The Longitudinal and Cross-Sectional Associations of Grief and Complicated Grief With Sleep Quality in Older Adults

ABSTRACT

Objective/Background: About 15% of grievers experience complicated grief. We determined cross-sectional and longitudinal relations of grief and complicated grief with sleep duration and quality in the general population of elderly adults. Participants: We included 5,421 men and women from the prospective population-based Rotterdam Study. Methods: The Inventory of Complicated Grief was used to define grief and complicated grief. We assessed sleep with the Pittsburgh Sleep Quality Index. Results: After 6 years, 3,511 (80% of survivors) underwent the follow-up interview. Complicated grief was cross-sectionally associated with shorter sleep duration and lower sleep quality. These associations were explained by the presence of depressive symptoms. The prospective analyses showed that sleep duration and sleep quality did not decline further during follow-up of persons who experienced grief or complicated grief. Conclusion: In community-dwelling, middle-aged and older adults, persons with normal and complicated grief had both a shorter sleep duration and a lower sleep quality, mainly explained by depressive symptoms. However, prospective analyses showed that sleep quality and sleep duration do not decline further in persons with normal grief and complicated grief.

The death of a loved one is a common life event in older adults (Boelen & Hoijtink, 2009; Boelen & van den Bout, 2008; M. K. Shear, 2015). Very few persons make it through old age without having to cope with this kind of loss, once or several different times. The loss of a partner, child, parent, or close family member can be very distressing (Monk, Germain, & Reynolds, 2008). However, even if it is experienced as a traumatic event, after a delimited period of grief, the majority of people recover. An estimated 15% of bereaved people continue to grieve for an extended period; they experience disbelief and are preoccupied by the deceased (Prigerson et al., 1995). This state is known as complicated grief (Prigerson et al., 2009). Complicated grief is an important mental health issue for the aging population, affecting social functioning and well-being (Newson, Boelen, Hek, Hofman, & Tiemeier, 2011). However, our knowledge about complicated grief is limited. Previous studies suggest that symptoms of complicated grief are distinct from those of depression and anxiety and have incremental validity predicting impairments in social and interpersonal daily functioning (Boelen, van de Schoot, van den Hout, de Keijser, & van den Bout, 2010; Newson et al., 2011; Prigerson & Jacobs, 2001). In addition, the severe emotional strain of the loss of a loved one can trigger profound changes in lifestyle. These changes often induce reductions in financial security, perceived personal safety, and freedom of action. All of these facets of grief could lead to changes in sleep patterns. Several studies (Hall et al., 1997; Kowalski & Bondmass, 2008; Monk, Begley, et al., 2008) suggest that grief is associated with significant sleep impairment. However, our knowledge regarding the associations of complicated grief with sleep is limited, as only a few studies with small sample size and a cross-sectional design have been conducted (Boelen & Lancee, 2013; Germain, Caroff, Buysse, & Shear, 2005; Maytal et al., 2007; Monk, Begley, et al., 2008; Purebl, Pilling, Konkoly, Bodizs, & Kopp, 2012; Spira, Stone, Beaudreau, Ancoli-Israel, & Yaffe, 2009). An exploratory study of the effects of complicated grief on sleep by McDermott et al. (1997) conducted analyses on 65 bereaved persons. The results showed mild subjective sleep impairment is associated with complicated grief, but no effect was detected using the electroencephalographic sleep measures. Germain et al. (2005) evaluated the severity of sleep disturbances in a group of 105 adults meeting criteria for complicated grief. They showed an association of complicated grief with an overall poor sleep quality. Comorbid depression (Adrien, 2002; Germain et al., 2005; Hall et al., 1997; Maytal et al., 2007; Monk, Begley, et al., 2008; Nutt, Wilson, & Paterson, 2008; Purebl et al., 2012; Spira et al., 2009), but not posttraumatic stress disorder, further worsened sleep quality.

Taking into account the lack of high-powered longitudinal studies in normal populations of elderly adults, we aim to determine whether in adults aged 55 years and above, grief or complicated grief was related to sleep duration and sleep quality, cross-sectionally and longitudinally.

We hypothesized that if studied cross-sectionally, persons with grief and complicated grief have shorter sleep duration and a lower sleep quality than persons who did not experience grief due to the stress that death of the loved one brings to person’s life. Second, we hypothesized that complicated grief remains a risk factor for further decline of sleep duration and poor sleep quality over time due to coping mechanisms that may not always be successful.

Methods

Settings and Study Population

This study was embedded in The Rotterdam Study, an ongoing prospective cohort of older adults designed to examine the occurrence and risk factors of chronic diseases. The study design and objectives are described in Hofman et al. (2013). The Rotterdam Study comprises two cohorts, which were combined in the current analysis. Between 2002 and 2005, complicated grief and sleep quality were assessed during a home interview, referred to as baseline. The baseline interview was conducted in 5,481 participants. Of these participants, 60 persons did not complete the grief or sleep questionnaire. This left 5,421 participants with assessment of grief and sleep characteristics for cross-sectional analysis. In part of the follow-up examination (2009– 2011), both components of sleep (duration and quality) were assessed at the research center. After an average of 6.33 years (SD = 0.42), 3,511 (80%) of the 4,601 surviving participants underwent the follow-up interview for sleep duration, and 3,003 (71%) for sleep quality. The Rotterdam Study has been approved by the Medical Ethics Committee of the Erasmus Medical Center (Erasmus MC) and by the Ministry of Health of the Netherlands, implementing the Wet Bevolkingsonderzoek: ERGO (Erasmus Rotterdam Gezondheid Onderzoek; Population Studies Act: Rotterdam Study). All participants provided written informed consent to participate in the study and to obtain information from their treating physicians.

Assessment of Complicated Grief

Complicated grief was diagnosed at the baseline examination (2002 and 2005), with a Dutch version of the 17-item Inventory of Complicated Grief (ICG) originally constructed by Prigerson et al. (which contains 19 items). First, participants were asked if they were currently grieving. If a positive answer was received, the ICG was administered, but if not, they were categorized as nongrievers (the reference group). The ICG is the most widely used instrument to measure complicated grief. Questions represent symptoms of complicated grief such as those in the most recent proposed criteria for the condition suggested by (Prigerson et al., 2009). Some of the symptoms include intense yearning for the lost person, anger over the death, distrust and detachment from others as a consequence of the death, survivor guilt, and loneliness. The measure has high internal consistency and convergent and criterion validity and it is considered the gold standard for measurement of complicated grief in older adults. The inventory is shown to represent a single underlying construct of complicated grief (Boelen & Hoijtink, 2009). The Dutch version of the Inventory of Complicated Grief contains 17 items and has been previously validated (Boelen et al., 2003). These 17 questions were asked and responses were provided on a 5-point scale to reflect an increase in severity (0-never, 1-seldom, 2-sometimes, 3-often, 4-always). In the current study one item from the original inventory, “I feel bitter over this person’s death,” was removed from the original ICG because a pilot study revealed that this sentiment had the same meaning within the Dutch language as the included item, “I feel anger over this person’s death.” Two further items (relating to seeing and hearing the deceased) were collapsed into one due to their similarity and to a pilot study indicating these symptoms were low in frequency and often overlapped (“I hear the voice of, or see, the person who died”). Several studies give further details on the interpretation of ICG (Boelen et al., 2010; Newson et al., 2011; Prigerson & Jacobs, 2001). We divided all interviewed participants into nongrievers (reference group), normal grievers, and complicated grievers. Complicated grief symptoms were assessed as present among participants who scored equal or greater than 22 on the ICG score and grieved longer than 6 months (Newson et al., 2011; Saavedra Perez et al., 2015).

Assessment of Sleep

Sleep duration and sleep quality were measured with the Pittsburgh Sleep Quality Index (PSQI), a self-reported questionnaire (Buysse, Reynolds, Monk, Berman, & Kupfer, 1989). The PSQI consists of 19 self-rated questions. Questions are grouped into seven component scores, each weighted equally on a 0–3 scale. The seven component scores are then summed to yield a global PSQI score, which is used in all further analyses. This score has a range of 0–21; higher scores indicate worse sleep quality. The seven components are subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medications, and daytime dysfunction. In the current study, we used total sleep time in hours to indicate sleep duration, and a total score of PSQI to indicate sleep quality. Finally, we presented a sample of PSQI (Supplement A).

Assessment of potential confounders

Age, sex, education, cognitive functioning, activities of daily living, body mass index (BMI), and depressive symptoms were considered as potential confounders. Education was assessed routinely in the home interview and subdivided into low, intermediate, and high education. Cognitive functioning was measured using the Mini-Mental State Examination (Folstein, Folstein, & McHugh, 1975) during one of the visits to our center. The ability to perform activities of daily living was measured with the Stanford Health Assessment Questionnaire (Bruce & Fries, 2003; Fries, Spitz, & Young, 1982). Height and weight were measured without shoes and heavy clothing to calculate the BMI (kg/m²). Depressive symptoms were measured with the Center for Epidemiological Studies Depression scale (CES-D). In our baseline table, we also showed the presence of depressive symptoms among participants who scored 16 or above, suggesting clinically relevant depressive symptoms on the Center for Epidemiological Studies Depression scale.

Statistical Analyses

To explore the association between grief and sleep parameters we used linear regression. Model 1 was adjusted for age and sex. Model 2 was additionally adjusted for education, cognitive functioning, activities of daily living, and BMI. Model 3 was further adjusted for depressive symptoms. In the longitudinal analyses, to examine whether grief status was prospectively associated with sleep duration and sleep quality, we used sleep duration and sleep quality assessed during the follow-up as outcomes. We selected the same covariates as in the cross-sectional analyses and adjusted for the respective baseline values of sleep duration or sleep quality.

We conducted a series of sensitivity analyses. First, we reran the analysis, not only for depressive symptoms, but to exclude all patients with major depression disorder at the baseline. We evaluated the presence of Major Depressive Disorder in those with a CES-D score, or above the established screening cutoff of 16, using the Schedules for Clinical Assessment in Neuropsychiatry (SCAN) interview (Wing et al., 1990). This semistructured clinical assessment was performed by trained clinicians to determine which participants fulfilled the DSM-IV major depressive disorder. We performed this sensitivity analysis in order to minimize the depressive disorders on sleep quality. Also, in our study, Major Depressive Disorder (MDD) was assessed at baseline only (prevalence of 2%). Second, to test the effect of the more recent sleep, we performed longitudinal analysis restricted to those who experienced the bereavement in the years prior to baseline assessment. Third, we repeated the cross-sectional linear regression only in those participants who had attended the follow-up assessment, to test whether any between–cross-sectional and longitudinal analysis reflected a selection effect. Adjustments for these analyses were conducted as in the main linear regressions. Fourth, since sleep duration component of PSQI, we performed a sensitivity analyses calculating the PSQI total without the component of sleep duration. Fifth, in order to explore the possibility of reverse causality, in the longitudinal analysis, we excluded people who had poor sleep at the baseline (defined as the total score of all components of PSQI greater than 5 points). Sixth, we explore whether, cross-sectionally and longitudinally, sleep duration and sleep quality differ between grievers (reference) and complicated grievers. We also explored the association between grief status and depressive symptoms (CES-D). In multivariable linear regression models, we examined whether baseline grief status was prospectively associated with CES-D at follow-up, further adjusted for the baseline value of depressive symptoms.

With respect to the remaining data, missing values were imputed using multiple imputations (Rubin, 2004). In the present study, for each missing value five draws were performed providing five substituted items of data, which in turn created five completed data sets. Analyses were performed separately on each completed data set and thereafter combined into one pooled estimate. The percentage of missing values within the population for the analyses was lower than 20% (ranging from 0 to 18%). Age and sex had no missing values, education had 18%, cognitive score had 8%, activities of daily living score had 1%, BMI had 14%, and CES-D had 0.2%. Analyses were performed using SPSS Statistics (version 20; SPSS, Chicago, IL, USA).

Results

Of the 5,421 eligible participants, 4,378 (80%) were classified as experiencing “no grief,” 795 (15%) as experiencing “normal grief,” and 248 (5%) as experiencing “complicated grief” at baseline. Table 1 presents the characteristics of the study population. Participants classified as experiencing complicated grief were older, were more likely to be female, had a lower level of education, and were more likely to have clinically relevant depressive symptoms. The main cause for grief was death of a partner (36% of those with normal grief and 56% of those with complicated grief), or a child (10% of those with normal grief and 22% of those with complicated grief).

TABLE 1 Baseline Characteristics of Study Participants 2002–2005 (N = 5,421)

	Nongrievers		Normal grievers		Complicated grievers
Characteristics	N = 4,378 (80%)		N = 795 (15%)		N = 248 (5%)
Age, years (SD)	72.4	(7.7)	73.4	(8.1)^a	74.7	(7.5)^b
Women (%)	55		73^a		72^b
Education^a.b
Primary (%)	19		22		30
Intermediate (%)	67		66		63
High (%)	14		12		7
Cognitive functioning, score (SD)	27.41	(2.58)	27.39	(2.52)^a	26.99	(3.04)^b,c
Depressive symptoms, score (SD)	5.42	(6.67)	8.62	(8.49)^a	14.12		(11.35)^b,c
Clinically relevant depressive symptoms (%)	4		8^a		25^b,c
Who died?
Partner (%)	-		36		56
Child (%)	-		10		22
Parent (%)	-		12		6
Brother/sister (%)	-		17		7
Others (%)	-		25		9
Activities of daily living, score (SD)	0.51	(0.57)	0.59	(0.57)^a	0.66	(0.60)^b
Body mass index (kg/m^2), (SD)	27.56	(4.03)	27.67	(4.34)	27.75	(4.38)
Sleep duration (hours), (SD)	6.91	(1.30)	6.69	(1.36)^a	6.44	(1.50)^b,c
Sleep quality, score (SD)	3.37	(2.91)	4.28	(3.20)^a	5.08	(3.67)^b,c

Group comparisons were performed with χ² (categorical variables) or t-test (continuous variables) for independent samples.

^aComparison of nongrieving participants with grieving participants (p < 0.05).

^bComparison of nongrieving participants with complicated-grief participants (p < 0.05).

^cComparison of grief participants with complicated-grief participants (p < 0.05).

Table 2 shows the cross-sectional associations of grief with sleep duration and sleep quality. In the age-and-gender adjusted analysis, we found a consistent association pattern of grief and complicated grief with sleep duration as well as with sleep quality. Further, adjustment for education level, activities of daily living, cognitive functioning, and body mass index did not change these associations. However, the association between grief and sleep indicators was explained by depressive symptoms (model 3).

TABLE 2 Cross-Sectional Associations of Grief and Complicated Grief With Sleep

		Model I				Model II				Model III
N		B	95% CI		p	B	95% CI		p	B	95% CI		p
Sleep quality (PSQI score)
Grieving status
Nongrievers (ref.)	4378	–	–	–	–	–	–	–	–	–	–	–	–
Grievers	795	–0.15	–0.25	0.05	0.003	–0.15	–0.25	–0.05	0.004	–0.05	–0.15	0.05	0.30
Complicated Grievers	248	–0.41	–0.57	–0.24	≤ 0.0001	–0.40	–0.57	–0.24	≤ 0.0001	–0.12	–0.30	0.05	0.16
Sleep quality (PSQI score)
Grieving status
Nongrievers (ref.)	4378	–	–	–	–	–	–	–	–	–	–	–	–
Grievers	795	0.64	0.39	0.90	≤ 0.0001	0.63	0.38	0.88	≤ 0.0001	0.15	–0.08	0.39	0.20
Complicated Grievers	248	1.52	1.09	1.95	≤ 0.0001	1.48	1.06	1.90	≤ 0.0001	0.08	–0.33	0.48	0.71

Abbreviations: PSQI (Pittsburgh Sleep Quality Index) Model I was adjusted for age and sex; Model II was adjusted for age, sex, education, activities of daily living, cognitive functioning, BMI, and the respective baseline values of sleep duration and Pittsburgh Sleep Quality Index to model change; Model III included factors in Model II and further adjustment for depressive symptoms.

In Table 3 we present the prospective association of grief with sleep duration and sleep quality (both assessed at follow up exam after 6.33 years on average (SD = 0.42)). We did not find an association of grief or complicated grief with changes in sleep duration or sleep quality, either in the age-and-gender adjusted or in the fully adjusted analyses. Next, we performed a series of sensitivity analyses. First, we excluded persons with major depression from our study population and reran the analysis; our result remained essentially unchanged. Then we limited the cases to those who experienced the bereavement leading to complicated grief in the last 2 years prior to baseline assessment (Supplement B). Our result showed no association between more recent lost event and sleep parameters in the longitudinal analysis. Also, to test whether the differences between cross-sectional and longitudinal analysis reflect a selection effect, we reran the cross-sectional analysis in participants who attended the follow-up assessment. Results remained essentially unchanged; the cross-sectional associations of complicated grief with sleep duration and sleep quality were similar to our original cross-sectional findings (data shown in Supplement C). Also, the results did not change when we reran the analysis calculating the PSQI score without including the sleep duration component (Supplement D). Next, in the longitudinal analysis, exclusion of subjects who had poor sleep quality at baseline did not change the results (Supplement E). Furthermore, we did not find any difference in sleep duration or sleep quality between grievers and complicated grievers in both cross-sectional and longitudinal analysis (data not shown). Last, we did not find an association between baseline grief status and depressive symptoms at follow-up (grievers: β = –0.22, 95% CI: –1.51—1.07, p = 0.74; complicated grievers: β = 1.00, 95% CI: –1.30—3.31, p = 0.39).

TABLE 3 The Longitudinal Associations of Grief and Complicated Grief With Sleep

		Model I				Model II				Model III
N		B	95% CI		p	B	95% CI		p	B	95% CI		p
Sleep duration (hours)
Grieving status
Nongrievers (ref.)	2876	–	–	–	–	–	–	–	–	–	–	–	–
Grievers	495	0.02	–0.10	0.14	p	0.02	–0.10	0.14	0.72	0.01	–0.11	0.13	0.91
Complicated Grievers	140	0.06	–0.15	0.27	0.57	0.06	–0.15	0.27	0.57	0.01	–0.20	0.23	0.91
Sleep quality (PSQI score)
Grieving status
Nongrievers (ref.)	2482	–	–	–	–	–	–	–	–	–	–	–	–
Grievers	413	–0.09	–0.39	0.21	0.55	–0.09	–0.39	0.20	0.53	–0.07	–0.37	0.23	0.66
Complicated Grievers	108	0.06	–0.49	0.61	0.83	0.07	–0.49	0.62	0.82	0.15	–0.42	0.71	0.61

Abbreviations: PSQI (Pittsburgh Sleep Quality Index). Model I was adjusted for age and sex; Model II was adjusted for age, sex, education, activities of daily living, cognitive functioning, BMI, and the respective baseline values of sleep duration and Pittsburgh Sleep Quality Index to model change; Model III included factors in Model II and further adjustment for depressive symptoms.

Discussion

In this large population-based study of middle-aged and elderly persons, we investigated whether persons with grief or complicated grief had a different sleep duration and sleep quality than participants without grief. Our cross-sectional findings showed that normal and complicated grief were associated with shorter sleep and lower sleep quality. These associations were mainly explained by the presence of depressive symptoms. No further changes in sleep duration and sleep quality between the groups were observed after an average follow-up of more than 6 years.

Complicated grief can be regarded as a bereavement situation for which sleep duration is likely to be affected (Monk, Germain, et al., 2008). A cross-sectional study of duration of sleep among unselected grievers, that is, most probably including persons with complicated grief, has been reported previously (Monk, Germain, et al., 2008). The authors conducted a laboratory study of sleep and circadian rhythm in 38 spousal bereaved seniors (≥ 60 years) observed 4 or more months after their loss event. On average, the bereaved seniors achieved only about 6 hr of sleep. In a large Japanese population-based prevalence study of 1,871 participants conducted by Doi, Minowa, Okawa, and Uchiyama (2000), the authors showed that being widowed or without a partner was associated with lower sleep quality.

There is evidence suggesting that behavioral changes associated with grief such as decreased activity levels or overall changes in social rhythm stability could lead bereavement to sleep disturbances. After the loss of a loved one, there are profound changes in lifestyle, often accompanied by reductions in financial security, perceived personal safety, and freedom of action (Monk, Germain, et al., 2008), all of which are likely to lead to sleep disruption. Also, the loss of a loved one is associated with psychological problems such as rumination or anxiety, which are shown to impair sleep (Carney, Edinger, Meyer, Lindman, & Istre, 2006; K. Shear et al., 2007). Sleep disturbances are particularly prevalent in depressed bereaved persons; even bereaved persons who fail to meet a formal diagnosis of depression have measurable sleep impairment (Reynolds et al., 1992). Indeed, our cross-sectional analysis showed that the association between grief and complicated grief with sleep indicators was largely explained by depressive symptoms. However, we ran multivariable linear regression models to see if baseline grief status was prospectively associated with CES-D at follow-up and found no association, providing support that depressive symptoms are not a mediator in the association between grief and sleep parameters. Further, reversed causality should be taken into account. Since relatively few studies have yet examined sleep difficulties as a risk factor for post-loss psychopathology, we cannot rule out that existing sleep problems make individuals vulnerable to more severe or prolonged grief or complicated grief. Indeed, as Boelen and Lancee (2013) pointed out, poor sleep quality is a known risk factor for many different forms of a psychopathology, including depression and PTSD.

We did not find a prospective association between grief and sleep parameters. Different explanations for these null findings are possible: First, the lack of findings can reflect the insufficient power to detect an association. Although fewer participants could be included in the longitudinal analyses, sufficient power to detect any effect similar to that observed in the cross-sectional analyses remained. Thus, these findings suggest that there was no further change in sleep duration and sleep quality once a person had reported bereavement at our baseline assessment. We carefully infer that the results could be explained with mechanisms of adaptive coping (S. S. Rubin, 1999; Stroebe & Schut, 1999) developed by the grieving participants during prolonged exposure to grief. Possibly, persons grieving reached a “stable state,” that is, with no further change of sleep quality, when participating in the follow-up assessment on average 6 years after the event. Sleep quality might have been affected before the occurrence of complicated grief. Due to the lack of prebereavement sleep assessment, it is not possible to evaluate the directional effects in the cross-sectional analysis, that is, whether bereavement triggered the decline of sleep duration and quality or whether sleep impairment preceded the grief reaction. However, our sensitivity analysis in which we excluded participants with poor sleep quality provides no evidence for reverse causality. Also, grief was assessed only at baseline. Consequently, we cannot account for the change in grief status, whether the feelings of grief remitted, persisted, or worsened.

However, the majority of clinical diseases and conditions are characterized by a progression of symptoms and their consequences; against this background, we had hypothesized a continuous decline of sleep problems, having in mind that sleep duration and quality among persons with complicated grief is of potential value for prognosis of grieving persons and potentially even of relevance for therapeutic interventions that rely on the cognitive behavioral interventions focused on sleep difficulties as discussed by Boelen and Lancee (2013). It suggests that the impact of grief on this important aspect of well-being is not accumulating over time and can potentially be overcome.

To the best of our knowledge, other longitudinal studies have not been performed previously in the general population. Also, our study is characterized by a long follow-up period and a large sample size. Furthermore, a middle-aged and elderly sample was used, which is the main vulnerable population for complicated grievers, as late-life loss of a loved one is among the most common life events. However, some limitations of the current study should be mentioned. First, in a population-based study, it is not feasible to ascertain grief and sleep directly after a loss event. Most important, complicated grief cannot be diagnosed if the event occurred less than 6 months before. Therefore, we performed sensitivity analyses restricting the study population to those who experienced the bereavement more recently (in the last 2 years before the baseline assessment). Second, we miss information on whether these persons are still suffering from complicated grief or MDD at follow-up. Further studies should be performed including this kind of prospective reassessment.

Conclusion

In community-dwelling middle-aged and older adults, persons with normal and complicated grief had both a shorter sleep duration and a lower sleep quality, mainly explained by depressive symptoms. However, prospective analyses showed that sleep quality and sleep duration do not decline further in persons with normal grief and complicated grief.

Acknowledgments

We want to thank all the persons who contributed to this article.

Supplemental Material

Supplemental data for this article can be accessed on the publisher’s website.

Additional information

Funding

The Rotterdam Study is supported by the Erasmus Medical Center and the Erasmus University Rotterdam; The Nederland’s Organization for Scientific Research (NWO); The Nederland’s Organization for Health Research and Development; the Ministry of Health, Welfare, and Sports; and the European Commission (DGXII). Jelena Milic was supported by an Erasmus Mundus (ERAWEB) grant. Henning Tiemeier was supported by the Vidi grant of the Netherlands Organization for Health Research and Development (ZonMw 2009-017.106.370). The funders had no role in the study design or data collection and analyses.