A Scoping Review of Care Trajectories across Multiple Settings for Persons with Dementia

Abstract Multiple transitions across care settings can be disruptive for older adults with dementia and their care partners, and can lead to fragmented care with adverse outcomes. This scoping review was conducted to identify and classify care trajectories across multiple settings for people with dementia, and to understand the prevalence of multiple transitions and associated factors at the individual and organizational levels. Searches of three databases, limited to peer-reviewed studies published between 2007 and 2017, provided 33 articles for inclusion. We identified 26 distinct care trajectories. Common trajectories involved hospital readmission or discharge from hospital to long-term care. Factors associated with transitions were identified mainly at the level of demographic and medical characteristics. Findings suggest a need for investing in stronger community-based systems of care that may reduce transitions. Further research is recommended to address knowledge gaps about complex and longitudinal care trajectories and trajectories experienced by sub-populations of people living with dementia.

The COVID-19 pandemic highlighted several challenges faced by older adults living with dementia in the community. Dementia increases the risk of exposure to the COVID-19 virus and to negative outcomes such as behavioural changes and delirium (Brown, Kumar, Rajji, Pollock, & Mulsant, 2020), thereby increasing the risk of hospitalization in this group (Alzheimer's Disease International, 2020). Unnecessary hospitalizations of people with dementia were already a concern in pre-pandemic Canada, with a hospitalization rate of 33 per 100 versus 20 per 100 people without dementia (Canadian Institute for Health Information, 2018a). Fear of exposure to the virus and social isolation in hospital prompted families to reconsider the risks and benefits of acute care (Alzheimer's Disease International, 2020). The pandemic also severely affected long-term care (LTC) homes in Canada, where 69% of residents live with dementia (Canadian Institute for Health Information, 2018a). In the first 3 months of the pandemic, more than 9,000 LTC staff were infected by the virus and 8 in 10 COVID-19 deaths across the country took place in LTC and retirement homes (Canadian Institute for Health Information, 2020).
The risk of admission to hospital or LTC for older adults with dementia therefore has serious consequences for this population and the larger Canadian health care system. A care transition refers to a physical move across locations that involves at least one overnight stay (Aaltonen, Rissanen, Forma, Raitanen, & Jylha, 2012) or consecutive days of care in different health care settings including home (Wang et al., 2017). Multiple care transitions are characterized by repeated or multiple moves between two or more care sites, or dynamic movement across multiple sites of care (Callahan et al., 2012(Callahan et al., , 2015. Care transitions across home, acute, and LTC settings by people living with dementia often result from increasingly complex care needs as function and health progressively decline (Alberta Health, Continuing Care, 2017;Fortinsky & Downs, 2014). In a United States study of older adults, the average number of care transitions between home, hospital, and LTC was two times higher among people with dementia than in those without dementia (Callahan et al., 2015). Furthermore, Callahan et al. (2015) found that 37% of older adults without dementia experienced no transitions during the observation period, compared with only 4% of individuals with dementia.
A care trajectory is defined in this review as a pathway that consists of care transitions across multiple settings. Care trajectories merit attention as they may signal gaps in appropriate community-based care and support, contributing to negative health outcomes for people with dementia and their care partners. Episodes of transition expose individuals to possible medication error (Deeks, Cooper, Draper, Kurrle, & Gibson, 2016) and loss of critical information such as advance directives and care plans (Canadian Medical Protective Association, 2018). Care transitions cause disruptions in daily schedules and care continuity that can be particularly stressful and detrimental for those living with dementia, contributing to lower physical and psychological well-being (Ryman et al., 2019). Negative outcomes for care partners such as stress, depression, and anxiety have also been attributed to care transitions, as well as disrupted self-care caused by a significant shift in focus to the individual with dementia (Sadak, Zdon, Ishado, Zaslavsky, & Borson, 2017). Emotional concerns and unmet needs for information and support among care partners have been shown to emerge throughout the care transition period; for example, before and after admission to LTC (Afram et al., 2015;Ray et al., 2015).
Although there is agreement that persons with dementia experience numerous transitions, there is less agreement on the standard classification of these transitions (Fortinsky & Downs, 2014). The purpose of this scoping review was to identify and classify care trajectories across multiple settings for people with dementia, and to gain an understanding of the prevalence of multiple transitions and related factors at the individual (demographic and medical) and organizational levels. The current review is intended to increase our knowledge of care trajectories in this population and to inform future research efforts by identifying key gaps in the literature and opportunities for research.

Methods
The review was guided by the Arksey and O'Malley (2005) fivestage scoping review methodology and additional steps for each stage as proposed by Levac, Colquhoun, and O'Brien (2010). Further, a review team met biweekly throughout the review process to consider decisions regarding the search strategy, study selection, and data extraction and analysis, as recommended by Levac et al. (2010).

Identifying the Research Question
This review aimed to answer the following research questions: What specific care trajectories involving multiple transitions are experienced by people with dementia in terms of care settings, and number and patterns of transitions? What is the prevalence of multiple transitions in this population? What are the factors associated with multiple transitions at the individual (demographic and medical) and organizational levels? For the purpose of this review, multiple transitions were defined as (1) two or more moves between at least three different care settings (e.g., home-hospital-LTC), (2) three or more moves between at least two different care settings that include home (e.g., home-hospital-home-hospital), or (3) two or more moves between two care settings that exclude home (e.g., LTC-hospital-LTC). We anticipated that the findings of this review would identify gaps in current knowledge and point to areas for future research. populations, with people 65 years of age and older accounting for approximately 97% of cases (Canadian Institute for Health Information, 2018a). A 10-year period was chosen for the search, given the increased number of relevant studies in recent years, the pace at which recommendations are being addressed, and to balance the large number of records retrieved.

Study Selection
As shown in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram in Figure 1 (Moher, Liberati, Tetzlaff, Altman, & The PRISMA Group, 2009), the search resulted in 16,704 records. This number was reduced to 10,870 after duplicates were removed. The records were exported to a reference program (Endnote) and a web-based systematic review program (DistillerSR), and further de-duplication reduced the number to 10,668 records.
In the first stage of study selection, four reviewers independently conducted title and abstract relevance screening based on initial inclusion/exclusion criteria. The first author (J.G.K.) screened all titles and abstracts, a second reviewer assessed half, and two additional reviewers each evaluated one quarter of the records. Inter-rater agreement was not calculated, as study selection was an iterative process involving ongoing deliberation among reviewers about the operationalization of "multiple transitions" as defined in the inclusion criteria (Table 1). After excluding 10,496 records, the second stage involved reviewing the full text of 172 articles using a screening form developed by the first author, based on the final inclusion and exclusion criteria. The reviewers first tested the form with 19 articles, and revisions were made based on reviewer feedback. Three reviewers then each independently evaluated the full text of one third of the articles and the first author assessed all articles. Conflicts were resolved by discussion at both stages. The second author (D.G.M.) made the final decision on 5 records when consensus was not possible in the first stage, and consensus was reached between the first author and reviewers on all articles in the second stage. As shown in Figure 1, 33 studies were included in the current review after excluding 139 articles during the second screening stage.

Charting the Data
The first author extracted the data using a Microsoft Excel spreadsheet that was continually refined throughout the abstraction process to adequately capture key data. The following information was extracted from each study, where available: authors and year of publication; study country and setting, objectives, design, timeframe and study period, intervention, outcome measures, sample characteristics [age at baseline, sex, ethnicity, residence (e.g., rural, live at home or in longterm care), method of dementia identification such as clinical diagnosis, timing of dementia identification such as before first transition, proportion of patients with dementia and type of dementia, and proportion of patients with other medical conditions]; transitions as a primary focus (yes/no); and study conclusions, key implications and recommendations, and reported limitations. Data charting also included identifying and extracting care trajectories based on transition patterns that involved two moves, three moves, or four or more moves across settings, the prevalence of transitions, and factors associated with transitions at the individual (demographic and medical) and organizational levels. In addition, length of stay in hospital was documented where applicable. Some studies reported more than one trajectory, and data relevant to all routes were extracted where this was the case.

Collating, Summarizing, and Reporting the Results
The data extracted to Microsoft Excel were collated in a series of Microsoft Word tables (not included here) as an intermediate step, and study characteristics were organized in a numerical summary (Table 2). A framework of care trajectories was developed based on transition patterns extracted at the charting stage ( Figure 2). This framework classifies trajectories based on starting location and other settings involved, number of transitions, and the pattern of movement across locations. For the purpose of this review, trajectories that began in an unspecified location or in hospital were considered to originate in location 'x'. We rationalized that hospitalization constitutes a temporary stay for individuals who have other permanent or long-stay living arrangements. In transitions that involved multiple unspecified locations, 'x' was counted as a single location. The main findings in the narrative synthesis centre on the most commonly reported trajectories (as examined in three or more included studies), prevalence of transitions, and factors associated with transitions. Table 3 summarizes the relevant findings including all identified care trajectories, prevalence of transitions, associated factors, and key implications and recommendations.

Study Characteristics
Characteristics of the 33 included studies are provided in Table 2. The majority of articles were published since 2015, and most studies were conducted in North America or Europe. Four studies included an intervention (Boltz, Chippendale, Resnick, & Galvin, 2015;Cassel et al., 2016;Gustafsson et al., 2017;Noel, Kaluzynski, & Templeton, 2017). All of the studies included a mixed-sex sample, with females accounting for at least 60% of the dementia sample in three quarters of the studies.
Most studies assessed the effect of dementia on the use and outcomes of health services in general, of which transitions across care settings were only one component. Time frames considered in the studies ranged from less than 30 days to 5 years, with short time frames of less than 1 year examined most frequently. Few studies focused on time frames of 3 years or longer (Bucher, Dubuc, von Gunten, & Morin, 2016;Callahan et al., 2012;Chang et al., 2015;Rudolph et al., 2010;Seematter-Bagnoud, Martin, & Bula, 2012).
Individuals with dementia were identified on the basis of administrative data in the majority of studies. In these studies, diagnosis was recorded from the data either directly after the first transition had occurred (Daiello, Gardner, Epstein-Lubow, Butterfield, & Gravenstein, 2014;Draper, Karmel, Gibson, Peut, & Anderson, 2011;Kasteridis et al., 2016;Oud, 2017;Zekry et al., 2009) or up to 1 year prior to the first transition (Davydow et al., 2014;Givens et al., 2013;Harvey et al., 2016;Mitchell et al., 2015;, 2017. Nevertheless, prevalent and incident cases were not differentiated in these studies, and the time since diagnosis was not provided. Only four studies, all based on administrative data, explicitly identified participants with incident (Callahan et al., 2012(Callahan et al., , 2015Hsiao et al., 2015;Sivananthan & McGrail, 2016) or prevalent dementia (Callahan et al., 2012). Other studies relied mainly on clinical diagnoses or screening tests for identification purposes, often diagnosing or assessing participants either directly after the first transition (Boltz et al., 2015;Chang et al., 2015;Fogg, Meredith, Bridges, Gould, & Griffiths, 2017;Ono, Tamai, Takeuchi, & Tamai, 2011;Seematter-Bagnoud et al., 2012) or at study enrollment before the first transition occurred (Davydow et al., 2014;Fong et al., 2012;Noel et al., 2017;Rudolph et al., 2010;Voisin, Sourdet, Cantet, Andrieu, & Vellas, 2009). In these studies, it is possible that some participants were previously diagnosed with dementia. In studies with long observation periods in which dementia had been diagnosed at enrollment, it is also possible that some participants had been living with dementia for some time before their first transition.

Inclusion criteria
Study participants may reside in any setting at the start of the study.
Study participants must be individuals with dementia (and caregivers) only, or individuals with dementia (and caregivers) and controls wherein those with dementia are analyzed as a separate group.
Transitions between settings must be a study outcome, not a predictor.
Care settings must not include emergency departments, physician offices, or any care setting that does not imply an overnight stay.

Exclusion criteria
Mixed-methods and qualitative studies Letters to the editor, opinion letters, commentaries, dissertations, reviews, study protocols, policy papers, reports, grey publications, and book chapters Studies involving care settings that do not require an overnight stay (e.g., emergency department, physician office, home care) 74 Julie G. Kosteniuk et al.

Prevalence of Transitions and Trajectories
We identified 26 distinct care trajectories experienced by individuals with dementia, based on the starting location of the trajectory and care settings involved, transition pattern, and number of transitions experienced ( Figure 2 and Table 3). In most trajectories, the first transition involved hospitalization (n = 24); in two trajectories, the first transition was from LTC to home. Considering all 26 trajectories, the destination of the second transition was LTC in 10 trajectories, home or other location in 6, unspecified in 6, hospice care in 2 and hospital in 2 trajectories. Described subsequently and organized by starting location are the seven most common trajectories, defined as those that were each examined in three or more of the included studies. These trajectories were examined in a total of 26 of the included studies. Included is the prevalence of each trajectory or the prevalence of specific transitions within the trajectories, as reported in the studies.

76
Julie G. Kosteniuk et al. Two transitions • 0.3% of LTC residents were discharged from hospital to a different LTC facility (i.e., experienced a "type 2" care transition reflecting a lack of continuity in LTC facility) (90-day time frame) Four or more transitions • 5.5% of LTC residents were hospitalized more than two times (i.e., experienced "type 3" care transition) (90-day time frame) Associated factors Two transitions • After adjusting for age, sex, year of death, and co-morbidity, the odds of "type 2" care transitions from a LTC facility were lower among the oldest than among the youngest age group (AOR 0.35; 95% CI 0.15, 0.82), higher among residents of sheltered housing at baseline than among those in a traditional nursing home (AOR 13.69; 95% CI 6.36, 29.44) and higher among residents of specialized LTC for people with dementia at baseline than among those in a traditional nursing home (AOR 15.31; 95% CI 5.68, 41.30) Four or more transitions • After adjusting for age, sex, year of death, and co-morbidity, the odds of "type 3" care transitions were lower among the oldest than among the youngest age group (AOR 0.73; 95% CI 0.58, 0.90), lower among females than among males (AOR 0.74; 95% CI 0.64, 0.85), higher among residents of sheltered housing at baseline than among residents of traditional nursing home (AOR 4.08; 95% CI 3.56, 4.69), and higher among residents of specialized LTC for people with dementia at baseline than among residents of traditional nursing home (AOR 3.04; 95% CI 2.33, 3.97) Key implications and recommendations • Traditional nursing homes are likely better equipped than sheltered housing facilities to treat medical conditions that may arise at end of life and to avoid multiple hospitalizations • Specialized LTC facilities for individuals with dementia should be further regulated to ensure that they provide proper end-of-life care and prevent multiple hospitalizations

Two transitions
• There was no difference in the proportion of patients transferred from hospital to a nursing home between the intervention and control groups (27% vs. 26%; p = 0.56) (18-month time frame) Three transitions • The intervention group had a lower 30-day hospital readmission rate than the control group (7% vs. 24%; p = 0.02) Intervention • The Family-Centred Function Focused Care intervention (Fam-FCC) was adapted to be "dementia capable" and is implemented by a registered nurse over 10-15 hours/ week. Components included environmental and policy assessment (baseline, 3 and 9 months), staff education (baseline, 2, 3, and 5 months), ongoing training and motivation of nursing staff (during 12 months), and family/patient education (during 12 months.). The control group received staff education. Key implications and recommendations • Reduced readmissions point with possible economic savings; however, further research such as a randomized controlled trial with a larger sample and longer intervention period is required Two transitions • 91% of patients living in a nursing home before hospitalization were discharged to a nursing home, none were discharged home, and 9% were discharged to an "other" location (3.5-year time frame) • 58.4% of patients living at home before hospitalization were discharged to a nursing home, 29.3% were discharged home, and 12.6% were discharged to an "other" location (3.5-yr time frame) • 50% of patients living in an "other" location before hospitalization were discharged to a nursing home, 22.7% were discharged home, and 27.3% were discharged to an "other" location (3.5-year time frame) Associated factors • Patients discharged from hospital to a nursing home were older than those discharged home (mean age 82.03 years vs. 78.5; p = 0.002) • Widowed patients were more likely than married or other patients to be discharged from hospital to a nursing home (76.9% versus 54%, 67%; p = 0.001) • After controlling for age and sex, the odds of being discharged from hospital to home vs. to a nursing home were lower for males (OR 0.86; 95% CI 0.93, 0.99) Key implications and recommendations • Limited availability of home care and other home support services may contribute to discharge from hospital to nursing home • Longer hospital stays among patients discharged to a nursing home compared with other locations may be partly attributed to limited availability of space in nursing homes • Males may be less likely than females to be discharged from hospital to home if they do not have a spouse living at home • Future research should take into account the effect of family caregivers and hospital care on living arrangements after hospital discharge Callahan et al., 2012 United States Prospective cohort Dementia identification: ICD diagnosis code at or prior to start of the study period (prevalent cases) or during the study period (incident cases) n = 4,197 primary care patients in one city (minimum age N/A) 524 prevalent dementia mean age = 75.0 AE 6.7 female = 64.7% 999 incident dementia mean age = 73.4 AE 6.3 female = 71.5% 2,674 no dementia mean age = 70.7 AE 5.3 female = 68.6% rural N/A 30 days, 5.2 years (mean)/7 years Two transitions

Two transitions
• 20% of all individuals with prevalent or incident dementia were transferred from LTC to hospital and back to LTC, 14% were transferred from LTC to home and then to hospital, and 10% were transferred from LTC to home and back to LTC (5.2 year mean) Three transitions • The 30-day re-hospitalization rate among individuals with prevalent or incident dementia discharged to LTC was 25.3%, compared with 21.8% among those discharged home with services, and 20.2% among those discharged home without services Key implications and recommendations • Individuals with dementia should receive monitoring as they move across care settings, such as from LTC facilities back into their homes • To reduce re-hospitalizations, LTC facilities should be considered "transitory" sites in light of the frequency with which individuals with dementia transition between acute and LTC sites • Clarifying the goals of care is equally important as examining care transitions, regardless of whether care is provided at home, in hospital, or in LTC, given that much of the substance of the care for individuals with dementia is similar across the three sites

Two transitions
Compared with individuals without dementia: • Individuals with moderate/severe dementia (rate ratio 2.85; 95% CI 2.43, 3.33) or mild dementia (rate ratio 2.29; 95% CI 1.99, 2.64) were more likely to transition from LTC to hospital and back to LTC (12-month time frame) • Individuals with moderate/severe dementia (rate ratio 1.33; 95% CI 1.13, 1.55) were more likely to transition from LTC to home without services, and then to hospital (12-month time frame) • Individuals with moderate/severe dementia (rate ratio 0.65; 95% CI 0.43, 1.00) or mild dementia (rate ratio 1.14; CI 0.84, 1.55) were as likely to transition from LTC to home with services, and then to hospital (12-month time frame) Three transitions Compared to individuals without dementia: • Individuals with mild dementia (rate ratio 1.08; 95% CI 1.01, 1.15) were more likely to transition from hospital to home without services, and back to hospital (12-month time frame) • Individuals with moderate/severe dementia (rate ratio 6.19; 95% CI 5.13, 7.48) or mild dementia (rate ratio 4.54; 95% CI 3.88, 5.32) were more likely to transition from hospital to LTC and back to hospital (12-month time frame) • Individuals with moderate/severe dementia (rate ratio 2.68; 95% CI 2.34, 3.06) or mild dementia (rate ratio 2.10; 95% CI 1.88, 2.36) were more likely to transition from hospital to LTC and then to home without services (12-month time frame) • Individuals with mild dementia (rate ratio 1.37; 95% CI 1.12, 1.67) were more likely to transition from hospital to home with services, and back to hospital (12-month time frame) Key implications and recommendations • Individuals with dementia including those with moderate to severe dementia, experienced compound transitions among LTC, hospitals, and other settings such as home, to a greater degree than those without dementia • Hospitals are the "front door" to LTC for many individuals with dementia, and a growing 'network of care' exists that involves formal and informal providers and settings across home, hospitals, and long-term care facilities • Improve support for individuals with dementia in formal and informal settings    • Including patients with heart failure, the rate of drug-related hospital readmission was lower among the intervention than the control group (5% vs. 11%; p =0.03) Three transitions (180-day time frame) • Including patients with heart failure, there was no difference in the rate of drugrelated hospital readmission among the intervention and control groups (18.9% vs. 23%; p = 0.28) • Excluding patients with heart failure, the rate of drug-related hospital readmission was lower among the intervention than the control group (11% vs. 20%; p = 0.02) Intervention • The intervention was conducted by three clinical pharmacists on existing hospital ward teams. The pharmacists participated in rounds and conducted medication reconciliation and review, meeting throughout the study to align the advice they provided to the health care teams regarding drug selection, dosage, and monitoring during rounds.
(Continued) Two transitions • Individuals with dementia were more likely to be discharged from hospital to a residential aged care facility than those without dementia (25.3% vs. 5.5%; p < 0.0001) (time frame unspecified) Three transitions • The 28-day hospital readmission rate was lower overall among individuals with dementia than among thos without (16% vs. 17.5%; p < 0.0001); however, there were exceptions according to injury type, specifically non-fracture arm injuries and toxic effects of non-medical substances (28-day time frame) Key implications and recommendations • The length of stay in hospital may be longer in individuals with dementia because they require a longer period of rehabilitation • Hospital readmission rates may be lower among individuals with dementia than among those without because the former are more likely to live with others or have better supports in place at discharge

Two transitions
• Individuals with dementia admitted to hospital with a primary diagnosis of dementia were more likely to be discharged to a care home than those with dementia admitted to hospital for treatment of an ambulatory care sensitive condition (19% vs. 14%, significance test not provided) (5-year time frame) Associated factors • Hospitalized patients with dementia had a greater risk of discharge to a care home with older age, female sex, incontinence, falls, hip fracture, cerebrovascular disease, cancer, and total number of other co-morbidities • Hospitalized patients with dementia had a lower risk of discharge to a care home if they lived in areas with a higher rate of unpaid care provision (50 or more hours/ week) and a higher proportion of guaranteed pension recipients Key implications and recommendations • The risk of care home placement may be reduced by addressing important risk (Continued)    • The risk of discharge from hospital to hospice care was higher among patients 80 years of age and older (AOR 1.8; 95% CI 1.7, 1.8), lower among individuals of "nonwhite" ethnicity (e.g., African-American AOR 0.67; 95% CI 0.65, 0.70), lower among those with health insurance (e.g., Medicaid AOR 0.41; 95% CI 0.37, 0.46), higher among patients with certain non-dementia co-morbidities (e.g., malignancy AOR 3.72, 95% CI 3.61, 3.84), higher among those with a greater number of failing organs (e.g., 4 or more AOR 6.18, 95% CI 5.61, 6.82), and higher among those on mechanical ventilation (AOR 1.12, 95% CI 1.06, 1.18) Key implications and recommendations • The negative association between health insurance and discharge from hospital to hospice care was unexpected and should be examined in future research • Future intervention research should take into account the factors associated with discharge from hospital to hospice care found in the present study Clinical diagnosis during hospitalization n = 1,764 patients admitted to one hospital (≥ 70 years of age) 1,038 cognitively intact mean age = 81.0 AE 6.6 female = 70% 301 CIND mean age = 83.7 AE 6.9 female = 73.1% 425 dementia mean age = 84.4 AE 6.3 female = 66.1% rural N/A 3 years/3 years Two transitions • x-hospital-LTC Two transitions • Patients with dementia were more likely than CIND and cognitively intact patients to be admitted to a nursing home after discharge from post-acute rehabilitation (28.8% versus 7.6%, 4.2%; p < 0.001) (3-year time frame) • After adjusting for demographic and clinical characteristics (e.g., depressive symptoms, formal home care before admission), patients with dementia had higher odds of institutionalization than cognitively intact patients (AOR 6.1; 95% CI 4.0, 9.3) and CIND patients (AOR 5.6; 95% CI 3.2, 9.7) (3-year time frame) Key implications and recommendations • Future research should address the possibility that longer hospital stays among patients with dementia, compared with those without dementia, are attributable to slower rehabilitation progress or complications from dementia • Interventions should be developed to advance earlier dementia diagnosis, such as cognitive screening in hospital, and to improve care for patients with dementia in hospital (Continued) Two transitions • In the first year of diagnosis, 5% of individuals with dementia already in LTC were hospitalized once and subsequently moved to an unspecified location (1-year time frame) Three transitions • In the first year of diagnosis, 20% of individuals with dementia in the community (i.e., home) were hospitalized two or more times (1-year time frame) • In the first year of diagnosis, 9.3% of individuals with dementia already in LTC were hospitalized two or more times (1-year time frame) Four or more transitions • In the first year of diagnosis, 2% of individuals with dementia already in LTC were hospitalized two times and moved to an unspecified location (1-year time frame) Key implications and recommendations • Care that is integrated across sectors is necessary to reduce transitions among individuals with dementia. • Further research is warranted to identify interventions to reduce transitions during the year of diagnosis and end of life   in Japan, that 23.3 per cent of patients first admitted to a ward for behavioural and psychological symptoms related to dementia, were readmitted to the same ward within 2 years. In a large retrospective cohort study of home care clients with dementia in one Canadian province, Mondor et al. (2017) reported that 13.5 per cent of all clients experienced two or more hospital admissions within 1 year. A separate retrospective cohort study in another Canadian province further demonstrated that 20 per cent of community-based individuals experienced two or more hospital admissions in 1 year; specifically, in the first year of the dementia diagnosis (Sivananthan & McGrail, 2016). Two separate intervention studies reported a reduction in hospital readmission rates among patients with dementia in an intervention group compared with a control group (Boltz et al., 2015;Cassel et al., 2016). In the first study, conducted in two United States hospitals, Boltz et al. (2015) found a lower 30-day hospital readmission rate among patients with dementia in the intervention compared with the control group (7% vs. 24%). Whereas the control group received staff education only, the intervention group was exposed to multiple elements: staff received family-centred education, ongoing training, and motivation; environmental and policy modifications were made, such as the addition of bedside white boards to increase communication between staff and patient/family; and patient/family education encouraged engagement in developing and communicating individualized goals and expectations. Similarly, the second study found that patients with dementia in a U.S. state who participated in a specialty palliative care program intervention had a lower rate of 30-day hospital readmission than non-participants (11% vs. 35%) (Cassel et al., 2016). The third common trajectory that started from home, which was investigated in three studies, involved a first hospitalization followed by readmission from an unspecified location (home-hospital-x-hospital) (Fong et al., 2012;Rudolph et al., 2010;Voisin et al., 2009). Fong et al. (2012 considered participants of a specialized US research centre with Alzheimer's disease (AD) residing at home in the community and admitted to hospital, finding that 61 per cent were readmitted to hospital within 1 year. Two articles reported the number of times that community-based individuals with AD were admitted to hospital, with Rudolph et al. (2010) finding that 47 per cent of patients attending a community-based United States research centre were admitted to hospital two or more times over an average of 4 years, and Voisin et al. (2009) finding that 19.8 per cent of participants of a prospective cohort study nationwide in France were hospitalized two times over 2 years.

From LTC
As investigated in five articles, the trajectory originating from LTC that was most often studied involved transfer to hospital and back to LTC (Aaltonen et al., 2014;Bucher et al., 2016;Callahan et al., 2012Callahan et al., , 2015Givens et al., 2013). Regarding this LTC-hospital-LTC route, Bucher et al. (2016) reported that 91 per cent of people with dementia living in Swiss nursing homes before hospitalization were discharged to LTC over 3.5 years. Callahan et al. (2012) observed that 20 per cent of primary care patients with dementia in a U.S. state were admitted from LTC to hospital and later discharged back to LTC over an average 5-year time frame. In a later study using national data from the United States Health and Retirement study, Callahan et al. (2015) reported that individuals with dementia compared with those without were more likely to experience this route over a 12-month time frame, regardless of the stage of dementia.
found in the studies that included control patients (Hsiao et al., 2015;Kedia et al., 2017). In two separate studies, each with a 1-year time frame, 2.2 per cent of individuals with dementia compared with 0.6 per cent of controls experienced three admissions in a national study of Taiwan (Hsiao et al., 2015), and 19.9 per cent of individuals with dementia and concurrent cancer across three U.S. states experienced three or more hospitalizations compared with 1.6 per cent with neither dementia nor cancer (Kedia et al., 2017). In a third study using United States national data, 10.7 to 12 per cent of individuals with dementia in the last 90 days of life underwent three or more hospital admissions compared with 13.2 to 19.9 per cent of those with cancer or chronic obstructive pulmonary disease (COPD): however, significance testing was not provided .

Factors Associated with Transitions and Trajectories
Eleven studies overall examined factors associated with transitions or care trajectories as a whole among individuals with dementia, with two of these studies examining variables related to more than one transition or trajectory (Aaltonen et al., 2014;Ono et al., 2011). Factors examined in each study are summarized in Table 3. Individual demographic and medical characteristics were the main focus in most studies, and both types of characteristics were examined in relation to re-hospitalization, repeated hospitalization, transition from hospital to hospice care, and transition from hospital to LTC. Organizational variables were examined in relation only to transition from hospital to LTC, and with respect to a trajectory that involved residence in two different LTC homes before and after hospitalization.
In terms of medical characteristics examined in eight studies, co-morbid dementia and depression were associated with greater odds of 30-day re-hospitalization from an unspecified location (Davydow et al., 2014). More severe cognitive impairment and increased medication use (Leung, Kwan, & Chi, 2013) increased the likelihood of re-hospitalization from LTC. Several medical factors were associated with hospitalization four or more times within 4 years of initial admission (repeated hospitalization), namely coronary artery disease, pneumonia, urinary tract infection, and fallrelated fracture recorded at index admission (Chang et al., 2015). Medical characteristics related to transition from hospital to hospice care included chronic co-morbidity, failing organs, and mechanical ventilation (Oud, 2017). Medical factors also increased the risk of transition from hospital to LTC in four studies, and included severe medical issues (Takacs et al., 2015); poor functional status and severe dementia (Zekry et al., 2009); a higher number of co-morbidities, incontinence, falls, hip fracture, cerebrovascular disease, and cancer (Kasteridis et al., 2016); and percutaneous endoscopic gastrostomy (PEG) tube insertion in hospital, greater functional ability, and diabetes (Givens et al., 2013).
Demographic factors examined in relation to care transitions in four studies mainly centred on age and sex; however, other variables were also considered. Ono et al. (2011) found sex differences in early re-hospitalization from home within 3 months of discharge from a hospital ward for patients with dementia. Ono et al. reported that having fewer cohabitants predicted readmission among males but not females, and that a longer stay in the index hospital predicted readmission among females but not males. Another study reported that the likelihood of three or more hospitalizations (repeated hospitalization) among LTC residents increased for those who were younger and male (Aaltonen et al., 2014). Demographics related to transition from hospital to hospice care included older age, "white" ethnicity, and lack of health insurance (Oud, 2017). Several demographic characteristics were associated with a higher risk of transition from hospital to LTC, including female sex (Kasteridis et al., 2016;Ono et al., 2011), male sex (Bucher et al., 2016), older age (Bucher et al., 2016;Kasteridis et al., 2016), and widowed marital status (Bucher et al., 2016). Patient demographics associated with a lower risk of transition from hospital to LTC included living in areas with a higher rate of unpaid care provision (50 or more hours/week) and a higher proportion of guaranteed pension recipients (Kasteridis et al., 2016). In examining the LTC1hospital-LTC2 trajectory, Aaltonen et al. (2014) found a relationship between younger age and residence in two different LTC homes pre-post hospitalization in the last 90 days of life.
Two studies also identified variables associated with care trajectories at the organizational level. The first study found that nursing home characteristics were associated with transition of residents from hospital to a skilled nursing facility, specifically large size (more than 100 beds), corporate chain status, for-profit structure, and urban versus rural location (Givens et al., 2013). The second study identified that living in sheltered housing or a specialized LTC versus a traditional nursing home was associated with residence in two different LTC homes pre-post hospitalization in the last 90 days of life (Aaltonen et al., 2014).

Discussion
This scoping review identified and classified care trajectories across multiple settings among people with dementia, and investigated the prevalence of multiple transitions and factors associated with transitions. We identified 26 distinct trajectories, including 7 trajectories that were each investigated in three or more studies and considered to be common pathways for the purpose of this review. Trajectories that involved either hospital readmission or discharge from hospital to LTC were most common. Dementia increased the likelihood of a single transition from hospital to LTC as well as hospital readmission. Factors associated with particular transitions were identified mainly at the individual level of medical and demographic characteristics. Complex care trajectories that involved numerous transitions over the course of several years were not typically considered in the studies, suggesting opportunities for future investigation.
Four of the most common trajectories involved three or more transitions consisting of hospital readmission with initial admission from home or an unspecified setting. In studies in which the overall trajectory was considered, prevalence ranged from 4.4 to 19.8 per cent for two admissions within 1-2 years (Hsiao et al., 2015;Voisin et al., 2009) and from 10.7 to 19.9 per cent for three or more admissions within 90 days to 1 year (Kedia et al., 2017;Teno et al., 2013). Variations across studies may be partly the result of differences in time frames and variations in health care systems across countries. Although studies with longer timeframes of 3 months to 1 year found that patients with dementia were more likely to be re-hospitalized than were controls (Draper et al., 2011;Hsiao et al., 2015;Kedia et al., 2017), studies with shorter periods of 28-30 days found mixed results (Daiello et al., 2014;Harvey et al., 2016;Kedia et al., 2017;Mitchell et al., 2015Mitchell et al., , 2016Mitchell et al., , 2017. These findings suggest a need to improve post-discharge care management for persons with dementia (Lin, Zhong, Fillit, Cohen, & Neumann, 2017) and a greater role for primary health care providers and community care providers in coordinating management (Austrom, Boustani, & LaMantia, 2018).
This review found that re-hospitalization factors were identified mainly in terms of medical and demographic characteristics. Previous research suggests that readmission risk may be reduced by effective discharge planning and provision of home health services after discharge, particularly for those who live alone (Lin et al., 2017). Specifically for persons in LTC, Leung et al. (2013) found that re-hospitalization risk increased with more severe cognitive impairment and medication use. Nearly half of all older adults living in LTC in Canada are prescribed at least 10 different drugs (Canadian Institute for Health Information, 2018b); however, awareness of the need to de-prescribe for older adults with dementia is growing among health care providers (Canadian Foundation for Healthcare Improvement, 2019).
Three other common trajectories involved two transitions consisting of discharge from hospital to LTC in the second transition. Among people with dementia living at home before hospitalization, the prevalence of a second transition to LTC ranged from 16 to 36 per cent compared with less than 9 per cent for those without dementia, over periods of 1-2 years (Fogg et al., 2017;Harvey et al., 2016;Kedia et al., 2017;Mitchell et al., 2016;Zekry et al., 2009). Included studies suggested that transition from hospital to LTC may follow from interactions among the cause of hospitalization, pre-existing conditions including cognitive and functional impairment, and the hospitalization experience itself. Cognitive impairment, physical assistance needs, and assessment for LTC admission performed in hospital are factors shown in prior research to increase the odds of admission to LTC (Canadian Institute for Health Information, 2017).
Several medical and demographic characteristics were also found to be associated with a single transition from hospital to LTC. Demographic factors included widowhood (Bucher et al., 2016), older age (Bucher et al., 2016;Kasteridis et al., 2016), and female sex (Kasteridis et al., 2016;Ono et al., 2011). Older females are more likely to live alone than older males (United Nations Department of Economic and Social Affairs, 2019), which may contribute to a greater risk of LTC admission. Several medical factors associated with LTC admission from hospital included severe dementia, severe medical issues, poor functional status, co-morbidity, falls, and hip fractures (Kasteridis et al., 2016;Takacs et al., 2015;Zekry et al., 2009). Older adults with dementia often have co-morbidities that can impair cognition and increase the challenge of management in the community (Austrom et al., 2018). Previous research shows that hospital administrators may follow policies that prioritize discharge to LTC rather than to home in the community, partly because of hospital capacity issues (Canadian Institute for Health Information, 2017). Initiatives to train and support care partners in combination with respite care may be effective in delaying LTC admission (Gresham, Heffernan, & Brodaty, 2018).
Interventions examined in included studies that were effective in reducing transitions focused exclusively on hospital readmissions, and were delivered either in-hospital or in-home (Boltz et al., 2015;Cassel et al., 2016;Gustafsson et al., 2017;Noel et al., 2017). Care partners were targeted in three of the four interventions, underscoring the importance of tailoring support and involving care partners in management. For example, reduced readmissions in patients with dementia were achieved after an in-hospital program trained care partners to take part in patient recovery in-hospital and after discharge (Boltz et al., 2015). Previous research advises health care providers to follow principles of family-centred care in ongoing management, such as involving care partners in conversations about medications, providing education about behaviors related to dementia, and soliciting care partners' reports about indications of pain (Austrom et al., 2018).
Findings point to opportunities for reducing the risks of transition to hospital and LTC by strengthening systems of community-based care for people living with dementia and care partners. The need for stronger community care systems is underscored by the significant effects of the COVID-19 pandemic on hospitals and LTC homes in Canada (Brown et al., 2020;Canadian Institute for Health Information, 2020). Further resources are necessary to train and support all segments of the community care system in dementia care, from paid care providers including personal care workers, first responders, and a wide range of health care professionals (e.g., family physicians and occupational therapists) to unpaid care partners who provide the majority of in-home care (Public Health Agency of Canada, 2019a). Optimal systems of community care also involve strategies for providing support with daily activities such as shopping, exercise, and technology; coordinating medical services and information sharing across paid providers; and making publicly funded home care and supportive housing widely available across jurisdictions (Boscart, McNeill, & Grinspun, 2019; Canadian Academy of Health Sciences, 2019; Canadian Institute for Health Information, 2018a). As the needs of people with dementia and their care partners vary and change as the condition progresses, practice guidelines and staffing to ensure early intervention and crisis prevention should also be widely available (Canadian Home Care Association, 2018; Canadian Institute for Health Information, 2018a).
Important opportunities for future research were revealed through this review. Complex care trajectories experienced by people living with dementia were not considered by studies in the review. In previous studies of older adults, researchers found between 131 and 240 unique transition patterns across several care settings with varying time frames (Abraham & Venec, 2016;Sato, Shaffer, Arbaje, & Zuckerman, 2010). Complex trajectories may signal fragmented care (Wang et al., 2017), particularly when occurring over a short time period. There is also a need for further study of care trajectories from the point of pre-diagnosis to end of life among people living with dementia (Boltz, 2016;Fortinsky & Downs, 2014). The longest transition time frame in the current review was 5 years (Callahan et al., 2012), signalling an opportunity for more longitudinal studies that may uncover patterns or group differences that point to discontinuity of care, such as a higher risk of transitions associated with demographic factors (Wang et al., 2017). Moreover, the risk of complex trajectories has been found to increase with the number of prior transitions experienced by people with dementia (Hathaway, 2019), further underscoring the importance of longitudinal research. Only one included study examined transition outcomes among female and male subpopulations separately (Ono et al., 2011). Sex and gender differences in care trajectories should be considered further, as sex and gender differences in neurodegenerative disorders have been observed across many studies (Tierney, Curtis, Chertkow, & Rylett, 2017). Moreover, no study in the review separately analysed transition patterns in rural and urban sub-populations. A larger share of older adults in Canada live in rural communities than in cities (Statistics Canada, 2017) yet face significant barriers in terms of accessing dementia-specific services (Morgan et al., 2015). Further investigation into rural and urban trajectories would make a substantial contribution to this research area. Finally, further research is needed concerning care trajectories among important subpopulations living with dementia. Traditionally under-represented in these studies are people with intellectual disability and youngonset dementia, and diverse ethnic groups (Boltz, 2016).

Limitations
The scoping review method allowed for a comprehensive search of the published literature and a subsequent classification of care trajectories experienced by people with dementia. The increase in published literature in recent years on the topic of multiple transitions among people with dementia fits well with the utility of scoping reviews for clarifying key concepts and knowledge gaps in emerging topics (Munn et al., 2018). It should be noted that this review did not include an assessment of the quality of included studies. Although quality assessment is not a common practice in scoping reviews (Peters et al., 2015), the lack of critical appraisal may nonetheless reduce uptake of findings into practice (Levac et al., 2010).
Grey literature and non-English articles were excluded, as were mixed-methods and qualitative studies, which may limit the generalizability of the findings. These studies were excluded because the purpose of this review was well suited to quantitative analysis, and less focused on evidence concerning the needs or experiences of people with dementia during the transition process. The latter is an important area that has been investigated with qualitative and mixed methods in previous reviews (Afram et al., 2015;Stockwell-Smith et al., 2018).
Studies included in the current review varied considerably, with some samples including those with Alzheimer's disease only, others including separate causes of dementia within the same study, and yet others not specifying the causes of dementia. Transition time frames varied from 28 days to 5 years, and studies from several different countries were included to cover the breadth of available evidence, consistent with scoping review methodology (Peters et al., 2015). These variations add to the challenge of drawing conclusions based on the findings.
Finally, it was not always possible to determine the location of settings from the information provided in the studies. Routes that began in hospital and involved a transition to a different care site directly afterward were considered to originate in an 'unspecified' location for the purpose of the review. We rationalized that individuals were admitted temporarily to hospital from permanent or long-stay living accommodations and that therefore this constituted the first transition. Unspecified locations may in fact be home, LTC, or even another hospital, thus potentially biasing the findings. It is also possible that individuals experienced more than one transition before the initial move to hospital. The time frames imposed on the observation of routes may reduce this possibility; nevertheless, it is a limitation of this study that the framework of multiple transition routes does not capture all possible patterns.

Conclusions
This review found that studies of care trajectories experienced by people with dementia most often involved hospital readmission and discharge from hospital to LTC. Studies reported that risk of hospital readmission and transition from hospital to LTC increased with dementia. Several variables contributing to transition risk were identified, mainly at the level of demographic and medical characteristics such as co-morbidity, as well as severity of both medical issues and dementia. Research opportunities exist to investigate more complex care trajectories, longitudinal trajectories of care, and trajectories experienced by sub-populations of people with dementia. Findings suggest that greater attention should be paid to care trajectories in this population, given the negative outcomes associated with transitions for people with dementia and their care partners. Efforts to strengthen community-based systems of care as part of a multifaceted strategy to reduce transitions are recommended. These efforts should involve considerable investment in comprehensive dementia-specific training to care providers and care partners, widely available home care and supportive housing, and practical support for people with dementia to live independently for as long as safely possible in their communities.
Supplementary Materials. To view supplementary material for this article, please visit http://doi.org/10.1017/S0714980821000167.