Differentiating Symptoms of Anxiety and ... - Semantic Scholar

Cogn itive Therapy and Research, Vol. 23, No. 1, 1999, pp. 53-74

Differentiatin g Sym ptom s of An xiety an d Depression in Older Adu lts: Distinct Cogn itive an d Affective Profiles? Ann e M. Sh ap iro,1,2 Joh n E. Roberts ,1 an d J. Gayle Beck1

Cogn itive an d affective dim ension s of sym ptom s of an xiety an d depression were exam in ed in a sam ple of 283 com m un ity-dwellin g older adu lts (ran gin g in age from 65 to 93 years). A prin cipal-axis factor analysis with varim ax rotation condu cted on the Cognition Checklist (CCL) revealed a factor structure different than that fou nd in you nger adu lts. Th ree factors emerged (Anxiou s, Social Loss, an d Negative Self-Evaluation/Worthlessness Cognitions) an d, in gen eral, these cognition s were not specifically related to anxiou s an d depressive sym ptom s. Instead, worthlessness cognition s were robu stly associated with both an xious an d depressive sym ptom s, inclu ding varian ce that was uniqu e to each. In terms of affective dim ension s, factor analyses revealed that on ly anxiety-related item s loaded on the negative affectivity subscale. Consequen tly, negative affectivity was stron gly related to varian ce that was uniqu e to an xious sym ptom s, but was only weakly related to varian ce that was uniqu e to depressive sym ptom s. On the oth er han d, po sitive affectivity was on ly weakly asso cia ted with bo th form s of sym ptom atolo gy. Fin din gs are d iscu ssed in term s of the cogn itive an d affective distinctions between older an d you nger adulthood. KEY WORDS: anxiety; older adults; cognition; depression; affect.

INTRODUCTION Although there is conside rable ove rlap be tween depression and anxie ty (Clark, 1989; Dobson, 1985; Stavrakaki & Vargo, 1986) , a numbe r of studie s have de monstrated that the se conditions can be diffe re ntiate d in terms of their cognitive and affe ctive profile s. Beck’s cognitive content-spe cificity hypothe sis state s that each type of e motional distre ss is accompanie d by its own unique cognitive products (i.e., automatic thoughts, and involuntary and re petitive image s; Be ck, 1976) . More specifically, de pre ssion is thought to be associate d with thoughts involving personal loss or failure and negative attitude s toward the past and future (Beck, Brown, Ste er, Eidelson, & Riskind, 1987; Beck, Rush, Shaw, & Emery, 1979) , whe re as anxi1

State University of New York at Buffalo, Buffalo, New York 14260. Address all correspondence to Anne M. Shapiro, De partment of Psychology, Park Hall, State University of New York at Buffalo, Buffalo, New York 14260; e -mail: ams8@ acsu.buffalo.edu.

2

53 0147-5916/99/0200-0053$16.00/0

Ó

1999 Plenum Publishing Corporation

54

Shap iro, Roberts , an d Beck

ety is thought to be characte rized by future -oriente d cognitions containing the mes of dange r (Be ck & Emery, 1985) . Furthe r, anxious pe ople are thought to misinte rpre t e xpe rie nce s as involving e ithe r a physical or psychosocial threat and to overestimate the probability and intensity of anticipate d harm (Beck e t al., 1987). To inve stigate diffe re nce s in the automatic thoughts characte ristic of anxie ty and de pre ssion, Be ck e t al. (1987) develope d the Cognition Che cklist (CCL). Results of factor analyse s sugge ste d that the CCL is compose d of two primary dimensions re flecting depressive and anxious cognitions (Beck e t al., 1987; Stee r, Be ck, Clark, & Beck, 1994) . The Depre ssive Cognition Subscale (CCL -D) is compose d of 14 items re flecting negative thoughts about one ’s se lf, past expe rience s, and future e xpe ctations, whe re as the Anxious Cognition Subscale (CCL -A) is compose d of 12 ite ms represe nting thoughts about pe rsonal and physical dange r. In support of the cognitive content-spe cificity hypothe sis, outpatie nts diagnose d according to the Diagnostic and Statistical Man ual of Men tal Disorders (3rd ed.) (DSM-III; American Psychiatric Association [APA], 1980) and DSM-III-R (APA, 1987) with anxie ty disorde rs, had highe r mean CCL -A score s than patie nts diagnose d with de pre ssion. In contrast, de pre ssed patie nts had highe r CCL -D score s than anxie ty patie nts (Beck e t al. 1987; Clark, Be ck, & Ste wart, 1990; Ste er et al., 1994) . Similarly, Clark, Be ck, and Brown (1989) found, in an outpatie nt sample , that anxious cognitions were unique ly relate d to highe r le ve ls of anxious symptoms (controlling for le ve ls of de pre ssive cognitions and symptom s) and that de pre ssive cogniti ons we re unique ly relate d to highe r le vels of de pre ssive symptoms (controlling for le vels of anxious cognitions and symptom s). Furthe r, depressive cognitions assessed with relate d cognitive measure s have be en shown to be more closely associate d with depre ssive mood than with othe r mood state s (Clark, 1986; Harrell, Chamble ss, & Calhoun, 1981; Thorpe , Barne s, Hunte r, & Hines, 1983) . Finally, Ingram, Kendall, Smith, Donne ll, and Ronan (1987) found that de pre sse d stude nts re porte d significantly more negative se lf-re fe re nt automatic thoughts than test-anxious and nondepressed/nonanxious stude nts, whe re as test-anxious stude nts reporte d significantly more automatic anxious (i.e., task-irre le vant) cognitions than depressed and nondepressed/nonanxious stude nts. In contrast to the cognitive content-spe cificity hypothe sis, Watson, Clark, and Carey (1988) posite d that dime nsions of affect are useful in diffe rentiating symptoms of depression and anxie ty. Watson and colle ague s found se lf-reporte d mood to be characte rize d by two dominant dime nsions: positive and ne gative affe ct (Te lle ge n, 1985; Watson, Clark, & Tellege n, 1984; Watson & Kendall, 1989; Watson & Te lle ge n, 1985) . Positive affe ctivity (PA) was defined as the “ exte nt to which a person avows a zest for life ” (Watson & Te lle ge n, 1985, p. 221) or the “ le ve l of pleasurable engage ment with the e nvironme nt” (Watson & Kendall, 1989, p. 10) . High PA is associate d with enthusiasm, joy, high e nergy, mental inte rest and ale rtness, and de te rmination, whe re as low PA refle cts fatigue and sadne ss (Watson & Kendall, 1989) . O n the othe r hand, negative affectivity (NA) was de fine d as the “ e xtent to which a person re ports fe e ling upse t or unple asantly arouse d ” (Watson & Te lle ge n, 1985, p. 221) . High NA include s such fe elings as nervousne ss, ange r, guilt, fear, and the like , whe re as low NA is associate d with fee lings of calmne ss and re laxation (Watson & Kendall, 1989). Watson, Clark, and Tellegen (1988) de-

Sym p tom s of An xiety an d Depress ion

55

velope d the Positive and Negative Affect Sche dule (PANAS) to serve as a measure of these two mood factors. Watson and colle ague s have hypothe sized that anxie ty e sse ntially is a state of NA, without a substantial PA compone nt, whe re as depression is a mixe d state of high NA and low PA. Thus, they have propose d a two-factor affectivity mode l claiming that (1) NA is a common characte ristic of both anxie ty and de pre ssion and, the re fore, is a major factor in producing the strong correlation betwee n them; and (2) low PA is unique to depression and conse que ntly may be important in the diffe rentiation of anxie ty and depression (Watson, Clark, & Care y, 1988) . Consiste nt with this model, a numbe r of studie s e mploying clinical and nonclinical sample s found that both anxie ty and de pre ssion ratings were correlate d with or had strong positive factor loadings on NA, but on ly de pre ssion ratings were corre lated with or had mode rate (ne gative ) factor loadings on PA (e.g., Te lle ge n, 1985; Watson, Clark, & Care y, 1988; Watson & Kendall, 1989). Se ve ral studie s have recently te ste d the validity of the cognitive and affe ctivity mode ls in conjunction with one anothe r. For e xample , Clark, Ste er, and Beck ( 1994) found support for both mode ls, utilizing both clinical and nonclinical sample s. Analyse s of re sponse s to the Be ck Anxie ty and Be ck Depre ssion Inventorie s yie lded two correlate d factors: Depre ssion and Anxie ty. As predicted by Watson ’s affe ctivity mode l, a large ge ne ral distre ss or NA factor was found to unde rlie the re lationship betwee n the se two first-orde r factors. Afte r controlling for NA, specific de pre ssion and anxie ty dime nsions e merge d, including depressive symptom s indicative of low PA (e.g., sadne ss and social withdrawal). Also, as hypothe sized by Be ck’s cognitive content-spe cificity model, cognitions de aling with personal loss and failure loade d significantly on the depression factor, but not on the anxie ty factor. Stee r, Clark, Be ck, and Ranie ri (1995) re plicate d the se findings in an outpatie nt sample . Similarly, Jolly, Dyck, Krame r, and Wherry (1994) found support for both cognitive and affe ctivity models among psychiatric outpatie nts. NA was associate d with both anxious and depressive symptom s, whe re as PA was found to be more strongly relate d to depressive symptoms than to anxious symptoms. In accord with the cognitive mode l, CCL -D scores corre late d more strongly with depressive symptoms than anxious symptoms. However, Jolly e t al. found that the strength of the correlations betwee n CCL -A score s and anxious and depressive symptom s were not significantly differe nt (see Ambrose & Rhole s, 1993, for similar re sults) . O verall, a numbe r of studie s have found support for the cognitive conte ntspe cificity hypothe sis and the affectivity mode l in differentiating symptoms of anxiety and de pre ssion. Howe ve r, no study to date has e xplore d the se mode ls in olde r adults. Because there is conside rable evide nce that late -life de pre ssion (and perhaps anxie ty) are phe nomenologically different from those occurring at othe r stage s of the life cycle (Blazer, George , & Lande rman, 1986; Gurian & Mine r, 1991), the ge ne ralizability of the cognitive and affectivity mode ls to olde r adults ne eds to be examine d. In addition to diffe re nce s in the pre vale nce of depressive and anxious disorde rs (Regier et al., 1988) and affe ctive states (Lawton, Kle ban, & Dean, 1993) , nume rous studie s have found that e lderly depressed individuals display somewhat differe nt symptom profile s than their younge r counte rparts.

56


A numbe r of studie s have found that de pre sse d olde r adults report fewer affe ctive symptoms. For e xample , dysphoric mood (Arean & Miranda, 1992; Craig & Van Natta, 1979; Gallo, Anthony, & Muthe n, 1994; Goldfarb, 1974) , self-blame (Muse tti et al., 1989), and guilt (Gallo et al., 1994) all have be en found to be less pervasive in olde r adults with depression than among younge r depressed persons. Instead, fe e lings of e mptine ss, impulse s to cry, and irritability see m to be more characte ristic of late -life depression (Blaze r, 1993) . Fee lings of he lple ssne ss and hope lessne ss about the future also have be en found to be more central to late -life depression (Blaze r, 1993) , but othe r studie s have found that suicidal ide ation occurs le ss in depre sse d olde r adults (Blaze r, Bachar, & Hughe s, 1987; Musetti e t al., 1989) . Compare d to the ir younge r counte rparts, de pre sse d olde r adults also have be en found to expe rience conside rably more somatic symptoms, including weight or appetite loss (Blazer e t al., 1987; Brodaty e t al., 1991; Muse tti et al., 1989) , constipation (Blazer et al., 1986; Blazer et al., 1987) , fatigue (Blazer et al., 1986; Gallo et al., 1994) , abdominal and back pain (Blaze r et al., 1986) , and insomnia (Gallo et al., 1994) . Studie s also have found support for both incre ase d psychomotor retardation (Brodaty et al., 1991; Gallo et al., 1994; Muse tti et al., 1989) an d psychomotor agitation (Brodaty e t al., 1991) in depressed olde r adults. Be cause somatic symptoms such as the se se em like ly to be a result of physical illne ss associate d with aging, Blaze r et al. ( 1986) note d that nonde pre sse d olde r adults were no more like ly to complain of the se symptoms than younge r adults (except for possibly constipation and insomnia) . Where as the re has bee n conside rable re se arch on symptom atology in late -life de pression, rese arch in a comple mentary are a in late -life anxie ty has be en ve ry sparse . It has bee n found that, of the anxie ty disorde rs, olde r adults most often suffe r from simple phobias, agoraphobia, and generalize d anxie ty disorde r (Blaze r, George , & Hughe s, 1991; Weissman et al., 1985) . Howe ve r, the re is little agre ement about what a classic prese ntation of anxie ty in the elde rly looks like (Gurian & Mine r, 1991; Shamoian, 1991) . Salzman (1982) , for example , claime d that anxie ty in the e lderly, especially anxie ty ove r declining health and pe rsonal loss, may be expe rience d as “ cognitive appre hension, be havioral agitation, or somatic symptoms with hypochondriacal compone nts ” (p. 71) . Others (e .g., Busse, 1975; Lade r, 1982) have emphasize d somatic symptom s, stating that olde r adults tend to become obsessed with the physical symptoms of anxie ty and ultimate ly may e xpe rience hypochondriacal symptom s. Still othe rs (e.g., Pfeiffer, 1979) have claime d that anxie ty in the elde rly is e xpre sse d more dire ctly as “ ove rt fear, panic, worry or be wilde rment, and without the intricate conve rsion mechanism customarily see n in younge r persons ” (Pfe iffe r, 1979, p. 28) . Recently, Be ck, Stanle y, and Ze bb ( 1996) found that late -life ge neralize d anxie ty disorde r is characte rize d by e levate d levels of anxiety, worry, social fears, and de pre ssion. Be cause these late -life symptom profile s have be en found to differ from those occurring at e arlie r stage s of life, it cannot be assumed that the cognitive conte ntspe cificity and affe ctivity mode ls can be e xte nde d to olde r adults. The curre nt study atte mpte d to de te rmine whe ther cognitions and affective state s are use ful in diffe rentiating symptoms of anxie ty and depression in olde r adults. First, assuming


57

the se mode ls can be extende d to olde r adults, it was pre dicte d that a factor analysis of the re sponse s of elde rly participants on the Cognition Checklist (CCL) would yield a two-factor structure similar to that found by Beck e t al. ( 1987) and Ste er et al. (1994) (i.e., anxious and de pre ssive cognitions) . Like wise, it was pre dicted that a factor analysis of the Positive and Negative Affect Schedule (PANAS) would yield a two-factor structure that re pre sents the two major mood dimensions: PA and NA (Watson, Clark, & Tellegen, 1988) . Second, consiste nt with the cognitive conte nt-spe cificity hypothe sis, it was e xpe cted that anxious cognitions would be more strongly pre dictive of anxie ty than depressive cognitions, whereas de pre ssive cognitions would be more strongly predictive of de pre ssion than anxious cognitions. Consiste nt with the affe ctivity mode l, it was e xpe cted that ne gative affe ctivity would be pre dictive of symptoms of both anxie ty and de pre ssion, whe re as positive affe ctivity would only be predictive of symptoms of de pre ssion.

METHOD Particip an ts an d Procedu re Participants were 283 adults 65 years of age and olde r (M = 73.99, SD = 5.76, range = 65 to 93) , who were re cruite d from se nior citize ns’ cente rs, church groups, and othe r similar social organizations. All participants were volunte ers who completed measure s in anonymity. Participants consiste d of 212 wome n ( 75% ) and 67 men (24% ), although four individuals faile d to report ge nde r. Eighty-e ight percent were Caucasian, 4% were African American, 0.7% were American Indian, 0.4% were Asian American, 0.4% were Hispanic, and 3% were of anothe r e thnicity. Nine individuals faile d to re port ethnicity. In te rms of the highe st le ve l of e ducation re ache d, 6% re ache d the e le me ntary le ve l, 14% comple te d at le ast some high school, 46% graduate d from high school, 16% comple te d some colle ge, 7% graduated from colle ge , and 7% comple ted at le ast some graduate work. Nine individuals faile d to re port le ve l of e ducation.

Measures Cognition Checklist (CCL) The CCL (Be ck e t al., 1987) is a 26-ite m se lf-re port measure consisting of two subscale s: the Depressive Cognition subscale (CCL -D) and the Anxious Cognition subscale (CCL -A), de signe d to measure the freque ncy of de pre ssive and anxious cognitions, re spe ctively. Score s on e ach of the CCL subscale s have bee n found to corre late significantly with clinician ratings of de pre ssion and anxie ty in young/middle-age adults and to discriminate de pre sse d young/middle -age adults from same age anxious adults (Beck et al., 1987; Ste er et al., 1994) . The validity of the CCL has ne ve r be en e xamine d in olde r adults.

58


Positive an d Negative Affect Schedu le (PANAS) The PANAS (Watson, Clark, & Tellegen, 1988) is a 20-ite m measure of affect consisting of two 10-ite m scales: PA and NA. Watson, Clark, and Tellegen ( 1988) demonstrate d high internal consiste ncy (for PA, alpha = .88; for NA, alpha = .85) and e xcelle nt conve rge nt and discriminant validity whe n used with colle ge stude nts. Howe ver, the consiste ncy and validity of this scale when used with olde r adults has ye t to be tested.

Beck Depression Inventory (BDI) The BDI (Be ck, Ste er, & Garbin, 1988) is a 21-ite m self-report measure develope d to asse ss the se verity of depressive symptoms. The BDI has bee n found to posse ss high inte rnal consiste ncy (alpha = .86) as well as high conve rgent and discriminant validity in young/middle -age adults (Be ck, Stee r, & Garbin, 1988). It also has bee n found to have high se nsitivity and specificity (Olin, Schne ider, Eaton, Z emansky, & Pollock, 1992) and high internal consiste ncy in olde r adult sample s (alpha = .91; Gallaghe r, Nie s, & Thompson, 1982) . G eriatric Depression Scale — Sh ort Form (G DS-SF) The GDS-SF (She ikh & Yesavage , 1986) is a self-report measure comprise d of 15 items from the original 30-ite m GDS (Ye savage e t al., 1983) . The GDS was spe cifically develope d for use in the e lde rly by excluding questions re late d to somatic complaints. Although depressed olde r adults expe rience somatic symptoms, as pre viously discusse d, the de gree to which these symptoms reflect de pre ssion is debatable because of the incre ase d rate of medical illne ss in this population. Inste ad, this instrume nt emphasize s mood, cognitive , and behavioral symptoms (e .g., “ Do you fee l that your life is e mpty? ” and “ Do you often fe el he lple ss? ” ). Also, the yes/no format of the GDS provide s a simple r task for olde r adults. The GDS-SF has bee n found to have a high sensitivity and specificity in e lde rly sample s when a cutoff score of 7 is employe d (Le she r & Berryhill, 1994) .

Beck Anxiety Inventory (BAI) The BAI (Be ck, Epstein, Brown, & Stee r, 1988) is a 21-ite m se lf-re port inve ntory de signe d to asse ss the severity of anxie ty symptoms. Fourte en items assess somatic symptoms, whereas se ve n asse ss cognitive and affe ctive symptoms of anxie ty. Initial studie s have indicate d that the BAI has high internal consiste ncy (alpha = .92) and ade quate conve rge nt and discriminant validity (Beck, Epstein et al., 1988) . Preliminary work has found the BAI to have high inte rnal consiste ncy (alpha = .87) and acceptable temporal stability (.78) when use d in olde r adults (Morin, Stone , Ling, & Trinkle , 1994) .


59

State-Trait Anxiety Inventory (STAI) The STAI (Spie lbe rger, Gorsuch, & Lushe ne, 1970) is a 40-ite m inve ntory consisting of two 20-ite m subscale s that measure state (STAI-S) and trait (STAI-T) anxie ty. Only the State subscale was include d in this study because curre nt e motional distress was of primary inte re st. The State subscale has be en found to be inte rnally consiste nt and valid when use d with both younge r and olde r adults (Patte rson, O ’Sullivan, & Spie lbe rger, 1980) .

RESULTS Factor An alys is of the Cogn ition Checklist A confirmatory factor analysis with LISREL 8.14 (J öreskog & S örbom, 1996) was conducte d to test the two-factor model sugge sted by Beck e t al. ( 1987) . Only participants who had comple te data on the CCL were include d in the se analyse s (n = 264) . This mode l provide d an inade quate fit to the data, c 2( 298, N = 264) = 1122.88, p < .001, c 2 /df ratio = 3.77; Goodne ss-of-Fit Inde x (GFI) = .73; adjusted GFI (AGFI) = .68. Conse quently, several differe nt exploratory factor analyses were conducte d on the CCL, including principal-comp one nts and alpha factor analyse s, using both oblique and orthogonal rotations. Although the factor loadings and eigenvalue s found in all approache s were similar in magnitude , the principalaxis factor analysis with varimax rotation was chose n be cause confirmatory factor analysis indicate d that it provide d the be st fit (see be low). The principal-axis factor analysis yielde d six factors with eigenvalue s gre ater than 1. Howe ver, a scree te st sugge ste d a three -factor solution. In addition, Factors 4 through 6 were uninte rpre table and were compose d of fewer than three ite ms. Conse que ntly, this analysis was re run se lecting for thre e factors. The thre e-factor solution accounte d for 49.5% of the total variance in CCL score s. Factor 1 accounte d for 33.6% of the variance and was compose d e xclusive ly of ite ms from Be ck e t al.’s ( 1987) original CCL -A subscale (see Table I). Factors 2 and 3 were compose d of items from Beck et al.’s (1987) original CCL -D subscale (e xce pt one CCL -A item that loade d on the third factor) , and accounte d for 9.9% and 6.0% of the variance , re spe ctively. Upon e xamination, it was determine d that Factor 1 consiste d of items pertaining to cognitions about anxie ty over health issue s. The re maining factors represente d two facets of de pre ssive thinking: cognitions concerning loss of social role and negative self-evaluation/worthle ssness cognitions. Subscale s were create d base d on this factor structure (CCL -AH, CCL -S, and CCL -W, re spe ctively) . In order to maximize the distinctive ne ss of the scales, inclusion was re stricted to items with a primary loading of at least .50 on a give n factor; ite ms with a secondary loading of .30 or greate r on any othe r factor were omitted. The final scales consiste d of 5, 6, and 4 ite ms, respective ly. Confirmatory factor analysis indicate d that this re vised three -factor solution provide d an ade quate fit to the data, c 2 (87, N = 264) = 241.55, p < .001, c 2/df ratio = 2.78; GFI = .89, AGFI = .85. Whe n three pairs of error

60

Shap iro, Roberts , an d Beck Table I. V arimax-Rotated Factor Loadings for Cognition Checklist (CCL) Items a CCL item

10. 11. 12. 13. 14. 15. 16. 17. 23. 25. 3. 4. 5. 6. 7. 8. 9. 18. 19. 20. 21. 22. 26.

What if I get sick and become an invalid? (A) Something might be happening that will ruin my appe arance. (A) I am going to be injured. (A) What if no one reaches me in time to help? (A) I ’m going to have an accident. (A) I might be trapped. (A) I am not a healthy pe rson. (A) The re ’s some thing ve ry wrong with me . (A) I ’m going to have a he art attack. (A) Something will happen to some one I care about. (A) Pe ople don’t re spect me anymore. (D) No one care s whether I live or die. (D) I ’m worse off than they are. (D) I don’t dese rve to be loved. (D) I ’ve lost the only friends I’ve had. (D) I ’m not worthy of people ’s atte ntion or affection. (D) The re ’s no one left to help me. (D) Life isn ’t worth living. (D) I ’m worthless. (D) I have become physically unattractive. (D) I will neve r overcome my problems. (D) Something awful is going to happen. (A) Nothing ever works out for me anymore . (D)

Factor 1

Factor 2

Factor 3

.60 .53 .62 .61 .77 .66 .48 .47 .53 .54

.30

.33 .33 .44 .31 .60 .67 .50 .65 .57 .76 .72

.43 .37

.34

.56 .64 .57 .71 .60 .52

a

Factor loadings of less than .30 were omitted. (A) = CCL Anxiety subscale, (D) = CCL De pre ssion subscale (Beck et al., 1987) .

te rms (Ite ms 10 and 13, 7 and 9, and 6 and 8) were allowe d to correlate , as sugge sted by the modification indice s, confirmatory factor analysis indicate d an e ven bette r fit, c 2 (84, N = 264) = 180.18, p < .001, c 2/df ratio = 2.15; GFI = .92, AGFI = .88. Scale items then were weighte d base d on their factor loadings. Factor An alys is of the Positive an d Negative Affect Schedu le A confirmatory factor analysis was also conducte d to te st the two-factor mode l sugge sted by Watson, Clark, and Tellegen (1988) . O nly participants who had complete data on the PANAS were include d in these analyse s (n = 243) . This mode l provide d an inade quate fit to the data, c 2 (169, N = 243) = 428.58, p < .001, c 2/df ratio = 2.54; GFI = .85; AGFI = .81. Conseque ntly, similar to the analysis of the CCL, a principal-axis factor analysis with varimax rotation was conducte d on the PANAS. The principal-axis factor analysis yie lde d four factors with e ige nvalue s gre ate r than 1. Howeve r, a scree te st sugge ste d a two-factor solution and Factors 3 and 4 were compose d of fe wer than three items. Conse que ntly, this analysis was re run se le cting for two factors. The two-factor solution accounte d for 44.9% of the total variance in PANAS score s. Factor 1 accounte d for 27.6% of the variance and was compose d e xclusive ly of ite ms re pre se nting positive affe ctivity. Factor 2 accounte d


61

for an additional 17.3% of the variance and was compose d e xclusive ly of ite ms re prese nting ne gative affe ctivity. The ite m composition of these factors was ide ntical to that found by Watson, Clark, & Te lle ge n (1988) . Howe ver, after e mploying our strict inclusion criteria to create the subscale s (i.e ., a primary loading of at le ast .50 with no se condary loading of .30 or greate r), thre e items ( “ guilty, ” “ hostile,” and “ ashame d ” ) were droppe d from the NA scale and one ite m ( “ e xcite d ” ) was droppe d from the PA scale . Confirm atory factor analysis indicate d that this re vise d factor solution provide d an ade quate fit to the data, c 2 (103, N = 243) = 241.68, p < .001, c 2/df ratio = 2.35; GFI = .89, AGFI = .85. Howe ve r, the e rror term of one ite m ( “ upse t” ) was found to correlate with four othe r e rror te rms, and was the re fore omitted from the scale. Like wise, the modification indice s sugge ste d that one pair of e rror te rms ( ‘‘scared ’’ and ‘‘afraid ’’) be allowe d to corre late . The final NA scale , Negative Affe ct — Revise d ( NA-R), consiste d of six ite ms ( “ nervous, ” “ jittery,” “ afraid, ” “ scared,” “ distre ssed,” and “ irritable ” ) and the final PA scale , Positive Affect—Revise d (PA-R), was compose d of nine ite ms ( “ inte re ste d ” “ strong, ” “ e nthusiastic, ” “ proud, ” “ alert,” “ inspire d,” “ de te rmine d,” “ atte ntive ,” and “ active ” ). Confirmatory factor analysis indicate d that this re vise d factor solution provide d a bette r fit to the data, c 2(88, N = 243) = 143.80, p < .001, c 2/df ratio = 1.63; GFI = .93, AGFI = .90. Scale ite ms were weighte d base d on their factor loadings. Descrip tive Statistics Univariate analyse s of variance were conducte d on the e ntire sample to assess the influe nce of demographic factors (including age , ge nde r, e thnicity, and e ducation) on our measure s of cognition, affect, and symptom atology. Females (M = 9.8) scored significantly highe r than did male s (M = 8.3) on the NA-R scale, t(164.7) = 3.60, p < .001, separate variance te st. Also, e ducation was corre lated with PA-R, r = .24, p < .001, and STAI-S, r = ¯ .13, p < .05. Individuals with highe r e ducation te nde d to report highe r le vels of positive affectivity and lowe r levels of anxie ty. Age was also corre late d with PA-R, r = ¯ .18, p < .01, indicating that olde r individuals tende d to re port lowe r leve ls of PA. No othe r demographic variable was significantly relate d to pre dictor or outcome variable s. Sp ecificity Analys es O nly participants who had comple te data on all of the que stionnaire s were include d in the subse quent analyse s (n = 189). Total means, standard de viations, and alpha coe fficients for each of the measure s are presented in Table II. Pe arson corre lations were used to examine the re lations among symptom s of anxie ty and de pre ssion and CCL and PANAS score s. Consiste nt with re se arch on younge r adults, measure s of de pre ssive symptom s were mode rately correlate d with measure s of anxious symptoms (rs range d from .47 to .57; se e Table II). There were also mode rate intercorre lations among the three cognition measure s (rs range d from .36 to .44) . The corre lation betwee n the PA-R and NA-R scales was low (r = ¯ .25)

.15

.15

.17

.27

.20

¯ .26

.34 d

.41 d

.32 d

.37 d

.33 d

PA-R

NA-R

BDI

GDS-SF

BAI

STAI-S

.54d

.53d

.52d

.46d

b

b

d

c

d

.39d

¯ .28

—

b

b

d

CCL-W

¯ .45

¯ .23

¯ .36

¯ .33

¯ .25

—

d

c

d

d

d

PA-R

.59d

.62d

.40d

.54d

—

NA-R

.56d

.57d

.64d

—

BDI

.52 d

.47 d

—

GDS-SF

.56 d

—

BAI

—

STAI-S

33.52

7.12

1.82

6.14

9.25

31.63

0.80

1.48

2.39

M

a

10.38

7.97

2.47

4.89

3.48

7.08

1.42

2.61

2.31

SD

.90

.90

.77

.77

.83

.88

.78

.84

.81

a

CCL-AH = Anxie ty subscale of the Cognition Checklist (CCL); CCL-S = Social Loss subscale of the CCL; CCL-W = Worthlessness subscale of the CCL; PA-R = revised Positive Affect scale of the Positive and Negative Affect Sche dule (PANAS); NA-R = revised Negative Affect scale of the PANAS; BDI = Beck Depression Inve ntory; GDS-SF = Geriatric Depression Scale — Short Form; BAI = Be ck Anxiety Inventory; STAI-S = State subscale of the State -Trait Anxiety Inve ntory. b p < .05. c p < .01. d p < .001.

a

¯ .17

.44 d

CCL-W

d

.39

.36 d

CCL-S

—

—

CCL-AH

CCL-S

CCL-AH

Measure

Table II. Means, Standard Deviations, Alpha Coefficients, and Z ero-Orde r Correlations Among Measures


63

and similar to that reporte d by Watson, Clark, and Tellegen (1988) . Finally, positive and ne gative affe ctivity were mildly to mode rate ly corre lated with the measure s of negative cognition (rs range d from | .15| to | .39| ). Zero-Order Correlations As a first te st of the spe cificity hypothe se s, the zero-orde r corre lation be twe e n e ach of the thre e cognition scale s and e ach symptom measure (anxie ty and de pre ssion) was compare d to its corre lation with e ach othe r symptom me asure by me ans of the t-statistic for de pe nde nt sample s (se e Table III). Inconsiste nt with findings from younge r sample s, the stre ngth of the corre lations be twe e n the thre e cognition measure s and the anxie ty symptom me asure s did not diffe r significantly from the corre lations be twe e n the cognition measure s and the de pre ssive symptom me asure s. In othe r words, e ach of the thre e cognition scale s was associate d with symptoms of de pre ssion to the same de gre e that it was associate d with symptom s of anxie ty.3 Like wise, the correlation be tween each affect scale (NA-R and PA-R) and each symptom measure (anxie ty and de pre ssion) was compare d to its corre lation with each othe r symptom measure by means of the t-te st statistic for depe nde nt sample s (se e Table III). The stre ngth of the corre lations betwee n the affe ct scale s and the anxie ty measure s did not diffe r significantly from the correlations betwee n the affect scales and the de pre ssion measure s. In othe r words, both NA-R and PA-R were

Table III. Te sts of Difference s Betwee n Z ero-O rde r Correlations a Scale

CCL-S

CCL-W

NA-R

PA-R

.41a

.15a

.54 a

.54a

¯ .33 a,b

GDS-SF

.32a

.17a

.53 a

.39b

¯ .36 a,b

BAI

.37a

.27a

.52 a

.65a

¯ .25 a

STAI-S

.33a

.20a

.48 a

.60a

¯ .47 b

a

3

CCL-AH

BDI

All corre lations are significant at p < .05. Within each column, different subscripts indicate significant differences in the magnitude of correlations (p < .05) . CCL-AH = Anxiety subscale of the Cognition Checklist (CCL); CCL-S = Social Loss subscale of the CCL; CCL-W = Worthlessne ss subscale of the CCL; NA-R = re vised Negative Affect scale of the Positive and Negative Affect Schedule (PANAS) ; PA-R = revised Positive Affect scale of the PANAS; BDI = Be ck De pression Inve ntory; GDS-SF = Geriatric De pre ssion Scale — Short Form; BAI = Beck Anxiety Inve ntory; STAI-S = State subscale of the State -Trait Anxiety Inve ntory.

Results were identical when these analyse s were rerun controlling for e ither positive or negative affectivity.

64


corre lated with symptoms of de pre ssion to the same e xtent that the y were correlate d with symptoms of anxie ty.4,5

Partial Correlations As a furthe r test of spe cificity, the unique variance of each type of symptomatology was e xamine d controlling for the opposite form of symptomatolo gy (see Table IV ). In othe r words, the unique variance in symptoms of depression was e xamine d by statistically controlling for variance in symptom s of anxie ty and, conversely, the unique variance in symptoms of anxie ty was e xamine d by statistically controlling for variance in symptoms of de pre ssion. In terms of the cognitive variable s, CCL -AH scores we re found to be associate d with the unique variance in symptom s of depression as measure d by the BDI (but not the GDS-SF) and with the unique variance in symptoms of anxie ty as measure d

Table IV. Tests of Differences between Partial Correlations Controlling for Opposite-Symptom Measures a Scale

CCL-AH

CCL-S

CCL-W

NA-R

PA-R

Partial correlations controlling for anxiety BDI

.21

GDS-SF

.13

c

¯ .02

.30

d

.16

.03

.33

d

.01

b

¯ .10 ¯ .17

b

Partial correlations controlling for depression BAI STAI-S

.18 .12

b

.21 .12

c

.27

d

.16

b

.49

d

¯ .05

.43

d

¯ .32

d

a

Both me asure s re flecting opposite symptomatology are partialed out. CCL-AH = Anxiety subscale of the Cognition Checklist (CCL); CCL-S = Social Loss subscale of the CCL; CCL-W = Worthlessne ss subscale of the CCL; NA-R = re vise d Ne gative Affect scale of the Positive and Ne gative Affect Sche dule (PANAS) ; PA-R = re vised Positive Affe ct scale of the PANAS; BDI = Beck Depression Inventory; GDS-SF = Geriatric Depression Scale — Short Form; BAI = Beck Anxiety Inventory; STAI-S = State subscale of the State -Trait Anxiety Inve ntory. b p < .05. c p < .01. d p < .001.

4

Less theore tically rele vant, the corre lations be tween NA-R and the BAI (r = .65) , t(186) = 4.50, p < .001, NA-R and the STAI-S (r = .60) , t( 186) = 3.78, p < .001, and NA-R and the BDI (r = .54) , t(186) = 2.86, p < .01, were all significantly greate r than the correlation between NA-R and the GDS-SF (r = .39). Also, the corre lation be tween PA-R and the STAI-S (r = ¯ .47) was significantly greater than the correlation between PA-R and the BAI (r = ¯ .25) , t( 186) = 3.52, p < .001. 5 Additional analyses were run in which the STAI-S was split into two scales, one that included only the positively valenced items (e.g., “I feel satisfied” ) and one that included only the negatively valenced items (e .g., “ I am worried”). Both the positively worded (r = ¯ .42) and negative ly worded (r = ¯ .38) subscales we re moderately corre lated with PA-R.


65

by the BAI (but not the STAI-S). Greater freque ncie s of negative health-re late d anxie ty cognitions were associate d with highe r le ve ls of de pressive and anxious symptomatology on these instrume nts. CCL -S score s were relate d to the unique variance in symptoms of anxie ty as measure d by the BAI (but not the STAI-S), but were not associate d with the unique variance in symptom s of depression as assessed by either the BDI or the GDS-SF. Greater fre quency of social loss cognitions was relate d to highe r le vels of anxie ty. Finally, after statistically controlling for the opposite form of symptomatology, the CCL -W was associate d with unique variance in both symptom s of de pre ssion and anxie ty across all four symptom measure s. Greater fre quency of worthle ssness cognitions was re lated to highe r leve ls of both depressive and anxious symptom s. In terms of the dimensions of affe ctivity, NA-R was associate d with the unique variance in both anxious and depre ssive symptoms. However, the associations between NA-R and unique variance in symptoms of anxie ty (as measure d by both the BAI and STAI-S) were moderate to strong in magnitude , whe reas the associations between NA-R and unique variance in symptoms of de pression were weak (in the case of the BDI) or nonsignificant (in the case of the GDS-SF). PA-R was negative ly associate d with the unique variance in depre ssive symptoms as measure d by the GDS-SF (but not the BDI), but it also was negative ly associate d with the unique variance in symptoms of anxie ty as measure d by the STAI-S (but not the BAI).

Regression Analyses As a final te st of specificity, hierarchical multiple -regression analyse s were conducted to de termine the degree to which e ach of the CCL and PANAS subscale s made unique contributions to the prediction of symptoms of anxie ty and depression, above that of the opposite form of symptomatology (see Table V). Regression analyse s were conducte d separate ly on each of the four symptom measure s (i.e., BDI, GDS-SF, BAI, and STAI-S). In each of the se analyse s, we used one of the four symptom measure s as the criterion variable and entered the two opposite symptom measure s at Step 1. The measure s reflecting the opposite form of symptomatology were entered first in order to statistically control for their high level of share d variance with the criterion variable . At Ste p 2, NA-R, PA-R, CCL -AH, CCL -S, and CCL -W were e nte red simultane ously as a block. This analysis provide s an e stimate of the unique contribution of each cognitive and affective dime nsion in the prediction of variance specific to either depre ssive or anxious symptomatology. The magnitude s of these effects are reporte d as standardize d betas ( b s) and partial correlations (prs). In the first two analyse s, the criterion variable s were symptom s of de pre ssion as measure d by the BDI (in the first analysis) and by the GDS-SF (in the second analysis) . In both analyse s, the BAI and STAI-S made significant contributions to the pre diction of depressive symptoms. The addition of the PANAS and CCL scales into the equation accounte d for an additional 9% of the variance above and beyond the effe cts of symptoms of anxie ty. Afte r controlling for all othe r variable s in the equation, NA-R, CCL -AH, CCL -S, and CCL -W were found to be significant predictors of BDI score s. Highe r levels of NA-R, CCL -AH, and CCL -W, but lower

66

Shap iro, Roberts , an d Beck Table V. Hierarchical Multiple Re gression Analyses Pre dicting Symptoms of Anxie ty and Depression a Predictor

b

pr

Step R2 change

t

Be ck De pre ssion Inventory Step 1 BAI STAI-S Step 2 NA-R PA-R CCL-AH CCL-S CCL-W

c

.37 .35

.30 .29

5.41 c 5.21

.16 ¯ .07 .12 ¯ .12 .27

.11 ¯ .06 .11 ¯ .11 .21

2.06 ¯ 1.09 b 2.00 b ¯ 2.08 c 4.02

b

.41c

.09c

Mode l R2 = .50, F(7, 181) = 25.69, p < .001 Geriatric Depression Scale — Short Form Step 1 BAI STAI-S Step 2 NA-R PA-R CCL-AH CCL-S CCL-W

c

.26 .37

.21 .31

3.54 c 5.17

¯ .01 ¯ .13 .04 .08 ¯ .33

¯ .01 ¯ .11 .03 ¯ .07 .25

¯ 0.11 b ¯ 1.98 0.56 ¯ 1.21 c 4.40

.32c

.09c

Mode l R2 = .41, F(7, 181) = 17.80, p < .001 Beck Anxiety Inve ntory Step 1 BDI GDS-SF

.65 .18

.37 .14

5.81 b 2.32

Step 2 NA-R PA-R CCL-AH CCL-S CCL-W

.44 .01 .03 .09 .17

.36 .01 .02 .08 .13

7.18 0.22 0.43 1.60 b 2.47

c

c

.34c

.20c

Mode l R2 = .54, F(7, 181) = 29.80, p < .001 State -Trait Anxiety Inve ntory— State Scale Step 1 BDI GDS-SF Step 2 NA-R PA-R CCL-AH

c

.38 .28

.29 .21

5.02 c 3.62

.37 .27 ¯ ¯ .01

.31 ¯ .22 ¯ .01

6.11 c ¯ 4.44 0.14 ¯

c

.36c

.18c


67 Table V. Continued

Predictor

b

pr

t

CCL-S CCL-W

.03 .08

.02 .06

0.48 1.15

Step R2 change

Mode l R2 = .54, F(7, 181) = 29.82, p < .001 a

B AI = B e ck Anxie ty Inve ntory; STAI-S = State subscale of the State -Trait Anxie ty Inve ntory; BDI = Be ck De pression Inve ntory; GDS-SF = Geriatric Depression Scale — Short Form; NA-R = revise d Ne gative Affect scale of the Positive and Negative Affect Schedule ( P ANA S) ; PA -R = re vise d Positive Affect scale of the PA NA S; CCL-AH = Anxie ty subscale of the Cognition Checklist (CCL); CCL-S = Social Loss subscale of the CCL; CCL-W = Worthlessness subscale of the CCL. b = standardize d be ta we ight; pr = partial corre lation. b p < .05. c p < .001.

le ve ls of CCL -S, were associate d with greate r de pre ssive symptomatology. In contrast, only PA-R and CCL -W made significant unique contributions to the pre diction of GDS-SF scores. Highe r le ve ls of CCL -W were associate d with highe r le ve ls of de pre ssive symptom atology, where as lowe r leve ls of PA-R were associate d with highe r leve ls of de pre ssive symptom atology. In the second two analyse s, the crite rion variable s were symptom s of anxie ty as measure d by the BAI (in the third analysis) and by the STAI-S (in the fourth analysis) . In both analyse s, the BDI and GDS-SF made significant contributions to the pre diction of anxious symptom s. The addition of the PANAS and CCL scales into the e quation accounte d for an additional 20% and 18% of the variance above and beyond the effe cts of symptoms of de pre ssion, re spe ctively. After controlling for all othe r variable s in the equation, NA-R and CCL -W were found to be significant predictors of BAI scores. Highe r leve ls of NA-R and CCL -W were associate d with gre ater anxious symptomatology. In contrast, NA-R and PA-R made significant contributions to the prediction of STAI-S scores. Highe r le ve ls of NA-R were associate d with greate r anxious symptom atology, whereas lower levels of PA-R were associate d with greate r anxious symptom atology. 6

DISCUSSION The present inve stigation e xamine d whether particular type s of cognitions and affe ctive state s are use ful in diffe re ntiating symptoms of anxie ty and depression in 6

All analyses, including both the regression analyses and the correlations, we re rerun controlling for health. This was to ensure that score s on measures were not being inflated because participants were e ndorsing symptoms due to physical illness. The Me dical Outcome Scale (MO S) Short-Form (Stewart, Hays, & Ware, 1988) was used to me asure physical functioning, pain, and health pe rceptions. After controlling for health in the regression analyse s, the CCL-AH and CCL-S scale s no longe r made significant unique contributions to the prediction of BDI score s. Otherwise, the results were largely identical to those re ported above .

68


community-dwe lling olde r adults, as the y have bee n found to be in younge r adults. These issues are important give n e vide nce that symptomatology of late-life depression and anxie ty might be phe nome nological ly distinct from that occurring in younge r adults. Although measure s of de pre ssion and anxie ty were mode rate ly correlate d in our sample as they are in younge r adults, the results from the prese nt study sugge st that the cognitive and affe ctive unde rpinnings of depression and anxiety are different in the elde rly. Furthe rmore, the prese nt findings sugge st that the cognitive and affective dimensions that are important in distinguishing de pre ssion and anxie ty among younge r sample s are not as useful among the e lderly. The re was little evide nce of distinct cognitive or affe ctive profile s that could diffe re ntiate symptoms of anxie ty or depression among our elde rly participants. In terms of the cognitive dimensions, confirmatory factor analysis indicate d that the two-factor mode l of de pre ssive and anxie ty cognition s e stablishe d in younge r sample s (Beck et al., 1987; Stee r e t al., 1994) provide d an inade quate fit to the data. Inste ad, three factors emerged from the CCL in an exploratory factor analysis. Although the first factor containe d only items from Be ck et al.’s ( 1987) original Anxious Cognition subscale , the Depressive Cognition subscale split into two separate scale s re flecting social loss and ne gative self-e valuation/worthle ssness cognitions. Appare ntly, the elde rly tend to make finer graine d distinctions among negative cognitions than younge r persons do. In othe r words, younge r persons tend to e xpe rience these de pre sse d cognitions as a single cluste r, whereas our e lde rly participants reporte d expe riencing more homoge nous “ subcluste rs” of ne gative cognitions. Pe rhaps of e ve n gre ate r importan ce, contrary to what has be e n found in younge r adults, the se cognitive dimensions were not spe cifically relate d to anxious and depressive symptoms in community e lders. First, whe re as score s on the CCL’s Anxie ty subscale have be en shown to be specifically predictive of anxious symptoms in younge r adults, the Anxie ty subscale (CCL -AH) in the current study was relate d to both anxious and de pressive symptoms. Ze ro-orde r correlations be tween this scale and symptoms of anxie ty and de pre ssion did not diffe r in magnitude . Furthe rmore, when anxious symptomatology was statistically controlle d in partial corre lations, the CCL -AH was associate d with BDI score s and, conve rse ly, whe n depressive symptomatology was controlle d, the CCL -AH was associate d with BAI score s. In furthe r contrast to the findings in younge r adults, re gression analyse s reveale d that, when all othe r cognitive and affective variable s were controlle d, anxious cognitions only were predictive of depressive symptoms as measure d by the BDI. Second, social loss cognitions generally were not re lated to anxious or depressive symptom s. Z e ro-orde r corre lations found that the se cognitions we re only weakly associate d with symptomatology. Furthe rmore, afte r statistically controlling for anxious symptomatology in partial correlations, the se cognitions were not significantly relate d to either of our two measure s of de pre ssive symptoms. Although social loss cognitions were re late d to anxious symptoms as measure d by the BAI afte r controlling for depressive symptoms, when all othe r cognitive and affe ctive variable s were controlle d in the regre ssion analysis, this association be came nonsignificant. This finding sugge sts that the association be tween social loss cognitions


69

and symptomatology among the elde rly is the result of share d variance with othe r cognitions or affective states rathe r than social loss cognitions pe r se. Finally, ne gative self-e valuation/worthle ssness cognitions were strongly re late d to symptoms of both anxie ty and de pre ssion e ve n afte r controlling for the opposite form of symptomatology in partial correlations. Furthe rmore , the se cognitions were predictive of both anxious and de pre ssive symptoms as measure d by three of our four measure s (with the e xce ption of the STAI-S), after controlling for all cognitive and affe ctive variable s. By definition, anxious cognitions consist of automatic thoughts re garding personal and/or physical dange r (e .g., “ What if I get sick and become an invalid? ” or “ I am going to be injure d ” ; Be ck e t al., 1987) . Howe ve r, in olde r adults, who are more like ly to suffe r from physical ailme nts, the se cognitions might reflect curre nt poor physical he alth rathe r than cognitive distortions. Be cause health proble ms have bee n repe atedly corre late d with depressive symptoms in e lderly sample s (Blaze r, 1989; Pfife r & Murre ll, 1986; Smith, Cole nda, & Espeland, 1994), the association betwee n “ anxious ” cognitions and depressive symptom s in our e lderly sample might have bee n due to a third variable — name ly, poor he alth. Consiste nt with this e xplanation, when health/physical functioning was statistically controlle d, the Anxie ty Cognition scale was no longe r pre dictive of de pre ssive symptoms (see footnote 6). The fact that social loss cognitions were only weakly predictive of symptoms of anxie ty and de pre ssion is some what surprising conside ring that the items on the scale (e .g., “ No one cares whe ther I live or die ” and “ There ’s no one left to he lp me ” ) see m to be particularly relevant during olde r adulthood. In ge neral, olde r adults go through multiple change s in social role s (e.g., re tirement, loss of a spouse ). Howe ver, de spite the se change s, rate s of de pre ssion re main re lative ly low compare d to those of younge r adults (Regier et al., 1988) . The fact that these type s of loss expe rience s are normative in the e lderly sugge sts that it is possible that change s in social role do not adve rse ly affect psychological well-be ing in the elde rly. Thus, it appe ars that cognitions pertaining to social role issues do not contribute to e motional distre ss in the e lde rly, and there fore are not use ful in diffe re ntiating symptoms of anxie ty and depression. Finally, negative se lf-evaluation/worthle ssne ss cognitions were robustly associated with score s on ne arly all measure s of e motional distre ss, even after statistically controlling for the othe r measure s of cognition and affect.7 Thus, negative selfevaluation/worthle ssness cognitions appe ar to be an important characte ristic of general psychological distress in the elde rly, particularly relative to the othe r cognitive dimensions examine d in the present research. The se findings contrast with rese arch on colle ge-age persons that has found that se lf-e ste em issue s (particularly temporal variability in se lf-e stee m) are specifically associate d with vulne rability to de pre ssive symptoms as compare d to anxie ty (Roberts & Gotlib, 1997) . It would be important 7

CCL-W was not significantly predictive of STAI-S score s in the regression analyses. Most likely, CCL-W was not as strong a pre dictor of anxiety symptoms (when compared to depressive symptoms) because of the significant ove rlap between NA-R and anxie ty. That is, because the revise d NA scale consisted primarily of anxie ty items (as will be discussed), NA-R accounted for a substantial portion of the variance in anxious symptomatology and left little variance to be e xplained by the other cognitive and affective variables.

70


for future rese arch to inve stigate the extent to which the se ne gative se lf-evaluation/worthle ssne ss cognitions play a role in vulne rability to emotional distre ss and symptomatology among the e lderly using prospe ctive de signs. In te rms of the affe ctive compone nts of late -life anxie ty and de pre ssion, re sults were partially consiste nt with findings from younge r sample s. Exploratory factor analysis of the PANAS re ve aled a two-factor solution that was identical in ite m composition to that found by Watson, Clark, and Te lle ge n (1988). Afte r several items that had relative ly small factor loadings (i.e., less than .50) were eliminate d from the scales, confirmatory factor analysis indicate d that the two-factor solution provide d a good fit to the data. However, the six remaining ite ms on the NA-R scale pre dominantly re fle cted anxie ty ( “ nervous,” “ jittery,” “ afraid, ” “ scare d,” “ distre sse d,” and “ irritable ” ) . Ite ms re fle cting conte nt othe r than anxie ty ( “ upse t,” “ ashame d,” “ hostile ,” and “ guilty ” ) faile d to cohe re on this factor in the same manner that they tend to in younge r sample s. Similar to our findings with the CCL, it appe ars that elde rly individuals tend to make fine r distinctions among their negative affe ctive states than younge r persons se em to do. O verall, the affe ctivity scales faile d to de monstrate the type of specificity freque ntly found in sample s of younge r persons. Although the zero-orde r correlations indicate d that ne gative affe ctivity was re late d to both de pre ssive and anxious symptoms, findings were somewhat diffe rent whe n the unique variance in de pre ssion and anxie ty were examine d in partial corre lation and re gression analyse s. In the se analyse s, the associations betwee n variance in NA and variance that was unique to anxie ty were moderate to strong in magnitude , whereas the associations be tween variance in NA and variance that was unique to depression were weak or nonsignificant depe nding on the particular measure . However, the implications of the se findings are difficult to de te rmine be cause the NA-R scale large ly containe d items reflecting anxious conte nt. In contrast to findings with younge r pe rsons, positive affectivity was not specifically associate d with depressive symptoms. First, zero-orde r correlations be tween PA and symptoms of depression were not large r in magnitude than those be tween PA and symptom s of anxie ty. Furthe rmore, partial correlation and re gression analyses reve ale d that PA was negative ly associate d with depressive symptom s as measure d by the GDS-SF (but not the BDI), but that it was also negative ly associate d with anxie ty symptoms as measure d by the State subscale of the STAI (but not the BAI).8 These results are intere sting give n that olde r adults with highe r leve ls of depressive symptoms may be less like ly than similarly depressed younge r adults to endorse fe elings of dysphoria (e.g., Gallo et al., 1994) . As pre viously discusse d, low PA is be st defined by descriptors refle cting fatigue (e .g., “ slee py” and “ sluggish ” ) and depression (e.g., “ sad ” and “ depressed ” ; Watson & Kendall, 1989) , sugge sting that PA and fee lings of dysphoria share conside rable variance . Pe rhaps the n, the olde r adults in the pre sent sample who were e xpe riencing othe r depressive symptoms did not report either dysphoria or low PA, there by making it difficult to difDiscrepancie s between the STAI-S and BAI might be due to differences in item content— the majority of the items on the BAI reflect somatic symptoms of anxiety, where as all of the items on the STAI-S reflect cognitive and affective symptoms of anxiety. In fact, Cox, Cohen, Dire nfeld, and Swinson ( 1996) have found evidence that the BAI may be a better measure of panic than of gene ral anxiety.

8


71

fe rentiate symptom s of de pre ssion from anxie ty based on PA score s. It is also possible that the nature of positive affectivity is diffe re nt in the elde rly. For example , Lawton e t al. ( 1993) found that arousal is a less important compone nt of PA in the e lderly compare d to younge r adults. It also is important to note that, in their earlie r work, Watson, Clark, and Tellege n (1988) found ne arly identical corre lations be tween PA and scores on the State subscale of the STAI (¯ .35) , and PA and score s on the BDI (¯ .36) , in colle ge stude nts. The y argue d that PA was re lated to state anxie ty (and not spe cific to depression) , in this case, because half of the STAI-S ite ms are “ re verse-ke ye d . . . reflecting pleasant or high PA state s” (e.g., fe eling joyful, ple asant, or self-confide nt, p. 1068) . Howe ver, as note d in footnote 5, in the prese nt data, the ne gative ly valenced STAI-S items also were associate d with PA. Finally, re lative ly low discriminant validity of the STAI re cently has be en found in othe r studie s (e .g., Creame r, Foran, & Be ll, 1995; Fydrich, Dowdall, & Chambless, 1992) , especially when compare d to the BAI. Be fore definitive conclusions can be made , our re sults ne ed to be replicate d with designs that addre ss the limitations of the prese nt study. The sample in this study was limite d to community olde r adults who were primarily recruited at active senior centers. It is possible that the unde rlying structure s of late -life anxie ty and depression may diffe r in e lde rly pe rsons who have sought treatme nt for de pre ssion or anxie ty disorde rs as compare d to our relative ly asymptomatic sample who had a relative ly restricte d range of score s. The se structure s might e ve n prove to differ in a sample of community adults who are le ss socially active and have le ss social support than our relative ly active and socially engage d participants. Also, all of the measure s e mploye d in this study were self-re port que stionnaire s. As has be en widely acknowle dge d, self-report measure s are limite d in that they assume individuals can accurate ly re port thoughts, emotions, and be haviors, and that individuals are willing to report socially unde sirable behavior. With regard to the present study, the ability of participants to accurate ly report automatic thoughts, in particular, see ms proble matic because, by de finition, the se cognitions are short-live d and involuntary. Also, the fact that individuals ofte n re spond to self-re port measure s in a socially desirable way is compounde d by the fact that measure s were comple te d in large groups, which might furthe r influe nce se lf-disclosure (Weisband & Kie sler, 1996) . We also should note that cognitive ite ms that are le ss directly tied to he alth issue s, social loss, and self-este em might result in greate r specificity. Future rese arch ne eds to be conducte d to e xamine the type s of cognitions and affe ctive state s that are unique ly associate d with late -life depression and anxie ty. In summary, the results of the prese nt study sugge st that the cognitive and affe ctive unde rpinnings of depression and anxie ty are diffe rent in olde r adults as compare d to younge r adults. In contrast to the two-factor structure that has be en found in younge r adults, ne gative automatic cognitions in the elde rly were composed of three factors: anxious, social loss, and worthle ssne ss cognitions. In ge neral, the se cognitions were not spe cifically re lated to anxious or depressive symptoms. Furthe rmore , although negative affectivity was relate d to both depression and anxiety, as in younge r sample s, positive affe ctivity generally was not use ful in differentiating de pre ssive from anxious symptom s. Be cause it is evide nt that the cognitions

72


and affective state s that have bee n used to differe ntiate anxie ty from de pre ssion in younge r adults are not use ful in olde r adults, future research nee ds to focus on elucidating the characte ristics that are spe cific to anxie ty and depression in olde r adults. In a time in which people are living longe r and the olde r adult population is growing, the unde rstanding of e motional distre ss in the elde rly is of paramount importance .

ACKNOWLEDGMENTS We would like to thank Laura Palumbo and J. Sidne y Shrauge r for their comments on an e arlie r ve rsion of this manuscript, as we ll as Nancy Collins for statistical advice .

REFERENCES Ambrose , B., & Rholes, W. S. (1993). Automatic cognitions and the symptoms of depression and anxiety in children and adole scents: An examination of the conte nt-specificity hypothesis. Cognitive Therapy and Research, 17, 153-171. American Psychiatric Association. (1980) . Diagnostic and statistical manual of m ental disorders (3rd e d.). Washington, DC: Author. American Psychiatric Association. (1987) . Diagnostic and statistical manual of mental disorders (3rd ed., re v.). Washington, DC: Author. Arean, P. A., & Miranda, J. ( 1992, Octobe r). Differen tial item functionin g of CES-D item s in older versus you nger m edical patients. Pape r pre se nted at the Sixth Annual NIMH International Re se arch Conference on Primary Care Mental He alth Re search: Conce pts, Methods, and Obstacles, Be thesda, MD. Beck, A. (1976) . Cognitive therapy and the em otional disorders. New York: New American Library. Beck, A. T., Brown, G., Steer, R. A., Eidelson, J. I., & Riskind, J. H. ( 1987) . Differentiating anxiety an d de pre ssion: A te st of the cognitive conte nt-specificity hypothe sis. Jo u rn al o f Abn orm al Psychology, 96, 179-183. Beck, A. T., & Eme ry, G. (1985) . Anxiety disorders and phobias: A cognitive perspective. New York: Basic Books. Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. ( 1988) . An inventory for measuring clinical anxiety: Psychometric properties. Journal of Consulting and Clinical Psychology, 55, 893-897. Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. ( 1979) . Cognitive therapy of depression. Ne w York: Guilford Pre ss. Beck, A. T., Steer, R. A., & Garbin, M. G. ( 1988) . Psychome tric prope rties of the Beck De pre ssion Inve ntory: Twenty-five years of evaluation. Clinical Psychology Review, 8, 77-100. Beck, J. G., Stanley, M. A., & Z ebb, B. J. ( 1996) . Characteristics of gene ralized anxiety disorder in older adults: A descriptive study. Behavior Research Therapy, 34, 225-234. Blaze r, D. G. (1989). Affective disorders in late -life. In E. W. Busse & D. G. Blazer (Eds.), G eriatric psychiatry (pp. 369-401) . Washington D.C.: Ame rican Psychiatric Press. Blaze r, D. G. (1993) . Depression in late life (2nd ed.). St. Louis: Mosby— Year Book. Blaze r, D., Bachar, J. R., & Hughes, D. C. (1987) . Major depression with me lancholia: A comparison of middle-aged and elderly adults. Journal of the Am erican G eriatrics Society, 35, 927-932. Blaze r, D., Ge orge, L. K., & Hughe s, D. ( 1991) . The e pidemiology of anxie ty disorders: An age comparison. In C. Salzman & B. D. Lebowitz (Eds.), Anxiety in the elderly (pp. 17-30). New York: Springer. Blaze r, D., George, L., & Landerman, R. (1986) . The phenomenology of late-life depression. In P. E. Be bbington & R. Jacoby (Eds.), Psychiatric disorders in the elderly (pp. 143-151) . London, England: Mental Health Foundation. Brodaty, H., Peters, K., Boyce, P., Hickie, I., Parker, G., Mitchell, P., & Wilhelm, K. (1991) . Age and depression. Journal of Affective Disorders, 23, 137-149.


73

Busse, E. (1975) . Aging and psychiatric diseases in late-life. In M. Reiser (Ed.), American handbook of psychiatry, 4. Ne w York: Basic Books. Clark, D. A. ( 1986) . Cognitive -affective inte raction: A te st of the “ spe cificity” and “ ge ne rality” hypothese s. Cognitive Therapy and Research, 10, 607-623. Clark, D. A., Beck, A. T., & Brown, G. (1989) . Cognitive mediation in general psychiatric outpatients: A test of the content-specificity hypothesis. Journal of Personality and Social Psychology, 56, 958-964. Clark, D. A., Beck, A. T., & Stewart, B. (1990) . Cognitive specificity and positive-negative affectivity: Complementary or contradictory views on anxiety and de pre ssion? Journal of Abnorm al Psychology, 99, 148-155. Clark, D. A., Steer, R. A., & Be ck, A. T. (1994). Common and spe cific dimensions of self-reported anxiety and de pre ssion: Implications for the cognitive and tripartite models. Journal of Abnorm al Psychology, 103, 645-654. Clark, L. A. ( 1989) . The anxie ty and depressive disorders: Descriptive psychopathology and differe ntial diagnosis. In P. C. Kendall & D. Watson (Eds.), An xiety and depression: Distinctive and overlaping features (pp. 83-129) . San Diego: Academic Press. Cox, B. J., Cohen, E., Direnfeld, D. M., & Swinson, R. P. ( 1996) . Does the Beck Anxiety Inventory measure anything beyond panic attack symptoms? Behaviour Research and Therapy, 34, 949-954. Craig, T. J., & V an Natta, P. A. (1979) . Influence of demographic characte ristics on two measures of de pre ssive symptoms: The re lation of pre vale nce and pe rsistence of symptoms with se x, age , education, and marital status. Archives of G eneral Psychiatry, 36, 149-154. Cre ame r, M., Foran, J., & B ell, R. (1995) . The Be ck Anxie ty Inve ntory in a non-clinical sample . Behaviour Research and Therapy, 33, 477-485. Dobson, K. S. ( 1985) . The relationship between anxiety and depression. Clinical Psychology Review 5, 307-324. Fydrich, T., Dowdall, D., & Chamble ss, D. L. (1992) . Re liability and validity of the Be ck Anxiety Inve ntory. Journal of Anxiety Disorder 6, 55-61. Gallagher, D., Nies, G., & Thompson, L. W. ( 1982) . Reliability of the Beck Depression Inve ntory with older adults. Journal of Consulting and Clinical Psychology, 50, 152-153. Gallo, J. J., Anthony, J. C., & Muthen, B. O. ( 1994) . Age differences in the symptoms of de pre ssion: A late nt trait analysis. Journal of G erontology, 49, 251-264. Goldfarb, A. I. (1974) . Masked depression in the e lderly. In Lesse (Ed.), Masked depression (pp. 236-249) . New York: Jason Aronson. Gurian, B. S., & Miner, J. H. (1991) . Clinical prese ntation of anxiety in the e lderly. In C. Salzman & B. D. Le bowitz (Eds.), Anxiety in the elderly (pp. 31-44). Ne w York: Springe r. Harre ll, T. H ., Chamble ss, D. L., & Calhoun, J. F. ( 1981) . Corre lational re lationships be twee n self-statements and affe ctive states. Cogn itive Therapy and Research, 5, 159-173. Ingram, R. E., Kendall, P. C., Smith, T. W., Donnell, C., & Ronan, K. (1987) . Cognitive specificity in emotional distress. Journal of Personality and Social Psychology, 53, 734-742. Jolly, J. B., Dyck, M., Kramer, T. A., & Wherry, J. N. (1994) . Integration of positive and negative affe ctivity and cognitive conte nt-specificity: Improved discrimination of anxious and de pre ssive symptoms. Journal of Abnorm al Psychology, 103, 544-552. J öreskog, K. G., & S örbom, D. (1996). Lisrel 8.14: User’s reference guide. Chicago: Scientific Software International. Lade r, M. (1982). Differential diagnosis of anxiety in the e lderly. Journal of Clinical Psychiatry, 43, 4-7. Lawton, M. P., Kleban, M. H., & Dean, J. ( 1993) . Affect and age : Cross-sectional comparisons of structure and prevalence . Psychology and Aging, 8, 165-175. Lesher, E. L., & Be rryhill, J. S. (1994) . V alidation of the Geriatric De pre ssion Scale — Short Form among inpatients. Journal of Clinical Psychology, 50, 256-260. Morin, C. M., Stone, J., Ling, W., & Trinkle, D. (1994, Novembe r). Th e Beck An xiety In ventory: Psychom etric properties with older adults. Poste r se ssion presented at the annual mee ting of the Association for the Advanceme nt of Behavior The rapy, San Die go, CA. Musetti, L., Perugi, G., Soriani, A., Rossi, V. M., Cassano, G. B., & Akiskal, H. S. (1989) . De pre ssion before and after age 65: A re-examination. British Journal of Psychiatry, 155, 330-336. Olin, J. T., Schneider, L. S., Eaton, E. M., Z emansky, M. F., & Pollock, V. E. (1992) . The Geriatric De pre ssion Scale and the Be ck De pre ssion Inventory as scree ning instruments in an older adult outpatient population. Psychological Assessm ent, 4, 190-192. Patterson, R. L., O ’Sullivan, M. J., & Spielberge r, C. D. ( 1980) . Me asure ment of state and trait anxiety in elde rly mental health clients. Journal of Behavioral Assessm ent, 2, 89-97. Pfeiffer, E. (1979). Interviewing the anxious patie nt. In Diagnosis and treatm ent of anxiety in the elderly, part II (pp. 23-33) . Nutley, NJ: Hoffman-La Roche.

74


Pfifer, J. F., & Murrell, S. A. (1986) . Etiologic factors in the onse t of depressive symptoms in older adults. Journal of Abnorm al Psychology, 95, 282-291. Re gier, D. A., Boyd, J. H., Burke, Jr., J. D., Rae, D. S., Myers, J. K., Kramer, M., Robins, L. N., George, L. K, Karno, M., & Locke, B. Z . (1988) . One-month prevalence of mental disorders in the United States, based on five epidemiologic catchment area sites. Archives of G eneral Psychiatry, 45, 977-986. Robe rts, J. E ., & G otlib, I. H. ( 1997) . Te mp oral variabi lity in gl obal se lf-este e m and spe cific self-evaluation as prospective pre dictors of e motional distress: Specificity in pre dictors and outcome . Journal of Abnorm al Psychology, 106, 521-529. Salzman, C. ( 1982) . A primer on geriatric psychopharmacology. American Journal of Psychiatry, 139, 67-74. Shamoian, C. A. (1991) . What is anxiety in the elderly? In C. Salzman & B. D. Lebowitz (Eds.), Anxiety in the elderly (pp. 3-15) . New York: Springer. She ikh, J. I. & Ye savage , J. A. ( 1986) . Ge riatric De pression Scale (GDS) : Re ce nt e vidence and developme nt of a shorter version. Clinical G erontologist, 5, 165-173. Smith, S. L., Colenda, C. C., & Espeland, M. A. (1994) . Factors determining the level of anxiety state in ge riatric primary care patients in a community dwelling. Psychosom atics, 35, 50-58. Spielberger, C., Gorsuch, R., & Lushene, R. ( 1970) . STAI manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press. Stavrakaki, C., & Vargo, B. (1986) . The re lationship of anxie ty and depression: A review of the literature . British Journal of Psychiatry, 149, 7-16. Steer, R. A., Beck, A. T., Clark, D. A., & Beck. J. S. (1994) . Psychometric properties of the cognition che cklist with psychiatric outpatients and unive rsity students. Clinical Assessm ent, 6, 67-70. Steer, R. A., Clark, D. A., Beck, A. T., & Ranieri, W. F. ( 1995) . Common and specific dime nsions of self-reported anxiety and depression: A replication. Journal of Abnorm al Psychology, 104, 542-545. Stewart, A. L., Hays, R. D., & Ware, J. E. ( 1988) . The MO S short-form general health survey. Medical Care, 26, 724-735. Tellege n, A. (1985). Structures of mood and personality and their relevance to assessing anxiety, with an e mphasis on self-report. In A. H. Tuma & J. D. Maser (Eds.), Anxiety and the anxiety disorders. (pp. 681-706) . Hillsdale, NJ: Erlbaum. Thorpe, G. L., Barnes, G. S., Hunter, J. E., & Hines, D. ( 1983) . Thoughts and fee lings: Correlations in two clinical and two nonclinical samples. Cognitive Therapy and Research, 7, 565-574. Watson, D., Clark, L. A., & Carey, G. (1988) . Positive and ne gative affe ctivity and their relation to anxiety and de pre ssive disorders. Journal of Abnorm al Psychology, 97, 346-353. Watson, D., Clark, L. A., & Te llegen, A. (1984) . Cross-cultural converge nce in the structure of mood: A Japane se replication and comparison with U. S. findings. Journal of Person ality and Social Psychology, 47, 127-144. Watson, D., Clark, L. A., & Tellegen, A. (1988) . Deve lopment and validation of brief measures of positive and negative affe ct: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063-1070. Watson, D., & Kendall, P. C. (1989) . Unde rstanding anxiety and de pre ssion: Their relation to negative and positive affe ctive state s. In P. C. Kendall & D. Watson (Eds.), Anxiety and depression: Distinctive and overlapping features (pp. 3-26) . San Diego: Academic Press. Watson, D., & Tellege n, A. (1985) . Toward a consensual structure of mood. Psychological Bulletin, 98, 219-235. Weisband, S., & Kiesler, S. ( 1996) . Self-disclosure on com puter form s: Meta-analysis and implications (Proceedings of CHI ‘96, ACM SIGCHI) New York: Association for Computing Machinery. Weissman, M. M., Myers, J. K., Tischler, G. L., Holze r, C. E., Leaf, P. J., Orvaschel, H., & Brody, J. A. ( 1985) . Psychiatric disorders (DSM-III) and cognitive impairme nt among the e lderly in a U.S. urban community. Acta Psychiatrica Scandinavica, 71, 366-379. Ye savage, J. A., Brink, T. L., Rose, T. L., Lum, O., Huang, V ., Ade y, M. B., & Leirer, V. O. (1983) . De velopment and validation of a geriatric depression scre ening scale: A preliminary report. Journal of Psychiatric Research, 22, 37-49.