Depression and physical illness among elderly general medical clinic patients

Journ~~l of Affrcme Elsevier

153

Lkvorders. IO (1986) 153-162

JAD 00367

Depression and Physical Illness among Elderly General Medical Clinic Patients W.A. Kukull ‘.‘, T.D. Koepsell I.*, T.S. Inui ‘,3, S. Borson 4.5, J. Okimoto 4, M.A. Raskind 4~5and J.L. Gale ’ ’ Northw’est HSR&D Field Program. Scuttle VA Medtcal Center, ’ Depurtment of Eprdemrologv, Unrwrsrty of Washrngton. ’ Deprrrtments of Medmne and Hmlth Serwces, Unwersity of Washrngton. ’ Gerratrtc Research and Educutron Clmrcal Center, Scuttle VA Medtcrrl Center. und i Department of P.yychratrv and Behaoiorol Scrences. Unroerstty of Washrngton. Secrttle. WA 98195 (U.S.A.) (Received 7 November, (Accepted 7 February.

1985) 1986)

Summary

In this study we conducted a resurvey at 33 months of elderly general medical clinic outpatients previously classified as depressed or not using the Zung Self-Rating Depression Scale. Resurvey results and review of medical records permitted characterization of the point prevalences of depression at the time of the initial and follow-up surveys, and identification of physical illness factors associated with depression. The point prevalences of depression were approximately equal (20%). although only about 10% were depressed at both occasions. Among the initially nondepressed, the number of new physical diagnoses during follow-up was the best predictor of depression at retest. Other factors associated with depression at one or both occasions were: alcohol abuse, obstructive pulmonary disease, and a relatively greater number of medical diagnoses. Thus, among elderly outpatients, depression appears common with roughly equal rates of remission and incidence; also, new medical illness may precipitate depression.

Key words:

Depression

~ Elderly ~ ‘Epidemiology

The prevalence of depression and/or depressive symptoms in clinic populations. particularly in the elderly, is substantial (Widmer and Cadoret 1978; Barsky 1979; Hoeper et al. 1979; Boyd and Supported in part by the Northwest HSR&D Field Program. Seattle VA Medical Center, Seattle, WA. U.S.A. Please send reprint requests to Dr. Walter A. Kukull. Department of Epidemiology. SC-36. University of Washington. Seattle, WA 98195, U.S.A. 0365-0327/86/$03.50 ’

1986 Elsevier

Science Publishers

~ Physical illness

Weissman 1981). However, the diagnosis of affective disorder in the elderly is complicated by physical illness and possibly other correlates of advancing age (Blazer 1980). Depression may be masked by somatic complaints (Salzman and Shader 1978; Barsky 1979; Nielsen and Williams 1980; Katon 1982) or it may be the result of a particular physical illness (Hall et al. 1978. 1980). Further, among medically ill patients the relationship between illness severity and depression is uncertain (Ripley 1947; Stewart et al. 1965:

B.V. (Biomedical

Division)

154

Schwab et al. 1967. 1968: Moffic and Paykel 1975). The course of depression in the elderly is unclear (Blazer 1980; Blazer and Williams 1980: Hankin and Locke 1982). For some patients with major depressive disorders. symptoms may persi.st in excess of 2 years. Others experience episodes of 6 months or less in duration and are more likely to recover (Keller and Shapiro 1981). Which of these subgroups predominate in elderly medical patients is not known. In the present study we conducted a 33-month follow-up of elderly VA medical clinic patients to determine first. whether there would be a change in the prevalence of depression at the beginning as compared to the end of follow-up. and. second whether the physical illness burden of depressed patients would be different from those not depressed. Methods

The study population (n = 396) consisted of all patients at the Seattle VA Medical Center general medical clinic who were at least age 60. and who provided complete responses to the Zung Self-Rating Depression Scale (SDS) (Zung 1965) on a survey conducted by Okimoto et al. (1982) in February, 1979. This survey is called Time 1 (T, ). Of the 396 initial subjects. 230 responded to a second survey completed in October, 1981. This survey is called Time 2 (T,). Of the remaining 166 patients, 71 had died. 4 were ‘too sick’, 42 refused, and 49 were lost to follow-up. Complete medical chart data were available for 202 of the respondents at Tz (62.2% overall). Live nonresponders did not differ to a statistically significant degree from responders with regard to marital status. employment, income, type of housing, home inhabitants, age or depression status as measured at T,. Education was significantly associated with response status (P < 0.02). lower education favoring nonresponse. Education was not associated with depression at T, or among T2 responders at either survey. however. The final group of responders consisted of 197 males and 5 females; the mean age (f SD) was 69.4 f 7.3 years at T,. Twenty percent of the 202 responders scored 60 or greater (i.e., depressed) on the SDS at T,.

The Zung SDS provides an index score which ranges from 25 to 100. Characteristics of the Zung SDS (e.g. internal consistency. factor structure) from a number of studies through 1978 were comprehensively reviewed by Hedlund and Vieweg (1979). They demonstrated the SDS to be of acceptable internal consistency and that it correlated well with other widely used depression scales (e.g. Beck Depression Inventory. Hamilton Rating Scale for Depression). In a study of normal aged (Zung 1967) Zung concluded that the elderly. in general. may have a higher baseline score than nonelderly due to a predominance of physical symptoms. Zung (1967) suggested that the index score indicating probable depression ( 2 50 in nonelderly) be increased. Steuer et al. (1980) have shown that while SDS somatic symptoms aubscale was significantly associated with patient health. as rated by a physician, the overall SDS index score was not. However. a specific cutoff value was not suggested. Two other studies (Kitchell et al. 1982; Okimoto et al. 1982) report that the value of 60 for the SDS index score indicating probable depression was specifically suggested by Dr. Zung based on his scale development work with both VA and non-VA elderly. Okimoto et al. (1982) determined that index scores of 60 or greater were highly predictive of a DSM-III (American Psychiatric Association 1980) diagnosis of Major Depressive Disorder based on a psychiatric interview. Okimoto’s conclusions were based on an independent sample of 55 patients from the same clinic population as the current study. Therefore. the present investigation has adopted the index score of 60 or greater as indicative of depression. The Seriousness of Illness Scale (Wyler et al. 196X) was used to assign a score for each patient corresponding to the one most serious diagnosis active during the study interval from medical record review. The Seriousness of Illness Scale was developed from a list of approximately 120 different problems which presented at an outpatient clinic. These problems ranged from dandruff to leukemia. Samples of physicians and lay persons ranked the individual problems on a scale (O&1000) given that peptic ulcer would have a value of 500. Scores assigned to the final scale were a combination of the physician or lay rankings which re-

155

fleeted the perceived life threat of the conditions. As an adjunct measure of illness severity for the present study. the case fatality rates for 3.50 diagnoses based on inpatients in Professional Activity Study (PAS) member hospitals (Commission on Professional and Hospital Activities 1975) were used. These case fatality rates were based on inpatient stays only and reflect a rough probability of death given hospital admission for a particular problem. For each study patient’s medical diagnoses. the highest case fatality rate corresponding to a single diagnosis was recorded. No attempt was made to estimate severity for combinations of diagnoses. or to estimate the patients’ perception of a particular condition. Procedure Study instruments were mailed to all persons identified by the initial survey (T,). Nonrespondents were sent a repeat mailing and if necessary were contacted by telephone to encourage survey completion. The final date for response was October 31, 1981. Washington State death records and VA medical records were searched to determine whether nonresponders had expired. Medical records of responders were reviewed to determine age. sex, number of visits to each clinic, ‘prevalent’ (i.e.. those active at T,) and ‘incident’ diagnoses (i.e., those with onset between T, and TZ ), diagnostic tests ordered in the general medicine clinic. number of hospitalizations and total length of stay, drugs taken for less than one month, and drugs taken for more than one month. Clinic visits, diagnostic tests, drugs, and hospitalizations are the subject of another analysis. Diagnoses were grouped according to ICD-9 (Commission on Professional and Hospital Activities 1978) major classification (infectious, neoplastic, endocrine. etc.) and also coded secondarily as more specific minor subgroups within each major class. Anu!vses The prospective analyses involved classifying subjects according to depression status at T,. Those with SDS scores of 60 or greater (depressed) at T, were called the ‘natural history’ group referring to the course of depression among those initially depressed. Those with lower SDS scores were called the ‘etiology’ group referring to the inci-

dence of depression among these initially not depressed. The etiology group (n, = 162) was tested for variables associated with development of a score in the depressed range at T?. Continuous variables representing age and seriousness of illness as well as number of diagnoses in each of 18 major ICD-9 groups, and categorical variables representing 62 disease subgroups within the initial 18 major diagnosis groups were screened for possible association with T, depression. The natural history group (n? = 40) likewise was tested for variables associated with an SDS score in the depressed range at Tz. In each case bivariate and multivariate (logistic regression) analyses were conducted. When considering the prospective analysis, one should note that depression status was measured by the SDS only at T, and T,, not throughout the 33-month interval. It is therefore possible that episodes of depression may have occurred in the interim or that persons normal at T, and Tz may include a subset of persons who experienced episodic depression within the interval. Thus. questions of clinical course and incidence which pertain to episode duration or onset cannot be answered definitively. Viewing the data in another way, cross-sectional analysis consisted of reclassifying subjects according to combined T, and Tz depression status, (n, = 140), once-depressed i.e.. never-depressed (n Z = 41) and twice-depressed (nj = 21). The latter two categories may serve to separate patients with short-term or labile depression from those with a more chronic form of depression. Both bivariate and multivariate statistical methods were again applied to detect variables associated with depression. Results Table 1 shows the depression status of persons completing both T, and T, surveys (n = 202). Roughly one-half of the persons depressed at T, had recovered, i.e., were not depressed. at Tz. Of those persons not depressed at T,. approximately 14% were depressed at T2. Since nearly equal numbers of persons changed status, McNemar’s chi-square test of change was nonsignificant. Simi-

156

TABLE

1

DEPRESSION

STATUS

AT T, AND T, ’ Total

T2 Depressed

Not depressed

21 22

19 140

40 162

43

159

202

T, Depressed Not depressed Total “ McNemar’s

chi-square:

0.097.

P > 0.75.

larly. the overall prevalence of depressive symptoms at T, and T, remains approximately unchanged, at 20%. Persons with an SDS score less than 60 at T,. the etiology group (n = 162) were screened as described above for variables associated with depression at T, (SDS > 60). Table 2 shows that the number of incident diagnoses and total diagnoses were significantly greater among those depressed at T:. Remarkably, neither age or seriousness of illness differed significantly between those patients who were depressed at T7 and those who were normal. One specific diagnosis ~ cataracts ~ was more common among those depressed at T, to a statistically significant degree: however. when considered as part of the multivariate model, eye problems failed to contribute significantly to explanation of Tz depression status.

Stepwise logistic regression was used to identify the variables from among those which contributed the most to explanation of depression status at T,. Age was entered a priori: then the physical illness variables shown in Table 2 were allowed to enter. in stepwise fashion. if their P-value to enter was less than 0.05. After the number of incident diagnoses was entered, no other variable added significantly to the model. Table 3 shows adjusted relative odds of predictors in that analysis in the left vertical columns (cross-sectional analyses are included in Table 3 also). The relative odds is a measure of the strength of association between depression and number of diagnoses or a specific diagnosis. A relative odds of one would imply that persons with a specific physical diagnosis were no more likely than persons without the diagnosis to exhibit depression at T, whereas a relative odds of 2 would indicate that they were twice as likely to be depressed. When interpreted from a logistic regression model. relative odds are adjusted for other included variables, i.e., the unique contribution of each variable to the association is measured given all others present in the model. According to this model (Table 3. left column) TABLE

3

ADJUSTED RELATIVE ODDS (95% CI) FOR PREDICTORS OF DEPRESSED STATES RESULTING FROM SEPARATE LOGISTIC REGRESSION MODELS Predictor

Etiology

model

Depressed/Not TABLE

2

Age

PHYSICAL ILLNESS VARIABLES T, DEPRESSION FOR ETIOLOGY PRESSED AT T, ) Variable

ASSOCIATED WITH GROUP (NOT DE-

P-VZllLX

Mean (SD)

Incident diagnoses Total diagnoses Injury diagnoses

Cataracts

Not depressed (n = 140)

Depressed (n = 22)

2.3 (2.0) 6.9 (2.9) 0.1 (0.4)

3.X (3.1) X.4 (4.1) 0.4 (0.9)

B (No.)

% (No.)

12.1 (17)

36.4 (8)

Incident dlagnors

1.03 (0.97- 1.09)

chl-square

models

Ever/Never depressed

Twice/Once

1.oo (0.95%1.05)

0.942 (0.87. 1.02)

depressed

1.27 (1.06-1.53)

Prevalent diagnoses

1.30

( 1.OO~ 1.70) 1.18 (1.07~131)

0.005 0.037 0.047 COPD 0.009 a (n = 162)

’ Corrected

Cro

T2

158

scores at T, and T2, (b) depressed SDS scores at either T, or T2 only, and (c) depressed SDS scores at both T, anri T,. The groups will be referred to as ‘never-depressed’, ‘once-depressed’ and ‘ twicedepressed’ respectively. The previous analysis (etiology group) referred only to persons who scored in the normal range (SDS < 60) at T, and followed them to T2. The results for the etiology group analysis should not be confused with those of the cross-sectional groups, since both numbers of subjects and group composition. re SDS scores. are different. Variables were screened as described above. Table 5 shows the variables for which significant P-values were obtained denoting group differences or associations. These variables were subjected to stepwise logistic regression with age being entered a priori; additional variables entered the model if they added significantly to prediction over and above those already included. Thus, variables highly correlated with those already in the model failed to enter. Two logistic models were constructed, one to differentiate never-depressed subjects from once-or-twice-depressed subjects (Table 3, center column), and the second to attempt to differentiate once-depressed from twice-depressed subjects (Table 3, right column). The logistic models presented in each column are the final models for each analysis. The relative odds reported are

TABLE

5

VARIABLES Variable

adjusted for age as well as the other variables in that particular model. After adjusting for age, a diagnosis of chronic obstructive pulmonary disease (COPD) and alcohol abuse and a greater total number of diagnoses were associated with scoring in the depressed range on the SDS on one or both occasions (Table 3. center column). Table 3 also shows the adjusted relative odds of those variables which entered the logistic model to predict twice-depressed status from among those depressed on at least one occasion (the right column). As noted on the previous model, alcohol abuse and illness burden (this time in terms of the number of prevalent diagnoses) were found to predict scoring in the depressed range on both occasions. As noted earlier. ‘prevalent’ in this context refers to conditions which were diagnosed prior to the study interval yet still active within the interval; ‘incident’ refers to conditions diagnosed during the period between T, and T2_ The sum of incident and prevalent is the total number of diagnoses. In order to determine whether cross-sectional depression status (never vs. once and twice) was due primarily to symptoms related to the existence of COPD or alcohol abuse, discriminant functions were constructed (as above with the ‘etiology group’) from T, SDS items and separately from T, SDS items. Separate functions were computed to

AND

P-VALUES

~’

FOR CROSS-SECTIONAL

GROUPS

Mean (SD)

P-\alue

Never depressed (n, = 140)

once depressed (n, = 41)

Tn ice deprCWd (rl-21)

Prevalent diagnoses Incident diagnoses Total diagnoses

4.6 (1.9) 2.3 (2.0) 6.9 (2.9)

5.1 (2.1) 3.6 (3.0) x.7 (3.9)

6.5 (2.6) 3.4 (2.7) 9.9 (3.7)

0.0003 0.0030 0.0000

Alcohol ahuse Cataracts COPD Intestinal diagnoses Esophageal/stomach

13.6 12.1 24.3 17.9 14.3

24.4 31.7 39.0 12.2 31.7

52.4 33.3 66.7 3x.1 28.6

0.0004 0.0035 0.0003 0.0413 0.0239

Percent

diagnoses

’ For categorical variables percent with the diagnosis and the P-value the P-value results from a one-way ANOVA. F-test.

of the chi-square

tet

are included.

For contmuous

vnrlahler

159

discriminate between patients with and without COPD. alcohol abuse, and depression. As with the ‘etiology group’ the purpose of the discriminant analysis was to describe item combinations. If we were to find that from among our pool of subjects the same (or substantially similar) SDS items were associated with COPD (or alcohol abuse) as were associated with depression, we might conclude that what we were witnessing as depression was only an effect of the COPD or alcoholism. To the extent that differenr items are significantly associated with COPD or alcohol abuse versus depression we could be more certain that the observed SDS score was not merely a reflection of the predominant physical problem. Table 4 shows the SDS items which provided significant discrimination for each condition. The leftmost column refers only to the etiology group analysis reported above, and should not be confused with the 3 columns to the right pertaining to the cross-sectional analysis. As explained earlier the etiology group is a specific subgroup of the total number of subjects included in the cross-sectional groups. Thus, items associated with being ‘once or twice depressed’ in the cross-sectional analysis may not correspond directly to items associated with onset among subjects who were initially normal. Table 4 shows that the exception of item 10, ‘fatigue’ (COPD) and item 18, ‘emptiness’ (alcohol), there are no other discriminating items which overlap with cross-sectional depression status. Depressive status is characterized as a combination of physical and psychological symptoms. Symptoms best describing concurrent COPD in the cross-sectional group are: loss of appetite. fatigue, and psychomotor retardation, the former 2 symptoms associated with physiological disturbance. Alcohol abuse in contrast, is characterized by symptoms of weight loss (physiological disturbance), as well as psychomotor agitation, personal devaluation and emptiness (psychological disturbances). Discussion

The present study was an attempt to examine some possible causes and consequences of depression among elderly patients attending a general medical clinic. with a focus on physical illness

factors. Since such individuals are already in contact with clinicians for care of other illnesses, the occurrence and clinical course of depression in this group is of interest because of the special opportunities they present for recognition and intervention. In addition, the frequency of medical illnesses in this patient population made it a particularly suitable group for study of the relationships between physical illness and depression. Among survivors at follow-up, the overall prevalence of depression remained approximately 2047, closely similar to the baseline prevalence of depression 33 months earlier. However, there proved to be considerable turnover in the depressed group over time: about half of those classified as depressed at baseline were no longer so classified at follow-up, but they had been replaced by an approximately equal number of newly depressed individuals. The changes in depression status over time can be viewed in at least 2 ways. First, those who changed status during the study may consist of some initially depressed individuals who recovered spontaneously or were successfully treated between assessments, plus another group of individuals who first became depressed during the interval. This viewpoint led to the ‘etiology’ and ‘natural history’ analyses. Second, those who changed status during the interval may consist of individuals who experience repeated temporary episodes of depression, to be distinguished from those with more chronic depression or no depression. This viewpoint led to the ‘cross-sectional’ analyses in which 3 patient groups were compared. Because it is possible that short episodes of depression or remission occurred between baseline and follow-up but were missed due to the length of time between surveys, both sets of analyses may have been subject to misclassification. A study in which more frequent surveys occurred repeatedly would be valuable in evaluating the extent of this effect, but such an approach was beyond the scope of the present investigation. Among patients classified as not depressed at baseline. both bivariate and multivariate analyses showed that measures of a patient’s physical illness burden were important predictors of depression status at follow-up. Considered singly, 4 summary measures were all positively associated with the risk of developing depression between the

160

baseline and follow-up assessments (number of incident diagnoses, total number of diagnoses, cataracts. and number of injury diagnoses). The most powerful predictor among these was the number of incident diagnoses: no other physical illness variables added significantly to a predictive multivariate model once the number of incident diagnoses had entered the model. The finding that incident diagnoses were more powerful predictors than prevalent diagnoses may indicate that the development of a new physical illness is a stronger risk factor for depression than is the existence of one or more chronic illnesses. This could in turn reflect either differences in the symptomstology and course of these groups of illnesses. or it could reflect greater opportunities to develop coping mechanisms for chronic diseases. Although a few specific diseases emerged as significantly associated with depression onset, there was no evident ‘common denominator’ among them. Given the number of individual diagnoses considered. chance alone could easily account for some of these associations. Seriousness of illness did not differ for depressed versus nondepressed patients at either level of analysis for the etiology group. contrary to results of earlier studies (Stewart et al. 1965; Schwab et al. 1967, 1968; Moffic and Paykel 197.5). The restricted range (homogeneity) of Seriousness of Illness Scale scores in this sample ~ most patients in the present study had scores toward the upper end of the scales employed may have contributed statistically to this lack of association. Also, since persons who died prior to T2 were excluded, the most seriously ill persons may have been eliminated from study, thus modifying any association. The PAS case fatality rates derived from inpatient experience may not have been directly applicable to persons visiting an outpatient clinic. Although a particular score was attached to a patient by one of the above methods it may not have reflected either the patient’s functional ability or how a specific diagnosis impacted the patient psychologically. Further, since illness burden in terms of number of diagnoses was shown to be associated with depressive status and only the one most serious illness was scored, the lack of effect could have been due to the inability to arrive at a suitable seriousness score for a patient

with several diagnoses. These general medical clinic patients were chronically ill. having several active diagnoses. The number of diagnoses. as shown in the Results. may be more important than any one seriousness score. If the study had been based on an index hospitalization and a seriousness score assigned to the principal diagnosis therein, the resulting effect of seriousness may have been different. Refinement and/or resealing of the seriousness scales to account for some of the problems noted above would provide a more substantial foundation on which to build conclusions about the seriousness of illness and depression. Discriminant analysis of SDS items associated with becoming depressed (Table 4, left column) showed that the physiological items: decreased libido and weight loss combined with affective and psychological items: pervasive depressed affect. hopelessness. suicidal rumination and dissatisfaction, contributed most to the prediction of depression at follow-up from among those persons who were initially not depressed. Whether these symptoms are antecedent or consequent to the excess of ‘incident’ physical illness cannot be dctermined from these data. It is clear. however. that persons classified as depressed at T, did not obtain that classification solely because of the expression of physical symptoms related to incident illness. Unfortunately. the relatively small number of individuals classified as depressed at baseline (n = 40) precluded drawing any firm conclusions about relationships between physical illness factors and the clinical course of depression in afflicted individuals. In the cross-sectional analyses, patients with depression at least once (at T, or Tl) were conpared with never-depressed individuals. and patients with depression on both assessments were compared with those depressed on only one assessment. These divisions were made in an attempt to segregate individuals with no depression. and longer-term deprexshort-term depression, sion. respectively. to permit examination of possible differences in the clinical profiles of the 3 groups with respect to physical illness factors. Because status at T, and Tz counted equally in this analysis. however. the temporal sequence of any relationship between physical illness and depres-

161

sion may be in either direction. As shown in Table 5, there was a fairly regular pattern of association between several summary measures of physical .illness and depression status: those with depression at least once had a greater number of diagnoses overall and in several organ system categories than did those never-depressed, and those with depression on both assessments carried a greater number of these diagnoses than those depressed only once. Multivariate analyses to distinguish ever- from never-depressed and twice- from once-depressed suggested that a diagnosis of alcohol abuse, in particular, was more common with depression and longer-term depression; chronic obstructive pulmonary disease was more weakly associated in the same direction. The association between depression and alcohol abuse could reflect 2 forms of response to a common set of precipitating factors, or one disorder could be a cause of the other. The specific diagnoses of COPD and alcohol abuse were not important predictors of depression onset, yet were associated with depression in the cross-sectional analysis suggesting a possible association with chronicity. Alcoholism and COPD are probably more common in this VA sample than in a non-VA population. For example, in a predominantly female, upper middle class sample alcohol abuse and COPD may be supplanted by other specific diagnoses. In a larger series of VA patients (Borson et al. 1985) a weak yet suggestive association between depressive symptomatology and COPD was noted, lending some support to current findings. The P-values obtained in the present study (Table 5) give us further confidence that the observed associations are unlikely to be the result of chance. A Bonferroni adjustment of these P-values for the number of disease categories screened (approximately 80) indicates that the ‘true’ P-values for alcohol abuse and COPD would still be less than 0.05. Whether this type of adjustment is necessary is in itself a matter for conjecture (Maclure 1985). Although the power of these analyses to detect other diagnosis-specific associations was low, the regularity of the pattern of associations across disease subgroups (Table 5) and the appearance of global measures of disease burden in both multivariate predictive models (Table 3) would seem to suggest broad associations between depression, its chro-

nicity, and number of physical illnesses. Illness burden may be the most generalizable factor associated with depression in this analysis. rather than the specific diagnosis. As shown by discriminant analyses (Table 4) SDS items associated with conditions of COPD or alcohol abuse overlap little with items associated with being depressed at least once in the cross-sectional analysis. The luck of overlap indicates that the observed SDS score was not due simply to signs and symptoms of COPD or alcohol abuse being reported. A combination of physiological and psychological SDS items were associated with ‘ever’ being depressed. Overall. the findings of this study suggest that depression is a common problem in elderly medical clinic patients. Rates of remission and incidence are substantial and about equal. The onset of new physical illness appears to be an important risk factor for development of depression. Moreover, there appear to be positive associations between the chronicity of depression and the number of coexisting physical diagnoses. References American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders. 3rd edn.. American Psychiatric Association. Washington, DC, 1980. Barsky. 111. A.J.. Patients who amplify bodily sensations. Ann. Intern. Med., 91 (1979) 63-70. Blazer, D.. The diagnosis of depression in the elderly. J. Am. Geriatr. Sot.. 28 (1980) 52-5X. Blazer, D. and Williams. C.D.. Epidemiology of dysphoria and depression in an elderly population. Am. J. Psychiatry. 134 (1980) 329-333. Boraon, S., Barnes. R.A.. Kukull. W.A.. Okimoto. J.T.. Veith. R.C., Inui. T.S., Carter. W.B. and Raskind. M.A.. Symptomatic depression in elderly medical outpatients. I. Prevalence, demography, and health service utilization. J. Am. Geriatr. Sot.. (1985) in press. Boyd. J.H. and Weissman, M.M., Epidemiology of affective disorders: a re-examination and future directions, Arch. Gen. Psychiatry, 3X (1981) 103991046. Commission on Professional and Hospital Activities. Professional Activities Survey, Mortality in PAS hospitals. 1974, Ann Arbor. MI. 1975. Commission on Professional and Hospital Activities. The International Classification of Diseases, 9th Revision. Clinical Modification. Ann Arbor, MI, 1978. Hall, R.C.W., Popkin. M.K.. DeVaul, R.A.. Faillace. L.A. and Stickney, SK.. Physical illness presenting as psychiatric disease, Arch. Gen. Psychiatry. 35 (1978) 1315-1320.

162

Hall. R.C.W.. Gardner. E.R.. Sticknq. S.K.. LeCann. .4-F. and Popkin, M.K.. Physical illness manifesting as psychiatric divas II. Analysis of a state hospital inpatlent population. Arch. Gen. Psychiatry. 37 (1980) 989-995. Hanhln. J.R. and Locke. B.Z.. The persistence of deprcsai\c symptomatolog~ among prepad group practice enrollees: an exploratory study. Am. J. Public Health, 72 (1982) 1000-1007. Hedlund. J.L. and Virweg. B.W.. The Zung Self-Rating Depresslon Scale: a comprehensive reviea. J. Operat. PsychIatry. IO (1979) 51-64. Hoeper. E.W.. NYU. G.R.. Cleary. P.D., Regler, D.A. and Goldberg. I.D.. Estimated prwalence of RDC mental disorder in primary medical care. Int. J. Ment. Health. X (1979) 6-15. Knton. W.. Depression: somatic symptom5 and medical di\orders in primary care. Compr. Psychiatry. 23 (1982) 274-287. Keller. M.B. and Shapiro. R.W.. Mqor depressive disorder: imtlal rraults from 3 one-year prospwtlve naturalistic follo\*-up study. J. New. Mcnt. Dis.. 169 (19X1) 761-768. Kltchell. M.A., Barnes. R.F., Veith. R.C., Okimoto. J.T. and Raskind, M.A.. Screening for depression In hospitalized geriatric medical patientA. J. Am. Gcriatr. Sot.. 30 (1982) 174-177. Maclure. M.. Popperjan refutation in epidemlolog). 4m. J. Epidrmiol.. 121 (19X5) 343-350. Moffic. H.S. and Paykel. E.S., Dcpres\lon in mrdlcal inpatient.\. Br. J. Psychlatr): 126 (1975) 3466353. Nlclsrn. A.C. und Williams. T.A.. DeprrsGon in ambulatory medxal patients: prewlencc h> self report questionnaire and recognition by nonpsychiatric phy\icxms, Arch. Grn. Psychiatry. 37 (19X0) 999-1004.

Okimoto. J.. Barnes, R.F.. Vrlth. R.C., Raskind. M.A.. Inui. T.S. and Carter. W.B.. Screening for depression 1n geriatric medical patients. Am. J. Psychiatry. 139 (1982) 799-802. Riple!. H.S.. Depressive reactions in a general hospital, J. New. Ment. Dis.. 105 (1947) 607-615. Salrman. C. and Shader. R.I.. Depression 1n the elderly. 1. Relatwnship between depression. paycholnglc drfenx mechanisms and physcal Illness, J. Am. Cierlatr. Sot.. 26 (197X) 253-260. Schwab. J.J.. Bialow. M.. Brown. J.M. and H&u. (‘YE.. Dagnwng depression m medlcal Inpatients. Ann. Intern. Med.. 67 (1967) 695-707. Schwab. J.J.. Brwn. J.M. and Holzrr, C.E.. Drpreahion in medical inpatients with gastrointestinal diseases, Am. J. Gaatroenterol.. 49 (1968) 146-152. Steucr. J.. Bank. L.. Olwn. E.J. and Jarwk. L.F.. Deprrwon. phyxal health. and somatic complamts 1n the elderlq: a study of the Zung Self-Rating Depresvon Scale. J. Gcrontol.. 35 (1980) 683-6X8. Stewart. M.A.. Drake. F. and Wlnokur. G.. Depresston among mrdlcally III patients. Dis. New. Syt.. 26 (1965) 479-4235. Wldmer. R.M. and Cadoret, R.J.. Depresson in primary care: changes tn pattern of patient wslta and complaints during ;I developing depression. J. Fam. Pratt.. 7 (197X) 293-302. Wyler. A.R.. Masuda. M. and Holmes. T.H.. Seriousness of illness rating scale. J. Psychosom. Rea.. 11 (196X) 363- 374. Zung. W.W.K.. A self-ratmg depression scale. Arch. C;en. Psychiatry. 12 (1965) 63-70. Zung. W.W.K.. Depreuion 1n the normal aged. Paychox>matlcs. X (1967) 289-292.