What Do We Know About Behavioral Crises in Dementia? A Systematic Review

Review questions

What are the factors associated with behavioral crises in dementia?

Protocol and registration

The protocol for this review is registered on PROSPERO (2016:CRD42016051002) [24].

Eligibility criteria and Information sources

We searched MEDLINE, CINAHL, PsycINFO, EMBASE, and AMED databases for articles published up to 20 June 2016. The search strategy covered three areas: behaviors, dementia, and crises (including institutionalization, referrals, and hospitalization) (full electronic search strategy used for MEDLINE in Supplementary Table 1). The search was piloted and then refined to increase the likelihood of obtaining relevant results. To maximize the findings, there were no limitations on the searches in relation to years considered, publication status, or study design. English and Spanish languages were considered. Searches were limited to humans. We also conducted a supplementary lateral search of key websites such as the Alzheimer’s Society, National Institute for Health and Care Excellence (NICE) Evidence Search, and Social Care Institute for Excellence (SCIE) Social Care Online websites, reference lists of included articles, and other databases including ALOIS, LILACS, and the Cochrane Library. Where necessary we contacted study authors to locate full texts.

Study selection process

The study selection occurred in two phases. First, two authors (TB and JC) individually screened all titles for potential eligibility. Titles were excluded if they were clearly not related to dementia or behaviors or if the article focused on children. All articles marked ineligible by both authors were excluded. Second, the abstracts were screened, and relevant full texts were obtained where articles were clearly relevant or to determine eligibility. Articles were read and assessed in regard to three questions. Does the research include PwD? Does the article include information on dementia-related behaviors? Is there a clear behavioral crisis (using the MacNeil Vroomen et al. definition [11])? Judgements were made about how the studies met the inclusion criteria, disagreements were discussed and resolved with a third person where necessary (TB, JC, EM).

Data collection processes and data items

A bespoke data extraction form was created, piloted using three articles to see if the data extracted was useful and covered all relevant and emergent aspects of crises. After review, the form was amended to add extra extraction categories such as ‘stages of dementia’ and ‘length of stay’ before completing extraction for all included articles. The extraction form included information on the authors; year; place of study; aims of study; methods used; participant group, number, and demographics; dementia diagnosis; crisis information including lead up to and resolution; behaviors and other factors assessed at the time of crisis; living situation pre- and post- crisis; and referral information. Data was extracted by TB in consultation with both JC and EM.

Risk of bias in individual studies

The Quality Assessment Tool for Quantitative Studies (EPHPP) [25] was used to assess risk of bias at study level. Two authors (TB and JC) reviewed articles independently. We controlled for assessment items that were not applicable to some study designs by calculating a percentage index of the ratings against the amount of applicable items for each study. Disagreements were resolved by discussion and then consensus to create an overall rating for each study. However, this information was only used to judge the quality of the studies included in the review and not to exclude studies since we wanted to maximize potential learning.

Synthesis of results

The ‘Guidance on the Conduct of Narrative Synthesis in Systematic Reviews’ [26] was used to frame a narrative descriptive synthesis of the data. Data extracted from the studies were transformed into a common rubric, vote counted where necessary, and tabulated by topic before being analyzed across the studies, also taking into account information not tabulated.

Classification and synthesis of behaviors leading to a dementia crises

Synthesizing the findings showing the behaviors causing the crises was challenging, due to many factors. We found that behaviors were not measured in the same way across studies. When the same measure had been used across studies, different versions, numbers of items, and scales were used [27–29] or results were not always reported [30, 31]. Studies also used different groupings and terminology for behaviors. Additionally, some patients had multiple behaviors at the time of crisis, which made it impossible to determine if one main behavior was the cause of the crisis or whether it was the combination of the behaviors. Therefore, we vote counted and charted all of the behaviors found at the time of crises from each article. The category agitation/aggression was combined due to multiple articles joining the two behaviors together. For example, both the Neuropsychiatric Inventory (NPI) and the Cohen Mansfield Agitation Inventory (CMAI) blend these categories. Therefore, the category agitation/aggression includes behaviors termed multiple ways such as assualtiveness, anger, violence, verbal aggression, restless agitation, and physical aggression. After discussion and consensus between the three authors, studies citing psychosis or psychotic symptoms were coded as both hallucination and delusion. Additionally, where two behaviors were joined as a category such as in the Behavioral Pathology in Alzheimer’s Disease rating scale (BEHAVE-AD) (apathy or depression), these were coded in both categories. One article [32] stated “behavioral problems” only and did not provide specific behaviors; however this article was linked to two other articles from the same study which did specify behaviors [31, 33].

Risk of bias across studies

Risk of bias across studies (publication bias, selective reporting) was assessed via a discussion of the included studies between TB and JC.

Study selection

The database search identified 5,544 records and a lateral search recognized two further records. After duplicates were removed, 3,737 were left and these were assessed for eligibility. 234 were retrieved for a full-text assessment. Ineligible articles were removed and 24 articles were included in the review (see Fig. 1). These articles described 18 distinct studies. Of the studies with more than one article relating to it, two articles per study were available for two studies and three articles per study available for two studies. The separate articles relating to one study reported: different or similar analyses for the same research methods and sites with overlapping time-points [31–33; 34–36; and 37, 38], and findings from the same patients [39, 40] therefore tabulated results are presented for both articles and studies.

OPEN IN VIEWER

Fig.1

Flow diagram of search results, screening, and eligibility.

Study characteristics

Table 1 presents the characteristics of the 18 studies. Studies covered a wide time-span (1990–2016) and multiple countries from Europe, North America, South East Asia, and Australia. Studies aimed to explore the reasons for institutionalization, or admittance to a dementia or psychiatric unit, or to assess an intervention. There was a mix of study designs including qualitative interviews, observational studies, chart reviews, and naturalistic and randomized trials. Studies took place in the community, long-term care facilities, and specialist dementia or psychiatric units. Participants were predominantly patients with dementia, but two studies consulted informal care informants [41, 42].

Risk of bias within studies

Differences in reviewer quality ratings were present for 7/24 articles. Two of the inconsistencies were due to oversight and five to differences in the interpretation of the study. Differences were discussed between TB and JC and consensus was reached without the need for a third person. Table 1 shows the quality ratings; 5 studies were rated as strong, 11 as moderate, and 2 as weak.

Synthesis of results

Which behaviors were present at crises?

As shown in Fig. 2, agitation/aggression was the most cited cause of crisis with all 18 (100%) studies describing these behaviors, all of which were among the three most prevalent behaviors found to be the cause of crises in each study. Delusions were cited by 13/18 (72%) of studies, wandering/absconding 12/18 (67%), and hallucinations 11/18 (61%). Although this review focusses on behavioral crises, some studies also included or mentioned patients with dementia reaching crisis points for other reasons, such as cognitive symptoms [29, 42–44] or caregiver factors [28, 38]. Differences in behaviors leading to crisis in different countries were highlighted in one study [42], with agitation/aggression being most common in France, wandering in England, and sleep behaviors in Sweden, Germany, and Holland. Marengoni, Calabrese, and Cossi [45] found somatic diseases, such as urinary tract infections, adverse drug effects, and pneumonia, underlying the behaviors in 93% of their patients.

OPEN IN VIEWER

Fig.2

Reported behaviors at the time of crisis by number and percentage of studies (n/18, percentage).

Summary of evidence

In this systematic review, we identified 18 studies (24 articles) that described behavioral crises in dementia. Agitation and aggression were reported as key behaviors at crisis point in all studies. Delusions, wandering/absconding, and hallucinations were also major behaviors contributing to crises. Behavioral crises predominantly happened in the severe stages of dementia (according to MMSE scores) and often led to admissions to psychiatric care or specialist dementia units, or to use of community services. People experiencing crises lived both in the community and in care homes. There was some evidence that crises are not being resolved long-term, since seven studies reported the same patients being re-admitted within the study timescale.

A recent online survey asked stakeholders in the United Kingdom to rank multiple risk factor categories for crises in dementia. Wandering, physical aggression, and sleep disturbances were perceived as the three main behavioral risk factors causing crises in dementia [51], while focus groups with carers found that wandering and physical aggression could be a cause of concern [52]. Our findings support these, but also highlight the role of delusions and hallucinations as contributors to behavioral crises in dementia, for which we cannot rule out the possibility of delirium.

This review has revealed there is little good quality evidence showing how best to manage behavioral crises in dementia. Interventions employed to resolve crises were often not clearly specified. Where they were specified, they were individualized to each patient making the actual intervention delivered difficult for authors to report and for this review to determine. Similarly, behaviors in the studies included here were not assessed or represented in a uniform way, e.g., behavior descriptions were assessed formally by clinicians, taken from patients’ notes, or reported by informal carers or care-home staff. Therefore, differences in the perception and categorization of behaviors is likely to have been present. This was reflected in the varied terminology used when reporting behaviors, which made the synthesis of them problematic. Additionally, from the information reported we could not determine whether behavioral crises occurred due to facility and/or caregiver variables or PwD variables. We recommend that studies use a validated behavior tool to assess and describe behaviors and that caregiver factors are also considered, with both reported clearly.

Related to the above, this review also uncovered the lack of consistency in management of agitation/aggression in dementia-related behavioral crises. Even though all studies identified aggression/agitation as a key BPSD leading to the crisis, a wide range of pharmacological and non-pharmacological approaches were used to manage behaviors, with half of the studies using a combination of the two. Perhaps clear and consistent international guidelines incorporating a multicomponent approach would not only help in the management of such challenging crises, but also inform future systematic investigations of the efficacy of such clinical approaches.

It is of note that MacNeil Vroomen et al. [11], when reviewing the definition of crisis in dementia care found that general practitioner (GP) management, increased information to GPs and informal caregivers about dementia, case management, and careful management of drug therapy were recommendations when BPSD were predictors or stressors of crises. While our findings included interventions similar to the latter two, they did not include the two former methods. This may have been due to our review only including actual incidents of crises which may have been too late for GP management or information to have an effect, whereas the MacNeil Vroomen et al. review included any emergency themes related to dementia care.

In 2007 in England, only one in six National Health Service Trusts provided crisis support to older people and only a tenth of areas provided crisis resolution team support to PwD [53]. Since then, dementia crisis teams have developed in some areas to support PwD and their carers with the aim to improve care and reduce admissions. Their effectiveness is currently being evaluated by the Achieving Quality and Effectiveness in Dementia care Using Crisis Teams (AQUEDUCT) study [54]. Findings from this ongoing study may complement this review by providing evidence relating to the role of dementia crisis teams in preventing and managing crises.

International differences were present amongst the studies included in the review. Differences were found in the measures used to assess behaviors across countries (i.e., CMAI and BPRS used in the USA, BEHAVE-AD in Japan, and the NPI in Italy, France, Canada, and the USA). The clear differences in practice between countries were not reflected in crisis interventions. Here the lack of consistency pervaded across all countries and regions with no patterns detected. One exception was Specialist Dementia and Geriatric Units, which were only found in Europe. An international approach to using the same standardized measures to assess diagnoses and behaviors would assist future comparisons and synthesis of results. Additionally, research clearly detailing the crisis interventions and care approaches used in particular institutions would enable shared learning to develop practices worldwide.

It is clear that more research is needed to determine interventions and models of care that may lead to fewer crises. These models of care would not have the scope to prevent all behavioral crises, but would likely address a great proportion of them by working early and continuously with the family and/or care setting accordingly. Continuity of care is also likely to be more cost effective [55, 56], not only from a financial perspective but also in regards to the wellbeing of the PwD and their carers.

As the prevalence of dementia increases across the globe and more people experience behavioral crises, services will need to adapt to help prevent and manage BPSD more effectively. It is important that these behaviors are managed as successfully as possible to reduce distress for the person with dementia and pressures on those caring for them. The findings from this review have relevance for multiple disciplines including neurology, dementia, gerontology, psychiatry and for community and acute psychiatric services.

Strengths and limitations

The strength of this review is the strict inclusion criteria used to gain information reporting factors on actual behavioral crises. Additionally, we used standardized procedures for extracting data from included studies. However, we did include studies using many different methodologies, covering different settings and using various assessments and measures which made the data difficult to synthesize and compare. It is also possible that not all relevant records were retrieved from our search due to the term “crisis” not being explicitly used in all research. Although we systematically judged the quality of articles, we did not exclude on the basis of low quality due to the exploratory nature of this review and our aim to maximize the information we could gather. Therefore, although we highlighted the quality of the individual articles, some evidence presented was rated as weak. Our review did not aim to investigate means of preventing crises, and this is largely still not well understood, and would be useful to inform models of care and services to prevent and manage crises in future.

Risk of bias across studies

The majority of the studies included in this review published data on admissions to psychiatric care generated through chart review so may not represent the general dementia population. Some behavioral crises may occur in the community, but have not been identified in our study due to the lack of community-based studies examining and reporting this issue. Additionally, smaller actions taken to resolve crises may not appear in our results.

Conclusions

Dementia-related behavioral crises are often precipitated by aggression and/or agitation. They predominantly occur in the severe stages of dementia and in both community and care home settings. Interventions and services to prevent and manage dementia-related behavioral crises should consider targeting people in the moderate to severe stages of dementia and their caregivers, focusing on people living in family homes and in care homes, and concentrate on reasons behind aggressive and agitated behaviors. More good quality research is needed to determine:

clear factors that could be addressed to prevent behavioral crises;