Prolonged grief disorder prevalence in adults 65 years and over: a systematic review

Background Prolonged grief disorder (PGD) is a recently recognised mental health disorder with an estimated prevalence of 10% in the bereaved adult population. This review aims to appraise and summarise evidence relating to PGD in older adults (≥65 years), a growing population group, most likely to experience bereavement and often assumed to cope well. Method Literature from Medline, PsycINFO, CINAHL, Cochrane Library and Web of Science was searched. Epidemiological and non-epidemiological studies including data on frequency of PGD in older adults bereaved by mainly natural causes were included and a descriptive analysis undertaken. Results From 2059 records, three epidemiological and six non-epidemiological studies were included. Most studies had good internal but not external validity. Conditional prevalence for PGD ranged between 3.2% and 48.8%. Heterogeneity in sample characteristics and study methodology contributed to this variability resulting in a descriptive analysis. The prevalence rate of 9.1% by Kersting et al was the best available estimate for PGD in older adults for western countries. The small number of epidemiological studies and the use of varying PGD-constructs which did not match International Classification of Diseases 11th Revision and Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition criteria were the main limiting factors. Conclusion This first review on PGD prevalence in older adults suggests that, despite studies’ methodological short comings, a similar proportion of older adults experience PGD as the general bereaved adult population (1:10). With older adults forming the largest subgroup among the bereaved, health and social care systems need to adapt their provision of care to address the specific needs of older adults.


BACKGROUND
The death of someone close confronts people with one of the most painful challenges of their lifetime.A bereavement can have an impact on health and psychological well-being. 1 2 Grief is the response to a bereavement; primarily an emotional reaction, but also involves cognitive, behavioural and physical reactions. 3Grief varies between individuals and between cultures. 4Most people 'grieve normally' and cope by drawing on their own resources and their informal support networks. 5'Grieving normally' includes acceptance of the death, processing related pain, adjustment to a life without the deceased and finding a way to stay connected while moving on in life.If these processes are not resolved, grief can become clinically relevant. 6This has recently been acknowledged by including 'persistent complex bereavement disorder' (PCBD) in the appendix of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) 7 and 'prolonged grief disorder' (PGD) in the International Classification of Diseases 11th Revision (ICD-11). 8 debate about precise criteria and how Systematic review to label clinically relevant grief is still ongoing (see the Method section).Given the similarity of both diagnostic constructs, 9 10 this review uses the term 'prolonged grief disorder' for both. 11he main characteristic of PGD is separation distress, described as intense yearning or longing for the deceased, emotional pain and preoccupation with the deceased person or the death.Other features include difficulties in accepting the death and continuing with life, emotional numbness, anger and avoidance of reminders of the loss. 7 8 10Table 1 details a full list of symptoms.While PGD symptoms do not differ from normal grief per se, they last longer, are more intense, distressing and impairing, compared with what is normal within an individual's culture. 10To avoid pathologising the initial often intense phase of grief, a diagnosis of PGD cannot be made until 6 months after the death according to ICD-11 and 12 months according to DSM-5.
By impairing mental health, and potentially also physical health, PGD can have a profound impact on the life of the bereaved and their families. 1 12 13he prevalence of PGD in the adult population is unknown.Reported prevalence rates vary widely (online supplemental material 1).Prevalence estimates for general population samples (bereaved and nonbereaved people) are 2.3% in Switzerland 14 and 3.7% in Germany. 15The prevalence rate for the bereaved adult population is estimated to be 10%. 16 17lder adults (aged 65+ years) 18 are more likely than younger people to experience the loss of someone close.The 'Changing Lives of Older Couples' study found 29% of widows in the study met criteria for PGD. 19Studies of bereaved family carers, who are often older adults, frequently report PGD prevalence rates above 20% (online supplemental material 1).
A common assumption is that grief in older adults is unproblematic and their support needs have in consequence been largely neglected. 20This may reflect ageist perspectives, 21 linked to assumptions that losing a loved one in the 'autumn of life' is easier to accept and that older people are able to handle grief successfully because they are likely to have experienced other bereavements. 20 22While research shows young age to be a risk factor for PGD, 23 we cannot conclude that older age is a protective factor: the common belief that bereavement is 'less' problematic in older age has been C. Since the death, at least six of the following on most days to a clinically significant degree, for at least 12 months after the death: 1. Marked difficulty accepting the death.2. Disbelief or emotional numbness over the loss.3. Difficulty with positive reminiscing about the deceased.4. Bitterness or anger related to the loss. 5. Maladaptive appraisals about oneself in relation to the deceased or the death (eg, self-blame).6. Excessive avoidance of reminders of the loss.7. A desire to die in order to be with the deceased.8. Difficulty trusting other people since the death.9. Feeling alone or detached from other people since the death.10.Feeling that life is meaningless or empty without the deceased, or the belief that one cannot function without the deceased.11.Confusion about one's role in life or diminished sense of one's identity.12. Difficulty or reluctance to pursue interests or to plan for the future (eg, friendships, activities) since the loss.Systematic review increasingly challenged. 23 24There is increasing awareness of the negative impact of bereavement on older individuals, with emerging evidence that older age is associated with increased risk of PGD. 16 25t needs to be considered, that older adults are potentially more vulnerable to PGD than other age groups.Changes in physical health, loss of home and occupation and reduction in social networks can reduce the ability to adapt to loss. 24 26-29][32] In addition, the COVID-19 pandemic has particularly caused deaths (and thus bereavements) among older adults. 33 34lthough globally the fastest growing population group 35 precise estimates of the prevalence of PGD in older adults are not available, limiting attempts to develop age-specific screening and treatments. 36 37We have reviewed the literature concerning this important and growing area of patient care.

Aims
To undertake a review to identify, appraise and summarise the literature concerning the prevalence of PGD in older adults.

METHOD
We followed recommendations of the Joanna Briggs Institute 38 and Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. 39udy constructs and terms 'Older age' was defined in accordance with the UK Office of National Statistics as people aged 65 and older. 18Studies were included if data were presented for a (sub)sample with a mean age of ≥65 or presented results in age categories including ≥65 or 'older old age' group (85+). 18ifferent labels and criteria have been suggested for clinically relevant grief.The American Psychiatric Association DSM-5 uses 'persistent complex bereavement disorder' (PCBD) 7 and the WHO ICD-11 uses 'prolonged grief disorder' (PGD). 8Previous criteriasets developed by authorities in the field included 'PGD-2009' 10 and 'Complicated Grief ' (CG) 17 and are compared in the literature. 40The constructs of ICD-11 and DSM-5 are similar enough to be seen as one diagnostic entity (online supplemental material 2) as is the PGD-2009.In contrast, CG identifies more people by including a broader set of grief reactions, including, but not restricted to, PGD. 9 40 The most prominent research measure used is the 19-item Inventory of Complicated Grief (ICG), 13 which is linked to development of PGD criteria sets. 10 41 42However, not all symptoms of PGD and PCBD are reflected by ICG and its successors: further assessment tools have been developed recently in response to the introduction of PGD and PCBD as new diagnostic entities within DSM-5 and ICD-11. 43revalence is the proportion of a population who have a specific characteristic in a given time period 44 : point prevalence refers to characteristics being present at a specific time.PGD studies often report 'conditional point prevalence' in the grieving population at a certain time after the death.We report prevalence as percentage and when provided, CIs with frequencies data transformed into percentages.

Data sources and search strategy
Searches of five databases (Medline via OVID, PsycINFO via Ebsco, CINAHL via Ebsco, Cochrane Library and Web of Science) were undertaken by an information scientist (IK).We sought to identify papers which focused on prevalence of PGD and included those that assessed PGD and hence frequency data.We searched and screened existing review, overview and meta-analysis papers, other synthesis papers and perspective, opinion and guideline pieces on PGD (referred to as summary papers).
The search strategy combined ( 1) key terms of PGD, (2) prevalence or epidemiology and (3) relevant study designs including reviews for the search strategy (online supplemental materials 4 and 5 presents search terms and a search procedure example).The literature was searched from January 2009 (when the conceptualisations of PGD became more consistent, 10 enabling a focus on the more recent literature) to November 2019.All age categories were included in order not to miss possible studies reporting results in those aged ≥65.References identified were downloaded into EndNote X7 and duplicates removed.Reference searching of the included papers was undertaken.
Applied inclusion criteria: studies published in English in peer reviewed journals that provided prevalence of PGD in adults who experienced the loss of a close person through mainly a non-violent causes of death. 16Studies were required to assess PGD with standardised, validated psychometric instruments at least 6 months post loss. 8][47][48] Titles and abstracts were independently screened by two authors (SH, PT).Full text papers were screened for sample age details (AS, SH) and included if the mean age was 65 and over or if age categories included old age.Full texts were then reviewed individually by PT, SH and AS to assess whether papers enabled estimates of prevalence rates based on PGD assessment, recruitment method and study design.Study authors were contacted for further information when needed.Results were discussed, compromise over disagreements reached by consensus and inter-rater reliability for abstract and title screening was assessed.Data Systematic review extraction used a predefined piloted data extraction sheet (online supplemental material 6).
Quality assessment was undertaken independently by PT and AS using the risk of bias (RoB) approach of Hoy et al 49 for epidemiological studies and an adapted version of the standard quality assessment criteria of Kmet et al 50 for other studies (see online supplemental material 7).
The study selection and screening process is summarised in figure 1. Inter-rater agreement was evaluated using Cohen's kappa statistic 51 following the interpretation of McHugh. 52As few epidemiological studies were found, a descriptive analysis and thematic synthesis was undertaken to optimise the inclusion of the diverse literature identified, focusing on factors with a possible impact on PGD. 15 53

RESULTS
The searches identified 2059 unduplicated titles, of which 1111 abstracts were reviewed, and 135 summary papers were included for reference screening.Of 326 papers screened for sample age, 78 met the age criterium.Review of study design, recruitment method and grief assessment identified nine papers for inclusion.Screening of their references and of the references of the 135 summary papers yielded 8 and 98 additional Systematic review records, respectively, with no further papers being included in analysis.The inter-rater reliability for the title and abstract screening was excellent with a kappa value of 0.87 (95% CI 0.81 to 0.92) and, respectively 0.82 (95% CI 0.73 to 0.92). 52tudy and study population characteristics are summarised in tables 2 and 3.The following text refers to included papers using the study numbers assigned in tables 2 and 3 in square brackets.
Type of bereavement investigated varied widely, at times being undefined [1, 2, 3, 4], for example, 'Are you currently grieving?' [2], or referred to 'the loss of a significant person' [1] or 'the most significant loss in your lifetime' [4].Some focused on specific losses such as spousal bereavement [6, 7, 8, 9] or the death of a person participants had cared for [5], most commonly a spouse.Spousal/partner bereavement was most common (46.5% of all participants, 31% of studies with undefined bereavement) followed by parental loss (32%).

RoB and quality assessment
Two studies had an RoB score of 8 [1, 2] indicating moderate RoB and one had a high RoB with a score of 5 [3] due to lack of external validity (online supplemental material 9).
Three non-epidemiological studies had a summary score of above 0.9 which indicated high methodology and reporting quality, one study had a score of 0.86 and one 0.75.The one RCT scored 0.67 mainly due to blinding issues, which do not affect estimates of prevalence rates based on baseline data.
RoB assessment and quality assessment were performed by two raters (AS, PT) independently with an inter-rater reliability at a moderate level, kappa=0.73(95% CI 0.49 to 0.98) and a strong level, kappa=0.87(95% CI 0.78 to 0.978), respectively. 52

Prevalence rates of PGD
The conditional prevalence rates of PGD for older adults in the three epidemiological studies range

Table 3 Continued
Systematic review between 3.4% and 26.2%.The rates for PGD derived from six non-epidemiological studies ranged between 6% and 48.7%.Online supplemental material 10 provides a detailed description of each study including prevalence rate, study methodology and sample characteristics.

DISCUSSION
Although older adults are a fast-growing population group and exposed to bereavement more than others, high-quality large-scale population-based studies of PGD prevalence in older adults are rare.We only found three epidemiological studies providing data on prevalence rates for PGD in older adults ≥65 years and only one included data for older old adults ≥85 years.The six included non-epidemiological studies focused on older adults, with data that enabled us to calculate PGD frequency.This lack of epidemiological studies of PGD might be explained by the relatively recent introduction of PGD into diagnostic classification systems, the preceding debate about pathologising normal grief reactions and ethical concerns about conducting bereavement research in general. 55 56Disagreements about the criteria for a grief-specific mental disorder has led some to focus research on CG rather than PGD. 57 58his review can therefore only provide a first estimate of prevalence of PGD in older adults and highlights the necessity of future research.
The overall quality assessment of most included studies was satisfactory.However, while nearly all studies were characterised by good internal validity many studies had external validity issues, as found in a previous meta-analysis. 16Selection biases were frequent: only one study was based on a nationally representative sample [1].Potential non-response biases limited the generalisability of study findings with response rates as low as 39%, dropout rates high as 72% and up to 28% participants being excluded due to missing data. 59 60Only three included papers provided partial non-response analyses [1, 4, 9].
The review reflects the known pattern that response rates decrease with increasing age, 61 62 are often low when assessing the bereaved 63 and non-responses are higher among those more impaired. 64The finding of Allen et al, 65 that relatively more non-participants than participants had higher symptoms of depression at baseline raises concerns that PGD prevalence rates in studies with high non-responses might be underestimated.
Major imbalances in terms of ethnicity, study location in terms of urban vs rural areas and gender was observed.Most studies were conducted in western countries representing white Americans or Europeans what leaves other ethnicities not or under-represented.Seven of nine studies were based solely in urban regions [2, 3, 4, 5, 8, 9].If PGD prevalence varies by setting as for depression, PGD prevalence might be overestimated if based on larger city studies. 66 67All studies had more female than male participants: maybe due to women's longer life expectancy 68 or due to them experiencing more bereavement by having wider social networks, 69 70 or because they are more likely to participate in survey studies. 71Because female gender is a risk factor for developing PGD [1] it is essential to look at prevalence rates separately for women and men.
Only one epidemiological study enabled calculation of the current prevalence rate of PGD in the general population of older western urban adults [2].With a rate of 5%, this is close to the 7% prevalence of depression in the general older population, which is recognised as one of the most common mental disorders in old age by the WHO. 72onditional prevalence rates of PGD for older adults ranged from 3.4% to 26.2% in epidemiological studies, increasing to 3.4%-48.7%when including non-epidemiological studies.
The variability can not be attributed to the different criteria sets of ICD-11 and DSM-5.A recent study reported prevalence rates of 8.3% and 5.8% with overlapping confidence intervals for PBCD and PGD. 40he wide range might be explained by the observed methodological heterogeneity including variations in measures and way of defining PGD cases and by variations in study sample characteristics.
Of the five studies with higher prevalence rates (17.6%-48.75%)four [2, 5, 6, 7] used the ICG as their study measure and all defined PGD cases by applying a cut-off score.Studies using the ICG-R15 and PG-13 following an algorithm approach for detecting PGD reported prevalence rates of 9.1% or lower [1, 3, 4, 9].These findings are in line with a recent meta-analysis of 14 papers; 16 three of which [1, 2, 4] are included in our review that shows that the ICG produces higher prevalence rates.The commonly used cut-off of 25, from the original ICG validation study, 13 might not be appropriate for samples with other characteristics. 73In addition, a simple sum cut-off score compared with an algorithm scoring method is less akin to diagnostic criteria and can lead to a higher prevalence rate. 53It is problematic that for the same measure different cut-offs and algorithm approaches exist, which rarely are based on validation studies: combined with different screening questions to identify the bereaved, this might account for much of the heterogeneity of results.It is also concerning that prevalence estimates in epidemiological studies were not based on clinical interviews the gold standard for assessing mental health disorders. 74Selfreport measures might be at risk of underreporting mental health diagnosis especially in older age adults as shown for depression by Eaton et al. 75 Variation in sample characteristics (country, relationship to the deceased, place of residence) may also contribute to the wide range of prevalence rates.The lowest conditional PGD prevalence rate of 3.4% reported in this review originated from an Asian study [3] which was much lower Systematic than prevalence rates from other studies.Mental health disorder prevalences are known to be consistently lower for countries within North and South East Asia than other regions. 76This might also be valid for PGD.Many of the included studies investigated spousal bereavement.It may be easier to recruit spouses.However, spousal bereavement is known as a risk factor for complications in grief. 23 57Four studies in this review which predominantly referred to spousal bereavement showed higher prevalence rates [5, 6, 7, 8] including the highest prevalence of 48.7% [6].The latter was a study of nursing home residents, a population confronted by multiple losses and therefore more vulnerable for PGD. 77Nursing home residents, were rarely represnted in the included studies, potentially leading to an underestimation of PGD in older adults.

CONCLUSION
While some of the variation in prevalence rates might be rooted in real differences such as cultural differences, a large proportion of variance can probably be explained by differences in methodology and sample characteristics.The heterogeneity of results makes it difficult to estimate the prevalence of PGD in older adults.Based on the low RoB of Kersting's study [1] and their representative population-based sample, which is likely to represent the whole spectrum of bereavement experience, we think their estimated conditional prevalence rate of PGD in older adults of 9.1% is the best estimate for western countries available at the moment.
Contrary to the assumption that bereavement is 'less problematic' in older adults, this indicates that older adults are at least as vulnerable as other adults.Considering the possibility of current studies underestimating PGD in older adults, older adults might even have a greater risk of developing PGD.Bereavement in old age is associated with negative outcomes including weight loss, sleep disturbances and increased use of health services. 24 78 79he main limitations of this review are linked to the PGD research area being an evolving field with yet different definitions for clinically relevant grief and shortcomings in its assessment including the use of self-report measures only and measures not accounting for all criteria.Nonepidemiological studies cannot provide strong evidence in terms of prevalence rates and exclusion of non-English publications introduced a western country bias.The use of different instruments for the quality assessment of epidemiological and non-epidemiological studies, did not allow a direct comparison of their quality.Although well undertaken most studies quality was rated 'moderate' demonstrating difficulties in this area of research.
For more exact estimates of PGD among the general population of older adults more large-scale population-based studies with good external validity, from different parts of the world, are required.Ideally these studies should report results by age groups including older-old age and match the gender ratio of the respective age groups.To improve generalisability of study findings research needs to include all groups of older adults (eg, rural and institutionalised residents) and must minimise inclusion and exclusion criteria (eg, type of death). 64In addition, efforts to reduce the number of non-responses from older adults are important.Studies should report efforts to minimise non-response and by default provide non-response analysis. 60The above requires a mutual definition and tailored validated measures for clinically relevant grief.This is the joint task of clinicians and researchers.Until a consensus on the criteria set for PGD is found studies should use measures which represent all symptoms of the various PGD diagnosis and apply all their diagnostic algorithms.To help evaluate the current evidence performance of self-reports for PGD should be evaluated against clinical interviews.
It is important for policy makers to be aware that PGD is an important aspect of the mental health of older adults, particularly given the risk of ageism 21 and that bereavement support is under-represented in end-of-life care policy. 80Only a subgroup of bereaved older adults develop PGD therefore effective identification is important to allocate resources to those most in need. 16nowing the possible proportion of patients who may present with PGD will help raise healthcare professionals' awareness of the diagnosis, distinguish PGD from other possible bereavement outcomes such as depression and PTSD and foster appropriate treatment.
Other frontline workers who engage with older adults such as clergy, care home staff and the bereaved themselves must also learn about PGD.This will improve the outlook of older adults suffering from PGD by enabling early detection and treatment.
Correction notice This article has been updated since it was first published.The article type has been changed to Systematic review.

Twitter Isla Kuhn @ilk21
Acknowledgements We would like to thank the Dowager Countess Eleanor Peel Trust for their financial support.We would like to thank Dr Linda Machin, Dr Annabel Price, Dr Sarah Hoare and Margaret Johnson for their help.They all contributed to this project at various stages.
Contributors SB and PT developed the project with PT leading on the project and writing the manuscript and SB providing support throughout.IK conducted the systematic literature search.SH, AS and PT screened the literature, SH and AS extracted the data and AS and PT conducted the quality assessment.SH, AS and TQ assisted PT in interpreting the results.All authors critically reviewed the manuscript and provided final approval for submission.

Funding
43 *ICD-11 criteria were ordered by the authors analogous to the DSM-5 criteria for better comparability.DSM-5, Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition; ICD-11, International Classification of Diseases 11th Revision.

e35
Thiemann P, et al.BMJ Supportive & Palliative Care 2023;13:e30-e42.doi:10.1136/bmjspcare-2020-002845Systematic review The project was funded by the Dowager Countess Eleanor Peel Trust.The funder was not involved in planning, conducting or publishing of the research or interpreting of results.SB is part funded by the National Institute for Health Research (NIHR) Applied Research Collaboration East of England (ARC EoE) programme.The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.

Table 1
Criteria sets of prolonged grief disorder as per ICD-11 and persistent complex bereavement disorder DSM-5

Prolonged grief disorder defined by ICD-11 8 * Persistent complex bereavement disorder defined by DSM-5 7
A. Disturbance following the death of a partner, parent, child or other person close to the bereaved A. Death of a close other B. Persistent and pervasive grief response characterised by longing for the deceased or persistent preoccupation with the deceased B. Since the death, at least one of the following on most days to a clinically significant degree, for at least 12 months after the death: (1) persistent yearning for the deceased; (2) intense sorrow and emotional pain in response to the death;(3) preoccupation with the deceased; (4) preoccupation with the circumstances of the death C. Accompanied by intense emotional pain, for example,

Table 2
Sample characteristics from epidemiological studies Based on paper and personal communication with authors.DoR, dropout rate total number of participants who participated at follow-up divided by the total number who participated at baseline; EP, excluded participant rate number of participants who did not provided data or sufficient data (defined by study) divided by all participants; F, female; M, male; n≥65, participants sample mean age ≥65 years; NA, not applicable; n ND, not derivable; n, total sample; NR, not reported; PGD, prolonged grief disorder; RR, response rate total number of participants who participated divided by the total number who were eligible or contacted.
*Based on paper and personal communication with authors.CG, complicated grief; DoR, dropout rate total number of participants who participated at follow-up divided by the total number who participated at baseline; EP, excluded participant rate number of participants who did not provided data or sufficient data (defined by study) divided by all participants; F, female; M, male; n≥65, participants sample mean age ≥65 years; NA, not applicable; n b 65, bereaved sample mean age ≥65 years; ; n b , bereaved participants sample; ND, not derivable; n, total sample; NR, not reported; PGD, prolonged grief disorder; RR, response rate total number of participants who participated divided by the total number who were eligible or contacted.*