Professional Documents
Culture Documents
0021-843X/93/S3.00
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
198
arately. Thus, the results do not provide firm evidence for lack
of descriptive validity.
It should be noted that the way in which subjects were assigned to the two groups in the present study means that the full
set of symptoms must occur more often in the MDE group than
in the contrast group; however, this need not be true for each
individual symptom. In the Clarkin et al. (1983) study, one of
the eight symptoms of borderline personality disorder (BPD)
did not occur significantly more often in cases of BPD than in
noncases (our calculation), even though the presence of five or
more symptoms was required for a diagnosis of BPD.
199
Method
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Subjects
The subjects were outpatients seeking psychiatric evaluation in a
county hospital in Dallas, Texas. The hospital, Parkland Memorial
Hospital, operates a 24-hr psychiatric emergency service for any individual who requests evaluation or is referred by court order. It serves a
wide variety of patients, but primarily those from lower socioeconomicstrata.
Subjects were selected for the present study from among 2,276 consecutive admissions (seen at all hours of the day or night) over a period
of 2.5 months. To be selected, individuals had to be at least 18 years old
and had to meet Criterion A for a DSM-III diagnosis of MDE: "dysphoric mood or loss of pleasure in all or almost all usual activities and
pastimes. . . [which is] prominent and relatively persistent" (American Psychiatric Association, 1980, p. 213). We added the restriction
that the mood disturbance must have lasted for at least 2 weeks.
Some subjects who might have met Criterion A were excluded from
this study as a result of certain screening rules. Patients who met Criterion A but who reported taking psychotropic medication at some time
during the 3 weeks prior to evaluation were excluded because medication may have eliminated some of their symptoms (n = 61). Similarly,
patients with a diagnosis of major affective disorder in remission (n =
4) were excluded because they would have fewer symptoms than during major depressive episodes. Patients diagnosed as suffering from
organic mental disorder (n = 40), or abuse of or dependence on alcohol
(n = 190) or other substances (n = 213), were also excluded unless any
depressive symptoms they might have had were clearly unrelated to the
use of the substance. Many of these patients were not fully evaluated
for depressive symptoms once these diagnoses were established because the diagnoses were known to be exclusion criteria for this study.
Furthermore, many of these patients were intoxicated on admission.
Patients were also excluded if their diagnoses did not permit clear
placement as cases or noncases of MDE. The latter included 9 patients
with a diagnosis of schizoaffective disorder and 4 cases of residual
schizophrenia that met the criteria for MDE and should thus be diagnosed as cases of atypical depression under the DSM-III. (The majority of these patients would have been excluded because they were
maintained on medication, in any case.) Finally, 193 patients classified
under the heading "additional codes" in the DSM-III were excluded.
These included two subgroups: (a) those who were so hostile (many
brought in on warrants) or so psychotic that the examiner believed that
accurate diagnostic data could not be obtained and (b) those judged to
have minor problems (often financial or social in nature) and to be free
of mental disorder.
After application of the screening rules, there were 168 cases that
met Criterion A and for which full information on depressive symptoms was available. Of these, 111 had a diagnosis of MDE, including
107 diagnosed as cases of major depression and 4 as cases of bipolar
disorder depressed (BP).1 (Data on the BP cases are not included in the
analyses because there were too few of them.) An additional 57 patients
met Criterion A but not Criterion B (non-MDE). The distribution of
diagnoses for the non-MDE group is shown in Table 1.
The sample was predominantly female (66%), White (58%), and between 20 and 40 years of age (75%). The MDE and non-MDE groups
did not differ significantly on any of these variables. For sex, 67% of
MDE and 63% of non-MDE cases were women, x2 (1, N= 164) = 0.284,
p > .50; for race, 55% of MDE and 63% of non-MDE cases were White,
X2(l, N= 164)= 0.978, p> .30; and forage, the means were 33.02 (SD =
10.04) years for MDE and 31.02 (SD = 10.32) for non-MDE, ;('62) =
1.31,p>.10.
Procedure
In the psychiatric emergency service, each patient is routinely evaluated by one or more psychiatric attending physicians, residents, or
Table 1
Frequency of Diagnoses in Non-MDE Subjects in Study
Sample (Principal Diagnosis Only)
Diagnosis
Substance use disorders
Schizophrenic disorders
Affective disorders
Other specific affective disorders
Dysthymic
Cyclothymic
Atypical affective disorders
Atypical depression
Anxiety disorders
Somatoform disorders
Psychosexual disorders
Adjustment disorders
V codes"
n
3.5
5.3
10
1
17.5
1.8
4
1
1
1
29
5
7.0
1.8
1.8
1.8
50.9
Reliability Interviews
As a check on the reliability of assessments, 43 subjects (26% of the
sample) were evaluated by two clinicians, an initial interviewer and
David Rudick-Davis.2 The initial interviewer always began the interview alone, and, if she or he believed that the subject met Criterion A
for MDE, asked the second interviewer to join the interview if he was
available (n = 10) or, if not (n = 33), to conduct a second interview with
the subject on the same day. In conjoint interviews, either interviewer
could gather any needed information. In all instances, both interviewers had access to case records, observations of behavior by other
staff members, and reports of informants; if one interviewer had access
to information but the other did not, this was communicated to the
other clinician. Reliability interviews were conducted at random times
(to reflect day, evening, night, and weekend samples), and data were
recorded independently. Whenever possible, reliability interviews
were conducted to represent the relative proportion of a rater's contribution to the total sample; that is, raters with many subjects were in1
200
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Reliability
For 39 of the 43 subjects (91 %), the second interviewer agreed
with the initial interviewer that Criterion A was met. Both interviewers agreed that none of the other 4 subjects could be
diagnosed with MDE because they all failed to meet Criterion
B. The disagreement found on Criterion A suggests that the
sample in this study may include some subjects who do not
meet Criterion A but who are most likely to be included in the
non-MDE group.
The two interviewers agreed on the diagnosis (MDE vs. nonMDE) in 38 of 43 cases (88%), K = .74, p < .001. This exceeds the
level (.70) that Spitzer (American Psychiatric Association, 1980)
described as "high" agreement. It should be noted that agreement on Criterion A and hence on overall diagnosis may be
spuriously high because the second interviewer knew that he
would not have been asked to participate in, or to conduct, an
interview unless the initial interviewer believed that the subject
met Criterion A.
Estimates of reliability were computed for each of the eight
symptoms of Criterion B and are reported in Table 2. The median percentage of agreement was 89% and the median K was
.732. Except for loss of interest or pleasure, which fell in the
"poor" range (<.40), all symptoms fell in the "good" (.60-74) or
the "excellent" (>.74) range of interrater reliability of specific
items using the terminology of Cicchetti and Sparrow (1981).
Heterogeneity of Cases
Table 3 shows the frequency distribution of number of symptoms per case. Overall, 65% of the subjects had four or more of
the symptoms of Criterion B and thus had cases of MDE. This
suggests that Criterion A, which was used to select the sample,
is a powerful screening criterion for the diagnosis of MDE, but
this must be qualified. The rate of MDE reflects the exclusions
made in selecting the sample of consecutive admissions and
may not apply to populations other than those seen at an emergency service. Within these limitations, simply knowing that a
person has a persistent and relatively prominent depressed
mood or pervasive anhedonia would enable a clinician to diagnose MDE and to be correct in 65% of instances.
The DSM-IH permits a diagnosis of MDE with as few as four
of the symptoms listed in Criterion B, but such cases are rare, at
least in the emergency service from which subjects were drawn.
Only 4 (4%) of 107 cases had four symptoms, as compared with
19 (18%) who had all eight symptoms and 81 (76%) who had six
or more. The modal number of symptoms in the MDE group
was 6 and the mean was 6.32. These figures can be compared
with data based on clinical examination of 486 routine admissions to inpatient and outpatient psychiatric services (Beck,
1967). We calculated means of 4.04, 5.94, and 6.94 for cases of
mild, moderate, and severe depression, respectively. Beck's
cases were grouped by ratings of depth of depression without
Table 2
Reliability of Individual Symptoms
Symptom
> Agreement
Appetite disturbance
Sleep disturbance
Psychomotor disturbance
Loss of interest or pleasure
Loss of energy
Feelings of worthlessness
Thinking difficulties
Thoughts of death or suicide
83
90
81
69
90
90
88
93
.642
.738
.619
.389
.757
.808
.121
.850
201
Table 3
Number of Criterion B Symptoms per Case
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Major
depressive
disorder/
no. of symptoms
Cases
8
7
6
5
4
Noncases
3
2
1
0
MDE
Frequency
19
27
35
22
4
18
25
33
21
4
28
17
9
3
49
30
16
5
in less than half of the cases (Clarkin et al., 1983; Widiger et al.,
1986).
Although all of the symptoms were common among cases of
MDE, sleep disturbance (98%), loss of energy (93%), and appetite disturbance (86%) were especially common, and the three
occurred together in almost 80% of cases. At the other extreme,
psychomotor disturbance and feelings of worthlessness were
the least frequent symptoms, each occurring in less than 70% of
the cases. A Cochran Q test (Myers, DiCecco, White, & Borden,
1982) applied to the frequency data showed a significant variation in frequencies, Q(l) = 67.77, p < .00001. (Using the ultraconservative Greenhouse-Geisser strategy [see Myers et al.,
1982] to guard against heterogeneity of covariances yielded p <
.01.) Follow-up Tukey tests (Levy, 1979) of the frequencies of
pairs of symptoms showed that sleep disturbance was more
frequent (p < .05) than motor disturbance, feelings of worthlessness, thoughts of death or suicide, loss of interest or pleasure, and thinking difficulties. Loss of energy was significantly
more frequent than all of these save the last, and appetite disturbance was more frequent than motor disturbance and feelings of worthlessness.
The present results differ in several respects from those presented by Beck (1967) for cases with moderate depression and
cases with severe depression. In both groups, loss of interest
ranked second and feeling inadequate third in frequency, and
both were more frequent than their counterparts in the present
study.4 Sleep disturbance and loss of appetite ranked relatively
low in both groups, and fatigability ranked sixth in the severe
group but first in the moderate group. Thus, in general, the
so-called "vegetative" symptoms were less common in Beck's
cases than in the current study. Interestingly, motor retardation,
as manifested by a reduction in spontaneous activity in the
severe group (87%) and by slow speech, reduced verbal output,
and so forth in the moderate group (72%), was more frequent
than the corresponding item, psychomotor agitation or retardation, in the current study, although it ranked sixth in both
groups. It seems likely that the frequency of loss of pleasure was
underestimated in the present study. This symptom was identified with poor reliability, and, in the reliability sample, the first
interviewers marked its presence in only 50% of the cases as
Symptom
Cases
(n = 107)
Noncases
(n = 57)
K"
Appetite disturbance
Sleep disturbance
Motor disturbance
Loss of interest or pleasure '
Loss of energy
Feelings of worthlessness
Thinking difficulties
Thoughts of death or suicide
86
98
62
75
93
68
78
74
32
47
11
19
28
25
18
44
.55
.57
.45
.52
.67
.40
.57
.30
compared with the second interviewer's 6 8%. The other discrepancies between the present results and Beck's may stem from
differences between the symptoms used by Beck's examiners to
decide that patients were moderately or severely depressed and
the criteria for MDE given in the DSM-III.
The frequency data are also somewhat inconsistent with clinical descriptions. For example, Hamilton (1982) asserted that
depressed mood, feelings of guilt, and suicidal thoughts are all
present in severe cases but that depressed mood, loss of interest, and anxiety (not assessed here) are the most common symptoms, followed by difficulty in falling asleep, loss of appetite,
loss of energy, and suicidal thoughts. Neither guilt (included in
feelings of worthlessness, self-reproach, or excessive or inappropriate guilt) nor loss of interest were found as frequently as
Hamilton's remarks suggest (however, see the earlier comments
on the frequency of loss of interest). It is somewhat difficult to
compare the present results with Kraepelin's (1921) classical
description of depression because he described symptoms separately for several variant depressive states. In general, however,
his descriptions suggest that thinking difficulties and motor
disturbance (retardation or agitation) should be more common
than they were in the present study.
Data on the frequency of individual symptoms for the nonMDE group are also shown in Table 4. In contrast to cases of
MDE, none of the symptoms occurred in as many as half of the
noncases. Only two symptoms, sleep disturbance (47%) and
Nelson, Mazure, Quinlan, and Jatlow (1984) and Miller and Nelson
(1987) tabulated frequencies separately for the three alternatives of the
Criterion B item (loss of pleasure, loss of interest, and decreased sex
drive) but found that one of these occurred both relatively and absolutely more often than in the present study. In contrast, they reported
feelings of worthlessness occurring with approximately the same frequency as in the present study, although they tabulated guilt separately.
Because of the way they fragmented the items in Criterion B, further
meaningful comparison with the present data is impossible.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
202
noses, Hj- must exceed the base rate of the disorder or of its
absence, whichever is larger (Meehl & Rosen, 1955). In the present study, predicting MDE in all instances would result in 65%
of diagnoses being correct. Table 6 shows that all symptoms
have values of Hp greater than .65; however, in some, or all,
instances the difference between .65 and H- may be due to
sampling error. To evaluate this possibility, a series of chisquare tests was carried out in which the observed frequencies
of correct and incorrect diagnosis based on single symptoms
were tested against expected frequencies based on the base
rates. For five symptoms, the null hypothesis was rejected at
.005 (two-tailed) or better, X 2 (l, N = 164) = 9.71 to 29.29; feelings of worthlessness, thoughts of death or suicide, and psychomotor disturbance were not significant at p < .05.
As Table 6 shows, values of H,- ranged from .85 for the most
efficient symptom, loss of energy, to .68 for the least efficient
symptom, thoughts of death or suicide. A Cochran Q test calculated to analyze the significance of variations in Hp yielded
(2(7) = 23.97, p = .01, using the procedure recommended by
Myers et al. (1982) to adjust the critical value for violation of
assumptions. Follow-up Tukey tests showed that loss of energy
was significantly more efficient than thoughts of death or suicide (p < .01), feelings of worthlessness (p < .05), and psychomotor disturbance (p = .05). Sleep disturbance was significantly more efficient than thoughts of death or suicide (p <
.05), but no other differences were significant. It should be
noted that differing base rates may produce different absolute
and relative values of Hp, so the results cannot be generalized to
populations with different base rates.
Identification of cases of MDE. The proportion of subjects
with a specified symptom who have cases of MDE is given by
Hp. Values of Hp are shown in Table 6. All are greater than the
base rate for MDE in the present sample. It can be shown algebraically that the value of Hp for a given symptom will be
greater than the base rate of the disorder whenever the symptom is more frequent among cases of the disorder than among
noncases. Thus, the evidence for descriptive validity cited earlier also indicates that each symptom can be used to identify
cases of MDE at a rate above random selection.
The values shown in Table 6 suggest that psychomotor disturbance is the best single symptom to use in identifying cases
of MDE (Hp = .92) and thoughts of death or suicide (Hp = .76) the
worst; however, the range of values is small, indicating that the
symptoms do not vary much in their ability to select cases of
MDE. Unfortunately, there is no satisfactory way to test
whether differences in Hp values are due to sampling variation.
The Cochran test cannot be used because it requires that each
subject be observed under all conditions, but Hp for a given
symptom is based only on subjects who had that symptom.
Tests for differences among independent groups would also not
be appropriate.
Identification of noncases. Values of H also appear in Table
6. This index is the proportion of subjects without a given
symptom who do not have cases of MDE. In a random sample
from the present population, we would expect 35% to be noncases. Again, the evidence for descriptive validity allows the
* This use of kappa was suggested by an anonymous reviewer.
203
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Symptom
1.
2.
3.
4.
5.
6.
7.
8.
Appetite disturbance
92/84
Sleep disturbance
97/54
Motor disturbance
100/66
Loss of interest or pleasure
95/82
Loss of energy
91/55
Feelings of worthlessness
99/64
Thinking difficulties
Thoughts of death or suicide 91/63
94/61
95/72
98/43
92/91 100/59
100/68
98/40
94/78 100/50
87/74 98/46
96/72
96/48
95/56
97/54
97/62
96/74
95/67
97/55
88/55 94/58
'b Conditional probability of MDE given the presence of the two symptoms.
Percentage of cases of MDE with the two symptoms.
conclusion that the absence of any given symptom can be used
to select noncases of MDE at a rate above random selection.
The best symptoms for identifying noncases of MDE are
(absence of) sleep disturbance (94%), loss of energy (84%), and
appetite disturbance (72%). The poorest symptoms are (absence
of) thoughts of death or suicide (53%), psychomotor disturbance (55%), and feelings of worthlessness (56%). It is noteworthy that sleep disturbance is one of the poorest indicators of
MDE and absence of psychomotor disturbance one of the
poorest indicators of non-MDE.
Although the numerical values of Hp and HN depend on the
base rates, the rank order of the symptoms in terms of either
one of these indices does not. The rank of a symptom based on
how well it can select cases (Hp) depends only on the ratio of the
frequency of the symptom among cases of MDE (true positives)
to the frequency of the symptom among noncases (false positives). Therefore, the values of Hp shown in Table 6 tend to be
inversely related to the frequency of each symptom among noncases. Similarly, rank based on how well a symptom can select
noncases (H,) depends only on the ratio of the frequency of
Table 6
Diagnostic Efficiency of Single Criterion B Symptoms for
Identifying Presence (Hp) and Absence (H^ of Major Depressive
Episode and Total Correct Diagnoses (Hj-)
Symptom
Appetite disturbance*
Sleep disturbance*
Psychomotor disturbance
Loss of interest or pleasure*
Loss of energy*
Feelings of worthlessness
Thinking difficulties*
Thoughts of death or suicide
HN
.84
.80
.92
.88
.86
.84
.89
.76
.72
.94
.55
.63
.84
.56
.66
.53
.80
.82
.71
.77
.85
.71
.79
.68
" For HT, the number of correct diagnoses based on each symptom was
tested against the number of correct diagnoses (107) that would be
made if major depressive episodes were diagnosed in every case using
the chi-square test (two-tailed). The values of HT varied significantly,
among symptoms, at .01 using a Cochran Q test.
* p < .005.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
204
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
References
American Psychiatric Association. (1980). Diagnostic and statistical
manual of mental disorders (3rd ed.). Washington, DC: Author.
American Psychiatric Association. (1987). Diagnostic and statistical
manual of mental disorders (Rev. 3rd ed.). Washington, DC: Author.
Beck, A. T. (1967). Depression: Clinical experimental and theoretical
aspects. New York: Paul B. Hoeber.
Cicchetti, D. V, & Sparrow, S. S. (1981). Developing criteria for establishing the interrater reliability of specific items in a given inventory:
Applications to the assessment of adaptive behaviors. American
Journal of Mental Deficiency, 86, 127-137.
Clark, L. A. (1989). The anxiety and depressive disorders: Descriptive
205