Psychosocial Predictors of Outcome in Depression

Robert M. A. Hirschfeld

Identification of predictors of outcome is useful because they provide clues to the etiology of depression and its pathogenesis. They are also useful clinically because they enable the physician to formulate a more accurate prognosis. Predictors of response to specific medications are especially helpful in selecting treatments.

The focus of this chapter is on psychosocial predictors of outcome in depression. Also included are general demographic predictors and a few prevalent clinical predictors, such as comorbid anxiety symptoms. The chapter is organized by type of predictor. For each predictor, the general relationship between the particular variables and depression is briefly presented, followed by research findings relevant to the variable's general utility as a predictor of general outcome and data available on the response to treatment, specifically pharmacotherapy.

After these presentations on predictors of outcome is a discussion of psychosocial outcomes. The issue of psychosocial outcomes, as contrasted with symptomatic outcomes, is of growing importance in medicine, as quality-of-life concerns are becoming part of investigations of all treatments. Fortunately, quality of life and other psychosocial measures have been of interest in psychiatry for a long time.

To be included in this review, the study had to use an objective measure of the particular variable, and the clinical sample had to be systematically ascertained and carefully evaluated, usually with a semistructured diagnostic interview (such as the SCID.)

Major depression and dysthymia are distributed across adulthood with a modest peak in early adulthood. However, there has been a trend over the last 50 years for the age of onset to decrease and the prevalence of depression to increase among young adults (39). Twice over a 6-year period, Coryell and his associates looked at major depression in a nonclinical sample of 965 relatives, controls, and spouses of affectively ill probands who had never been mentally ill when first examined. They found that subjects younger than 40 years were three times more likely to develop depression than were older subjects (17).

Robinson and Starkstein found that the older age of onset for a first major depressive episode was associated with greater likelihood of an undiagnosed neurological (or other general medical) disorder (64). Also, an older age of onset lessened the probability of atypical personality or a family history of depression (8).

With regard to recovery, the findings are mixed. In several studies, age was found to be positively associated with speed of recovery (36). However, Akiskal (3), Keller et al. (35, 37), and Lewinsohn et al. (41), along with several others, did not find a relationship between age and recovery. Also, in a naturalistic study of 248 outpatients with bipolar disorder, O'Connell et al. found that neither current age nor age of onset was predictive of outcome (53).

In a community study of 119 married respondents with depression, McLeod and her colleagues found that age at episode onset was the only significant predictor of recovery in the total sample among other demographic variables, such as education, sex, and income. Specifically, persons who were older than 30 years at onset experienced longer episodes than did persons who were younger. Additionally, age was found to be more strongly predictive of recovery during early stages of an episode of depression, but not so much from later stages. Once an episode had lasted six or more weeks, age was no longer a predictive factor for the probability of recovery (47). With regard to medication, Brown and his coworkers found that older age predicted poorer response to tricyclic antidepressants in nonpsychotic, endogenously depressed patients (9).

In summary, age is clearly related to the prevalence of depression. With regard to recovery and other outcome variables, though, the relationship between age and depression is not clear.

With the exception of bipolar disorders, mood disorders occur twice as frequently in women as in men (12, 76). This well-established finding suggests that gender plays an important role in the etiology and/or pathogenesis of depression.

Gender does not appear to be a predictor of recovery or relapse when treatment is uncontrolled, as in longitudinal studies (37, 38). In the Zurich study of a community sample of young adults first interviewed at age 21, Vollrath and Angst found that gender did not predict recovery rates or response to medication (70).

The literature on differential sex response to antidepressant medications is rather limited. Hamilton and her colleagues reported a modest differential response to clorgyline by sex (24). In a study of 82 depressed subjects, Croughan et al. found that the clinical differences between male and female patients were minimal with regard to response to imipramine or amitriptyline. The women had equally good response to treatment, but that response tended not to be as strongly correlated with the clinical predictors as it was for men (19). In a study of atypical depression with 151 inpatients and outpatients, Davidson and Pelton found a gender difference when comparing responses to TCAs and monoamine oxidase inhibitors. Atypically depressed women responded better to MAOIs than to TCAs; and atypically depressed men showed a greater response to TCAs than to MAOIs (21).

In an examination of patients used for two multihospital collaborative studies of drug treatment in depression, Raskin found that younger women (i.e., those under the age of 40 years) responded less well to imipramine than did older women and men. Men under 40 years of age were seen to have responded better to chlorpromazine, but to have responded than younger women to phenelzine (60). However, O'Connell and his coworkers did not find any significant differences in outcome between the sexes in response to long-term treatment with lithium (53).

The findings with regard to the sex differences in pharmacokinetics of antidepressants, particularly amitriptyline and nortriptyline, have been inconsistent. Zeigler and Biggs (78) found no significant gender differences in plasma levels, but Preskorn and Mac (73) found higher plasma levels of either imipramine or amitriptyline in women. These differences may have occurred because of the coadministration of medications, including oral contraceptives, which could affect tricyclic antidepressant plasma levels (1). In their review of gender differences in pharmacokinetics and pharmacodynamics of psychotropic medication, Yonkers and colleagues concluded that there is little evidence to support claims of sex differential metabolism or actions of tricyclic antidepressants.

However, some differences have been reported for lithium. There appears to be a relationship between menstrual physiology and lithium pharmacology (77). Some studies have shown that adverse reactions to lithium may occur more frequently in women (7). This may be because lithium-induced hypothyroidism is more common in women (54). In addition, rapid-cycling bipolar illness is more prevalent in women (72); and this has been associated with thyroid abnormalities (54).

In summary, although there are a few scattered reports of modest or minimal gender effects predicting response to medication, the great majority of clinical studies did not find any evidence of gender relating to either outcome or medication response.

Depression has been found to be much higher among the nonmarried than among the married population. In the Epidemiology Catchment Area Study (ECA) studies, rates of major depression were lowest among the married, particularly those who were never divorced; whereas those who were divorced or cohabiting had the highest rates (67). This was also true of bipolar disorder in which married people had lower rates than did the divorced and never married.

Onset of depression is also more likely among the unmarried. In the Psychobiology of Depression Collaborative Study, Coryell and colleagues reported that onset of depression was most likely among single women. Divorce or separation increased the likelihood of a first-depressive episode (17). Coryell and associates (18), also found that both unipolar and bipolar patients were half as likely to marry as the comparison group. Bipolar patients with a predominance of manic episodes were also more likely not to be married, as reported by Romans and McPherson (65) in their study of 58 bipolar patients.

In their study of 101 patients with major depressive disorder in the Psychobiology Collaborative Study, Keller et al. (36) found a nonsignificant trend toward lower recovery rate in divorced or separated patients compared to married and single patients (45%, 63%, and 61%, respectively).

In the overwhelming majority of pharmacotherapy clinical trials in depression, marital status did not predict response to treatment (37, 38, 52, 68). However, in one study, O'Connell and colleagues found that unmarried status was associated with poor outcome in 248 bipolar patients enrolled in an outpatient lithium program (53).

In summary, marital status is greatly associated with onset and prevalence of depression. As with gender, the majority of studies found that marital status is not a significant predictor of outcome. However, one study did find that an unmarried status was associated with poor outcome to medication response.

The quality of marital and family relationships and their effect on outcome in depressed patients has long been of interest. However, methodological problems have limited the utility and generalizability of these reports. For example, many studies have confounded individual characteristics that influence outcome (e.g., personality features, severity, and duration of symptoms) with familial factors.

An emotional atmosphere of calm and control at the time of discharge from the hospital was predictive of steady improvement in patients (48). If the environment lacked these positive features, it would predispose the patient to dangerous or undesirable activities, such as alcohol or drug abuse (4).

Expressed emotion is a key measure of intimate relationships within the family. The presence of expressed emotion has been noted to result in closeness and better functioning within the family setting (56). O'Connell and colleagues found that patients whose families showed high expressed emotion were overrepresented in the poorer outcome groups (53). Hooley and Teasdale reported that along with high expressed emotion, marital distress and the depressed person's perception of the amount of criticism of his or her spouse was predictive of relapse (30).

Similarly, Hooley and her associates (29) reported that depressed patients with relatives with high levels of expressed emotion had a poor course of illness. Those depressed patients from dysfunctional families whose relatives had high expressed emotion were three times more likely to relapse within the first 9 months following depression than were those patients whose relatives had low levels.

Miklowitz and his partners (48) found that the emotional atmosphere of the family during the postdischarge period was an important predictor of the clinical course of bipolar disorder. In their sample of 23 manic patients recruited from two inpatient wards, the risk of relapse for patients from high-expressed-emotion homes was 5.5 times that of patients from low-expressed emotion homes. The risk of relapse for patients from negative affective style homes was 5.9 times that of similar patients from benign affective-style homes. Thus, if either high rates of expressed emotion or negative affective style were present in a patient's home, relapse was highly likely (94%); whereas patients whose families demonstrated low stress had low relapse rates (17%).

In a study of 68 inpatients with major depression who had been categorized as belonging to functional or dysfunctional families at index, Miller et al. (49) found that impaired family functioning was an important predictor of the course of depressive illness. At the 12-month follow-up, depressed patients with dysfunctional families had a significantly poorer course of illness, as was apparent by higher levels of depression, lower levels of overall adjustment, and a lower proportion of recovered patients (34.9% with dysfunctional families and 69.6% rate with functional families). Keitner and his colleagues found that better family functioning was one of the five most important factors related to recovery in major depression (34).

In general, supportive and positive response to a patient's depression predicts a more rapid recovery than nonsupport. In a community study of 119 married men and women in which one partner had experienced a depressive episode, McLeod and her colleagues (47) found that recovery was more rapid in those relationships where the nondepressed spouse was perceived as being compassionate and warm toward the depressed individual. Positive feedback seemed to be especially important for predicting recovery, especially during the early stages of an episode.

In the same study, when McLeod and her affiliates (47) examined the impact of six indicators of social support on recovery, the only ones that reached significance in the total sample were conflict with friends and spouse warmth. Even though conflict with a spouse did not consistently predict speed of recovery as did conflictual relations with friends (i.e., less rapid recovery), spouses' reports of their reactions to the respondents' depression did. Patients whose spouse reported feeling positively toward them (i.e., feeling warmth and compassion) were much more likely to recover in any given time period than were patients whose spouse did not report those feelings.

Impaired family functioning (i.e., negative affective style) predicted poor outcome from treatment with lithium (53). For patients to be able to have a good outcome, their family relationships and home environment needed to be benign as opposed to critical and overinvolved (53).

In summary, in spite of many methodological concerns, three studies reported that an atmosphere of calm and control (expressed emotion) and spousal acceptance was predictive of steady improvement, closeness, and better family functioning within a depressed patient's family. This led to a greater possibility that the patient would recover and with a more rapid recovery. Six other studies consistently reported that high levels of expressed emotion and impaired family functioning was predictive of poorer outcome (i.e., more likely to relapse and lower proportion of recovered patients). Within those six studies, one study showed that impaired family functioning was predictive of poorer response to medication treatment with lithium.

Similar to marital and familial support, social support has been of interest as a predictor of outcome in depressed patients. Several studies have examined the link of social support to depression. Clayton and her colleagues (15) have reported that a close, confiding relationship and physical proximity (i.e., social support) offers protection against the development of depression in persons in stressful situations. Warheit (71) provided evidence that individuals with low social support are at much greater risk of developing depressive symptoms. In their study of 44 outpatients with unipolar depression, Flaherty and colleagues found that patients with high social support had significantly better depression rating scores than did patients with low social support (22).

In general, the presence of social support has predicted successful outcome. In their study of 60 bipolar patients treated with lithium for one year, O'Connell and associates found positive social support, among other demographic measures such as sex, marital status, race, age, or education, to be most strongly correlated with a good outcome (52). Brugha and colleagues found that social support predicted clinical improvement in depressed patients in psychiatric hospitals, even when other potential risk factors such as age, sex, diagnosis, and severity of depression were controlled (11). In their more recent study of 248 bipolar outpatients, O'Connell et al. found that persons who had poor outcome to lithium treatment had less adequate social support than did those with better response to lithium (53).

Level of function in depressed subjects influences outcome (16, 68). In depressed individuals, low social dysfunction was predictive of good response to interpersonal psychotherapy; high work dysfunction was predictive of good response to imipramine; and high impairment of function was predictive of good response to combined treatment (68).

In summary, social support has a strong connection to depression. Quality social support reduces the likelihood of developing depression. Furthermore, the presence of positive social support predicts successful outcome, and its absence has been associated in one study to poor outcome to lithium treatment.

Whether socioeconomic status is a predictor of outcome from pharmacotherapy is unresolved. With regard to prevalence, persons meeting poverty level criteria had higher rates for major depression than those persons who were not in poverty (10). People of low socioeconomic status were more likely to have chronic or recurrent depression than others in higher socioeconomic strata (50).

In a study of bipolar disorder, O'Connell found that lower socioeconomic status was associated with poor outcome for treatment with lithium (53). On the other hand, in the Biological Psychobiology Collaborative Study, those with lower annual income were more likely to recover (19). This latter study was of very severely ill inpatients.

In summary, depression is slightly more prevalent in individuals of lower socioeconomic status. The findings with regard to socioeconomic status as a predictor of outcome are mixed.

In general, racial and cultural variables have not been reported as predictors of outcome in clinical trials and other studies, although one study did address this issue.

In a study of 159 black and 555 white depressed patients looking at the differential effects of chlorpromazine, imipramine, and a placebo, Raskin and Crook reported some differences in response between racial groups. They found that black patients evidenced a higher improvement rate at one week, irrespective of treatment, than did the white patients (61).

The relationship between personality and depression has received much attention in the literature (26). Four models have been described regarding the relationship between personality and depression: (a) predisposition, (b) complication, (c) spectrum, and (d) pathoplasty. The predisposition model proposes that certain personality characteristics render an individual vulnerable to depression and, in specific situations, can lead to depression. The complication model proposes that the experience of depression itself can cause personality change in an individual following recovery. The spectrum model proposes that certain personality characteristics may be genetically related to forms of depression. The pathoplasty model suggests that certain personality types may be associated with specific symptoms in a depressive episode (e.g., a histrionic personality may be associated with hostility, anger, and complaining during a depressive episode) (28).

To assess the connection between premorbid personality characteristics and onset of depression, Hirschfeld and his associates (27) looked at a sample of first-degree relatives, spouses, and controls of patients with mood disorders. Overall, differences were not found on measures of interpersonal dependency or extraversion. However, younger age (17 to 30 years old) predicted first onsets, both alone and in interaction with personality measures. With the younger subjects, personality variables did not significantly discriminate between the groups. Older subjects (31 to 41 years old), however, had decreased emotional strength, increased interpersonal dependency, and increased thoughtfulness. In another study, Maier et al. found that individuals with high levels of rigidity (men and women) and neuroticism (men only) were at risk for onset of depression (44).

Several studies of personality traits have been reported to predict response to antidepressant medication. For example, depressed individuals with assertive, independent, and competitive personality features were more likely to respond to medication treatment than were others who lacked these traits (33). High neuroticism in depressed persons predicted poor response to tricyclic antidepressants (33) and to lithium (45). In contrast, Davidson and his colleagues found that high neuroticism did not predict a poorer response to antidepressants (20).

Other personality traits that predicted outcome include sociotropy and autonomy. Sociotropy, a quality similar to interpersonal dependency, is a personality trait in which the individual seeks to have needs fulfilled by other people. An individual high in sociotropy is concerned about rejection from others and frequently acts in ways to please them to secure a strong interpersonal attachment. Autonomy refers to 14 qualities directed toward self-sufficiency. A highly autonomous individual is one who is concerned about personal failure and who strives to maximize control over his or her environment. Subjects with high autonomous and low sociotropic traits were more likely to show a greater response to antidepressants than were those subjects who displayed low autonomous and high sociotropic traits (58).

Depressed individuals with an associated personality disorder did not respond as well to antidepressants as those without personality disorders (33). In contrast, Davidson and his colleagues found that the presence of a personality disorder did not predict a poorer response to antidepressants (20).

In summary, personality traits and disorders have been found to be highly predictive of response to antidepressant medication. High neuroticism predicts a poor response to medication. Positive predictors of response to pharmacotherapy include high autonomy–low sociotropy, assertiveness, independence, and competitiveness.

The findings with respect to personality disorders are not as clear. One study reported that in depressed individuals with associated personality disorders, response to antidepressants was reduced. Another study did not find this.

Life events have been associated with the onset of depression for many years. Paykel and his colleagues found that depressed patients, when compared with matched general population controls, had three times the number of life events in the 6-month period preceding the onset of depression (57). Other studies conducted more recently have been reasonably consistent in supporting this association (6, 43).

In their longitudinal study of 65 depressed college students, Needles and Abramson found that individuals who displayed a reduction in hopelessness and experienced positive events were more likely to have a remission of depressive symptoms and subsequent recovery (51). In corroboration, Paykel and Cooper reported that, when events occurred during treatment, negative events were more predictive of poorer outcome and neutral events were more predictive of better outcome (56).

Other studies have examined the relationship between life events and relapse. Hunt and his associates found that in approximately one-third of their 62 bipolar patients, a large number of severe life events had occurred during the month immediately preceding relapse. However, the actual number of manic and depressive relapses were not different (31).

O'Connell et al. found a trend for the poorer outcome groups to have experienced more life events in the year before a depressive episode (53).

Some specific drug and life event interactions have been reported. In a study of 116 patients suffering from major depression with melancholia, those individuals treated with imipramine who had an absence of life events during a 6-month follow-up had a better outcome at 6 months. For individuals treated with phenelzine, however, an absence of life events prior to the onset of an episode predicted a poorer outcome at 6 months (69).

In summary, life events are associated with depression, with relapse, and with prediction of response to medication. The more negative a life event, the more possibility of a relapse into a depressive episode.

The prevalence of depression is increased among the bereaved, especially among those who have lost a spouse or a child. Widows and widowers were found to be at an increased risk for developing major depression compared with married women and men (79). In their study of widowhood, Clayton et al. found that 35% of individuals had enough symptoms to meet criteria for major depression 1 month after the death of a spouse. At 4 months, 25% met the criteria, and at 1 year, 17% met the criteria (15).

In an analysis of 350 widows and widowers at 2 and 7 months following the death of their spouses, Zisook and Shuchter found that subjects who were most susceptible to depression were those with a history of major depression and who were younger (79).

In treating depression associated with bereavement, two studies have shown that the depressive symptoms improve with antidepressant treatment (32, 55). However, even though the depressive symptoms do improve substantially, there is little, if any, improvement in the manifestation of bereavement (e.g., the intensity of grief). These differential responses to treatment suggest that major depression and bereavement are separate and distinguishable phenomena (55) and that treatment of depression may foster the process of grieving, not block it or "cover it over."

Whether depressions associated with bereavement respond better to pharmacotherapy or to psychotherapy is not known.

The chapter, thus far, has addressed demographic and psychosocial predictors of outcome. This section will address two important clinical predictors of response to pharmacotherapy—anxiety and substance abuse. These two clinical factors were selected because of their high comorbidity with depression.

Anxiety disorders are the second most common psychiatric disorders in the United States and often cooccur with depression. Between 15% and 33% of depressed patients have frank panic attacks (14).

The presence of anxiety symptoms in patients with major depression predicts more intense severity and greater likelihood of chronicity of depressive episodes, a decreased ability to respond to treatment, an increase in family prevalence of anxiety and/or depression, and increased impairment in social and vocational functioning (42). Outpatients with unipolar depressive disorder with higher ratings for anxiety recovered more slowly than those with lower levels of anxiety (14).

McLeod and colleagues found that a lifetime history of anxiety disorders significantly predicted a slower rate of recovery in their sample of outpatients with depression. They also found that a lifetime history of substance abuse predicted a slower rate of recovery (47).

In summary, anxiety disorders frequently cooccur with depression. Their cooccurrence predicts a slower recovery rate.

According to the ECA, the lifetime prevalence rate of alcoholism is 13.8% in the total population (25), and 6.2% had a history of drug abuse/dependence (5). The prevalence rates for major depressive disorder for alcohol abuse, drug abuse, and other mental disorders was 2.3% at 1 month, 3.0% at 6 months, and 5.9% for lifetime (62).

A history of alcohol or substance abuse in patients with depression increases the likelihood of hospitalization, committing suicide, and not complying with treatment (4).

Current alcohol and drug abuse in depressed individuals is known to hamper active treatment and is predictive of poor outcome in response to antidepressant treatment (3, 53). In the O'Connell et al. study of bipolar disorder, 36% of the people in the poor outcome group were found to have substance abuse problems but only 7% of the people in the good outcome group (53). Akiskal found that concurrent sedative or alcohol abuse with depression was more likely to be associated with a poorer response to antidepressants (3).

In summary, the cooccurrence of substance abuse with depression is very common. Alcohol and drug abuse, in combination with depression, are predictors of poor outcome to medication treatment.

Earlier attention in this chapter has been on psychosocial variables at index, their relation to depression, and their value as predictors of treatment response. Now, the focus is on these variables as outcome measures in depressed individuals.

In several follow-up studies, patients with major depression were found to be at higher risk of impairment in physical, social, and role functioning which resulted in lower levels of overall functioning (46, 75).

Wells and colleagues demonstrated that outpatients with major depression had poorer physical functioning and feelings of well being than did patients with chronic medical conditions, including arthritis, hypertension, diabetes, and back pain. The only patients who scored lower than depressive patients on these measures were those with current heart conditions (75).

Antidepressant therapy resulted in a significant improvement in functioning, generally, and a significant improvement in work and house functioning, specifically (2). Improved social and vocational functioning was seen in responders following treatment with imipramine (40).

When compared with symptom improvement, work functioning appeared to have a delayed reaction. After taking medication for approximately 6 weeks, symptoms in depressed patients began to lessen; but the bulk of improvement in work functioning did not occur until the patient had been taking medication for 6 to 24 weeks (23).

In summary, work functioning improves with successful treatment of depression, particularly antidepressants; but there appears to be a delay in improvement in work functioning compared with symptomatic improvement.

Persons with major depression were found to have more marital and family problems than those without the disorders. In a study by Rounsaville et al. (66) of 76 moderately depressed married women who received outpatient maintenance treatment for depression, approximately 25% of those with marital disputes at treatment onset had a substantial improvement in their marriage over the course of 8 months of treatment. Those whose marriages did not improve were more symptomatic than those whose marriages improved. The married women in the dispute group at index were found to be more socially impaired (in terms of adjustment) than the married women in the nondispute group. Despite the poor prognosis for depressed women with marital disputes, those who did show improvement in their marital relationships were more likely to experience an improvement in depressive symptoms and overall social functioning by the end of their treatment.

In a follow-up study, persons with major depression were at an increased risk for not getting along with their partners and for not being able to confide in their partners (46). In support of this, Weissman showed that there was an approximate sixfold increase in the likelihood that persons with major depression would have trouble functioning in their marriage or within their families (73).

Sociodemographic and psychosocial variables clearly are related to depression and may be important in its etiology. There has been a trend in the last 50 years for the age of onset of depression to become younger. Nearly all types of depression are two times more prevalent in women than in men. Depression is more prevalent among the unmarried, and the onset of depression is more likely among the unmarried. There is some increase in prevalence of mood disorders (i.e., depression) in the lowest socioeconomic classes. However, these same variables have not been found to predict outcome generally or to influence the response to pharmacotherapy. There have been scattered findings reported in the literature, but no consistent pattern has emerged, that would suggest that these variables are helpful in predicting response to pharmacotherapy.

Although marital status per se does not predict outcome, the quality of marital, familial, and social relationships does seem to be an important predictor of outcome and response. A number of studies have found that calm, positive, steady support is associated with and predictive of a better outcome, a better response to medication, and often a speedier recovery. On the other hand, high criticism, distress, less support, and high expressed emotion is predictive of a poorer outcome in general and a poorer response to specific treatment. In addition, these same variables are predictive of relapses.

Other important predictors are personality features and disorders. High neuroticism, low autonomy, high sociotropy, are predictive of a poor response to tricyclic antidepressants. The presence of personality disorders also has an adverse affect on response to treatment. Comorbid anxiety disorders and substance abuse complicates the course of depression and reduces positive outcomes.

Depression rarely produces nonsymptomatic outcomes. Patients with depression appear to suffer from impairments in physical, social, and role functioning and a variety of other problems, including many common general medical illnesses. Successful treatment is associated with an improvement of psychosocial outcomes, but improvement in these outcomes tends to temporally lag behind symptomatic outcomes. This makes sense intuitively in that interpersonal relationships in role functioning often take time to improve.

Whether patients achieve nonclinical outcomes and what the effects of depression are on quality of life over time are very important questions, and should receive more attention in the future.

The author would like to express his thanks and appreciation to Ms. Victoria Trimm whose assistance in the preparation of this chapter was invaluable.

published 2000