The important role of a supportive social environment in preventing the onset of unipolar depression now seems widely accepted. Social support also seems to have a beneficial effect on the course of depressive disorders, while negative social relationships increase the probability of relapses after remission from clinical depression.
However, not all positive, supportive relationships are beneficial. In a series of recent studies of depressed psychiatric in-patients, we showed that the quality and structure of social support networks encountered after remission and discharge from hospital affected the further course of the disorder in quite specific ways: a patient’s general satisfaction with available support had a beneficial effect, whereas a large kin support network was associated with a subsequent increase in depressive symptoms. (The size of the support network of friends and acquaintances on the other hand, although cross-sectionally correlated with indicators of depression, had no appreciable prospective effects.) Such a separation of effects was made possible by a differentiated assessment of supportive environments. The results, however, as well as those of Billings and Moos (1985), were obtained with continuous questionnaire scores as measures of depression, and this study examines whether the same relationships hold when diagnosable cases of depression are to be predicted.
Over-all-Design. This study is part of an ongoing major prospective research program on psychosocial predictors of the course of clinical depression after discharge from psychiatric in-patient treatment. Patients are interviewed 1 and 25 months, after discharge, when depressive symptomatology, potential and actual social support factors, psychosocial needs, stressful life events, habitual coping tendencies and actual coping behaviours are assessed. Instruments and Variables. At both Ti and at T2 the following instruments (among others) were used: the Inventory to Diagnose Depression (IDD; Zimmermann and Coryell 1987), the Present State Examination (PSE-9; Wing et al. 1982) (questions added to allow DSM-III diagnoses of a major depressive episode (MDE)), and the Mannheim Interview on Social Support (MISS; Veiel 1987, 1990). The IDD is a self-report questionnaire which also covers symptoms relevant for a DSM-III diagnosis of MDE and from which a symptom sum score was derived. Based on the expanded PSE two dichotomous variables were defined: “caseness” at Ti (MDE/no MDE), which was used as a selection criterion (see below), and “relapse/no relapse”, which was defined by the reappearance of depressive symptoms meeting MDE criteria at any time after T3 (up to and including T2). The MISS is a structured interview which maps a person’s supportive environment. Three global parameters were derived from the MISS: (a) the patient’s general satisfaction with available social support, averaged over 12 prototypical support functions; (b) the size of the kin support network (excluding spouses but including in-laws); and (c) the size of the support network of friends and acquaintances. The variables were named “Satisfaction”, “Kin Network Size”, and “Non-kin Network Size”, respectively. Sample Selection. The sample was based on a cohort of 138 consecutively admitted 18 to 60-year-old patients of the Psychiatric Clinic of the Central Institute of Mental Health in Mannheim, FRG, who met the criteria for a DSM-III diagnosis of MDE or who scored 25 or higher on the IDD (corresponding to one standard deviation above the mean of normal samples). Patients with bipolar disorder, current or past schizophrenic spectrum symptoms, primary substance abuse, or organic brain damage were excluded. One hundred and eleven patients were interviewed at T1; and 99 were followed-up at T2. Of these, 35 patients who were unremitted at Ti were excluded. Thus the final sample consisted of 64 discharged inpatients who had remitted from unipolar depression.
Analyses and Results
Linear regression analyses with the IDD-derived, continuous symptom sum score at T2 as the dependent variable, and logistic regression analyses with PSE-derived relapse as the dichotomous criterion were performed in parallel. In both analyses, the following variables which had shown substantial predictive power on their own were forced into the prediction equation; the patients’ sex, the IDD sum score at T\, and the sum of all PSE items at Ti on which the MDE diagnosis was based. All three support variables were then entered into the regression equations and, by a backward-stepping procedure, subsequently removed if their regression coefficient had an associated chance probability of P 2 0.10. Thus, the power of the social support variables to predict T2 depression was evaluated after the effects of the patients’ sex and of the Ti symptom level had been removed. In order to make the results of the two analyses comparable, the net effects of the predictor variables are represented as partial correlations with the respective criteria. For the regression analyses, straightforward partial product-moment correlations are used. For the logistic regression, the chi-square associated with each predictor variable has been transformed into a partial phi correlation coefficient. The social support variables significantly predicted general depressive symptom levels at T2. As expected, satisfaction was negatively, and the size of the kin network was positively correlated with the T2 symptom score. Clinical relapses, however, were not predicted by the support variables. This discrepancy in predictive power could have been caused by any of the following four factors: (a) the difference between self-reported (IDD) and interviewerrated (PSE) symptomatology: (b) the emphasis on severe levels of depression in the relapse criterion (MDE), as opposed to the broad range of symptom levels reflected in the sum score; (c) the difference between a symptom count (IDD sum score) and a non-summative diagnostic algorithm (PSE-MDE); (d) the emphasis on different kinds of symptoms. In order to assess (a), the linear regression analysis was repeated with a sum score of the PSE depression port variables on symptomatic changes was not restricted to low-to-medium symptom levels, nor was it specific to self-reported symptoms. In other words, it was not due to scaling or assessment artefacts. In order to decide between the remaining two possibilities (diagnostic algorithm vs. symptom count, and emphasis on different kinds of symptoms), the IDD sum score was split into a sum score of mood-related items (12) and a sum score of the remaining items (10), comprising mainly vegetative (appetite, sleep etc.) and non-specific (e.g. irritability) symptoms. The split was made roughly in accordance with the distinctions obtained by Goldberg et al. (1987). Table 3 shows the results when the “mood” sum score and the “vegetative” sum score were predicted separately. Compared with the total symptom score, the support variables predicted the “mood” score even better, but the correlations were lower for the “vegetative” score. In a second set of analyses, the “vegetative” and “mood” scores at T2 were predicted after removing their common variance, i.e. by forcing the “vegetative” score at T2 into the prediction equation for the “mood” score at T2, and vice versa. In Table 3, the figures in brackets represent the effects of the support variables on the “purified” scores; while significant for the “mood” score, they are practically zero for the “vegetative” score. (It is also interesting to note that the associations of the other predictor variables — Tj symptom scores and sex — with mood and with vegetative symptoms had opposite signs.) Thus the effect shown in the left column of Table 1 must be attributed to the mood component of the IDD sum score, and the failure to predict clinical relapses is because mood-related symptoms are weighted much more heavily in symptom sum scores than in MDE diagnoses.
Among the various facets of depression the mood component seems to be primarily responsive to psychosocial influences. This finding is consistent with the view held by many clinicians that clinical or “major” depression is qualitatively as well as quantitatively different from “normal” variations of mood. Two recent prospective studies also found different predictors for interview-based clinical diagnoses of depression and for symptom scores (Le- winsohn et al. 1988; O’Hara et al. 1984). Neither study, however, has examined the details of this difference. The authors — as indeed most clinical authors — have apparently assumed a kind of hierarchical relationship between depressed mood and “real”, severe, or major depression, where the former appears as a precursor as well as a symptom of the latter. Lewinsohn et al. (1988), for example, postulated that negative affect mediates the effect of cognitive vulnerability factors on major depression. Such a hierarchical organization seems only plausible, however, if one accepts the notion of a homogeneous entity “depression”, which is reflected in a clinical diagnosis. Given the symptomatic and nosological diversity of depressive disorders, however, a non-hierarchical conceptualization is perhaps more adequate. The complex psychological and physiological changes associated with depressive disorders may be organized in clusters reflecting the activity of distinct, relatively autonomous, and only moderately correlated psycho-physiological systems, each with its specific set of associated triggers and vulnerability factors. One such subsystem, whose activation is experienced by the individual himself or herself and by his or her environment as depressed mood and anhedonia, may be particularly responsive to (negative and positive) social experiences. Other sub-systems are conceivable, in particular one primarily responsive to stress experiences and involving arousal symptoms associated with changes in the adrenal hormonal system (such as tension, irritability, sleeplessness, and, eventually, exhaustion and inactivity — cf. Selye 1974). While the notion of autonomous subsystems with possibly different degrees of specificity for depression is intriguing and carries some potential for clarifying such issues as “double depression” (Keller and Shapiro 1982) and the overlap of anxiety and depressive disorders, it is still speculative. Research with a variety of other predictors (stress, physiological parameters) is needed, and partly under way, to obtain a clearer view.