Context
We examined the relationship between socioeconomic status and response to treatment for depression among older adults.
Methods
Secondary analysis of pooled data from the open-label phase of 2 National Institute of Mental Health–funded clinical trials of nortriptyline hydrochloride or paroxetine combined with interpersonal psychotherapy (N = 248). We used Cox proportional hazards regression analyses to examine the association between socioeconomic status, indexed by census tract median annual household income and the subject’s educational attainment, and treatment response and remission according to the Hamilton Depression Rating Scale. The association between socioeconomic status and suicidality or depressed mood reported at each week of treatment was examined using repeated-measures generalized logit models.
Results
Subjects residing in middle-income census tracts were significantly more likely to respond to antidepressant treatment than subjects residing in low-income census tracts (adjusted hazard ratio, 1.80; 95% confidence interval [CI], 1.18-2.75]). Throughout the course of antidepressant treatment, subjects in the middle- and high-income census tracts were significantly less likely to report suicidal ideation (adjusted odds ratios, 0.48 [95% CI, 0.27-0.94] and 0.39 [95% CI, 0.16-0.94], respectively). No association was found between socioeconomic status and remission.
Conclusion
Residence in a low-income census tract is associated with a less favorable course of depression among older adults receiving a combination of pharmacologic and psychosocial treatment.
Socioeconomic inequalities in health have been observed for nearly 200 years1,2; on average, persons with lower socioeconomic status (SES) have elevated rates of morbidity and mortality relative to persons of higher SES.3,4 These inequalities have been demonstrated at all stages of life, from early childhood5 through old age,6 and for physical7 and mental disorders.8 An extensive literature exists about the influence of the components of SES—income,9 education,10 and occupation11—and the factors that mediate the association (eg, psychosocial environments,12-14 biological processes,15 and ethnicity16,17).
A number of studies have demonstrated the association between indicators of low SES and prevalence of depression.18-20 These social differentials, as demonstrated in a meta-analysis by Lorant et al,21 may be attributable to the tendency of individuals with lower SES to be at increased risk for onset of first depression and the likelihood of these individuals to have a more severe course of depression (in terms of episode duration and recurrence) compared with individuals with higher SES. Given the factors that are associated with these inequalities (eg, low income, lack of education, and stressful social environments), it seems reasonable to assume that these same factors may also vitiate the effects of pharmacologic and psychosocial interventions and produce social inequalities in response to treatment.22,23 However, research on social factors and treatment response has been inconclusive. In a series of studies published in the late 1960s and early 1970s, Rickels et al22,24 and Downing and Rickels25 found that low SES was associated with poorer response to pharmacotherapy. In contrast, another study26 found no association between SES and treatment outcomes. Although some research has suggested that high educational status is associated with positive clinical outcomes27,28 and improved social adjustment,29 other work did not.30 Findings about the effects of employment have also been inconsistent. One study28 found that employment was associated with improved outcomes, whereas other studies26,29 found no such association. Although low income is predictive of increased rates and persistence of depression,21 its association with treatment outcome is less clear.31
Integrating the findings of previous research on SES and antidepressant response is difficult in view of inconsistencies in the measures of SES used and differences in treatment modalities and settings across studies. For example, only 1 of these studies31 collected direct information on family income, whereas measures of educational status were variously based on number of years in school26 or whether an individual had graduated from high school27 or college.30 The findings of Rickels et al,22,24 Downing and Rickels,25 and others23,28 were based on treatments administered in multiple settings (ie, outpatient clinic, general medical practice, private psychiatric practices, and primary care clinics), with possible differences in the quality of care offered in particular practices or by particular physicians.
The present study builds on previous research by providing an examination of whether 2 core components of SES, income and education,32 have an effect on response to antidepressant treatment among a sample of older subjects receiving standardized care for major depressive disorder. We hypothesized that even in the context of highly controlled treatment, lower SES will be associated with a lower likelihood of treatment response and higher levels of dysphoria and suicidality during the course of treatment.
Subjects were participants in 1 of 2 studies funded by the National Institute of Mental Health and conducted at the University of Pittsburgh, Pittsburgh, Pa, on the maintenance treatment of late-life depression. The first of these studies investigated the efficacy of combined nortriptyline hydrochloride and interpersonal psychotherapy in 169 subjects 59 years and older.33 The second study evaluated the efficacy of combined paroxetine and interpersonal psychotherapy among 116 subjects 69 years and older.34 Both studies began with open-label (ie, nonblinded and standardized treatments within each study) short-term and continuation treatments for major depressive disorder before randomization to 1 of several maintenance treatment conditions. The present analyses are based on data from the short-term and continuation phases, which together lasted up to 26 weeks. Combining data from the 2 studies resulted in a demographically and clinically diverse sample.
The analysis sample for the current study included 248 subjects (145 from study 1 and 103 from study 2) whose census tract median annual household incomes were available from the US Census Bureau.35
Depressive symptoms were measured weekly with the 17-item Hamilton Rating Scale for Depression (HRSD).36 Scores on the HRSD were used to generate indicators of treatment efficacy for the present study as follows. Response to treatment was defined as HRSD scores of 10 or less for at least 3 consecutive weeks; remission of major depressive disorder was defined as HRSD scores of 6 or less for at least 3 consecutive weeks.37 The depressed mood and suicide items from the HRSD were used in analyses of differential treatment response at the symptom level. As core symptoms of depression, we believed these 2 items would give the best indication of whether response to treatment was associated with SES. For depressed mood (HRSD item 1), a rating of 0 (subject denies any feelings of sadness, hopelessness, helplessness, or worthlessness) was coded as not depressed. Subjects were coded as depressed if they reported gloomy attitude, pessimism, or hopelessness on questioning (rating of 1); reported these feelings spontaneously (rating of 2); or communicated them nonverbally (rating of 3). For suicidality (HRSD item 3), a rating of 0 (subject denied any degree of suicidality) or of 1 (subject said life is empty or not worth living) was coded as nonsuicidal. Subjects who reported recurrent thoughts of death or wishes to be dead (rating of 2), had active suicidal ideation (rating of 3), or had attempted suicide (rating of 4) were classified as suicidal.
Income and years of education were used as the 2 measures of SES in the current study. Income was based on the median annual household income of the census tracts in which subjects lived,38 categorized as less than $25 000 (reference category), $25 000 to $50 000, and more than $50 000. The 3 income groups correspond to tertiles of median household income in Allegheny County, Pennsylvania,35 the county of residence of 88% of the subjects. Census tract data have been used in many investigations of social inequalities in health in the United States.39-48 Information on subjects’ years of education was collected at baseline.
Demographic variables included in the analyses have been shown to be associated with depression, including age in years,49 sex,50 race/ethnicity (nonwhite vs white),16 and marital status (married, widowed, and separated/divorced/never married).51
The baseline clinical characteristics included in the analyses were first vs recurrent episode of depression, age at first lifetime onset of depression, duration in weeks of the current episode, the subject’s concurrent medical burden as measured by the Cumulative Illness Rating Scale–Geriatric,52 and the total 17-item HRSD score.
Cox proportional hazards regression analyses53 were used to evaluate the association of SES with time to response and time to remission. Data from nonresponders were censored at 26 weeks or at the time of nonresponse, ie, when they were in need of nonprotocol treatment. We modeled the association between SES and the presence of depressed mood and suicidality at each clinical assessment during the first 20 weeks of the trial using repeated-measures generalized logit regression.54 For these analyses, we used data from the first 20 weeks of the trial only because, after that point, the number of observations fell off sharply.
Regression models for time to response/remission, depressed mood, and suicidality were estimated sequentially. The first regression model included 2 indicator variables for income (middle and high income vs low income) and 3 indicator variables for education (12, 13-15, and ≥16 years vs <12 years). In addition to the income and education variables the second model included sociodemographic variables, and the third model included baseline clinical characteristics. Results of the unadjusted and adjusted models of treatment response are shown; for the remaining outcomes, only the final adjusted models are shown because the effects of SES did not vary substantively between models.
The demographic and baseline clinical characteristics of subjects in the 2 studies are presented in Table 1. On average, subjects in study 2 were older and had higher levels of education. Severity of depression was more pronounced among study 1 subjects, as indicated by higher levels of recurrent depression (an inclusion criterion for study 1), higher HRSD scores, and a higher prevalence of suicidality. Controls for demographic and baseline clinical characteristics in the multivariate models serve to adjust the effect estimates for SES differences in the samples enrolled in the 2 studies. In the pooled sample, the only significant difference in the demographic and baseline clinical characteristics of subjects in the different income groups was that the low-income group had a greater proportion of nonwhite subjects (χ22 = 10.42; P = .005).
Treatment response rates among study subjects in the low-, middle-, and high-income groups were 80% (32 subjects), 77% (122 subjects), and 78% (39 subjects), respectively. Median times to response (HRSD score of ≤10 for 3 weeks) were 9.1, 7.0, and 7.4 weeks in the low-, middle-, and high-income groups, respectively.
Unadjusted and adjusted hazard ratios (HRs) comparing the likelihood of treatment response among individuals in the middle- and high-income groups relative to those in the low-income group are shown in Table 2. Unadjusted HRs were obtained from separate models containing only income or education. These effects were then entered into a model together (adjusted model 1); subsequently, the effects of income and education were adjusted for demographic factors (adjusted model 2) and baseline clinical characteristics (adjusted model 3). In the unadjusted models and in adjusted model 1, income and education were not significantly associated with response to combined pharmacologic and psychosocial treatment. However, after accounting for demographic factors (model 2) and baseline clinical characteristics (model 3), we observed a significant difference in the likelihood of treatment response between subjects in the middle-income category relative to those in the low-income category. Model 3, which included baseline clinical characteristics in addition to the SES and demographic variables, revealed that middle-income subjects were significantly more likely to respond to treatment than were low-income subjects (HR, 1.80; 95% confidence interval [CI], 1.18-2.75) and that high-income subjects were marginally although not significantly more likely to respond to treatment (HR, 1.25; 95% CI, 0.76-2.05). Combining middle- and high-income subjects into a single category demonstrated that, in the aggregate, they were significantly more likely to respond to treatment than were low-income subjects (HR, 1.61; 95% CI, 1.07-2.42).
Other results included were that (1) years of education was not a significant covariate in any of the models and (2) in model 3, nonwhite race/ethnicity was significantly associated with greater likelihood of treatment response (HR, 2.02; 95% CI, 1.16-3.54).
Remission rates in the low-, middle-, and high-income groups were 48% (19 subjects), 47% (74 subjects), and 44% (22 subjects), respectively. As in the analyses for response, unadjusted and adjusted Cox proportional hazards models were estimated to examine the relationship between SES and remission of depressive symptoms (HRSD score of ≤6 for 3 consecutive weeks). Census tract income and education were not significantly associated with remission.
The final sets of analyses evaluated the association between SES and the HRSD items for suicide and depressed mood as recorded each week during the first 20 weeks of treatment. The weekly pattern of suicidality by income group is shown in the Figure. Although the percentage of subjects who endorsed the suicide item declined during the treatment period, suicidal ideation was consistently higher in the low-income group than in the middle- and high-income groups. Repeated-measures logistic analyses were used to quantify the average SES differences in suicidality during 20 weeks of treatment. Even after controlling for baseline clinical characteristics, middle- and high-income subjects were significantly less likely to endorse suicidal ideation than were low-income subjects (Table 3). The adjusted odds of suicidality for the middle-income group were 0.48 times that of the low-income group (95% CI, 0.27-0.94) and for the high-income group were 0.39 times that of the low-income group (95% CI, 0.16-0.94). These odds ratios suggest an inverse relationship in the association between income and suicidality during treatment for major depression. When the middle- and high-income groups were combined, the average odds of suicidality were 0.46 times that of the low-income group (95% CI, 0.24-0.87). Years of education was not related to suicidality during treatment.
Szanto et al34 reported that suicidal ideation is associated with a significantly lower likelihood of treatment response in the current sample. In view of this and the present findings that census tract median household income of subjects is associated with both suicidality and likelihood of response during treatment, we conducted several exploratory analyses to evaluate the extent to which suicidality contributes to the link between census tract median household income and treatment response. We performed these analyses by adding indicators to model 3 (Table 2) for suicidal ideation at baseline (HR, 1.04; 95% CI, 0.71-1.52) and at any time during the treatment period (HR, 0.75; 95% CI, 0.55-1.04). However, the HRs for income remained statistically significant and virtually unchanged. Together, these findings indicate that the association between income group and treatment response is not attributable to suicidal ideation, but rather both SES and suicidality independently predict the likelihood of responding to antidepressant therapy.
The repeated-measures analyses showed that the depressed mood item from the HRSD was not differentially endorsed by subjects in the 3 income groups (Table 3).
The analyses revealed 2 significant associations between SES (as indicated by census tract median household income) and response to treatment for major depressive disorder in older adults. First, subjects who resided in low-income census tracts were less likely than other subjects to respond to treatment. This effect was strongest for the comparison of low- and middle-income subjects. One possible explanation for this finding comes from research that demonstrated that the persistence of depressive episodes was more pronounced among individuals with lower SES.21 That is, if individuals with lower SES tend to have depressive episodes of greater duration, rates of difficult-to-treat depression may also be elevated in this population.
Although subjects who lived in middle- and high-income census tracts were, in the aggregate, more likely to respond to treatment than subjects who lived in low-income census tracts, we must note that the nonlinearity of the results contradicts our expectation that subjects who lived in high-income census tracts would be the most likely to respond. We are not able to provide a definitive explanation for this but can offer possible reasons. First, the imprecise correspondence between the census tract and individual measures of SES may have weakened the dose-response relationship. For example, some subjects of modest means may have become home owners before their neighborhoods underwent significant increases in property values and became characterized as high-income census tracts. In contrast, home ownership in a high-income neighborhood might reflect past SES but not present circumstances. Under either of these conditions, census tract information would not accurately reflect individual measures of SES and, thus, would weaken the association between household income and treatment response.
Census tract income was a significant predictor of response to treatment only when accounting for individual demographic and baseline clinical characteristics. Ultimately, data on individual and neighborhood-level SES will be needed to correctly distinguish the effects of neighborhoods from those of individual characteristics when considering predictors of response to antidepressant treatment.
The second association between SES and treatment outcomes was that, compared with subjects who lived in low-income census tracts, subjects who lived in middle- and high-income census tracts had significantly lower levels of suicidality during the clinical trials. Low-income subjects were about twice as likely as middle-income subjects and about 2½ times as likely as high-income subjects to report suicidal ideation during treatment, suggesting an inverse relationship between the median household income of the neighborhoods in which subjects resided and suicidality. Furthermore, an exploratory analysis provided no evidence that suicidal ideation mediated the association between residence in a low-income census tract and treatment response. Rather, income and suicidality were independently related to treatment response.
We did not observe an association between SES and the likelihood of remission. The relatively small proportion of subjects who achieved remission (48%, 47%, and 44% in the low-, middle-, and high-income groups, respectively) might account for this, but it is also possible that predictors of remission and response (as defined in this study) are not the same. Further research is necessary to answer this question.
To the extent that educational status usually serves as a principal component of SES, is generally correlated with income,10 and has been shown to be a predictor of response to treatment,27,29 we had expected it to be a significant covariate in our analyses. However, in our models, educational status was not a predictor of response to treatment or of endorsement of the HRSD suicide item. This might be explained by a cohort effect. It is only in recent decades that the economy of the Pittsburgh region has shifted from heavy industry to higher education, medicine, and services. However, when the older subjects in our study came of age, economic and social success in the industrial economy was not so dependent on education. Thus, we speculate that years of education is a less accurate measure of SES in this sample.
Racial/ethnic disparities in health have garnered a great deal of attention.55 Given that some of the worst disparities in health are seen as being embedded in the close relationship between minority status and low SES,17 we would have welcomed the opportunity to examine this relationship and its influence on treatment response. However, the small number (n = 19) of minority-status subjects in our sample did not permit this. The analyses did suggest that race/ethnicity and SES are not synonymous. To the contrary, being nonwhite appeared to be associated with a greater likelihood of response. This may reflect lower lifetime rates of depression among nonwhite subjects, as has been observed in national epidemiology surveys,8,18 or it may reflect that, although minority persons generally lack adequate care for depression, once they receive it, they respond well.56 Further research, with more representative samples, is needed to disentangle the independent and interdependent effects of SES and race/ethnicity17,57 on depression and response to treatment.
Although the research reported herein is broadly consistent with results of previous work,22-25 there are several limitations. First, the use of census tract data on median household incomes is potentially problematic insofar as this may be an inaccurate representation of an individual’s own income.58,59 Without data on individuals’ incomes, we cannot determine whether the associations between census tract median household income and response to treatment are due to the nature of the census tracts in which subjects lived, to the influence of subjects’ actual household incomes, or to factors correlated with both. However, area-based measures of SES have been shown to be fairly stable over time,59 have been shown to predict a range of health outcomes in general60-65 and for older adults in particular,64,66 and have been argued to provide valid approximations of social inequalities in health when individual data are not available.39,65,66 In particular, results from the Hampshire Depression Project23 are suggestive of neighborhood-level variation in antidepressant treatment. In that study, attendees of primary care practices that were located in low-SES neighborhoods were more likely to have depressive symptoms at baseline and 6 months.
Second, inferences about SES differences in response to treatment are constrained by the lack of a placebo arm. Thus, we cannot replicate the work of Rickels et al,22,24 who found relatively small differences between responses to antidepressant medication or placebo among low-SES patients in an outpatient clinic and relatively large medication-placebo differences among middle-SES general practice patients. However, given that their findings may have been biased by unknown factors in the different treatment settings, the question of whether SES patterns in response to active interventions are different from or simply mirror the “natural” course of depression more generally must await further investigation.
Third, one must be cautious in generalizing the results of these analyses. The psychological, biological, and social worlds of children, adolescents, and young, midlife, and older adults are all different, and, thus, we would expect that the relationship between SES and response to treatment, as well as the relative contributions of neighborhood effects and the characteristics of individuals,1 are likely to vary across age groups.
Despite its limitations, this study suggests that social worlds, in interaction with the demographic and clinical characteristics of individuals, may affect response to antidepressant treatment, even in clinical trials when subjects receive optimal pharmacologic and psychosocial treatments. Therefore, we suggest that future clinical trials routinely gather data on individual income, educational degrees earned, occupation,32,67 and aspects of the broader social environment such as social capital.68 However, to transform evidence into knowledge that will inform the treatment of depression, it is essential that future research examines all of the factors (eg, neighborhoods,69 stress,70 social support,71 race/ethnicity,57 or income inequalities72) that may mediate the association between SES and clinical outcomes.
Correspondence: Alex Cohen, PhD, Department of Social Medicine, Harvard Medical School, 641 Huntington Ave, Boston, MA 02115 (alex_cohen@hms.harvard.edu).
Submitted for Publication: November 29, 2004; final revision received May 14, 2005; accepted June 3, 2005.
Funding/Support: This study was supported in part by grants P30 MH52247, RO1 MH43832, RO1 MH37869, R25 MH60473, and P60 MD000207-03 from the National Institute of Mental Health, Rockville, Md.
1.MacIntyre
SEllaway
A Neighborhoods and health: an overview. In:Kawachi
IBerkman
LFeds. Neighborhoods and Health New York, NY Oxford University Press Inc2003;20- 42
2.Adler
NEBoyce
TChesney
MACohen
SFolkman
SKahn
RLSyme
SL Socioeconomic status and health: the challenge of the gradient.
Am Psychol 1994;4915- 24
PubMedGoogle ScholarCrossref 5.Gortmaker
SLWise
PH The first injustice: socioeconomic disparities, health services technology, and infant mortality.
Annu Rev Sociol 1997;23147- 170
PubMedGoogle ScholarCrossref 6.Breeze
EFletcher
AELeon
DAMarmot
MGClarke
RJShipley
MJ Do socioeconomic disadvantages persist into old age? self-reported morbidity in a 29-year follow-up of the Whitehall Study.
Am J Public Health 2001;91277- 283
PubMedGoogle ScholarCrossref 7.Marmot
MBobak
MSmith
GD Explanations for social inequalities in health. In:Amick
BC
IIILevine
STarlov
ARWalsh
DCeds. Society and Health New York, NY Oxford University Press Inc1995;172- 210
8.Kessler
RCMcGonagle
KAZhao
SNelson
CBHughes
MEshleman
SWittchen
HUKendler
KS Lifetime and 12-month prevalence of
DSM-III-R psychiatric disorders in the United States.
Arch Gen Psychiatry 1994;518- 19
PubMedGoogle ScholarCrossref 10.Winkleby
MAJatulis
DEFrank
EFortmann
SP Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease.
Am J Public Health 1992;82816- 820
PubMedGoogle ScholarCrossref 11.Marmot
MRyff
CDBumpass
LLShipley
MMarks
NF Social inequalities in health: next questions and converging evidence.
Soc Sci Med 1997;44901- 910
PubMedGoogle ScholarCrossref 12.Adler
NESnibbe
AC The role of psychosocial processes in explaining the gradient between socioeconomic status and health.
Curr Dir Psychol Sci 2003;12119- 123
Google ScholarCrossref 13.Williams
RB The role of psychosocial factors in human disease: lessons from animal models.
Acta Physiol Scand Suppl 1997;640100- 102
PubMedGoogle Scholar 14.Siegrist
JMarmot
M Health inequalities and the psychosocial environment—two scientific challenges.
Soc Sci Med 2004;581463- 1473
PubMedGoogle ScholarCrossref 15.McEwen
BSSeeman
T Protective and damaging effects of mediators of stress: elaborating and testing the concepts of allostasis and allostatic load.
Ann N Y Acad Sci 1999;89630- 47
PubMedGoogle ScholarCrossref 16.US Department of Health and Human Services, Mental Health: Culture, Race, and Ethnicity. Rockville, Md Office of the Surgeon General, Public Health Service, US Dept of Health and Human Services2001;
17.Mechanic
D Policy challenges in addressing racial disparities and improving population health.
Health Aff (Millwood) 2005;24335- 338
PubMedGoogle ScholarCrossref 18.Kessler
RCBerglund
PDemler
OJin
RKoretz
DMerikangas
KRRush
AJWalters
EEWang
PSNational Comorbidity Survey Replication, The epidemiology of major depressive disorder: results from the National Comorbidity Survey Replication (NCS-R).
JAMA 2003;2893095- 3105
PubMedGoogle ScholarCrossref 19.Roberts
RELee
ES Occupation and the prevalence of major depression, alcohol, and drug abuse in the United States.
Environ Res 1993;61266- 278
PubMedGoogle ScholarCrossref 20.Wilson
KCChen
RTaylor
SMcCracken
CFCopeland
JR Socio-economic deprivation and the prevalence and prediction of depression in older community residents: the MRC-ALPHA Study.
Br J Psychiatry 1999;175549- 553
PubMedGoogle ScholarCrossref 21.Lorant
VDeliege
DEaton
WRobert
APhilippot
PAnsseau
M Socioeconomic inequalities in depression: a meta-analysis.
Am J Epidemiol 2003;15798- 112
PubMedGoogle ScholarCrossref 23.Ostler
KThompson
CKinmonth
ALKPeveler
RCStevens
LStevens
A Influence of socio-economic deprivation on the prevalence and outcome of depression in primary care: the Hampshire Depression Project.
Br J Psychiatry 2001;17812- 17
PubMedGoogle ScholarCrossref 24.Rickels
KJenkins
BWZamostien
BRaab
EKanther
M Pharmacotherapy in neurotic depression: differential population responses.
J Nerv Ment Dis 1967;145475- 485
PubMedGoogle ScholarCrossref 25.Downing
RWRickels
K Predictors of response to amitriptyline and placebo in three outpatient treatment settings.
J Nerv Ment Dis 1973;156109- 129
PubMedGoogle ScholarCrossref 26.Keitner
GIRyan
CEMiller
IWNorman
WH Recovery and major depression: factors associated with twelve-month outcome.
Am J Psychiatry 1992;14993- 99
PubMedGoogle Scholar 27.Hirschfeld
RMRussell
JMDelgado
PLFawcett
JFriedman
RAHarrison
WMKoran
LMMiller
IWThase
MEHowland
RHConnolly
MAMiceli
RJ Predictors of response to acute treatment of chronic and double depression with sertraline or imipramine.
J Clin Psychiatry 1998;59669- 675
PubMedGoogle ScholarCrossref 28.Ronalds
CCreed
FStone
KWebb
STomenson
B Outcome of anxiety and depressive disorders in primary care.
Br J Psychiatry 1997;171427- 433
PubMedGoogle ScholarCrossref 29.Spillmann
MBorus
JSDavidson
KGWorthington
JJ
IIITedlow
JRFava
M Sociodemographic predictors of response to antidepressant treatment.
Int J Psychiatry Med 1997;27129- 136
PubMedGoogle ScholarCrossref 30.Perlis
RHAlpert
JNierenberg
AAMischoulon
DYeung
ARosenbaum
JFFava
M Clinical and sociodemographic predictors of response to augmentation, or dose increase among depressed outpatients resistant to fluoxetine 20 mg/day.
Acta Psychiatr Scand 2003;108432- 438
PubMedGoogle ScholarCrossref 31.Croughan
JLSecunda
SKKatz
MMRobins
EMendels
JSwann
AHarris-Larkin
B Sociodemographic and prior clinical course characteristics associated with treatment response in depressed patients.
J Psychiatr Res 1988;22227- 237
PubMedGoogle ScholarCrossref 32.Adler
NENewman
K Socioeconomic disparities in health: pathways and policies: inequality in education, income, and occupation exacerbates the gaps between the health “haves” and “have-nots.”
Health Aff (Millwood) 2002;2160- 76
PubMedGoogle ScholarCrossref 33.Reynolds
CF
IIIFrank
EPerel
JMImber
SDCornes
CMiller
MDMazumdar
SHouck
PRDew
MAStack
JAPollock
BGKupfer
DJ Nortriptyline and interpersonal psychotherapy as maintenance therapies for recurrent major depression: a randomized controlled trial in patients older than 59 years.
JAMA 1999;28139- 45
PubMedGoogle ScholarCrossref 34.Szanto
KMulsant
BHHouck
PDew
MAReynolds
CF
III Occurrence and course of suicidality during short-term treatment of late-life depression.
Arch Gen Psychiatry 2003;60610- 617
PubMedGoogle ScholarCrossref 37.Frank
EPrien
RFJarrett
RB
et al. Conceptualization and rationale for consensus definitions of terms in major depressive disorder: remission, recovery, relapse, and recurrence.
Arch Gen Psychiatry 1991;48851- 855
PubMedGoogle ScholarCrossref 38.US Census Bureau, Census 2000 Basics. Washington, DC US Census Bureau2002;
39.Krieger
NChen
JTWaterman
PDSoobader
MJSubramanian
SVCarson
R Choosing area based socioeconomic measures to monitor social inequalities in low birth weight and childhood lead poisoning: the Public Health Disparities Geocoding Project (US).
J Epidemiol Community Health 2003;57186- 199
PubMedGoogle ScholarCrossref 40.Krieger
NZierler
SHogan
JWWaterman
PDChen
JTLemieux
KGjelsvik
A Geocoding and measurement of neighborhood socioeconomic position: a US perspective. In:Kawachi
IBerkman
LFeds.
Neighborhoods and Health New York, NY Oxford University Press Inc2003;147- 178
Google Scholar 41.Perez-Smith
ASpirito
ABoergers
J Neighborhood predictors of hopelessness among adolescent suicide attempters: preliminary investigation.
Suicide Life Threat Behav 2002;32139- 145
PubMedGoogle ScholarCrossref 42.Cesaroni
GFarchi
SDavoli
MForastiere
FPerucci
CA Individual and area-based indicators of socioeconomic status and childhood asthma.
Eur Respir J 2003;22619- 624
PubMedGoogle ScholarCrossref 43.Schwartz
KLCrossley-May
HVigneau
FDBrown
KBanerjee
M Race, socioeconomic status and stage at diagnosis for five common malignancies.
Cancer Causes Control 2003;14761- 766
PubMedGoogle ScholarCrossref 44.Huang
SSFinkelstein
JARifas-Shiman
SLKleinman
KPlatt
R Community-level predictors of pneumococcal carriage and resistance in young children.
Am J Epidemiol 2004;159645- 654
PubMedGoogle ScholarCrossref 45.Boardman
JD Stress and physical health: the role of neighborhoods as mediating and moderating mechanisms.
Soc Sci Med 2004;582473- 2483
PubMedGoogle ScholarCrossref 46.Robert
SAStrombom
ITrentham-Dietz
AHampton
JMMcElroy
JANewcomb
PARemington
PL Socioeconomic risk factors for breast cancer: distinguishing individual- and community-level effects.
Epidemiology 2004;15442- 450
PubMedGoogle ScholarCrossref 47.Klassen
ACCurriero
FCHong
JHWilliams
CKulldorff
MMeissner
HIAlberg
AEnsminger
M The role of area-level influences on prostate cancer grade and stage at diagnosis.
Prev Med 2004;39441- 448
PubMedGoogle ScholarCrossref 48.Singh
GKMiller
BAHankey
BFEdwards
BK Persistent area socioeconomic disparities in US incidence of cervical cancer, mortality, stage, and survival, 1975-2000.
Cancer 2004;1011051- 1057
PubMedGoogle ScholarCrossref 49.Gildengers
AHouck
PMulsant
BHPollock
BGMaxumdar
SMiller
MDDew
MAFrank
EKupfer
DJReynolds
CF
III Course and rate of antidepressant response in the very old.
J Affect Disord 2002;69177- 184
PubMedGoogle ScholarCrossref 52.Miller
MDParadis
CFHouck
PRMazumdar
SStack
JARifai
AHMulsant
BReynolds
CF
III Rating chronic medical illness burden in geropsychiatric practice and research: application of the Cumulative Illness Rating Scale.
Psychiatry Res 1992;41237- 248
PubMedGoogle ScholarCrossref 53.Cox
DR Regression models and life-tables.
J R Stat Soc Ser B 1972;34187- 220
Google Scholar 55.Smedley
BDedStith
AYedNelson
ARed Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington, DC National Academy Press2003;
56.Miranda
JChung
JYGreen
BLKrupnick
JSiddique
JRevicki
DABelin
T Treating depression in predominantly low-income young minority women.
JAMA 2003;29057- 65
PubMedGoogle ScholarCrossref 58.Geronimus
ATBound
JNeidert
LJ On the validity of using census geocode characteristics to proxy individual socioeconomic characteristics.
J Am Stat Assoc 1996;91529- 537
Google ScholarCrossref 59.Geronimus
ATBound
J Use of census-based aggregate variables to proxy for socioeconomic group: evidence from national samples.
Am J Epidemiol 1998;148475- 486
PubMedGoogle ScholarCrossref 60.Soobader
MLeClere
FBHadden
WMaury
B Using aggregate geographic data to proxy individual socioeconomic status: does size matter?
Am J Public Health 2001;91632- 636
PubMedGoogle ScholarCrossref 61.Pickett
KEPearl
M Multilevel analyses of neighbourhood socioeconomic context and health outcomes: a critical review.
J Epidemiol Community Health 2001;55111- 122
PubMedGoogle ScholarCrossref 62.Ostir
GVEschbach
KMarkides
KSGoodwin
JS Neighbourhood composition and depressive symptoms among older Mexican Americans.
J Epidemiol Community Health 2003;57987- 992
PubMedGoogle ScholarCrossref 63.Krieger
NWaterman
PDChen
JTSoobader
MJSubramanian
SV Monitoring socioeconomic inequalities in sexually transmitted infections, tuberculosis, and violence: geocoding and choice of area-based socioeconomic measures—the Public Health Disparities Geocoding Project (US).
Public Health Rep 2003;118240- 260
PubMedGoogle Scholar 64.Krieger
NChen
JTWaterman
PDSoobader
MJSubramanian
SVCarson
R Geocoding and monitoring of US socioeconomic inequalities in mortality and cancer incidence: does the choice of area-based measure and geographic level matter? the Public Health Disparities Geocoding Project.
Am J Epidemiol 2002;156471- 482
PubMedGoogle ScholarCrossref 65.Krieger
NChen
JTWaterman
PDRehkopf
DHSubramanian
SV Race/ethnicity, gender, and monitoring socioeconomic gradients in health: a comparison of area-based socioeconomic measures—the Public Health Disparities Geocoding Project.
Am J Public Health 2003;931655- 1671
PubMedGoogle ScholarCrossref 66.Krieger
NGordon
D Re: “Use of census-based aggregate variables to proxy for socioeconomic group: evidence from national samples [comment].
Am J Epidemiol 1999;150892- 896
PubMedGoogle ScholarCrossref 69.Kawachi
IedBerkman
LFed Neighborhoods and Health. New York, NY Oxford University Press Inc2003;
70.Dohrenwend
BP The role of adversity and stress in psychopathology: some evidence and its implications for theory and research.
J Health Soc Behav 2000;411- 19
PubMedGoogle ScholarCrossref 71.Harris
TBrown
GWRobinson
R Befriending as an intervention for chronic depression among women in an inner city, I: randomised controlled trial.
Br J Psychiatry 1999;174219- 224
PubMedGoogle ScholarCrossref 72.Weich
SLewis
GJenkins
SP Income inequality and the prevalence of common mental disorders in Britain.
Br J Psychiatry 2001;178222- 227
PubMedGoogle ScholarCrossref