Grief in family carers of people living with dementia: A systematic review

ABSTRACT Objectives: Grief research in family carers of people with dementia has increased. We aimed to report the prevalence of pre-death and post-death grief and to synthesize associated factors and the relationship between pre-death factors and post-death grief and services used to manage grief. Design: (Prospero protocol: CRD42020165071) We systematically reviewed literature from PsycINFO, MEDLINE, CINAHL, and ASSIA until April 2020. Effectiveness of intervention data and studies not written in English were excluded; qualitative studies were additionally excluded during study selection. Study quality was assessed using the Mixed Methods Appraisal Tool. Evidence was narratively summarized. Participants: Family non-paid carers of somebody with any dementia type. Measurements: Validated measures of pre-death and/or post-death grief. Results: We included quantitative data from 55 studies (44 rated as high quality). Most included solely spouse or adult child carers. Forty-one studies reported pre-death grief, 12 post-death grief, and 6 service use; eight were longitudinal. 17% met the Prolonged Grief Disorder criteria pre-death (n = 1) and 6–26% (n = 4) of participants met the Complicated Grief criteria post-death. Being a spouse, less educated, caring for somebody with advanced dementia, and greater burden and depression were associated with higher pre-death grief. Lower education level and depression were predictive of higher post-death grief. Pre-death factors found to influence post-death grief were grief and depression. Limited service use evidence was reported. Conclusion: Awareness of characteristics which increase the likelihood of higher grief can help identify those in need of support. Future research should focus on what supports or services are beneficial to grief experiences.


Background
Dementia is the leading cause of death in the UK and the seventh commonest cause globally (World Health Organization, 2017). More than 80 billion hours of care a year are provided annually by informal carers. Dementia caregiving can have significant emotional implications for family carers, and the experience of grief while the person with dementia is still alive, known as pre-death grief (Blandin and Pepin, 2017;Lindauer and Harvath, 2014), is particularly common.
As knowledge about pre-death grief in the context of dementia caring has increased in the literature, terms used to reflect this experience have also evolved moving from the concept of "anticipatory grief" to pre-death grief, which can also be referred to as "dementia grief." There are many overlaps between anticipatory grief and pre-death grief; however, pre-death grief relates to losses experienced rather than anticipated and is thought to better encompass the important facets of pre-death grief for this population (Blandin and Pepin, 2017). Predeath grief has been defined as "the emotional and physical response to the perceived losses in a valued care recipient. Family caregivers experience a variety of emotions (e.g. sorrow, anger, yearning and acceptance) that can wax and wane : : : from diagnosis to the end of life" (Lindauer and Harvath, 2014). Pre-death grief can occur due to the lengthy and uncertain dementia trajectory and can be triggered by losses associated with dementia such as compromised communication and changes in relationship quality and carer freedom (Lindauer and Harvath, 2014).
Bereavement and grief are a normal part of life; however, for a minority of people grief can interfere with everyday life and involve long-term severe reactions to the loss that impact on functioning. Researchers and clinicians have been attempting to differentiate between normative bereavement and pathological or disordered grief since the 1990s. A debate has ensued involving competing theoretical conceptualizations, diagnostic criteria, and psychometric measurement. The first diagnostic criteria for a bereavement related disorder were termed pathological grief (Horowitz et al., 1993) which was then updated to Complicated Grief (CG) (Horowitz et al., 1997). Different terminology has been used over time, but the terms CG and Prolonged Grief Disorder (PGD) have been most commonly used. Higher levels of grief prior to death are associated with PGD or CG after death (Chan et al., 2013;Romero et al., 2014). PGD is characterized by symptoms such as longing for and preoccupation with the deceased, emotional distress, and significant functional impairment for at least 6 months after the loss (Killikelly and Maercker, 2018). CG, while a very similar concept to PGD, is characterized by intense grief that lasts longer than would be expected according to social norms and impairment in daily functioning.
PGD has been associated with poor physical health, suicidality, reduced quality of life, and functional impairment (Boelen and Smid, 2017). Although distinct from other mental health disorders, PGD can co-occur with depression, posttraumatic stress disorder, and anxiety (Boelen and Smid, 2017). While pre-death grief is not PGD, the intensity and duration of pre-death grief experienced by some may be consistent with definitions of PGD, and therefore research has begun to explore this using adapted versions of PGD measures Moore et al., 2017) In response to advancing research evidence, grief disorders have been included in two diagnostic classification systems. The DSM-5 introduced Persistent Complex Bereavement Disorder under the conditions for further study, which is a combination of PGD and CG criteria, while the International Classification of Disease 11 th revision (ICD-11) introduced PGD as a disorder based largely on the PGD criteria proposed by Prigerson et al. (2009). For the purpose of this paper, post-death grief will be reported using the terminology used in the original studies, that is, CG or prolonged grief.
The most recent systematic review that synthesizes the prevalence and associated factors of predeath, post-death, and prolonged/CG was published in 2013 (Chan et al., 2013). The review included 31 studies, many of which were of poor quality, and included only 1 study reporting the prevalence of PGD. Studies exploring the relationship between pre-and post-death grief were also limited within this review. They found that moderate to severe stage of dementia predicted pre-death grief, while being a spousal carer and being depressed were the biggest predictors of both normal post-death grief and prolonged grief post-death. Poor quality evidence suggested that between 47% and 71% experienced pre-death grief, and around 20% experienced CG. Since this review of studies published until 2009, research has further explored the experience of grief. This, in turn, adding to our understanding of the prevalence and predictors of grief.
We do not know how many carers need support either before or after the death of the person with dementia. Current grief services tend to target those who have experienced a recent death. The current bereavement model in the UK suggests that most people manage with support from family and friends, and without the need for professional intervention. However, it is unclear whether this model meets the needs of carers of people living with dementia (National Institute for Clinical Excellence, 2004). While the prevalence of CG is estimated between 10 and 20% (Lobb et al., 2010), one in three carers of people living with dementia was found to access bereavement services (Bergman et al., 2011) suggesting a higher proportion of carers seek professional support than the bereavement model indicates. It is also less known if carers of people living with dementia seek services for pre-death grief, although a recent study of current carers found that 30% had accessed formal counseling (Moore et al., 2020).
In light of the newer definitions regarding pre-death grief, CG and PGD, and the wealth of research exploring these experiences, we aimed to update and extend the review by Chan (2013). We aimed to seek answers to the following review questions: In family carers of people living with dementia: 3. In longitudinal studies, what is the relationship between pre-death factors and post-death prolonged/CG? 4. What services do carers use to manage grief? This review does not examine effectiveness of grief interventions as this was addressed in a recent review (Wilson et al., 2017).

Methods
The review protocol CRD42020165071 was registered on PROSPERO international prospective register of systematic reviews and followed PRISMA guidelines (Moher et al., 2009).
We initially planned to include qualitative studies and gray literature; however, due to the large volume of literature available we decided to limit our inclusion criteria to quantitative studies during full-text review.

Inclusion criteria
• Type of studies: All quantitative studies or quantitative data from mixed methods studies. Studies were not excluded based on quality. • Topic: Grief prevalence, relationship between preand post-death grief, factors associated with grief and services used to manage grief. • Participants: Family or friend non-paid carers (aged 18 or over) of people with dementia. • Setting: Participants were providing care or support for somebody living with any type and severity of dementia in the community or in long-term care facilities. Bereaved carers were also included.

Exclusion criteria
• Effectiveness of intervention data • Studies not written in English • Paid/professional carers • Qualitative data (excluded at full-text review) • Gray literature (excluded at full-text review).

Search strategy
We searched PsycINFO, MEDLINE, CINAHL, and ASSIA to April 2020. The search strategy was refined through test searches using medical subject headings and free-text terms. The search included keywords and terms associated with dementia, grief and family carers as shown in Supplementary File 1.

Selection of studies
Abstracts of identified citations were independently screened by two reviewers (either SC and KM, or SC and NK) to ensure consistency when applying the inclusion criteria. Interrater reliability of full-text selection was calculated using Cohen's kappa (K) and ranged from moderate (93.3% agreement) to nearly perfect (97.8% agreement) between the author combinations (Landis and Koch, 1977). Full texts of citations were checked for eligibility by two reviewers, and any discrepancies resolved through discussion with all three reviewers.

Data extraction
Characteristics of the studies were extracted by SC into a table developed for this review. Extracted data included country of origin, study design, details of grief measurement tools used, participant characteristics, and results such as relationship between factors associated with and prevalence of pre-and post-death grief, and services used by carers to manage grief. Two authors (KM, NK) independently checked 20% of data extraction.

Quality assessment
The Mixed Methods Appraisal Tool (MMAT) 2018 revised version with improved content validity was used to assess the quality of included studies (Hong et al., 2018). The MMAT appraises and describes the quality for three methodological domains: mixed, qualitative, and quantitative (subdivided into three sub-domains: randomized controlled, non-randomized, and descriptive). Authors are advised to present how studies meet the quality criterion appropriate to their study type. All studies were assessed using the tool, and 20% were checked by a second author (KM, NK, EW) with any discrepancies discussed and resolved.

Data analysis
The Cochrane framework for summarizing study characteristics and synthesizing data was implemented. At protocol stage, questions were defined, and planned analyses proposed; evidence for Q1 was synthesized based on the measure of grief and the cut-off scores used in studies. Q2 was addressed by summarizing associations with grief and exploring subgroup differences such as differences in carer and the person living with dementia characteristics and experience of grief such as gender, ethnicity, age, relationship to the person living with dementia, and dementia severity. Longitudinal evidence for Q3 was summarized to describe the impact of carer and care-related factors assessed before the death of the person living with dementia on post-death grief. Services used to manage grief were described to address question Q4.
Grief in family carers of people living with dementia 479 Associations were tabulated and a narrative summary was provided of evidence from studies which met at least four out of five of the MMAT quality criteria. Associations were discussed within the narrative summary if there were at least three studies reporting a factor for pre-death grief, while all factors were discussed for post-death grief as there were fewer included studies. Where associations were reported, factors were discussed within domains which were identified from the evidence: demographic carer factors; psychosocial characteristics; person living with dementia and care-related factors; and bereavement factors.

Results
We identified 771 unique citations after removing 134 duplicates, of which 230 met our inclusion criteria for full-text review. Fifty-five quantitative and mixed methods studies were included. Only quantitative data were included from mixed methods studies as demonstrated in the PRISMA diagram (Moher et al., 2009) (Figure 1).

Sample size and characteristics
The majority of studies were conducted in the USA (n = 33), seven in Europe, six in Singapore, three in Hong Kong, three in Canada, and one study each in Australia, Puerto Rico, and South Korea. Studies mainly consisted of spouse or adult child carers. Thirty-four studies reported dementia severity of the care recipient; 13 studies included moderate to severe dementia, 3 reported moderate, 3 reported advanced, and 10 included mild, moderate, and severe dementia. Forty studies reported pre-death grief, 11 post-death grief, 3 both pre-and postdeath grief, and 6 service use. Eight of the 55 studies included longitudinal data. Pre-death grief was measured using the Marwit Meuser Caregiver Grief Inventory (MMCGI) (n = 11), the MMCGI-Short Form (SF) (n = 14), the PGD Scale pre loss (PG-12) (n = 1), or the Inventory of PGD Scale short form pre-loss (n = 1). Disordered post-death grief was measured using the Inventory of Complicated Grief (n = 8) and the PGD Scale (PG-12) (n = 1), while normal postdeath grief was reported using the Texas Revised Inventory of Grief (n = 3) (Table 1).

Quality appraisal
As only quantitative data were included from mixed methods studies, the appropriate quantitative section was completed. Similarly, for RCT studies, as only data reporting grief prevalence or associated factors were included, either the non-randomized quantitative study or quantitative descriptive study component of the MMAT was more appropriate to complete than the RCT component. For the purpose of this review, a high-quality study was determined by studies meeting four or five of the MMAT criteria. Of the six studies where the quantitative descriptive component of the MMAT was completed, five studies were rated high quality. No studies met the criteria related to the sample being representative of the target population. Of the 49 studies assessed using the non-randomized quantitative study component, 16 studies met all the criteria, 23 studies met four of the five criteria, and 10 met 3/5. Twenty-five studies did not meet the criteria regarding representativeness of sample, 4 studies did not use appropriate measures to assess grief, and 12 studies did not control for confounders. Two studies did not have complete data (Table 2).
Q1. What is the prevalence of pre-death and prolonged/CG and when does it become a clinical disorder? Pre-death grief: Four studies (Chan Wei Xin et al., 2019;Liew and Yap, 2018;Sanders and Adams, 2005;Ott et al., 2007) reported the prevalence of pre-death grief data; 10-18% of participants were reported to be at risk of high grief based on the MMCGI or MMCGI-SF cut-off criteria of scores being one standard deviation above the mean (Marwit and Meuser, 2002). One study reported 16.7% met the criteria for PGD as assessed using the PG-12 before death ;  used 11 items of the PG-12 and reported a low rate of participants met the criteria for PGD. Moore et al. (2017) reported 38% had a high occurrence of symptoms as assessed by the ICG short-form preloss version. See Supplementary File 2 for detailed prevalence data. Complicated grief: 20-26% of participants were reported to meet the criteria of CG as assessed by the 19 item ICG . Two studies used revised versions of the ICG; Moore et al. (2017) reported 22% met the criteria of CG using a 16 item version, and Romero et al. (2013) reported 6% of participants met the criteria using a 15 item version.
Prolonged grief: Givens et al. (2011) reported a low rate or participants met the criteria for PGD as assessed using 11/12 item PG-12 (see Supplementary File 2) Q2. What factors associated with pre-death and prolonged/CG? Pre-death and post-death associations are shown in Table 3.

PRE-DEATH A SSOCIATIONS ( N = 31)
Carer demographic factors Relationship type: The evidence indicates being a spousal carer is associated with higher pre-death grief than adult children or other relationship types with 10/16 studies reporting significant findings. Two studies found interaction effects: Cheung et al. (2018) found spouses caring for someone in later stages had the highest grief, and Ott et al. (2007) found being a spouse was only associated with higher grief when the carer did not live with the person living with dementia.
Lower education was found to be associated with higher grief in most of the studies which explored education (Liew, 2015;Liew et al., 2019a;Passoni et al., 2015;Kiely et al., 2008). However, one study (MacCourt et al. 2017) reported contradicting findings that not having a university education predicted lower grief and Marwit and Meuser (2002) found no association.
Gender was not found to be associated with grief Passoni et al., 2015;Sanders and Adams, 2005;Warchol-Biedermann et al., 2014;Liew, 2015;Meichsner et al., 2016;Meichsner and Wilz, 2018;Sanders et al., 2008;Liew et al., 2019b) with the only significant association reported by Passoni et al. (2015) who found being female was significantly associated with higher grief when gender was the only predictor.
Ethnicity: The included studies reported mixed evidence for the impact of ethnicity on grief. Ross and Dagley (2009) in a US-based study found African-Americans reported higher grief than white carers. Similarly, two Singapore-based studies found being of Malay ethnicity was associated with higher grief than Chinese/Indian/Eurasian/ other ethnicities (Liew, 2015;Liew et al., 2019b). However, two US studies with a majority of white sample found no association with ethnicity Sanders and Adams, 2005).
Age: Carer age was not independently associated with total grief scores, with only one study reporting an association between older carer age and grief when age was combined into a demographic variable with education and caring time . with the person before institutionalization , and currently living with the person Meichsner et al., 2016;Meichsner and Wilz, 2018) were found to be associated with higher grief. Whereas no associations were found between grief and living with the person in two studies (Cheng et al., 2019;Warchol-Biedermann et al., 2014) or whether the person lived at home or in residential settings and grief (Sanders and Adams, 2005;Sanders et al., 2008).
Primary carer: Being the primary carer was mainly found to be associated with higher grief Liew et al., 2019a;Chan et al., 2017).
Amount of care provided: The studies report mixed evidence for an association between providing daily care or amount of time spent providing care and grief. Liew et al. (2019b) found providing daily care was associated with higher grief and Kiely et al. (2008) found an association with providing a minimum of 7 hours of care a week. However, no association was found between perceived amount of care provided and grief (Liew, 2015;Marwit and Meuser, 2002).
Coping styles: Dysfunctional coping was found to be positively associated with grief (Romero et al., 2013;Ott et al., 2010). Additionally, Ott et al. (2007) found greater use of coping by emotional venting was associated with grief when the person lived at home, and coping by planning and self-blame were positively associated with grief when the person did not.
Social support: Elements of social support appear to have a positive impact on grief. Negative associations were found between grief and perceived social support (Chan et al., 2017;Marwit and Meuser, 2002;Romero et al., 2013) and an association was found between greater satisfaction with social relationships and lower grief (Meichsner et al., 2016). Support was reported to buffer the effect of grief and mediate the negative relationship between grief and psychological well-being in a study by Park et al. (2019). Social network size was not found to be associated with grief (Cheng et al., 2019).
Carer health problems: The studies report mixed evidence as to whether greater health problems were associated with higher grief; two found significant positive associations (Walker and Pomeroy, 1997) and two reported no association (Cheng et al., 2019;Kiely et al., 2008) while Sanders and Adams (2005) reported an association for the MMCGI subscale HSL only.

Declining satisfaction with life
Walker 1997  Person living with dementia related factors Dementia severity: Mixed findings were reported for the impact of dementia severity and grief. Eleven studies explored severity, and associations between greater dementia severity and higher grief were found in seven (Adams and Sanders, 2004;Liew et al., 2019b;Liew et al., 2019a;Marwit and Meuser, 2002;Sanders et al., 2008;Warchol-Biedermann et al., 2014;Chan Wei Xin et al., 2019). One study (Chan et al., 2017) found dementia severity was only associated with grief for spouse carers and four studies found no association (Meichsner et al., 2016;Ott et al., 2007;Passoni et al., 2015;Strong and Mast, 2013) Behavioral problems/neuropsychiatric symptoms: There was some indication that behavioral problems or neuropsychiatric symptoms were associated with higher grief; severe behavioral problems (Liew et al., 2019a;Liew et al., 2019b) and disruptive behaviors and psychotic symptoms (Cheng et al., 2019) were associated with higher grief.
Age of person living with dementia: Evidence was mixed regarding younger age of the person living with dementia and carer grief.
Gender of person living with dementia was also not associated with grief Liew, 2015;Liew et al., 2019b).

ASSOCIATIONS WITH CG/PGD (N = 6)
Carer demographic and care-related factors There was no association between carer gender and CG. Less education was mainly found to be associated with higher CG. Ethnicity was explored in five studies, with the evidence suggesting no association with grief. There was mixed evidence regarding whether being a spouse carer was associated with higher CG than adult children or other relationship types.
Carer health and psychosocial factors post loss Postdeath social support: Surprisingly,  found participants with higher grief were more likely to participate in social activities and less likely to pursue negative interactions. Social support and satisfaction with support were not significant. Bergman et al. (2011) found satisfaction with support was strongly negatively associated with grief, and having less people in their social network was moderately associated with higher grief.
Post-death depression: Higher post-loss depression was associated with higher CG in all studies Bergman et al., 2011;Givens et al., 2011).
Bereavement factors Mixed findings were reported regarding an association between time since death and grief. Schulz et al. (2006) and Holland et al. (2009) reported strong evidence that grief improved over time, while Stahl and Schulz (2018) and Romero et al. (2013) did not find associations. Retrospective reporting of being unprepared for the death (Hebert et al., 2007;Stahl and Schulz, 2018) and lower levels of relief (Stahl and Schulz, 2018; were associated with higher grief.
Q3. In longitudinal studies, what is the relationship between pre-death factors and post-death prolonged/CG? Associations from six longitudinal studies relating to this research question are shown in Table 4.

PRE-L OS S CAR ER FA CTORS
Two studies explored pre-death grief Romero et al., 2013) and both report strong evidence of an association between high pre-death and high post-death grief even when accounting for confounders. Romero et al. (2013) reported 34% of post-death grief was accounted for by pre-death grief. The evidence indicates pre-loss depression is associated with CG post-death, with all studies finding associations between higher depression and higher grief (Holland et al., 2009;Bergman et al., 2011;Givens et al., 2011;Romero et al., 2013;Schulz et al., 2006;Boerner et al., 2004).
Of the three studies Bergman et al., 2011;Hebert et al., 2007) which explored elements of religiosity, a significant negative association was only found between religious attendance and grief (Hebert et al., 2007).
Two studies measured social support before and after the death. Romero et al. (2013) found no association between grief and social support. Hebert et al. (2007) found an increase in social integration from pre-loss to post-loss was associated with fewer grief symptoms. An increase in satisfaction with support pre-and post-death was not associated with grief.
Amount of care provided, believing the person had at least 6 months to live, and carer's understanding of the complications of dementia were explored in one study, and no associations with grief were found .

PRE-LOSS PERSON LIVING WITH DEMENTIA
FACTOR S Schulz et al. (2006) found that younger age, greater dementia severity, and a higher dependence in activities of daily living (ADLs) were all significant independent predictors of higher grief. When demographic and bereavement confounders were Grief in family carers of people living with dementia 501 controlled for, however, greater dementia severity and dependence in ADLs, and the person being female emerged as significant predictors of higher grief. However, Givens et al. (2011) and Stahl and Schulz (2018) both found dementia severity was not associated with grief, with Givens et al. (2011) reporting younger age, having lived with the person prior to nursing home admission, and the person having been hospitalized in the last 90 days of life to be associated with higher grief.  Walker and Pomeroy (1997) and Ott et al. (2010) reported support group attendance (36% and 35%, respectively). Ott et al. (2010) found nearly a third of participants used resources from dementia-related organization and 60% attended an adult care program. Loos and Bowd (1997) reported perceived helpfulness of services; family assistance was the most helpful (60%), followed by physicians (57%), friends (43%), Alzheimer's Society dementia charity (32%), support groups (22%), and nurses (22%). Kobiske et al. (2018) reported 56% of carers of someone living with young onset dementia had not received professional counseling. Post-death: Bergman et al. (2011) reported 30% accessed at least one service (i.e. counseling, support group or psychotropic medication), 13% received either individual, family, or pastoral counseling, and 13% accessed a bereavement support group. Crespo et al. (2013) found 98% accessed professional help and 84% received support from non-formal sources. While only 16% accessed bereavement services, 38% reported a need to attend a bereavement-related service.

Discussion
This review synthesizes quantitative data from an extensive and disparate body of international literature. We attempted to address four key research questions, however, the bulk of the evidence focused on determining associations with grief.
Q1. What is the prevalence of pre-death and prolonged/CG and when does it become a clinical disorder? The most commonly used measure of grief was the MMCGI and MMCGI-SF. The majority of studies reported mean grief scores for the whole sample which does not indicate whether individual participants scored at risk of high grief. From the studies that did report individual risk, 10-18% scored above this normative cut-off score, which fits the statistics of the original study assessing grief using this measure (Marwit and Meuser, 2002). This suggests that a subsample of carers may need support at this stage. This is also likely to be at a time where grief is not recognized by society or family and friends and can lead to complex grief situations and feelings of isolation .
An important finding is that, despite an abundance of research into dementia carer grief, we are unable to determine the prevalence of carers experiencing high pre-death grief. This is reflective of pre-death grief being a described concept without diagnostic criteria, and lack of a clinical tool to assess the experience. The MMCGI and MMCGI-SF allow comparisons with a normative sample score, which is statistically and not clinically driven. In the absence of a gold-standard screening tool, using this approach can identify carers who may be at risk of experiencing higher grief who are in need of further assessment, but may miss others in need of support but who score lower than the top 18%.
Indicators of disordered post-death grief are identified in the literature by the use of validated measures developed against a defined criteria; the ICG was most commonly used and determines indicators of pathological grief. The prevalence of CG ranged from 6 to 26% from four studies (although there is no validated cut-off score for the ICG-r). One study used a modified version of the PG-12 measure to assess PGD, and found a low rate of participants met the criteria. Subtle but key differences exist among the different criteria for PGD or CG and the algorithms applied to determine prevalence, and while PGD is a classified disorder in the ICD-11, there is not currently a validated tool which assesses all of the proposed criteria. Therefore, reported prevalence should be interpreted with caution and within the context of individual studies' criteria and assessment of grief (Lenferink et al., 2019;Eisma et al., 2020).
Q2. What factors are associated with pre-death and prolonged/CG? Our findings build on the work of Chan 2013 and highlight that being a spouse carer, less educated, caring for somebody at a more severe stage of dementia, and higher levels of burden and depression are associated with greater pre-death grief. Studies exploring associations with post-death grief scores reported using the measures described above found that higher levels of pre-death grief and depression were predictive of higher post-death grief. Not being prepared for the death of the person and lower levels of carer education were also indicative of higher grief scores post-death. No studies conducted analysis to determine factors that were associated only with those who met the criteria for CG or PGD. There is mixed evidence for relationship type and post-death grief in comparison to predeath grief, which may suggest that bereavement factors and other demographic or psychosocial variables have a stronger role post-death. Bereavement factors were less frequently explored in the reported studies, and it is unclear from the evidence whether time since death is associated with grief. Evidence for both pre-death and post-death grief suggests there is no relationship between carer gender and grief. Research in diverse samples is needed to Grief in family carers of people living with dementia 503 further understand the relationship between ethnicity and grief.
There is, however, a difficulty of determining which factors are most associated with grief, which lies within the complexity and interplay of different variables. Evidence is limited to the factors included in studies and the type of analyses carried out. Variations in associations could be in part attributed to study methodology and study participants, particularly as 12 studies did not meet the MMAT criteria for attempting to account for confounders. For example, in contrast to much of the evidence, Passoni et al. (2015) found relationship type had no direct impact on grief, and instead suggested the higher probability of spouses developing PGD can be attributed to sociodemographic or psychophysical features rather than being a spouse or adult child carer. Additionally, few studies explored how anxiety impacts on grief, but as evidence suggests carers do experience anxiety (Meichsner et al., 2016;Passoni et al., 2015;Bergman et al., 2011;Schulz et al., 2006) often at the level of a clinical disorder (Moore et al., 2017), future studies should be including it as a potentially influencing factor, particularly exploring direction of causality and the role of predisposing factors such as personality type.
The different measures used to assess grief and associated variables can also make it difficult to interpret or generalize findings. This is particularly relevant where different elements of a concept are explored under a shared term. For example, various aspects of social support were measured ranging from one question determining network size to a 20-item scale designed to measure the extent to which the individual perceives their needs for support, information, and feedback are fulfilled. It is therefore unsurprising that there is mixed evidence about the impact of social support. However, there is indication that elements of social support have a positive impact on grief, and exploring the role of social support domains and grief will increase our knowledge on how to support carers with grief. A recent study exploring grief in family carers in palliative care found that while there were no differences in pre-death or post-death grief in relation to social support, social support moderated the relationship between them. This was significantly stronger for those with lower social support, suggesting that those with high predeath grief and low social support were more likely to have high post-death grief (Axelsson et al., 2020).
Q3. In longitudinal studies, what is the relationship between pre-death factors and post-death prolonged/CG? While only two studies Romero et al., 2013) explored the relationship between pre-death and post-death grief, they provide strong evidence that higher pre-death grief is associated with higher grief post-death. Higher pre-death depression was also associated with higher grief post-death. This highlights the importance of recognizing that carers may benefit from support with grief while they are still providing care. The need for grief support pre-death suggests that current grief and bereavement programs which target post-death grief may not be meeting the needs for this population. Further exploration of the role of social support and post-death grief is needed; Hebert et al. (2007) found an increase in social integration and having fewer negative social interactions were associated with lower post-death grief, while  found greater social activities were associated with higher grief.  interprets this finding within the context of the dual-process model, where social activities during bereavement may be indicative of a coping strategy.

Q4. What services do carers use to manage grief?
Due to the little evidence on service use in the included studies, it is difficult to draw definitive conclusions to this question. Those that explored service use pre-death found that over a third of participants received counseling or attended a support group. An older study by Loos and Bowd (1997) found participants reported support from family as being the most helpful. Typically, grief services are offered post-death, with pre-death grief being less frequently screened for by services. Therefore, increasing awareness and understanding of pre-death grief could be a promising step in being able to support those experiencing it. A recent study exploring the usefulness and acceptability of an animation to raise awareness of grief found some benefits of recognizing and identifying experiences as grief in carers of people living with dementia (Scher et al., 2021). Post-death service use was explored in just two studies where bereavement support groups and counseling were reported. Perhaps the most important finding was that selfreported need for bereavement services more than double the number of participants who used them (Crespo et al. 2013). Service evaluations may provide further insight into what services are being provided, who utilizes services, and the effectiveness of them.

STRENGTHS AND L IMITATIONS
We excluded non-English studies, qualitative studies and gray literature which may have meant we missed useful evidence. However, this study is able to provide an up-to-date reflection of what demographic and psychosocial factors are important to the experience of grief. While the included studies were conducted in a range of countries, studies from lower middle-income countries are underrepresented. However, research from different cultures is emerging, and differences in grief experiences being reported; relationship type was not found to be significant for Polish carers, and the authors suggest this is reflective of three-generational living and emotional ties between adult children and their parents remaining strong . Similarly, Asian carers expressed more worry and felt isolation than the normative sample which was conducted in the USA (Liew, 2015). There were also very few longitudinal studies that explored grief over time or into bereavement, which would provide a richer understanding of the grief experience. In some studies by the same authors, it was difficult to determine if the same samples were being reported so there may be some over-reporting factors associated with grief (e.g. Liew and Yap (2018) and Liew et al. (2019a). The included studies were of high quality as determined by the MMAT, with 44/55 studies meeting at least 4 of the 5 quality criteria. However, 25 studies did not meet the criteria related to samples being representative as most cohort studies recruited from dementia-related clinical or information services rather than general population samples. CLINICAL A ND RESEARCH IMPLIC ATIONS Tools such as the MMCGI may provide a useful screen to identify those at risk of high pre-grief, but awareness that certain demographic, psychosocial, and care recipient-related factors can influence the experience of grief will also be beneficial. While predeath grief interventions are in their infancy, evidence from pilot studies indicates that interventions should be multifaceted and not increase carer burden due to the unique clinical presentation of predeath grief in this population (Wilson et al., 2017).
While there is a huge body of research reporting cross-sectional data on the factors associated with grief, a shift is needed toward research building on the evidence from the few promising intervention studies undertaken (Wilson et al., 2017). For carers of people living with dementia living in care homes, a grief management intervention that involved group sessions with dementia education, communication skills, conflict management, and grief management skills demonstrated a reduction in the heartfelt sadness and longing subscale of the MMCGI (Paun et al., 2015). Another intervention, aimed at carers caring for the person living at home, involved one to one counseling and reported promising grief-related benefits (Ott et al., 2010). Future research should continue to build on these findings to identify what individual components of grief interventions are most beneficial and who they are most beneficial for with regard to particular demographic and psychosocial factors. Utilizing standardized measures to assess grief and reporting those who are at risk of higher grief or who meet the criteria for Complicated or Prolonged Grief across the dementia caregiving trajectory will also further our understanding of who needs support and when.

Conclusion
This review builds on the previous findings of Chan et al. (2013) and synthesizes quantitative data exploring the grief experience of carers of people living with dementia. The findings indicate that particular demographic features and psychosocial characteristics play a role in grief for these family carers. Awareness of factors that increase the likelihood of experiencing higher levels of grief can help to identify those in need of support. Future research should consider the interplay of such factors and focus on potentially modifiable elements such as social support. There is limited evidence regarding service use and grief, and future research should also focus on what components of support or service provision are important for carers with regard to grief.

Conflict of interest
None.

Description of authors' roles
The review was designed and conducted by SC, KM, and NK. EW supported with the quality appraisal of studies, and ES contributed to the protocol. The manuscript was prepared by SC and critically reviewed and approved by all authors.