associations among social support, income, and symptoms of ...uncahs.org/articles/brummett 2003 int...

INTERNATIONAL JOURNAL OF BEHAVIORAL MEDICINE, 10(3), 239–250Copyright © 2003, Lawrence Erlbaum Associates, Inc.

Associations Among Social Support,Income, and Symptoms of Depression

in an Educated Sample: The UNC Alumni Heart Study

Beverly H. Brummett, John C. Barefoot, Peter P. Vitaliano,and Ilene C. Siegler

It has been suggested that the inverse association between social support and de-pression may be stronger in persons with lower income. This study tested the sup-port × income hypothesis in a sample of 2,472 individuals enrolled in the UNCAlumni Heart Study. The income was examined as a moderator of the relation be-tween support and self-reported ratings of symptoms of depression. The appraisalsubscale of the Interpersonal Support Evaluation Scale, household income level,and their interaction were modeled as predictors of depression ratings. The sup-port × income interaction term was significantly associated with symptoms ofdepression, F(2, 2471) = 4.71, p = .007. Social support was more strongly associ-ated inversely with depression ratings in persons with lower income as comparedto those with higher income. The present results extend previous work regardingthe moderating effect of income to a sample of relatively high education andincome level.

Key words: income, social support, depression, distress

Beverly H. Brummett, John C. Barefoot, and Ilene C. Siegler, Duke University Medical Center,Durham, NC, USA; Peter P. Vitaliano, Department of Psychiatry and Behavioral Sciences, Universityof Washington, Seattle, WA, USA.

This research was supported by Grants R01-AG12458 from the National Institute on Aging, R01-HL55356 and P01-HL36587 from the National Heart Lung and Blood Institute, and R01-MH57663from the National Institute of Health.

Correspondence concerning this article should be addressed to Beverly H. Brummett, Duke Uni-versity Medical Center, Box 2969, Durham, NC 27710, USA. E-mail: [email protected]

It is generally accepted that perceptions of social support are associated with psy-chosocial and health outcomes (Berkman, 1995; Horsten, Mittleman, Wamala,Schenck-Gustafsson, & Orth-Gomer, 2000; Kaplan & Toshima, 1990; Ruberman,Weinblatt, & Goldberg, 1984; Williams et al., 1992). Likewise, research has shownthat level of income is related to general well-being and mortality (Backlund, Sorlie,& Johnson, 1999; Diener, Suh, Lucas, & Smith, 1999; Kaufman, Long, Liao,Cooper, & McGee, 1998). Much of the past research, however, has been built onmain effects models, and more complex interactional perspectives are needed totake the next step in our understanding (Taylor & Seeman, 1999).

It has been hypothesized that the association between social support and dis-tress may be stronger in persons with lower income in part because of their poten-tial lack of resources. The resources hypothesis is based on the proposition thatpsychological distress in the presence of environmental stressors is a negativefunction of a person’s coping resources and a positive function of their vulnera-bility (Vitaliano, Dougherty, & Siegler, 1994; Vitaliano, Russo, Young, Teri, &Maiuro, 1991). Support for this hypothesis has been found in diverse samples. So-cial support was a stronger predictor of recovery from depression among cardiacpatients who had lower income compared to those with higher income (Barefootet al., 2000). Similarly, in a sample of spouse caregivers of persons withAlzheimer’s disease and matched controls social support was more beneficial inpersons with lower income with respect to physiological responses (Vitaliano et al., 1999, 2001). In addition, variation in blood pressure among individuals ofdifferent social classes has been associated with perceptions of social support(Dressler, Grell, Gallagher, & Viteri, 1992). Finally, research conducted byStrogatz and James (1986) has shown that the relation between hypertension andemotional support among Blacks is specific to persons with low income.

This study used data from the University of North Carolina Alumni HeartStudy (UNCAHS) to further examine the hypothesis that the association betweensocial support and distress is dependent on household income level. With respectto the present sample the lack of resources hypothesis predicts that distress (i.e.,ratings of depressive symptoms) will be a multiplicative function of social supportand income level. Unlike the above-noted studies that have examined this hypoth-esis, the UNCAHS sample consists primarily of college-educated individuals liv-ing in households reporting moderate to higher income, thereby affording an op-portunity to explore the boundaries of the lack of resources theory. Demonstrationof the hypothesized finding in the present sample would suggest that the associa-tion between support and distress may not be unique to persons at the lower endsof the income distribution.

In addition to income level, two other financial measures (i.e., sufficiency of in-come and financial pressure ratings) were used to explore the lack of resourceshypothesis. Specifically, if insufficient financial resources underlie the need forsupport in the lower income groups, the measures of perceived financial adequacy

240 BRUMMETT, BAREFOOT, VITALIANO, SIEGLER

(i.e., sufficiency and financial pressure) should follow a pattern similar to that ofincome, for example, individuals reporting financial concerns should benefit morefrom social support as compared to those reporting fewer financial worries.Finally, studies have shown that other measures of socioeconomic status (SES)may have differential relations to health behavior, mortality, and distress (Liberatos,Link, & Kelsey, 1988). Therefore, the potential effects of two additional measuresof SES (i.e., education level and occupational prestige) were also explored in thisstudy. Thus, this study provides an opportunity to significantly extend previousfindings.

METHODS

Participants

The UNCAHS is an ongoing prospective study of coronary heart disease andcoronary heart disease risk that started its prospective data collection in1986–1987 when alumni were in their early 40s (Siegler et al., 1992a; Siegler,Peterson, Barefoot, & Williams, 1992b). This study sample consisted of 2,472UNCAHS participants who had data available for the primary measures of inter-est (i.e., ratings of depression, social support, and income). This sample reflectsthe sociodemographic characteristics of the UNC student population in the 1960s,with minority enrollment less than 1%.

Measures

Psychosocial and demographic measures examined in this study were drawn fromUNCAHS follow-up data. Table 1 shows the timeline for the collection of allmeasures analyzed in this article. Table 2 provides the sample characteristics

INCOME, SUPPORT, AND DEPRESSION 241

TABLE 1Time Table for Collection of Study Measures

Study Measure Time (Wave) of Data Collection

Social support Study Wave 3 (1989–1990)Income Study Wave 3 (1989–1990)Sufficiency of income Study Wave 3 (1989–1990) Education level Study Wave 3 (1989–1990) Occupational prestige Study Wave 3 (1989–1990) Financial concerns Study Wave 2 (1987–1988) Symptoms of depression Study Wave 6 (1994–1995)

Note. The dropout rate for individuals who had the required data at Wave 3 but did not providedata at Wave 6 was 1%.

including the means and standard deviations, or frequencies where appropriate, ofthe study variables described later. Finally, Table 3 presents the relations amongthe independent measures.

Social support. In the UNCAHS social support was assessed using the10-item appraisal subscale of the Interpersonal Support Evaluation Scale (ISEL)(Cohen, Mermelstein, Kamarck, & Hoberman, 1985). The ISEL was designed toassess perceptions of the availability of social resources. The appraisal subscaleprimarily captures perceptions regarding the availability of someone to discuss hisor her problems with, for example, “There are several people who I trust to helpsolve my problems.” The ISEL appraisal subscale has demonstrated adequate in-ternal reliability (α coefficient = .77–.92) in general population samples (Cohen et al., 1985). Appraisal scores range between 10 and 40. Higher scores onthe ISEL indicate positive perceptions of social support.

Household income. Income was assessed using an 11-level categoricalitem that asked participants to categorize their annual household income betweenthe following extremes: 1 = below $10,000 to 11 = $300,000 and above. Two


TABLE 2Sample Characteristics

Characteristic Value

Gender 77.8% maleAge, mean (SD) 42.8 (1.3)Marital status 80.2% marriedEducation level

Some college, no degree 6.9%College degree 18.3%College degree + training 27.2%Masters degree 24.1%Doctorate/Law/Medical 23.5%

Symptoms of depressionCES-D, mean (SD) 8.2 (7.7)Scored 16 or above on CES-D 14.1%

Appraisal of support, mean (SD) 32.3 (6.1)Income

Low: below $50,000 21.9%Moderate: $50,000–99,999 50.4% High: $100,000 and above 27.7%

Sufficiency of income, median 7Financial pressure, median 3Occupational prestige, mean (SD) 61.0 (18.1)

Note. Range for sufficiency of income and financial pressure is 1–10, higher scores representgreater sufficiency and pressure.

statistics reported by the U.S. Census are helpful to put the present income ratings (Table 2) in context. During 1990 the median annual income reported forall households was $29,943 and the median annual income for men 45–54 yearsof age with Bachelor’s degree was $62,174 (U.S. Census Bureau: http:/www.census.gov/hhes/www/income.html). Thus, although the household income re-ported in this sample is likely to be high as compared to values reported by a ran-dom population sample, it is consistent with what would be expected given thecharacteristics of this sample (i.e., primarily college-educated men).

Symptoms of depression. The Center for Epidemiologic Studies Depres-sion Scale (CES-D) (Radloff, 1977) is a 20-item self-report scale designed tomeasure depressive symptomatology in a general population. The items werescored on a 4-point scale, with the total score ranging between 0 and 60. Higherscores on the CES-D represent depressive responses and a score of 16 or above isgenerally considered suggestive of a depressive disorder. Test–retest correlationsfor the CES-D range between .45 and .70 (Radloff, 1977).

Sufficiency of income. Along with reports of income, UNCAHS partici-pants also responded to a measure indicating their perceptions of the sufficiencyof their income that read as follows: “How sufficient is your income when consid-ering your commitments, responsibilities and the lifestyle choices you havemade?” Responses were rated from 1 (not at all sufficient) to 10 (more thansufficient).

Financial pressure. The study participants also reported their perceptionsregarding the amount of pressure in their lives due to financial issues by rating thefollowing item (from 1 [no pressure] to 10 [extreme pressure]): “pressure causedby financial issues.”


TABLE 3Pattern of Correlations Among Independent Measures

Variable 1 2 3 4 5 6

1 Social support —2 Income .04* —3 Occupational prestige .02 .15* —4 Education level .03 .21* .53* —5 Sufficiency of income .12* .48* .07* .10* —6 Financial pressure –.06* –.16* –.10* –.07* –.41* —

Note. For sufficiency of income and financial pressure, higher scores represent greater suffi-ciency and pressure.

*p < .05.

Education. In this sample education was treated as a 5-level categoricalvariable in analyses (see Table 2 for distribution).

Occupational prestige. Occupational prestige was derived from a census-based classification scheme that provides a score indicating the level of occupa-tional SES (Stevens & Cho, 1985; Stevens & Featherman, 1981). Specific occu-pations across a broad range of categories receive a score that falls between 14 and91, with higher scores indicating greater occupational prestige. UNCAHS codersclassified occupational prestige scores from occupations reported at StudyWave 3.

Analyses

To examine the primary hypothesis a general linear model was estimated withhousehold income level, social support, gender, and the two- and three-way inter-action terms as predictors of CES-D ratings. Nonsignificant interaction termswere removed from final models. It was predicted that associations between sup-port and symptoms of depression would be stronger among individuals reportinglower income. Statistical significance was set at p < .05 (two-tailed test) for thisstudy. An identical analytic approach was applied to the additional measures of fi-nancial pressure, income sufficiency, education level, and occupational prestige.

RESULTS

The main effects and significant interaction results are provided in Table 4. Withrespect to the main effects, social support, financial concerns, and sufficiency ofincome ratings were significantly related to symptoms of depression. In contrast,level of education and occupational prestige ratings were unrelated to ratings ofdepression.

The support × income interaction term was a significant predictor of CES-Dratings such that support was more strongly associated with symptoms of depres-sion for participants of lower household income compared to those who reportedhigher income levels. Figure 1 portrays the pattern of the support × income inter-action. Other two- and three-way interactions (gender × income, gender × support,gender × income × support) were nonsignificant (ps .78 –.98).

The support × financial adequacy interactions examined were statisticallysignificant and followed a pattern similar to that of income. Figure 2 portrays thepattern of results for these findings. Analyses also revealed that neither the socialsupport × education interaction term nor the social support × occupational pres-tige reached the level of statistical significance as a predictor of CES-D ratings ( p = .91; p = .584, respectively).


245

TABLE 4Results of Main Effects and Interactions of Theoretical Interest Associated

With Symptoms of Depression

Main Effects Results

Social support F(1, 2471) = 101.6, p < .001Income F(2, 2471) = 16.5, p < .001Financial pressure F(1, 2381) = 83.1, p < .001Sufficiency of income F(1, 2463) = 70.1, p < .001Education F(1, 2471) = 1.7, p < .200Occupational prestige F(1, 2244) = 0.61, p < .436Interactions

Social support × income F(2, 2471) = 4.86, p = .007Social support × financial pressure F(1, 2381) = 14.32, p = .001Social support × income sufficiency F(1, 2463) = 13.58, p = .001Social support × education level F(1, 2461) = 0.75, p = .387Social support × occupational prestige F(1, 2244) = 0.42, p = .517

Note. All models adjusted for gender.

FIGURE 1. Social support (SS) × household income interaction as a moderator of symptomsof depression (CES-D ratings).

246

FIGURE 2. Social support (SS) × income adequacy interactions as moderators of symp-toms of depression (CES-D ratings).

DISCUSSION

The present results extend previous findings by demonstrating a social support ×income interaction in middle-aged persons of relatively high income and educationlevel. Indeed, examination of these findings showed that social support seemed in-consequential with respect to CES-D scores only in the very high income group.For some this may call into question the feasibility of a lack of resources hypothe-sis. However, in this “baby boom” sample it is possible that relatively high levels ofincome are necessary to perceive that there is no need for a social buffer with re-spect to monetary concerns. Finally, in addition to providing support for the sup-port × income interaction hypothesis, the present findings suggest that social sup-port, income, income sufficiency, and concerns regarding financial pressure have asubstantial main effect with respect to symptoms of depression.

Unlike prior research in this area, this article provides support for the support ×income interaction by analyzing income from three different perspectives. Specif-ically, when our measure of absolute household income was replaced with meas-ures that are somewhat more subjective and may more closely capture the relativenature of income, persons with lower ratings of income sufficiency and higher rat-ings of financial pressure benefited more from social support as compared to thosewith lower ratings. Such findings may provide indirect support for Wilkinson’stheory that perceptions of an individual’s place in the social hierarchy, and the re-sulting health effects of such perceptions, are based on the relative position of theincome level rather than the absolute level (Wilkinson, 1996). Indeed, results ofpost hoc analyses that included the interactions of all three indicators with socialsupport simultaneously showed that the measure of financial pressure was thestrongest predictor of symptoms of depression.

As previously mentioned, different measures of SES may have differential re-lations with health outcomes (Davey-smith et al., 1998; Liberatos et al., 1988).Such findings may be expected given the differential attributes of SES measures.For example, occupational prestige and income are likely to fluctuate during alifetime, whereas education level is likely to remain stable following the decade ofthe 20s (Davey-smith et al., 1998; Liberatos et al., 1988). With respect to bothmain effects and interaction models, education level and occupational prestige didnot demonstrate a pattern of results similar to that of the income in this study. Thislack of association has also been demonstrated by Vitaliano (Vitaliano et al.,2001) who found that, across time, social support was more beneficial in personswith lower income, but not with lower education, with respect to physiological re-sponses. Thus high levels of income may offer a feeling of security with respect tofewer needs for social support that is not provided by other SES indicators. How-ever, it is also possible that the lack of a significant relationship for the measuresof education and occupational prestige in this study were due to the restrictedrange of these measures in this relatively high SES sample.


A well-known alternative hypothesis exists that may also account for the pres-ent pattern of findings. Cohen and Wills (1985) proposed a stress-buffering hy-pothesis with respect to the positive association between support and well-being.The buffering hypothesis suggests that support is protective during times of stress.Thus, in this study low income might be viewed as a stressor that is moderated bysocial support. Indeed, related research in older adults has shown that informa-tional support buffers the effects of chronic financial strain with respect to de-pressive symptoms (Krause, 1987). In this article it is empirically impossible todetermine which interpretation is most correct.

There are several points that should be considered with regard to interpretingthese findings. It should be noted that these findings may not be generalized tosample of lower income and education levels. In addition, few of our participantsreported high ratings on the CES-D, nor did they report extremely low levels ofsocial support. However, the pattern of results was fairly consistent across sev-eral outcome measures. Also, group differences larger than 3 points on a meas-ure with a standard deviation of 7.7 are typically considered sizable. In addition,this study included only the appraisal subscale of the ISEL, therefore these find-ings may not be generalized to other areas of support (i.e., self-esteem, belong-ing, and tangible support). Furthermore, it has been suggested that symptoms ofdepression influence how a person rates his or her social support, as opposed tosocial support influencing ratings of depression. It is not possible to refute thisentirely, however, results of prospective studies suggest that this is probably notthe case (Blazer & Hughes, 1991; Krause, Liang, & Yatomi, 1989; Lin & Dean,1984).

Adler, Boyce, Chesney, Folkman, and Syme (1993) noted that the relation be-tween SES and health outcomes is not limited to those of low income. It is notclear, however, whether the mechanisms that underlie the SES and health associ-ation are the same at the lower and the upper ends of the SES spectrum. Prior re-search examining the present association between social support and distress atdifferent levels of income has demonstrated this finding in primarily lower in-come individuals and individuals who are experiencing significant life stress(i.e., patient and caregiver samples). Thus, these findings contribute to the litera-ture in this area by suggesting that the relation between psychological distressand social support occurs at moderately high levels of SES and therefore maymoderate the association of SES and health even at the upper ends of the SESgradient.

REFERENCES

Adler, N. E., Boyce, T., Chesney, M. A., Folkman, S., & Syme, S. L. (1993). Socioeconomic inequali-ties in health. Journal of the American Medical Association, 269(24), 3140–3145.


Backlund, E., Sorlie, P. D., & Johnson, N. J. (1999). A comparison of the relationships of education andincome with mortality: The National Longitudinal Mortality Study. Social Science and Medicine,49, 1373–1384.

Barefoot, J. C., Brummett, B. H., Clapp-Channing, N. E., Siegler, I. C., Vitaliano, P. P., Williams, R. B.,& Mark, D. B. (2000). Moderators of the effect of social support on depressive symptoms in cardiacpatients. American Journal of Cardiology, 86, 438–442.

Berkman, L. F. (1995). The role of social relations in health promotion. Psychosomatic Medicine, 57,245–254.

Blazer, D., & Hughes, D. C. (1991). Subjective social support and depressive symptoms in major de-pression: Separate phenomena or epiphenomena. Journal of Psychiatric Research, 25(4), 191–203.

Cohen, S., Mermelstein, R., Kamarck, T., & Hoberman, H. (1985). Measuring the functional compo-nents of social support. In I. G. Sarason & B. R. Sarason (Eds.), Social support: theory, research andapplication. The Hague, Holland: Martinus Nijhoff.

Cohen, S., & Wills, T. A. (1985). Stress, social support, and the buffering hypothesis. PsychologicalBulletin, 98, 310–357.

Davey-Smith, G., Hart, C., Hole, D., MacKinnon, P., Gillis, C., Watt, G., Blane, D., & Hawthorne, V.(1998). Education and occupational social class: Which is the more important indicator of mortal-ity risk? Journal of Epidemiology and Community Health, 52, 153–160.

Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective well-being: Three decades ofprogress. Psychological Bulletin, 125, 276–302.

Dressler, W., Grell, G., Gallagher, P., & Viteri, F. (1992). Social factors mediating social class differ-ences in blood pressure in a Jamaican community. Social Science Medicine, 35, 1233–1244.

Horsten, M., Mittleman, M. A., Wamala, S. P., Schenck-Gustafsson, K., & Orth-Gomer, K. (2000). De-pressive symptoms and lack of social integration in relation to prognosis of CHD in middle-agedwomen. The Stockholm Female Coronary Risk Study. European Heart Journal, 21, 1043–1045.

Kaplan, R. M., & Toshima, M. T. (1990). The functional effects of social relationships on chronicillnesses and disability. In B. R. Sarason, I. G. Sarason, & G. R. Pierce (Eds.), Social support: An interactional view (pp. 427–453). New York: Wiley.

Kaufman, J. S., Long, A. E., Liao, Y., Cooper, R. S., & McGee, D. L. (1998). The relation betweenincome and mortality in U.S. blacks and whites. Epidemiology, 9, 147–155.

Krause, N. (1987). Chronic financial strain, social support, and depressive symptoms among olderadults. Psychology and Aging, 2, 185–192.

Krause, N., Liang, J., & Yatomi, N. (1989). Satisfaction with social support and depressive symptoms:A panal analysis. Psychology and Aging, 4, 88–97.

Liberatos, P., Link, B. G., & Kelsey, J. L. (1988). The measurement of social class in epidemiology.Epidemiologic Reviews, 10, 87–121.

Lin, N., & Dean, A. (1984). Social support and depression: A panel study. Social Psychiatry, 19,83–91.

Radloff, L. S. (1977). The CES-D scale: A self report depression scale for research in the generalpopulation. Applied Psychological Measurement, 1(3), 385–401.

Ruberman, W., Weinblatt, E., & Goldberg, J. D. (1984). Psychosocial influences on mortality aftermyocardial infarction. New England Journal of Medicine, 311, 552–559.

Siegler, I. C., Peterson, B. L., Barefoot, J. C., Harvin, S. H., Dahlstrom, W. G., Kaplan, B. H., Costa, P. T. J., & Williams, R. B. J. (1992a). Using college alumni populations in epidemiologic research:The UNC Heart Alumni Study. Journal of Clinical Epidemiology, 45, 1243–1250.

Siegler, I. C., Peterson, B. L., Barefoot, J. C., & Williams, R. B. (1992b). Hostility during late adoles-cence predicts coronary risk factors at mid-life. American Journal of Epidemiology, 136, 146–154.

Stevens, G., & Cho, J. H. (1985). Socioeconomic indexes and the new 1980 census occupationalclassifications scheme. Social Science Research, 14, 142–168.


Stevens, G., & Featherman, D. L. (1981). A revised socioeconomic index of occupational status. Social Science Research, 10, 364–395.

Strogatz, D. S., & James, S. A. (1986). Social support and hypertension among blacks and whites in arural, southern community. American Journal of Epidemiology, 124, 949–956.

Taylor, S. E., & Seeman, T. E. (1999). Psychosocial resources and the SES-health relationship. Annalsof the New York Academy of Sciences, 896, 210–225.

Vitaliano, P., Dougherty, C., & Siegler, I. C. (Eds.). (1994). Biopsychosocial risks for cardiovasculardisease in spouse caregivers of persons with Alzheimer’s disease. New York: Springer.

Vitaliano, P. P., Russo, J., Young, H. M., Teri, L., & Maiuro, R. D. (1991). Predictors of burden inspouse caregivers of individuals with Alzheimer’s disease. Psychology and Aging, 6, 392–401.

Vitaliano, P. P., Scanlan, J. M., Brummett, B., Barefoot, J., Zhang, J., Savage, M. V., & Siegler, I. C.(2001). Are the salutogenic effects of social supports modified by income? A test of an “added valuehypothesis”. Health Psychology, 20, 155–165.

Vitaliano, P. P., Scanlan, J. M., Zhang, J. P., Savage, M. V., Barefoot, J. C., Brummett, B. H., & Siegler,I. C. (1999). Are the salutogenic effects of social supports modified by income? Paper presented atthe American Psychosomatic Society, Vancouver, Canada.

Wilkinson, R. G. (1996). Unhealthy Societies: The Afflictions of Inequality. London: Routledge.Williams, R. B., Barefoot, J. C., Califf, R. M., Haney, T. L., Saunders, W. B., Pryor, D. B., Hlatky,

M. A., Siegler, I. C., & Mark, D. B. (1992). Prognostic importance of social and economic resourcesamong medically treated patients with angiographically documented coronary artery disease. Journal of the American Medical Association, 267, 520–524.


associations among social support, income, and symptoms of ...uncahs.org/articles/brummett 2003 int...

Documents