Prolonged grief in caregivers of community-dwelling dementia patients

Abstract

Objectives:

To estimate the frequency of prolonged grief disorder (PGD) in 90 primary caregivers of patients with dementia who live in the community dwelling in Milan and to identify the relationship between grief intensity (GI) and other caregiver variables; another aim was to clarify the role of the objective cognitive and functional impairment of the patients and the level of deterioration perceived by caregivers.

Design:

Cross-sectional study.

Results:

In all, 16.7% of caregivers had PGD. Caregiver variables increasing GI and the risk of developing PGD were burden, anxiety, and some sociodemographic features. The objective level of patient’s deterioration was irrelevant for PGD probability/GI, while the deterioration level perceived by the caregiver increased PGD probability and GI.

Conclusions:

Caregivers of patients with dementia may experience grief symptoms that are associated with low educational level, high level of burden and anxiety, and high perceived deterioration of their demented relatives’ cognitive and functional abilities.

Introduction

Caring for people with dementia has a substantial emotional toll on family caregivers. Prolonged care of patients with dementia is associated with somatic and physical symptoms that characterize a caregiver syndrome that culminates in high risk of developing affective disorders, ^1,2 physical illness, emotional strain, and mental health problems. ^3,4 Much attention has been paid to burden among caregivers of people with dementia, ^{5 –8} and higher level of burden predicts prolonged grief in this population along with depression and greater cognitive impairment of the care-recipient. ⁹ However, little attention has been paid to grief of caregivers of patients with dementia, although grief becomes a serious health concern for a relative so as to meet criteria for complicated grief, also referred to as prolonged grief disorder (PGD). ¹⁰

Prolonged grief disorder is a form of grief characterized by intense longing and yearning for the person who died, recurrent intrusive and distressing thoughts about his or her absence that make it difficult to concentrate and to move beyond an acute state of mourning, and inability to form other interpersonal relationships and engage in potentially rewarding activities. Formal diagnostic criteria for PGD were proposed for inclusion in the Diagnostic and Statistical Manual of Mental Disorder (Fifth Edition) ^10,11 : according to these criteria, the diagnosis of PGD requires that the bereaved person has persistent and disruptive yearning, pining, and longing for the person who died and that the distress symptoms lasted for at least 6 months while significantly impairing his or her functioning in important life domains. Complicated grief often co-occurs with other disorders such as major depression and posttraumatic stress disorder (PTSD); however, PGD is a specific condition ¹² that better describes caregivers’ reality independent of the diagnosis of depression or PTSD.

Although the PGD diagnostic criteria were developed for bereavement, they can also be applied to other significant losses. In Holly Prigerson’s work regarding the consensus criteria for traumatic grief, the authors wrote “[…] once criteria for traumatic grief in response to the death of a significant other have been formulated and tested, it would be possible to test the criteria on those grieving over other losses (eg, terminal illness, divorce),” (p. 68) and they conclude that future works could examine the extent to which this syndrome emerges with other type of losses. ¹³ Indeed some PGD research focused on family members of patients with terminal illnesses (cancer, vegetative state), finding a subgroup of caregivers with high levels of grief symptoms. ^11,14,15

The PGD literature already provided evidence that there are significant issues of loss inherent in caring for a loved one with dementia ^{16 –18} : indeed not only caregivers must cope with the stressful demands of addressing the needs of a loved one who can no longer adequately care for himself or herself but

they must also cope with the interpersonal loss of a deep relationship.

Meuser et al ¹⁹ showed that grief is a primary reaction to taking care of a person with dementia: they define grief as “the caregiver's emotional, cognitive, and behavioral reactions to the recognition of personally significant loss.” They argue that it is true grief that is relatively indistinguishable from postdeath grief in personal impact and meaning.

Noyes et al ²⁰ suggested a new descriptive grief stress model of caregiving, linking the grief process to health consequences in caregiving and clarifying the consequences of relationship losses. This model assumes that depending on the type and the significance of the losses one of several outcomes could be anticipated: caregivers who experienced relationship loss (companionship, communication, support, and relationship dynamic change) are more likely to experience grief than caregivers who primarily experience caregiving losses (personal freedom, social and work opportunities, personal health, and recreations opportunities) which are more likely to produce feelings of stress and burden.

Therefore, grief is a constant companion of caregiving of people with dementia, ²¹ because the preillness emotional bond changes and weakens with the progress of the disease. These changes create a sense of ambiguity in the relationship because despite the physical presence of the patient, she or he is psychologically absent. Such phenomenon was described by Boss ²² and Dupuis ²³ as “ambiguous loss”: painful reactions can occur, with the pain being very similar to that caused by a real mourning.

Indeed, family members of patients with dementia experience multiple losses, sometimes referred to as “triple grief”: the loss of the patient’s personhood, the loss at the time of nursing home admission, and, finally, the loss at the patient’s death. More specifically, the primary grief, experienced by the loss of the patient’s personhood before their actual bodily death, has been called “anticipatory grief” or “predeath grief.” ²⁴

Previous studies have attempted to identify predictors of PGD among caregivers of patients with dementia ^9,24 : in general, the focus was on sociodemographic factors, on the characteristics of the caregiving experience and on psychiatric morbidities. However, some issues remain unresolved, ²⁰ like, for example, the role of the functional and cognitive impairment of the care recipient in the caregiver’s grief response.

In our study, (1) we estimated the frequency of PGD predeath grief symptomatology in primary caregivers of patients with dementia who live in the community dwelling; to do so we employed a PGD scale that has been specifically devised to assess preloss symptoms ²⁵ ; (2) we studied the relationship between PGD and several caregiver variables, such as sociodemographic characteristics, emotional features, and caregiving burden; and (3) we tried to clarify the role of the patient’s cognitive and functional impairment on the caregivers’ grief.

Method

Patients

The Alzheimer Evaluation Unit (UVA) of Niguarda Ca’ Granda Hospital provides diagnosis of dementia following the Diagnostic and Statistical Manual of Mental Disorder (Fourth Edition) criteria ²⁶ and monitors the patients’ cognitive level every 6 months by means of neurological and neuropsychological batteries. We planned to recruit at least 85 caregiver–patient pairs since such sample size provides 80% power in detecting interesting predictive patterns (absolute r ≥ .3). We initially recruited 115 patients and the caregivers with whom they accessed the UVA; we then excluded patient–caregiver pairs when the caregiver accessing the UVA (1) was not the patient’s primary caregiver, (2) did not have adequate educational level (conventionally <3 years), (3) was not a native Italian speaker, or (4) did not have a link of kinship with the patient. Ninety pairs survived this selection and entered the study, approved by the local ethics committee. Patient inclusion criteria were assessed by the physician during clinical evaluation; psychological assessment of the caregivers was performed by a psychologist during the control visits of the patients.

Measures

Table 1 reports ranges and pathology directions.

Table 1.

Demographic Characteristics and Test Variable Scores (Mean, SD) of Caregivers and Patients.

	Min	Max	Mean	SD
Caregiver variables
Age, years	30	88	63.36	14.48
Education, years	1	5	3.26	0.97
Caring time (1-6)	1	6	4.64	2.02
VAS perc speed (0-10)	1	10	5.67	2.30
VAS perc severity (0-10)	1	10	6.02	2.19
CBI (0-96)	3	77	36.94	18.34
Anxiety (0-30)	4	30	15.21	6.21
Depression (0-15)	0	13	4.99	3.34
Patient variables
Age, years	43	92	76.28	8.12
Illness duration, months	4	204	47.47	40.81
MMSEcorr (0-30)	0	30	19.56	6.52
ADL (0-6)	0	6	4.90	1.34
IADL (0-8)	0	8	3.80	2.31

Abbreviations: ADL, activities of daily living; CBI, Caregiver Burden Inventory; IADL, instrumental activities of daily living; Max, maximum; Min, minimum; MMSEcorr, Mini Mental State Examination corrected; SD, standard deviation; VAS, Visual Analogue Scale.

Caregiver

Caring time was quantified according to the scale: 1 = discontinuous care (less than once a day); 2 = nighttime care; 3 = daily care, <3 hours/d; 4 = 4 to 6 hours/d; 5 = 7 to 12 hours/d; and 6 = continuous care (24 hours/d).

Anxiety and Depression Scale-Reduced Form

To assess the caregivers’ psychological status, the Anxiety and Depression Scale-Reduced Form was administered, which consists of the 15 items (range: 0-15) of the Depression Questionnaire Reduced Form ^27,28 and the 10 items (range: 10-40) of the State Anxiety Inventory-Reduced Form. ^29,30

Caregiver Burden Inventory

Burden was quantified by the Caregiver Burden Inventory (CBI), a multidimensional instrument ³¹ including 5 sections assessing different burden aspects: time dependence burden (items 1-5), developmental burden (items 6-10), physical burden (items 11-14), social burden (items 15-19), and the emotional burden (items 20-24). Caregiver Burden Inventory’s 24 items yield a score ranging from 0 to 96.

Prolonged grief disorder scale

There are 2 versions of the PGD Scale: the preloss one and the postloss one. ³² The preloss PGD Scale (PG-12) measures predeath grief symptoms and assesses criteria for PGD diagnosis. ²⁴ We used the Italian version of the preloss scale, which was translated into Italian and validated by Chiambretto et al. ²⁵ Cronbach’s α from the Italian sample (α = .88) allows clinical use of the scale. ³³ The test includes 12 items; respondents are asked to rate how often they experienced specific grief symptoms over the past month on a Likert-type scale ranging from 1 (almost never) to 5 (always); symptoms refer to yearning, bitterness, interpersonal disengagement, and a sense of meaninglessness. A respondent is diagnosed with PGD when she or he meets the following 4 criteria: she or he has experienced bereavement (event criterion, A); she or he endorses item 1 or 2 in the separation domain at a frequency of often or always (Likert-type scale 4-5; separation distress criterion, B); she or he endorses at least 5 items in the cognitive, emotional, and behavioral domain at a frequency of once a day or quite a bit (Likert-type scale 4-5; cognitive, emotional, and behavioral symptoms criterion, C; questions 3-11); and she or he has significant functioning impairment due to grief in the social, occupational, or other important areas (impairment criterion, D, item 12). Beyond such yes–no diagnostic classification, we also considered a continuous measure of grief symptoms intensity, taken from the very same items of the PG-12 test. Such a measure was obtained by summing the Likert-type scale scores of each of the item subsets corresponding to criteria B, C, and D mentioned earlier, rescaling each of the 3 sums to 0 to 1 and averaging them out. Hence, the final score, called grief intensity (GI), ranged from 0—the minimal intensity of PGD-related symptoms—to 1—the maximum possible intensity. This continuous scale will provide more powerful statistical analyses (see also Kiely et al ²⁴ and Givens et al ³² ).

Visual Analogue Scale

The Visual Analogue Scale (VAS) is usually applied in the evaluation of pain. ³⁴ Caregivers were asked to point to a graduated horizontal line (a Likert-type 0-10 scale) to judge their subjective perception of the severity of the disease of their relatives and of its relative speed of evolution.

Patients

Mini-Mental State Examination

Mini-Mental State Examination (MMSE), ^35,36 a 30-point screening test, was administered to assess the patient’s state of dementia; it evaluates various cognitive functions such as memory and orientation. The cutoff for the Italian population is 24 of 30; the raw scores were corrected for age and education ³⁶ (MMSEcorr).

Instrumental activities of daily living and activities of daily living

Instrumental activities of daily living (IADL ³⁷ ; range score: 0-8) and activities of daily living (ADL ³⁸ ; range score: 0-6) were used to estimate the degree of autonomy in basic daily living activities and in caring of one’s own person.

Statistical Analyses

We used SPSS 17.0 (Chicago, Illinois) to reduce dimensionality among predictors and perform logistic regression/general linear model (GLM) analyses accounting for effects on PGD/GI.

Results

The 90 caregivers (30 male) were most often the spouses (N = 52) or the sons/daughters (N = 33) of the patients. Table 1 reports demographic characteristics and score distributions on all measures of both caregivers and patients. In the table, the anxiety scale was rescaled from the original 10 to 40 range to 0 to 30 range. The direction of pathology is VAS, CBI, anxiety, depression toward high scores and MMSEcorr, ADL, IADL toward low scores.

Of the 90 caregivers, 15 (16.7%) were classified as having PGD according to the standard criterion. However, even without being given a formal PGD diagnosis, virtually all caregivers had some degree of grief as witnessed by their GI scores. These covered the whole 0 to 1 range (min: 0.03; max: 0.94), with a mean of 0.532 (standard deviation [SD] 0.245). Caregivers with a diagnosis of PGD had a mean GI score of 0.864 (SD 0.053); caregivers without a PGD diagnosis scored 0.465 (SD 0.212). Before studying the effects of various potential predictors on PGD/GI, we studied the structure of the correlations between them.

Correlations Between Continuous Predictors

Table 2 reports the Pearson’s correlations between 13 potential PGD predictors, 8 referring to caregivers and 5 referring to patients. Many significant correlations are of straightforward interpretation, for example, the older the caregiver , the less educated (r = −.457), the more hours a day she or he can care (she or he is more likely to be retired; r = +.563); the older the caregiver, the older the patient (the caregiver is typically the spouse, or the son/daughter of the patient; r = +.321). Such network of relationships generates many spurious correlations; for example, the (negative) correlation between caregiver Education and Caring Time (r = −.441) is likely to be an artifact of caregiver’s age: older caregivers are both more available for caregiving and have fewer years of education. In general, the high number of significant correlations raises the issue of collinearity. In order to avoid this problem, we derived a minimal set of independent predictors by means of factor analysis (see subsequently). Some correlations were interesting in their own right, like, for example, the surprisingly insignificant, or marginally significant, correlation between the objective cognitive deterioration of the patient (MMSEcorr) and the deterioration perceived by his or her caregiver (VAS; perceived speed: −.114, P = .287; perceived severity: −.22, P = .037).

Table 2.

Correlations Between Predictors.^a

	Caregiver Variables								Patient Variables
	Age	Ed	Cr T	VAS PSp	VAS PSv	CBI	Anx	Depr	Age	Ill D	MMSE corr	ADL
Caregiver variables
Ed	−.457b
	<.001
Cr T	.563b	−.441b
	<.001	<.001
VAS PSp	.137	−.204	.097
	.199	.054	.362
VAS PSv	.210c	−.098	.152	.744b
	.047	.356	.154	<.001
CBI	.185	−.118	.309b	.323b	.418b
	.081	.266	.003	.002	<.001
Anx	.153	−.193	.236c	.231c	.223c	.424b
	.150	.069	.025	.028	.035	<.001
Depr	.308b	−.277b	.395b	.210c	.230c	.658b	.589b
	.003	.008	<.001	.047	.029	<.001	<.001
Patient variables
Age	.321b	−.022	−.049	−.087	.002	.068	−.049	.130
	.002	.837	.645	.412	.984	.523	.649	.222
Ill D	.061	.003	.081	.080	.321b	.298b	.026	.195	−.023
	.571	.974	.450	.452	.002	.004	.811	.065	.827
MMSE corr	−.017	−.235c	.057	−.114	−.220c	−.121	.016	.141	−.142	−.093
	.872	.026	.592	.287	.037	.256	.880	.184	.180	.383
ADL	−.033	−.110	−.100	−.098	−.233c	−.194	−.047	−.010	−.166	−.227c	.576b
	.758	.302	.348	.357	.027	.067	.658	.923	.119	.032	<.001
IADL	−.039	.079	−.013	−.112	−.119	−.132	−.026	−.031	−.149	−.211c	.507b	.657b
	.719	.462	.903	.293	.263	.214	.808	.773	.160	.046	<.001	<.001

Abbreviations: ADL, activities of daily living; CBI, Caregiver Burden Inventory; IADL, instrumental activities of daily living; VAS, Visual Analogue Scale; Ed, education; Cr T, caring time; VAS PSp, VAS perceived speed; VAS PSv, VAS perceived severity; Anx, anxiety; Depr, depression; Ill D, illness duration.

^aEach cell reports the Pearson’s r coefficient and its 2-tailed P value.

^b <.01.

^c <.05.

Removing Collinearity Between Predictors

One way to eliminate collinearity is to use factor analysis (our STV ratio is 6.92, ie, >5). ³⁹ Principal components analysis was used as extraction method; rotation was performed by means of Kaiser-normalized Varimax, after preliminary Oblimin analysis showed that the independence assumption was tenable. Factors were extracted whose eigenvalues exceeded 1. Table 3 reports factor loadings that lead to straightforward interpretation. Five independent factors (columns in Table 3) were found to neatly underlie 13 variables (rows in Table 3). Factor 1 has very high loadings on MMSE, IADL, ADL, thus clearly reflecting the (objective) level of deterioration of the patient (pathology direction: the higher the factor value, the less deteriorated the patient); factor 2 has high loadings with caregivers’ burden, anxiety, and depression: hence we called it Caregiver’s psychophysical status (the higher the factor value, the worse the psychophysical status); factor 3 has positive loading on caregiver’s age, negative on caregiver’s education (which is known to be negatively related to age), and positive on caring time (older caregivers are more likely to be retired, thus having more time to dedicate to caregiving). Hence, we called it Caregiver’s sociodemographic features (high factor values indicate caregivers who are older, less educated, and dedicate more time to caregiving). We called factor 4 perceived deterioration: it has very high loadings on both the VAS perceived-deficit scales (the higher the factor value, the more severe the perceived deterioration). Factor 5 just reflects patient’s age (loading: +.91).

Table 3.

Factor Loadings Obtained by Factor Analysis.^a

	Factors
	1	2	3	4	5
Variables	Patient’s Objective Deterioration	Caregiver’s Psychophysical Status	Caregiver’s Sociodemographic Features	Perceived Deterioration	Patient’s Age
Caregiver variables
Age	−.043	.112	.772	.127	.385
Education	−.122	−.055	−.802	−.105	.044
Caring Time	−.071	.273	.792	−.023	−.129
VAS perc speed	−.028	.130	.097	.920	−.053
VAS perc severity	−.190	.223	.077	.882	−.030
CBI	−.183	.794	.078	.253	.000
Anxiety	.055	.748	.090	.101	−.026
Depression	.056	.876	.257	.046	.088
Patient variables
Age	−.165	.076	.046	−.058	.913
Illness duration	−.382	.297	.030	.103	−.336
MMSEcorr	.759	.079	.169	−.168	−.165
ADL	.878	−.064	.002	−.051	−.037
IADL	.838	−.005	−.091	−.001	−.008

Abbreviations: ADL, activities of daily living; CBI, Caregiver Burden Inventory; IADL, instrumental activities of daily living; perc, perceived; VAS, Visual Analogue Scale.

^a(Absolute) factor loadings > .7 are reported in italic.

Categorical Predictors

Available categorical predictors were patient's gender, caregiver's gender, patient-caregiver kinship. The first two are strongly redundant (in 52 of 90 cases the caregiver was the patient's spouse), so we only analyzed the effects of caregiver's gender and kinship (we only included sizable categories, ie, spouses and sons/daughters) on the 5 continuous factors obtained from factor analysis (see earlier) by means of GLM. Both variables influenced psychophysical status (kinship: F _1,81 = 6.145, P = .015 and gender: F _1,81 = 13.695, P < .001): female caregivers had higher levels of burden, anxiety, and depression than male caregivers; spouses had the same disadvantage with respect to sons or daughters. Not surprisingly, spouses had higher scores on sociodemographic features than sons or daughters (the former are older, less educated, and have more time to dedicate to caregiving; F _1,81 = 149.972, P < .001).

In the next sections, we used the 5 factors, together with kinship and caregiver's gender, as predictors of PGD probability (logistic regression) or GI (GLM).

Predictors of PGD Diagnosis: Logistic Regression

Table 4 reports logistic regression results. Beta values indicate increasing (+) or decreasing (−) risk of PGD diagnosis with increasing values of the predictor: for example, β = +.763 (P = .042) for perceived deterioration means that PGD risk increases with increasing perceived deterioration scores (ie, the more deteriorated the patient is perceived by the caregiver).

Table 4.

Logistic Regression Analysis (Y = Presence/Absence of PGD Diagnosis).^a

	β	SE (β)	Wald	df	P
Patient real deterioration	−0.242	0.329	0.541	1	.462
Caregiver's psychophysical status	1.305	0.484	7.276	1	.007b
Caregiver’s sociodemographic features	1.576	.699	5.085	1	.024c
Perceived deterioration	0.763	0.374	4.149	1	.042c
Patient’s age	−0.116	0.322	0.130	1	.719
Caregiver’s gender	−0.159	0.837	0.036	1	.850
Caregiver/patient kinship	1.368	1.381	0.981	1	.322
Intercept	−4.038	2.287	3.116	1	.078

Abbreviations: PGD, prolonged grief disorder; SE: standard error

^a2-tailed P values are reported.

^b<.01.

^c<.05.

Caregiver’s gender, kinship, and patient's age have no effect on PGD’s probability. “Caregiver’s psychophysical status” significantly increases the risk of developing PGD as does “caregiver’s sociodemographic features.” Interestingly, the objective level of patient’s deterioration is irrelevant for PGD probability: by contrast, the level of deterioration that is perceived by the caregiver is a significant predictor of such probability. Note that objective and perceived deterioration are unrelated with one another (see Table 2).

Predictors of GI: GLM Analyses

The more powerful analysis using GI as dependent variable fully confirmed the results obtained on PGD diagnosis (Table 5). Objective deterioration of the patient was confirmed to be an irrelevant predictor of GI, while perceived deterioration was relevant. Massive effects were observed by caregiver’s sociodemographic features and by caregiver's psychophysical status. Kinship, gender, and patient’s age had no significant effect on GI.

Table 5.

General Linear Model Analysis on Grief Intensity.^a

	SS	df	MS	F	P
Corrected model	2.399	7	0.343	10.623	<.001b
Intercept	18.486	1	18.486	573.11	<.001b
Patient’s real deterioration	0.005	1	0.005	0.146	.704
Caregiver's psychophysical status	0.753	1	0.753	23.353	<.001b
Caregiver’s sociodemographic features	0.270	1	0.270	8.384	.005b
Perceived deterioration	0.280	1	0.280	8.674	.004b
Patient’s age	0.000	1	0.000	0.008	.930
Caregiver’s gender	0.100	1	0.100	3.088	.083
Caregiver/patient kinship	0.007	1	0.007	0.209	.649
Error	2.484	77	0.032
Total	29.937	85
Corrected total	4.882	84

Abbreviations: MS, mean square; SS, sums of square.

^a2-tailed P values are reported.

^b<.01.

If gender and kinship are introduced as the only predictors, they produce a massively significant effect on GI (F _{1 ,82} = 20.719, P < .001, F _1,82 = 11.808, P = .001 respectively). These effect disappear in the overall design (Table 5), witnessing that they rather depend on differences in psychophysical status and/or sociodemographic features.

In the next, we took each significant factor and tested whether some of its composing measures (the variables with high loadings on it) have a predominant role over the others.

Subcomponents of Caregiver’s Sociodemographic Features

This factor is actually a mixture of the caregiver’s age, education, and caring time (Table 3). In a new GLM, we replaced the factor with those 3 variables. Only caregiver’s education reached significance (F _1,84 = 9.814, P = .002), with caregiver’s age and caring time being far from it (P > .144). So, the less educated the caregiver, the more intense his or her grief.

Subcomponents of Caregiver’s Psychophysical Status

Several variables compose this factor: burden, depression, and anxiety. Caregiver Burden Inventory in turn is composed of 5 subscales. So we replaced the factor with the 5 CBI subscales plus depression and anxiety in a new GLM. Only physical burden (CBI subscale; F _1,80 = 4.968, P = .029) and anxiety (F _1,80 = 4.837, P = .031) reached significance.

Subcomponents of Perceived Deterioration

Finally, we decomposed the effect of perceived deterioration in its 2 VAS components: no significant effect was found (an artifact of the high collinearity between VAS scales, r = +.744; Table 2).

Discussion

The aim of the present work was to address the presence, and measure the frequency, of PGD among caregivers of patients with dementia as well as outlying possible predictors of such a condition. In identifying predictors, we ensured that collinearity was avoided and that independent contributions from different predictors could be estimated with maximal power.

Confirming the results of the previous literature regarding grief among caregivers of nursing home residents with advanced dementia, ^24,32 we obtained formal PGD diagnosis in 16.7% of caregivers; however, PGD symptoms were ubiquitous in the population: indeed, by looking at the intensity of predeath grief symptoms (GI), this averaged at .532 (on a 0-1 scale).

Therefore, grief is a widespread companion of this type of disease ²¹ : the preillness emotional bond with the patient changes and weakens with the progress of dementia, so family members gradually lose the essence of their loved one because of the degenerative nature of the disease.

Some studies attempted to identify PGD predictors among caregivers of patients with dementia ^9,24 and found many variables to have a role, for example, sociodemographic factors, characteristics of the caregiving experience, psychiatric morbidities (anxiety, depression, etc), patient’s deterioration, and patient’s age. In our study, we introduced further variables (critically, the VAS subscales for perceived deterioration) that lead to an interesting pattern of results (exemplified in Figure 1).

Figure 1.

Diagram of between-variables “pure” relationships (see text for details).

Figure 1 reports variables in boxes, with connections illustrating significant relationships that were obtained partialling out the effect of other variables (“pure” effects). Therefore, the absence of connections between any 2 variables does not necessarily mean that they are not statistically correlated; it means that a correlation, if present, does not survive covariation with other variables and is therefore likely to be spurious. The five boxes in the central column report the independent factors that were obtained when we removed between-predictors collinearity. We failed to find significant effects by patient’s age or by patient’s objective cognitive–functional deterioration on grief (PGD/GI). By contrast, it was the level of deterioration perceived by the caregiver (which, in turn, was not correlated to objective deterioration) that increased the probability and intensity of grief. Other predictors were the caregivers’ sociodemographic features (caregivers who are older, less educated, and have more time to dedicate to caregiving, have higher levels of PGD/GI) and their psychophysical condition (the higher the levels of burden, anxiety, depression, the higher the levels of PGD/GI). Interestingly, kinship had no direct impact on grief (no direct connection to PGD/GI in Figure 1). This means that, although spouses have massively higher probability to develop PGD than sons or daughters (see “Predictors of GI: GLM analyses” above), such an effect could be attributed to sociodemographic (age, education, caregiving time) and psychophysical (burden, anxiety, depression) differences between the 2 groups rather than to the condition of being the patient’s spouse or the patient’s son or daughter per se (which was insignificant after covariation, see Tables 4 and 5; note that such conclusions rely on the (common) linearity assumption).

Similar reasoning can be proposed for caregiver’s gender: female caregivers have massively higher grief levels than male caregivers, but this effect disappears when psychophysical status is introduced as a covariate. Female caregivers have lower psychophysical integrity than male caregivers. Hence gender per se, over and above differences in psychophysical integrity, seemed to be irrelevant. In the following, the 3 main independent factors related to grief will be separately discussed.

Caregiver’s Psychophysical Status

We found that some subcomponents of this factor are particularly linked to grief, that is, the levels of physical burden (describing feelings of chronic fatigue and damage to physical health) and anxiety. This confirms previous findings of a link between grief and considerable morbidity and burden: caring for people with dementia for a long time is associated with somatic and physical symptoms, high risk of developing affective disorders, ^1,2 physical illness, emotional strain, and mental health problems. ^3,4

A prospective study of bereavement ⁹ found that caregivers who showed postloss complicated grief had shown preloss higher levels of caregiving burden. Although the relationship between grief and burden/anxiety status is solid, the direction of causality is uncertain.

Some literature has reported that grief is also statistically related to depression ^9,12 ; Kiely et al ²⁴ and Prigerson et al ¹² confirmed that PGD and major depression often co-occur; however, they maintain they are separate illnesses, with PGD better describing the mental condition of caregivers of patients with dementia. Our results confirm such a view: a superficially positive relationship between GI and depression (r = .535, P < .001) disappears when all other variables are inserted as covariates (Table 5).

Sociodemographic Features

Caregivers who are older, less educated, and have more time to dedicate to caregiving (the “sociodemographic” factor) have higher probability to develop PGD and have more intense grief. This does not seem to be related to the mere fact of being the spouse versus the son/daughter of the patient. Of the 3 component variables, education proved dominant. Kiely et al ²⁴ also found that preloss grief was significantly associated with the caregivers’ educational level. A low educational level might be related to lower ability to cope with the burden.

Perceived Deterioration

The most interesting predictor of grief is the level of deterioration of the patient’s condition as perceived by the caregiver (VAS subscales). Previous studies found that the patient’s objective cognitive impairment predicts complicated postloss grief ^9,40 ; by contrast, we found that the patient’s objective level of deterioration (as measured by means of MMSE, IADL, ADL scales) does not predict preloss PGD probability and GI. Such an apparent, and counterintuitive, independence between objective and perceived deterioration is also witnessed by their weak mutual correlation (coefficients ranged from −.098 to −.233 and were never significant, see Table 2) and by their factor analysis loadings that were on separate factors (respectively, 1 and 4 in Table 3). Hence, caregivers misperceive their relative’s deterioration, and their grief depends on such misperception and not on the objective cognitive/behavioral deficits. Of course, the nature of the available evidence does not allow us to tell in which direction causality flows: we do not know whether perceived deterioration causes PGD, or PGD causes a misperception of the patient’s conditions, or both. What we do know is that such relationship does not depend on the other factors we studied: kinship, gender, sociodemograpic, and psychophysical features had been partialled out by covariation or by design. Future research will help clarify the nature of misperception of one’s relative’s cognitive and functional damage. One possibility will be to refer to Noyes et al’s ²⁰ theoretical proposal according to which caregivers who mainly focus on the loss of a significant relationship are more likely to develop PGD, while caregivers who focus on caregiving responsibility and pragmatical issues are more likely to produce feelings of stress and burden. The features of the former phenomenon, that is, the loss of an emotionally fundamental relationship, might help understanding misperceptions of the patient’s cognitive deficit as well as its link to grief. Measuring the caregiver’s personality profiles as well as the premorbid quality of the patient–caregiver relationship might also shed light over such intricate interaction.

Conclusion

Our data confirmed that although the criteria of PGD were originally developed for bereavement, they can also apply to significant losses that are different from death ¹³ , like relationship losses in caregivers of patients with dementia. Grief seems to be more related to some of the caregivers’ variables, namely, physical burden, anxiety, low educational level, and, a result which to our knowledge has never been reported before, the perception of disease severity, which dissociated completely from the objective measures of disease. The reasons of such misperception need further research to be clarified and might be modulated by the caregiver’s emotional and personality profile as well as by the quality of the premorbid relationship with the care recipient. If grief appears to be a relevant component of the caregiving experience, and the misperceived deterioration is an important factor in the genesis of grief, further studies are needed to better investigate these interactions and to allow the development of grief-targeted interventions.