CES-D symptoms and DSM criteria: A method for classifying levels of se
http://www.100md.com
《中华医药杂志》英文版
CES-D symptoms and DSM criteria: A method for classifying levels of self-reported depressive symptoms in the elderly
1 Departments of Biostatistics, Biomathematics & Bioinformatics and Psychiatry, Georgetown University School of Medicine, Washington, D.C.20007，U.S.
2 Department of Psychiatry, University of Texas Southwestern Medical Center, Dallas TX,U.S.
3 Population and Health Research Center, Bureau of Health Promotion, Taichung, Taiwan
Correspondence to Rochelle E. Tractenberg, Department of Biostatistics, Biomathematics and Bioinformatics, M7202, 7 East Main Georgetown University Hospital, 3800 Reservoir Road, NW, Washington, D.C. 20007,U.S.
Tel:2024448748,Fax:2024444114
E-mail:ret7@georgetown.edu
[Abstract] Objective Even without reaching a diagnostic threshold, symptoms of depression are associated with worse function, greater sleep disturbance, and lower life satisfaction. In large-scale studies, cutoff scores are frequently used to identify “depression”. These cutoff scores tend to have suboptimal sensitivity and specificity; this can be more pronounced for non-major depressions. Here we propose and demonstrate a classification scheme combining the number of reported depression symptoms in an elderly cohort with their frequency ratings.Methods Eight questions were analyzed from a self-report depression instrument (CES-D), in Taiwan elderly (ages 65+ in 1989;N=3874) at three-year intervals from 1989～1999. Depressive symptom-levels were based on both number of symptoms and their respective frequency ratings. Total scores and symptom-level membership were traced triennially over ten years in this community cohort, and compared to cutoff scores.Results The number-frequency combination approach revealed distinct symptom-level groups with scores distributed around cutoff values. Women reported significantly more symptoms than men, but within the symptom-level groups, total scores did not differ by sex. Conclusions Discrepancies between diagnoses of major/minor depression and cutoff scores on self-reported instruments may be due, in part, to distinguishable subgroups (symptom-levels) with score distributions around the cutoff value. The number-frequency combination method does not require additional data and may improve identification of clinically relevant symptom levels with self-report instruments. Without confirmation of depression, sensitivity and specificity of our classifications could not be established. However, the approach does explain low sensitivity and specificity for cutoff scores relative to formal diagnostic criteria.
[Key words] elderly;depression;CES-D;longitudinal
INTRODUCTION
Depression is a highly prevalent disorder[1], associated with decreased quality of life[2] and loss of productive work time[3]. It is also associated with increased morbidity and mortality in many illnesses, with mortality in nursing home residents[4], and with increased mortality in elders independent of concomitant medical illness[5]. Of greatest importance is that depression is treatable[7] and that, in addition to improving function and quality of life[8], treatment of depression may have great impact on mortality, as has recently been shown in a study of post-stroke depression[9]. In the elderly, “subthreshold” levels of depression[10] have been associated with lower levels of function[11, 12] and life satisfaction[13]. Subthreshold depression is likely more frequently missed than major depression, and so might represent an important, as yet unquantified, health problem, particularly in the elderly.
When depression is not the main focus of a clinical or epidemiological study, its quantification may still be critical in the accurate evaluation of risks and trends, particularly in the elderly. Without expert evaluations, it is difficult to correctly characterize levels of depressive symptoms. The utility of cutoff scores on depressive symptom inventories, such as the Center for Epidemiologic Studies-Depression Scale (CES-D)[14], has been reported to be excellent[15], with good sensitivity and specificity in elderly as compared to a diagnosis by American Psychiatric Association’s Diagnostic and Statistics Manual (DSM)[16] criteria. However, high false positive rates[17], low positive predictive value[18, 19], and reduced sensitivity and specificity for minor depression[15] have also been reported for the CES-D when cutoff scores are used[18, 20].
Low sensitivity and specificity reflect discrepancies between DSM-based and cutoff-score-based labels of “depression”. The purpose of the CES-D and other self-report depression instruments is to measure depressive symptoms and not to provide diagnosis[17, 20, 21], yet there is a high degree of dependence on self reports in psychiatric epidemiology in particular[22] and in epidemiologic studies in general (i.e., the CES-D was developed for these contexts)[14]. Psychiatrists’ judgments about depressive symptoms (i.e., application of DSM criteria) necessarily reflect a different perspective than do self reports such as the CES-D[23]. In this study we describe a combination of elements of the DSM diagnostic approach and the CES-D symptom list to derive symptom-level categories that can be considered to reflect both qualitatively and quantitatively different types of depressive symptoms. To evaluate this method, we applied it to CES-D responses from a large cohort of elderly over four survey interviews during a ten-year period. Rather than using these categories to ‘diagnose’ respondents in this elderly cohort, the purpose of this study was to characterize levels of depressive symptoms over time and compare these results to those derived using cutoff scores.
MATERIAL AND METHODS
Subjects Individuals were sampled for the “National Survey of Health in the Elderly”, sponsored by “the Taiwan Ministry of Health” in 1988～1989. The sampling and study design are described elsewhere[24]. Briefly, a representative sample of over 4,000 persons in Taiwan over the age of 65 was recruited for the survey. This sample has the advantages of size, representativeness, and a focus on the elderly, in whom more subtle depressive symptoms may be of greatest concern[11~13]. The focus of the survey was not depression but general aspects of health and specific aspects of economic and social functioning.
Instrument These analyses focus on eight depression-related symptoms (poor appetite; doing anything is exhausting; poor/restless sleep; terrible mood/feel sad; feel lonely; feel people are unfriendly; feel anguished/depressed; no will to do anything/could not get going) that were translated from the CES-D and incorporated into the survey[24]; two other items (feel joyful; feel life is going well) were not included in our analyses. The eight negative items were rated for their frequency in the week previous to the interview (not at all=0; rarely (one day)=1;, sometimes (2～3 days)=2 or often/chronically (4～7 days)=3). These items were selected for inclusion when the survey was designed without specific reference to psychiatric literature but they overlap generally with the symptom list most frequently used as the ‘gold standard’ for establishing the diagnosis of depression (see below).
DSM-IV-TR[16], the most recent edition of the American Psychiatric Association Diagnostic and Statistics Manual, provides a list of symptoms of which five or more must be present in a patient to meet criteria for major depressive episode. These are: depressed mood or markedly diminished interest/pleasure in activities (at least one of which is required); significant unintentional weight loss or change in appetite; significant increase or decrease in sleep; restlessness or lethargy; fatigue or loss of energy; feelings of worthlessness; indecisiveness or inability to concentrate; and recurrent thoughts of death or suicide (p. 356). The presence of these symptoms must represent a change from previous function, which is usually not established in large-scale epidemiologic or survey studies.
The CES-D8 symptoms do not completely overlap with the DSM symptom list, nor are similar levels of frequency or the change from previous functional levels possible to establish; however, both DSM criteria and CES-D have been used across cultures for detecting depression[13, 19, 21, 25~29]. We combined internationally utilized CES-D items with the descriptive rigor inspired by the DSM, as described below, to characterize the level of depressive symptoms in this cohort over time.
Procedure CES-D items and DSM criteria: We established combination endorsement-frequency levels to classify respondents’ level of depressive symptoms as follows: According to the DSM-IV-TR[16], five or more symptoms must be present “nearly every day” for the two weeks prior to the assessment for an individual to meet the criteria for a major depressive episode (p. 356). The highest frequency rating possible for CES-D symptoms is “4 or more days in the past week”; therefore, individuals with 5～8 of the symptoms on the survey, all rated at the highest frequency (3), were characterized as “high depressive symptoms” (HDS). Because the symptom “felt people were unfriendly” does not appear in the DSM symptom list, individuals who met the HDS criterion with “unfriendly” as one of just five symptoms were not included as HDS but were classified at the next (lower) symptom level. Individuals classified as HDS were considered to have symptoms closest to the clinical criteria for major depressive episode.
DSM-IV-TR[16] specifies that for research purposes, a diagnosis of dysthymic disorder applies to individuals who have three or more symptoms “more days than not” (p. 775). Therefore, respondents rating 3～4 symptoms on the survey, all with the highest frequency (3), were characterized as “medium-high depressive symptoms” (MDS). Although the diagnosis of dysthymic disorder requires the presence of these symptoms for two years and we did not have that information, respondents classified as MDS were considered to have symptoms closest to the clinical (research) criteria for dysthymic disorder.
Individuals who rated 3~8 symptoms as occurring rarely or sometimes (1～2) were characterized as “medium-low depressive symptoms” (MLDS). These respondents were considered most similar to persons who might meet the criteria for subthreshold depression[10].
Individuals who rated 1～2 items at any frequency (i.e., not ‘never’) were characterized as “low depressive symptoms” (LDS), and individuals reporting no symptom were characterized as “no depressive symptoms” (NDS).
Depressive symptoms over time: Because we had no formal psychiatric evaluations for these respondents, we generally characterized depressive symptoms as the number of CES-D8 symptoms endorsed (having occurred at least one day in the previous week, 0～8) and the CES-D8 score (sum of frequency ratings, 0～24). We also determined respondents’ depressive symptom-level group membership (i.e., HDS, MDS, MLDS, LDS or NDS) at each wave; persons meeting two definitions were put into the higher-level group. Finally, we examined the distribution of CES-D8 scores for the depressive symptom levels at each wave. All statistical analyses were carried out using SPSS 11.5 (SPSS Inc. Chicago, IL). Multiple comparisons (across survey waves) were adjusted for using Holm’s method (a modified Bonferroni method[30]), and adjusted P-values less than 0.05 were considered significant.
RESULTS
Background (demographic) values: Table 1 presents the background and descriptive/demographic variables for the full cohort at the first survey (1989, “baseline”). The proportion of the sample falling into the depressive symptoms groups at baseline is also presented.
Table 1 Background/Demographics overall and by Sex at First (Baseline) Survey (1989) Mean (±SD) or percent
TP1
Notes:NDS: No depressive symptoms. LDS: “Low” depressive symptoms; 1~2 items at any frequency. MLDS: “Medium-low” depressive symptoms; 3~7 symptoms rarely or sometimes. MDS: “Medium-high” depressive symptoms; 3~4 symptoms rated as always present. HDS: “High” depressive symptoms; 5~8 symptoms rated as always present. “No group”: Individuals who did not fall into any of the five a priori symptom level groups. Individuals falling into more than one group were assigned to the higher-level group. * Individuals reporting literacy but no formal education were assigned the value 0.5 years of education to differentiate them from those reporting no formal education and no literacy
The respondents were 57.1% male, 42.9% female. Their mean± SD age was 68.1 ± 6.5, educational attainment was 3.9 ± 4.5 years (with “literate, but no formal education” assigned a value of 0.5). Women were slightly, but significantly older (68.8 ± 6.7 years) than men (67.6 ± 6.3 years) (z=-5.6, P<0.001) and had significantly less education (z=-28.5,P< 0.001). Women also tended to rate their health worse at every interview (all chi square >64, all P< 0.001) and in each year, more men than women tended to rate their heath better than, or the same as, the past year (all chi square> 14, all P≤ 0.001).
When we compared individuals with valid CES-D8 scores at all four waves (“completers”) to those missing at least one score, we found that completers were younger (65.7 ± 4.7 vs. 69.3 ± 6.9 years), had more education (4.4 ± 4.6 vs. 3.6 ± 4.5 years) and lower baseline CES-D8 scores (2.4 ± 3.8 vs. 3.6 ± 4.5) (all t-tests significant at P<0.001). Higher baseline symptom severity level was also associated with an incomplete survey record (chi square=43.6,P<0.001): 40.2% of NDS, 31.4% of LDS, 28.7% of MLDS, 31.2% of MDS and 22.2% of HDS had complete records. Thirty-two percent of both men and women had complete data (P=0.9).
Age, education and depressive symptoms: We stratified the sample by sex and explored the association between age and education and the two continuous depressive symptom variables, CES-D8 scores and number of symptoms endorsed at baseline, using Spearman’s (nonparametric) correlation. Weak positive association was observed between age and both depressive symptom variables (both rho< 0. 075,P<0.05 for men,P<0.01 for women). Similarly, weak negative association was observed between education and both depressive symptom variables (-0.13 < both rho <-0.11). Although these values reached significance (P<0.001 for men, P<0.001 for women), age and education each accounted for less than 1% of the variability in either of these depressive-symptom variables regardless of respondent sex. CES-D8 scores for men and women with valid responses (i.e., all 8 items rated) are shown at each visit in Figure 1.
Figure 1 Mean score on CES-D8 (range: 0~24) over four survey waves, by sex
Figure 1 shows that over the three follow-up surveys, average CES-D8 scores increased for both male and female respondents (a similar pattern was observed for complete cases (N=1,782, 30.1% female); data not shown). We present all available data at each visit in Figure 1 because individuals with incomplete records had significantly higher CES-D8 scores on average (see above) and so although the trends were the same for completers and non-completers, without these non-completers the distributions would be biased downward.
For both men and women, lower education was significantly associated with higher levels of depressive symptoms at baseline (both chi square >20, both P<0.001). Age was not significantly different across depressive symptom levels (men: chi square=4.5,P=0.31; women: chi square=8.9,P=0.06) although at baseline, respondents who were older tended to have higher depressive symptom levels.
Depressive symptom levels: There were valid CES-D8 responses (i.e., all 8 items rated) for 3,874 of the 4,049 participants (93.5%) in the baseline survey. Of these, 1.4% (n=54) were classified as “high” depressive symptoms, and a further 2.9% (n=109) met our criteria for medium-high depressive symptoms. Over 70% of the respondents had low (33.3%) or no (40.5%) depressive symptoms in the week preceding the 1989 survey. Figures 2A and 2B present the distributions of the symptom severity groups for male and female respondents, respectively, with valid responses at all four waves (men:N=989; women:N=793).
Figures 2 2A and 2B. Distribution of depressive symptom levels (“severity group”) over four survey waves in men (2A, N=989) and women (2B, N=793) with valid responses at all four waves
Notes: NDS: No depressive symptoms. LDS: “Low” depressive symptoms; 1~2 items at any frequency. MLDS: “Medium-low” depressive symptoms; 3~7 symptoms rarely or sometimes. MDS: “Medium-high” depressive symptoms; 3~4 symptoms rated as always present. HDS: “High” depressive symptoms; 5~8 symptoms rated as always present. “No group”: individuals not otherwise classified. Individuals falling into more than one group were assigned to the higher level group
A sixth grouping of depressive symptoms was observed that did not fit with any of the five groups we defined a priori. Of the 3,874 individuals with valid CES-D8 responses at baseline, 113 (2.9%) fell into this sixth category. Nonparametric means comparisons revealed that, at every visit, individuals in this “no group” group had significantly greater symptoms on average than the LDS group and significantly less symptoms on average than the MLDS group (symptoms measured as number of symptoms endorsed, number rated “always”, mean symptom frequency rating, and CES-D8 score; all adjusted P<0.001 for both two-group comparisons); therefore these individuals clearly did not fit within the five-group symptom level structure derived from the DSM diagnostic categories. At the baseline visit, their average (± SD) CES-D8 score was 7.3 (± 1.5) points, with a range from 5~10. On average these individuals endorsed 3.2 (±0.41) of the eight depressive symptoms at baseline. This group is represented in Table 1 and Figures 2A and 2B (also in Figure 3 in terms of the distribution of their CES-D8 scores at each survey).
Sex differences in depressive symptoms: We found significant differences between male and female respondents in terms of all indicators of depressive symptoms. CES-D8 scores and number of items endorsed were significantly higher for women than men (collapsed across severity levels) at each survey visit. In terms of depressive symptom levels at baseline, more men (46.5%) than women (32.6%) demonstrated no depressive symptoms (NDS), but nearly identical proportions had low levels (men: 33.3%; women: 33.2%). More women than men were classified as medium-low (M:17.3%; F: 28.0%), medium-high (M: 1.8%; F: 4.3%), and high (M:1.1%;F:1.9%) levels of symptoms at baseline. In fact, the distributions of men and women falling into the five depressive symptom level groups differed significantly at every interview (all chi square > 79, all P<0.001). The distributions appeared to be stable across the four survey waves in respondents who contributed data at each of the four surveys (N=1,782; see Figures 2A and 2B).
Nonparametric means comparisons were carried out to evaluate differences in symptoms at baseline between men and women. While women overall had significantly higher levels of depressive symptoms by every measure than men did, when the data were broken down by symptom level at baseline, women in only the low symptoms group (LDS) had significantly higher numbers of items endorsed (z=-3.03, P<0.01), higher average frequency rating (z=-2.7,P<0.01), and higher CES-D8 (z=-2.7,P<0.01) than men in the same group. In the medium-low symptom level group (MLDS) at baseline, the differences between men and women in terms of average frequency ratings and CES-D8 scores approached significance (both z=-1.8, both P= 0.07). None of these values was significantly different (all P≥ 0.10) within the other depressive symptom grouping levels at baseline, nor did men and women in the sixth (a posteriori group) group differ significantly at the baseline visit (all P>0.3).
Depressive Levels over Time
The distributions of CES-D8 scores for the five symptom levels plus individuals in the sixth a posteriori (“no group”) group are presented at each survey wave in Figure 3. The plot in Figure 3 includes reference lines for two CES-D short form cutoffs: 8 and 10 points[21, 31].
Figures 3 Depressive symptom score distributions across survey waves by depressive symptom level. Box represents scores in the 25th～75th percentiles with solid line at median value; whiskers represent largest/smallest scores that are not outliers. Circles represent scores 1.5～3 box lengths from upper whisker (outliers). Reference lines at CES-D8 scores=10 and=8 are included to highlight the distinct depressive symptom levels present above and below the cutoffs
Notes: NDS: No depressive symptoms, not shown on this plot (all CES-D8 scores=0). LDS: “Low” depressive symptoms;1～2 items at any frequency. MLDS: “Medium-low” depressive symptoms; 3～7 symptoms rarely or sometimes. MDS: “Medium-high” depressive symptoms; 3～4 symptoms rated as always present. HDS: “High” depressive symptoms; 5～8 symptoms rated as always present. Individuals falling into more than one group were assigned to the higher level group. “no group”: these individuals did not meet the criteria for any of the five depressive symptom severity (“severity”) groups
At every wave, some individuals classified as MDS fell below the 10-point cutoff. Individuals classified as MLDS, which we established to correspond to “subthreshold” depression, tended to have CES-D8 scores at or very close to the cutoffs. As many as 50% of this group would have been missed in a screening effort for significant depressive symptoms if either the 10-or the 8-point cutoff had been used in any given year.
Over the ten-year observation period, 50.7% of incomplete cases and 41.0% of complete cases were classified only as NDS or LDS at each recorded visit, reflecting stability of these low levels of depressive symptoms over participation in the survey. Among all respondents, at one or more survey waves, 5.3% were characterized as HDS; CES-D8 scores at this level are uniformly high at all visits; similarly, few of the 9.4% who were characterized as MDS at one or more survey waves would have had CES-D8 scores below the 10-point cutoff at any survey wave.
DISCUSSION
The mean CES-D8 score increased over the ten year period we analyzed; although men had significantly lower scores than women at every survey wave, this increasing trend was observed for both. A sizeable proportion of this cohort was experiencing some level of depressive symptoms at every visit; however, respondents falling into the low level and no-group groups would not have been identified as “depressed” if cutoff scores had been used. Further, our results suggest that individuals who might fit a minor depressive or subthreshold depressive profile (i.e., MLDS) could be missed as much as 50% of the time when cutoff scores are used; among all persons failing to meet the cutoff score were two qualitatively different groups, the MLDS with relatively low total scores and individuals who did not fit any of the five DSM-based categories, our a posteriori sixth group. Thus, in addition to differences in perspectives between DSM criteria and self reports such as the CES-D[21], our results suggest that low sensitivity and specificity observed when using cutoff scores may also be due to qualitative differences, namely, that the DSM criteria include both number and frequency of symptoms whereas the CES-D is simply the sum of ratings.
Combining the number of symptoms endorsed and their frequency ratings to create depressive symptom levels based loosely on DSM thresholds resulted in five distinct levels of severity at each of four surveys in these elderly respondents. Without increasing the burden of reporting (i.e., using an 8-or 10-item CES-D short form), the number-and-frequency approach increases the available amount of information about depressive symptoms; this is particularly true for individuals not classified (sixth group) and MLDS whose totals fall below a cutoff score. The overlap in scores for these qualitatively different groups represents another difficulty for the use of cutoff scores in screening; individuals like these may be part of the reason that score-based characterizations and DSM-based diagnoses do not always agree[17~20].
The application of DSM diagnostic criteria and the use of the CES-D in the Taiwan Survey cohort assumes that both of these methods reflect culturally universal affective and somatic symptoms of depression; however, the combination of DSM-based criteria and CES-D symptoms that we describe is not dependent on this assumption. That is, if the CES-D is or has been administered, then the method we describe here will pertain: it is a classification scheme. The approach can be used with short or long forms of any self-reported depressive inventory (including those in extant databases) and might lead to improved sensitivity and specificity of these instruments. Because even low levels of depressive symptoms can adversely impact the health and well-being of the elderly, validation of the groupings described here with formal psychiatric evaluations is warranted.
REFERENCES
1. Kessler RC, Berglund P, Demler O, Jin R, Koretz D，et al. The epidemiology of major depressive disorder: results from the National Comorbidity Survey Replication (NCS-R). JAMA,2003,289(23):3095-3105.
2. Ruo B, Rumsfeld JS, Hlatky MA, Liu H, Browner WS, Whooley MA. Depressive symptoms and health-related quality of life: the Heart and Soul Study. JAMA, 2003，290(2): 215-221.
3. Stewart WF, Ricci JA, Chee E, Hahn SR, MorgansteinD. Cost of lost productive work time among US workers with depression. JAMA, 2003，289(23): 3135-3144.
4. Rovner BW, German PS, Brant LJ, Clark R, Burton L, Folstein MF. Depression and mortality innursing homes. JAMA，1991,265(8): 993-996.
5. Takeshita J, Masaki K, Ahmed I, Foley JD, Li YQ, et al. Are depressive symptoms a risk factor for mortality in elderly Japanese American men?: the Honolulu-Asia Aging Study. Am J Psychiatry, 2002，159(7): 1127-1132.
6. Wells KB, Stewart A, Hays RD, Burnam MA, Rogers W，et al. The functioning and well-being of depressed patients. Results from the Medical Outcomes Study. JAMA, 1989，262(7): 914-919.
7. Schulberg HC, Katon W, Simon GE & Rush AJ. Treating major depression in primary care practice. Archives of General Psychiatry, 1998，55:1121-1127.
8. Lin EH, Katon W, Von Korff M, Tang L, Williams JW Jr，et al. Effect of improving depression care on pain and functional outcomes among older adults with arthritis: a randomized controlled trial. JAMA, 2003，290(18): 2428-2429.
9. Jorge RE, Robinson RG, Arndt S and Starkstein S. Mortality and poststroke depression: a placebo-controlled trial of anti-depressants. Am J Psychiatry, 2003，160(20): 2823-1829.
10. Hybels CF, Blazer DG & Pieper CF. Toward a threshold for subthreshold depression: An analysis of correlates of depression by severity of symptoms using data from an elderly community sample. The Gerontologist, 2001，41(3)：357-365.
11. Bruce ML, Seeman TE, Merrill SS & Blazer DG. The impact of depressive symptoms on physical disability: MacArthur Studies of Successful Aging. American Journal of Public Health, 1994，84:1796-1799.
12. Penninx BW, Leveille S, Ferrucci L, van Eijk JTM, & Guralnik JM. Exploring the effect of depression on physical disability: longitudinal evidence from the Established Populations for Epidemiologic Studies of the Elderly. American Journal of Public Health, 1999，8:1346-1352.
13. Chi I. Mental health of the old-old in Hong Kong. Clinical Gerontologist, 1995，15(3): 31-44.
14. Radloff LS. The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1977，1:385-401.
15. Lyness JM, Noel TK, Cox C, King DA Conwell Y & Caine ED. Screening for depression in elderly primary care patients: a comparison of the Center for Epidemiologic Studies-Depression Scale and the Geriatric Depression Scale. Archives of Internal Medicine, 1997，157:449-454.
16. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders (4th ed. Text rev.). Washington, DC: Author, 2000.
17. Papassotiropoulos A & Heun R. Screening for depression in the elderly: A study on misclassification by screening instruments and improvement of scale performance. Progress in Neuro-Psychopharmacology & Biological Psychiatry, 1999,23:431-446.
18. Boyd JH, Weissman MM, Thompson D & Myers JK. Screening for depression in a community sample. Archives of General Psychiatry, 1982, 39:1195-1200.
19. Beekman ATF, Deeg DJH, van Limbeek J, Braam AW, de Vries MZ, & van Tilburg W.Criterion validity of the Center for Epidemiologic Studies Depression scale (CES-D): results from a community-based sample of older subjects in the Netherlands. Psychological Medicine, 1997, 27:231-235.
20. Myers JK & Weissman MM. Use of a self-report symptom scale to detect depression in a community sample. American Journal of Psychiatry, 1980, 137(9)：1081-1084.
21. Boey KW. Cross-validation of a short for m of the CES-D in Chinese elderly. International Journal of Geriatric Psychiatry, 1999,14:608-617.
22. Eaton WW, Neufeld K, Chen L-S & Cai G. A comparison of self-report and clinical diagnostic interviews for depression: Diagnostic Interview Schedule and Schedules for Clinical Assessment in Neuropsychiatry in the Baltimore Epidemiologic Catchment Area Follow-up. Archives of General Psychiatry, 2000,57:217-222.
23. Plutchik R & van Praag HM. Interconvertability of five self-report measures of depression. Psychiatry Research, 1987, 22:243-256.
24. Chang MC & Hermalin AI. Taiwan Provincial Institute of Family Planning and Population Studies Center and Institute of Gerontology, University of Michigan (1989) Survey of Health and Living Status of the Elderly in Taiwan: Questionnaire and Survey Design, Comparative Study of the Elderly in Four Asian Countries, Research Report No. 1. Ann Arbor, MI: University of Michigan, 1989.
25. Geerlings MI, Schoevers RA, Beekman ATF, Jonker C, Deeg DJH, et al. Depression and risk of cognitive decline and Alzheimer’s disease. British Journal of Psychiatry, 2000,176:568-575.
26. Foley KL, Reed PS, Mutran EJ & DeVellis RF. Measurement adequacy of the CES-D among a sample of older African Americans. Psychiatry Research, 2002,109:61-69.
27. Frerichs RR, Aneshensel CS & Clark VA. Prevalence of depression in Los Angeles county. American Journal of Epidemiology, 1981,113(6)：691-699.
28. Iwata N & Roberts RE. Age differences among Japanese on the Center for Epidemiologic Studies Depression Scale: An ethnocultural perspective on somatization. Social Science and Medicine, 1996, 43(6)：967-974.
29. Beekman ATF, Geerlings SW, Deeg DJH, Smit JH, Schoevers RS, et al. The natural history of late-life depression. Archives of General Psychiatry, 2002,59:605-611.
30. Holm S. A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 1979,16:65-70.
31. Andresen EM, Carter WB, Malmgren JA, & Patrick DL. Screening for depression in well older adults: Evaluation of a short form of the CES-D. American Journal of Preventive Medicine, 1994,10:77-84.
(Editor Jaque)(Rochelle E. Tractenberg1, Myron F. Weine)
1 Departments of Biostatistics, Biomathematics & Bioinformatics and Psychiatry, Georgetown University School of Medicine, Washington, D.C.20007，U.S.
2 Department of Psychiatry, University of Texas Southwestern Medical Center, Dallas TX,U.S.
3 Population and Health Research Center, Bureau of Health Promotion, Taichung, Taiwan
Correspondence to Rochelle E. Tractenberg, Department of Biostatistics, Biomathematics and Bioinformatics, M7202, 7 East Main Georgetown University Hospital, 3800 Reservoir Road, NW, Washington, D.C. 20007,U.S.
Tel:2024448748,Fax:2024444114
E-mail:ret7@georgetown.edu
[Abstract] Objective Even without reaching a diagnostic threshold, symptoms of depression are associated with worse function, greater sleep disturbance, and lower life satisfaction. In large-scale studies, cutoff scores are frequently used to identify “depression”. These cutoff scores tend to have suboptimal sensitivity and specificity; this can be more pronounced for non-major depressions. Here we propose and demonstrate a classification scheme combining the number of reported depression symptoms in an elderly cohort with their frequency ratings.Methods Eight questions were analyzed from a self-report depression instrument (CES-D), in Taiwan elderly (ages 65+ in 1989;N=3874) at three-year intervals from 1989～1999. Depressive symptom-levels were based on both number of symptoms and their respective frequency ratings. Total scores and symptom-level membership were traced triennially over ten years in this community cohort, and compared to cutoff scores.Results The number-frequency combination approach revealed distinct symptom-level groups with scores distributed around cutoff values. Women reported significantly more symptoms than men, but within the symptom-level groups, total scores did not differ by sex. Conclusions Discrepancies between diagnoses of major/minor depression and cutoff scores on self-reported instruments may be due, in part, to distinguishable subgroups (symptom-levels) with score distributions around the cutoff value. The number-frequency combination method does not require additional data and may improve identification of clinically relevant symptom levels with self-report instruments. Without confirmation of depression, sensitivity and specificity of our classifications could not be established. However, the approach does explain low sensitivity and specificity for cutoff scores relative to formal diagnostic criteria.
[Key words] elderly;depression;CES-D;longitudinal
INTRODUCTION
Depression is a highly prevalent disorder[1], associated with decreased quality of life[2] and loss of productive work time[3]. It is also associated with increased morbidity and mortality in many illnesses, with mortality in nursing home residents[4], and with increased mortality in elders independent of concomitant medical illness[5]. Of greatest importance is that depression is treatable[7] and that, in addition to improving function and quality of life[8], treatment of depression may have great impact on mortality, as has recently been shown in a study of post-stroke depression[9]. In the elderly, “subthreshold” levels of depression[10] have been associated with lower levels of function[11, 12] and life satisfaction[13]. Subthreshold depression is likely more frequently missed than major depression, and so might represent an important, as yet unquantified, health problem, particularly in the elderly.
When depression is not the main focus of a clinical or epidemiological study, its quantification may still be critical in the accurate evaluation of risks and trends, particularly in the elderly. Without expert evaluations, it is difficult to correctly characterize levels of depressive symptoms. The utility of cutoff scores on depressive symptom inventories, such as the Center for Epidemiologic Studies-Depression Scale (CES-D)[14], has been reported to be excellent[15], with good sensitivity and specificity in elderly as compared to a diagnosis by American Psychiatric Association’s Diagnostic and Statistics Manual (DSM)[16] criteria. However, high false positive rates[17], low positive predictive value[18, 19], and reduced sensitivity and specificity for minor depression[15] have also been reported for the CES-D when cutoff scores are used[18, 20].
Low sensitivity and specificity reflect discrepancies between DSM-based and cutoff-score-based labels of “depression”. The purpose of the CES-D and other self-report depression instruments is to measure depressive symptoms and not to provide diagnosis[17, 20, 21], yet there is a high degree of dependence on self reports in psychiatric epidemiology in particular[22] and in epidemiologic studies in general (i.e., the CES-D was developed for these contexts)[14]. Psychiatrists’ judgments about depressive symptoms (i.e., application of DSM criteria) necessarily reflect a different perspective than do self reports such as the CES-D[23]. In this study we describe a combination of elements of the DSM diagnostic approach and the CES-D symptom list to derive symptom-level categories that can be considered to reflect both qualitatively and quantitatively different types of depressive symptoms. To evaluate this method, we applied it to CES-D responses from a large cohort of elderly over four survey interviews during a ten-year period. Rather than using these categories to ‘diagnose’ respondents in this elderly cohort, the purpose of this study was to characterize levels of depressive symptoms over time and compare these results to those derived using cutoff scores.
MATERIAL AND METHODS
Subjects Individuals were sampled for the “National Survey of Health in the Elderly”, sponsored by “the Taiwan Ministry of Health” in 1988～1989. The sampling and study design are described elsewhere[24]. Briefly, a representative sample of over 4,000 persons in Taiwan over the age of 65 was recruited for the survey. This sample has the advantages of size, representativeness, and a focus on the elderly, in whom more subtle depressive symptoms may be of greatest concern[11~13]. The focus of the survey was not depression but general aspects of health and specific aspects of economic and social functioning.
Instrument These analyses focus on eight depression-related symptoms (poor appetite; doing anything is exhausting; poor/restless sleep; terrible mood/feel sad; feel lonely; feel people are unfriendly; feel anguished/depressed; no will to do anything/could not get going) that were translated from the CES-D and incorporated into the survey[24]; two other items (feel joyful; feel life is going well) were not included in our analyses. The eight negative items were rated for their frequency in the week previous to the interview (not at all=0; rarely (one day)=1;, sometimes (2～3 days)=2 or often/chronically (4～7 days)=3). These items were selected for inclusion when the survey was designed without specific reference to psychiatric literature but they overlap generally with the symptom list most frequently used as the ‘gold standard’ for establishing the diagnosis of depression (see below).
DSM-IV-TR[16], the most recent edition of the American Psychiatric Association Diagnostic and Statistics Manual, provides a list of symptoms of which five or more must be present in a patient to meet criteria for major depressive episode. These are: depressed mood or markedly diminished interest/pleasure in activities (at least one of which is required); significant unintentional weight loss or change in appetite; significant increase or decrease in sleep; restlessness or lethargy; fatigue or loss of energy; feelings of worthlessness; indecisiveness or inability to concentrate; and recurrent thoughts of death or suicide (p. 356). The presence of these symptoms must represent a change from previous function, which is usually not established in large-scale epidemiologic or survey studies.
The CES-D8 symptoms do not completely overlap with the DSM symptom list, nor are similar levels of frequency or the change from previous functional levels possible to establish; however, both DSM criteria and CES-D have been used across cultures for detecting depression[13, 19, 21, 25~29]. We combined internationally utilized CES-D items with the descriptive rigor inspired by the DSM, as described below, to characterize the level of depressive symptoms in this cohort over time.
Procedure CES-D items and DSM criteria: We established combination endorsement-frequency levels to classify respondents’ level of depressive symptoms as follows: According to the DSM-IV-TR[16], five or more symptoms must be present “nearly every day” for the two weeks prior to the assessment for an individual to meet the criteria for a major depressive episode (p. 356). The highest frequency rating possible for CES-D symptoms is “4 or more days in the past week”; therefore, individuals with 5～8 of the symptoms on the survey, all rated at the highest frequency (3), were characterized as “high depressive symptoms” (HDS). Because the symptom “felt people were unfriendly” does not appear in the DSM symptom list, individuals who met the HDS criterion with “unfriendly” as one of just five symptoms were not included as HDS but were classified at the next (lower) symptom level. Individuals classified as HDS were considered to have symptoms closest to the clinical criteria for major depressive episode.
DSM-IV-TR[16] specifies that for research purposes, a diagnosis of dysthymic disorder applies to individuals who have three or more symptoms “more days than not” (p. 775). Therefore, respondents rating 3～4 symptoms on the survey, all with the highest frequency (3), were characterized as “medium-high depressive symptoms” (MDS). Although the diagnosis of dysthymic disorder requires the presence of these symptoms for two years and we did not have that information, respondents classified as MDS were considered to have symptoms closest to the clinical (research) criteria for dysthymic disorder.
Individuals who rated 3~8 symptoms as occurring rarely or sometimes (1～2) were characterized as “medium-low depressive symptoms” (MLDS). These respondents were considered most similar to persons who might meet the criteria for subthreshold depression[10].
Individuals who rated 1～2 items at any frequency (i.e., not ‘never’) were characterized as “low depressive symptoms” (LDS), and individuals reporting no symptom were characterized as “no depressive symptoms” (NDS).
Depressive symptoms over time: Because we had no formal psychiatric evaluations for these respondents, we generally characterized depressive symptoms as the number of CES-D8 symptoms endorsed (having occurred at least one day in the previous week, 0～8) and the CES-D8 score (sum of frequency ratings, 0～24). We also determined respondents’ depressive symptom-level group membership (i.e., HDS, MDS, MLDS, LDS or NDS) at each wave; persons meeting two definitions were put into the higher-level group. Finally, we examined the distribution of CES-D8 scores for the depressive symptom levels at each wave. All statistical analyses were carried out using SPSS 11.5 (SPSS Inc. Chicago, IL). Multiple comparisons (across survey waves) were adjusted for using Holm’s method (a modified Bonferroni method[30]), and adjusted P-values less than 0.05 were considered significant.
RESULTS
Background (demographic) values: Table 1 presents the background and descriptive/demographic variables for the full cohort at the first survey (1989, “baseline”). The proportion of the sample falling into the depressive symptoms groups at baseline is also presented.
Table 1 Background/Demographics overall and by Sex at First (Baseline) Survey (1989) Mean (±SD) or percent
TP1
Notes:NDS: No depressive symptoms. LDS: “Low” depressive symptoms; 1~2 items at any frequency. MLDS: “Medium-low” depressive symptoms; 3~7 symptoms rarely or sometimes. MDS: “Medium-high” depressive symptoms; 3~4 symptoms rated as always present. HDS: “High” depressive symptoms; 5~8 symptoms rated as always present. “No group”: Individuals who did not fall into any of the five a priori symptom level groups. Individuals falling into more than one group were assigned to the higher-level group. * Individuals reporting literacy but no formal education were assigned the value 0.5 years of education to differentiate them from those reporting no formal education and no literacy
The respondents were 57.1% male, 42.9% female. Their mean± SD age was 68.1 ± 6.5, educational attainment was 3.9 ± 4.5 years (with “literate, but no formal education” assigned a value of 0.5). Women were slightly, but significantly older (68.8 ± 6.7 years) than men (67.6 ± 6.3 years) (z=-5.6, P<0.001) and had significantly less education (z=-28.5,P< 0.001). Women also tended to rate their health worse at every interview (all chi square >64, all P< 0.001) and in each year, more men than women tended to rate their heath better than, or the same as, the past year (all chi square> 14, all P≤ 0.001).
When we compared individuals with valid CES-D8 scores at all four waves (“completers”) to those missing at least one score, we found that completers were younger (65.7 ± 4.7 vs. 69.3 ± 6.9 years), had more education (4.4 ± 4.6 vs. 3.6 ± 4.5 years) and lower baseline CES-D8 scores (2.4 ± 3.8 vs. 3.6 ± 4.5) (all t-tests significant at P<0.001). Higher baseline symptom severity level was also associated with an incomplete survey record (chi square=43.6,P<0.001): 40.2% of NDS, 31.4% of LDS, 28.7% of MLDS, 31.2% of MDS and 22.2% of HDS had complete records. Thirty-two percent of both men and women had complete data (P=0.9).
Age, education and depressive symptoms: We stratified the sample by sex and explored the association between age and education and the two continuous depressive symptom variables, CES-D8 scores and number of symptoms endorsed at baseline, using Spearman’s (nonparametric) correlation. Weak positive association was observed between age and both depressive symptom variables (both rho< 0. 075,P<0.05 for men,P<0.01 for women). Similarly, weak negative association was observed between education and both depressive symptom variables (-0.13 < both rho <-0.11). Although these values reached significance (P<0.001 for men, P<0.001 for women), age and education each accounted for less than 1% of the variability in either of these depressive-symptom variables regardless of respondent sex. CES-D8 scores for men and women with valid responses (i.e., all 8 items rated) are shown at each visit in Figure 1.
Figure 1 Mean score on CES-D8 (range: 0~24) over four survey waves, by sex
Figure 1 shows that over the three follow-up surveys, average CES-D8 scores increased for both male and female respondents (a similar pattern was observed for complete cases (N=1,782, 30.1% female); data not shown). We present all available data at each visit in Figure 1 because individuals with incomplete records had significantly higher CES-D8 scores on average (see above) and so although the trends were the same for completers and non-completers, without these non-completers the distributions would be biased downward.
For both men and women, lower education was significantly associated with higher levels of depressive symptoms at baseline (both chi square >20, both P<0.001). Age was not significantly different across depressive symptom levels (men: chi square=4.5,P=0.31; women: chi square=8.9,P=0.06) although at baseline, respondents who were older tended to have higher depressive symptom levels.
Depressive symptom levels: There were valid CES-D8 responses (i.e., all 8 items rated) for 3,874 of the 4,049 participants (93.5%) in the baseline survey. Of these, 1.4% (n=54) were classified as “high” depressive symptoms, and a further 2.9% (n=109) met our criteria for medium-high depressive symptoms. Over 70% of the respondents had low (33.3%) or no (40.5%) depressive symptoms in the week preceding the 1989 survey. Figures 2A and 2B present the distributions of the symptom severity groups for male and female respondents, respectively, with valid responses at all four waves (men:N=989; women:N=793).
Figures 2 2A and 2B. Distribution of depressive symptom levels (“severity group”) over four survey waves in men (2A, N=989) and women (2B, N=793) with valid responses at all four waves
Notes: NDS: No depressive symptoms. LDS: “Low” depressive symptoms; 1~2 items at any frequency. MLDS: “Medium-low” depressive symptoms; 3~7 symptoms rarely or sometimes. MDS: “Medium-high” depressive symptoms; 3~4 symptoms rated as always present. HDS: “High” depressive symptoms; 5~8 symptoms rated as always present. “No group”: individuals not otherwise classified. Individuals falling into more than one group were assigned to the higher level group
A sixth grouping of depressive symptoms was observed that did not fit with any of the five groups we defined a priori. Of the 3,874 individuals with valid CES-D8 responses at baseline, 113 (2.9%) fell into this sixth category. Nonparametric means comparisons revealed that, at every visit, individuals in this “no group” group had significantly greater symptoms on average than the LDS group and significantly less symptoms on average than the MLDS group (symptoms measured as number of symptoms endorsed, number rated “always”, mean symptom frequency rating, and CES-D8 score; all adjusted P<0.001 for both two-group comparisons); therefore these individuals clearly did not fit within the five-group symptom level structure derived from the DSM diagnostic categories. At the baseline visit, their average (± SD) CES-D8 score was 7.3 (± 1.5) points, with a range from 5~10. On average these individuals endorsed 3.2 (±0.41) of the eight depressive symptoms at baseline. This group is represented in Table 1 and Figures 2A and 2B (also in Figure 3 in terms of the distribution of their CES-D8 scores at each survey).
Sex differences in depressive symptoms: We found significant differences between male and female respondents in terms of all indicators of depressive symptoms. CES-D8 scores and number of items endorsed were significantly higher for women than men (collapsed across severity levels) at each survey visit. In terms of depressive symptom levels at baseline, more men (46.5%) than women (32.6%) demonstrated no depressive symptoms (NDS), but nearly identical proportions had low levels (men: 33.3%; women: 33.2%). More women than men were classified as medium-low (M:17.3%; F: 28.0%), medium-high (M: 1.8%; F: 4.3%), and high (M:1.1%;F:1.9%) levels of symptoms at baseline. In fact, the distributions of men and women falling into the five depressive symptom level groups differed significantly at every interview (all chi square > 79, all P<0.001). The distributions appeared to be stable across the four survey waves in respondents who contributed data at each of the four surveys (N=1,782; see Figures 2A and 2B).
Nonparametric means comparisons were carried out to evaluate differences in symptoms at baseline between men and women. While women overall had significantly higher levels of depressive symptoms by every measure than men did, when the data were broken down by symptom level at baseline, women in only the low symptoms group (LDS) had significantly higher numbers of items endorsed (z=-3.03, P<0.01), higher average frequency rating (z=-2.7,P<0.01), and higher CES-D8 (z=-2.7,P<0.01) than men in the same group. In the medium-low symptom level group (MLDS) at baseline, the differences between men and women in terms of average frequency ratings and CES-D8 scores approached significance (both z=-1.8, both P= 0.07). None of these values was significantly different (all P≥ 0.10) within the other depressive symptom grouping levels at baseline, nor did men and women in the sixth (a posteriori group) group differ significantly at the baseline visit (all P>0.3).
Depressive Levels over Time
The distributions of CES-D8 scores for the five symptom levels plus individuals in the sixth a posteriori (“no group”) group are presented at each survey wave in Figure 3. The plot in Figure 3 includes reference lines for two CES-D short form cutoffs: 8 and 10 points[21, 31].
Figures 3 Depressive symptom score distributions across survey waves by depressive symptom level. Box represents scores in the 25th～75th percentiles with solid line at median value; whiskers represent largest/smallest scores that are not outliers. Circles represent scores 1.5～3 box lengths from upper whisker (outliers). Reference lines at CES-D8 scores=10 and=8 are included to highlight the distinct depressive symptom levels present above and below the cutoffs
Notes: NDS: No depressive symptoms, not shown on this plot (all CES-D8 scores=0). LDS: “Low” depressive symptoms;1～2 items at any frequency. MLDS: “Medium-low” depressive symptoms; 3～7 symptoms rarely or sometimes. MDS: “Medium-high” depressive symptoms; 3～4 symptoms rated as always present. HDS: “High” depressive symptoms; 5～8 symptoms rated as always present. Individuals falling into more than one group were assigned to the higher level group. “no group”: these individuals did not meet the criteria for any of the five depressive symptom severity (“severity”) groups
At every wave, some individuals classified as MDS fell below the 10-point cutoff. Individuals classified as MLDS, which we established to correspond to “subthreshold” depression, tended to have CES-D8 scores at or very close to the cutoffs. As many as 50% of this group would have been missed in a screening effort for significant depressive symptoms if either the 10-or the 8-point cutoff had been used in any given year.
Over the ten-year observation period, 50.7% of incomplete cases and 41.0% of complete cases were classified only as NDS or LDS at each recorded visit, reflecting stability of these low levels of depressive symptoms over participation in the survey. Among all respondents, at one or more survey waves, 5.3% were characterized as HDS; CES-D8 scores at this level are uniformly high at all visits; similarly, few of the 9.4% who were characterized as MDS at one or more survey waves would have had CES-D8 scores below the 10-point cutoff at any survey wave.
DISCUSSION
The mean CES-D8 score increased over the ten year period we analyzed; although men had significantly lower scores than women at every survey wave, this increasing trend was observed for both. A sizeable proportion of this cohort was experiencing some level of depressive symptoms at every visit; however, respondents falling into the low level and no-group groups would not have been identified as “depressed” if cutoff scores had been used. Further, our results suggest that individuals who might fit a minor depressive or subthreshold depressive profile (i.e., MLDS) could be missed as much as 50% of the time when cutoff scores are used; among all persons failing to meet the cutoff score were two qualitatively different groups, the MLDS with relatively low total scores and individuals who did not fit any of the five DSM-based categories, our a posteriori sixth group. Thus, in addition to differences in perspectives between DSM criteria and self reports such as the CES-D[21], our results suggest that low sensitivity and specificity observed when using cutoff scores may also be due to qualitative differences, namely, that the DSM criteria include both number and frequency of symptoms whereas the CES-D is simply the sum of ratings.
Combining the number of symptoms endorsed and their frequency ratings to create depressive symptom levels based loosely on DSM thresholds resulted in five distinct levels of severity at each of four surveys in these elderly respondents. Without increasing the burden of reporting (i.e., using an 8-or 10-item CES-D short form), the number-and-frequency approach increases the available amount of information about depressive symptoms; this is particularly true for individuals not classified (sixth group) and MLDS whose totals fall below a cutoff score. The overlap in scores for these qualitatively different groups represents another difficulty for the use of cutoff scores in screening; individuals like these may be part of the reason that score-based characterizations and DSM-based diagnoses do not always agree[17~20].
The application of DSM diagnostic criteria and the use of the CES-D in the Taiwan Survey cohort assumes that both of these methods reflect culturally universal affective and somatic symptoms of depression; however, the combination of DSM-based criteria and CES-D symptoms that we describe is not dependent on this assumption. That is, if the CES-D is or has been administered, then the method we describe here will pertain: it is a classification scheme. The approach can be used with short or long forms of any self-reported depressive inventory (including those in extant databases) and might lead to improved sensitivity and specificity of these instruments. Because even low levels of depressive symptoms can adversely impact the health and well-being of the elderly, validation of the groupings described here with formal psychiatric evaluations is warranted.
REFERENCES
1. Kessler RC, Berglund P, Demler O, Jin R, Koretz D，et al. The epidemiology of major depressive disorder: results from the National Comorbidity Survey Replication (NCS-R). JAMA,2003,289(23):3095-3105.
2. Ruo B, Rumsfeld JS, Hlatky MA, Liu H, Browner WS, Whooley MA. Depressive symptoms and health-related quality of life: the Heart and Soul Study. JAMA, 2003，290(2): 215-221.
3. Stewart WF, Ricci JA, Chee E, Hahn SR, MorgansteinD. Cost of lost productive work time among US workers with depression. JAMA, 2003，289(23): 3135-3144.
4. Rovner BW, German PS, Brant LJ, Clark R, Burton L, Folstein MF. Depression and mortality innursing homes. JAMA，1991,265(8): 993-996.
5. Takeshita J, Masaki K, Ahmed I, Foley JD, Li YQ, et al. Are depressive symptoms a risk factor for mortality in elderly Japanese American men?: the Honolulu-Asia Aging Study. Am J Psychiatry, 2002，159(7): 1127-1132.
6. Wells KB, Stewart A, Hays RD, Burnam MA, Rogers W，et al. The functioning and well-being of depressed patients. Results from the Medical Outcomes Study. JAMA, 1989，262(7): 914-919.
7. Schulberg HC, Katon W, Simon GE & Rush AJ. Treating major depression in primary care practice. Archives of General Psychiatry, 1998，55:1121-1127.
8. Lin EH, Katon W, Von Korff M, Tang L, Williams JW Jr，et al. Effect of improving depression care on pain and functional outcomes among older adults with arthritis: a randomized controlled trial. JAMA, 2003，290(18): 2428-2429.
9. Jorge RE, Robinson RG, Arndt S and Starkstein S. Mortality and poststroke depression: a placebo-controlled trial of anti-depressants. Am J Psychiatry, 2003，160(20): 2823-1829.
10. Hybels CF, Blazer DG & Pieper CF. Toward a threshold for subthreshold depression: An analysis of correlates of depression by severity of symptoms using data from an elderly community sample. The Gerontologist, 2001，41(3)：357-365.
11. Bruce ML, Seeman TE, Merrill SS & Blazer DG. The impact of depressive symptoms on physical disability: MacArthur Studies of Successful Aging. American Journal of Public Health, 1994，84:1796-1799.
12. Penninx BW, Leveille S, Ferrucci L, van Eijk JTM, & Guralnik JM. Exploring the effect of depression on physical disability: longitudinal evidence from the Established Populations for Epidemiologic Studies of the Elderly. American Journal of Public Health, 1999，8:1346-1352.
13. Chi I. Mental health of the old-old in Hong Kong. Clinical Gerontologist, 1995，15(3): 31-44.
14. Radloff LS. The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1977，1:385-401.
15. Lyness JM, Noel TK, Cox C, King DA Conwell Y & Caine ED. Screening for depression in elderly primary care patients: a comparison of the Center for Epidemiologic Studies-Depression Scale and the Geriatric Depression Scale. Archives of Internal Medicine, 1997，157:449-454.
16. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders (4th ed. Text rev.). Washington, DC: Author, 2000.
17. Papassotiropoulos A & Heun R. Screening for depression in the elderly: A study on misclassification by screening instruments and improvement of scale performance. Progress in Neuro-Psychopharmacology & Biological Psychiatry, 1999,23:431-446.
18. Boyd JH, Weissman MM, Thompson D & Myers JK. Screening for depression in a community sample. Archives of General Psychiatry, 1982, 39:1195-1200.
19. Beekman ATF, Deeg DJH, van Limbeek J, Braam AW, de Vries MZ, & van Tilburg W.Criterion validity of the Center for Epidemiologic Studies Depression scale (CES-D): results from a community-based sample of older subjects in the Netherlands. Psychological Medicine, 1997, 27:231-235.
20. Myers JK & Weissman MM. Use of a self-report symptom scale to detect depression in a community sample. American Journal of Psychiatry, 1980, 137(9)：1081-1084.
21. Boey KW. Cross-validation of a short for m of the CES-D in Chinese elderly. International Journal of Geriatric Psychiatry, 1999,14:608-617.
22. Eaton WW, Neufeld K, Chen L-S & Cai G. A comparison of self-report and clinical diagnostic interviews for depression: Diagnostic Interview Schedule and Schedules for Clinical Assessment in Neuropsychiatry in the Baltimore Epidemiologic Catchment Area Follow-up. Archives of General Psychiatry, 2000,57:217-222.
23. Plutchik R & van Praag HM. Interconvertability of five self-report measures of depression. Psychiatry Research, 1987, 22:243-256.
24. Chang MC & Hermalin AI. Taiwan Provincial Institute of Family Planning and Population Studies Center and Institute of Gerontology, University of Michigan (1989) Survey of Health and Living Status of the Elderly in Taiwan: Questionnaire and Survey Design, Comparative Study of the Elderly in Four Asian Countries, Research Report No. 1. Ann Arbor, MI: University of Michigan, 1989.
25. Geerlings MI, Schoevers RA, Beekman ATF, Jonker C, Deeg DJH, et al. Depression and risk of cognitive decline and Alzheimer’s disease. British Journal of Psychiatry, 2000,176:568-575.
26. Foley KL, Reed PS, Mutran EJ & DeVellis RF. Measurement adequacy of the CES-D among a sample of older African Americans. Psychiatry Research, 2002,109:61-69.
27. Frerichs RR, Aneshensel CS & Clark VA. Prevalence of depression in Los Angeles county. American Journal of Epidemiology, 1981,113(6)：691-699.
28. Iwata N & Roberts RE. Age differences among Japanese on the Center for Epidemiologic Studies Depression Scale: An ethnocultural perspective on somatization. Social Science and Medicine, 1996, 43(6)：967-974.
29. Beekman ATF, Geerlings SW, Deeg DJH, Smit JH, Schoevers RS, et al. The natural history of late-life depression. Archives of General Psychiatry, 2002,59:605-611.
30. Holm S. A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 1979,16:65-70.
31. Andresen EM, Carter WB, Malmgren JA, & Patrick DL. Screening for depression in well older adults: Evaluation of a short form of the CES-D. American Journal of Preventive Medicine, 1994,10:77-84.
(Editor Jaque)(Rochelle E. Tractenberg1, Myron F. Weine)