Distinguishing Bereavement from Depression in DSM-5: Evidence from Longitudinal Epidemiologic Surveys

Abstract

Following the earlier versions of the diagnostic manual, the DSM-IV set a higher symptom and duration threshold for the diagnosis of major depression in individuals who have experienced recent bereavement—the bereavement exclusion criterion. This criterion excludes a diagnosis of major depression among those whose symptoms persist for less than two months, as long as they do not have marked functional impairment, morbid preoccupation with worthlessness, suicidal ideation, psychotic symptoms, or psychomotor retardation. The DSM-5 committee, however, has proposed to remove this criterion from the upcoming DSM-5. The committee's decision was based on reviews of past literature. However, few past studies directly compared DSM-excluded bereavement-related depression to other major depressive disorder in representative population samples and had adequate power to detect differences. The results of these studies, therefore, did not provide strong evidence for the validity of bereavement exclusion. In this paper, we review three recently published analyses based on large epidemiologic samples that found significant differences between those with bereavement-excluded episodes and episodes meeting major depression criteria with regard to short-term risk of future depressive episodes, psychiatric comorbidity and other clinical and socio-demographic characteristics. In follow-ups ranging from 1 to 3 years, individuals with bereavement-excluded depressive episodes were significantly less likely to experience new episodes than those who met criteria for depression, and were not more likely to experience future episodes than those without any past history of depression. The findings from these new studies support the validity of the DSM-IV bereavement exclusion criterion and argue for preserving it in the new edition of the manual.

The DSM-IV definition of mental disorder states that a disorder is not “merely an expectable and culturally sanctioned response to a particular event” and that it represents a “dysfunction in the individual” (1). Because adverse experiences are a part of life and negative emotions arise as natural and adaptive responses to these experiences, much of the psychological distress that humans experience is considered expectable, culturally sanctioned, and functional. Thus, sadness is not a mental disorder in the DSM. However, adverse experiences can trigger negative emotional reactions that are so extreme that they are considered pathological. A large body of research has demonstrated the link between stressful life events and major depression (2-4), which is the main operationalization of pathological sadness in the DSM. If the symptoms that the individuals experience meet the manual's criteria for a major depressive episode (MDE), the context in which the symptoms present does not influence diagnosis, with one exception: bereavement.

DSM-IV acknowledges that individuals experiencing bereavement may present with symptoms resembling a MDE. For that reason, the threshold between normal distress and disorder is somewhat elevated for the bereaved. For the diagnosis of MDE to be applied in these individuals, symptoms need to persist longer than two months after the loss (as opposed to two weeks for the non-bereaved), or the individual has to experience marked functional impairment, psychotic symptoms, psychomotor retardation, suicidal ideations or morbid preoccupations with worthlessness. Bereavement-related depressive episodes lasting 2 months or longer or presenting any of these characteristics are sometimes referred to as cases of “complicated bereavement” (5). It should be noted that this term refers to depressive episodes that occur in the context of bereavement but are still considered MDEs. It should not be confused with “complicated grief”, which is a distinct syndrome characterized by prolonged, debilitating, painful emotions such as intense yearning, bitterness, shock, loneliness, and emptiness (6, 7). In recent years, the bereavement exclusion has come under criticism, and some experts have advocated for its removal (8-10). These experts question the validity of bereavement exclusion, maintaining that criteria for MDE should be applied without regard to context. In line with this view, the DSM-5 committee proposed removing the bereavement exclusion from the DSM-5—a proposal that has led to other criticism (11-14).

A number of articles published in recent years, including literature reviews and original analyses, have addressed the issue by seeking empirical evidence for the validity of the bereavement exclusion (6, 14-25). These studies pose the question whether individuals who would otherwise meet criteria for major depression, but for whom this diagnosis is precluded because of the bereavement exclusion, have the same disorder as those who are not excluded. The studies generally test the construct validity of the exclusion by comparing individuals with excluded bereavement (EB) to those with major depression on indicators such as risk factors, symptom profiles, and recurrence.

Some of the recent reviews and empirical studies have concluded that the available evidence supports removal of the bereavement exclusion from the DSM-5 (6, 15, 16, 20). However, there are problems with the data on which these conclusions are based. In their critical review of past studies of the validity of the bereavement exclusion, Wakefield and First (14) conclude that the past studies largely failed to provide a conclusive answer due to methodological limitations. They noted that in many studies purportedly supporting the removal of the bereavement exclusion, the bereaved group was comprised of both “complicated” and “uncomplicated” bereavement cases, greatly limiting their ability to test the validity of the exclusion as it exists in the DSM. Other studies have used clinical samples, lacked adequate power, inappropriately applied the exclusion criteria, and/or addressed the related but separate question of how bereavement compares to other losses (14). However, three recently published studies (two of which are discussed by Wakefield and First) are less affected by these limitations of past literature (18, 19, 23). These studies used data from large scale longitudinal epidemiologic surveys and directly compared EB cases with cases of major depression. The purpose of this paper is to provide a summary of the evidence for the validity of the bereavement exclusion based on these studies.

STUDY DESIGN AND METHODS

Data sources

All three studies described here presented analyses from longitudinal, population-based, epidemiologic surveys of mental disorders in the United States (See Table 1). Mojtabai (2011) and Gilman et al. (2012) analyzed waves 1 and 2 of the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) conducted in 2001-2002 and 2004-2005, respectively. Wakefield and Schmitz (2012) analyzed waves 1 and 2 of the Epidemiologic Catchment Area study (ECA, 1980-1985), conducted one year apart. Psychiatric disorders were measured in the NESARC using the Alcohol Use Disorder and Associated Disabilities Interview Schedule (AUDADIS), and in the ECA using the Diagnostic Interview Schedule (DIS). Both instruments are structured, lay-administered interviews that generate DSM diagnoses. Details of the NESARC and ECA can be found elsewhere (26-28).

Table 1.

Sample, methods and results of three recently published studies of the validity of bereavement exclusion based on longitudinal epidemiologic surveys.

	Mojtabai 2011	Gilman et al. 2012	Wakefield and Schmitz 2012
Data source	• NESARC (2 waves)	• NESARC (2 waves)	• ECA (2 waves)
Sample	• 43,093 wave 1 respondents	• 43,093 wave 1 respondents	• 19,291 with responses at both waves
Depression measure	• AUDADIS	• AUDADIS	• DIS
Comparison groups of interest (n)	• No history of depression (36,821) • Single, brief, bereavement-related episode excluding “complicated bereavement” (162) • Single, brief MDE (including bereavement-unrelated MDE and “complicated bereavement”) (1037)	• No history of depression (34,467) • Bereavement-excluded depression (196) • “Complicated bereavement” (566) • MDE (7,864)	• No history of MDEs (18,239) • DSM bereavement-excluded MDEs (25) • All other brief MDEs (446) (including bereavement-unrelated MDE and “complicated bereavement”) • Nonbrief MDEs (581)
Factors compared on	• Wave 1 characteristics: demographics, history of depression in first-degree relatives, comorbid mental disorders, functional impairment, treatment seeking, prescribed medication for depression • Depressive symptom profiles • New depressive episodes at wave 2 that were not bereavement-related	• Antecedent indicators: family history loading of depression and alcoholism, prior history of mental disorder (panic, GAD, social phobia, alcohol dependence), and personality disorders • Number of lifetime depressive episodes • Psychosocial impairment • Treatment seeking • Risk of mental disorders (MDE, panic, GAD, social phobia, and alcohol dependence) between waves 1 and 2	• Prevalence of 1-year MDEs at Wave 2
Criteria for excluded bereavement (EB) classification	• Met AUDADIS MDE criteria but did not have a history of manic, mixed, or hypomaic episodes • Single episode that began just after death of a loved one • Duration <2 months • No marked impairment • No psychomotor slowing • No feelings of worthlessness • No suicidal ideation	• Met AUDADIS MDE criteria • Symptoms began just after someone close to them died • Duration <2 months • No marked functional impairment • No suicidal ideation • No psychomotor retardation	• Met full DIS MDEs criteria • All episodes were bereavement-related • Duration of 2 months or less • No suicidal ideation • No psychomotor retardation • No sense of worthlessness
“Complicated bereavement”	• Not used as a comparison group • Included with single, brief MDE group in comparisons presented below • Psychotic symptoms not assessed	• Used as comparison group • Defined as MDE following the death of a loved one persisting for at least 2 months and accompanied by marked functional impairment, suicidal ideation, or psychomotor retardation • Psychotic symptoms and morbid preoccupation with worthlessness not assessed	• Not used as a comparison group • Included with other brief MDE group • Marked impairment and psychotic symptoms not assessed
Differences between EB and depression groups	• EB group less likely than single, brief MDE group to have age of onset in 20-29 age range (17.8% vs. 33%) and more likely to have onset in 50+ age range (24.7% vs. 13.9%) • EB group less likely to have received any treatment for depression (19.5% vs. 33.3%) than Single, brief MDE group • EB group less likely to have received prescription medications (9.1% vs. 23.9%) • EB group less likely to have symptoms of increased sleep (20.4% vs. 41.8%) and more likely to have decreased sleep (86.9% vs. 72.5%) • Recurrence of depression during follow up was not different between the EB group and the no history group [OR .48, 94% CI (.23, 1.03)] • Recurrence was higher for the single, brief MDE group than for the EB group [OR 3.03, 95% CI (1.36, 6.74)] and the no history group [OR 1.47, 95% CI (1.14, 1.89)]	• Social phobia [OR .44, 95% CI (.20, .97)] and avoidant [OR .24, 95% CI (.08, .70)], obsessive-compulsive [OR .54, 95% CI (.33, .88)], and schizoid [OR .20, 95% CI (.07, .54)] PDs less likely to precede EB than MDE • EB group had fewer lifetime depressive episodes [rate ratio .37, 95% CI (.30, .46)] and less psychosocial impairment [mean difference −.77, 95% CI (−.98, −.57)] than MDE group • EB group had lower odds of seeking any treatment [OR .18, 95% CI (.12, .30)] for depression than MDE group • EB group also had lower odds of treatment from a mental health professional [OR .15, 95% CI (.09, .24)], hospitalization for depression [OR .09, 95% CI (.02, .42)], ER visits for depression [OR .10, 95% CI (.02, .58)], and prescribed medication for depression [OR .23, 95% CI (.14, .39)] • EB group had significantly lower odds of depression [OR .33, 95% CI (.16, .65)] and social phobia [OR .15, 95% CI (.04, .68)] at wave 2 • No significant differences in ORs for wave 2 disorders between EB and no history of depression group	• Prevalence of MDEs at wave 2 was not different between those with EB at wave 1 and those with no MDE history at wave 1 [3.7% (−3.6%, 11.0%) vs. 1.7% (1.4°%, 1.9%)] • Prevalence of MDEs at wave 2 was lower among those with EB at wave 1 than among those with brief MDEs at wave 1 [3.7% (−3.6%, 11.0%) vs. 14.4% (11.0%, 17.9%), p<.05]and those with nonbrief MDEs at wave 1 [16.2% (12.7%, 19.7%), p<.05] • Similar results reported using original ECA syndrome severity criteria and after eliminating episodes that were ongoing during wave 1 • When using alternative 6-month threshold for uncomplicated bereavement-related depressive episodes, wave 2 recurrence rate for the bereavement-related group decreased to 2.4% (−2.4%, 7.2%)

Identification of bereavement-excluded cases

The three studies used largely similar criteria for identifying cases of EB, with slight variations. Cases were identified that met criteria for major depression, with symptoms beginning after the death of a loved one and lasting less than two months, and that lacked indicators of “complicated bereavement”. Wakefield and Schmitz used suicidal ideation, psychomotor retardation, and worthlessness feelings as such indicators, while Gilman et al. used suicidal ideation, psychomotor retardation, and marked functional impairment. Mojtabai presented two sets of analyses: one for episodes lasting less than two months regardless of other indicators, and a second removing those with suicidal ideation, psychomotor retardation, feelings of worthlessness, or marked functional impairment (indicators of “complicated bereavement”) from the EB group and placing them in the MDE group. To facilitate comparison with the other two studies, this second set of results of Mojtabai's study is discussed here. None of the studies was able to include psychotic symptoms as an indicator of disorder.

Mojtabai included only EB cases with a single lifetime episode at wave 1. Wakefield and Schmitz included only those who reported that all MDEs were bereavement-related. Gilman et al. included individuals whose worst MDE was bereavement-related, regardless of whether other lifetime bereavement-related or bereavement-unrelated MDEs were present.

Comparison groups

Mojtabai compared the EB to four other groups: those with single, brief (<2 months) MDEs that were bereavement-unrelated, those with single non-brief MDEs, those with two or more MDEs, and those with no lifetime history of depression at wave 1. The first of these comparison groups provided the main comparison of interest and also included those with “complicated bereavement”. We refer to this group as the “single, brief MDE” group. Similarly, Wakefield and Schmitz compared the EB group to those with other brief MDEs (either bereavement-unrelated or “complicated bereavement”), those with MDEs lasting more than two months, and those with no history of MDEs at wave 1. Gilman et al. compared the EB group to those with bereavement-unrelated MDEs, those with “complicated bereavement,” and participants with no lifetime history of MDEs at wave 1.

Comparisons

The three studies compared groups on a number of different characteristics that were selected as indicators of pathology or because of phenomenological coherence with major depression. These included family history of depression and personal history of selected mental disorders, symptom profiles, impairment, treatment, recurrence risk, and risk of other disorders at wave 2. Additionally, Mojtabai presented comparisons on age, sex, race/ethnicity, and age of onset.

RESULTS

The results of comparisons between EB and MDE groups are presented below. Differences between EB and MDE groups are summarized in Table 1.

Demographics

Mojtabai compared demographic characteristics across depression groups. Compared to those with single, brief MDE, those with EB were more likely to be in the oldest age group (50+) and less likely to be in the 20-29 years age group at the onset of their symptoms (13.9% vs. 24.6% and 33.0% vs. 17.8%, respectively).

Family history and comorbid disorders

Gilman et al. and Mojtabai investigated differences in family history and prior or concurrent mental disorders between depression groups. Gilman et al. reported that social phobia was less likely to precede EB cases than MDE [OR 0.44, 95% CI (0.20, 0.97)]. Similarly, lifetime avoidant [OR 0.24, 95% CI (0.08, 0.70)], obsessive-compulsive [OR 0.54, 95% CI (0.33, 0.88)], and schizoid [OR 0.20, 95% CI (0.07, 0.54)] personality disorders were less common among participants with EB compared to MDE. Gilman et al. did not find significant associations with prior panic disorder or generalized anxiety disorder, or paranoid, histrionic, or antisocial personality disorders. Mojtabai did not find differences in wave 1 rates of lifetime dysthymia or alcohol dependence. Neither Mojtabai nor Gilman et al. found differences by family history of depression and Gilman et al. did not find differences by family history of alcoholism.

Symptoms, impairment, and number of episodes

Mojtabai compared symptom profiles between participants with EB and single, brief MDE and found that those with EB were less likely to report increased sleep and more likely to report decreased sleep, than those with single, brief MDE (20.4% vs. 41.8% and 86.9% vs. 72.5%, respectively). Psychomotor slowing, morbid preoccupations with worthlessness, and suicidal ideation were not included in the comparison of symptom profiles because they were used to define EB. The two groups did not differ in prevalence of depressed mood, concentration difficulties, anhedonia, fatigue, appetite problems, excessive guilt, or agitation.

Gilman et al. reported that those with EB had less psychosocial impairment than those with MDEs [mean difference -0.77, 95% CI (-0.98, -0.57)]. Mojtabai did not assess differences in impairment because impairment was used in the definition of EB.

Gilman et al. assessed differences in number of lifetime depressive episodes and found that EB was associated with fewer episodes than MDE [rate ratio 0.37, 95% CI (0.30, 0.46)]. Mojtabai did not assess such differences because the EB cases and the main bereavement-unrelated comparison group were limited to cases with single episodes.

Treatment

Mojtabai and Gilman et al. investigated differences in treatment experience between groups. Mojtabai reported that compared to those with single, brief MDE, individuals with EB were less likely to have received any treatment for depression (19.5% vs. 33.3%) and were less likely to have been prescribed medication for depression (9.1% vs. 23.9%). Similarly, Gilman et al. reported that, compared to those with MDEs, EB cases were less likely to receive any treatment [OR 0.18, 95% CI (0.12, 0.30)], less likely to have sought help from a mental health professional [OR 0.15, 95% CI (0.09, 0.24)], less likely to have been hospitalized for depression [OR 0.09, 95% CI (0.02, 0.42)], less likely to have visited an emergency room for depression [OR 0.10, 95% CI (0.02, 0.58)], and less likely to have been prescribed medication for depression [OR 0.23, 95% CI (0.14, 0.39)].

Risk of depression at follow-up

All three studies compared the risk of future depressive episodes among the depression groups and also included a comparison with participants with no history of depression at wave 1. All three studies found that the risk of depression at wave 2 was not significantly different among participants with EB and those with no history of depression at wave 1. In contrast, risks in other depression groups were elevated compared to those with no history of depression. Wakefield and Schmitz compared the 1-year prevalence of MDE at wave 2 and found that the percent with these future episodes in the EB group [(3.7%) (0%, 11.0%)] was not different from that in participants without a depression history at wave 1 [1.7% (1.4%, 1.9%)], and was lower than among those with brief MDEs [14.4% (11.0%, 17.9%)] or non-brief MDEs [16.2% (12.7%, 19.7%)]. Similarly, Mojtabai reported that the odds of new depressive episodes during follow-up were not different between those with EB and those without a history of major depression [OR 0.48, 95% CI (0.23, 1.03)], but that those with single, brief MDE had a higher odds than the EB cases [OR 3.03, 95% CI (1.36, 6.74)] or than participants without a history of major depression [OR 1.47, 95% CI (1.14, 1.89)]. Finally, Gilman et al. reported that compared to those with MDEs at wave 1, EB cases were less likely to experience depression at wave 2 [OR 0.33, 95% CI (0.16, 0.65)], and that a test comparing the ORs for MDEs at wave 2 between EB cases and those with no depression at wave 1 was not statistically significant.

Risk of other disorders at follow-up

In addition to the risk of depression, Gilman et al. compared risks of panic disorder, social phobia, GAD, and alcohol dependence at wave 2 between the groups. They reported that like depression, social phobia at wave 2 was significantly less likely among those with EB compared to those with MDEs [OR 0.15, 95% CI (0.04, 0.68)] and that EB was not associated with increased risk of disorders at wave 2 when compared to those with no history of depression at wave 1. The comparisons of wave 1 EB and MDEs with regard to the risk of wave 2 panic disorder, GAD, and alcohol dependence were in the same direction as comparisons for risk of social phobia but not statistically significant.

DISCUSSION

We have reviewed the results of three recently published, longitudinal, large-scale epidemiologic studies investigating the validity of the DSM MDE bereavement exclusion. Taken together, the results indicate that compared to other types of depressive episodes, the EB episodes are associated with a lower prevalence of comorbid mental disorders, have different symptom profiles, lower likelihood of treatment in a variety of settings, fewer lifetime depressive episodes, and lower risk of future depressive episodes. The finding that is most consistent across these studies is that of a lower risk of future depressive episodes among the EB cases. All three studies found that the risk of future episodes among EB cases was not different than the risk of future depressive episodes among those with no prior depression history, and that it was significantly lower than the recurrence risk among people with MDEs.

The decision of whether some psychological or behavioral phenomenon should be considered a disorder has major consequences, and it can be made on a variety of potential bases. It has been the position of the American Psychiatric Association that the DSM should be founded upon the best available empirical evidence (1). We would agree with Maj that removing the bereavement exclusion “requires strong and unequivocal research evidence” (13). The studies described here constitute some of the strongest available data on the validity of the bereavement exclusion, and serve as an appropriate basis for evaluating its inclusion. These studies do not support removal of bereavement exclusion from the DSM; at the very least, they indicate that the data thus far are inconclusive.

By themselves, the studies discussed here may be limited in their ability to determine the validity of bereavement exclusion. These are only three studies drawn from two distinct data sources. The Mojtabai and Gilman et al. analyses were based on the same data, and should not be regarded as independent lines of evidence. Although these studies’ finding that future risk of depression is lower for the EB cases was replicated by Wakefield and Schmitz, the remaining results have yet to be replicated in independent samples. There is clearly a need for more high-quality studies to further investigate the validity of bereavement exclusion.

The ability of any study to test the validity of the bereavement exclusion depends on how well the DSM-IV construct is operationalized. Although the studies reviewed here provided improvements over previous studies, none were able to fully capture the construct because of data limitations. For example, none of the studies excluded cases from the EB group based on presence of psychotic symptoms. Furthermore, marked impairment in role functioning is not clearly defined in the DSM-IV and, therefore, cannot be clearly operationalized for research. It is possible that a more precise operationalization of the bereavement exclusion would produce different results. However, the effect of any misclassification of cases of genuine depression as EB cases would be to attenuate differences between these groups. Thus, improved classification of cases would likely lead to greater differences between EB and MDE cases. In addition, studies are limited by the validity indicators that are used. For example, some studies have indicated that about 50% of people with a first episode of major depression will recover without experiencing future episodes (29). If major depression is not necessarily a recurrent disorder, then presence of future episodes may not be the most useful disorder indicator. Despite these limitations, the studies discussed here address some of the major limitations of prior studies and provide support for validity of the DSM-IV bereavement exclusion criterion.

In this paper we focused on studies that tested the bereavement exclusion as it exists in the DSM-IV, using the most complete operationalization of the exclusion possible. One potential drawback of this approach is that it does not address the question of whether depressive episodes that come after the death of a loved one are different from those that do not. It is possible that the differences in recurrence rates observed between those with EB and those with MDD are merely due to differences in duration, impairment, and other EB defining symptoms, and do not reflect a difference between bereavement-related and bereavement-unrelated depressive episodes. The field may benefit from more epidemiologic studies of the natural history of bereavement and depression, which could provide insight into distinction between pathological and non-pathological responses to the death of a loved one. In addition, a separate, but related, question is whether the bereavement exclusion should be extended to other types of loss and stressful events. This question has generated some interesting results thus far (5, 21) and is worthy of future study.

Conclusion

There is a scarcity of research on the validity of the bereavement exclusion that is epidemiologically-based, longitudinal, adequately powered, and that appropriately separates EB other bereavement-unrelated depressive episodes (14). Here we have summarized the results of three recently published studies that satisfy these conditions. Overall, the results do not support the proposal to remove the bereavement exclusion from the DSM. There is evidence that those with EB have a risk of future depressive episodes that is lower than that of individuals with major depression, and similar to those with no baseline history of depression. There is also preliminary evidence that individuals with EB may differ from those with major depression in symptomatology, number of lifetime episodes, and treatment engagement. However, these findings need to be corroborated in future research. We would argue for maintaining the bereavement exclusion in the DSM until a substantial body of high-quality research contradicts the available evidence.