Complicated grief after death of a relative in the intensive care unit

Introduction

Over the last two decades, the use of intensive care at the end of life has increased significantly [1–3]. Studies have described intensive care unit (ICU) practices at the end of life [3–7], shed light on treatment-limitation decisions [8], identified specific needs of bereaved relatives, and determined what affects the quality of the dying experience for relatives, physicians and nurses [9–17].

Relatives of patients who die in the ICU experience a considerable burden of harm [18]. Immediately after the loss, they frequently exhibit symptoms of anxiety and depression [19]. 3 months later, half of them have post-traumatic stress disorder (PTSD) symptoms [20] and quality-of-life alterations [21]. Ill effects identified within the year following the death include financial difficulties [11–13] and psychiatric illness [22]. Studies have identified targets for improving the quality of dying and death, as well as the relatives' experience of bereavement [14, 15, 23]. Nevertheless, specific data are needed on the grieving process after the death of a relative in the ICU.

Sadness, anger, and symptoms of depression are normal manifestations of grief that usually resolve within a few months [24, 25]. Complicated grief, a well-defined syndrome of grief manifestations persisting >6 months [24, 25], occurs in ∼10% of bereaved relatives [26, 27] and adversely affects quality of life [25, 28–31]. The persistent manifestations may include intense yearning for the deceased, a sense of disbelief regarding the death, anger, bitterness, intrusive preoccupying thoughts of the deceased, avoidance of reminders of the loss and difficulty in moving on with life [24, 25]. Features of depression and PTSD may exist in relatives with or without complicated grief, raising diagnostic challenges [32], although some distinctions have been established. For example, one of the key distinguishing features of complicated grief is that it tends to come in waves, unlike depression, which tends to be unremitting and persistent. Complicated grief can result in physical and mental health problems or in the worsening of these problems. In a study by Prigerson et al. [25] outside the intensive care setting, complicated grief after the death of the spouse predicted negative health outcomes such as cancer, heart trouble, high blood pressure, suicidal ideation and changes in eating habits at 13- or 25-month follow-up. Complicated grief is also associated with substantial impairment of work and social functioning, sleep disturbance, suicidal ideations and behaviour, increase use of tobacco and alcohol, and impairment in relationship functioning [33]. Many studies have focused on complicated grief in elderly patients [34, 35]. Complicated grief is considered more common in those experiencing disasters [33]. It has also been found to be overrepresented in family members of patients with life-threatening illnesses and suicides or homicides [36]. Few studies have assessed complicated grief during the year following the death of an adult relative in the ICU. In two small, single-centre studies based on the Inventory of Complicated Grief (ICG), the prevalence of complicated grief was 5% (two out of 41) after 3–12 months [22] and 46% (six out of 13) after 6 months [37]. However, these studies provide no information on potential links between complicated grief and other components of the post-ICU burden in bereaved relatives, nor do they indicate how specific ICU practices may affect the risk of complicated grief.

Here, we report a prospective observational study in 41 ICUs that was designed to determine the prevalence and risk factors of complicated grief during the first year following the death of an adult relative in the ICU. We also evaluated links between complicated grief and symptoms of PTSD and depression.

Results

Figure 2 is a flow chart of the study design. Among the 4607 patients admitted to the 41 participating ICUs during the study period, 875 (19%) died, including 228 who met exclusion criteria, 104 for whom the opportunity for inclusion was missed and 68 for whom the relative refused participation. The remaining 475 (54%) were included. Table 1 presents the circumstances of the deaths. The characteristics of the ICUs (table S1), patients and relatives (table S2) are presented in the online supplementary material.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

Flow chart of patients and relatives. ICU: intensive care unit. ^#: patients either died before ICU day 3 or did not have a relative visiting them before the time of death.

Response rates were 90.5%, 81.3%, 59.4% and 45.2% on days 21, 90 and 180, and at 1 year, respectively. Among all the collected variables (related to ICUs, centres and patients), none was significantly different between relatives who did not respond and those who actually participated.

Determinants of complicated grief and PTSD-related symptoms at 6 months

Table 3, table S3 and figure 3 present factors associated with ICG >25 and/or IES-R >32 after 6 months. 6 months after the loss, factors independently associated with complicated grief only were patient-related (i.e. patient's refusal of treatment was associated with a decreased risk of presenting complicated grief symptoms) and physicians-related (i.e. intensivist board certification before 2009 and relatives reporting poor quality communication with physicians). Interestingly, none of the ICU characteristics predicted complicated grief. Perception by the relative that the patient could not breathe peacefully was only associated with a higher risk of PTSD symptoms. Variables associated with both complicated grief and PTSD related symptoms were family-related (relative of female sex and relative living alone), related to end-of-life management (patient died while intubated, the relative being present at time of death and the relative not having said goodbye to the patient). In the bivariate analyses, involvement of the relative in treatment-limitation decisions was associated with decreased complicated grief whereas disagreement between the relative and intensivist about goals of care was associated with increased PTSD symptoms. However, neither covariate had a significant effect in the multivariate analysis.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3

Plot distribution of independent determinants of complicated grief and post-traumatic stress disorder (PTSD) symptoms identified by multivariate analysis. Data are presented as odds ratios. Whiskers represent 95% confidence intervals.

In order to confirm these results, and to analyse the relationship between complicated grief and the explanatory variables, a focused principal-component analysis was performed (fig. 4). Three clusters of explanatory variables are depicted in this figure. First, the year of intensivist's board certification and the fact that patients died intubated are positively correlated. A second cluster (patient's refusal of treatment, female relative, relative living alone following the loss and patient could not breathe peacefully) is diametrically opposite and therefore negatively correlated to the first one. The third cluster includes the fact that relatives witnessed the death and did not say goodbye to the dying patient.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4

Focused principal-component analysis of determinants of post-traumatic stress disorder (PTSD) symptoms and symptoms of complicated grief (CG) at 6 months. This graphical display is focused on CG and its correlation with other variables. This representation allows both the pattern of relationships between CG and the dependencies between the responses to be seen. The relative positions of the predictors may distinguish some clusters of explanatory variables. CG is at the center of the diagram and the distance of this point to another variable may be translated into correlation. When two points are close to one another, then the two underlying variables are positively correlated. When two points form a right angle with the origin, the two underlying variables are uncorrelated. When two points are diametrically opposed, then the two underlying variables are negatively correlated. The variables inside the red circle are significantly correlated with the dependent variable at the 5% level. Three clusters of explanatory variables are depicted in this figure: first, the year of intensivist's board certification and the fact that patients died while intubated are positively correlated; a second cluster (patient's refusal of treatment, female relative, relative living alone following the loss and patient who could not breathe peacefully) is diametrically opposite and therefore negatively correlated to the first one; the third cluster includes the fact that relatives witnessed the death and did not say goodbye to the dying patient.

Figure S1 shows the end-of-life process and its relationships with complicated grief and PTSD symptoms.

Discussion

This multicentre study is the first to report the incidence of complicated grief in a large number of relatives of patients who died in the ICU. Sadly, over half of these bereaved family members presented with complicated grief symptoms at 6 months, a proportion that persisted essentially unchanged for a total of 12 months. This result is alarming considering that complicated grief is a condition that exposes bereaved relatives to additional burden, such as symptoms of yearning and failure to adapt. PTSD was also high in this study, with 43.6% of relatives presenting symptoms after 6 months, but, unlike complicated grief ,this proportion diminished significantly in the following 6 months. Among study variables, some were associated with complicated grief only and some with both complicated grief and PTSD. This study is important because it provides new targets for improving end-of-life practices with the ultimate goal of decreasing the burden on bereaved relatives: taking into account patients' wishes, timely extubation, closure at the end of life and high-quality communication.

Complicated grief is distinct from depression and PTSD, yet may produce symptoms of both. In this study, presence of symptoms of depression at 3 months was associated with complicated grief at 6 months. In addition, up to 36.9% of the patients presented with both complicated grief and PTSD, suggesting that detecting complicated grief does not add to PTSD. However, some clinical symptoms are specific to complicated grief (such as symptoms of yearning and failure to adapt) and may require different coping strategies. Also, high levels of PTSD symptoms have been identified for over 10 years and still remain high, suggesting that current ways of preventing PTSD may not be effective. This study clearly shows that if up to one-third of relatives present with both complicated grief and PTSD, ∼15% present complicated grief only and 6% PTSD only. Interestingly, some of the patients who had both complicated grief and PTSD after 6 months showed an improvement in one of these two scores and not in the other. In keeping with the relative's follow-up after an ICU stay, our data report some differences in PTSD prevalence based on the time of assessment (3 months versus ≥6 months), the difference between bereaved and nonbereaved relatives, and study designs (control groups from interventional studies) [20, 22, 23, 37].

Complicated grief is the persistence of specific symptoms >6 months after the death [24–25] and severely affects quality of life [24–27, 49]. Some of the factors significantly associated with complicated grief suggest specific explanations. For instance, relatives of patients who died while intubated may have been unaware of the exact time of death, which may have led to a feeling of disbelief about the death. Relatives of patients who died while not intubated may have perceived the death as more natural and may have received greater support from the ICU staff, in particular during obvious manifestations of respiratory distress, gasps or a need for titrating sedation. Not saying goodbye to the patient may contribute to feelings of yearning for the deceased, as well as to anger and bitterness. Actively helping the relatives understand that the patient is dying and helping them find closure by saying goodbye to their loved one is crucial. Therefore, the ICU staff must regularly assess the relatives' feelings and experience, and offer to remain with the relatives at the patient's bedside. Being present at time of death increased the risk of presenting complicated grief symptoms in our study. This finding may seem surprising, as relatives often express the wish to be with their loved one at the time of death [12] and witnessing the death has been identified as a marker for good-quality end-of-life care [47]. However, relatives may wish to witness the death then find the experience difficult to handle, at least in the ICU setting. Inadequate management of patient comfort and dignity can impact on relatives' experience of the dying process. This puts forward the necessity for improving perceived suffering by patients during the dying process. Furthermore, inadequate preparation related to poor communication and lack of support can adversely affect the relatives 6 months after the death. Relatives who wish to be present at the time of death need compassionate support before, during and after this experience.

We identified three areas for improvement. The first one is relevant to end-of-life practices and affects three domains: taking into account patients' wishes (treatment refusal), timely extubation and closure at the end of life, all three of which are important components of high-quality end-of-life care in the ICU, in keeping with earlier data [15, 16, 50–53]. However, our study is the first to show a significant association between end-of-life practices and complicated grief symptoms up to 1 year after the death. The second area of improvement is communication: interestingly, suboptimal communication with the intensivist was associated with complicated grief symptoms but not with PTSD-related symptoms. High-quality communication involves adequate information that death is imminent, to ensure that the family is aware of the reality of death [23] and can say goodbye to the patient; another component of good communication is adequate support for relatives who wish to witness the death. That intensivist board certification before 2009 predicted complicated grief may reflect recent changes in the critical care curriculum in France, which now includes specific training on care, ethical issues and communication at the end of life. An alternative explanation is that communication with young physicians may be less stressful for the relatives, as suggested by a previous study [54]. The third area of improvement is post-ICU care: living alone after the death was associated with the presence of complicated grief symptoms, underlining the importance of providing bereaved relatives with effective social support. Previous studies reported that complicated grief was increased among subjects with an absent or unhelpful family [55]. Interventions to evaluate the impact of targeting these three domains with the goal of diminishing the frequency of complicated grief are warranted.

This study has several limitations. First, all participating ICUs were in France, and whether our findings apply to other countries is unclear. However, the proportions of relatives with complicated grief and/or PTSD-related symptoms were consistent with earlier data [37]. In addition, previous studies in France, Canada and the USA reported that the prominent model for decision making remains the shared decision-making model [56], with surrogate decision makers empowered to be able to actively be part of the decisions. In addition, the large sample size, varied case-mix and ethnic diversity of our patient population suggest applicability of our findings to other settings. Second, in this study, the lack of a control group of bereaved relatives after the death of a patient outside the ICU meant that we were not able to state that complicated grief occurs more frequently in this setting. However, the present study focused on identifying targets for improvement of end-of-life care in the ICU and to alert hospital clinicians, as well as general practitioners, that these bereaved relatives are at high risk of complicated grief. Third, nearly half the relatives were lost to follow-up before the end of the year, suggesting a potential for estimation bias. However, the 60% response rate at 6 months is among the highest in similar studies [14, 23, 57]. In addition, the reasons for loss to follow-up indicate clearly that a 100% response rate after 6 or 12 months is not realistic in studies of bereavement. ICG and IES-R scores at 6 months showed no significant differences between the groups with and without assessments at 12 months. Fourth, complicated grief and PTSD were assessed using questionnaires, as opposed to semi-structured interviews with psychologists or psychiatrists. However, these questionnaires have been validated in the ICU setting and proven reproducible across studies [20, 23, 58]. The ICG is a good screening instrument for complicated grief and scores >25 predict a range of negative health outcomes in many studies [26, 27]. We believe qualitative data would help to give meaning to the findings from the present study. Finally, we did not use the Revised ICG as no French translation of the scale was valid at the time of the study.

In practice, this study depicts the grieving process in individuals having lost an adult family member in the ICU. The five-fold higher frequency of complicated grief compared with studies in the general population [24, 25] indicates that community healthcare professionals must maintain a high index of suspicion for complicated grief, and discuss possible treatment and support with psychiatrist and/or psychologist. This conclusion applies also to depression and PTSD symptoms. The manifestations of normal grief and complicated grief should be taught in medical and nursing schools. Identification of risk factors associated with management of end-of-life care in the ICU helps target interventions that may help decrease bereaved families' burdens: encouraging, when possible, extubation; encouraging families to say goodbye; and adequate preparation and support during end-of-life care for families who wish to be present. The end of life in the ICU can still be improved both by enhancing communication strategies and by introducing palliative care at an early stage to help relatives make meaning of the patient's death.