Research report
Discrepancies between self and observer ratings of depression: The relationship to demographic, clinical and personality variables

https://doi.org/10.1016/S0165-0327(99)00156-1Get rights and content

Abstract

Background: The observer-rated Hamilton depression scale (HamD) and the self-report Beck Depression Inventory (BDI) are among the most commonly used rating scales for depression, and both have well demonstrated reliability and validity. However, many depressed subjects have discrepant scores on these two assessment methods. The present study evaluated the ability of demographic, clinical and personality factors to account for the discrepancies observed between BDI and HamD ratings. Method: The study group consisted of 94 SCID-diagnosed outpatients with a current major depressive disorder. Subjects were rated with the 21-item HamD and completed the BDI and the NEO-Five Factor Inventory. Results: Younger age, higher educational attainment, and depressive subtype (atypical, non-melancholic) were predictive of higher BDI scores relative to HamD observer ratings. In addition, high neuroticism, low extraversion and low agreeableness were associated with higher endorsement of depressive symptoms on the BDI relative to the HamD. In general, these predictive variables showed a greater ability to explain discrepancies between self and observer ratings of psychological symptoms of depression compared to somatic symptoms of depression. Limitations: The study does not determine which aspects of neuroticism and extraversion contribute to the observed BDI/HamD discrepancies. Conclusions: Depression ratings obtained with the BDI and HamD are frequently discordant and a number of patient characteristics robustly predict the discrepancy between these two rating methods. The value of multi-modal assessment in the conduct of research on depressive disorders is re-affirmed.

Introduction

Rating scales for depression play an essential role in depression research, and have great utility for clinical assessment of the severity of depressive illness. Two of the most commonly used rating scales for depression are the Beck Depression Inventory (BDI) – a self-report instrument (Beck et al., 1961; 1978) and the Hamilton Rating Scale for Depression (HamD) – an observer rated scale (Hamilton, 1960). Although studies have consistently shown that scores on these two measures are significantly correlated, estimates of the size of the correlation in acutely depressed samples have varied considerably, from as low as r=0.16 (Schnurr et al., 1976) to as high as r=0.73 (Davies et al., 1975). Most studies have reported correlations between r=0.4 to r=0.7 in depressed samples (Richter et al., 1998).

Although concordance rates between these self and observer ratings of depression are generally acceptable, significantly discordant ratings are obtained in a substantial number of patients (Bailey and Coppen, 1976, Domken et al., 1994, Prusoff et al., 1972, Sayer et al., 1993). Several explanations have been suggested for such discrepant ratings. It has been noted that the item content of the BDI and the HamD scales differ considerably, with the BDI emphasizing psychological and subjective experiences of depression and the HamD emphasizing somatic and vegetative symptoms (Brown et al., 1995, Moran and Lambert, 1983, Steer et al., 1987). A number of investigators have noted that the correlation between BDI and HamD scores increases during the course of antidepressant treatment (Bailey and Coppen, 1976, Sayer et al., 1993, Senra and Polaino, 1993) though this most likely represents a statistical effect resulting from broadening of the range of scores (Senra and Polaino, 1993). One prior study found that a subgroup of severely depressed patients with high HamD scores and low BDI scores accounted for most of the discrepancies in ratings (Sayer et al., 1993). However, in our outpatient mood disorders clinic we have been more impressed with a subgroup of patients who appear to have modest levels of depression symptoms by observer rating, but indicate a high level of symptoms of depression on self-report.

Factor analytic studies of the HamD and BDI have consistently demonstrated that the components of depression assessed by the two scales are substantially different (eg. Brown et al., 1995, Sayer et al., 1993, Steer et al., 1987). In fact, one combined factor analysis of the BDI and HamD found that the factor solution reflected primarily the different measurement approaches (self-report versus observer rating) (Steer et al., 1987). While these analyses clearly demonstrate that the BDI and HamD measure different aspects of depressive pathology, they do not identify the characteristics of the patients who differ substantially on self versus observer ratings.

Several previous studies have evaluated the relationship between patient characteristics and discrepancies between self and observer ratings of depression. Two recent studies of depressed patients have reported that BDI scores were negatively associated with age while HamD scores were not significantly correlated with age (Lyness et al., 1995, Wallace and Pfohl, 1995), indicating that older patients have a tendency to under-report depressive symptoms on self-rating relative to observer ratings. Studies of depressed adolescents have demonstrated significantly higher correlations between self and observer ratings of depression in girls than in boys (Ambrosini et al., 1991, Shain et al., 1990), perhaps because of greater variability in adolescent boys’ willingness to endorse depressive symptoms on self-report measures. Preliminary findings from one study (Sayer et al., 1993) indicated that less education was associated with relatively lower endorsement of self-report (BDI) depression symptoms compared with observer ratings (HamD). Three studies using different self rating scales have found that the non-endogenous subtype of depression was associated with greater endorsement of self-reported depression symptoms relative to observer ratings (Domken et al., 1994, Rush et al., 1987, White et al., 1984). Rush et al., 1987, suggested that these findings were “consistent with the idea that anxious, atypical, somaticizing depressives view themselves as more severely depressed than do clinicians.”

Using a computerized literature search we were only able to identify one article that reported the relationship between personality factors and discrepancies between self and observer ratings of depression. Domken et al., 1994, evaluated the relationship between a number of personality factors (neuroticism, self-esteem, and dysfunctional attitudes) and the discrepancy between self and observer rated forms of the Inventory for Depressive Symptomatology (IDS) in a group of 48 hospital treated patients with major depressive disorder. They found that low self-esteem and high levels of neuroticism and dysfunctional attitudes were associated with inflated self-report IDS scores relative to observer ratings. Their regression analysis showed that collectively these three variables accounted for 48% of the variance in rating discrepancies.

The present study sought to evaluate the relationship between broad higher order personality dimensions and discrepancies between the two most commonly used self and observer rated depression rating scales (the BDI and the HamD respectively) in a group of outpatients with carefully diagnosed major depressive disorder. In view of the disparate item content of these scales (noted earlier), analyses were conducted separately for ‘psychological’ and ‘somatic’ symptoms. Studies reported by most previous authors have examined demographic and clinical variables and the difference between self and observer ratings of depression in relative isolation. In addition to evaluating personality dimensions, we attempted to replicate the findings of other authors regarding the effects of demographic variables (age, gender, educational attainment) and clinical variables (depressive subtype, number and duration of depressive episodes) on the difference between self and observer ratings of depression. To evaluate the importance of personality factors in explaining discrepancies between self and observer ratings, we controlled for the effect of significant clinical and demographic variables.

Section snippets

Subjects

The sample consisted of 94 adult outpatients (52 women and 42 men, mean age=43.68 years, SD=13.43). All patients had a DSM-IV (American Psychiatric Association, 1994) diagnosis of major depression and were non-psychotic. The most common co-morbid diagnoses were panic disorder (22.3%), social phobia (22.3%), and dysthymia (24.5%). None of the subjects had a current substance abuse disorder. Seventy six percent of subjects had completed high school and 19.1% had a university degree.

Materials and procedure

Patients were

Results

The mean BDI score was 30.1 (S.D.=9.0) with a range from 11 to 49. The mean HamD score was 23.1 (S.D.=4.5) with a range from 10 to 34. The correlation between HamD and BDI scores in the study group was r=0.40, P<.001. Correlations between depression discrepancy scores, demographic, clinical and personality factors are presented in Table 1. Cohen (1992) has specified effect size criteria to describe the degree of correlation. Cohen suggested that an r≥0.3 indicates a medium sized correlation and

Discussion

The present study demonstrates that a number of demographic, clinical and personality features of depressed patients are robustly related to discrepancies between self and observer ratings of depressive symptom severity. While a significant correlation was observed between total scores on the BDI and the HamD (r=0.40), it should be borne in mind that this correlation indicates that only 16% of the variance in scores on one measure can be explained by scores on the other measure. This highlights

References (31)

  • J Cohen

    A power primer

    Psychol. Bull.

    (1992)
  • P.I Costa et al.

    Revised NEO Personality Inventory (NEO-PI-R) and NEO Five Factor Inventory (NEO-FFI): professional manual

    (1992)
  • B Davies et al.

    A comparative study of four depression rating scales

    Aust. N.Z. J. Psychiatry.

    (1975)
  • M.B First et al.

    Structured clinical interview for DSM-IV Axis I disorders — patient edition

    (1995)
  • M Hamilton

    A rating scale for depression

    J. Neurol. Neurosurg. Psychiatry.

    (1960)
  • Cited by (0)

    View full text