Public stigma toward prolonged grief and COVID-19 bereavement: A vignette-based experiment

Abstract

We investigated the effects of cause of death and the presence of prolonged grief disorder (PGD) on eliciting public stigma toward the bereaved. Participants (N = 328, 76% female; M_age = 27.55 years) were randomly assigned to read one of four vignettes describing a bereaved man. Each vignette differed by his PGD status (PGD diagnosis or no PGD diagnosis) and his wife’s cause of death (COVID-19 or brain hemorrhage). Participants completed public stigma measures assessing negative attributions, desired social distance, and emotional reactions. Bereavement with PGD (versus without PGD) elicited large and significantly stronger responses across all stigma measures. Both causes of death elicited public stigma. There was no interaction between cause of death and PGD on stigma. With increased PGD rates expected during the pandemic, the potential for public stigma and reduced social support for people bereaved via traumatic deaths and people with PGD requires mitigation.

Introduction

The COVID-19 pandemic is a public health crisis that has caused over six million deaths worldwide (World Health Organization [WHO], 2022). Each death leaves approximately nine family members bereaved, resulting in over 54 million bereaved individuals to date (Verdery et al., 2020). The nature of COVID-19 deaths has created unique bereavement conditions and new potential risk factors. Some of these circumstantial risk factors associated with loss include forced separation during the dying process and disrupted mourning rituals. These factors are associated with an increased risk of impaired social and work functioning, psychological distress, and poor mental health for the bereaved (Breen, Lee, et al., 2022; Lee & Neimeyer, 2022; Neimeyer & Lee, 2022; Schneider et al., in press). Therefore, early into the pandemic, the individual, community, and global experience of death, dying, and bereavement was predicted to be negatively impacted by adverse bereavement outcomes such as increased grief intensity or prolonged grief responses (Eisma et al., 2021; Mayland et al., 2020).

Prolonged grief disorder (PGD) is characterized by a pervasive, enduring preoccupation with and longing for the deceased, substantial functional impairment, and intense grief symptoms, persisting for six (WHO, 2019) or 12 months (American Psychiatric Association, 2022; Prigerson et al., 2021) beyond bereavement. People who meet PGD criteria before 12 months of bereavement are at increased risk of receiving a PGD diagnosis (Boelen & Lenferink, 2022). Before the pandemic, approximately 10% of people bereaved by natural deaths met the criteria for PGD (Boelen & Smid, 2017; Lundorff et al., 2017); however, evidence from early in the pandemic suggested an association between bereavement due to COVID-19 and an increased risk of PGD (see Eisma & Boelen, 2021; Stroebe & Schut, 2021). Given the negative impacts of PGD on functioning etc., this association warrants further investigation.

A pioneering study by Eisma et al. (2021) compared acute grief symptoms using the Traumatic Grief Inventory Self Report (TGI-SR; Boelen & Smid, 2017) between people bereaved by natural deaths (n = 1,182), unnatural deaths (n = 210), and COVID-19 (n = 49) and found that the latter group reported higher levels of acute grief than the natural death group. Multiple studies of COVID-19 bereaved participants in the United States assessed grief symptoms using the Pandemic Grief Scale (PGS; Lee & Neimeyer, 2022) and have shown high proportions of dysfunctional grief, depression, anxiety, and functional impairment (Breen et al., 2021; Lee & Neimeyer, 2022; Lee et al., 2021; Lee et al., 2022; Neimeyer & Lee, 2022; Schneider et al., in press). A recent study with COVID-19 bereaved participants from the United Kingdom assessed grief symptoms using the PGS and also showed high levels of dysfunctional grief (Breen, Mancini, et al., 2022). A study with Chinese participants bereaved by COVID-19 assessed grief symptoms using the International ICD-11 Prolonged Grief Disorder Scale (WHO, 2019) and showed that 38% met the criteria for PGD (Tang & Xiang, 2021). A comparative study from the Netherlands (Eisma & Tamminga, 2022) showed that COVID-19 deaths yielded higher grief levels than natural deaths (but not unnatural deaths), as measured by the Traumatic Grief Inventory Self Report Plus (TGI-SR+; Lenferink et al., 2022). A study of American participants assessed grief symptoms using the PGS and showed no significant difference between bereavement by COVID-19, natural death, and violent death (Breen, Lee, et al., 2022). Similarly, Gang et al. (2022) reported a substantially elevated frequency of probable PGD (67%) in a sample from the United States; with the potential risk of PGD for COVID-19 death being higher than some natural death, such as dementia, but lower than unnatural deaths.

The recognition of PGD as a psychiatric diagnosis, as it has been added to the ICD-11 and DSM-5-TR, has elicited concerns that the diagnosis will lead to stigmatization (e.g., Breen et al., 2015; Dietl et al., 2018; Ogden & Simmonds, 2014). Stigma is the co-occurrence of labeling, separation, stereotyping, discrimination, and status loss, within a context of power inequity (Link & Phelan, 2001) and has two linked components: public stigma and self-stigma (Eisma et al., 2019; Livingston & Boyd, 2010). Public stigma occurs when members of the general population support stereotypical beliefs, endorse prejudicial reactions, and discriminate against a stigmatized group (Corrigan & Watson, 2002). Self-stigma occurs when individuals internalize public stigma (Corrigan & Watson, 2002). From a mental health perspective, public stigma is associated with diminished help-seeking (Bracke et al., 2019). Stigma related to PGD may lead to less social support being received by the bereaved, which is essential for coping with bereavement (Gonschor et al., 2020).

Indicators of public stigma in relation to the bereaved include negative attributions about them, increased desired social distance, and negative emotional reactions toward them (Eisma et al., 2019). These negative effects of stigma associated with PGD can be substantial and might explain why bereaved people who experience more severe grief symptoms also report experiencing increased negative reactions from their loved ones (Johnson et al., 2009). Recent vignette-based studies with Dutch, German, and Australian participants demonstrated a robust finding that a bereaved individual who meets PGD criteria elicits public stigma, with more negative attributions, stronger desired social distance, and increased emotional reactions elicited for people with PGD (vs. without PGD; Dennis et al., 2022; Eisma, 2018; Eisma et al., 2019; Gonschor et al., 2020). Further, in determining whether the diagnostic label of PGD itself increases stigma, one study showed that public stigma did not differ for people with both PGD symptoms and diagnosis compared to people only experiencing PGD symptoms (Gonschor et al., 2020).

An understudied potential source of public stigma is whether bereaved people are stigmatized due to COVID-19 as the cause of death (Somse & Eba, 2020; WHO, 2020). Previous viral pandemics have shown that viruses and prejudice spread simultaneously (Demirtaş-Madran, 2020). This was documented in the Ebola virus epidemic, where bereaved family members were stigmatized and discriminated against (Kamara et al., 2017). A recent survey investigated the COVID-19 pandemic’s impact on bereavement care in the United Kingdom and showed that some families reported concerns about being stigmatized from having COVID-19 recorded on the deceased person’s death certificate (Pearce et al., 2021). Therefore, understanding the risks of public stigma being elicited toward people bereaved by COVID-19 is a current and important concern.

Studies of public stigma according to the cause of death reveal mixed findings. A systematic review of bereavement studies showed more guilt attributions, blame, shame, and difficulty interacting with the bereaved individual were elicited for bereavement from suicide than an accident, natural causes, and illness (Logan, Thornton, & Breen, 2018). In contrast, a recent experimental study with Dutch participants showed no difference in public stigma elicited for bereavement from suicide than natural causes or homicide (Eisma et al., 2019). Further experimental studies with participants from Australia and the United States also showed no difference in public stigma elicited for bereavement from voluntary assisted dying/medical assisted dying than long-term illness but found more public stigma elicited when the deceased was younger than older (Philippkowski et al., 2021; Singer et al., 2023). Although previous studies show mixed findings for public stigma according to the cause of death, it is unknown whether COVID-19 is a source of stigma for the bereaved and is therefore important to investigate.

The current study

Given that previous research has not investigated public stigma related to COVID-19 bereavement and PGD, and the significant potential for adverse outcomes, a better understanding of stigma being implicated in COVID-19 bereavement is important. We hypothesized an interaction effect between cause of death and PGD on eliciting public stigma, in that public stigma would be intensified for people bereaved by COVID-19 with PGD. We predicted that people bereaved by COVID-19 death would elicit more stigmatizing responses than brain hemorrhage bereavement. We further predicted that bereaved people with PGD would elicit more stigmatizing responses than people without PGD.

Method

Research design

A randomized factorial 2 × 2 (COVID-19 death vs. brain hemorrhage; PGD vs. without PGD) between groups, vignette-based experimental design was used to investigate the impact of cause of death and PGD diagnosis on levels of public stigma. Nine dimensions of public stigma were examined: attributions (competent, warm, emotionally stable, dependent, and sensitive), desired social distance, and emotional reactions (anger, prosocial, and fear).

Participants

Participant inclusion criteria included being aged 18 years and over, and literate in the English language. Participants were unpaid and recruited using convenience and snowball sampling via social media (e.g., Facebook). The information included a weblink to access the study online. An a priori power analysis (Faul et al., 2007) determined that 343 participants were needed to detect a small-to-moderate (ƒ² = 0.03) 2-way interaction at an alpha level of 0.05, based on the effect size for PGD/no PGD reported by Dennis et al. (2022). Our final sample included 328 participants, falling shy of the a priori sample size, but sufficient to detect moderate-to-large effects. The sample was predominantly female (76%), ranging from 18 to 82 years old (M = 27.55, SD = 14.55), and most were university-educated (71%), and mainly living in Australia (75%) and the United States (22%) as shown in Table 1. Participants who were significantly impacted by grief were mainly residents of Australia (74%) and the United States (24%); had received COVID-19 diagnosis lived mostly in the United States (83%) and Australia (17%); and had knowledge of others with COVID-19 lived in Australia (52%) and the United States (47%).

Table 1. Sociodemographic characteristics of participants (N = 328) by groups.

Characteristic	COVID/PGD	Brain/PGD	COVID/NoPGD	Brain/NoPGD	Total
	n = 89	n = 78	n = 77	n = 84	n (%)
Age (years), M (SD)	28.02 (14.36)	29.19 (15.66)	26.0 (14.38)	26.95 (13.89)	27.55 (14.55)
Gender, n (%)
Women	70 (79)	60 (77)	54 (70)	65 (77)	249 (76)
Men	19 (21)	18 (23)	23 (30)	19 (23)	79 (24)
Education, n (%)
Primary	0 (0)	1 (1)	1 (1)	1 (1)	3 (1)
Secondary	15 (17)	17 (22)	16 (21)	10 (12)	58 (18)
Vocational	9 (10)	12 (16)	5 (6)	9 (11)	35 (10)
Undergraduate	30 (34)	29 (37)	26 (34)	39 (46)	124 (38)
Postgraduate	35 (39)	19 (24)	29 (38)	25 (30)	108 (33)
Employment, n (%)
Home duties	5 (6)	3 (4)	3 (4)	3 (4)	14 (4)
Full-time	36 (40)	36 (46)	47 (61)	38 (45)	157 (48)
Part-time	25 (28)	21 (27)	14 (18)	20 (24)	80 (24)
Retired/volunteer	11 (12)	11 (14)	3 (4)	7 (8)	32 (10)
Student	8 (9)	4 (5)	8 (10)	9 (11)	29 (9)
Unemployed/disability	4 (5)	0 (0)	0 (0)	1 (1)	5 (2)
Unemployed	0 (0)	1 (1)	2 (3)	2 (2)	5 (1)
Other	0 (0)	2 (3)	0 (0)	4 (5)	6 (2)
Country, n (%)
Australia	67 (75)	56 (72)	61 (79)	63 (75)	247 (75)
United States	21 (24)	19 (24)	15 (20)	18 (21)	73 (22)
Other	1 (1)	3 (4)	1 (1)	3 (4)	8 (3)
COVID diagnosis, n (%)
No, myself	88 (99)	74 (95)	77 (100)	83 (99)	322 (98)
Yes, myself	1 (1)	4 (5)	0 (0)	1 (1)	6 (2)
No, do not know others	46 (52)	45 (58)	45 (58)	56 (67)	192 (58)
Yes, know others	43 (48)	33 (42)	32 (42)	28 (33)	136 (42)
Bereavement status, n (%)
Not bereaved	18 (20)	15 (19)	15 (19)	16 (19)	64 (19)
Bereaved, not impacted	33 (37)	29 (37)	27 (35)	31 (37)	120 (37)
Bereaved, impacted	38 (43)	34 (44)	35 (46)	37 (44)	144 (44)

Note. COVID: COVID-19 death; PGD: Prolonged Grief Disorder present; NoPGD: Prolonged Grief Disorder absent; Brain: brain hemorrhage death.

Measures

Demographic questions

Demographics included participants’ gender, age, highest education level, main employment, country of residence, personal bereavement experience, personal diagnosis with COVID-19, and knowing someone diagnosed with COVID-19.

Vignettes

Four fictional vignettes were adapted from Eisma et al. (2019), describing a fictional character Robert who was bereaved by his wife, Margaret (see Table 2). Vignette conditions differed by Robert’s PGD diagnosis (with PGD vs. without PGD) and Margaret’s cause of death (COVID-19 vs. brain hemorrhage). As per Eisma et al. (2019), the vignettes containing a PGD diagnosis included two PGD criteria (grief continuing longer than the six months minimum and daily functioning impairment) and four PGD symptoms (struggling to accept the death, longing for the deceased, struggling to engage in activities, and guilt; Eisma et al., 2019; WHO, 2019). Spousal bereavement was selected due to its regular occurrence, social acceptance of strong grief reactions (Logan, Thornton, & Breen, 2018), and use in other grief and stigma research (Dennis et al., 2022; Eisma, 2018; Eisma et al., 2019; Gonschor et al., 2020). The vignette’s time since death was ten months to create distance from the six months minimum grief period required for PGD diagnosis in the ICD-11 (WHO, 2019). Although some other studies have used two years, this period could be considered an anniversary reaction by participants (see Eisma, 2018).

Table 2. Content of vignettes varying conditions of cause of death and PGD.

Vignette condition	Vignette description
PGD diagnosis/COVID-19 death	Robert is aged seventy years old, and his wife Margaret died ten months ago due to COVID-19. Robert is not functioning well at work or at home, and is finding being bereaved extremely difficult. Since Margaret’s death, he has been yearning strongly for her. Robert has difficulty accepting her death, and has been experiencing strong feelings of guilt. Robert has withdrawn socially and is not engaging in many activities. He has been diagnosed with Prolonged Grief Disorder from a mental health professional, due to this behavior.
PGD diagnosis/brain hemorrhage death	Robert is aged seventy years old, and his wife Margaret died ten months ago due to a brain hemorrhage. Robert is not functioning well at work or at home, and is finding being bereaved extremely difficult. Since Margaret’s death, he has been yearning strongly for her. Robert has difficulty accepting her death, and has been experiencing strong feelings of guilt. Robert has withdrawn socially and is not engaging in many activities. He has been diagnosed with Prolonged Grief Disorder from a mental health professional, due to this behavior.
No PGD/COVID-19 death	Robert is aged seventy years old, and his wife Margaret died ten months ago due to COVID-19. Although Robert was extremely sad after Margaret’s death and strongly yearned for her, he has learnt to live with her death. Robert is now able to function well at home and work. He has further accepted Margaret’s death, is able to participate in meaningful activities and is experiencing less feelings of guilt.
No PGD/brain hemorrhage death	Robert is aged seventy years old, and his wife Margaret died ten months ago due to a brain hemorrhage. Although Robert was extremely sad after Margaret’s death and strongly yearned for her, he has learnt to live with her death. Robert is now able to function well at home and work. He has further accepted Margaret’s death, is able to participate in meaningful activities and is experiencing less feelings of guilt.

Stigma

Participants completed measures of three stigma components: attributions (competent, warm, emotionally stable, dependent, sensitive), desired social distance, and emotional reactions (anger, prosocial emotion, fear).

Attributions

The attribution items were previously used by Eisma (2018) and Eisma et al. (2021) based on research of depression stigma (Angermeyer & Matschinger, 2003). Participants were asked to assess the characteristics associated with Robert. Using a 4-point Likert scale ranging from 1 (completely disagree) to 4 (completely agree), participants indicated the extent to which each attribute (competent, warm, emotionally stable, dependent, sensitive) represents Robert (e.g., “Robert is competent”), with higher scores indicating a higher level of the attribute.

Desired social distance

The 7-item Social Distance Scale (SDS; Link et al., 1987) measures participants’ desired social distance from an individual by asking participants how willing they are to interact with them. In the present study, items referred to Robert, and were adapted to reflect more practical scenarios (e.g., “having your children marry someone like the person in the story” was changed to “having Robert marry a family member”). A 4-point Likert scale, ranging from 1 (definitely willing) to 4 (definitely unwilling) was used with higher scores indicating a larger desired social distance from Robert. The SDS is internally consistent, with a Cronbach’s alpha of 0.85 in previous research (Eisma et al., 2019) and 0.84 in this study.

Emotional reactions

Emotional reactions toward Robert were measured using an adaptation of von dem Knesebeck et al.’s (2017) 9-item scale, which assesses common reactions to individuals with mental illness (anger, fear, and prosocial emotions; Angermeyer & Matschinger, 2003). Eisma et al.’s (2019) adaptation involved adding five items and removing one to improve construct validity. This 13-item version comprises three sub-scales: anger (four items, e.g., “I feel irritated”), prosocial emotion (four items, e.g., “I take pity”), and fear (five items, e.g., “I feel tense”). Each item is rated on a 4-point Likert scale, from completely agree (4) to completely disagree (1), with higher scores indicating stronger reactions. Internal consistency was demonstrated by Eisma et al. (2019) for each subscale (anger, α = 0.82; pro-social emotion, α = 0.75; and fear, α = 0.85). Cronbach’s alpha was also good for each subscale in the current study (anger, α = 0.80; prosocial emotion, α = 0.81; and fear, α = 0.88).

Manipulation check

A post-manipulation check was used to assess participants care with reading the questions and including “What was the cause of death for Robert’s wife?” and “Was Robert diagnosed with a mental health disorder?”.

Procedure

We obtained ethics approval from the Curtin University Human Research Ethics Committee [HRE2021-079]. Interested participants were directed to the Qualtrics survey, which first displayed the participant information sheet before asking participants to provide informed consent via a check box. Participants were randomly assigned to read one of four fictional vignettes before being asked to complete three measures in relation to the vignette character (Robert). Finally, participants were asked to complete the demographic questionnaire and manipulation check questions. Participation took approximately 10 minutes to complete the survey.

Data analyses

The Statistical Packages for the Social Science (SPSS; Version 27) was used for all analyses. We collected 425 participant responses from 17 July to 19 September 2021 (9 weeks). The overall missing data was 23% and missing completely at random, χ² (df, N = 213) = 210.54, p = .535 (Little, 1988). Consequently, listwise case deletions (N = 97) were conducted where we could not perform imputation methods (Tabachnick & Fidell, 2013) for cases: without any data (n = 71); missing significant single-item questionnaires and/or whole missing constructs (n = 24); and, although gender is not binary, there were too few participants indicating that they were not women and men to include in comparisons (n = 2).

Normality was assessed by visually inspecting the histograms and boxplots, and using Shapiro-Wilk tests. The histograms and boxplots indicated univariate outliers present for some of the variables, though none were extreme (>3 SDs) and were within the possible range of scores and therefore retained. As anticipated, the Shapiro-Wilk results indicated that the data distribution differed significantly from a perfect normal distribution (as p < .05). However, visual inspection of the histograms and boxplots, in conjunction with mostly acceptable skewness and kurtosis statistics (z scores between −1.96 and + 1.96), and the large sample size suggested that the results of the MANOVA could be interpreted meaningfully (Pituch & Stevens, 2016). Also, as a measure of best practice, separate non-parametric tests (i.e., the Kruskall-Wallis ANOVA) were performed on those variables that were the most skewed and/or kurtosis, with these results revealing a similar pattern of findings and statistical significance, further supporting the interpretability of the MANOVA. The presence of multivariate outliers were indicated using the maximum Mahalanobis Distance; however, Cook’s Distance was less than 1, indicating a low influence. Multicollinearity was found to be appropriate by examining the dependent variables (DVs) correlations. The scatterplots indicated the DVs relationships were roughly linear. Box’s M was significant at α = <.001; however, a MANOVA analysis is robust against this violation with larger participant group samples. Levene’s Test of Equality of Error Variances was based on the median due to having a reduced bias from outliers (Field, 2018) and was non-significant for all DVs except pro-social, indicating equality of variance (Field, 2018).

A two-way multivariate analysis of variance (MANOVA) was used to test the hypotheses with a two-tailed significance test (α = 0.05). Cause of death (COVID-19, brain hemorrhage) and PGD (with PGD, without PGD) were included as independent variables, and attributions (competent, warm, emotionally stable, dependent, and sensitive), desired social distance, and emotional reactions (anger, fear, and prosocial) were the DVs. Significant MANOVA results were followed up with ANOVAs. Partial eta squared ($n_p^2$) was used to determine the effect sizes, with 0.01 considered a small effect, 0.06 a medium effect, and 0.14 a large effect (Cohen, 2013). Confidence Intervals (95%) were used to assist with interpreting results.

Results

Preliminary findings

Analysis of variance (ANOVAs) and chi-square tests of independence (χ²) were used to assess successful participant randomization. Between the four vignette groups, no significant differences were found based on: gender, χ² (3, N = 328) = 1.92, p = .590; age, F(3, 324) = 0.70, p = .553; education level, χ² (12, N = 328) = 12.83, p = .382; employment, χ² (21, N = 328) = 30.39, p = .085; country of residence, χ² (9, N = 328) = 8.02, p = .532; bereavement experience, χ² (6, N = 328) = 0.17, p = 1.000; COVID-19 diagnosis (self), χ² (3, N = 328) = 6.60, p = .086; COVID-19 diagnosis (know others), χ² (3, N = 328) = 4.03, p = .258. These results indicate successful randomization. The manipulation check showed most participants correctly identified cause of death (88%) and PGD diagnoses (74%). As per convention (Dennis et al., 2022; Eisma, 2018), we reported the full sample’s findings as there were no differences in the full and partial samples’ resulting patterns. Normality assumptions were assessed for the nine DVs and were found to be appropriate. Statistical assumptions relevant to the MANOVA were assessed and appropriately met (Tabachnick & Fidell, 2013).

Main findings

The MANOVA identified a non-significant interaction between cause of death and PGD diagnosis, indicating the effect of cause of death did not differ depending on PGD diagnosis. The means and standard errors for the four conditions are listed in Table 3. A non-significant main effect for cause of death was found, F(9, 321) = 1.61, p = .319, $n_p^2$ = .03, λ = 0.97, indicating COVID-19 death did not elicit more public stigma than brain hemorrhage death.

Table 3. Means, standard errors, and confidence intervals for stigma variables (N = 328).

Measure	COVID/PGD		Brain/PGD		PGD total	COVID/NoPGD		Brain/NoPGD		NoPGD total
	M	(SE)	M	(SE)	M (SE) [95% CI]	M	(SE)	M	(SE)	M (SE) [95% CI]
Attributions
Competent	2.87	(0.051)	2.81	(0.056)	2.84 (0.038) [2.76, 2.92]	2.99	(0.056)	3.16	(0.054)	3.07 (0.039) [2.99, 3.15]
Warm	2.96	(0.056)	2.91	(0.060)	2.93 (0.041) [2.85, 3.02]	2.90	(0.061)	2.99	(0.058)	2.94 (0.042) [2.86, 3.03]
Emotionally stable	2.23	(0.057)	2.18	(0.061)	2.20 (0.042) [2.12, 2.28]	2.94	(0.062)	3.05	(0.060)	2.99 (0.043) [2.91, 3.08]
Dependent	2.54	(0.062)	2.48	(0.068)	2.51 (0.046) [2.42, 2.60]	2.27	(0.068)	2.32	(0.065)	2.30 (0.047) [2.20, 2.39]
Sensitive	3.10	(0.058)	3.20	(0.063)	3.15 (0.043) [3.07, 3.23]	2.71	(0.064)	2.83	(0.061)	2.77 (0.044) [2.69, 2.86]
Social distance scale	14.85	(0.314)	14.52	(0.340)	14.68 (0.231) [14.23, 15.14]	13.34	(0.345)	12.50	(0.330)	12.92 (0.239) [12.45, 13.39]
Emotional reactions
Anger	5.58	(0.199)	5.68	(0.216)	5.63 (0.147) [5.34, 5.92]	5.91	(0.219)	5.58	(0.210)	5.75 (0.152) [5.45, 6.04]
Prosocial	12.33	(0.231)	12.20	(0.251)	12.27 (0.170) [11.93, 12.60]	9.47	(0.254)	8.58	(0.243)	9.03 (0.176) [8.68, 9.37]
Fear	8.76	(0.272)	8.73	(0.295)	8.75 (0.200) [8.36, 9.14]	7.83	(0.298)	7.45	(0.286)	7.64 (0.207) [7.24, 8.05]

Note. 95% Confidence intervals within square brackets. COVID: COVID-19 death; PGD: with Prolonged Grief Disorder; NoPGD: without Prolonged Grief Disorder; Brain: brain hemorrhage death.

A significant main effect for PGD diagnosis was found, indicating more public stigma is elicited for those with PGD than without PGD, F(9, 321) = 44.83, p = < .001, $n_p^2$ = .56, λ = .44, indicating that the vignette with the person with PGD elicited more stigma than the person without PGD. Follow-up univariate ANOVAs indicated that Robert was perceived to be: less competent, F(1, 329) = 17.94, p = < .001, $n_p^2$ = .05; less emotionally stable, F(1, 329) = 172.97, p = < .001, $n_p^2$ = .35; more dependent, F(1,329) = 10.36, p = .001, $n_p^2$ = .03; and more sensitive, F(1,329) = 36.91, p = < .001, $n_p^2$ = .10 in the vignettes with PGD compared to without PGD. Likewise, the vignette describing Robert with PGD elicited more social distance, F(1, 329) = 28.23, p = < .001, $n_p^2$ = .08; more prosocial emotions, F(1, 329) = 175.41, p = < .001, $n_p^2$ = .35; and more fear, F(1, 329) = 14.80, p = < .001, $n_p^2$ = .04 than the vignette without PGD. No differences were observed for warm, F(1, 329) = .02, p = .893, $n_p^2$ = .00; or anger, F(1, 329) = .29, p = .589, $n_p^2$ = .00.

Post hoc analyses

Half of the sample (44%) reported having experienced a significantly impacting bereavement experience, which might have affected warm ratings. We tested differences between the reported warm mean scores according to bereavement status (bereaved vs. non-bereaved) with a two-tailed t-test (α = 0.05). The results of the t-test comparing the difference in warm scores reported by participants who were bereaved (n = 264; M = 2.95, SD = .54) to those who were not bereaved (n = 64; M = 2.92, SD = .51) was not significant, t(326) = .388, p =. 698, two-tailed, 95% CI of the mean difference [−.175, .117]. As a measure of best practice, we conducted a separate non-parametric test (i.e., the Mann–Whitney U) due to violated normality assumptions and kurtosis; both analyses revealed the same result, supporting the interpretability of the t-test.

Discussion

Our overall study aim was to investigate the effects of cause of death and PGD on eliciting public stigma for the bereaved. Consistent with our expectations, participants reported more stigma toward a bereaved person with PGD than without. The significant main effect detected for PGD diagnosis on public stigma supports previous research findings (Dennis et al., 2022; Eisma, 2018; Eisma et al., 2019; Gonschor et al., 2020). Specifically, participants assessed the bereaved person with PGD as being less competent and emotionally stable, but more dependent and sensitive than the bereaved person without PGD. Participants also expressed increased desired social distance, more prosocial emotions, and more fear toward the person with PGD than without PGD. These results indicate the public may experience mixed feelings of wanting to help the bereaved while also feeling fearful of them and wanting to avoid them (Dennis et al., 2022).

However, our PGD findings contrast with previous research in two ways. First, previous research showed bereaved people with PGD elicited significantly more anger than people without PGD (Dennis et al., 2022; Eisma, 2018; Eisma et al., 2019; Gonschor et al., 2020). Our sample size was adequate and similar to previous research that detected a moderate effect (Eisma, 2018; Eisma et al., 2019). In our sample, anger toward the bereaved was elicited equally irrespective of PGD, which aligns with previous research showing that bereaved people elicited anger without PGD present (Philippkowski et al., 2021). Second, unlike previous studies (Dennis et al., 2022; Eisma, 2018; Eisma et al., 2019; Gonschor et al., 2020), bereaved people with PGD were not perceived by our sample as significantly less warm. Given that our sample size was adequate and similar to or larger than previous research that achieved a moderate effect (Dennis et al., 2022; Eisma, 2018; Eisma et al., 2019), the contrasting results could have been because almost half of our sample reported experiencing a significantly impacting bereavement experience. Other studies (e.g., Dennis et al., 2022) reported a smaller proportion of the sample as significantly impacted by bereavement (28%) or did not report whether participants’ bereavement experience was significantly impacting (Eisma, 2018; Gonschor et al., 2020). Previous research has shown that personal experience with bereavement and interacting with bereaved individuals can promote an understanding of bereavement (Costa et al., 2007; Jorm & Oh, 2009), but has no effect on intentions to provide support to bereaved people (Logan, Thornton, Kane, et al., 2018). However, our analysis showed no difference between for warm mean scores according to bereavement status (bereaved experience vs. nonbereaved experience), indicating the limited impact that personal bereavement history has on social support variables following bereavement (Logan, Thornton, Kane, et al., 2018).

Overall, our findings suggest that bereaved people with PGD are significantly more likely to be stigmatized than people without PGD. A major consequence of this stigmatization relates to support and treatment; stigma may prevent potential supporters from helping bereaved individuals and limit support seeking (Corrigan, 2005). The latter is particularly noteworthy in that people with PGD symptoms are less likely to seek counseling or mental health services than other grieving persons (Lichtenthal et al., 2011). This finding contributes to a growing body of research across countries and languages that the role of PGD in eliciting stigma is a robust effect that is generalizable across Western cultures (Dennis et al., 2022; Eisma, 2018; Eisma et al., 2019; Gonschor et al., 2020).

In contrast to our expectations, we found no significant interaction between cause of death and PGD diagnosis, and stigma did not differ based on cause of death (COVID-19 and brain hemorrhage). However, the mean stigma scores in our sample showed high levels of stigma for both death causes (COVID-19 and brain hemorrhage). Although COVID-19 deaths are potentially stigmatizing (Pearce et al., 2021) and traumatic in nature (Doka, 2021), brain hemorrhage deaths may also be experienced as traumatic due to being unpredicted and sudden (Penman et al., 2014). COVID-19 and brain hemorrhage deaths may have been perceived as similarly traumatic causes of deaths.

Interestingly, a comparison of our COVID-19 stigma results with those reported in a previous study (Eisma et al., 2019) shows that participants in this study reported the strongest stigma mean scores. Specifically, our stigma means were higher than those reported by Eisma et al. (2019), using the same stigma measures, such that, compared to people bereaved by suicide, stroke, or homicide, those bereaved by COVID-19 were perceived as less warm, and emotionally stable, but more dependent, and elicited more anger and fear; there were no differences in attributions of sensitivity or prosocial emotions. Together, these comparisons suggest that, although people bereaved by COVID-19 did not elicit significantly stronger stigma than brain hemorrhage deaths in our sample, both causes elicited higher stigma levels than other causes of death studied previously (i.e., suicide, stroke, homicide; Eisma et al., 2019). It is possible that the unique pandemic conditions contribute to public stigma of the bereaved, irrespective of whether COVID-19 is the cause of death. Currently, however, no other COVID-19 bereavement and public stigma research exists for comparison.

Strengths, limitations, and future directions

Strengths of this research include experimentally investigating COVID-19 bereavement and PGD eliciting public stigma and contributing to the body of research. Our study used a robust experimental design, participant randomization to vignettes, participant manipulation checks, and multiple comprehensive stigma measures. Limitations of the study included participants being mainly female, highly educated, and young, potentially limiting the generalizability of the findings. However, previous research conducted in different cultures with varying sexes/genders, education, and ages showed similar results (Dennis et al., 2022; Eisma, 2018; Eisma et al., 2019; Gonschor et al., 2020; Singer et al., 2023). Future research could use other recruitment methods to recruit a broader participant range and extend the generalizability of the findings, including non-Western samples. For example, one study from mainland China showed that nearly one-quarter were concerned that labeling grief as a disorder will subject bereaved persons to public stigma (Tang et al., 2020). Despite previous evidence that public stigma does not differ for people with both PGD symptoms and diagnosis compared to people only experiencing PGD symptoms (Gonschor et al., 2020), it is possible that the PGD label itself was stigmatizing, and future research could address the impact of the diagnosis from the widower’s struggle, whether diagnosed or not. There is evidence that emphasizing underlying health conditions of the deceased could elicit stigma (Gronholm et al., 2021), and future studies could investigate if such conditions elicit stigma. Although using vignette-based experiments for stigma approximates real-life responding, it is not well-established how generalizable the results are to public stigma in real situations (Dennis et al., 2022).

Conclusions

Results indicate COVID-19 deaths do not elicit significantly more public stigma for the bereaved than brain hemorrhage deaths, but both elicit high levels of stigma. Further, this study supports the robust finding that PGD elicits public stigma. With an increased risk of elevated PGD rates expected during the pandemic, further research is needed to understand factors that elicit PGD stigma and enable timely support for the bereaved in need.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.