Effectiveness of interventions on feeding difficulties among people with dementia: A systematic review and meta‐analysis

This systematic review and meta‐analysis aimed to examine non‐pharmacological interventions for helping people with dementia who experience feeding difficulties in order to improve their nutritional status.


| INTRODUCTION
Over 55 million people live with dementia worldwide. This is a staggering figure, made all the more striking as it rises on a daily basis, with forecasts that this may reach 78 million by 2030. 1 Individuals with dementia have a gradual, irreversible deterioration of cognitive functioning and behavioural abilities which eventually affect a person's abilities to perform everyday activities independently. 2 The actual prevalence of eating/feeding difficulties among residents with dementia is high and ranges from 41%, 3 45%, 4 60% 5 to 61%. 6,7 Feeding difficulties lead to reduced food and fluid consumption, unintentional weight loss and, ultimately, malnutrition. 7,8 Studies report that at least 25% of residents with dementia in residential aged care facilities (RACFs) were malnourished while up to 75% were at risk of malnutrition. 9 Another study indicated that only 9% of the residents included in that research had a healthy nutritional condition. 10 Malnutrition is associated with poor quality of life, high risk of pressure injuries, exacerbation of ill health, increased hospital admissions 11 and a high mortality rate. 12 Various definitions of eating and feeding issues among people with dementia have been developed. The term 'feeding', according to Siebens et al., 13 specifically describes the capability of moving food from a plate to the mouth, then swallowing. In contrast, 'eating' generally indicates the quality and quantity of food a person chooses to consume. 14 Watson 14 used the term 'difficulty' to describe any feeding-related behaviour which can result in a reduction in food intake. Chang and Roberts 2 stated that issues with feeding arise in the interaction between people with dementia and the staff assisting them, highlighting the challenges in feeding assistance.
Eating behaviour can change during ageing due to physiological, psychological and social changes. Changes in dietary behaviour are associated with healthy ageing, as well as with age-associated dementia. Individual factors influencing eating behaviour and food choice are based on both physiological (e.g. hunger, satiety and any innate preference for sweet foods) and psychological processes (e.g. learned food preferences, knowledge, motivations, attitudes, values, personality traits, cognitive processes and self-regulation). 15 Eating and feeding difficulties are multidimensional phenomena influenced by different factors. 16 Moreover, people with dementia experience neurodegeneration that induces increased difficulties in communicating their needs and increases their vulnerability to stressors. 17 Those changes not only manifest challenging behaviours at mealtimes but also make feeding more challenging for staff. Secondly, the dining environment and mealtime behaviours may also influence food intake. 18 The nature of the dining environment can stimulate physical, social, cultural and sensory aspects that can potentially invoke an individual's emotional response or arouse physical reactions. [19][20][21][22] In addition, environmental stimulation that is tailored to fit one's personal or cultural needs can improve their eating performance. 22 Adversely, an unsupportive dining environment manifests unmet needs and contributes to feeding difficulties in people with dementia. 23,24 Therefore, 'mealtime difficulties' is conceptualised as a theoretical term that encompasses eating and feeding behaviours, aversive eating and implicit environmental, interpersonal and sociocultural components. 25 Weight loss and malnutrition are important quality indexes in RACFs. 11,26 In order to counteract feeding difficulties in people with dementia, effective interventions that promote adequate food and fluid consumption are needed. 27 A study by Aukner et al. 28 suggested that effective interventions should focus on impaired cognitive function, nutritional intake, staff training and environmental modifications. This recommendation is in line with the Australian Aged Care Quality Standards K E Y W O R D S dementia, eating, education, feeding behavior, professional

Practice Impact
Direct feeding assistance and indirect interventions may have a positive impact on eating behaviours and higher food intake in people living with dementia.

Policy Impact
Measuring feeding difficulty using a validated tool and a better understanding of the effects of interventions are fundamental for developing evidence-based practice guidelines for managing feeding difficulty in residential aged care facilities (RACFs).
(ACQS), which highlights menu design and review, meal planning and preparation, mealtime experience and environment, assistance with feeding/drinking and mealtime support options which includes policies and procedures. 29 A growing body of evidence examined the effectiveness of mealtime interventions either from single or multifaceted aspects. 25,[30][31][32][33][34][35] Recent studies also further examined the effects of oral nutritional supplements on improving the nutritional status of people with dementia with mealtime difficulties. 36 Although these reviews mainly focused on the level of food and fluid consumption and/or nutritional status as primary outcome measurements, the effects on eating, feeding or mealtime difficulties were not discussed.
People with dementia in RACFs often demonstrate behavioural and psychosocial symptoms during mealtimes that are perceived by staff as challenging behavioural responses, which have been demonstrated to increase the work pressure and strain on staff. 37 Modifications to the mealtime environment such as group conversation, food service and music interventions had only weak positive effects on disturbed eating behaviours. However, this study did not examine the degree of eating/feeding difficulties and nutritional outcomes. 24 Staff need to be able to use validated tools to identify changes in behaviour, to understand how to provide appropriate care to assist with food intake and to encourage older persons with dementia who are experiencing difficulty at mealtimes. 38 The literature review conducted by Spencer and colleagues 39 identified several tools that were used for assessing problem feeding behaviours of individuals with dementia. Two validated tools were found, the Edinburgh Feeding Evaluation in Dementia (EdFED) and the Feeding Difficulty Index (FDI), which showed moderate-to-excellent reliability.
There are no available systematic reviews investigating the effectiveness of mealtime interventions on the degree of eating/feeding/mealtime difficulties in people with dementia living in RACFs using the Edinburgh Feeding Evaluation in Dementia (EdFED) 40 or the Feeding Difficulty Index (FDI). 41 The EdFED and FDI were primarily used in the research field to determine the prevalence of feeding difficulty among individuals with dementia in RACFs and health-care settings. 42

| Aims
This systematic review and meta-analysis aimed to assess the effectiveness of interventions in reducing feeding difficulties and improving the nutritional status of individuals with dementia.

| METHODS
The systematic review was conducted following Cochrane Collaboration guidelines, 43 and the protocol for this review was registered with the National Institute for Health Research -PROSPERO registration number: CRD42020201016. The PRISMA guidelines and checklist were used in this systematic review and meta-analysis.

| Search strategy
Five databases, PsycINFO, Medline, PubMed, CINAHL and Cochrane, were searched to identify eligible articles. The following keyword combinations were developed with assistance from the university librarian: (food intake or oral intake or eating or feeding or mealtime) AND (difficulty) AND (dementia or Alzheimer's or cognitive impairment) AND (experimental or quasi-experimental) AND (intervention). The search only included full-text articles published in peer-reviewed journals in the English language. Additionally, other sources involved manually searching reference lists of identified papers and Google Scholar. The full search strategy is presented in Appendix S1.

| Inclusion and exclusion criteria
Only those studies that met the following inclusion criteria were selected: 1. Types of studies: Experimental studies including randomised controlled trials (RCTs) and non-randomised studies with a concurrent comparator group and preand poststudies were eligible. 2. Types of participants: Studies of older people (older than 60 years) with dementia living in RACFs who were experiencing feeding difficulties. No restrictions were placed on participants' stage of dementia (range from mild to severe dementia), gender, ethnicity or other demographic characteristics. 3. Types of interventions: Studies that examined the effectiveness of non-pharmacological interventions on feeding difficulties were eligible. 4. Types of outcome measures: Assessed mealtime difficulties using validated tools, such as the Edinburgh Feeding Evaluation in Dementia (EdFED) 44 or Feeding Difficulty Index (FDI), 45 were considered eligible.
Qualitative papers such as single-case studies, a series of single-case studies, interview studies and observational studies were excluded from this review. Interventions that used food-or drink-based supplements were also excluded.

| Data abstraction and quality assessment
Two independent reviewers extracted data from studies that met the inclusion criteria by employing the Cochrane data extraction form for RCT and non-RCTs. The data extraction form was completed using the Cochrane guidance for process evaluations. Two reviewers extracted the following information: title, authors, year of publication, year of intervention, country, study design, sample size, sample characteristics, intervention (timing and duration), comparator/control, outcome measures and results. Concurrently, the Cochrane riskof-bias tool for randomised trials 46

and the Risk Of Bias
In Non-randomised Studies of Interventions (ROBINS-I) tool 47 were used to assess for the risk of bias in the RCT and non-RCT studies. Risk estimates were calculated as odds ratios (OR) with 95% confidence intervals (95% CI). Any discrepancies were resolved by a consensus-based decision, or if necessary, by a discussion with a third reviewer.

| Data synthesis
Included studies were classified by intervention type for meta-analysis and narrative synthesis. Data analysis was conducted using the Cochrane Review Manager (RevMan 5.4) software if two or more studies under a specific outcome were available and suitably comparable. Heterogeneity was quantified using I 2 . 44 The randomeffect model was used if the I 2 value was higher than 50%. The confidence interval was set at 95%, and the level of significance was set at less than 5%. Where insufficient data were reported, attempts were made to contact authors by email requesting data and clarification.

| Study selection
The initial search identified 1038 articles in total. After de-duplication, 918 articles were screened with title and abstract, of which 871 full-text articles were excluded with reasons. A total of 47 papers were retrieved for full assessment, while 40 articles were removed because they had not assessed feeding difficulty levels as outcome measurements. Seven articles were included for review of which four were excluded for meta-analysis due to insufficient data reported ( Figure 1). Thus, we included seven articles in the systematic review and three in the meta-analysis.

| Study characteristics
Participant demographics, interventions, results and outcome assessments are detailed in Table 1. Two studies originated from the USA, 47 three from Taiwan 48-50 and two were from China. 51,52 All studies were conducted in nursing homes or dementia special care units in RACFs. All studies used EdFED to measure feeding difficulties. Experimental studies using FDI were not available. The type of dementia was not always specified but only two studies included participants with Alzheimer-type dementia (AD) and the rest of the studies included participants with various dementia types. Five studies [48][49][50][51][52] involved mixed-stage dementia from mild to moderate, and two studies 47,53 recruited people with severe dementia. Four studies 47,50,51,53 excluded potential participants if they had medical conditions impacting their swallowing capabilities such as traumatic brain injury or tumours of the head, neck or oesophagus.

| Intervention characteristics
Included interventions varied in length, ranging from several days to 6 months. The considered interventions could be generally categorised into three groups: eating training for people with dementia (n = 4); staff training (n = 1); and feeding assistance and support (n = 2). Eating training interventions involved hand exercises, 52 spaced retrieval (SR), Montessori-based activities or a combination of SR and Montessori interventions. [48][49][50] Spaced retrieval training is a method that can improve learning and information retention by recalling details or instructions over increasing time intervals. Montessori-based interventions stress task coordination using sensory stimulations and emphasise the practices of existing skills with tasks progressing from the concrete to the abstract. The staff training intervention was web-based dementia feeding skills training, with coaching. 53 Feeding assistance studies assessed the effect of hand techniques on feeding behaviours. 47 The multicomponent interventions involved preparation of residents, environment, food and utensils, appropriate assistance during mealtime, psychological care and nursing care postmeal. 51

| Outcome measurements
Edinburgh Feeding Evaluation in Dementia scale was used to measure feeding behaviours to scale the level of feeding difficulty in all included studies. Self-feeding time was assessed in three studies where the articles measured the time it took for people with dementia to eat food by themselves. Two studies assessed the time spent providing feeding assistance. All studies also reported nutritional outcomes. Food intake, quantified as either weight or portion of food consumed as calories, was the most common nutritional outcome assessed, followed by body weight, a nutritional assessment tool or biochemical markers.

| Study quality
Among the studies retrieved, two were RCTs, one was a crossover study, two were quasi-experimental studies and two had an experiment design without control groups. None of the included RCTs met all the criteria outlined by the Cochrane risk-of-bias tool for randomised trials. Eligibility criteria were clearly specified and met in all of the studies. The majority of the studies appeared to  The order of the superscript alphabets is determined by the alphabetical order of the author surnames, as well as the type of training intervention being compared. † selectively report primary outcome data. All studies had small samples, and none used power-size calculations.

| Studies examining the effects of eating training for people with dementia on level of feeding difficulty
Four studies 48-51 examined the effect of eating training programs for people with dementia. Three interventions for resident training were Montessori-based activities, spaced retrieval and a hand exercise program. [48][49][50] Two studies conducted in Taiwan implemented Montessoribased activities related to eating, including hand-eye coordination, scooping, pouring, squeezing and matching. 48,49 Both studies applied the intervention for at least 30 minutes per session, three times per week and over 8 weeks. Lin et al. 49 reported that the Montessori group had a significant reduction in the EdFED score, with mean difference −1.57 ± 3.41 (p = 0.01). The mean score of EdFED in the Montessori group was also significantly less than the control group, with a p-value of 0.008. While the study of Lin et al. 48 reported a reduction in mean score of EdFED postintervention, its p-value was not reported. Additionally, Lin et al. found the Montessori intervention had a positive effect on participants' eating behaviour, self-feeding frequency and self-feeding time. Participants spent 3.86 min more on self-feeding after receiving Montessori activities. 49 Similarly, a study by Lin et al. 48 also found that participants in the Montessori group had an increase in self-feeding time. The average self-feeding time in the Montessori group was 6.21 min longer than the control group, with a p-value <0.05. Lin et al., 48 meanwhile, observed that the intervention group increased the amount of meal consumption postintervention. Despite these positive effects, nutritional status and weight did not improve. Both studies found limited effects brought by verbal assistance, physical assistance and feeding by staff.
Spaced retrieval intervention was implemented in the study of Lin et al. 48 and found a reduction in the EdFED score postintervention, but a p-value was not reported. The study found participants receiving space retrieval had a significant increase in the average score of nutritional status with p < 0.01. However, there was no significant change in weight postintervention. This intervention was further combined with Montessori-based activities in a study by Wu et al. 50 Participants were allocated into two groups that received standardised and individualised training sessions of SR combined with Montessori-based activities. The study reported that both intervention groups had a significant decrease in EdFED scores by 0.39 (standardised group) and 0.30 (individualised group) compared with the control group. Positive effects on body weight and meal consumption in both intervention groups were observed in the posttest. The lasting effects were measured. The meal consumption for the standardised and individualised groups increased significantly by 6% (p < 0.001) and 3% (p = 0.07), respectively, when compared with the meal consumption variable of the control group 6 months after the interventions ceased. Body weight also increased significantly in the standardised and individualised groups by 0.99 kg (p < 0.001) and 0.72 kg (p = 0.001), respectively, when compared with the control group at the 6-month follow-up. 50 Three experimental studies were entered into a metaanalysis. [48][49][50] The interventions included Montessori-based activities or the addition of spaced retrieval. In these studies, an intervention was administered for 8 weeks in populations generalisable in health status, health-care setting and age. The results (see forest plots in Figure 2A) indicated a significant effect of Montessori-based activities on feeding difficulty (EdFED), with a weighted mean difference of −1.36 (95% CI: −1.84 to −0.89, p < 0.001). Montessori-based activities were also shown to have a positive effect on self-feeding time; weighted mean difference was 10.58 (95% CI: 5.78 to 15.38, p < 0.001) (see forest plot in Figure 2C). The pooled results of SR interventions showed a positive effect on feeding difficulty (EdFED) (MD: −1.53, 95% CI: −2.06 to −1.00, p < 0.001) (see forest plot in Figure 2B) but no significant effect on the amount of food consumed (MD: 1.78, 95% CI: −3.97 to 7.52, p = 0.54) (see Figure 2).

| Studies examining the effects of feeding assistance and support on the level of feeding difficulty
Two studies examined the effect of feeding assistance and support from direct care staff. 51,53 Staff applied three different types of feeding techniques, namely: Direct Hand, Over Hand, and Under Hand. The Direct Hand technique specifies that the staff member holds the cup or cutlery to give fluids or food personally to the resident without any direct action or involvement from the latter. The Over Hand technique has the staff placing their hand on top of the resident's hand to guide, assist or support them with eating. Lastly, the Under Hand technique involves the staff holding the cup or cutlery and positioning their hand under that of the patient or resident to assist with eating. 47 Outcome measures were time spent providing assistance, food intake and EdFED. The outcomes showed that techniques in hand feeding had a significant effect on feeding behaviours per meal, measured by EdFED (p = 0.02). The mean score for the Over Hand technique (8.3) implemented per meal was higher than Direct Hand (8.0, p = 0.04) and Under Hand (7.7, p = 0.001). This was interpreted as showing that feeding behaviours were more frequent with the Over Hand technique. Direct Hand and Under Hand techniques were found to have higher food consumption than Over Hand. These three techniques did not have an effect on the time spent providing feeding assistance.
Another study reported a multicomponent intervention, involving preparation of residents, environment, food and utensils, appropriate assistance during mealtime, psychological care and nursing care postmeal (e.g. oral care, handwashing and position). 51 The outcomes showed that the multicomponent intervention significantly reduced the EdFED score (p < 0.001), increased upper arm circumference and skinfold thickness (p < 0.001) and increased the level of serum albumin (p < 0.001). Despite the significant outcomes, there was no control group for comparison.

| Studies examining the effects of staff training on the level of feeding difficulty
One study explored the effectiveness of a staff training program on feeding difficulty, food consumption and mealtime length. 53 The training content consisted of strategies for dealing with circumstances around feeding difficulties and hand-feeding techniques. The results revealed that residents increased the amount of food intake and the time of eating the meal was extended while they received feeding assistance from trained staff. There was no effect on reducing the EdFED score. The effect of hand-feeding techniques was not measured in this study.

| DISCUSSION
The main assessed outcome was the level of feeding difficulty. The review suggests that it is possible to reduce feeding difficulty and/or improve feeding behaviours during mealtimes through a variety of interventions. Direct training for people with dementia resulted in positive findings. [48][49][50]52 In addition, spaced retrieval, Montessoribased activities and hand exercises helped people with dementia to initiate eating and the time of automatous self-feeding, and promoted eating independence.  In addition, of the included studies, five mentioned the degree of dementia of the participating residents ranging from mild to severe, and two studies did not mention the dementia stage of the residents concerned. The residents at each mealtime were not identified, which meant we could not control for the impact of the dementia stage or severity of symptoms on care practices or interactions between staff and residents. This should be addressed in future studies, particularly given that cognitive ability has been found to influence a resident's level of mealtime engagement and eating ability. 54 Food consumption is directly related to nutritional status. The evidence suggested that residents receiving spaced retrieval, a Montessori intervention or hand exercises did not increase the amount of food consumed. Body weight, BMI and nutritional status scores were not affected postintervention. Unexpected hospitalisations and the type of food provided during the experiment may be possible reasons for this. 48,49 However, a combination of spaced retrieval and Montessori activities, regardless of whether they were standardised or individualised in design, increased meal consumption as well as weight gain. 50 These findings resulted in an inconclusive outcome that can be generalised by various explanations, such as the type of meal and length/intensity of the intervention. This requires more studies with rigorous designs to retest the effectiveness of food consumption and nutritional status.
Direct feeding assistance and indirect interventions such as resident preparation, environmental support and postmeal care resulted in positive effects on feeding behaviours, which resulted in lower EdFED scores and higher food intake. 47,51 One review pointed out that a supportive dining environment plays an important role in enhancing a positive mealtime experience. 18 One study also found that people with cognitive impairment and their carers have to tackle mealtimes despite this lack of evidence. 55 Despite having a positive outcome, only three studies were included in the meta-analysis and there was no robust evidence to explore the effect due to the limited number of studies with similar interventions, insufficient data and study design lacking control groups.
In one study, training/education programs demonstrated moderate evidence to increase eating time and decrease feeding difficulty. 56 However, in this study, staff training programs had no effect on reducing feeding difficulties. 53 This result was similar to the study of Chang and Lin, 57 which reported that the EdFED score in their study was not reduced after the application of a feeding skills training program for nursing staff. Regardless, residents' food consumption and mealtime length increased when they were assisted by trained staff, 53 suggesting that trained staff spent more time assisting residents with their meals. This result may be associated with a positive change in staff attitude and knowledge after they had received education and training. 57,58 Therefore, overall, while the training program for staff had a limited effect on managing the residents' feeding behaviours, it still had a positive influence on staff knowledge and skills to support residents during mealtimes.

| Strengths and limitations
This systematic review is a pioneer in examining the significance of interventions on feeding difficulty in people with dementia. The researchers implemented meta-analyses inclusive of studies with suitably similar participants, interventions and outcomes, although this was only appropriate for Montessori-based and spaced retrieval studies. The non-pharmacological interventions included in this review may help improve self-feeding ability and increase food consumption in people with dementia. Perhaps these outcomes may be specifically beneficial for persons with dementia who are attempting to improve their nutrition status. Considering that the data in this review are derived from experimental studies, information on the long-term effects of these interventions on food intake and level of feeding difficulty is limited.
In addition, while the experiment study baseline measurement was designed to monitor the reliability of data collected prior to the intervention, variations in the culture of RACFs, the nature of the resident, the status of dementia and staff knowledge and training were difficult to control. Future studies may consider more stringent experimental controls.
Although a comprehensive search strategy was attempted, it is possible that some relevant studies were overlooked due to some restrictive eligibility criteria. The principal constraint of this review is the limited quantity of included studies for narrative synthesis and minimal generated evidence. Further, only three studies were extracted and available for data pooling for the meta-analysis. The majority of the studies were either too small in sample size or too short in time to detect any significant changes in nutritional outcomes.
A few interventions discussed, including hand exercise, feeding support, staff education and multicomponent intervention, were only reported individually. These were unable to present collective evidence and it was also difficult to compare statistically the effects between different interventions. More research is needed to refine assessment measures, develop interventions and portray the mechanisms by which they affect feeding difficulties.
Although we could not report definitive evidence on interventions for improving dementia feeding difficulties and nutritional status, reviewing all available studies and suggesting evidence on possible effective interventions could hold merit for the direction of future research.

| CONCLUSIONS
None of the included RCTs met all the criteria outlined by the Cochrane risk-of-bias tool for randomised trials. Considering the small number of studies included, incorporating multicomponent interventions into everyday practice is a novel approach to maintaining nutritional status and providing a positive mealtime experience among people with dementia in RACFs. This may reduce the level of frustration experienced by staff while providing feeding assistance, and may increase the quality of life for people living with dementia. However, high-quality trials are required to establish the full efficacy of such interventions as well as the cost implications.