Abstract
Purpose:
Depressive symptoms relapse and remit over time, perhaps differentially by race and income. Few studies have examined whether time-varying depressive symptoms (TVDS) differentially predict mortality. We sought to determine whether race (white/black) and income (</≥$35,000) moderate the association between TVDS and mortality in a large cohort.
Methods:
The REGARDS study is a prospective cohort study among community-dwelling U.S. adults aged 45 years or older. Cox proportional hazard models were constructed to separately analyze the association between mortality (all cause, cardiovascular death, noncardiovascular death, and cancer death) and TVDS in race and income stratified models.
Results:
Point estimates were similar and statistically significant for white (aHR = 1.24 [95% CI: 1.10, 1.41]), black (aHR = 1.26 [95% CI: 1.11, 1.42]), and low-income participants (aHR = 1.28 [95% CI: 1.16, 1.43]) for the association between TVDS and mortality. High-income participants had a lower hazard (aHR = 1.19 [95% CI: 1.02, 1.38]). Baseline depressive symptoms predicted mortality in blacks only (aHR = 1.17, 95% CI: [1.00, 1.35]).
Conclusions:
We found that TVDS significantly increased the immediate hazard of mortality similarly across race and income strata. TVDS may provide more robust evaluations of depression impact compared with the baseline measures, making apparent racial disparities cited in the extant literature a reflection of the imperfection of using baseline measures.
Keywords: Depressive symptoms, REGARDS, Time-varying, Race, Income, Mortality, CVD mortality
Introduction
Research has shown that individuals with elevated depressive symptoms have an increased risk of cardiovascular disease and all-cause mortality (mortality).[1–7] Given that racial minority status and financial difficulty are associated with both depressive symptoms and poor health outcomes,[8–19] it is plausible that the relationship between depressive symptoms and mortality is moderated by race and income. However, observational studies have been largely inconsistent.[1–4,20–22] Recent literature suggests a higher hazard of mortality attributed to the baseline depressive symptoms in whites compared with blacks,[21,22] after adjusting for income, access/insurance, and health status.[22]
Results by income have similarly been inconsistent. Sumner et al. found that concurrent depressive symptoms and stress were associated with excess hazard of mortality and cardiovascular mortality in low-income individuals while adjusting for race and other risk factors.[6] Recent research suggests that the excess mortality attributed to depression may not differ by income.[23]
The extant literature is complex and inconsistent around the differential impact of depressive symptoms on mortality by race and income.[9,10,24–27] Given that depressive symptoms relapse and remit, particularly in racial minorities[9–11,28] and those experiencing financial strain,[12–19] time-varying analyses (TVA) have the potential to elucidate these relationships.[2] Our investigators demonstrated that time-varying depressive symptoms (TVDS) are associated with an increased hazard of all-cause, cardiovascular, and cancer mortality,[1–4] after adjusting for demographics, medical comorbidities, and behavioral and physiologic risk factors.[1,2] To our knowledge, few have examined time-varying associations between depressive symptoms and mortality through the lens of racial and income disparities.[1,4] Any assessment of race or income requires understanding the complex way in which race, socioeconomic status, and mental health interact to affect those outcomes[8,10,11]; and no prior study has further explored income interactions, and separately race interactions, in the relationship between depressive symptoms and mortality.
The Reasons for Geographical and Racial Differences in Stroke (REGARDS) study is one of the largest ongoing cohort studies of community-dwelling black and white individuals in the United States and has the potential to clarify time-varying associations and inconsistent findings around whether income and race moderate the relationship between depressive symptoms, cardiovascular disease (CVD), and mortality.[1–3,29,30] We sought to determine if there is an association between TVDS and mortality separately among black versus white and low-versus high-income participants. Because depression is associated with increased hazard of cardiovascular and all-cause mortality among those with coronary artery disease,[31–35] we further assessed whether race and income moderated the effect of TVDS on CVD and mortality among those with existing coronary heart disease (CHD). Such findings have the potential to improve our understanding of social determinants[36] of health and may have implications on treatment.
Methods
REGARDS is a prospective cohort study focused on cardiovascular outcomes among community-dwelling adults aged 45 years and older who live in the contiguous United States. The sample was balanced on race and oversampled individuals from the Southeast. Sampling procedures, inclusion and exclusion criteria, and other study methodology have been previously described.[37] Of note, participants who had an active cancer diagnosis at enrollment were excluded from the study. Participants were recruited and enrolled between January 2003 and October 2007 by mail through commercially available lists.
Written informed consent was obtained from all participants. The University of Alabama at Birmingham Institutional Review Board, and all participating institutions, approved the main REGARDS protocol. The University of Alabama at Birmingham and Weill Cornell Medicine Institutional Review Boards approved the ancillary study procedures.
Study procedures
Computer-assisted telephone interviews, self-administered questionnaires, and an in-home examination were used to collect baseline demographic, medical, and behavioral risk factor data.[37] At the baseline and subsequent follow-up periods, trained research staff administered telephone interviews to collect demographic data, medical history and risk factors, hospitalizations, and associated medical records.[37] The in-home examination collected physical measurements, medication inventory, and blood and urine specimens.[37] Median time between the baseline telephone interview and in-home examination was 28 (inter quartile range: 21, 42) days.
Primary outcomes
Four primary outcomes were considered for these analyses: mortality, CVD death, non–cardiovascular disease mortality (nonCVD), and cancer mortality (all body sites). NonCVD mortality included death from cancer, accidents, injury, suicide, homicide, liver disease, infection, sepsis related death, dementia, chronic obstructive pulmonary disease (COPD), pulmonary embolism, and other causes. CVD death included participants who died from coronary heart disease, stroke, heart failure, sudden cardiac death, vascular pathology, and other CVD causes. Deaths were ascertained by next-of-kin report, online agencies (e.g., Social Security Death Index) or the National Death Index. A panel of experts adjudicated each participant’s cause of death and CVD outcomes from death certificates; medical records; autopsy reports, baseline medical history, physiologic variables, and medications; adjudicated study outcomes; and next-of-kin reports.[38]
Depressive symptoms
The main predictor was elevated depressive symptoms defined by a score greater than or equal to 4 on the 4-item Center for Epidemiologic Studies Depression scale (CES-D-4).[39] The CES-D-4 uses a four-point (0–3) scale to record the presence and frequency of specific symptoms of depression experienced during the preceding week. Participants are asked to rate the number of days over the last week in which they had felt depressed, lonely, sad, and had crying spells. Response options included less than 1 day, 1–2 days, 3–4 days, and 5–7 days (0, 1, 2, and 3 points, respectively). Total scores range from 0 (no symptoms) to 12. As previously described, the Cronbach’s alpha for the CES-D-4 in the total sample was 0.80.[2] The reliability and validity of the CES-D-4 is similar to the original 20-item instrument.[39] Each participant could contribute up to 3 measurements of CES-D-4 over the follow-up period: baseline with the initial telephone call, and two additional telephone surveys approximately 5 and 7 years later.
Covariates
The baseline demographic data included self-reported age, gender, race (black or white), education (dichotomized as less than high school and high school graduate, and some college, college graduate and above), annual income (dichotomized as <$35,000 and ≥$35,000), insurance status (yes/no), and stroke region (categorized as stroke belt, stroke buckle, and nonbelt).[2,37] Stroke buckle was defined as coastal regions within the states of North Carolina, South Carolina, and Georgia. Stroke belt was defined as the states of Alabama, Arkansas, Louisiana, Mississippi, Tennessee, and the noncoastal regions within the states of North Carolina, South Carolina, and Georgia.[37] Clinical baseline characteristics included diabetes diagnosis, systolic and diastolic blood pressures (average of two standard measurements in mm Hg), measured body mass index, albumin to creatinine ratio (logarithmically transformed), high-density lipoprotein cholesterol, and total cholesterol.[40,41] Diabetes was defined as fasting blood glucose greater than or equal to 126 mg per dL, random glucose greater than 200 mg/dL, or oral hypoglycemic or insulin use. Baseline medication usage (yes/no) for aspirin, antidepressants (serotonin and norepinephrine reuptake inhibitors, selective serotonin reuptake inhibitors, and tricyclic antidepressants), statins, and antihypertensives were also included. COPD was defined as use of inhaled beta-2 adrenergic agonists, leukotriene inhibitors, inhaled corticosteroids, combination inhalers, or other pulmonary medications such as ipratropium, cromolyn, aminophylline, and theophylline.
Self-reported baseline clinical characteristics collected included history of CHD, self-reported stroke, peripheral vascular disease or aneurysm, cognitive impairment (on the 6-item screener of global cognitive function),[41] and chronic lung disease. History of CHD included the following CHD events: self-reported history of myocardial infarction, coronary revascularization procedure, or evidence of past myocardial infarction on ECG.[2,37]
Baseline behavioral characteristics included self-reported current cigarette usage, pack-years of cigarette usage, physical activity, alcohol use, and medication nonadherence. Physical activity was assessed through the question, “How many times per week do you engage in intense physical activity, enough to work up a sweat?” response options included: none, 1–3 times per week and 4 or more times per week. Alcohol use was queried by asking participants “How many alcoholic beverages do you drink?”; response options included none, moderate defined as 1 drink per day for women and 2 drinks per day for men.[37] Medication nonadherence was assessed with a 4-item medication adherence scale (≥1).[42] Two physiological risk factors were included in this analysis: high-sensitivity C-reactive protein and perceived stress. Perceived stress was ascertained from patient responses to the 4-item Perceived Stress Scale (score of ≥5 compared with <5).[43]
Statistical analyses
Baseline characteristics of participants with and without elevated depressive symptoms were compared using χ2 tests for categorical variables, Student’s t tests for normally distributed continuous variables, and Kruskal-Wallis tests for non-normally distributed continuous variables.
Depressive symptoms were included as time-varying covariates in all models.[2] Follow-up time was calculated from the date of the in-home examination to the earliest date of one of the following: death, last telephone follow-up, or censored at the assigned data set end date of December 31, 2012.
Cox proportional hazard regression models were constructed to separately analyze the association between TVDS (CES-D-4≥4) and mortality (each type of mortality examined separately) in models stratified by race and, separately income, overall models to test interactions. Models were constructed to control for demographic, clinical, behavioral, and psychological factors. The fully adjusted model adjusted forage, gender, education, health insurance, region, systolic blood pressure, high-density lipoprotein cholesterol, total cholesterol, aspirin usage, antihypertensives or statins, body mass index, albumin to creatinine ratio, diabetes, CVD (defined as stroke or heart disease), antidepressants, COPD, cognitive impairment, current smoking, self-reported alcohol use, physical inactivity, medication nonadherence, log-transformed C-reactive protein, and perceived stress. A formal test for interaction was performed between depressive symptoms and race and depressive symptoms and income in fully adjusted models. The interactions were found to be statistically significant (P < .1), for all mortality outcomes except CVD death. All models were stratified for race (black vs. white) and income (<$35 K and ≥$35 K).
Proportional hazards assumptions were assessed visually by examining the Schoenfeld and log-log plots and with a Schoenfeld residual test. All proportional hazards assumptions were met.
Missing data/multiple imputation
Few covariates had missing data (<1%). Cognitive impairment had 19.3% missing and medication adherence had 9.2% missing. To eliminate potential bias from complete case analyses, we performed multiple imputation with chained equations.[44] Equations were created to impute missing values and were composed of all variables used in the fully adjusted models.[44] Five imputations were computed based on current guidelines in the literature and computational power.[44] Diagnostic measures were performed to check the fitness of the generated data sets, including the use of the Stata command midiagplots.[45] All of the results presented in this article are from the multiply imputed data set.
Sensitivity analyses
Two types of sensitivity analyses were conducted: 1) to construct non–time-varying models that used baseline depressive symptoms and 2) to calculate TVDS models among participants with a history of CHD at the baseline. Baseline CHD was defined as either a self-reported history of myocardial infarction, electrocardiogram evidence of a myocardial infarction, or history of a percutaneous coronary intervention or coronary artery bypass surgery.[3]
Stata version 14.0 and SAS version 9.0 were used to perform the analyses reported in this article.
Results
Participant characteristics
The average age was 65.0 [standard deviation 9.4] years old; 55.1% were female; 41.1% black; 12.5% had less than a high school education (Tables 1 and 2). Participants with elevated depressive symptoms (vs. nonelevated depressive symptoms) at follow-up were significantly more likely to be female, be black, have less than a high-school education, and have an annual income less than $35,000. Baseline prevalence of depressive symptoms was higher among low-income (16.1%) compared with high-income (7.4%) and black (14.0%) participants. In addition, recurring depressive symptoms was higher among low-income individuals (19.2%) and blacks (17.0%). Among participants without baseline depressive symptoms, 5.6% (n = 1468) and 3.3% (n = 867) had newly elevated depressive symptoms at the first and second follow-up, respectively. Antidepressant usage was higher among whites (17.2%) than blacks (9.0%) and low-income (15.1%) than high-income (13.0%) individuals.
Table 1.
Baseline characteristics of REGARDS participants according to race and baseline depression
| Characteristics | Black (n = 12,124) | White (n = 17,353) | ||||
|---|---|---|---|---|---|---|
| CES-D<4 (n = 10,422) | CES-D ≥ 4 (n = 1702) | P-value | CES-D<4 (n = 15,802) | CES-D ≥ 4 (n = 1551) | P-value | |
| Sociodemographics | ||||||
| Age, M (SD) | 64.5 (9.2) | 62.8 (9.5) | <.001 | 65.7 (9.4) | 63.9 (10.0) | <.001 |
| Female, n (%) | 6326 (60.7%) | 1215 (71.4%) | <.001 | 7655 (48.4%) | 1041 (67.1%) | <.001 |
| Male, n (%) | 4096 (39.3%) | 487 (28.6%) | – | 8147 (51.6%) | 510 (32.9%) | – |
| Less than high-school education, n (%) | 1877 (18.0%) | 543 (31.9%) | <.001 | 1035 (6.5%) | 237 (15.3%) | <.001 |
| High-school education and above, n (%) | 8545 (82.0%) | 1159 (68.1%) | 14,767 (93.5%) | 1314 (84.7%) | – | |
| Annual household income, n (%) | – | – | <.001 | – | – | <.001 |
| <$35,000, n (%) | 5278 (50.6%) | 1172 (68.9%) | – | 5186 (32.8%) | 830 (53.5%) | – |
| ≥$35,000, n (%) | 5144 (49.4%) | 530 (31.1%) | – | 10,616 (67.2%) | 721 (46.5%) | – |
| No health insurance, n (%) | 934 (9.0%) | 242 (14.2%) | <.001 | 597 (3.8%) | 152 (9.8%) | <.001 |
| Health insurance, n (%) | 9475 (90.9%) | 1459 (85.7%) | – | 15,194 (96.2%) | 1398 (90.1%) | – |
| Missing, n (%) | 13 (0.1%) | 1 (0.1%) | – | 11 (0.1%) | 1 (0.1%) | – |
| Region, n (%) | – | – | <.001 | – | – | <.001 |
| Stroke belt* | 3366 (32.3%) | 663 (39.0%) | – | 5603 (35.5%) | 557 (35.9%) | – |
| Stroke buckle† | 1836 (17.6%) | 343 (20.2%) | – | 3599 (22.8%) | 408 (26.3%) | – |
| Nonstroke belt or buckle | 5220 (50.1%) | 696 (40.9%) | – | 6600 (41.8%) | 586 (37.8%) | – |
| General health and medical conditions | ||||||
| Self-reported general health, n (%) | – | – | <.001 | – | – | <.001 |
| Poor, fair, good | 6674 (64.0%) | 1404 (82.5%) | – | 6537 (41.4%) | 1118 (72.1%) | – |
| Excellent, very good | 3724 (35.7%) | 295 (17.3%) | – | 9236 (58.4%) | 430 (27.7%) | – |
| Missing | 24 (0.2%) | 3 (0.2%) | – | 29 (0.2%) | 3 (0.2%) | – |
| History of CHD‡ | 1487 (14.3%) | 360 (21.2%) | <.001 | 2993 (18.9%) | 363 (23.4%) | <.001 |
| Missing | 225 (2.2%) | 42 (2.5%) | – | 252 (1.6%) | 39 (2.5%) | – |
| Cardiovascular disease (CVD), n (%)§ | 2162 (20.7%) | 519 (30.5%) | <.001 | 3669 (23.2%) | 467 (30.1%) | <.001 |
| Diabetes, n (%)∥ | 2978 (28.6%) | 610 (35.8%) | <.001 | 2322 (14.7%) | 337 (21.7%) | <.001 |
| Missing | 460 (4.4%) | 67 (3.9%) | – | 507 (3.2%) | 51 (3.3%) | – |
| COPD, n (%) | 760 (7.3%) | 192 (11.3%) | <.001 | 1545 (9.8%) | 211 (13.6%) | <.001 |
| Physical component score on SF-12 scale, M (SD.) | 45.9 (10.3) | 39.8 (11.5) | <.001 | 47.8 (10.0) | 41.6 (12.9) | <.001 |
| Impaired cognitive status (cognitive score ≤ 4) | 911 (8.7%) | 240 (14.1%) | <.001 | 630 (4.0%) | 106 (6.8%) | <.001 |
| Missing | 2024 (19.4%) | 281 (16.5%) | – | 3150 (19.9%) | 226 (14.6%) | |
| Elevated perceived stress (PSS≥5) | 2937 (28.2%) | 1252 (73.6%) | <.001 | 3341 (21.1%) | 1056 (68.1%) | <.001 |
| Depressive symptoms | ||||||
| Depression at 5 y follow-up (CES-D-4) n = 20,925 | 652 (6.3%) | 400 (23.5%) | <.001 | 816 (5.2%) | 385 (24.8%) | <.001 |
| Missing | 3472 (33.3%) i | 750 (44.1%) | – | 3794 (24.0%) | 536 (34.6%) | – |
| Depression at 7 y follow-up (CES-D-4) n = 12,448 | 320 (3.1%) | 180 (10.6%) | <.001 | 547 (3.5%) | 210 (13.5%) | <.001 |
| Missing | 6500 (62.4%) | 1205 (70.8%) | – | 8335 (52.7%) | 989 (63.8%) | – |
| Physiological risk factors | ||||||
| Body mass index, kg/m2, M (SD) | 30.7 (6.6) | 31.6 (7.3) | <.001 | 28.2 (5.5) | 29.4 (6.6) | <.001 |
| Systolic blood pressure, mm Hg, M (SD) | 130.6 (17.1) | 131.8 (19.0) | .008 | 125.4 (15.7) | 125.3 (16.5) | .94 |
| Total cholesterol, mg/dL, M (SD) | 192.8 (40.8) | 193.9 (43.0) | .30 | 191.1 (39.1) | 195.4 (42.9) | <.001 |
| High-density lipoprotein, mg/dL, M (SD) | 53.4 (15.8) | 54.3 (17.0) | .028 | 50.7 (16.3) | 50.5 (15.3) | .73 |
| QT interval, corrected for heart rate, ms, M (SD) | 406.3 (23.6) | 410.3 (24.6) | <.001 | 407.7 (23.5) | 409.6 (23.5) | .003 |
| High-sensitivity C-reactive protein, mg/L, median, IQR | 2.8 (1.2, 6.3) | 3.7 (1.4, 8.3) | <.001 | 1.8 (0.8, 4.1) | 2.5 (1.1, 5.7) | <.001 |
| Albumin to creatinine ratio, mg/g, median, IQR | 7.9 (4.7,19.9) | 8.5 (5.0,24.2) | .002 | 7.1 (4.6,13.8) | 8.1 (5.1,16.7) | <.001 |
| Medications | ||||||
| Antihypertensive medication use, n (%) | 6537 (62.7%) | 1147 (67.4%) | <.001 | 6742 (42.7%) | 760 (49.0%) | <.001 |
| Missing, n (%) | 124 (1.2%) | 23 (1.4%) | – | 163 (1.0%) | 23 (1.5%) | – |
| Statin use, n (%) | 3005 (28.8%) | 527 (31.0%) | .066 | 5235 (33.1%) | 520 (33.5%) | <.001 |
| Missing, n (%) | 30 (0.3%) | 8 (0.5%) | – | 31 (0.2%) | 1 (0.1%) | – |
| Aspirin use, n (%) | 3963 (38.0%) | 700 (41.1%) | .015 | 7409 (46.9%) | 714 (46.0%) | <.001 |
| Missing, n (%) | 5 (<1%) | 0 (0.0%) | – | 9 (0.1%) | 2 (0.1%) | – |
| Antidepressant use, n (%) | 748 (7.2%) | 342 (20.1%) | <.001 | 2414 (15.3%) | 579 (37.3%) | <.001 |
| Missing, n (%) | 30 (0.3%) | 8 (0.5%) | – | 31 (0.2%) | 1 (0.1%) | – |
| Behavioral risk factors | ||||||
| Self-reported smoking, pack years, M (SD) | 10.0 (18.1) | 12.0 (20.5) | <.001 | 15.3 (25.1) | 12.2 (21.3) | <.001 |
| Current smoking, n (%) | 1654 (15.9%) | 441 (25.9%) | <.001 | 1806 (11.4%) | 358 (23.1%) | <.001 |
| Former smokers, n (%) | 3959 (38.0%) | 537 (31.6%) | – | 6774 (42.9%) | 563 (36.3%) | – |
| Never smokers, n (%) | 4757 (45.6%) | 717 (42.1%) | – | 7173 (45.4%) | 624 (40.2%) | – |
| Missing, n (%) | 52 (0.5%) | 7 (0.4%) | – | 49 (0.3%) | 6 (0.4%) | – |
| Alcohol use, n (%) | – | – | .18 | – | – | <.001 |
| Heavy | 236 (2.3%) | 50 (2.9%) | – | 807 (5.1%) | 79 (5.1%) | – |
| Moderate | 2582 (24.8%) | 403 (23.7%) | – | 6202 (39.2%) | 437 (28.2%) | – |
| None | 7348 (70.5%) | 1190 (69.9%) | – | 807 (5.1%) | 79 (5.1%) | – |
| Missing, n (%) | 256 (2.5%) | 59 (3.5%) | – | 227 (1.4%) | 35 (2.3%) | – |
| Physical inactivity, n (%) | 6546 (62.8%) | 897 (52.7%) | <.001 | 10,788 (68.3%) | 808 (52.1%) | <.001 |
| Missing, n (%) | 153 (1.5%) | 23 (1.4%) | – | 240 (1.5%) | 22 (1.4%) | – |
| Medication nonadherence, n (%) | 2709 (26.0%) | 593 (34.8%) | <.001 | 4107 (26.0%) | 545 (35.1%) | <.001 |
| Missing, n (%) | 1109 (10.6%) | 142 (8.3%) | – | 1354 (8.6%) | 99 (6.4%) | – |
Indicates a statistically significant P-value, ns, indicates a nonstatistically significant P-value.
IQR = interquartile range; M = mean; n = total number assuming no missing data; SD = standard deviation; PSS = Perceived Stress Scale.
Defined as a score ≥4 on the 4-item Center for Epidemiologic Studies Depression scale at one of the 2 follow-up periods, 5 and 7 y after enrollment in REGARDS.
Stroke belt defined as the states of Alabama, Arkansas, Louisiana, Mississippi, Tennessee, and the noncoastal regions within the states of North Carolina, South Carolina, and Georgia.
Stroke buckle defined as coastal regions within the states of North Carolina, South Carolina, and Georgia.
History of CHD was defined as either a self-reported history, echocardiogram evidence of a myocardial infarction, history of a percutaneous coronary intervention, or coronary artery bypass surgery.
CVD defined as baseline coronary heart disease, stroke, periphery artery disease, or aortic aneurism.
Diabetes defined as fasting blood glucose ≥126 or random glucose >200 mL per dL or oral hypoglycemic or insulin use.
Table 2.
Baseline characteristics of REGARDS participants according to income and baseline depressive symptoms
| Characteristics | Income <$35,000 (n = 12,466) | Income ≥$35,000 (n = 17,011) | ||||
|---|---|---|---|---|---|---|
| CES-D<4 (n = 10,464) | CES-D ≥ 4 (n = 2002) | P-value | CES-D<4 (n = 15,760) | CES-D ≥ 4 (n = 1251) | P-value | |
| Sociodemographics | ||||||
| Age, M (SD) | 67.4 (9.2) | 63.9 (9.7) | <.001 | 63.8 (9.2) | 62.4 (9.8) | <.001 |
| Female, n (%) | 6398 (61.1%) | 1410 (70.4%) | <.001 | 7583 (48.1%) | 846 (67.6%) | <.001 |
| Male, n (%) | 4066 (38.9%) | 592 (29.6%) | – | 8177 (51.9%) | 405 (32.4%) | – |
| African American, n (%) | 5278 (50.4%) | 1172 (58.5%) | <.001 | 5144 (32.6%) | 530 (42.4%) | <.001 |
| White, n (%) | 5186 (49.6%) | 830 (41.5%) | – | 10,616 (67.4%) | 721 (57.6%) | – |
| Less than high-school education, n (%) | 2050 (19.6%) | 602 (30.1%) | <.001 | 862 (5.5%) | 178 (14.2%) | <.001 |
| High-school education and above, n (%) | 8414 (80.4%) | 1400 (69.9%) | – | 14,898 (94.5%) | 1073 (85.8%) | – |
| No health insurance, n (%) | 997 (9.5%) | 313 (15.6%) | <.001 | 534 (3.4%) | 81 (6.5%) | <.001 |
| Health insurance, n (%) | 9459 (90.4%) | 1689 (84.4%) | – | 15,210 (96.5%) | 1168 (93.4%) | – |
| Missing, n (%) | 8 (0.1%) | 0 (0.0%) | – | 16 (0.1%) | 2 (0.2%) | – |
| Region, n (%) | – | – | <.001 | – | – | <.001 |
| Stroke belt* | 3835 (36.6%) | 790 (39.5%) | – | 5134 (32.6%) | 430 (34.4%) | – |
| Stroke buckle† | 2090 (20.0%) | 436 (21.8%) | – | 3345 (21.2%) | 315 (25.2%) | – |
| Nonstroke belt or buckle | 4539 (43.4%) | 776 (38.8%) | – | 7281 (46.2%) | 506 (40.4%) | – |
| General health and medical conditions | ||||||
| Self-reported general health, n (%) | – | – | <.001 | – | – | <.001 |
| Poor, fair, good | 6427 (61.4%) | 1630 (81.4%) | – | 6784 (43.0%) | 892 (71.3%) | – |
| Excellent, very good | 4016 (38.4%) | 369 (18.4%) | – | 8944 (56.8%) | 356 (28.5%) | – |
| Missing | 21 (0.2%) | 3 (0.1%) | – | 32 (0.2%) | 3 (0.2%) | – |
| History of CHD‡ | 2044 (19.5%) | 495 (24.7%) | <.001 | 2436 (15.5%) | 228 (18.2%) | <.001 |
| Missing | 188 (1.8%) | 50 (2.50%) | – | 289(1.8%) | 31 (2.5%) | – |
| Cardiovascular disease (CVD), n (%)§ | 2785 (26.6%) | 661 (33.0%) | <.001 | 3046 (19.3%) | 325 (26.0%) | <.001 |
| Diabetes, n (%)∥ | 2673 (25.5%) | 642 (32.1%) | <.001 | 2627 (16.7%) | 305 (24.4%) | <.001 |
| Missing | 433 (4.1%) | 80 (4.0%) | – | 534 (3.4%) | 38 (3.0%) | – |
| COPD, n (%) | 1005 (9.6%) | 249 (12.4%) | <.001 | 1300 (8.2%) | 154 (12.3%) | <.001 |
| Physical component score on SF-12 scale, M (SD) | 44.6 (10.9) | 39.1 (11.8) | <.001 | 48.7 (9.3) | 43.2 (12.6) | <.001 |
| Impaired cognitive status (cognitive score ≤ 4) | 830 (7.9%) | 240 (12.0%) | <.001 | 711 (4.5%) | 106 (8.5%) | <.001 |
| Missing | 2304 (22.0%) | 359 (17.9%) | – | 2870 (18.2%) | 148 (11.8%) | – |
| Elevated perceived stress (PSS ≥ 5) | 3026 (28.9%) | 1488 (74.3%) | <.001 | 3252 (20.6%) | 820 (65.5%) | <.001 |
| Depressive Symptoms | ||||||
| Depression at 5-y follow-up (CES-D-4) n = 20,925 | 723 (6.9%) | 485 (24.2%) | <.001 | 745 (4.7%) | 300 (24.0%) | <.001 |
| Missing | 3471 (33.2%) | 878 (43.9%) | – | 3795 (24.1%) | 408 (32.6%) | – |
| Depression at 7-y follow-up (CES-D-4) n = 12,448 | 392 (3.7%) | 257 (12.8%) | <.001 | 475 (3.0%) | 133 (10.6%) | <.001 |
| Missing | 6333 (60.5%) | 1393 (69.6%) | – | 8502 (53.9%) | 801 (64.0%) | – |
| Physiological risk factors | ||||||
| Body mass index, kg/m2, M (SD) | 29.6 (6.5) | 30.8 (7.4) | <.001 | 28.9 (5.7) | 30.1 (6.5) | <.001 |
| Systolic blood pressure, mm Hg, M (SD) | 130.0 (17.2) | 130.2 (18.6) | .71 | 125.7 (15.7) | 126.4 (17.1) | .17 |
| Total cholesterol, mg/dL, M (SD) | 192.4 (41.4) | 195.1 (44.7) | .011 | 191.3 (38.6) | 193.8 (40.0) | .032 |
| High-density lipoprotein, mg/dL, M (SD) | 51.7 (15.9) | 52.3 (16.4) | .12 | 51.8 (16.4) | 52.8 (16.2) | .040 |
| QT interval, corrected for heart rate, ms, M (SD.) | 408.8 (24.6) | 410.7 (24.6) | .001 | 406.1 (22.7) | 408.8 (23.1) | <.001 |
| High-sensitivity C-reactive protein, mg/L, median, IQR | 2.6 (1.1, 5.9) | 3.4 (1.3, 7.7) | <.001 | 1.9 (0.8, 4.3) | 2.5 (1.1, 5.9) | <.001 |
| Albumin to creatinine ratio, mg/g, median, IQR | 8.6 (5.1,20.1) | 9.1 (5.3, 23.0) | .031 | 6.7 (4.4, 13.4) | 7.4 (4.7, 15.6) | <.001 |
| Medications | ||||||
| Antihypertensive medication use, n (%) | 6007 (57.4%) | 1253 (62.6%) | <.001 | 7272 (46.1%) | 654 (52.3%) | <.001 |
| Missing, n (%) | 126 (1.2%) | 23 (1.1%) | – | 161 (1.0%) | 23 (1.8%) | – |
| Statin use, n (%) | 3324 (31.8%) | 645 (32.2%) | .67 | 4916 (31.2%) | 402 (32.1%) | .49 |
| Missing, n (%) | 21 (0.2%) | 6 (0.3%) | – | 40 (0.3%) | 3 (0.2%) | – |
| Aspirin use, n (%) | 4556 (43.5%) | 900 (45.0%) | .24 | 6816 (43.2%) | 514 (41.1%) | .14 |
| Missing, n (%) | 7 (0.1%) | 1 (<1%) | – | 7 (<1%) | 1 (0.1%) | – |
| Antidepressant use, n (%) | 1330 (12.7%) | 550 (27.5%) | <.001 | 1832 (11.6%) | 371 (29.7%) | <.001 |
| Missing, n (%) | 21 (0.2%) | 6 (0.3%) | – | 40 (0.3%) | 3 (0.2%) | – |
| Behavioral risk factors | ||||||
| Self-reported smoking, pack years, M (SD.) | 15.0 (24.9) | 16.6 (26.0) | .010 | 12.1 (21.2) | 13.7 (23.0) | .010 |
| Current smoking, n (%) | 1746(16.7%) | 569 (28.4%) | <.001 | 1714 (10.9%) | 230 (18.4%) | <.001 |
| Former smokers, n (%) | 4153 (39.7%) | 631 (31.5%) | – | 6580 (41.8%) | 469 (37.5%) | – |
| Never smokers, n (%) | 4532 (43.3%) | 798 (39.9%) | – | 7398 (46.9%) | 543 (43.4%) | – |
| Missing, n (%) | 33 (0.3%) | 4 (0.2%) | – | 68 (0.4%) | 9 (0.7%) | – |
| Alcohol use, n (%) | – | – | .14 | – | – | <.001 |
| Heavy | 288 (2.8%) | 69 (3.4%) | – | 755 (4.8%) | 60 (4.8%) | – |
| Moderate | 2542 (24.3%) | 460 (23.0%) | – | 6242 (39.6%) | 380 (30.4%) | – |
| None | 7431 (71.0%) | 1410 (70.4%) | – | 8483 (53.8%) | 780 (62.4%) | – |
| Missing, n (%) | 203 (1.9%) | 63 (3.1%) | – | 280(1.8%) | 31 (2.5%) | – |
| Physical inactivity, n (%) | 6357 (60.8%) | 1006 (50.2%) | <.001 | 10,977 (69.7%) | 699 (55.9%) | <.001 |
| Missing, n (%) | 182 (1.7%) | 21 (1.0%) | – | 211 (1.3%) | 24 (1.9%) | – |
| Medication nonadherence, n (%) | 2780 (26.6%) | 719 (35.9%) | <.001 | 4036 (25.6%) | 419 (33.5%) | <.001 |
| Missing, n (%) | 882 (8.4%) | 147 (7.3%) | – | 1581 (10.0%) | 94 (7.5%) | – |
IQR = interquartile range; M = mean; n = total number assuming no missing data; PSS = Perceived Stress Scale; SD = standard deviation.
Indicates a statistically significant P-value.
Indicates a nonstatistically significant P-value.
Defined as a score ≥4 on the 4-item Center for Epidemiologic Studies Depression scale at one of the 2 follow-up periods, 5 and 7 y after enrollment in REGARDS.
Stroke belt defined as the states of Alabama, Arkansas, Louisiana, Mississippi, Tennessee, and the noncoastal regions within the states of North Carolina, South Carolina, and Georgia.
Stroke buckle defined as coastal regions within the states of North Carolina, South Carolina, and Georgia.
History of CHD was defined as either a self-reported history, echocardiogram evidence of a myocardial infarction, history of a percutaneous coronary intervention, or coronary artery bypass surgery.
CVD defined as baseline coronary heart disease, stroke, periphery artery disease, or aortic aneurism.
Diabetes defined as fasting blood glucose ≥126 or random glucose >200 mL per dL or oral hypoglycemic or insulin use.
Mortality outcomes stratified by race
TVDS were significantly associated with all-cause mortality in both black (aHR = 1.26, [95% CI: 1.11, 1.42]) and white participants (aHR = 1.24, 95% CI: [1.10, 1.41]) (Table 3). For CVD death, TVDS in a fully adjusted model was not statistically significant in blacks (aHR = 1.17, 95% CI: [0.95, 1.43]) or whites (aHR = 1.07, 95% CI: [0.86, 1.35]). In a fully adjusted model for nonCVD mortality, TVDS were statistically significant for both blacks (aHR = 1.31, 95% CI: [1.13, 1.53]) and whites (aHR = 1.33, 95% CI [1.15,1.53]). Cancer mortality, the largest subgroup of nonCVD death, was evaluated as a separate outcome; however, none of the fully adjusted models for blacks or whites were statistically significant (Table 3).
Table 3.
Association of elevated time-varying depressive symptoms with mortality, stratified by race and income, for REGARDS participants
| Overall (n = 29,477) | Black (n = 12,124) | White (n = 17,353) | Income <$35,000 (n = 12,466) | Income ≥$35,000 (n = 17,011) | |
|---|---|---|---|---|---|
| HR (95%CI) for categorical CES-D (score ≥4 vs. <4) | |||||
| All-cause mortality | |||||
| Events, n | 3660 | 2008 | 1652 | 2147 | 1513 |
| Crude | 1.66 (1.53–1.80) | 1.56 (1.39–1.74) | 1.71 (1.53–1.91) | 1.39 (1.26–1.53) | 1.65 (1.43–1.90) |
| Fully adjusted* | 1.25 (l.15–1.37) | 1.26 (1.11–1.42) | 1.24 (1.10–1.41) | 1.28 (1.16–1.43) | 1.19 (1.02–1.38) |
| Interaction term P-value† (depression × race) | – | 0.68 | – | ||
| Interaction term P-value† (depression × income) | – | – | 0.76 | ||
| CVD death | |||||
| Events, n | 1628 | 801 | 827 | 972 | 656 |
| Crude | 1.61 (1.40–1.84) | 1.57 (1.31–1.89) | 1.53 (1.25–1.88) | 1.36 (1.15–1.60) | 1.49 (1.16–1.93) |
| Fully adjusted* | 1.14 (0.98–1.33) | 1.17 (0.95–1.43) | 1.07 (0.86–1.35) | 1.18 (0.99–1.42) | 1.05 (0.79–1.38) |
| Interaction term P-value† (depression × race) | – | 0.54 | – | ||
| Interaction term P-value† (depression × income) | – | – | 0.77 | ||
| NonCVD death | |||||
| Events, n | 3137 | 1823 | 1314 | 1859 | 1278 |
| Crude | 1.69 (1.53–1.86) | 1.55 (1.35–1.78) | 1.79 (1.57–2.05) | 1.41 (1.25, 1.59) | 1.72 (1.46, 2.04) |
| Fully adjusted* | 1.31 (1.18–1.46) | 1.31 (1.13–1.53) | 1.33 (1.15–1.53) | 1.35 (1.18, 1.54) | 1.25 (1.05, 1.50) |
| Interaction term P-value† (depression × race) | – | 0.37 | – | ||
| Interaction term P-value† (depression × income) | – | – | 0.63 | ||
| Cancer death | |||||
| Events, n | 1461 | 846 | 615 | 753 | 708 |
| Crude | 1.25 (1.07–1.47) | 1.18 (0.94–1.49) | 1.30 (1.05–1.62) | 1.18 (0.98–1.43) | 1.05 (0.78–1.39) |
| Fully adjusted* | 1.12 (0.94–1.32) | 1.13 (0.89–1.45) | 1.12 (0.89–1.42) | 1.24 (1.01–1.53) | 0.89 (0.66–1.22) |
| Interaction term P-value† (depression × race) | – | 0.65 | – | ||
| Interaction term P-value† (depression × income) | – | – | 0.19 | ||
Each participant contributes to up to 3 time-varying CES-D measures. End of follow-up, December 31, 2012.
All results presented are from multiply imputed models.
SF-12 = Short-Form Health Survey.
Fully adjusted model includes sociodemographics (age, gender, region, income, health insurance, education, race, or income depending on the model of interest), medical conditions, physiological factors, and medication use (systolic blood pressure, total cholesterol, high-density lipoprotein cholesterol, use of aspirin, statins, antihypertensives, antidepressants, body mass index, logarithmically transformed albumin to creatinine ratio; diabetes, cardiovascular disease, medication use as a proxy for chronic obstructive pulmonary disease, and cognitive impairment), behavioral risk factors (pack-years of cigarette smoking, self-reported alcohol use, physical inactivity, medication nonadherence), and other factors (physical health component score of SF-12, log-transformed high-sensitivity C-reactive protein, and perceived stress).
Interaction term P-value is from the fully adjusted model and consists of depression and race or depression and income depending on the model of interest.
Mortality outcomes stratified by income
TVDS was significantly associated with mortality in low-income (aHR = 1.28, 95% CI: [1.16, 1.43]) and high-income individuals (aHR = 1.19,95% CI: [1.02,1.38]) (Table 3). For CVD death, the overall and the income stratified models were not statistically significant (Table 3). TVDS were found to be statistically significant for nonCVD death for both low-income (aHR = 1.35, 95% CI [1.18, 1.54]) and high-income individuals (aHR = 1.25, 95% CI [1.05, 1.50]). None of the fully adjusted models for cancer mortality and income were statistically significant (Table 3).
Baseline depressive symptom sensitivity analysis
Baseline depressive symptoms was associated with mortality among black participants in fully adjusted models (aHR = 1.17, 95% CI: [1.00,1.35]), but not white (aHR = 1.02 95% CI: [0.87,1.19]), high-income (aHR = 1.08 95% CI: [0.88,1.31]) or low-income (aHR = 1.09 95% CI: [0.96, 1.23]) individuals (Table 5). For nonCVD mortality, only black participants were found to have a significant association with baseline depressive symptoms (aHR = 1.25, 95% CI: [1.02, 1.51]). Findings for CVD and cancer mortality were not significant.
Table 5.
Association of elevated time-varying depressive symptoms with mortality, stratified by race and income, among REGARDS participants with a history of CHD at the baseline
| Overall (n = 5203) | Blacks (n = 1847) | Whites (n = 3356) | Income <$35,000 (n = 2539) | Income ≥$35,000 (n = 2664) | |
|---|---|---|---|---|---|
| HR (95%CI) for categorical CES-D (score ≥4 vs. <4) | |||||
| All-cause mortality | |||||
| Events, n | 1203 | 445 | 758 | 702 | 501 |
| Crude | 1.48 (1.29, 1.69) | 1.27 (1.03, 1.56) | 1.63 (1.37, 1.94) | 1.31 (1.12, 1.54) | 1.45 (1.13, 1.84) |
| Fully adjusted* | 1.20 (l.04, 1.40) | 1.20 (0.95, 1.53) | 1.23 (1.01, 1.49) | 1.22 (1.02, 1.47) | 1.15 (0.87, 1.50) |
| Interaction term P-value† (depression × race) | 0.42 | – | |||
| Interaction term P-value† (depression × income) | – | 0.67 | |||
| CVD death | |||||
| Events, n | 661 | 279 | 382 | 385 | 276 |
| Crude | 1.41 (1.15, 1.72) | 1.22 (0.90, 1.65) | 1.52 (1.15, 2.00) | 1.25 (0.98, 1.59) | 1.28 (0.86, 1.91) |
| Fully adjusted* | 1.11 (0.88, 1.39) | 1.10 (0.78, 1.56) | 1.09 (0.80, 1.49) | 1.14 (0.87, 1.49) | 0.97 (0.63, 1.51) |
| Interaction term P-value† (depression × race) | 0.82 | – | |||
| Interaction term P-value† (depression × income) | – | 0.98 | |||
| NonCVD death | |||||
| Events, n | 825 | 263 | 562 | 476 | 349 |
| Crude | 1.53 (1.29, 1.82) | 1.32 (0.99, 1.75) | 1.72 (1.38, 2.14) | 1.37 (1.11, 1.70) | 1.56 (1.15, 2.12) |
| Fully adjusted* | 1.28 (1.05, 1.56) | 1.29 (0.94, 1.80) | 1.33 (1.04, 1.71) | 1.29 (1.02, 1.65) | 1.27 (0.90, 1.79) |
| Interaction term P-value† (depression × race) | 0.39 | – | |||
| Interaction term P-value† (depression × income) | – | 0.64 | |||
| Cancer | |||||
| Events, n | 351 | 112 | 239 | 158 | 193 |
| Crude | 1.09 (0.80, 1.49) | 1.11 (0.68, 1.83) | 1.12 (0.75, 1.67) | 1.10 (0.76, 1.60) | 0.85 (0.46, 1.57) |
| Fully adjusted* | 1.03 (0.73, 1.45) | 1.16 (0.67, 2.05) | 1.00 (0.65, 1.55) | 1.18 (0.78, 1.79) | 0.76 (0.39, 1.46) |
| Interaction term P-value† (depression × ace) | 0.97 | – | |||
| Interaction term P-value† (depression × income) | – | 0.37 | |||
Each participant contributes to up to 3 time-variant CES-D measures. End of follow-up, December 31, 2012.
SF-12, Short-Form Health Survey.
All results presented are from multiply imputed models.
Fully adjusted model includes sociodemographics (age, gender, region, income, health insurance, education, race, or income depending on the model of interest), medical conditions, physiological factors, and medication use (systolic blood pressure, total cholesterol, high-density lipoprotein cholesterol, use of aspirin, statins, antihypertensives, antidepressants, body mass index, logarithmically transformed albumin to creatinine ratio; diabetes, cardiovascular disease, medication use as a proxy for chronic obstructive pulmonary disease, and cognitive impairment), behavioral risk factors (pack-years of cigarette smoking, self-reported alcohol use, physical inactivity, medication nonadherence), and other factors (physical health component score of SF-12, log-transformed high-sensitivity C-reactive protein, and perceived stress).
Interaction term P-value is from the fully adjusted model and consists of depression and race or depression and income depending on the model of interest.
CHD sensitivity analysis
Among individuals with a history of CHD, the association between TVDS and mortality was not statistically significant for black participants (aHR = 1.20, 95% CI: [0.95, 1.53]). For white participants, TVDS predicted mortality in fully adjusted models (aHR = 1.23, 95% CI: [1.01,1.49]) (Table 6). Although the point estimates are similar for all-cause mortality among low-income (aHR = 1.22, 95% CI: [1.02, 1.47]) and high-income participants (aHR = 1.15, 95% CI: [0.87, 1.50), only the low-income participants were statistically significant.
Table 6.
Number and type of events that compose the non-cardiovascular disease (nonCVD) outcome in the REGARDS study
| Causes of death | Overall | Race | Income | ||
|---|---|---|---|---|---|
| N (%) | Black N (%) | White N (%) | <$35,000 N (%) | ≥$35,000 N (%) | |
| Cancer | 1226 (44.3%) | 508 (43.5%) | 718 (44.9%) | 656 (42.6%) | 570 (46.5%) |
| Accidents/injury/suicide/homicide | 164 (5.9%) | 55 (4.7%) | 109 (6.8%) | 86 (5.6%) | 78 (6.4%) |
| Liver disease | 56 (2.0%) | 25 (2.1%) | 31 (1.9%) | 32 (2.1%) | 24 (2.0%) |
| Infection | 498 (18.0%) | 213 (18.2%) | 285 (17.8%) | 273 (17.7%) | 225 (18.3%) |
| ESRD | 118 (4.3%) | 75 (6.4%) | 43 (2.7%) | 78 (5.1%) | 40 (3.3%) |
| Dementia | 187 (6.8%) | 75 (6.4%) | 112 (7.0%) | 100 (6.5%) | 87 (7.1%) |
| COPD | 247 (8.9%) | 87 (7.4%) | 160 (10.0%) | 160 (10.4%) | 87 (7.1%) |
| Pulmonary embolism | 38 (1.34%) | 27 (2.1%) | 11 (0.7%) | 26 (1.7%) | 12 (1.0%) |
| Other | 232 (8.4%) | 103 (8.8%) | 129 (8.1%) | 128 (8.3%) | 104 (8.5%) |
| Missing | 263 | 126 | 137 | 165 | 98 |
For white participants, TVDS were significantly associated with nonCVD mortality (aHR = 1.33, 95% CI: [1.04,1.70]) but not among black participants (aHR = 1.29, 95% CI: [0.94,1.80]), although point estimates were similar. Similarly, TVDS were associated with nonCVD mortality among low-income participants (aHR = 1.29, 95% CI: [1.02, 1.65]) but not for high-income participants (aHR = 1.27, 95% CI: [0.90,1.79]). The effects of TVDS on CVD and cancer mortality were not statistically significant regardless of the race or income strata (Table 6).
Conclusion
In a large cohort of diverse, middle-aged adults across the United States, we found that while baseline depressive symptoms appear to predict mortality in blacks only, TVDS were associated with an increased hazard of mortality similarly among black, white, low- and high-income participants. To our knowledge, this is the first study to use TVA to examine whether race and income moderate the relationship between depressive symptoms and mortality, allowing for more accurate, powered analyses. The literature on whether race and income influence the relationship between depressive symptoms and mortality have been inconsistent, with some suggesting that depressive symptoms predict mortality in whites and not blacks.[21,22] Depression relapses, remits, and often persists; prior studies have poorly accounted for temporality of depression symptoms.[11,31,46] Recent research have called for TVA in examining the relationship between depressive symptoms and mortality.[47–50] Our findings demonstrate that depressive symptoms are associated with mortality across race and income, including those with CHD.
Effects on mortality also appeared to be largely driven by nonCVD death, comprised mostly cancer deaths, among those with and without CHD. The mortality risk we demonstrate across race and income strata suggests unmeasured factors may be contributing to the risk of nonCVD mortality, including hormonal and thrombotic factors.[51] In addition, we cannot say for certain whether discrepancies in our results reflect long-term risk. Our nonsignificant findings for CVD mortality are consistent with prior research suggesting behavioral risk factors mediate the relationship between depressive symptoms and CVD mortality.[52,53] Future research is needed to elucidate the risk between depressive symptoms and nonCVD mortality, which we were underpowered to assess by subgroup (Table 3).
Our baseline analyses suggest that depressive symptoms significantly predict mortality in blacks.[5,6,50,54] Prior inconsistencies around the moderating role of race in the extant literature may have been attributed to small sample sizes underpowered to assess subgroups or differential adjustment for medical, behavioral, and physiologic factors. Examining baseline analyses through the lens of TVA suggest effects in blacks may be driven by the lower depression treatment rates we found in blacks compared with whites. Several theories have emerged to examine the high hazard of mortality in certain groups. Habituation to stressful environments including unsafe neighborhoods, low-paying jobs, and food insecurity plus differential access to treatment might be both protective and chronically deleterious to mortality.[6,25,55,56] Gender and social support differences are protective for mortality, which may differ by race and socioeconomic status.[25,55,56] Research is needed to elucidate how these factors differentially influence individuals with elevated depressive symptoms.
This study has several strengths and limitations. To our knowledge, this is the largest cohort used to study the associations between TVDS and mortality stratified by race and income. All deaths were adjudicated by experts studying medical records and data from the National and Social Security Death Index. Because REGARDS is a prospective cohort study, we were able to collect follow-up CES-D-4 scores from participants, which strengthened the data for analysis. Owing to the sampling structure used to recruit participants, our findings might not be generalizable. Another limitation is our use of the CES-D-4 measure.[57,58] Other measures might be more robust, yet the CES-D-4 is still widely used and cited allowing for our study to be compared with others.[39,59,60] In addition, this study might be underpowered. The number of cancer deaths was low, especially after stratification, which might be attributed to the design of REGARDS which excludes active cancer patients.[37,59,61] Despite excluding active cancer participants at the baseline, we were unable to elucidate how receiving a subsequent cancer diagnosis influenced depressive symptom trajectories and mortality outcomes.
Our results contribute to the literature by elucidating how race and income interact with depressive symptoms to affect mortality. More research is needed to examine the association between TVDS in other subcohorts and to elucidate effects on nonCVD mortality. In addition, research is needed to formally ascertain how behavioral, medical, and other factors differentially mediate the relationship between depressive symptoms and mortality.
Supplementary Material
Table 4.
Association of elevated baseline depressive symptoms with mortality, stratified by race and income, for REGARDS participants
| Overall (n = 29,477) | Black (n = 12,124) | White (n = 17,353) | Income <$35,000 (n = 12,466) | Income >$35,000 (n = 17,011) | |
|---|---|---|---|---|---|
| HR (95%CI) for categorical CES-D (score >4 vs. <4) | |||||
| All-cause mortality | |||||
| Events, n | 770 | 355 | 415 | 545 | 225 |
| Crude | 1.17 (1.06, 1.29) | 1.20 (1.05, 1.36) | 1.17 (1.02, 1.35) | 1.02 (0.92, 1.15) | 1.23 (1.03, 1.46) |
| Fully adjusted* | 1.09 (0.98, 1.22) | 1.17 (1.00, 1.35) | 1.02 (0.87, 1.19) | 1.09 (0.96, 1.23) | 1.08 (0.88, 1.31) |
| Interaction term P-value† (depression × race) | 0.11 | – | |||
| Interaction term P-value† (depression × income) | – | 0.42 | |||
| CVD death | |||||
| Events, n | 338 | 144 | 194 | 252 | 86 |
| Crude | 1.23 (1.05, 1.45) | 1.33 (1.08, 1.64) | 1.10 (0.85, 1.43) | 1.04 (0.86, 1.26) | 1.31 (0.95, 1.79) |
| Fully adjusted* | 1.09 (0.91, 1.31) | 1.17 (0.92, 1.50) | 0.95 (0.71, 1.26) | 1.08 (0.87, 1.34) | 1.14 (0.79, 1.63) |
| Interaction term P-value† (depression × race) | 0.04 | ||||
| Interaction term P-value† (depression × income) | – | 0.49 | |||
| NonCVD death | |||||
| Events, n | 580 | 262 | 318 | 420 | 160 |
| Crude | 1.14 (1.01, 1.29) | 1.16 (0.98, 1.38) | 1.16 (0.97, 1.38) | 1.01 (0.87, 1.17) | 1.20 (0.96, 1.50) |
| Fully adjusted* | 1.11 (0.97, 1.28) | 1.25 (1.02, 1.51) | 1.02 (0.84, 1.24) | 1.11 (0.94, 1.31) | 1.08 (0.84, 1.39) |
| Interaction term P-value† (depression × race) | 0.36 | – | |||
| Interaction term P-value† (depression × income) | – | 0.74 | |||
| Cancer death | |||||
| Events, n | 226 | 95 | 131 | 155 | 71 |
| Crude | 0.89 (0.73, 1.09) | 0.88 (0.66, 1.17) | 0.95 (0.71, 1.26) | 0.87 (0.68, 1.10) | 0.74 (0.49, 1.01) |
| Fully adjusted* | 0.97 (0.78, 1.21) | 1.05 (0.76, 1.44) | 0.92 (0.67, 1.26) | 1.00 (0.77, 1.30) | 0.81 (0.51, 1.26) |
| Interaction term P-value† (depression × race) | 0.91 | – | |||
| Interaction term P-value† (depression × income) | – | 0.41 | |||
End of follow-up, December 31, 2012.
All results presented are from multiply imputed models.
SF-12 = Short-Form Health Survey.
Fully adjusted model includes sociodemographics (age, gender, region, income, health insurance, education, race, or income depending on the model of interest), medical conditions, physiological factors, and medication use (systolic blood pressure, total cholesterol, high-density lipoprotein cholesterol, use of aspirin, statins, antihypertensives, antidepressants, body mass index, logarithmically transformed albumin to creatinine ratio; diabetes, cardiovascular disease, medication use as a proxy for chronic obstructive pulmonary disease, and cognitive impairment), behavioral risk factors (pack-years of cigarette smoking, self-reported alcohol use, physical inactivity, medication nonadherence), and other factors (physical health component score of SF-12, log-transformed high-sensitivity C-reactive protein, and perceived stress).
Interaction term P-value is from the fully adjusted model and consists of depression and race or depression and income depending on the model of interest.
Acknowledgments
The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions.
This research project is supported by a cooperative agreement U01 NS041588 and cofounded by the National Institute of Neurological Disorders and Stroke, National Institutes of Health (NIH), Department of Health and Human Service. The REGARDS-MI study was supported by NIH grants R01 HL080477 and K24 HL111154. This work was also supported by funds from the NIH grant K23 HL121144. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or the NIA. Representatives of the NINDS were involved in the review of the manuscript but were not directly involved in the collection, management, analysis, or interpretation of the data. The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at https://www.uab.edu/soph/regardsstudy/.
Footnotes
Uncited table
References
- 1.Moise N, Khodneva Y, Richman J, Shimbo D, Kronish I, Safford MM. Elucidating the association between depressive symptoms, coronary heart disease, and stroke in black and white adults: the reasons for geographic and racial differences in stroke (REGARDS) Study. J Am Heart Assoc 2016;5(8):e003767. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Moise N, Khodneva Y, Jannat-Khah D, Richman J, Kronish IM, Shaffer J, et al. An observational study of the differential impact of time-varying depressive symptoms on all-cause and cause-specific mortality by health status in community dwelling adults: The REGARDS study. BMJ Open 2017;8(1): e017385. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ye S, Muntner P, Shimbo D, et al. Behavioral mechanisms, elevated depressive symptoms, and the risk for myocardial infarction or death in individuals with coronary heart disease. J Am Coll Cardiol 2013;61(6):622–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Cummings DM, Kirian K, Howard G, et al. Consequences of comorbidity of elevated stress and/or depressive symptoms and incident cardiovascular outcomes in diabetes: results from the reasons for geographic and racial differences in stroke (REGARDS) Study. Diabetes Care 2016;39(1):101–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Talala KM, Huurre TM, Laatikainen TKM, Martelin TP, Ostamo AI, Prättälä RS. The contribution of psychological distress to socio-economic differences in cause-specific mortality: a population-based follow-up of 28 years. BMC Public Health 2011;11(1):138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Sumner JA, Khodneva Y, Muntner P, et al. Effects of concurrent depressive symptoms and perceived stress on cardiovascular risk in low- and high-income participants: findings from the reasons for geographical and racial differences in stroke (REGARDS) study. J Am Heart Assoc 2016;5(10):e003930. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Van der Kooy K, van Hout H, Marwijk H, Marten H, Stehouwer C, Beekman A. Depression and the risk for cardiovascular diseases: systematic review and meta analysis. Int J Geriatr Psychiatry 2007;22(7):613–26. [DOI] [PubMed] [Google Scholar]
- 8.Assari S Social determinants of depression: the intersections of race, gender, and socioeconomic status. Brain Sci 2017;7(12):156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Barry LC, Thorpe RJ, Penninx BWJH, et al. Race-related differences in depression onset and recovery in older persons over time: the health, aging, and body composition study. Am J Geriatr Psychiatry 2014;22(7):682–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Cohen CI, Goh KH, Gustave M. A prospective study of outcome and predictors of subclinical and clinical depression in an older biracial sample of psychiatric outpatients. J Affect Disord 2010;121(3):204–11. [DOI] [PubMed] [Google Scholar]
- 11.González HM, Tarraf W, Whitfield KE, Vega WA. The epidemiology of major depression and ethnicity in the United States. J Psychiatr Res 2010;44(15): 1043–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Krsteska R, Pejoska VG. The association of poor economic condition and family relations in childhood with late-life depression. Psychiatr Danub 2013;25(3): 241–7. http://www.ncbi.nlm.nih.gov/pubmed/24048391. [Accessed 13 September 2018]. [PubMed] [Google Scholar]
- 13.Mota Pereira J Financial crisis increases the risk of depression relapse. J Psychiatr Res 2015;61:235–6. [DOI] [PubMed] [Google Scholar]
- 14.Najman JM, Hayatbakhsh MR, Clavarino A, Bor W, O’Callaghan MJ, Williams GM. Family poverty over the early life course and recurrent adolescent and young adult anxiety and depression: a longitudinal study. Am J Public Health 2010;100(9):1719–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Mehta K, Kramer H, Durazo-Arvizu R, Cao G, Tong L, Rao M. Depression in the US population during the time periods surrounding the great recession. J Clin Psychiatry 2015;76(4):e499–504. [DOI] [PubMed] [Google Scholar]
- 16.Tani Y, Fujiwara T, Kondo N, Noma H, Sasaki Y, Kondo K. Childhood socioeconomic status and onset of depression among japanese older adults: The JAGES prospective cohort study. Am J Geriatr Psychiatry 2016;24(9):717–26. [DOI] [PubMed] [Google Scholar]
- 17.Melchior M, Chastang J-F, Leclerc A, Ribet C, Rouillon F. Low socioeconomic position and depression persistence: longitudinal results from the GAZEL cohort study. Psychiatry Res 2010;177(1–2):92–6. [DOI] [PubMed] [Google Scholar]
- 18.Nicolosi GT, Falcäo DV, Batistoni SST, et al. Depressive symptoms in old age: relations among sociodemographic and self-reported health variables. Int Psychogeriatr 2011;23(06):941–9. [DOI] [PubMed] [Google Scholar]
- 19.Schaakxs R, Comijs HC, van der Mast RC, Schoevers RA, Beekman ATF, Penninx BWJH. Risk factors for depression: differential across age? Am J Geriatr Psychiatry 2017;25(9):966–77. [DOI] [PubMed] [Google Scholar]
- 20.Weinberger AH, Gbedemah M, Martinez AM, Nash D, Galea S, Goodwin RD. Trends in depression prevalence in the USA from 2005 to 2015: widening disparities in vulnerable groups. Psychol Med 2017;48:1308–15. [DOI] [PubMed] [Google Scholar]
- 21.Assari S, Moazen-Zadeh E, Lankarani MM, Micol-Foster V. Race, depressive symptoms, and all-cause mortality in the United States. Front Public Heal 2016;4:40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Assari S, Lankarani MM, Burgard S. Black-white difference in long-term predictive power of self-rated health on all-cause mortality in United States . Ann Epidemiol 2016;26(2):106–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Brandão DJ, Fontenelle LF, da Silva SA, Menezes PR, Pastor-Valero M. Depression and excess mortality in the elderly living in low- and middle-income countries: systematic review and meta-analysis. Int J Geriatr Psychiatry 2019;34(1):22–30. [DOI] [PubMed] [Google Scholar]
- 24.Houle JN. Depressive symptoms and all-cause mortality in a nationally representative longitudinal study with time-varying covariates. Psychosom Med 2013;75(3):297–304. [DOI] [PubMed] [Google Scholar]
- 25.Beydoun MA, Beydoun HA, Mode N, et al. Racial disparities in adult all-cause and cause-specific mortality among us adults: mediating and moderating factors. BMC Public Health 2016;16(1):1113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.McDougall GJ, Morgan S, Vaughan PW. Sixteen-month evaluation of depressive symptomatology in older adults. Arch Psychiatr Nurs 2012;26(2):e13–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Brown C, Bromberger JT, Schott LL, Crawford S, Matthews KA. Persistence of depression in African American and Caucasian women at midlife: findings from the Study of Women Across the Nation (SWAN). Arch Womens Ment Health 2014;17(6):549–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Glymour MM, Yen JJ, Kosheleva A, Moon JR, Capistrant BD, Patton KK. Elevated depressive symptoms and incident stroke in Hispanic, African-American, and White older Americans. J Behav Med 2012;35(2):211–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Kronish IM, Carson AP, Davidson KW, Muntner P, Safford MM. Depressive symptoms and cardiovascular health by the American Heart Association’s definition in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study. PLoS One 2012;7(12):e52771. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.McQueen A, Tower S, Sumner W. Interviews with “vapers”: implications for future research with electronic cigarettes. Nicotine Tob Res 2011;13(9):860–7. [DOI] [PubMed] [Google Scholar]
- 31.Leung YW, Flora DB, Gravely S, Irvine J, Carney RM, Grace SL. The impact of premorbid and postmorbid depression onset on mortality and cardiac morbidity among patients with coronary heart disease. Psychosom Med 2012;74(8):786–801. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Egede LE, Nietert PJ, Zheng D. Depression and all-cause and coronary heart disease mortality among adults with and without diabetes. Diabetes Care 2005;28(6):1339–45. http://www.ncbi.nlm.nih.gov/pubmed/15920049. [Accessed 13 September 2018]. [DOI] [PubMed] [Google Scholar]
- 33.Stenman M, Holzmann MJ, Sartipy U. Association between preoperative depression and long-term survival following coronary artery bypass surgery — A systematic review and meta-analysis. Int J Cardiol 2016;222:462–6. [DOI] [PubMed] [Google Scholar]
- 34.Dao TK, Chu D, Springer J, et al. Clinical depression, posttraumatic stress disorder, and comorbid depression and posttraumatic stress disorder as risk factors for in-hospital mortality after coronary artery bypass grafting surgery. J Thorac Cardiovasc Surg 2010;140(3):606–10. [DOI] [PubMed] [Google Scholar]
- 35.Versteeg H, Hoogwegt MT, Hansen TB, Pedersen SS, Zwisler A-D, Thygesen LC. Depression, not anxiety, is independently associated with 5-year hospitalizations and mortality in patients with ischemic heart disease. J Psychosom Res 2013;75(6):518–25. [DOI] [PubMed] [Google Scholar]
- 36.Havranek EP, Mujahid MS, Barr DA, et al. Social determinants of risk and outcomes for cardiovascular disease. Circulation 2015;132(9):873–98. [DOI] [PubMed] [Google Scholar]
- 37.Howard VJ, Cushman M, Pulley L, et al. The reasons for geographic and racial differences in stroke study: objectives and design. Neuroepidemiology 2005;25(3):135–43. [DOI] [PubMed] [Google Scholar]
- 38.Halanych JH, Shuaib F, Parmar G, et al. Agreement on cause of death between proxies, death certificates, and clinician adjudicators in the reasons for geographic and racial differences in stroke (REGARDS) study. Am J Epidemiol 2011;173(11):1319–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Melchior LA, Huba GJ, Brown VB, Reback CJ. A short depression index for women. Educ Psychol Meas 1993;53(4):1117–25. [Google Scholar]
- 40.Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12(3):189–98. http://www.ncbi.nlm.nih.gov/pubmed/1202204. [Accessed 7 February 2018]. [DOI] [PubMed] [Google Scholar]
- 41.Callahan CM, Unverzagt FW, Hui SL, Perkins AJ, Hendrie HC. Six-item screener to identify cognitive impairment among potential subjects for clinical research. Med Care 2002;40(9):771–81. [DOI] [PubMed] [Google Scholar]
- 42.Morisky DE, Green LW, Levine DM. Concurrent and predictive validity of a self-reported measure of medication adherence. Med Care 1986;24(1):67–74. http://www.ncbi.nlm.nih.gov/pubmed/3945130. [Accessed 7 February 2018]. [DOI] [PubMed] [Google Scholar]
- 43.Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J Health Soc Behav 1983;24(4):385–96. http://www.ncbi.nlm.nih.gov/pubmed/6668417. [Accessed 8 August 2014]. [PubMed] [Google Scholar]
- 44.White IR, Royston P, Wood AM. Multiple imputation using chained equations: issues and guidance for practice. Stat Med 2011;30(4):377–99. [DOI] [PubMed] [Google Scholar]
- 45.Eddings W, Marchenko Y, Eddings W, Marchenko Y. Diagnostics for multiple imputation in Stata. Stata J 2012;12(3):353–67. http://econpapers.repec.org/article/tsjstataj/v_3a12_3ay_3a2012_3ai_3a3_3ap_3a353-367.htm. [Accessed 8 September 2017]. [Google Scholar]
- 46.Gilsanz P, Walter S, Tchetgen Tchetgen EJ, et al. Changes in depressive symptoms and incidence of first stroke among middle-aged and older US adults. JAm Heart Assoc 2015;4(5):e001923. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Szpakowski N, Bennell MC, Qiu F, et al. Clinical impact of subsequent depression in patients with a New Diagnosis of Stable Angina. Circ Cardiovasc Qual Outcomes 2016;9(6):731–9. [DOI] [PubMed] [Google Scholar]
- 48.Win S, Parakh K, Eze-Nliam CM, Gottdiener JS, Kop WJ, Ziegelstein RC. Depressive symptoms, physical inactivity and risk of cardiovascular mortality in older adults: the Cardiovascular Health Study. Heart 2011;97(6):500–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Sullivan MD, O’Connor P, Feeney P, et al. Depression predicts all-cause mortality: epidemiological evaluation from the ACCORD HRQL substudy. Diabetes Care 2012;35(8):1708–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Finlay AK, Oliva EM, Timko C, Moos RH, Cronkite R. Predictors of 30-year mortality in depressed and comparison samples. J Affect Disord 2014;165: 114–9. [DOI] [PubMed] [Google Scholar]
- 51.Burg MM, Shimbo D, Shaffer J, et al. The ‘Perfect Storm’ and acute coronary syndrome onset: do psychosocial factors play a role? Prog Cardiovasc Dis 2013;55(6):601–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Lett H, Ali S, Whooley M. Depression and cardiac function in patients with stable coronary heart disease: findings from the heart and soul study. Psychosom Med 2008;70(4):444–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Elderon L, Whooley MA. Depression and cardiovascular disease. Prog Cardiovasc Dis 2013;55(6):511–23. [DOI] [PubMed] [Google Scholar]
- 54.Briley M, Lépine M. The increasing burden of depression. Neuropsychiatr Dis Treat 2011;7(Suppl 1):3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Peters A, McEwen BS. Stress habituation, body shape and cardiovascular mortality. Neurosci Biobehav Rev 2015;56:139–50. [DOI] [PubMed] [Google Scholar]
- 56.McNally RJ. Resolving the vulnerability paradox in the cross-national prevalence of posttraumatic stress disorder. J Anxiety Disord 2018;54:33–5. [DOI] [PubMed] [Google Scholar]
- 57.Ware J, Kosinski M, Keller SD. A 12-item short-form health survey: construction of scales and preliminary tests of reliability and validity. Med Care 1996;34(3): 220–33. http://www.ncbi.nlm.nih.gov/pubmed/8628042. [Accessed 7 February 2018]. [DOI] [PubMed] [Google Scholar]
- 58.Schulz R, Drayer RA, Rollman BL. Depression as a risk factor for non-suicide mortality in the elderly. Biol Psychiatry 2002;52(3):205–25. http://www.ncbi.nlm.nih.gov/pubmed/12182927. [Accessed 18 February 2018]. [DOI] [PubMed] [Google Scholar]
- 59.Pinquart M, Duberstein PR. Depression and cancer mortality: a meta-analysis. Psychol Med 2010;40(11):1797–810. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Radloff LS. The CES-D scale. Appl Psychol Meas 1977;1(3):385–401. [Google Scholar]
- 61.Wulsin LR, Vaillant GE, Wells VE. A systematic review of the mortality of depression. Psychosom Med 1999;61(1):6–17. http://www.ncbi.nlm.nih.gov/pubmed/10024062. [Accessed 18 February 2018]. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.