Development and validation of a brief

observer-rated screening scale for

depression in elderly medical patients
MARGARET F. HAMMOND, SHAUN T. O'KEEFFE
1
, DAVID H. BARER
2
Geriatric Medicine, University Clinical Department, The Duncan Building, Daulby Street, Liverpool L69 3GA, UK
1
Department of Geriatric Medicine, St Michael's Hospital, Dun Laoghaire, County Dublin, Republic of Ireland
2
Department of Clinical Geriatric Medicine, Queen Elizabeth Hospital, Sheriff Hill, Gateshead,
County Durham NE9 65X, UK
Address correspondence to: M. F. Hammond. Fax: (q44) 151 706 4064. Email: mhammond@liverpool.ac.uk
Abstract
Objective: to develop a depression screening scale that does not rely on verbal communication.
Setting: an acute geriatric unit in a teaching hospital.
Subjects: 96 patients (mean age 81 years, range 6892, 59 women); 40% of the initial study group of 50 and 22%
of the validation group of 46 were diagnosed as depressed.
Methods: we devised a scale using nine items which could be rated by an observer; we determined inter-rater
reliability, sensitivity, specicity and predictive values for each item compared with a Geriatric Mental State-AGECAT
diagnosis of depression; we validated a nal scale of six items.
Results: inter-rater reliability was poor for two items (irritability and sleep disturbance) while two items (sleep
disturbance and night sedation) had poor sensitivity; we omitted these items in a revised scale. Re-analysis of data from
the initial study showed that a cut-off of 03 on the revised scale gave a sensitivity of 83%, a specicity of 95%, a
positive predictive value of 0.89 and a negative predictive value of 0.90. Spearman's correlation coefcient between the
six-item questionnaire and the Hamilton rating scale was 0.79. In the validation study, the cut-off score of 03 on the
revised six-item scale had a sensitivity of 90%, specicity of 72%, a positive predictive value of 0.69 and a negative
predictive value of 0.96.
Conclusions: this simple, short, observation-based screening scale completed by nurses is sensitive and specic in
identifying depression in elderly medically ill patients, and may be a useful addition to clinical practice.
Keywords: depression, older people, screening scales
Introduction
Studies of elderly medical inpatient populations have
found prevalence rates for depression of up to 50%
[15]. Depression in this population has been shown to
be associated with increased mortality and duration of
hospital stay [6, 7]. Nevertheless, it is often unrecognized
[4, 8].
Satisfactory sensitivity and specicity have been
reported for self-rated `paper and pencil' screening
tests for depression in medical inpatients [3, 9, 10]. As
a result, it has been suggested that these tests should
be used in routine clinical practice. The Geriatric
Depression Scale [11] is recommended by the Royal
College of Physicians and the British Geriatrics Society
for screening for depression in elderly inpatients [12].
However, although usually acceptable to patients, these
tests are rarely used by physicians on busy hospital wards
[10]. Anecdotal evidence suggests the low use of
depression screening scales may be due to the reluctance
of clinical staff to ask `difcult' or `sensitive' questions,
the time taken to complete scales and the inappropriate-
ness of `paper and pencil' instruments in clinical practice.
In addition, although depression is strongly asso-
ciated with cognitive impairment and communication
disorders in geriatric patients [3] and in stroke patients
[13, 14] (particularly those with non-uent aphasia [15]),
assessment methods which rely on patient replies are not
suitable for use in sick patients with communication
disturbances [16, 17]. Patients with dysphasia, dysarthria,
organic brain syndromes or visual or hearing impair-
ment, or who are illiterate constitute up to one-third of
Age and Ageing 2000; 29: 511515
#
2000, British Geriatrics Society
511
the population of elderly medical patients and have been
excluded from systematic studies of depression because
of the problems in rating [3, 8, 18].
An alternative approach to detecting depression,
which has not yet been examined in detail, is to use the
observational skills of nurses who are in daily contact
with patients. In this study, we sought to develop and
validate a brief observer-based screening test for
depression.
Patients and methods
Development of scale
We constructed an initial scale using observable
behaviour from the Diagnostic and Statistical Manual,
third edition, revised, criteria for depression [19] and
from symptoms known to discriminate depression in
physically ill people [2022]. We examined nursing and
medical notes for phrases and wording used by staff
when describing behaviour of patients with depression.
We asked nurses to comment on the behaviour of named
patients identied as depressed. Items considered
included crying, restlessness, agitation, irritability, un-
cooperativeness, hostility, aggressiveness, disinterest in
food, social withdrawal, lack of reactivity, difculty
getting to sleep, disturbed sleep, early waking, diurnal
mood change and weight loss.
Early evaluation suggested that questions on appetite,
food intake or weight loss would be of little value in
screening for depression in our inpatient population
because of their poor specicity. We decided in advance
that the scale should have a yes/no format to avoid the
ambiguity inherent in multiple-response scales.
The initial nine-item questionnaire is shown in
Table 1.
Patients
We planned sample sizes of 50 for both the initial and
the validation studies. We considered consecutive
patients admitted to the acute geriatric wards for
inclusion. Patients with communication problems (e.g.
deafness, aphasia), severe cognitive impairment (dened
as an Abbreviated Mental Test [23] score (5/10)
or delirium (according to the Confusion Assessment
Criteria) [24] and patients taking anti-depressant medica-
tions were excluded, as were patients who were unwilling
or were considered too ill to participate.
Methods
In the initial study, the screening scale was completed
separately by two nurses familiar with the patient, one
of whom was always the primary nurse responsible
for the patient's care. Patients were then interviewed,
using the Geriatric Mental State Schedule (GMS) [25], by
an experienced research psychologist who was unaware
of the nurses' assessments, and a diagnosis was made
using the GMSAGECAT diagnostic syndrome case
level criteria. The Hamilton Depression Rating Scale
[26] was completed by this interviewer for the rst 24
patients assessed.
We used the k statistic to determine inter-rater
reliability between the two nurses for each of the indi-
vidual items on the nine-item scale; k has a value of 1.0
when agreement is perfect, 1.0 when disagreement is
absolute, and a value of 0 when agreement is no better
than chance. Guidelines for interpreting values between
0 and 1.0 are: -0.20=poor agreement; 0.210.40=fair
agreement; 0.410.60=moderate agreement; 0.610.80=
good agreement; and 0.811.0=excellent agreement [27].
We calculated the sensitivity, specicity and positive and
negative predictive values of individual items using
standard formulae from the ratings of the primary nurse
only. We omitted items with poor inter-rater reliability or
with sensitivity or specicity of -50% from the revised
scale. We selected optimal cut-off points to maximize the
sum of sensitivity and specicity.
The revised scale was tested in a further 46
consecutive patients admitted to the acute geriatric
wards. The same procedure was followed, except that the
screening scales were completed only by the patients'
primary nurses.
Results
Details of patients considered for the study and reasons
for exclusions are shown in Table 2. The scores on the
scale did not differ between the 29 patients for whom
a subsequent psychological assessment was not per-
formed for logistical reasons and the remaining 96
patients. The mean (SD) age of the 96 patients with
complete assessments was 81 (6) years; there were 59
women and 37 men. Primary diagnoses in these patients
included cardiac disease (25), respiratory disease (19),
gastrointestinal disease (16), cancer (10), cerebrovascular
disease (eight) and falls (six).
Table 1. Original questionnaire (items retained in the
nal questionnaire are shown in bold print)
1. Does the patient sometimes look sad, miserable or
depressed?
2. Does the patient ever cry or seem weepy?
3. Does the patient seem agitated, restless or anxious?
4. Is the patient lethargic or reluctant to mobilize?
5. Is the patient demanding or irritable?
6. Does the patient need a lot of encouragement to do
things for him/herself ?
7. Does the patient seem withdrawn, showing little
interest in the surroundings?
8. Is the patient having problems sleeping?
9. Is the patient taking night sedation?
M. F. Hammond et al.
512
Twenty patients (40%) in the initial study group were
diagnosed as being depressed (median age 80, range
6892; 14 women). Inter-rater agreement, sensitivity,
specicity and predictive values for the nine items
included in the initial scale are shown in Table 3. A cut-
off of 04 on this questionnaire gave a sensitivity of
90%, a specicity of 77%, a positive predictive value of
0.72 and a negative predictive value of 0.92. Spearman's
correlation coefcient between the nine-item question-
naire and the Hamilton Depression Rating Scale was
0.76.
Inter-rater reliability was poor for item 5 (on
irritability) and item 8 (on sleep disturbance), while
item 8 and item 9 (night sedation) had poor sensitivity
for depression. Accordingly, we omitted these three
questions in a revised six-item scale. A re-analysis of data
from the initial study showed that a cut-off of 03 on the
revised scale gave a sensitivity of 83%, a specicity
of 93%, a positive predictive value of 0.89 and a negative
predictive value of 0.90. Spearman's correlation coef-
cient between the six-item questionnaire and the
Hamilton Depression Rating Scale was 0.79.
Ten (22%) of the 46 patients in the validation group
were depressed (median age 78, range 6790 years; six
women). In this group, the cut-off of 03 on the six-item
scale gave a sensitivity of 90%, a specicity of 72%,
a positive predictive value of 0.69 and a negative
predictive value of 0.96.
Discussion
Although self-rated screening scales have proved
effective in identifying depression in medically ill
patients, our experience and that of others is that they
are rarely used by physicians [4, 10]. Observations by
nurses, who are most intimately involved in day-to-day
patient care, might be a useful approach [4]. Indeed, in
units that do not use formal screening tests, recognition
of depression is usually triggered by staff observations.
In most cases, an initial suspicion of depression can be
further investigated by questioning the patient. However,
in patients with stroke or major communication or
cognitive disturbance due to another cause, antidepres-
sant medications are frequently started solely on the
basis of the observational impression of staff.
Studies that have examined nurses' identication of
depression have had generally disappointing results.
When patients' key nurses were asked to make an
assessment of depression along a four-point scale from
`denitely not depressed' to `denitely depressed' in a
study of 59 elderly medical inpatients, their ratings of
`denite or probable' depression had a sensitivity of only
38% and a specicity of 79% relative to AGECAT
diagnoses [4]. Similarly, research nurses who visited
stroke patients at home had a sensitivity of 55% and a
specicity of 93% for the detection of depression [17].
In oncology patients, the concordance between nurses'
ratings of depression severity and patients' own
estimates were little better than chance (k=0.17) [28].
Poorly developed observational skills are not neces-
sarily the reason for the unsatisfactory levels of
recognition of depression in these previous studies.
Staff may lack condence in making a diagnosis of
depression, or symptoms consistent with depression may
be discounted as normal effects of illness or of
hospitalization [10]. In at least some cases of depression
not recognized by nurses, the signs and symptoms of
Table 2. Patients considered for inclusion, numbers and
reasons for exclusions
Study
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Initial Validation
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Patients considered 107 98
Exclusions
Cognitive impairment 15 13
Antidepressants 12 11
Critical illness 5 6
Aphasia 4 3
Deafness 4 2
Refusal 2 3
Incomplete assessments 15 14
Patients studied 50 46
Table 3. k values, sensitivity, specicity and positive and negative predictive values of
individual questions for the detection of depression
Predictive value
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Question k (% agreement) Sensitivity (%) Specicity (%) Positive Negative
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 0.77 (86%) 85 80 0.74 0.89
2 0.51 (77%) 60 87 0.75 0.76
3 0.66 (83%) 70 73 0.64 0.79
4 0.57 (80%) 80 60 0.59 0.86
5 0.15 (66%) 50 77 0.59 0.70
6 0.37 (69%) 70 73 0.64 0.79
7 0.55 (77%) 80 83 0.76 0.86
8 0.13 (66%) 40 73 0.50 0.65
9 0.77 (86%) 15 60 0.20 0.51
An observer-rated screening scale for depression
513
depression had indeed been identied but had been
considered by nurses to be understandable in view of
patients' social problems [1]. Similarly, Jackson and
Baldwin noted that nurses seemed able to recognize the
symptoms, if not the syndrome, of depression [4].
One way of improving nurses' recognition of
depression would be to provide explicit training
regarding the features of depression, although in primary
care, education alone has been found to be insufcient in
effecting improvements in treatment [29]. An alternative
approach, which the results of the present study suggest
might be applicable in clinical practice, would be to ask
nursing staff for their observations of specic behav-
iours or symptoms, rather than seeking a diagnosis. Our
study indicates that a simple scale completed by primary
nurses is sensitive and specic in identifying depression
in elderly medically ill patients. Inter-rater reliability for
the individual items was reasonable. The revised scale
takes less than 1 min to complete.
Nurses are able successfully to incorporate screening
for depression, among other routine assessments, into
their usual practice [30]. However, although the
systematic use of reliable screening tools is an efcient
method of enhancing recognition, identication of
depressive symptoms should be supported by clinical
management protocols. These should specify guidelines
for identication by screening, which is followed by
further assessment of important symptoms, treatment
and referral options, and follow-up procedures. Although
the evidence-base on efcacy of treatment with anti-
depressants in this population is meagre [31, 32], the
implementation of algorithms which systematically dene
structured management and intervention steps may
remove the ambiguity in the treatment for depression of
physically ill elderly patients and signicantly improve
response [33, 34].
The prevalence of depression in our study population
was high, in accordance with previous reports. The
diagnosis was made using a standardized and well-
validated diagnostic instrument by an experienced
examiner. However, there are some important limitations
to this study. Our unit had recently conducted studies
into the treatment of depression: nursing awareness of
the importance of individual features of depression may,
therefore, have been greater than in other units. We
acknowledge that there are patients who admit to
depression on questioning but do not necessarily exhibit
immediately obvious behavioural clues. Conversely,
studies of stroke patients suggest that some patients
with some features of depression (e.g. crying or apathy)
deny low mood [13, 14].
Although a purely observer-rated scale would be of
most value in detecting depression in patients with
communication problems due to stroke, deafness or
major cognitive impairment, such patients were, of
necessity, omitted from the present study. Our scale does
require validation in these populations. However, the
behavioural and motivational manifestations of depres-
sion [2022] and a depressed appearance [14] have been
noted by others to reliably indicate depression in
physically ill patients, suggesting that the scale would
be useful in these patients. There are also depressed
patients who, on rst impression, may not appear
depressed but may manifest behavioural cues that, with
closer acquaintance, can be identied by the primary
nurse.
The sensitivity and specicity of this scale compare
very favourably with the Geriatric Depression Scale [5].
An observation-based screening scale that can be quickly
completed as part of nursing care may be a valuable
contribution to the recognition of depression in elderly
hospital inpatients.
Key points
.
Nurses reliably recognize signs of depression in their
elderly patients.
.
The six-item observation-based scale shows good
sensitivity and specificity for depression in geriatric
inpatients.
References
1. Koenig HG, Meador KG, Cohen HJ et al. Depression in
elderly hospitalized patients with medical illness. Arch Intern
Med 1988; 148: 192936.
2. O'Riordan TG, Hayes JP, Shelley R et al. The prevalence of
depression in an acute geriatric medical assessment unit. Int J
Geriatr Psychiatry 1989; 4: 1721.
3. Ramsay R, Wright P, Katz A et al. The detection of
psychiatric morbidity and its effect on outcome in acute elderly
medical admissions. Int J Geriatr Psychiatry 1991; 6: 8616.
4. Jackson R, Baldwin B. Detecting depression in elderly
medically ill patients. Age Ageing 1993; 22: 34953.
5. Shah A, Herbert R, Lewis S et al. Screening for depression
among acutely ill geriatric inpatients with a short geriatric
depression scale. Age Ageing 1997; 26: 21721.
6. Katona CLE. Depression in Old Age. Chichester: John
Wiley, 1994.
7. Murphy E. Increased mortality rates in late-life depression.
Br J Psychiatry 1988; 152: 34753.
8. Rapp S, Parisi S, Walsh D et al. Detecting depression in
elderly medical in-patients. J Cons Clin Psychol 1988; 56:
50913.
9. Shah A, Phongsathorn V, Bielawska C et al. Screening for
depression among geriatric inpatients with short versions of
the Geriatric Depression Scale. Int J Geriatr Psychiatry 1996;
11: 91518.
10. Meakin CJ. Screening for depression in the medically ill.
The future of paper and pencil tests. Br J Psychiatry 1992; 100:
21216.
M. F. Hammond et al.
514
11. Yesavage JA, Brink TL, Rose TL et al. Development and
validation of a geriatric depression scale: a preliminary report.
J Psychiatr Res 1983; 17: 3749.
12. Royal College of Physicians. Standardised Assessment
Scales for Elderly People. Report of joint workshops of the
Research Unit of the Royal College of Physicians and the
British Geriatrics Society. London: Royal College of Physicians,
1992.
13. Ramasubbu R, Kennedy SH. Factors complicating the
diagnosis of depression in cerebrovascular disease. Can J
Psychiatry 1994; 39: 596607.
14. Kishi Y, Robinson RG, Kosier JT. The validity of observed
depression as a criteria for mood disorders in patients with
acute stroke. J Affect Disord 1996; 40: 5360.
15. Robinson RG, Benson DF. Depression in aphasic stroke
patients: frequency, severity and clinical-pathological correla-
tions. Brain Lang 1981; 14: 28291.
16. Hermann M, Bartels C, Wallesch C-W. Depression in acute
and chronic aphasia: symptoms, pathoanatomicalclinical
correlation and functional implication. J Neurol Neurosurg
Psychiatry 1993; 56: 6728.
17. House A, Dennis M, Hawton K et al. Methods of
identifying mood disorders in stroke patients: experience in
the Oxfordshire Community Stroke Project. Age Ageing 1989;
18: 371379.
18. Allman P. Depressive disorders and emotionalism follow-
ing stroke. Int J Geriatr Psychiatry 1991; 6: 37783.
19. American Psychiatric Association. Diagnostic and
Statistical Manual of Mental Disorders, 3rd edition, revised.
Washington DC: American Psychiatric Association, 1987.
20. Rapp SR, Vrana S. Substituting nonsomatic for somatic
symptoms in the diagnosis of depression in elderly male
medical patients. Am J Psychiatry 1989; 146: 1197200.
21. Okimoto JT, Barnes RF, Veith RC et al. Screening for
depression in geriatric medical patients. Am J Psychiatry 1982;
139: 799802.
22. Clark DC, Cavanagh SV, Gibbons RD. The core
symptoms of depression in medical and psychiatric patients.
J Nerv Ment Disord 1983; 171: 70513.
23. Hodkinson M. Evaluation of a mental test score for
assessment of mental impairment in the elderly. Age Ageing
1972; 1: 2338.
24. Inouye SK, van Dyck CH, Alessi CA et al. Clarifying
confusion: the confusion assessment method. A new method
for detection of delirium. Ann Intern Med 1990; 113: 9418.
25. Copeland JMR, Dewey ME, Grifths-Jones HM.
Computerized psychiatric diagnosis and case nomenclature
for elderly subjects: GMS and AGECAT. Psychol Med 1986;
16: 8999.
26. Hamilton M. Development of a rating scale for primary
depressive illness. Br J Soc Clin Psychol 1967; 6: 27896.
27. Altman DG. Practical Statistics for Medical Research.
London: Chapman and Hall, 1991.
28. McDonald MV, Pakkis SD, Dugan W et al. Nurses'
recognition of depression in their patients with cancer.
Oncol Nurs Forum 1999; 26: 5939.
29. Thompson C, Kinmonth AL, Stevens L et al. Effects of a
clinical-practice guideline and practice-based education on
detection and outcome of depression in primary care:
Hampshire Depression Project randomised controlled trial.
Lancet 2000; 355: 18591.
30. Philp I, Goddard A, Connell et al. Development and
evaluation of an information system for quality assurance. Age
Ageing 1994; 23: 1503.
31. Evans M, Hammond M, Wilson K et al. Placebo-
controlled treatment trial of depression in elderly physically
ill patients. Int J Geriatr Psychiatry 1997; 12: 81724.
32. Evans M, Hammond M, Wilson K et al. Treatment of
depression in the elderly: effect of physical illness on response.
Int J Geriatr Psychiatry 1997; 12: 118994.
33. Simon GE, Von Korff M, Rutter C et al. Randomised trial
of monitoring, feedback, and management of care by telephone
to improve treatment of depression in primary care. Br Med J
2000; 320: 5504.
34. Nakano T, Kugaya A, Akechi T et al. Algorithm for the
treatment of major depression in patients with advanced
cancer. Psychiatry Clin Neurosci 1999; 53 (suppl.): S615.
Received 2 December 1999; accepted in revised form 8 June
2000
An observer-rated screening scale for depression
515