Abstract
The lifetime prevalence of major depressive disorder in the general population is approximately 17% [1], and is even greater among the medically ill [1, [2], [3], [4]]. Depression is associated with considerable morbidity [5, [6]] and mortality [7], and in those with medical illness it may also exacerbate the effects of their medical disorder.
The problems of diagnosis of depression in the medically ill are well documented [8, [9], [10], [11], [12]] and the diagnosis of depression in this population is often missed [13, [14], [15]]. Difficulties with diagnosis occur for three main reasons: reluctance to make a diagnosis of disorder in a situation where mood disturbance is ‘understandable’; difficulty determining the boundary between ‘normal’ sad feelings and the syndrome of major depression; and difficulty determining which symptoms (e.g. anorexia, weight loss, fatigue) are due to physical illness and which are due to depression.
Research aiming to improve the diagnosis of depression in medically ill patients has followed two broad directions: modification of diagnostic criteria which have been developed in medically well populations, and identification of those symptoms which best discriminate between depressed and non-depressed individuals.
There have been four major approaches to the evaluation of depression in the medically ill: inclusive, aetiological, substitutive and exclusive [16]. The inclusive approach used by Rifkin and colleagues [17] is simplest. All symptoms are counted in making a diagnosis of depression whether or not they may be attributable to a primary physical problem. This approach has high inter-rater reliability but is associated with a risk of a high false positive rate.
The aetiological approach, advocated by Spitzer and the developers of the third edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-III) [18], and subsequently DSM-III-R [19] and DSM-IV[20], requires the interviewer to attribute causality to somatic symptoms of depression: thus, sleep disturbance, appetite disturbance, fatigue, psychomotor agitation or retardation and changes in concentration may only be counted towards a diagnosis of depression if not ‘clearly due to a physical illness’. This is the decision rule in the Structured Clinical Interview for DSM-III (SCID) [21] as well as the Diagnostic Interview Schedule [22]. This approach may lead to findings of low rates of depression and a high false negative rate. A further criticism of this approach is that although psychiatrists working in specialised liaison settings may be able to reliably differentiate symptoms due to medical illness from those due to depression, this may not be so for the general psychiatrist or non-psychiatric physicians. For example, in a group of studies in stroke patients by Robinson and colleagues [23], this approach was used and although the authors reported high inter-rater reliability, it is uncertain whether other psychiatrists would be able to recognise and make the same decisions.
The exclusive approach, used for example by Holland's group researching depression in cancer patients [24, [25]], simply eliminates the symptoms fatigue and appetite disturbance from the diagnostic criteria. This approach is relatively ‘clean’, but it is harder for patients to meet diagnostic criteria, and so it is likely to result in high false negative rates.
A fourth approach, the substitutive approach, is to substitute other symptoms of depression when DSM symptoms are problematic. For example, Endicott [26] suggested replacing appetite disturbance with fearfulness or depressed appearance; sleep disturbance with social withdrawal or decreased talkativeness; fatigue with brooding, self-pity or pessimism; and difficulty thinking or concentrating with mood that is not reactive (i.e. cannot be cheered up). While this approach has heuristic appeal, a criticism has been that Endicott failed to indicate the source of these substitute criteria so that it is unclear whether her list would lead to over or under diagnosis of depression.
Key features of Cavanaugh (1991) criteria for major depressive disorder in the medically ill (modified DSM-III-R criteria)
As for other medically ill patients, the reported rates of depression in patients with human immmunodeficiency virus (HIV) disease vary markedly [29]. There are several possible reasons for this: the setting of the study (inpatient, outpatient, etc.); the geographical location; the HIV exposure risk category of individuals studied; and the problems identified in the diagnosis of depression in other medically ill groups. Given the morbidity [5, [6]] and mortality [7] associated with depression, it is essential that clinicians use the diagnostic approach most likely to maximise accurate detection of cases and minimise false negatives. False positive cases should also be minimised to avoid unnecessary and inappropriate treatment.
This pilot study was undertaken with the following aims: to devise a set of substitutive diagnostic criteria for use in patients with HIV disease; to compare the utility (rate of detection, false positive and false negative rates) of four different sets of diagnostic criteria for depression in patients with HIV disease; to examine the performance of four different depression rating scales in patients with HIV disease. The study was conducted over a 10-month period; patient recruitment was limited by this time frame.
Method
Setting
At the time of the study, Fairfield Hospital was a 138-bed hospital for the treatment of patients with infectious diseases in Melbourne. Forty beds were designated for the treatment of patients with HIV/acquired immune deficiency syndrome (AIDS). As previously described [30], patients were admitted to an inpatient unit for management of HIV-related medical disorders, for medical procedures requiring inpatient admission (including non-acute medical care, respite care, and palliative care) and for general surgery. In addition, a small number of patients were admitted to the medical wards for treatment of a comorbid psychiatric disorder; this reflected patient preference and the desirability of joint medical/psychiatric management. All such patients were managed on the medical wards by the consultation-liaison psychiatry service. During the study period, Fairfield Hospital was the major service for the majority of patients with HIV/AIDS in Victoria, caring for approximately 60% of HIV-positive individuals in Melbourne and providing inpatient treatment for approximately 80% of those infected.
Instruments
Diagnostic criteria sets
DSM-III-R diagnostic criteria for major depressive episode (MDE) were modified to develop substitutive, exclusive and inclusive criteria. The substitutive criteria were developed using the approach described by Endicott [26] and Cavanaugh [10]. Items identified for substitution were those noted by us in our clinical work, and by others [10, [17], [26]] to be difficult to identify as clearly due to depression rather than physical illness (i.e. appetite disturbance and/or weight loss; psychomotor agitation or retardation; loss of energy. Alternative symptoms substituted were: sense of failure; sense of punishment; and frequent crying). These three items were chosen from the seven suggested by Cavanaugh etal. [28] as we had previously found them to differentiate between HIV-positive patients with high and low depression scores measured on the Beck Depression Inventory (BDI) [31].
The exclusive criteria rules used were those recommended by Bukberg et al. [24]. We omitted the items fatigue and appetite disturbance and required patients to have five of the remaining seven symptoms to meet diagnostic criteria for MDE. For inclusive criteria, we followed Rifkin and colleagues' [17] approach and counted all symptoms towards a diagnosis of depression.
Symptom rating scales
Four rating scales were chosen for the study: two self-rated scales, the Beck Depression Inventory (BDI) [32] and the Centre for Epidemiological Studies Depression Scale (CES-D) [33]; and two clinician-rated scales, the Hamilton Depression Rating Scale (HDRS) [34], and the Montgomery Asberg Depression Rating Scale (MADRS) [35]. The BDI contains seven somatic and 14 cognitive/affective items. It has been used frequently in studies of depression in the medically ill and its use in patients with HIV disease has previously been reported by our group [29, [30]]. The CES-D has been used in several studies of patients with HIV [36, [37]]. CES-D scale items are mostly cognitive/affective symptoms of depression. The two clinician-rated scales both contain a mix of somatic and cognitive/affective items. The HDRS is more weighted to somatic items than the MADRS and so has been suggested by some researchers [38] to be less suitable for use in medically ill patients.
Subject
Patients were included in the study if they were: male; currently an inpatient in the HIV/AIDS unit; referred to consultation-liaison psychiatry service for assessment and/or management of ‘depression’; and gave written informed consent. Patients were not included if they were: non-English speaking; unable to complete questionnaires due to physical disability; or if they had dementia.
Assessment
All patients and controls were assessed by standard clinical interview to obtain details of developmental history, current and past psychiatric history. Assessment included completion of symptom checklists for both presence and absence of affective symptoms and their possible causation (yes/no) by physical illness. Details of medical history were available from case notes. Formal cognitive assessment was undertaken to identify any patients with dementia. Level of physical disability was assessed using the Karnofsky Scale [39] which is an observerrated scale giving a single percentage rating from 0% (moribund) to 100% (fully able).
Diagnosis of psychiatric disorder was made according to DSM-III-R criteria (i.e. aetiological criteria). Patients who were ‘depressed’, according to DSM-III-R criteria, were also rated using our inclusive, substitutive and exclusive criteria and severity of depression was assessed using the HDRS, MADRS, BDI and CES-D.
Controls matched according to age (++ 5 years), Centre for Disease Control (CDC) stage of illness, and CD4 cell count (++ 50) were recruited from two sources: patients referred to the consultation-liaison service in whom a diagnosis of depression was not made; and inpatients in the HIV/AIDS unit not referred to the consultation-liaison service. Consecutive patients from these two sources were approached to obtain the 17 required to match the depressed patients.
Statistical analysis
Clinical, diagnostic and rating scale data were collated, and using the Statistical Package for Social Sciences, (SPSS) for Windows [40], data were averaged and standard deviations calculated for the relevant data. Differences for parametric scores were assessed by analysis of variance, and for non-parametric scores were assessed by Chi-squared tests (non-continuous variables) or with the Kruskal-Wallis method.
Methods
During the 10-month period of the study, 24 depressed patients met inclusion criteria for the study. Three refused to participate and four were excluded from the study (severe dementia, n=2; cognitive impairment due to cerebral toxoplasmosis, n=2). Seventeen controls were assessed during this time. None of the consecutive control patients approached refused assessment.
The 17 depressed patients were all male, had a mean ± SD age of 40.6 ± 10.3 years, and one-third (n=7) had CDC IV disease (AIDS). Mean ± SD CD4 count was 149 ± 180.3 and mean ± SD Karnofsky score was 68.2 ± 13.8. There were no statistically significant differences between patients and controls on any of these variables.
Comparison of different methods of diagnosis
Using the aetiological approach (DSM-III-R criteria), 17 patients were diagnosed with major depression. All of the 17 patients meeting the criteria for major depression using the aetiological approach also met the substitutive and inclusive criteria, but only 15 met the exclusive criteria. The two patients who met the DSM-III-R criteria and not the exclusive criteria were patients in whom exclusion of the physical symptoms of weight loss and fatigue then left insufficient symptoms to meet the criteria for diagnosis (= five of the remaining seven symptoms). Both patients had six of the original nine DSM-III-R symptoms but when weight loss and fatigue were omitted, could not meet the exclusion criteria.
Of the 17 controls (i.e. patients who did not meet the DSM-III-R criteria for major depression), two met the substitutive criteria for depression, five met inclusive criteria, and one met the exclusive criteria for depression. These controls can be considered ‘false positive’ cases when using DSM-III-R criteria as the ‘gold’ standard.
Control 1 was diagnosed as depressed by the substitutive, inclusive and exclusive approaches but not by DSM-III-R. This patient had advanced HIV disease (CDC category IV, CD4 count of 35, and Karnofsky score of 60) with a combination of marked depressive and physical symptoms. Although he met eight of the nine criteria for depression in DSM-III-R, a number of symptoms were ‘excluded’ as weight loss, insomnia, loss of energy, apathy, decreased interest and poor concentration could all be explained by the severity of his HIV-related illness. Using the exclusive approach, only two symptoms (weight loss and fatigue) were excluded so that the patient met the six of the seven remaining criteria. Using the inclusive and substitutive criteria, this patient was also diagnosed as depressed.
Control 2 met the substitutive and inclusive criteria for depression, but not the DSM-III-R criteria. This patient had a markedly depressed affect and feelings of hopelessness, sense of failure and tearfulness. This patient had multiple severe symptoms related to physical illness (Karnofsky score 60), so these could not be counted towards a DSM-III-R diagnosis of depression.
Controls 3–5 all met inclusive criteria but did not meet the aetiological, substitutive or exclusive criteria. All three patients had physical symptoms that were thought to be due to medical problems rather than depression. These included symptoms of fatigue, weight loss, poor concentration, apathy and loss of interest. The three patients had Karnofsky scores of 70,75, and 60, all had intercurrent illnesses, and CD4 cell count < 40.
Rating scale scores
The 17 patients diagnosed with depression according to DSM-III-R criteria and the 17 controls were all assessed by clinical interview using HDRS and MADRS and all completed the two self-rating scales (BDI and CES-D). Mean ± SD scores for the patients and the controls were significantly different on all four scales: BDI 35.9 ± 5.2 versus 22.7 ± 10.8 (p<0.05); CES-D 39.9 ± 6.6 versus 16.0 ± 13.1 (p<0.05); MADRS 35.4 ± 5.2 versus 12.5 ± 7.8 (p <0.05); HDRS 26.2 ± 5.2 versus 11.1 ± 5.5 (p<0.05).
Analysis of rating scale items
For each of the rating scales, comparisons were made between the mean scores on each scale item between the depressed (DSM-III-R) and control groups.
Montgomery Asberg Depression Rating Scale
Significant differences (p<0.01) in mean item scores were shown between depressed patients and controls on both somatic/vegetative symptoms (appetite, sleep) and cognitive/affective symptoms (apparent and reported sadness, concentration, inability to feel, pessimistic thoughts and suicidal thoughts). For only two items (lassitude and inner tension), no significant differences were demonstrated.
Hamilton Depression Rating Scale
Significant differences (p<0.01) were found in mean item scores for the two groups for both somatic/vegetative symptoms (initial, middle and late insomnia, psychomotor retardation, gastrointestinal symptoms, somatic symptoms) and cognitive/affective symptoms (depression, guilt, suicide, work and interests). By comparison, no significant differences were shown for three items depicting anxiety symptoms (psychomotor agitation, anxiety-somatic, anxiety-psychic) with low scores in both groups. Four items, three related to physical health (loss of libido, hypochondriasis, and loss of weight) and the item regarding maintenance of insight were equally common in both groups.
Beck Depression Inventory
Comparison of mean scores of Beck Depression Inventory (BDI) items for depressed and non-depressed patients
Centre for Epidemiological Studies Depression Rating Scale
Significant differences in mean item scores (p<0.01) between the two groups were found on 15 of the 20 CES-D items (bothered by things, appetite poor, blues, trouble keeping my mind on what I was doing, depressed, everything was an effort, life a failure, fearful, sleep restless, happy, lonely, enjoy life, crying, sad, and could not get going). Items which did not show a significant difference between the depressed patients and controls were ‘felt just as good as others, hopeful, talk less, people unfriendly, and felt people dislike me’.
Discussion
This study tested four approaches to the diagnosis of depression in a sample of HIV-positive patients. Of note, some of the HIV-positive patients were admitted solely for the management of their psychiatric symptoms and were not necessarily acutely medically unwell at the time of the assessment. However, all patients were medical inpatients managed by a consultation-liaison psychiatry service.
Here, as was our general approach, we used the DSM-III-R criteria and diagnostic rules as the ‘gold standard’. Having developed experience in the treatment of patients with HIV infection, we feel confident in apportioning causation to physical versus psychological illness. Using different diagnostic criteria sets and rules, up to five additional ‘depressed’ patients were identified from the control group. However, on review, and with follow-up, we (the treating clinicians), did not believe that any of the patients with ‘false positive’ diagnoses made by the substitutive and/or inclusive and/or exclusive approach were suffering from depression. All five patients who met diagnostic criteria for depression using these three approaches, and not when using DSM-III-R criteria had physical symptoms which appeared to account for their meeting diagnostic criteria. For clinicians not familiar with the manifestations of HIV disease, use of DSM-III-R criteria may not be easy, and so other approaches should be considered.
For clinical purposes the approach which maximises sensitivity (all patients with possible depression are evaluated and treated), is the inclusive approach. For the clinician who is not experienced in assessment of HIV-positive patients this may be the most acceptable. However, the disadvantage of this approach is that a number of patients may receive unnecessary treatment. In this study, five out of 17 controls were found to have ‘depression’ according to the inclusive criteria. This is an unacceptably high number of ‘wrong’ diagnoses. In addition to inappropriate use of resources, this may produce unnecessary morbidity for patients already suffering from a number of physical problems (e.g. side effects of medication and possible drug interactions). For research purposes the exclusive approach can be considered the ‘cleanest’. This approach ensures that a more clear cut group of depressed patients are identified. However, in the clinical setting this approach may miss ‘cases’ of depression who could benefit from treatment as would have been the case for two patients in our group had we adopted these criteria. Use of the substitutive approach, which has been suggested to be more suitable than the other three approaches for use in the medically ill, also has limitations. In this study, this approach led to the diagnosis of depression in two patients not found to be depressed using the aetiological approach. Of note, on clinical review neither was considered to have a depressive illness (i.e. were regarded as false positives). This potential for creating false positives, and the need to choose appropriate substitute items according to the nature of the physical illness studied are major criticisms of this approach. However, in this study the approach produced only a small number of false positives (2/17) and no false negatives, suggesting it may be of more practical value than the inclusive approach. If this approach to diagnosis is to be used in populations of medically ill patients, further consideration will need to be given to the most appropriate way of choosing and validating substitutive criteria.
The 17 patients found to be depressed using DSM-III-R criteria were all moderately to severely depressed (mean ± SD HDRS score 26.2 ± 5.2). While this may be expected in a psychiatric inpatient setting, this is not often the case in the consultation-liaison setting.
This bias towards more severe depression may account for the fact that contrary to our expectations and data in the literature, the four rating scales with their different emphasis on somatic and cognitive symptoms equally differentiated (mean ± SD score) between depressed (DSM-III-R) and non-depressed patients. Also contrary to findings in the literature [10, [25], [28]] somatic/vegetative symptoms appeared to discriminate as well as cognitive/affective symptoms between those patients with and without depression. Sleep, appetite, psychomotor retardation and agitation were all significantly different between the depressed and non-depressed patients. By contrast, loss of libido, fatigue, hypochondriasis and loss of weight were poor discriminators between the depressed and non-depressed groups. This is not surprising in a group of patients where fatigue, concern about and focus on one's physical health, weight loss and the use of a wide variety of drugs are the norm. These findings require further investigation in samples of medically ill patients with diagnoses other than HIV disease and with varying severity of depression.
If we look to the rating scale items to provide some guidance as to what symptoms would be most useful as part of the diagnostic criteria for depression in this group of medically ill patients it appears that a mixture of cognitive/affective symptoms, suicidal ideation, sleep and appetite disturbance may be the most helpful. This symptom mix is similar to that contained in DSM-III-R and used in medically well populations.
Finally, when considering our results, it is important to note that the findings of this study must be evaluated by taking into account a number of methodological difficulties. The sample size was small (limited by the study duration) and given the admission policy of the hospital (i.e. to admit patients to the medical wards for treatment of a psychiatric disorder) cannot be seen as ‘typical’ of those patients usually seen in a consultation-liaison setting. The recruitment process was such that the interviewer only assessed patients for depression if they had been seen by their physician and identified as being ‘depressed’. We do not know anything of those patients who were depressed and not identified as such by their physician (e.g. those with milder forms of depression or those with an atypical clinical picture). Inclusion of such patients in the sample may well have changed the results and may in part account for the differences between our study and that of others previously reported in the literature [10, [17], [26]].
Conclusions
Concern has been expressed about the appropriateness of using DSM criteria in the consultation-liaison setting. Thus, it is of note that in this study of HIV-positive patients, we found the DSM-III-R (aetiological approach) criteria to be clinically useful for identifying patients with depression. This maybe due to the familiarity of the clinicians involved in this study with the physical manifestations of HIV disease and their confidence in attributing symptoms to either a physical or psychological cause. Where clinicians are not so familiar with the physical illness, substitutive criteria maybe of most use. However, these criteria need to be developed by clinicians who are familiar with the physical illness of interest and need to be validated for each physical illness in which they are used.
We found all four rating scales tested here to be of use. However, as this is contrary to the findings of previous studies this requires further investigation.
Studies in larger samples of patients with HIV infection and depression will be required to further assess the comparative utility of the four different approaches to diagnosis and the comparative utility of commonly used depression rating scales.
Footnotes
Acknowledgements
The authors are grateful to the patients of Fairfield Hospital for their participation and to the staff of the Hospital for their cooperation and encouragement during the study. We thank Dr John Lloyd and Dr Andrew Gibbs for their advice regarding neuropsy chological assessment.