The influence of cognitive variables on recovery in depressed inpatients

Journal of Affective Disorders 43 (1997) 207–212

Research report

The influence of cognitive variables on recovery in depressed inpatients a b, Robert Bothwell , Jan Scott * a

Formerly Lecturer in Psychiatry, Newcastle University, Now Consultant Psychiatrist Groote Schuur Hospital, Cape Town, South Africa b University Department of Psychiatry, Royal Victoria Infirmary, Newcastle upon Tyne NE1 4 LP UK Received 21 May 1996; revised 13 January 1997; accepted 13 January 1997

Abstract Forty two unipolar depressed inpatients were assessed on admission to hospital and again two years after the onset of the index episode. Fifty seven per cent of the sample (n 5 24) fulfilled NIMH recovery criteria within this period. Older age, female gender, severity of index episode, median prior duration of episode, higher levels of dysfunctional attitudes and low self-esteem significantly predicted chronicity of depression. Backward stepwise logistic regression identified that three of these variables measured at the time of admission: severity of index depression, higher levels of dysfunctional attitudes related to the need for approval, and low self-esteem provided a robust logistic model for predicting outcome. However, the small sample size and statistical analysis employed means that replication of our research is required. Larger scale studies could also include a factor analysis so that the common elements within the different instruments may be detected. If the association between cognitive dysfunction and chronicity is confirmed, we would recommend that inpatient treatment strategies are revised to incorporate more overt psychosocial interventions.  1997 Elsevier Science B.V.

1. Introduction In the past 15 years there have been a number of attempts to identify vulnerability factors for persistent depression. These studies fall into two broad categories: those which retrospectively identify a group of operationally defined chronic depressives and compare and contrast them with non-chronic major depressives (eg, Akiskal et al., 1981; Hirschfeld et al., 1986; Scott, 1988), and those which prospectively follow a cohort of major depressives *Corresponding author.

and analyse the pattern of recovery over time (eg, Andrew et al., 1993; Keller et al., 1982, 1984, 1986, 1992; Scott et al., 1992). The latter are now more common and allow systematic identification of the demographic and clinical variables which predict poor outcomes in depression. In a series of papers, Keller and colleagues reported that factors such as inpatient status, low family income, secondary subtype of depression, and ‘double depression‘ (major depression superimposed on an underlying dysthymia) were predictors of chronicity. Greater symptom severity, the presence of psychotic phenomena, longer prior episode length, and inadequate

0165-0327 / 97 / $17.00  1997 Elsevier Science B.V. All rights reserved PII S0165-0327( 97 )01431-6

208

R. Bothwell, J. Scott / Journal of Affective Disorders 43 (1997) 207 – 212

antidepressant treatment have also been associated with worse outcome (Coryell et al., 1990; Keller et al., 1982; Parker et al., 1992; Scott et al., 1992). In addition to broad-based demographic and clinical factors, a number of social and psychological variables have also been found to predict persistence of depression. Brown et al. (1994) noted that depressed individuals reporting interpersonal difficulties were at significantly greater risk of chronic depression than those without such difficulties. Premorbid neuroticism has repeatedly been associated with a risk of more protracted depression in retrospective and prospective community and hospital studies (for a review see Scott, 1988). The combination of high levels of neuroticism and melancholia may have a particularly adverse effect on prognosis (Duggan et al., 1990). In community and outpatient samples of mild to moderately severe depression, cognitive variables such as low self-esteem and dysfunctional attitudes both predict depressive chronicity (Brown et al., 1986; Miller et al., 1989; Dent and Teasdale, 1988; Peselow et al., 1990). Less data is available on severely depressed inpatients, but Williams et al. (1990) showed that individuals with higher baseline levels of dysfunctional beliefs were more likely to have persistent depressive symptoms at 6 weeks. In a smaller 12-month follow-up study (including 17 inpatients), Scott et al. (1995) showed that length of major depressive episode was significantly correlated with higher levels of premorbid neuroticism and dysfunctional beliefs on admission. Andrew et al. (1993) demonstrated that low selfesteem was the only significant psychosocial variable to predict one-year outcome in a sample of 59 female patients with depression. No study so far has looked prospectively at the influence of all three variables (neuroticism, dysfunctional attitudes and self-esteem) simultaneously on the outcome of inpatient depression. This study examines the role of these and other demographic and clinical factors in predicting depressive chronicity or recovery.

2. Method The study represents a follow up of 48 depressed inpatients who previously partcipated in a study of

the relationship between subjective and observer views of illness severity (Domken et al., 1994). The patients were recruited from a cohort of 60 consecutive admissions to inpatient units within Newcastle. Entry criteria were: evidence of appropriate and adequate antidepressant treatment (as defined by Cole et al., 1993), the presence of a non-psychotic, unipolar major depression meeting DSMIIIR criteria without any other co-existing Axis I diagnosis, absence of drug or alcohol abuse, and no cognitive impairment. At initial screening, 12 individuals were excluded, five failed to meet the entry criteria (one was unable to give informed consent, three failed to meet major depression criteria, one demonstrated cognitive impairment) and seven either refused to take part or failed to complete the required selfratings. As reported previously, the excluded groups did not differ significantly in demographic or illness variables from the final sample. Demographic and illness characteristics including endogenicity as measured on the Newcastle Diagnostic Index (NDI, Carney et al., 1965), length of illness episode and number of previous illness episodes were recorded as described previously (Domken et al., 1994; Scott et al., 1992). Observer ratings of severity of depression were assessed using the 17item version of the HRSD (Hamilton, 1960). Subjects then completed the 21-item BDI (Beck et al., 1961) and the N Scale of the Eysenck Personality Questionnaire (Eysenck and Eysenck, 1975) with an instruction to ensure that scores referred to how the patient regarded themself premorbidly (Kendell and Discipio, 1968). In addition, individuals completed the Rosenberg Self-Esteem Questionnaire (SEQ; Rosenberg, 1965) and the 40-item Dysfunctional Attitudes Scale (DAS; Weissman, 1979). The SEQ comprises 5 positive and 5 negative items rated on a 1–4 scale. Positive and negative ratings are summated to give the SEQ score with higher negative scores indicative of poor self-esteem. The DAS asks subjects to rate how strongly they agree or disagree with each belief statement on a 7-point Likert scale. Scores range from 40–280 with higher scores representing greater dysfunction. Two subscale scores measuring perfectionism and need for approval can also be derived. All ratings were completed at initial interview and at follow-up assessment. As with our previous study (Scott et al., 1992), patients were followed until recovery or until two

R. Bothwell, J. Scott / Journal of Affective Disorders 43 (1997) 207 – 212

years after the onset of the index episode (at this point the disorder is classified as chronic depression). The median length of inpatient stay was five weeks and the length of follow-up ranged from nine to 19 months. The psychiatric case files of all subjects were examined and permission to re-interview the patient was sought from the responsible psychiatrist or general practitioner. Recovery was defined according to NIMH criteria, that is ‘no or minimal symptoms of depression for eight consecutive weeks‘ (Keller et al., 1982; Frank et al., 1991). Treatment was not controlled, but no patient was being prescribed less than 100 mg of a tricyclic antidepressant (or its equivalent) per day and the median equivalent dose was 172 mg. Thirty patients were prescribed other medications at some time, the majority (n 5 22) receiving major tranquillisers. Descriptive statistics were used to identify the characteristics of the sample. The initial set of demographic, clinical, personality and cognitive variables were then analysed using backwards stepwise logistic regression to determine the best combination of predictors for logistic regression. This was undertaken by a medical statistician using the ‘MINITAB‘ statistical package.

3. Results During the course of the follow-up period, two patients could not be traced and the diagnosis had altered in four cases (to bipolar affective disorder in two cases, schizophrenia in one case and one individual with SLE also developed an organic brain syndrome). These six patients were excluded from the final analysis. Although three patients died during the course of the follow-up, it was possible to use contemporaneous information recorded in the outpatient casenotes to assign the patients to the appropriate outcome category. Of the 42 subjects included in the analysis, 15 were male and 27 were female. At initial interview, the mean age of the sample was 41.0 years (SD 13.9), 19 subjects were married or cohabiting. The mean scores on the HRSD and BDI were 20.7 (SD 6.1) and 32.3 (SD 11) respectively. Fourteen patients met criteria for endogenous depression and the median duration of the index depressive episode at the time of the assessment was six months. The

209

median number of previous depressive episodes was one (range 1–10) whilst the mean age at which the patients suffered their first ever episode was 31.3 years (SD 15). The mean N score for the sample was 15.5 (SD 5.3). Scores on the DAS ranged from 77–256 with a mean of 164.6 (SD 45). Mean SEQ score for the sample was 2 5.6 (range 2 20 to 1 9). There were no significant gender differences on any of the ratings undertaken. At the end of the follow-up period, 24 patients met criteria for remission, whilst 18 patients (43%) continued to suffer from depressive symptoms. Outcome as it relates to age, gender and baseline observer and self-ratings is shown in Table I. Female gender ( p , 0.05) and greater initial severity of index episode as measured on the HRSD ( p , 0.024) and BDI ( p , 0.0002) significantly predicted chronicity of symptoms. Longer duration of depressive episode prior to initial interview was also significantly associated with persistent depression ( p , 0.01). Endogenicity as measured on the NDI and N score as measured on the EPQ were not significantly related to outcome in this analysis, but initial high DAS score ( p , 0.027) and low SEQ score ( p , 0.0003) were both significantly associated with persistent major depression. High N score was significantly correlated with a high score on the DAS (r 5 0.5.; p , 0.001) and with a low SEQ score (r 5 0.4; p , 0.01). Low SEQ score was also significantly correlated with high global DAS score (r 5 0.7; p , 0.001) and high scores on both subscales (approval: r 5 0.5, p , 0.001; perfectionism: r 5 0.6; p , 0.001). Change in HRSD scores for the whole sample between the two interviews (mean reduction 8.4) significantly correlated with change in DAS score (r 5 0.37; p , 0.05), but not with SEQ or N score. The initial set of variables was analysed using backwards stepwise logistic regression to determine the best combination of predictors of outcome. The independent variables used were: age, gender, episode duration, number of previous episodes and HRSD, NDI, BDI, DAS approval subscale, DAS perfectionism subscale, N and SEQ scores. Backwards stepwise logistic regression revealed that a model including severity as measured at admission on the HRSD, global DAS score (the combined subscale scores), SEQ, gender and age significantly predicted outcome for 35 out of 42 (83%) subjects.

210

R. Bothwell, J. Scott / Journal of Affective Disorders 43 (1997) 207 – 212

This model had a sensitivity of 84% and a specificity of 83%. However, a model using only initial SEQ score had an odds ratio of 0.84 (95% CI 0.75, 0.94; p , 0.002). Using the actual coefficients from the SEQ logistic model, the probability of an individual not recovering given a range of SEQ scores was calculated. If the SEQ score was 1 10, the individual had only a 4% probability of chronicity, however, if the SEQ score was 2 10, the individual had a 60% probability of persistent depression. Although the specificity of this model was reasonable (78%), its sensitivity was only 68%. The most robust logistic model incorporated only three variables: the baseline scores on HRSD, DAS approval subscale and SEQ. It had a sensitivity of 82%, a specificity was 87.5% and correctly classified 36 individuals (85%) to either the recovered or chronically depressed outcome category. According to this model, an individual with a DAS approval score of 50, an SEQ score of 2 5, and an HRSD score of 15 has a 76% probability of recovery. However, an individual with the same DAS approval score but an SEQ score of 2 15 and an HRSD score of 30 has only a 2% probability of recovery. As this model has a negative predictive value of 87.5% it will correctly classify about 9 out of 10 cases to the recovery category.

4. Discussion Before discussing the results of this paper in detail, methodological limitations, notably the relatively small sample size, the naturalistic design, cross-sectional assessment procedure, and the number of variables assessed need to be borne in mind. In addition, we have previously noted that excluding involuntary and psychotic patients may have biased our sample towards less severe cases (Domken et al., 1994). The most serious limitation is that patients were only followed up for a total of two years after onset of the index depressive episode. Thus subjects with a longer episode duration prior to admission (and therefore before entry into the study) were followed for a shorter time after discharge. We attempted to overcome this problem by controlling for prior episode duration in our the statistical analyses. Also, it should be noted that the minimum

length of time between admission and final follow-up was over six months, suggesting that there was a significant time period in which response to the more intensive treatment offered by a hospital unit might have occurred. Although an ideal study would have followed subjects with persitent depression for longer, it is still apppropriate to make some judgements about the impact of hospital treatment on the outcome of depression. It is acknowledged that the odds ratio findings we report are modest and repetition of this study is required in a larger sample. Also, a more realistic measure of the sensitivity and specificity of a logistic model can only be obtained by applying it to different data sets rather than the sample from which it was originally derived. Our outcome data do compare reasonably with previous studies. Keller et al. (1982), employing NIMH criteria for recovery and life table analyses found that about 50% of 121 depressed in- and outpatient subjects recovered in the first year of prospective follow-up and a further 28% in the second year. In our study, 57% of the sample met recovery criteria at two years. The higher proportion of subjects having persistent depression probably reflects the severity and complexity of symptoms and problems afflicting an inpatient sample. It may also reflect inadequate treatment (or noncompliance), however, median dosage of antidepressant medication received by patients in this study was higher than that reported by Keller et al. (1982). This study demonstrates that illness severity, dysfunctional attitudes, particularly those relating to the need for approval, and low self-esteem can predict persistence of depressive symptoms after inpatient care even when prior duration of episode is taken into account. The results support previous research on outpatient samples. The findings of previous studies (including our own) that age, gender, prior duration of illness episode, past history and severity of index episode independently predict chronicity, were also confirmed. Although we failed to demonstrate that high N score predicted outcome, we did show that high N score was significantly correlated with other identified predictors namely low self-esteem and high levels of dysfunctional attitudes. These findings may support the view that when assessing neuroticism, clinicians are actually tapping into cognitive factors that enhance vul-

R. Bothwell, J. Scott / Journal of Affective Disorders 43 (1997) 207 – 212

nerability to persistent depression (Teasdale, 1988). In future, measuring dysfunctional attitudes and utilising self-esteem questionnaires may be a more specific way of assessing this vulnerability. It would be particularly useful to undertake a larger scale study focusing on a factor analysis of the different questionaires. If the common components of these tools could be identified, it may be possible to devise a more reliable and valid measure of cognitive vulnerability. Clinicians dealing with depressed individuals in inpatient settings are usually focused initially on illness severity and the need for biological treatments of the disorder. The inpatients in this study are very similar to depressed individuals admitted elsewhere, they received comprehensive pharmacological treatments and generic psychosocial input, but none were referred for specific psychological therapies such as cognitive therapy or interpersonal therapy. Episode severity and DAS score were important predictors of outcome. As the severity of the depression subsided, some reduction in dysfunctional attitudes occurred. However, higher scores on the need for approval subscale of the DAS were associated with nonrecovery even when severity of depression was controlled for. These DAS findings parallel those reported by Reda et al. (1985) who found that patients treated with cognitive therapy or with antidepressant drugs both showed significant reductions in HRSD scores and DAS scores over a period of 12 months. However, in the drug-treated group, there were certain ‘drug resistant‘ dysfunctional attitudes which did not change, and which may require additional psychological interventions. Our study suggests that targeting an individual’s ‘approval‘ schema may be particularly beneficial. This study confirms previous research that suggests that self-esteem is an important factor in predicting onset and recovery from depression (Andrew et al., 1993; Brown et al., 1990). We found that the probability of a patient ‘not recovering‘ if the initial SEQ score was 2 20 was about 90%. Although low self-esteem may be both a cause and a consequence of depressive disorder, this study again reinforces the need to target clinical interventions at improving self-esteem if we are to enhance the prospects of full recovery. Recent discussions about the treatment of depres-

211

sion have suggested that there is a need to identify patients who might benefit from a combined pharmacological and psychological approach to their depression as opposed to receiving either treatment in isolation (Scott, 1995). In this study, it was found that more severely depressed inpatients with low self-esteem and high levels of dysfunctional attitudes treated with adequate and appropriate pharmacotherapy, generic inpatient care and routine outpatient follow-up, were at high risk of remaining depressed. Our results suggest that inpatients with this profile, especially those who are admitted over a year after the onset of their depression, may particularly benefit from a combined approach with medication that initially targets severe vegetative symptoms and the use of more systematic psychological therapies that target self-esteem and dysfunctional attitudes.

Acknowledgments Robert Bothwell was supported by a grant from the MRC during the preparation of this paper. We wish to thank Dr Marc Domken (psychiatrist) and Dr Peter Kelly (statistician) for their input to this project.

References Andrew, B., Hawton, K., Fagg, J., Westbrook, D. (1993) Do psychosocial factors influence outcome in severely depressed psychiatric inpatients. Br. J. Psychiatry 163, 747–54. Akiskal, H.S., King, D., Rosenthal, T., Robinson, D. and ScottStrauss, A. (1981) Chronic Depressions. Part I. Clinical and familial characteristics in 137 probands. J. Affective Disorders 3, 297–315. Beck, A.T., Ward, C., Mendelson, M., Mock, J. and Erbaugh, J. (1961) An inventory for measuring depression. Arch. Gen. Psychiatry 4, 561–571. Brown, G.W., Andrews, B., Harris, T.O., Adler, Z., and Bridge, L. (1986). Social support, self-esteem and depression. Psych. Med. 16, 813–831. Brown, G.W., Bifulco, A., Andrews, B. (1990) Self esteem and depression: IV. Effect on course and recovery. Soc. Psychiatry Psychiatr. Epidemiol. 25, 244–49. Brown, G.W., Harris, T., Hepworth, C., Robinson, R. (1994) Clinical and psychosocial origins of chronic depressive episodes. Br. J. Psychiatry 165, 457–65. Carney, M., Roth, M., Garside, R. (1965) The diagnosis of

212

R. Bothwell, J. Scott / Journal of Affective Disorders 43 (1997) 207 – 212

depressive syndromes and the prediction of ECT response. Br. J. Psychiatry 111, 659–674. Cole, A., Scott, J., Eccleston, D., Ferrier, I. (1993) Patterns of resistance in bipolar affective disorder. Acta Psychiatrica Scandinavica 88, 121–123. Coryell, W., Endicott, J., and Keller, M. (1990) Outcome of patients with chronic affective disorder: a five year follow-up. Am. J. Psychiatry, 147 (12), 1627–1633. Dent, J., Teasdale, J. (1988) Negative cognition and the persistence of depression. J. Abnormal Psychol. 97, 29–34. Domken, M., Scott, J., and Kelly, P. (1994) What factors predict discrepancies between self and observer ratings of depression? J. Affective Disorders, 31, 253–259. Duggan, C., Lee, A., and Murray, R. (1990) Does personality predict long-term outcome in depression? Br. J. Psychiatry 157, 19–24. Eysenck, H.J., and Eysenck, S.B.G. (1975). Manual of the Eysenck Personality Questionnaire. London: London University. Frank, E., Prien, R., Jarret, R., et al. (1991). Conceptualisation and rationale for consensus definitions of terms in major depressive disorder. Remission, recovery, relapse and recurrence. Arch. Gen. Psychiatry 48, 851–55. Hamilton, M. (1960). A rating scale for depression. J. Neurol. Neurosurg. Psyhciatry 23, 56–62. Hirschfeld, R., Klerman, G., Andreasen, N., et al. (1986). Psychosocial predictors of chronicity in depressed patients. Br. J. Psychiatry 148, 648–54. Keller, M.B., Shapiro, R., and Lavori, P.W., et al. (1982). Recovery in major epression. Analysis with the life table and regression model. Arch. Gen. Psychiatry 39, 905–10. Keller, M., Klerman, G., and Lavori, P., et al. (1984). Long-term outcome of episodes of major depression. Clinical and public health significance. J. Am. Med. Assoc. 252, 788–92. Keller, M., Lavori, P., Rice, J., et al. (1986). The persistent risk of chronicity in recurrent episodes of non-bipolar major depressive disorder: a prospective follow–up. Am. J. Psychiatry 143, 24–28. Keller, M., Lavori, P., Mueller, T., et al. (1992) Time to recovery, chronicity and levels of psychopathology in major depression:

a five year prospective follow-up of 431 subjects. Arch. Gen. Psychiatry 49, 809–16. Kendell, R.E., and Discipio, W.J. (1968) Eysenck Personality Inventory scores of patients with depressive illness. Br. J. Psychiatry 114, 767–770. Miller, P., Kreitman, N., Ingham, J. and Sashidharan, S. (1989) Self esteem, life stress and psychiatric disorder. J. Affective Disorders 17, 65–75. Parker, G., Hadzi-Pavlovic, D., Brodaty, H., Boyce, P., Mitchell, P., Wilhelm, K., and Hickie, I. (1992) Predicting the course of melancholic and non-melancholic depression: a naturalistic comparison study. J. Nervous and Mental Dis. 180, 693–702. Peselow, E.D., Robins, C., Block, P., Barouch, F., and Fieve, R. (1990) Dysfunctional attitudes in depressed patients before and after recovery and in normal control subjects. Am. J. Psychiatry 147, 439–44. Reda, M., Carpiniello, B., Secchiaroli, M., Blanco, S. (1985) Thinking, depression and antidepressants: modified and unmodified depressive beliefs during treatment with Amitriptyline. Cognitive Therapy and Res. 9, 135–43. Rosenberg, M. (1965) The Measurement of Self-esteem. Society and the Adolescent Image. Princeton: Princeton University Press. Scott, J. (1988). Chronic depression. Br. J. Psychiatry, 153, 287–297. Scott, J. (1995). Editorial: Psychological treatment of depression: An update. Br. J. Psychiatry 167, 289–92. Scott, J., Boyes, R., Eccleston, D. (1992) Can we predict the persistence of depression? Br. J. Psychiatry 161, 633–37. Scott, J., Williams, J.M.G., Brittlebank, A. and Ferrier, I.N. (1995) The relationship between premorbid neuroticism, cognitive dysfunction and persistence of depression: a one year followup. J. Affective Disorders 33, 167–72. Teasdale, J.D. (1988). Cognitive vulnerability to persistent depression. Cognition and Emotion 2, 247–74. Weissman, A.N. (1979). The dysfunctional attitudes scale: a validation study. Dissert. Abstr. Intern. 40, 1389–1390. Williams, J.M.G., Healey, D., Teasdale, J., White, W., and Paykel E.S. (1990) Dysfunctional attitudes and vulnerability to persistent depression. Psychol. Med. 20, 375–81.