intractable.
Carer education
Disturbed behaviours might also be remedied by means of educational programs directed at family or professional carers. Of the seven papers on this topic, all except one appeared efficacious but only two had strong or moderate validity ratings (Table 1).
Bourgeois et al. [49] reported that training seven family carers to use individualised written cues to prompt memory significantly decreased repetitive verbalisations. Following a 3-h workshop on behaviour management principles, a therapist worked with caregivers in 11 1-h weekly home visits to construct a series of individually tailored cards to remind patients of frequently forgotten material. As an example, the carer of a man who repeatedly asked to go out was encouraged to say, ‘Today we are going to (location) at (time). Here is a card which tells you this so that you can remember when and where we are going. If you forget, look at this card’. If he repeated himself, the carer said, ‘Here is your card that tells you when we are going out’ and walked away. Use of the card was encouraged and praised. The mean daily frequency of repetitive statements or questions fell from 21.9 at baseline to 11.2 12 weeks later and then to 8.6 at 6-month follow-up, a statistically significant decline. Carers unanimously rated the home visits as very helpful.
Education appears to work best, at least with respect to its impact on behaviour, when the proposed intervention is highly specific [50]. In a large controlled trial of 54 family carers, the provision of verbal and videotaped material about dementia, the reasons for behavioural problems and lists of possible interventions led to a slight but statistically insignificant reduction in subjects' scores on a behaviour disturbance scale (compared with a slight increase for the control group). Adding written material outlining possible behavioural strategies and modelling such responses with the caregiver had no additional benefit. Despite this, approximately half of the suggested approaches were rated as frequently or always effective by 75% of the carers.
Training might be of special benefit to professional careworkers. When three 30-min in-service programs covering dementia, the reasons for aggression, care principles and role-plays were offered to employees of a large extended care hospital, incidents of resident aggression dropped to a statistically significant degree from 182 in one week at the beginning of the study to 93 in the week after all willing staff (87%) had attended the final session [51]. It was notable that teaching was offered to all staff members, including housekeeping and recreational staff, some of whom knew little or nothing at commencement about dementia and its consequences.
In a more intensive study, staff in the dementia wing of a large nursing home were given workshops and literature on dementia care and encouraged to adopt a more client-orientated approach in which residents' wishes, whether expressed verbally or non-verbally, were respected [53]. Residents were given the freedom to rise from bed, take their meals and participate in activities as they wished. Meals were still served at scheduled times but set aside for those who were absent. In the case of severely confused residents, nurses helped them out of bed if they were awake and cooperative but, if they showed resistance, staff allowed them to re-settle and returned later.
Findings from this study were inconsistent. There were no changes in measures of aggression and non-physical agitation; daytime sleep increased, and levels of verbal disruption were decreased but only in the morning. The authors noted that staff on the afternoon shift were less flexible in their approach, perhaps because of a lower staff-resident ratio and the greater use of agency personnel at these times. In addition, afternoon staff complained that, because residents were allowed to sleep longer in the mornings, some activities of daily living previously conducted on the early shift were deferred to the late shift, thus resulting in increased staff-resident interaction and a consequent rise in levels of agitation.
Conclusions
The majority of the studies covered by this review reported improvements in disruptive behaviours as a result of practical, clinical and educative interventions. However, even when treatments appeared ineffective, it would be premature to dismiss them as lacking value since replication using more rigorous methods with larger numbers of subjects and in a wider range of settings might produce different outcomes.
Studies which altered the physical environment by masking exits and making spaces more stimulating and congenial were relatively inexpensive to implement and required little staff time. Most were directed at specific behaviours (e.g. exiting and aggression) and had positive results. Concealing exits is one of the few non-pharmacological strategies to have been replicated to date [15,17]. Importantly, none of the strategies employed in the studies described here seemed likely to compromise the safety of residents and staff.
Activity programs also proved successful but the interventions varied greatly in style and content. It is unclear, therefore, how much of the reported benefits derived from a program's special content and how much from the stimulation, diversion and social engagement implicit in any structured pastime. Studies in which one activity is compared directly with another will, therefore, be of special interest. While activity programs require time and some expertise to set up and run, an evening walking program run by volunteers was practical, effective and readily implemented [21]. Such positive outcomes reinforce the notion that people with dementia benefit from structured and well-planned activities of an appropriate kind.
Music, which is offered in most residential communities, has been studied relatively intensively. Personalising music to reflect individuals' tastes and cultural backgrounds appears promising but, given the demands this entails for staff, it will be of value to compare generic with individually selected music, live with recorded music, music heard alone with music heard in company and so on. Audiotapes of relatives recounting family stories and reciting familiar poems and prayers helped to maintain a link between residents and family members and allow families to contribute actively to the relief of distress and behavioural disturbance [35]. Since relatives will usually require guidance to select suitable material and produce tapes of acceptable quality, it should be demonstrated that this extra effort produces an added return.
Exposure to bright light appears to decrease daytime sleep, decrease ‘sundowning’ and increase night-time sleep, at least in people with moderate to severe dementia. Some subjects complained of excessive brightness [44] and use of this modality is limited by the need to supervise residents closely or else restrain them in chairs for lengthy periods. It has yet to be determined what minimum dose and duration of treatment is needed to produce a clinically useful effect. Light therapy shows promise, however, as nocturnal disturbance is a major source of stress to carers and co-residents.
Education and training of family and professional caregivers require specific expertise and staff time to implement. While the results were encouraging, with all but one study reporting decreases in target behaviours, these programs can be expensive to run and their long-term effects on care practices have yet to be demonstrated.
With respect to research design, most studies to date have been of limited rigour. Of the 43 considered here, only one was rated as having strong validity. Researchers should seek to improve the strength of their study design, instrumentation and analysis if clinical care is to be guided by scientific investigation. Admittedly, some aspects of study design are hard to control when working with people at risk of relocation, illness and death but published reports should include detailed descriptions of the target group, data collection methods and statistical analyses.
Studies in this area would be strengthened greatly by more widespread use of rigorous trials conducted in more than one facility with multivariate analyses of multiple outcome measures (e.g. direct observations, behaviour rating scales and carer questionnaires). It may not be practicable to blind observers to treatment allocation since interventions are often obvious and must be delivered in subjects' usual place of residence. However, data collection bias can be reduced by means of computer-assisted real time observations [27]. This technology and coding system requires lengthy training to master but videotapes, audiotapes and mechanical activity monitors can also provide useful objective data. Whatever methods are adopted, it is important that staff receive adequate training and that the final report provide sufficient details to permit replication.
Few of the studies reviewed here measured the efficacy of interventions simultaneously from the viewpoints of persons with dementia, their carers and other stakeholders. Most focused solely on the frequency and severity of behaviour with little acknowledgement that interventions might have untoward effects for one or more parties. It cannot be assumed that non-pharmacological strategies are innocuous or that a good outcome for professional care staff invariably equates to a good outcome for people with dementia or their relatives. Future studies should also address the broader aspects of quality of life for people with dementia and their carers and consider all potentially competing interests. As a minimum, reports should list the number and magnitude of adverse events attributable to an intervention to ensure that subjects' and carers' interests are preserved.
There are many factors to consider when implementing strategies for behavioural difficulties in dementia, such as the person's individual needs, the multifaceted nature of his or her problem and the available physical and social resources. Future research should build on some of the successful studies to date, improving methodologies where necessary, extending outcome measures to include all stakeholders and adding to the available bank of evidence. There are still many kinds of interventions that require rigorous evaluation to help guide the choice of aged care workers. Finally, it will be of value to make direct comparisons of the efficacy of pharmacological and non-pharmacological strategies.