Systematic review of the current psychosocial interventions for people with moderate to severe dementia

Dementia, a global epidemic, currently affects 40 to 50 million individuals worldwide. There are currently limited effective treatments for moderate to severe dementia, and most treatments focus on reducing symptoms rather than improving positive factors. It is unclear if improvements are not possible due to disease severity. This review examines the efficacy of the current psychosocial interventions for people with moderate to severe dementia, focusing on improving cognition and quality of life to evaluate what treatments are working and whether improvements are possible.


| INTRODUCTION
Dementia, a global epidemic, currently affects 50 million individuals worldwide. 1 While 12.5% of people with Alzheimer's disease have severe dementia, there is currently no cure and limited effective treatments for moderate to severe dementia. 2 Only five drugs are approved by the Food and Drug Administration to reduce its symptoms. 3 Therefore, psychosocial interventions are frequently used to promote cognition and quality of life (QoL), because of the various adverse effects and contraindication of pharmacological treatments for dementia. [4][5][6] Despite the increase in psychosocial interventions in recent years, there is little guidance and understanding of how to treat moderate to severe dementia. The National Institute for Health and Care Excellence (NICE) recommends cognitive stimulation therapy (CST), one of the most established psychosocial interventions, for improving cognitive function for people with mild to moderate dementia because its effects are comparable to antidementia drugs. 7,8 While NICE does ask professionals to consider using multisensory stimulation (MSS) for people with moderate to severe dementia, there is insufficient evidence for an official recommendation. It is unclear if and how MSS and other current psychosocial interventions are effective for people with moderate to severe dementia. Past systematic reviews of dementia have largely been intervention focused rather than stage-specific. 9 For example, Cochrane reviews of exercise, aromatherapy, reminiscence therapy (RT), music therapy, and MSS interventions. [10][11][12][13][14] Even when systematic reviews were stage-specific, the emphasis was mostly on mild cognitive impairment and mild to moderate dementia. [15][16][17][18] A recent review assessed nonpharmacological interventions (NPIs) for moderate to severe dementia; yet, they did not breakdown the different types of NPIs, so we do not know what kind was effective. 19 While psychosocial interventions and NPI are often used synonymously, NPI do not explain what an intervention is, just what it is not, and has a strong sense of symptom management. Psychosocial interventions focus on improvement and can go beyond basic problem-management. Boote et al. 20 conducted a review on psychosocial interventions-almost 20 years ago. Both reviews included various study designs, publication types, which were low to moderate in quality with inconsistent evidence. An up-to-date systematic review on psychosocial treatments with a robust methodology is warranted to indicate what interventions are effective for the target population.
Previous literature on moderate to severe dementia focuses on reducing problem behaviors rather than examining if positive factors, cognitive function and QoL, can be maintained or improved, 21 leading to an implicit assumption that improvements might not be possible for people with moderate to severe dementia. Cognitive function and QoL are, however, the main outcomes of CST, meaning that these improvements are possible for earlier stages of dementia. For later stages, some argue that the focus should be on a dignified life instead of enhancing cognitive function or QoL. However, improved cognition, for example, can honor people with dementia (PwD) by enhancing their functional abilities, communication, interpersonal relationships, which allows for better QoL. Past studies have shown that changes in cognition mediate the effects of QoL. 22 Theories on brain plasticity and the biopsychosocial model have also suggested disease outcomes to be modifiable intrinsically (biological factors) and extrinsically (social factors). 23 Cognition is a defining attribute of dementia, for it is a major neurocognitive disorder characterized by a decline in one or more of the six cognitive domains in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition. 24 Interventions have aimed to improve cognition and prolong disease progression. Commonly used interventions include computerized cognitive training, music therapy, CST. 8,25,26 However, most of these interventions, like CSTrecommended by NICE 7 for improving cognition-target people with mild to moderate dementia.
QoL is also a critical outcome as it is an essential aspect of health.
It is defined as "an individual's perception of their position in life in the context of culture and value systems in which they live and in relation to their goals, expectations, standards and concerns." 27 According to the WHO's constitution, health is "a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity." 28 Improving well-being is just as important as enhancing cognition, and QoL is a core outcome in dementia research. 29 As a result, a number of past interventions for mild to moderate dementia, such as cognitive stimulation, cognitive training, and cognitive rehabilitation, 17,30 have measured QoL.
This review aimed to evaluate whether the current psychosocial interventions can improve or maintain cognition and QoL for people with moderate to severe dementia, specifically considering (1) the types of effective interventions for the affected population; (2) and their significant outcomes with their respective effect sizes. We also aimed to evaluate other significant outcomes in the selected psychosocial interventions, so we could provide a comprehensive view of what is effective for the affected population.

| Search strategies
The systematic review was performed in a prespecified protocol, in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) transport reporting of systematic reviews recommendations. 31  We used keywords and database-specific subject headings in the title and abstract to perform the search. The MeSH terms varied between databases. In some databases, such as Web of Science, MeSH terms were not available. The search terms were developed in conjunction with a university research subject librarian. The keywords were reviewed by three authors. The first search was conducted on 4 January 2020 and replicated on 1 April 2020, with no relevant additional or relevant publications found.

| Inclusion criteria
� Study design: RCTs published in an English peer-reviewed journal within the period of 2000-2020.
� Participants: people with a diagnosis of moderate to severe dementia, aged 18 years of older. If a study did not specify the stage of dementia, then the participant's mean Mini-Mental State Examination (MMSE) score plus two standard deviations ≤ 20 was used. This cut-off score was selected, because according to the NICE guidelines, people with moderate and severe dementia have an MMSE score of 10-20 and <10, respectively. The Clinical Dementia Rating (CDR) score of two or more, and Global Deterioration Scale of five or more were also accepted.
� Interventions: psychosocial interventions were defined as physical, cognitive, or social activities that may maintain or improve "functioning, interpersonal relationships and well-being in people with dementia." 32 � Comparisons: any control group; for example, treatment as usual. 33 � Outcome(s): a cognitive or/and QoL outcome, with corresponding between-group p values, included as a minimum (Table 1).

| Other outcomes and missing information
Significant effects of other outcome measures, such as depressive symptoms, agitation, activities of daily living (ADL), and other behavioral and psychiatric symptoms, were also included. We included these outcomes in order to evaluate all positive effects of current psychosocial interventions. Authors were contacted for missing data, such as details of randomization, means, and standard deviation.

| Data extraction
E. Hui examined the titles and abstracts against the inclusion criteria.
Subsequently, the full-text reports were screened to confirm if the articles met the inclusion requirements, especially in terms of outcome measures. To resolve questions on eligibility where necessary, an additional study author, A. Spector, was sought. The reasons for excluding studies were recorded.

| Quality assessment
The methodological quality of the studies was independently assessed by two reviewers (E. Hui  differences-and evaluation of statistical information sufficiencyintention-to-treat analysis (ITT), between-group and point measures. Each criterion was rated as "1" for yes, "0" for no, and "unclear" where there was insufficient detail reported in the study, and the corresponding authors were contacted. If the authors did not respond within seven working days, then a "0" was given. The scale has a maximum score of 10, where five is the cut-off point.
High-quality papers had a score greater or equal to 5.
We referenced the CONSORT guidelines for further information, such as power analysis, effect size, method of statistical analysis, randomization sequence generation, and allocation concealment mechanism. 35 Records identified through database searching (n = 4191)

Eligibility Identification
Additional records identified through other sources (n = 2) Records after duplicates removed (n = 3195) Records screened (n = 3195)

Studies included in qualitative synthesis (n = 14)
Full-text articles assessed for eligibility (n = 74)

| Description of selected studies
The systematic search produced 4191 articles, with two identified additionally through bibliography reviews. Only peer-reviewed articles, written in English between 2000 and 2020 were included. Duplicates, 997 studies, were removed among the databases. After the title and abstract screening of 3195 articles, 74 full-text articles were reviewed. Sixty studies were excluded for including the incorrect disease (n = 1), incorrect population (n = 9), incorrect intervention (n = 13), incorrect publication type (n = 3), incorrect study design (n = 7), incorrect outcomes (n = 19), and missing data (n = 8).
Fourteen RCTs were included in this review. Please see Figure 1 for the PRISMA flow diagram.

| Participants and settings
The included studies contained a total of 1161 participants, ranging from 21 to 189 participants per study (

| Methodological quality of the included studies
The overall quality of the studies was moderate (see Table 2). The scores ranged from four to seven on the PEDro scale, with 5.86 as the average score. The cut-off score of the 10-point scale was 5, where a score below five denoted low quality. 37 Ten studies had a low risk of bias, scoring above 5; three had moderate risk, scoring 5; and one was high risk. The most frequent methodological shortcomings were lack of blinding and high attrition rate.
All studies met the standard criteria of PEDro, as they specified the eligibility criteria and use of randomization. However, only eight studies had concealed allocation. Four studies' control and intervention groups were significantly different at baseline.
None of the studies were double-blinded, except one on aromatherapy. 38 However, the lack of double blinding is not indicative of the studies' quality. Due to the nature of most psychosocial interventions, the therapists could not be blinded because they were delivering the treatment. As a result, only single-blind was possible.
Eight RCTs blinded the participant or assessor.
Over half of the included studies had an 85% attrition rate in at least one key outcome. 33,[38][39][40][41][42] All but six studies included an ITT analysis. 33,38,40,[43][44][45] All included between-group statistical measures with their respective point measures except for two studies. 38,46 To assess other potential biases, additional information on poweranalysis, effect size, and follow-up were obtained (see Table 2). Out of the 14 included studies, 5 included a power analysis, and 8 were sufficiently powered. While all studies reported effect sizes, only 11 studies indicated a specific p value, where five were accompanied by a confidence interval.
Five studies included follow-up assessments indicating that cognitive and QoL effects were maintained for 3-6 months, meaning that once improvements are present, the positive effects can endure for some time. 33,41,42,45,47 However, since most of the studies did not have follow-ups longer than 6 months, it is unclear if and for how long those studies' intervention effects were maintained.

| Multisensory stimulation
Many types of MSS interventions were included in this review, such as Snoezelen rooms, aromatherapy, robotic pets, and music therapy (see Table 2 for intervention details). Only aromatherapy, however, reported significant improvements in QoL (p = 0.01) postintervention. 38 This study was sufficiently powered and is of high quality (PEDro = 7). The dosage was 15 min, twice a day for 4 weeks.
Robotic pets also had significant effects for QoL in its subgroup analysis of people with moderate to severe dementia (p = 0.01).
While this study was of high quality, the sample size of the subgroup was too small for results to be conclusive (n = 24).

| Exercise
Two exercise interventions were included in this review. 36,44 Other studies with exercise had multiple components, involving other types of activities, so they were included as multicomponent programs.
Both studies found significant improvements in cognition. The statistical significance can be attributed to the slight improvement in cognition postintervention in the experimental group and the decline in the control group. 36,44 Both interventions were of moderate quality (PEDro = 6) and included aerobic exercises, but they differed in terms of dosages and sample sizes (see Table 2). Unlike the cycling intervention, 36 the walking program was insufficiently powered. 44 The cycling program was 15 min daily for 15 months, and the walking program was 30 min daily for 6 months.

| Multicomponent
Multicomponent programs refer to interventions with more than one type of program in this review. Henskens et al. 47 for example, had exercise as well as ADL in the experimental group. Out of the five selected studies, one reported improvement in cognitive function that was statistically significant. 40 This study included a variety of programs in its intervention: aerobic exercise, memory games, and music therapy. Its quality was the lowest among the ones with significant results (PEDro = 5), and its sample size was small (N = 36).
The music and movement intervention were moderate in quality (PEDro = 6) and was sufficiently powered; however, only a positive trend in cognitive improvement was reported (p = 0.055). 46

| Reminiscence
The only RT study included in this review enhanced the QoL (p < 0.01). 42 This paper was sufficiently powered (N = 135) and had high quality ratings.

| Other outcomes
Apart from cognitive and QoL outcomes, we also presented other outcomes in Table 2. Many other outcomes were investigated, but the five main ones with statistically significant improvements were: neuropsychiatric symptoms, agitation, depressive symptoms, mobility, and ADL. Several studies reported significant positive effects on neuropsychiatric symptoms and used the same measure-Neuropsychiatric Inventory. 36,38,41 In terms of agitation, two MSS interventions had a significant impact. 38,41 Another MSS study, Raglio et al. 49 and an exercise program, Cancela et al. 36 found significant differences in the pretest and posttest scores, where the intervention group had fewer depressive symptoms than the control. Two studies with aerobic exercise components were effective in improving mobility. 36,40 Exercise was also associated with significant improvements in ADL, meaning that people with moderate to severe dementia became better at performing routine activities after the intervention. 36,44 Other outcomes, such as emotions and social engagement, were also explored. The use of RT was associated with enhanced social engagement. 42 Robotic animals reduced and increased people with moderate to severe dementia's negative and positive emotions, respectively. 43 Details of the results are in Table 2.

| Impact on outcomes
Aside from the design and content of an intervention, the power sufficiency and quality of RCTs also impacted study outcomes. Results from the five studies with moderately high quality and sufficient power were less prone to type I error. 36,38,41,42,49 It was unclear if dosage influenced the intervention outcomes as the studies were too heterogenous for comparisons to be meaningful.

| DISCUSSION
This systematic review was designed to evaluate the efficacy of the current psychosocial treatments for moderate to severe dementia, focusing on cognition and QoL to see whether improvements in these areas are possible. We believe this is the most comprehensive review to date that emphasizes these positive aspects for this population while also taking other significant outcomes into account. Comparing to the most recent review, which included three studies on cognition and one on QoL, we provided a more comprehensive and updated evaluation with 10 RCTs on cognition and 4 on QoL. 19 In this review, aromatherapy and RT showed the strongest evidence in improving QoL. There was some evidence a multicomponent study also advanced QoL, and aerobic exercise enhanced cognition. 36,40,42,44 However, the aerobic exercise study was insufficiently powered, and the multicomponent program was poor in quality. 40,42,44 Overall, the included RCTs varied greatly in factors such as intervention duration, methodological quality, study settings, and outcomes, making it difficult to draw firm conclusions.

| Effectiveness of intervention based on intervention types
Results suggest that stimulating multiple senses simultaneously exerts positive effects on cognition and QoL for people with moderate to severe dementia, because of the statistically significant improvements in one aromatherapy, and one multicomponent intervention. 38,40 The multicomponent intervention included many types of programs in its experimental group-music therapy, memory games, and aerobic exercises. As a result, it was unclear what activity was effective. This study also had a small sample size and was of low quality. Out of all the MSS studies, aromatherapy had the most solid HUI ET AL.
-13 evidence in improving QoL; it had sufficient power and the highest quality rating in this review. Our findings align with the previous Cochrane systematic reviews on aromatherapy, aerobic exercises, and music therapy, indicating positive trends in the use of MSS for QoL and cognition. [10][11][12][13] Results from other Snoezelen, robotic animals, and multicomponent intervention studies, however, did not have statistical significances-only positive trends were observed. 33,39,43,46 Our findings support past studies, where Snoezelen provided some evidence of the immediate improvements in cognition in small studies.
However, the promising effects were not observed between groups and in systematic reviews with methodologically weak studies. 50 There was insufficient evidence to suggest that these interventions-Snoezelen, robotic animals, music with movement-were beneficial for moderate to severe dementia.
We found RT to be advantageous for people with moderate to severe dementia, not just dementia in general. While previous studies on the effects of RT varied, there was some evidence that it could improve QoL, cognition, communication, and mood. 12 Most past studies were not stage-specific, and if they were, they focused on the mild to moderate population; for example, the majority of the studies in the most recent Cochrane review of RT were on mild to moderate dementia. 12 Knowing that RT is applicable to moderate to severe dementia is imperative for the development of future interventions.
Two exercise interventions had positive findings for cognitive outcomes. 36,44 However, the statistical significance was mainly due to the major decline in the control group's cognitive abilities for one study, 36 and the other RCT was insufficiently powered. 44 Other interventions with exercise components did not exhibit the same positive effect. 45 Whether exercise enhanced cognition is unclear.
There is conflicting evidence on exercise's role in cognition in our findings and previous studies. Walking or cycling for 30 min a day might not be physically possible for some with severe dementia due to disease progression. 51 Previous studies have suggested exercise to enhance physical and cognitive abilities. 10,52 Dementia and Physical Activity, a recent multicentered RCT, however, claimed exercise to be damaging for PwD. Twenty-five adverse events (AE) were reported during this trial. Not only did exercise not improve the QoL, but it could potentially damage cognition. 53 Many other outcomes were investigated in the 14 RCTs (see Table 2). As mentioned previously, the five main areas with significant outcomes were neuropsychiatric symptoms, agitation, depressive symptoms, mobility, and ADL. The positive effects of MSS on agitation support previous reviews. 11,38,41,54 Current practice supports the use of antipsychotic medications for agitation and neuropsychiatric symptoms. 7 Due to the emergence of AE and the high risk of contraindications of pharmacological treatments, effective NPIs would be impactful. 4,5 This provides the rationale for the current study.
While we found significant improvements in depressive symptoms, mobility, and ADL from various interventions in this review, there is conflicting evidence in the existing literature. 36,40,44,46,47,49 Depressive and dementia symptoms can overlap and be hard to distinguish, making positive effects in mood difficult to evaluate. 55 While some exercise interventions appear to increase mobility and improve ADL, others suggest it can be potentially harmful or ineffective. 53 Publications to date have also not provided the mechanism of how exercise influences cognitive function in dementia.

| Limitations
This review was limited by including RCTs with outcome measures that may not be sensitive to people with moderate to severe dementia. While the MMSE is the gold standard for assessing cognitive impairment, whether it is effective in measuring cognitive changes in moderate to severe dementia is unclear. 56 MMSE is not a staging tool, so for studies that did not use an additional staging tool, it is unclear if people with moderate to severe dementia were accurately selected for participation. 36,43 Results of this systematic review were limited by the fact that there is no standard assessment of QoL. Each study used a different instrument, making it hard to compare the effects across studies as they have various rating systems ( Table 2). The validity and reliability of some are questionable. For example, one study used a selfreported test, Self-reported Quality of Life. 42 Since people with moderate to severe dementia often have limited communication and language abilities, self-rating would be challenging; whether responses are accurate is questionable. Only one study used a QoL instrument specifically for moderate to severe dementia. 33 Our systematic review also could not present the most comprehensive data for people with moderate dementia. We excluded one of the most established psychosocial interventions for mild to moderate dementia, CST, because we could not separate the moderate dementia results from the mild. CST studies also did not meet our inclusion criteria, designed to ensure most of the sample have moderate to severe dementia.
Publication bias and including studies with insufficient power, and/or lack of statistical information may also threaten the validity of this review. We did not include any gray literature, and trials with positive findings are more likely to be published. 57 Several studies were insufficiently powered to detect changes in effect sizes. For studies with sufficient power, there was a lack of statistical information. 40,41,43,44,48 While all studies reported between-group pvalues and effect sizes, most p values did not include a confidence interval, and it was hard to compare the effect sizes due to the heterogeneity in study design and outcome measures.

| Future work
Current pharmacological treatments can reduce symptoms of de-

| Conclusion
Results from the present systematic review show aromatherapy and RT to improve QoL, and positive trends in the use of multicomponent programs for QoL and aerobic exercise for cognition. Due to the limited studies per intervention type, methodological weaknesses, and heterogeneity, results must be interpreted with caution.
Since dementia is currently a global epidemic, and the world's aging population is increasing, developing effective treatment is a public health priority. Various types of current psychosocial interventions can enhance the health of people with moderate to severe dementia. However, future research with well-defined and quality outcome measures with a rigorous study design is required before recommendations can be made regarding the use and prescription of MSS, RT, exercise, or multicomponent interventions for people with moderate to severe dementia.