A systematic review of interventions to improve acute hospital care for people with dementia

Background: People with dementia are confronted with negative consequences due to hospital stays. When developing new interventions to improve care for people with dementia in acute care hospitals (by intervention we understand organisational strategies, programmes, and/or treatments and actions performed by healthcare professionals), it is crucial to have a comprehensive overview of previous interventions tested in this eld and to know the outcomes. Methods: We conducted a systematic review to provide an overview of interventions targeting the care situation of patients with dementia in acute care hospitals. We included trials with interventional study designs and systematic reviews without restrictions regarding the outcomes. We searched ve electronic databases, conducted a hand search of journals and performed forward and backward citation tracking of included studies. For risk of bias assessment, we used the Cochrane Collaboration’s tool for assessing risk of bias in randomised trials, ROBANS (for non-randomised controlled trials) and AMSTAR (for systematic reviews). We provide tabular and narrative summaries of the ndings. Results: Twenty studies met the inclusion criteria. The ndings indicated a broad range of interventions and outcomes. We categorised the interventions into eight intervention types: educational programmes, special non-pharmacological interventions, delirium management programmes, inpatient rehabilitation interventions, family-/person-centred programmes, use of specially trained nurses, volunteer programmes and special care units. Staff outcomes were investigated in most of the studies, followed by patient outcomes. Outcomes concerning relatives were reported in four studies. Educational programmes were the most commonly reported intervention type and resulted in improved staff outcomes. Family-/person-centred care programmes, use of specially trained nurses and delirium management programmes were effective in improving patient-related outcomes, e.g. delirium severity and functional performance. Rehabilitation interventions after hip fractures and special care units for people with dementia hardly showed any signicant effects. Conclusions: Although the included interventional studies reported improved outcomes concerning patients, relatives and staff, the evidence is insucient to declare which interventions are effective in improving care for people with dementia in acute care hospitals. Future research should focus on relevant patient and family caregiver outcomes. Thereby, it is necessary to consider the complexity of the interventions. Interventions addressing relatives, interprofessional collaboration and the entire practice culture are needed.

with dementia in acute care hospitals, it is, however, crucial to have a comprehensive overview of previously investigated interventions and reported outcomes in this eld. This review is part of a larger project that aims to develop a care model for persons with dementia in the acute hospital setting in Switzerland.
Recent knowledge about the investigated interventions will inform the development of this care model and will be important when developing and evaluating other initiatives, programmes or interventions in this eld.

Objective
The aim of this review was to provide a systematic overview of interventions targeted to improve the care of patients with dementia and cognitive impairment on regular wards (all wards except intensive care units and emergency departments) in acute care hospitals. This review will show which kinds of interventions were tested in this eld, which outcomes were reported, and which effects were achieved. This knowledge can contribute to developing improved and evidence-based care for people with dementia in acute care hospitals.
We conducted a systematic review of interventional studies and systematic reviews to gather the best possible evidence for developing research-based dementia care in acute care hospitals.
This review is part of a larger project aiming to develop a care model for persons with dementia in the acute hospital setting in Switzerland. Recent knowledge about investigated interventions will inform the development of this care model and will be important for elaborating and evaluating other initiatives, programmes or interventions in this eld.

Methods
Criteria for considering studies for this review We included articles with (1) interventional study design (randomised controlled trials, controlled clinical trials, before-after studies) and systematic reviews of interventional studies assignable to a speci c intervention type, (2) published in journals, books or as grey literature in (3) English or German (4) since 2009. It was our aim to get the most up to date body of evidence from the last ten years. This is necessary since dementia care has changed in the last decade, e.g.
regarding the development of national dementia strategies, nursing education or initiatives to improve hospital care for people with dementia. We included studies if they (5) were conducted in the acute hospital setting and (6) examined any intervention targeting the care situation of (7) patients with dementia (including all types and stages and all ages) in this setting. We also included studies focussing on patients with cognitive impairment in general since the dementia diagnosis is often not known in the hospital setting [1]. Outcomes of interest were any patient outcomes and any formal or informal caregiver outcomes. Since patient outcomes may be in uenced by various factors (e.g. caregivers` knowledge, interactions between the person and the formal or informal caregiver), patient-related outcomes as well outcomes related to formal and informal caregivers were of interest.
We excluded trials with (1) no interventional study design, as well as (2) abstracts and (3) conference proceedings. We did not include studies focussing on interventions outside the acute hospital setting (interventions to avoid hospital admissions or post-discharge interventions), studies addressing pharmacological interventions or exclusively persons with delirium (without dementia) (4). We also excluded studies investigating interventions conducted in the psychiatric setting, special dementia units (except if the special dementia unit was the intervention), intensive care units or emergency departments.
Two reviewers independently screened the titles and abstracts. Afterwards, we obtained the full texts of the remaining references and the two reviewers read them independently. We solved all con icts by discussion. The ow chart ( Figure 1) shows the selection process depicting the number of identi ed, excluded and included references in detail.

Search strategy
We conducted a comprehensive literature search from January to March 2019. In June 2020, we updated the search. The following electronic databases were searched (from 01/2009 until 06/2020): CINAHL, MEDLINE via PubMed, Web of Science, PsycINFO via Ovid and the Cochrane Library. Furthermore, we performed forward and backward citation tracking of included studies using the Scopus database. Additionally, we conducted a free web search in Google Scholar as well as a hand search of relevant journals not indexed in the databases (Journal/ISSN: P egezeitschrift/2520-1816, HeilberufeScience/2190-2100, Angewandte Gerontologie Appliquée/2297-5179 and Klinische P egeforschung/2365-7863).
We used the following search terms in combination with the operators AND and OR: dementia, Alzheimer, hospital, acute care, acute hospital care, inpatients, hospital units, experimental studies, quasi-experimental studies, randomized controlled trial, controlled clinical trial, randomised, random allocation, trial, experiment, quasi-experiment, clinical trial, intervention, before-after, pretest-posttest, nonrandomized, evaluation and repeated measures design. We used MesH terms where available and used them as search terms as well. With this sensitive approach we aimed to nd all relevant interventional studies in the eld of dementia in hospitals. The search strings for all databases are available in Additional le 1.
One author developed the search strategy and another author checked it by means of the PRESS-Guideline [18] before we conducted the search. Both authors are experienced in developing search strategies, particularly in the eld of dementia care research. We registered the review protocol in PROSPERO (https://www.crd.york.ac.uk/PROSPERO/): CRD42018111032.
Due to the heterogeneity of interventions and outcome measures, we did not undertake a meta-analysis. We performed a narrative synthesis of included studies.

Risk of bias assessment
To assess the risk of bias of randomised controlled trials, we used the Cochrane Collaboration's tool for assessing risk of bias in randomised trials [19]. For non-randomised controlled trials and before-after studies, we used ROBANS [20]. Risk of bias of systematic reviews was assessed with AMSTAR 2 [21]. Two authors independently assessed risk of bias of included studies. We solved discrepancies by discussion.
Further studies investigated a delirium management programme for people with cognitive impairment [27] (n = 1) and a special non-pharmacological intervention [22] (n = 1). The systematic reviews focussed on educational interventions [15,35] (n = 2), on volunteer programmes [38] (n=1), on special care units for people with dementia [16] (n = 1), on palliative care consultation services for people with advanced dementia [39] (n=1), on interventions to improve caregiver readiness [37] (n=1), and on inpatient rehabilitation interventions for people with dementia after hip fracture on surgical units [36] (n = 1). Table 1 provides information about the characteristics of included studies.

Risk of bias of included studies
The main weakness of the non-randomized interventional studies was the selection of participants since most studies did not use a control group [23][24][25][26][28][29][30][31]. Confounders were considered in four of the twelve studies [24,26,30,33]. Blinding of outcome assessments was only ful lled in two studies [22,28]. The risk for selective outcome reporting was rated as high or unclear in all non-randomized studies. Furthermore, we estimated the overall risk of bias for the randomized controlled trial as high. One systematic review showed very good quality in nearly all domains [36]. The others were of good quality in several domains with concerns regarding some domains. A detailed description and risk of bias table is provided in additional le 2.

Educational programmes
Intervention characteristics Three of the educational interventions for healthcare staff were aimed at different hospital professions (nurses, doctors, allied health professionals, occupational/physiotherapists) [23,28,33]. One focussed on both clinical and non-clinical staff [31]. Two addressed communication skills and used experiential as well as interactive strategies [21,26]. One addressed person-centred care, communication, self-protection and team/patient/family debrie ng [33]. Another intervention aimed to prepare nurses and change leader nurses for implementing a systematic nurse-caregiver conversation [30]. One intervention introduced key communication strategies and screening for cognitive impairment [31]. Furthermore, a visual cue was placed at the bedside to help staff identifying patients with cognitive impairment [31]. One educational intervention was an online training programme for nursing assistants and allied healthcare workers. It contained information on dementia-friendly principles, communication, wandering and falls [26]. The volunteer educational programme reported by Eggenberger et al. [29] addressed life situations of older people, communication, dying in hospital, activities with people affected by dementia as well as the role and identity of volunteers. The two systematic reviews about educational interventions included interventions targeted at healthcare staff from different professions, with nurses representing the largest group [15,35]. Details about intervention characteristics are displayed in Table 1.

Reported outcomes and effects
The following outcomes were assessed in the studies about educational interventions: knowledge, self-e cacy and con dence of staff, staff comfort, staff beliefs and attitudes, burnout, satisfaction with care for people with dementia, perceived organisational support, perceived equipment of hospital environment, perceived di culty when caring for patients with dementia and their families, number of palliative care consultations and changes in aggressive incidents.
Four interventional studies identi ed a signi cant effect regarding knowledge of healthcare staff [26,28,30] or volunteers [29]. There was a signi cant increase in knowledge after the end of a two-day communication training course for nurses, doctors and allied health professionals [28] and after an online training tool for nursing assistants and allied health workers [26]. The systematic reviews by Abley et al. [15] and Scerri et al. [35] showed mixed results, though most included studies showed signi cant improvement in knowledge. Furthermore, self-rated evaluations of sense of competence, con dence and self-e cacy signi cantly improved in four interventional studies describing training programmes for healthcare professionals [23,28,31,33]. One study with clinical and non-clinical staff reported signi cantly enhanced staff comfort [31]. In the systematic review by Scerri et al. [35], staff con dence was signi cantly higher in ve included studies, and self-e cacy was signi cantly elevated in two studies. In the review by Abley et al. [15], both studies measuring dementia con dence reported signi cant increase in con dence immediately post-intervention. However, at 120 days post-intervention, con dence levels signi cantly declined.
Self-assessed communication skills did not improve in the volunteer training study [29]. In another study only three of 11 communication behaviours showed signi cant change [28]. Scerri et al. [35] described improved documentation and risk assessment as well as a reduced use of sedations. One study included in the review by Scerri et al. [35] described less aggressive incidents pre-training compared to post-training. Table 2 provides an overview of all reported outcomes and effects. Detailed information about ndings of included studies and statistical data are included in additional le 3.

Special non-pharmacological interventions
Intervention characteristics One study focused on "simple pleasures" interventions for hospital patients with late stage dementia. To reduce identi ed expressions of unmet needs, patients received, for example, balls lled with rice or eece-covered warm water bottles [22].

Reported outcomes and effects
The study on "simple pleasures" interventions only reported outcomes related to agitation/aggression [22]. Wierman et al. [22] were not able to show a signi cant effect of "simple pleasures" interventions. However, they reported a trend towards an improvement regarding agitation. Table 2 provides an overview of all reported outcomes and effects.
Delirium management programme

Intervention characteristics
Hasemann et al. [27] investigated a nurse-led interdisciplinary programme to detect and treat delirium in patients with cognitive impairment. The programme contained an educational package, systematic screening of patients over 70 for cognitive impairment and the implementation of interdisciplinary interventions when delirium occurred.

Reported outcomes and effects
The study reported on outcomes regarding delirium severity/duration and benzodiazepine use [27]. Delirium severity decreased signi cantly on the adherent wards. This effect was evident from the rst to the second day of delirium and over the complete course of the delirium. Table 2 provides an overview of all reported outcomes and effects.

Inpatient rehabilitation intervention
Intervention characteristics The systematic review of rehabilitation strategies after hip fracture surgery included patients with any form of dementia. It investigated enhanced interdisciplinary inpatient (and home-based) rehabilitation and care models as well as geriatrician-led inpatient management [36].

Reported outcomes and effects
Smith et al [36] reported on patient outcomes, such as functional performance and mortality, discharge destination and length of hospital stay. The results demonstrate that enhanced interdisciplinary rehabilitation showed a non-signi cant trend towards better functional performance due to the intervention. A greater proportion of participants in the intervention group regained pre-fracture walking levels and better ADL performance at 3 and 12 months but not at 24 months [36]. Mortality did not differ between the groups in two studies at 3 or 12 months [36].
In two studies comparing enhanced interdisciplinary inpatient and home-based rehabilitation with conventional rehabilitation, a difference between the groups in favour of the intervention was measured regarding discharge destination at 3 months but not at 12 months. There was a signi cant difference for people with mild or moderate dementia at 3 months but not at 12 months. No difference was shown for persons with severe dementia [36]. Table 2 provides an overview of all reported outcomes and effects.

Intervention characteristics
Two interventional studies addressed family-centred programmes based on the 'TOP5 strategy'. To enhance person-centred care, staff de ned up to ve personalised care strategies with family caregivers [24,25]. A systematic review focused on interventions (e.g. nurse education sessions) supporting family caregivers of inpatients with dementia. The aim was to assist caregivers in resuming their role as caregivers after discharge [37].

Reported outcomes and effects
Two interventional studies and one systematic review focussing on family-or person-centred programmes reported on a broad range of outcomes, e.g. use of one-to-one nursing care, antipsychotic medication, staff con dence, average length of stay, anxiety, quality of life or family caregiver burden.
One-to-one nursing care used "in cases where regular sta ng levels are not equipped to provide care, leaving the patient, other patients or staff at risk of negative outcomes" [25], signi cantly decreased in two studies [24,25].
One study showed signi cantly improved staff con dence in caring for patients with dementia after introducing the intervention compared to preimplementation. This effect lasted between 6 and 12 months [24]. Another study showed no difference in the level of con dence at three time points and no improvement of staff comfort in engaging with family carers [25]. The systematic review focusing on interventions to prepare caregivers for ful lling their role after hospital discharge showed no signi cant improvement in quality of life, anxiety, depression or family caregiver burden [37]. The family-centred programme investigated by Luxford et al. improved family caregivers' satisfaction regarding communication with staff [24]. Table 3 provides an overview of all reported outcomes and effects.

Use of specially trained nurses and consultation services
Intervention characteristics One study investigated the upskill of registered nurses to so-called "cognition champions" who should develop action plans for practice change [34]. Another study compared a transition programme conducted by advanced practice nurses with the help of 'resource nurses' (upskilled registered nurses) and an augmented standard care programme (research assistants communicating cognitive screening results to staff) [32]. One systematic review focused on interventions providing palliative care consultation services by specialised healthcare professionals for people with advanced dementia with the aim to in uence advance care planning [39].

Reported outcomes and effects
The programme studied by Naylor et al. [32] included an intervention for the transition from hospital to home, led by an advanced practice nurse. The study focused on outcomes regarding rehospitalisation of patients with cognitive impairment. Besides patient outcomes, such as agitation or confusion, Travers et al. [34] also reported on staff outcomes, e.g. staff behaviour and assessment of cognitive impairment. The systematic review by Kelly et al. addressed the behaviour of staff and relatives regarding advanced care planning and health system processes like hospital length of stay or emergency visits [39].
The "cognition champions" intervention led to a signi cant increase in the number of patients assessed for cognitive impairment by means of a standardized tool at the time of hospital admission. The frequency of nurses' informal assessment of patients' cognitive impairment showed no signi cant effect [34]. Furthermore, no signi cant change was observed in nurses' behaviours on acute care hospital wards based on ve pre-de ned behaviours, e.g. explaining an action in easily understandable terms [34]. Regarding patient outcomes, Travers et al. [34] did not observe a signi cant change in patient activity (number of patients doing nothing versus number of patients engaged in some activity), signs of agitation or pain in the "cognition champions" programme. Concerning the outcomes for relatives of people with dementia, palliative care consultation services resulted in improved satisfaction with end-of-life care and in a change in advanced care planning behaviour [39].
Patients receiving a transitional care model intervention led by an advanced practice nurse experienced a longer time to rst rehospitalisation or death than patients in the augmented standard care group or in the resource nurse care programme. This difference was statistically signi cant at 30 and 60 days between augmented standard care and the transitional care model [32]. With regard to system-related processes (e.g. hospital length of stay, emergency visits or hospice referrals), two studies in the systematic review by Kelly et al. showed an improvement, whereas two studies revealed no improvement [39]. Table 3 provides an overview of the outcomes and the effects.
Volunteer programme

Intervention characteristics
The systematic review addressing volunteer-delivered programmes to support people with dementia and/or delirium in acute hospitals included, for example, support with regard to orientation and interaction, mobilisation, provision of non-pharmacological approaches to manage distress or assisting with hydration and nutrition [38].

Reported outcomes and effects
The systematic review reported staff outcomes (level of hope) and a range of patient outcomes like mortality, length of stay, delirium incidence and delirium severity. Regarding staff outcomes, volunteer programmes showed improved levels of hope in volunteers, whereas staff level of hope did not change. Patient outcomes improved regarding delirium incidence, delirium severity and functional status. Furthermore, patients received more analgesics. No improvement was reported for mortality, length of stay, number of falls and residential aged care placement after hospital discharge [38]. Table 3 provides an overview of the outcomes and the effects.
Special care units for people with dementia

Intervention characteristics
The systematic review comparing special care units in general hospitals for patients with dementia to standard care wards included acute care hospital inpatients of any age with any form of dementia [16].

Reported outcomes and effects
Patient outcomes, such as mortality, length of stay, delirium incidence or use of antipsychotic medications, as well as staff outcomes, e.g. documentation of treatment decisions, drug history or discharge plans, were reported. Family caregiver outcomes were strain and psychological well-being. No signi cant differences regarding mortality were found between special care units and standard care wards [16].
Special care units showed non-signi cant improvements in readmission rates. Patients on special care units were more likely to be discharged to their own home than to a care home (although this was not statistically signi cant) [16]. Although patients on special care units were signi cantly often in a positive mood or engaged, no signi cant difference was found regarding the rates of behavioural and psychological symptoms of dementia [16]. Incidence of delirium was slightly but not signi cantly higher on the special care units. New prescriptions of antipsychotic medications were moderately but not signi cantly higher [16]. Furthermore, no signi cant differences were shown for carer strain and carer psychological well-being [16]. Table 3 provides an overview of the outcomes and the effects.

Discussion
This systematic review provides a comprehensive overview of interventional studies and systematic reviews of interventions targeting the care situation of patients with dementia in acute care hospitals. The ndings indicate a broad range of different interventions, outcomes and measurements. We categorised the interventions into eight intervention types: educational programmes, special non-pharmacological interventions, delirium management programmes, inpatient rehabilitation interventions, family-/person-centred programmes, use of specially trained nurses, volunteer programmes and special care units.
Educational programmes were the most commonly reported intervention type among the included studies. Six interventional studies and two systematic reviews (including 17 studies relevant to this review) focused on dementia education for healthcare professionals in the acute hospital setting. Due to high risk of bias in some domains (e.g. no control group and no blinding of outcome assessment), results of the studies on educational programmes should be interpreted cautiously.
Studies addressing educational interventions predominantly reported staff-related outcomes, e.g. knowledge, sense of competence, con dence, beliefs and attitudes. These outcomes signi cantly improved in most of the studies. With regard to more complex staff outcomes, such as a change of behaviour or practice, the majority of studies did not show any improvements. This demonstrates that enhanced knowledge does not necessarily mean better clinical expertise. As Benner et al.
[40] pointed out, it is not possible to address interpersonal responsibilities and situational discernment by means of technique or formal knowledge alone. Furthermore, patient outcomes were reported only in one study, while outcomes related to relatives were not reported at all. Thus, the impact of these interventions for applied practice remains unclear. The results do not demonstrate to what amount healthcare professionals' accumulated knowledge and improved attitudes reached patients with dementia and their relatives. A critical synthesis of the evidence about dementia care education in hospitals pointed out that sustainability of effects was "largely unknown" due to poor follow-up outcomes [41]. The results of our review con rm this statement. Furthermore, Abley et al. [15] demonstrated that staff knowledge and staff con dence remained stable in three hospitals after four months but signi cantly declined in one hospital. Scerri et al. [35] reported that long-term sustainability of educational interventions was unclear. In contrast, Sampson et al. [23] showed improved sense of competence in healthcare staff three months after education. Another study included in our review indicated increased knowledge of healthcare staff after six months [30].
Updating staff education seems important to promote long-term sustainability of the intervention effect. The optimal duration of education updates is still unclear. Future studies should focus on this issue by investigating long-term follow-up of educational interventions in dementia care in the acute hospital setting.
In addition, to ensure that staff knowledge and con dence as well as modi ed practice reach the patients, it is crucial to combine educational interventions with other interventions. Our review provides an overview of such interventions. Three included studies reported on family-/person-centred programmes in which important care principles or care pathways were de ned and discussed with families. It is notable that one study did not investigate family caregiver outcomes at all [25]. Another study described family caregiver outcomes regarding satisfaction with staff communication.
Satisfaction improved due to the implemented programme [24]. However, high risk of bias in some domains have to be considered when interpreting these results (no control group, no blinding, selective outcome reporting (e.g. satisfaction with care, involvement in care, well-being and creating a therapeutic culture) were not su ciently considered in the included primary studies.
The systematic review by Pritchard et al. [37] about interventions to support family caregivers' readiness for discharge home indicated that family caregiver outcomes (e.g. quality of life, anxiety, depression and caregiver burden) did not improve. Reasons for this might be the complexity of the intervention, different in uencing factors (personal characteristics of involved persons, and attitudes of healthcare professionals), and organisational culture [37]. Thus, the complexity of interventions for people with dementia in the acute care hospital should be considered when developing and implementing new programmes. Across all studies, outcomes related to family caregivers were rarely reported. Only two interventional studies and one systematic review investigated caregiver-related outcomes.
A meta-synthesis of qualitative studies shows that relatives of people with dementia experience many negative aspects of care in the hospital.  [2]. Thus, important consequences of a hospital stay for people with dementia are covered by these outcomes. Regarding the clinically relevant outcome of delirium, specially trained nurses were not able to reduce acute confusion [34]. However, volunteer programmes resulted in lower delirium incidence and severity [38]. The delirium management programme for people with dementia reduced delirium severity but not delirium duration [27]. Inpatient rehabilitation interventions showed no improvement regarding delirium incidence [36]. Furthermore, specially trained nurses were not able to reduce agitation [34]. Rates of behavioural symptoms did not improve in special care units for people with dementia [16].
When interpreting these results, the high risk of bias in some domains has to be considered. The programmes were probably not su ciently effective to improve clinically relevant patient outcomes. The di culty of considering the complexity of the intervention and the in uencing factors could be one reason for this. Furthermore, it is questionable whether these studies measured outcomes that could be really in uenced by the intervention and whether they focused on outcomes relevant for patients.
A meta-synthesis of qualitative studies focussed on the perspective of persons with dementia on their acute care hospital stay. Hospital care and the hospital environment caused anxiety, distress, lack of control, worries about the future, distrust, feelings of social exclusion, and uncertainty [50]. These aspects were not addressed in the reported outcomes. Thus, the patient perspective has not been su ciently considered. When developing new interventions in this eld, relevant outcomes should be considered from different perspectives. Given the identi ed aspects for good quality of care for persons with cognitive impairment in acute care hospitals in recent literature [9], it becomes clear that the interventions in our review are related to following aspects: staff knowledge and con dence (educational programmes), family-and person-centredness (family-/person-centred programmes), clinical experts (specially trained nurses) and adequate physical environment (special care units). Furthermore, Naef et al. [9] described caring relationships (addressed in the family-/person-centred care interventions), adequate assessments (addressed in the use of specially trained nurses and educational interventions) and interventions (educational interventions) as well as continuous and collaborative care (addressed in the use of specially trained nurses, family-/person-centred programmes and volunteer programmes). Although different healthcare professionals were involved, interprofessional collaboration was not speci cally addressed in educational interventions. According to Naef [9], supportive clinical leadership and practice culture are important for providing high quality of care for persons with dementia. These aspects were not addressed in the included interventional studies. The existing practice culture with rigid task-oriented routines may have a substantial impact on the effectiveness of the interventions. If adequately educated professionals are not supported in practice, this will negatively in uence the effectiveness of educational interventions. Rigid routines make it impossible to effectively implement person-centred interventions. Previous studies have shown that healthcare professionals nd it particularly di cult to provide adequate care for people with dementia due to the existing practice culture [51]. Thus, without changing the existing practice culture on the whole, the effectiveness of singular interventions aiming to improve person-centred care (e.g. TOP 5 strategy) or special non-pharmacological interventions could be at risk. This underlines the complexity of interventions in this eld. Therefore, future research should consider the complexity and the in uencing factors. It is necessary to question the suitability of common study designs to evaluate complex interventions in the given context. Therefore, it seems reasonable to choose highly exible study designs taking into account different relevant perspectives. A well-planned evaluation including a qualitative part and a process evaluation seems to be essential.

Strengths and limitations
A strength of this review is the broad and comprehensive search with a study selection independently performed by two researchers. With this sensitive approach, we were able to provide a systematic overview of interventional studies in dementia care in the acute hospital setting. Since complex interventions are frequently not evaluated by means of conventional methodological approaches (e.g. randomised controlled trials [43]), we opened our inclusion criteria for all studies implementing an intervention and testing it with quantitative methods.
However, due to excluding qualitative ndings, cohort studies and case studies, we probably were not able to identify the most complex interventions in this eld. The inclusion of different intervention types was restricted by our selection of study designs. Therefore, it was possible to include only those intervention types which were evaluated in the included studies. Furthermore, the included studies are very heterogenous, thereby rendering comparisons di cult and data pooling impossible.
To avoid excluding relevant studies, we decided to include also studies with persons affected by "cognitive impairment" and representing symptoms of dementia without diagnosis because in the acute hospital setting, the dementia diagnosis is not always known. However, inclusion and exclusion of studies concerning this criterion was not always highly selective.

Conclusions
This systematic review of interventions to improve hospital care for people with dementia provides a comprehensive overview of reported interventions, outcomes and effects. Interventions, such as the use of specially trained nurses, family-/person-centred programmes and delirium management programmes, proved to have an in uence on some outcomes related to organisation, staff, family caregiver and patients. Educational interventions were commonly investigated and showed improvements, particularly in staff outcomes, including knowledge and con dence. However, several reported results should be interpreted with caution due to high risk of bias. The evidence is not su cient to declare which interventions are effective in improving care for people with dementia in acute care hospitals.
Future research should focus on developing multicomponent interventions, thereby considering various relevant perspectives. Furthermore, research should report relevant patient and family caregiver outcomes. The high complexity of interventions and different in uencing factors (e.g. organisational culture) should be taken into account in this eld. Moreover, interventions addressing the family or relatives of persons with dementia are needed. Focusing interprofessional collaboration and the entire practice culture is required as well. Availability of data and materials: All data generated or analysed during this study are included in this published article and its supplementary information les.

Abbreviations
Competing interests: The authors declare that they have no competing interests.
Funding: We obtained funding from Nursing Science Foundation Switzerland (Stiftung P egewissenschaft Schweiz). The funding body had no in uence on the study design, the collection and analysis process, the interpretation of the data or the writing of the manuscript.
Authors' contributions: All authors were responsible for the conception of the study design. MK and AZ designed the search strategy. MK and AS conducted the literature search, the selection of the articles and the risk of bias assessment. MK and AS were responsible for data extraction. Data analysis was conducted by MK and discussed with AS, AZ and HM. MK was responsible for writing the rst draft of the paper. All authors substantively edited the manuscript. All authors read and approved the nal manuscript.      The "Dem Del" was a nurse-led intervention with six components: (1) An educational package for nurses and physicians regarding cognitive impairment and delirium (professions separated from each other). Support was provided by specially trained ward nurses and case conferences by the study team.
(2) On admission, nurses screened all patients aged 70 and over for cognitive impairment. Signs of cognitive impairment were discussed with the physicians within a short period of time.
(3) Based on the nurses' screening results, the physicians prescribed or revised PRN medication for delirium according to the prepared recommendation of the hospital.
(4) A priority list of evidence-based interdisciplinary measures to prevent and treat delirium in patients at risk was provided to the staff.   We have only reported on quantitative findings. Data were only extracted for dementia-specific issues, e.g. if a systematic review included studies about delirium interventions, we have not reported these results. x Agitation/Aggression # = significant improvement " = no significant effect or mixed effects (significant on only single items, improvement described but not significant) $ = significant decline #g = significant improvement in some but not all of the included studies (concerning SR) or significant improvement just for one investigated ta X = Outcome was assessed, but significance was not described / =Determination of the effect on this outcome was not possible because of imprecision in the results