Introduction

Dementia is an impending healthcare crisis that affects more than 55 million individuals worldwide. With the ageing population, this number is expected to reach approximately 78 million by 2030 and 139 million by 2050 [1]. The global financial burden of dementia is estimated to be USD 1.3 trillion and may reach USD 2.8 trillion by 2030. Dementia is a progressively degenerating disease that causes multiple cognitive deterioration, which results in disability, dependence, institutionalisation and mortality. Long-term care of patients with dementia is mostly provided by family caregivers, who provide unpaid or informal attention and play an essential role for the persons they are taking care of and for the overall healthcare system. Family dementia caregivers are more stressed, suffer more serious anxiety and depressive symptoms and are at higher risk of cardiovascular diseases than non-dementia caregivers [2]. These over-burdened caregivers are more likely to give up their caregiver role by institutionalisation of their care recipient [3]. All of these factors are projected to significantly increase the costs of healthcare and long-term care in the future. Therefore, effective interventions should be designed to reduce the negative effects of caregiving on family dementia caregivers.

A recent systematic review and meta-analysis compared the effects of 131 randomised controlled trials (RCTs) for dementia caregivers with community-dwelling care recipients; the result showed that multicomponent interventions had the most significant effect among nonpharmacological interventions on reducing the burdens on family dementia caregivers, decreasing their stress levels and enhancing their subjective well-being [4]. An earlier meta-analysis reported that multicomponent interventions are effective in delaying or preventing the institutionalisation of people with dementia [5]. Despite the favourable outcomes of multicomponent interventions, the effectiveness of individual components is concealed with the black box approach in traditional RCTs. Caregiving support interventions are often complex and consist of multiple components in variable combinations. For example, an intervention may include components, such as education on dementia and caregiving education, self-care skills, behavioural symptom management, stress management and support group. Traditional RCTs have only examined the effectiveness of various combinations of interventions as a package. The effectiveness of individual components of a multicomponent intervention has been rarely tested. As such, studies of broadly conceived family caregiver programmes consisting of different components have produced highly varied results in terms of effect size, which ranges from small to medium [4]. Components and behavioural skills that can effectively alleviate caregiving burden and enhance psychosocial outcomes among dementia family caregivers and to whom they are suitable for remain to be identified.

The multiphase optimisation strategy (MOST) developed by Collins and her colleagues [6] is an innovative approach for examining which components of a multicomponent intervention are effective in achieving a particular outcome, singly and in combination with one another. The value of MOST has been established in recent research in several areas, including cessation of smoking, promotion of physical activity, reducing obesity, reducing alcohol use and unsafe sex and conducting education interventions [7]. However, to our knowledge, MOST has not been applied in gerontology, making our proposed study the first of its kind. MOST consists of three stages: (1) preparation to conduct an optimisation trial, (2) optimisation to reveal what constitutes an optimised intervention and (3) evaluation of the optimised intervention relative to an established intervention in an RCT. In this study, we will focus only on the preparation and optimisation phases due to resource and time constraints.

In the preparation phase, we will use six core psychoeducational components, namely, (1) education on dementia and caregiving, (2) self-care skills, (3) behavioural symptom management, (4) behavioural activation, (5) modified mindfulness-based cognitive therapy and (6) social support. Components 1 to 4 and 6 were adapted from the ‘Resources for Enhancing Alzheimer’s Caregiver Health’ (REACH) intervention [8,9,10], which is one of the most studied multicomponent interventions for informal dementia caregivers. Two multicomponent intervention studies of REACH that considered cultural perspective in caregiving [11] were conducted in Hong Kong and reported positive preliminary findings. In the 2010 study [9], Au adapted and evaluated the multicomponent intervention called Coping with Caregiving, which is one version of REACH; they found that the intervention enhanced caregiving self-efficacy and effective coping strategies. In the 2015 study [10], Cheung conducted an implementation study of the translated version of REACH II by using a quasi-experimental single-group pre − post treatment design. They found that it had significant effects on the perception of positive aspects of caregiving, reduction in depressive symptoms, subjective burden, bother and caregiving risks among caregivers and abatement in behavioural problems among care recipients. Promises in reach and adoption were demonstrated by the participation of 85 interventionists from 11 non-governmental organisations across 18 districts and 243 caregivers of various demographic characteristics. In a 2020 meta-analysis [4], multicomponent interventions for dementia caregivers had limited positive effect on depressive symptoms. By contrast, mindfulness-based interventions significantly reduced anxiety and depressive symptoms [4, 12,13,14,15]. Hence, this proposed study will incorporate the modified mindfulness-based cognitive therapy, which has been tested effectively among dementia caregivers, as a stress management component [16, 17].

Following the MOST framework, we will select the treatment components carefully to ensure that they are conceptually and operationally distinct from one another, so each component can be evaluated independently. However, we do not assume that the treatment components do not interact. Interaction effects may exist between individual components and influence primary outcomes. For example, during support group sessions, caregivers may exchange skills and information on managing the behavioural symptoms of care recipients by sharing and discussion; as such, the caregivers may build strong bonds and a sense of social support. As a result, behavioural symptom management component may moderate and strengthen the relationship between the support group component and perceived social support. In this regard, social support had a significant effect only when incorporated in multicomponent interventions instead of other domain-specific interventions [18]. Therefore, in the optimisation phase, we will use a factorial approach with effect coding to concurrently test the main effects of individual components and their interaction effects. All experimental conditions will involve education on dementia and caregiving as the core component, since it has been identified as the most ubiquitous component for caregiver support programmes as well as its ease of implementation. This factional factorial design will reduce the total number of conditions from 32 to 16, which retains the benefits of a factorial design whilst allowing a more logistically manageable and feasible study.

Another major limitation of studies on multicomponent interventions is that researchers often omit to report whether components have been implemented with fidelity; thus, the implementation fidelity of individual components remains unknown [19]. Assessment of implementation fidelity is an important first step in determining the effectiveness of a treatment component. Moreover, to our knowledge, no previous study has examined the underlying mechanism of individual components of multicomponent interventions. The factorial design of single intervention components will enable us to determine not only which components are the most strongly associated with changes in primary outcomes but also the fidelity of the implementation of each component, the components that lead to improvements in corresponding proximal outcomes (i.e. the goal of the component) and whether the proximal outcomes mediate the effects of their corresponding components on primary outcomes (i.e. understand the underlying mechanism of each component).

Multicomponent interventions can be tailored to meet the specific needs of individual caregivers given the variability inherent in caregiving situations [8]. Specific treatment modules or strategies and techniques may be applied according to the risk profiles of caregivers to provide personalised support. However, whether such matching between the risk profiles of caregivers and individual components works better than a one-size-fits-all approach remains unknown because the relative efficacy of a specific component on a specific outcome has yet to be examined systematically. Hence, this study will also determine the moderating effects of the baseline scores of primary outcome measures in the relationship between intervention components and primary outcomes.

Although theoretical and empirical evidence suggests the positive effects of multicomponent interventions, a limited number of dementia caregiving support interventions have looked into the “black box” to understand which intervention components work or do not work, and how they work. To optimise the efficacy and scalability of multicomponent interventions, investigations should cover implementation fidelity, the main effect of individual components and their interactions and the underlying mediating/moderating mechanism. This study aims to address these research gaps. The findings will provide a wider repertoire of evidence-based personalised multicomponent interventions to support dementia caregivers.

Objectives

This study aims to:

1.

determine the implementation fidelity of each intervention component,

2.

assess the effect of each component on corresponding proximal outcomes,

3.

examine the effect of each component on primary outcomes (e.g. depression, burden),

4.

determine if the proximal outcomes of each component mediate the components’ effects on primary outcomes,

5.

determine the interaction effect between components on their proximal and primary outcomes and

6.

determine the moderating effect of the baseline scores of primary outcomes in the relationship between each intervention component and the corresponding outcome.

The hypotheses to be tested are as follows:

1.

The implementation fidelity of all six components is equally high.

2.

Components 2 (self-care skills), 3 (behavioural symptom management), 4 (behavioural activation), 5 (modified mindfulness-based cognitive therapy) and 6 (support group) will have significant effects on proximal outcomes, namely, self-care, dementia caregiving strategy, engagement in pleasurable activity, mindfulness and satisfaction with support group, respectively.

3.

Components 2–6 will have a stronger effect on their corresponding primary outcomes, namely, physical health status, stress/burden, psychological well-being, anxiety/depressive symptoms and social support, respectively, than the other components.

4.

Self-care, dementia caregiving strategy, engagement in pleasurable activity, mindfulness and satisfaction with support group mediate the relationship between component 2 and physical health status, between component 3 and stress/caregiving burden, between component 4 and psychological well-being, between component 5 and anxiety/depressive symptoms and between component 6 and social support, respectively.

5.

Component 3 moderates the relationship between component 6 and social support.

6.

The baseline scores of physical health status, stress/burden, psychological well-being, anxiety/depressive symptoms and social support moderate the effects of components 2–6 on changes in scores between the baseline and follow-up assessments of the corresponding outcomes.

Trial design

This study adopts a prospective, assessor-blinded, randomised controlled trial with fractional factorial design by using the MOST principles to evaluate the effects of the five individual components and their two-way interactions (including self-care skills, behavioural problem management, behavioural activation, mindfulness yoga and support group; Table 1). A complete factorial experiment of the five factors would have 25 = 32 experimental conditions (ECs). To conserve resources and reduce logistical complexity, we select a 25–1 fractional factorial design and decrease the number of ECs from 32 to 16 [20]. Our fractional factorial design is made up of a strategically selected subset of 32 ECs based on prioritising the estimation of intervention component main effects and two-way interactions [21]. All the included ECs are listed in Table 1. The study contains a subset of ECs, so the participants will not be randomised into a usual care/control group but every EC can be served as a control condition under different circumstances to test the effect of each component. For example, the effect of component 2 on primary outcomes could be examined through the comparison of participants in EC1 to EC8 and those in EC9 to EC16. In addition, the study assumes the presence of overlapping components within the 16 ECs. These overlaps will be recognised and carefully measured through fidelity monitoring. This study will conform with the Declaration of Helsinki.

Table 1 Experimental conditions of the fractional factorial RCTMethods: participants, interventions and outcomesStudy participants

Eligible family caregivers will be recruited if they satisfy the following inclusion criteria: (i) Hong Kong Chinese citizens aged 18 years or above; (ii) spouse, adult child or child-in-law of a care recipient; (iii) without cognitive impairment (i.e. HK-MoCA 5-Min with a cut-off score that is respective to the age and educational level); (iv) primary family caregiver for an individual with dementia (at least 20 h per week) for at least 1 year; individuals are considered as caregivers if they have assisted with the activities of daily living (ADLs) and instrumental activities of daily living (IADLs) of people with dementia; and (v) caregivers with a certain degree of depression or feeling of burden (Patient Health Questionnaire-9 > 9 or Zarit Burden Scale > 18) to ensure a homogeneous sample [22, 23].

Study setting, sampling and recruitment strategies

We will reach potential participants by using traditional and social-networking recruitment strategies. Traditional methods include newspaper and radio advertisements as well as referrals made through psychiatric and psychogeriatric clinics, non-governmental organisations providing elderly community services, Alzheimer’s associations and university campuses. Social-networking methods include regular postings on sites, such as Facebook, and contextual targeting methods for identifying and directly targeting potential participants (based on their social media comments) with recruitment advertisements.

Allocation concealment, assignment of interventions and blinding

After baseline assessment, all participants will receive component 1 (dementia and caregiver education) as the core intervention. Participants will be randomly assigned to one of the 16 ECs (Table 1) by using the fractional factorial MOST design. Computer-generated random numbers will be used for allocation. Participants will be informed of their assigned group by an independent person not involved in the assessment. The allocation list will be computer-generated by an independent researcher and concealed from other researchers and participants until the time of assignment. Outcome assessors will be blinded to participant group allocation. Unblinding should only occur for participant safety concern when a series adverse event transpires and knowledge of the assignment of intervention could mitigate the health risk (Fig. 1) [24].

Fig. 1

SPIRIT figure — schedule of enrolment, interventions, and assessments

Components of the multicomponent intervention

In the MOST preparation phase, we develop a conceptual model (Fig. 2) based on the theoretical framework model for the stress-health process of informal caregivers for PWDs [25]. Figure 3 illustrates how each individual component affects its proximal outcome, which in turn affects the primary outcome.

Fig. 2

Supposed impact of the different intervention components on the stress-health process of the informal caregiver (adapted from Schulz [25])

Fig. 3

Mechanism of the multicomponent intervention

Component 1: dementia and caregiving education. This core component will provide general knowledge about dementia, information on skills for communicating with PWDs, and common problems related to caregiving and seeking help. In line with the methods employed by Kwok [26], three 30-min sessions will be delivered via telephone and information will be provided in the form of presentation slides (also available in audio and video formats). Following safety walkthroughs (via Zoom) during the first two sessions, advice will be offered about ensuring home safety (e.g. through home modifications).

Component 2: self-care skills. Participants will receive information related to the importance of self-care and its associated skills. This information will be provided in the form of presentation slides (also available in audio and video formats). Caregivers will be reminded of the importance of maintaining good health and continuing healthy behaviour for themselves and their care recipients. They will also be taught to use health passports to keep records of health conditions and doctor appointments of themselves and their care recipients.

Component 3: behavioural symptom management. This component is adopted from previous studies conducted by our team members [27, 28], which will be delivered via telephone by well-trained research assistants. The antecedents–behaviour–consequence model [29] will be introduced to caregivers to manage the behavioural symptoms of care recipients. They will be taught to observe and record the antecedents and consequences of problem behaviour and to use weekly records to keep track of behavioural changes. They will also be taught how to develop plans to change precipitating events or their reactions to these problem behaviours. Home practices will be introduced, and caregivers will be encouraged to complete them regularly.

Component 4: behavioural activation. In this component, we will encourage participants to schedule their engagement in pleasant activities into their daily routines by following the telephone-based protocol used in previous studies of Chinese caregivers conducted by our team members [27, 28]. Specifically, caregivers will learn about the principles of BA and participate in sessions on activity monitoring, activity scheduling, reinforcing or modifying a pleasant event and activity rescheduling based on changes after modification.

Component 5: modified mindfulness-based cognitive therapy. The programme will be led by a mindfulness practitioner and consists of seven 2-h group sessions to be delivered online via video-conferencing software. Each session will have a theme and related activities. For instance, in the first session, the theme is waking up from automatic pilot and participants will be required to scan their body and perform raisin-eating meditation.

Component 6: support group. This component consists of six 1-h group sessions delivered online, in which six major themes will be covered [30]: introduction of dementia caregiving and development of a mutual support group; improvement of home care skills and interpersonal relationships; awareness of caregivers’ mental health; reaching out to community resources; experience sharing on behavioural symptom management; and conclusion and review. These sessions will involve sharing information and engaging participants in group discussion as well as the provision of psychological support and participation in problem-solving exercises. An experienced social worker will act as a facilitator, and two peer leaders will be elected by group members.

Outcome measures

All outcome measures have been validated in local context and will be used in participants. Figure 3 lists all the primary and proximal outcomes.

Primary outcome measures

Physical health status will be measured by the validated 12-item Short-Form Health Survey (SF-12) [31]. SF-12 consists of eight domains, and the participant’s health condition is rated on a 4-point Likert scale. Only the self-rated and physical component will be used to measure the primary outcome, and a high score indicates a good health condition. The Cronbach’s alpha of the physical component in the Chinese version is 0.68 [32].

Caregiver burden and general stress will be measured by the 12-item Zarit Burden Interview scale (ZBI) [33] and the 10-item Perceived Stress Scale (PSS) [34]. ZBI uses a 5-point Likert scale, and a high score indicates a high caregiver burden. The Cronbach’s alpha of the validated Chinese version is 0.84 [33]. PSS consists of the positive element and negative element subscale and measures the subjective perception of stress in a 5-point Likert scale. A high score indicates a high stress level. The Cronbach’s alpha is 0.75 (0.82 for negative subscales; 0.87 for positive subscales) [35].

Psychological well-being will be measured by the four subscales in Ryff’s Psychological Well-Being Scale in 24-item version [36]: self-acceptance (4 items), positive relations to others (4 items), purpose in life (4 items) and personal growth (4 items). The 16 items of subjective psychological well-being will be measured on a 6-point Likert scale. The Cronbach's alpha of the four subscales ranges from 0.77 to 0.88 [36].

Anxiety and depressive symptoms will be measured by the 7-item Chinese version of the Anxiety Subscale of Hospital Anxiety and Depression Scale (HADS-A) [37] and the 9-item Chinese version of the Patient Health Questionnaire (PHQ-9), respectively [22]. HADS-A uses a 4-point Likert scale to measure anxiety symptoms over the past week. The Cronbach’s alpha is 0.77 for the anxiety subscale. PHQ-9 uses a 4-point Likert scale to assess depressive symptoms over the past week [38]. A high score indicates severe depressive symptoms. The Cronbach’s alpha is 0.91 [39].

Perceived adequacy of functional social support will be measured by the 20-item Medical Outcomes Study Social Support Survey (MOS-SSS) [40]. MOS-SSS measures function aspects of perceived social support in four domains by using a 5-point Likert scale: tangible support, emotional-information support, positive social interactions and affectionate support. The Cronbach’s alpha of the overall scale is 0.98 [41].

Proximal outcomes

Self-care will be measured by the 14-item self-care subscale in Risk Appraisal Measure (RAM-SC) [42, 43]. The subscale illustrates the unique risk profile of caregivers in terms of self-care. A high score indicates a low risk of caregivers.

Caregiving strategies will be measured by the 34-item Dementia Management Strategies Scale (DMSS) [42, 43]. The scale uses a 5-point Likert scale and consists of three domains, namely, criticism, encouragement and active management. The Cronbach’s alpha of the three subscales ranges from 0.86 to 0.90 [42, 43].

Engagement in pleasurable activities will be measured by recording the frequency (times per week) and duration (minutes in each time) of each engagement in the past 2 weeks.

Mindfulness will be measured by the 20-item version of the Five Facet Mindfulness Questionnaire (FFMQ) [44]. Using a 5-point Likert scale, it measures five facets of mindfulness, namely, observing (4 items), describing (4 items), acting with awareness (4 items), nonjudging to inner experience (4 items) and non-reacting to inner experience (4 items). A high score reflects a high level of mindfulness. The Cronbach’s alpha values are 0.83 in the community sample and 0.80 in the clinical sample [44].

Social support will be measured by the 12-item Multidimensional Scale of Perceived Social Support (MSPSS) [45]. The scale uses a 7-point Likert scale and consists of three sources of support: family, friends and significant other. The Cronbach’s alpha of the three subscales ranges from 0.85 to 0.91. [46].

Satisfaction with the support group will be measured on a 7-item scale during the post-intervention. Participants will report their satisfaction with specific components and quality of service [46].

Moderators (control variables)

Positive caregiving appraisal will be measured by the 11-item Positive Aspect of Caregiving (PAC) [47]. The two domains, namely, enriching life and affirming self, are measured on a 5-point Likert scale. The Cronbach’s alpha is 0.89.

Caregiving self-efficacy will be measured by the 15-item Revised Scale of Caregiving Self-Efficacy [