Unadjusted rates of significant depressive symptoms (ie, Center for Epidemiologic Studies Depression Scale score ≥5) within 0 to 2 years of each new medical condition.
Polsky D, Doshi JA, Marcus S, Oslin D, Rothbard A, Thomas N, Thompson CL. Long-term Risk for Depressive Symptoms After a Medical Diagnosis. Arch Intern Med. 2005;165(11):1260-1266. doi:10.1001/archinte.165.11.1260
This study examines the risk of development of significant depressive symptoms after a new diagnosis of cancer, diabetes, hypertension, heart disease, arthritis, chronic lung disease, or stroke.
The study used 5 biennial waves (1992-2000) of the Health and Retirement Study to follow a sample of 8387 adults (aged 51 to 61 years and without significant depressive symptoms in 1992) from 1994 to 2000. Time-dependent Cox regression models estimated adjusted hazard ratios (HRs) for an episode of significant depressive symptoms after a new diagnosis for each of the 7 medical conditions.
Within 2 years of initial diagnosis, subjects with cancer had the highest hazard of depressive symptoms (HR, 3.55; 95% confidence interval [CI], 2.79-4.52), followed by subjects with chronic lung disease (HR, 2.21; 95% CI, 1.64-2.79) and heart disease (HR, 1.45; 95% CI, 1.09-1.93). The hazard for depressive symptoms for most of these diseases decreased over time; however, subjects with heart disease continued to have a higher risk for depressive symptoms even 2 to 4 years and 4 to 8 years after diagnosis, and a significantly higher hazard for depressive symptoms developed for persons with arthritis 2 to 4 years after diagnosis (HR, 1.46; 95% CI, 1.11-1.92).
The findings identify several high-risk patient groups who might benefit from depression screening and monitoring to improve health outcomes in this vulnerable population facing new medical illnesses.
Despite its heavy disease burden, depression is often unrecognized and undertreated.1- 7 Almost one third to half of all depressed individuals in primary health care situations fail to receive a diagnosis from their physicians.2,5,6 Because treatment of major depression has been shown to decrease morbidity and mortality,8- 10 to reduce use of health care resources and costs,11 and to improve work productivity and quality of life,12- 14 a better understanding of the onset of depressive symptoms and major depressive episodes could lead to earlier diagnosis and higher rates of treatment.
Among the various factors that may influence the onset of depressive symptoms, the contribution of medical comorbidities has received a great deal of attention during the past decade. The increase in the risk for depression may be directly or indirectly related to the biological, psychological, or social impact of medical illness. Research has indicated a positive association between depression symptoms and medical conditions such as diabetes,15,16 stroke,17 myocardial infarction,18,19 congestive heart failure,20 and cancer.21
The literature assessing the impact of medical conditions on depression has 3 important limitations. First, most studies have limited their focus on the association of depressive symptoms to a single specific medical condition. However, the association may be confounded owing to other coexisting medical illnesses. Second, most of the studies are cross-sectional and do not allow us to elucidate the causal order of the relationship between the medical condition and depression.22- 25 The few longitudinal studies available have limited their assessments of depression incidence to a short time frame, and so we have no understanding of whether certain chronic medical conditions are associated with the risk of immediate or distant onset of depressive symptoms. To our knowledge, no study has examined the timing of multiple new medical diagnoses and the onset of depressive symptoms. Finally, few studies have examined this issue in patients in late middle age who are at a stage in life where the incidence of chronic and acute medical conditions starts an upward trend.
This report uses a prospective longitudinal survey of a nationally representative sample to examine the risk and correlates of development of significant depressive symptoms during an 8-year period after a new diagnosis of cancer, diabetes, hypertension, heart disease, arthritis, chronic lung disease, or stroke among adults aged 51 to 61 years.
Our study cohort comes from the Health and Retirement Study, a nationally representative, longitudinal survey of age-eligible respondents and their spouses sponsored by the National Institute on Aging, Bethesda, Md, and conducted by the Institute for Social Research at the University of Michigan, Ann Arbor. The purpose of the survey was to explore the relationship between health and retirement decisions. The study was begun in 1992 and used a complex survey sampling design to enroll noninstitutionalized adults in 48 US states who were born from 1931 through 1941, with oversampling of black and Hispanic subjects and Florida residents. The study also interviewed the spouses of married individuals. A sample of 12 652 subjects (9772 age-eligible respondents and 2880 non–age-eligible spouses) was drawn in 1992, and these subjects have been followed up prospectively with follow-up interview waves every 2 years. For patients who died between interview waves, exit interviews were conducted with a proxy. Extensive information on each respondent was collected ranging from demographics, health status, housing, family structure, work history and current employment, retirement plans, income, and health and life insurance. The detailed survey sampling procedures and contents have been described elsewhere.26
Our current study uses data from 5 biennial waves of the Health and Retirement Study from 1992 (wave 1) to 2000 (wave 5). A prospective cohort design is used to follow up respondents who were aged 51 to 61 years in 1992 (n = 9772). We excluded those with no follow-up survey (n = 399) and those with significant depressive symptoms at baseline to ensure a causal-order relationship between a new diagnosis of medical condition and episode of depressive symptoms (n = 986). Our final sample consisted of 8387 respondents who were followed up from 1992 until their first episode of depressive symptoms, death, loss to follow-up, or the end of the study period (ie, 2000).
The Health and Retirement Study measures depressive symptoms with a subset of 8 items of the standard Center for Epidemiologic Studies Depression Scale (CES-D). Respondents were asked about their depressive affect (“I felt depressed,” “I felt lonely,” and “I felt sad”), well-being (“I was happy” and “I enjoy life”), and somatic symptoms (“I felt everything I did was an effort,” “My sleep was restless,” and “I could not get going”). Previous studies have reported that this modification results in little loss of the structure and precision of the original scale.27,28 Participants indicated whether they did or did not experience (yes/no) each of the 8 symptoms much of the time in the past week. All items worded in the positive direction are reverse scored. A summary score is created by summing the number of yes answers across the 8 items and then dichotomizing the score to define subjects with and without significant depressive symptoms. Previous studies have used a cutoff score of 3 or 4, and a cutoff point of 3 or more has been found to have a sensitivity of 71% and specificity of 79% for major depression.29,30 We selected a cutoff score of 5 in our analysis to offer more stringent evidence of a high burden of depressive symptoms, but we also tested whether the results are sensitive to the cutoff of 3 used elsewhere.
In every survey wave, respondents were asked if, since the last interview, a physician had told them they had a specific medical condition or disease. We focused on all medical conditions measured in the survey as potential risk factors for an episode of depressive symptoms. These included cancer (excluding minor skin cancers), diabetes, hypertension, heart disease (heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems), arthritis or rheumatism (hereafter called arthritis), lung disease (chronic lung disease such as chronic bronchitis or emphysema, not including asthma), and stroke. A new diagnosis for each of these conditions was identified by determining the follow-up wave in which the medical condition was first reported, ie, it was not reported in previous waves. In terms of reliability of self-reports, previous studies have shown good agreement between administrative claims or medical records and self-reports for chronic conditions.31- 33
Using the medical conditions questions at each of the 5 waves of panel surveys, the diagnosis of each of the 7 medical conditions was defined in relation to the onset of every depressive symptom episode using 3 time-dependent dummy variables (newly diagnosed within the last >0-2, >2-4, and >4-8 years of significant depressive symptoms in the sample), resulting in a total of 21 key covariates.
Other study covariates included baseline socioeconomic and demographic variables such as age, sex, race, education, marital status, assets, and income. Health characteristics included baseline CES-D score, self-reported health status, and activities of daily living limitations.
We estimated the unadjusted rates of episodes of significant depressive symptoms within 0 to 2 years of a new diagnosis of each of the 7 medical conditions. For the multivariate analysis, we used time-dependent Cox regression models to estimate the relative hazard of significant depressive symptoms for patients with a new medical diagnosis (within the past >0-2, >2-4, or >4-8 years) to those without that medical diagnosis while controlling for confounders. Our Cox regression analysis modeled the time to first event of significant depressive symptoms. In other words, patients were considered no longer at risk after their first episode of significant depressive symptoms. Hence, the hazard ratios (HRs) for depressive symptoms within more than 2 to 4 and more than 4 to 8 years of a new medical diagnosis are conditional on not having experienced a depressed mood event within 0 to 2 and 0 to 4 years, respectively. All results are weighted to represent the national estimates, and standard errors were adjusted for the complex survey design using the method of Potthoff et al.34 A 2-tailed P value of .05 was considered statistically significant.
About 80% of the sample had complete follow-up until the end of the study period (8% died and 12% were lost to follow-up). About half the sample reported receiving a new diagnosis of at least 1 of the 7 medical conditions. The rate of new diagnosis of medical conditions (before depressed symptoms, if any) ranged from 4% (stroke) to almost 21% (arthritis) during the study period. Table 1 presents the baseline characteristics of the overall study cohort and for persons with new medical conditions diagnosed before onset of significant depressive symptoms. The mean (SD) baseline age of the study cohort was 55.8 (3.2) years; 51.8% were female, and 12.8% were nonwhite. Although the mean age was quite similar across all medical condition groups, adults with a new stroke, diabetes diagnosis, and chronic lung disease had a higher proportion of risk factors for depressive symptoms such as low income and assets and fair to poor health at baseline.
About 16% of all respondents had an episode of significant depressive symptoms during the entire follow-up period. The Figure presents the unadjusted rates of significant depressive symptoms within 0 to 2 years of a new diagnosis for each of the 7 medical conditions. Overall, 5.7% of this population sample aged 51 to 61 years in 1992 reported significant depressive symptoms within 0 to 2 years of the start of follow-up. Among the 7 medical conditions, significant depressive symptom rates were the highest in persons with new diagnoses of cancer (13%) or chronic lung disease (13%), followed by those with stroke (9%). Persons with hypertension (3.5%) or arthritis (4.4%) had among the lowest rates of significant depressive symptoms within 0 to 2 years of diagnosis of these conditions.
Table 2 presents the adjusted HRs with 95% confidence intervals (CIs) for onset of significant depressive symptoms relative to the timing of the onset of each of the 7 medical conditions; Table 3 presents the adjusted HRs with 95% CIs for the baseline socioeconomic, demographic, and health variables. Persons with a new diagnosis of cancer in the last 0 to 2 years had more than 3 times the hazard of significant depressive symptoms compared with persons without a diagnosis of cancer (HR, 3.55; 95% CI, 2.79-4.52). Similarly, persons with a recent diagnosis (in last >0-2 years) of chronic lung disease (HR, 2.21; 95% CI, 1.64-2.97) and heart disease (HR, 1.45; 95% CI, 1.09-1.93) had a higher hazard of having significant depressive symptoms than persons without a diagnosis of the respective conditions. Although the hazard for significant depressive symptoms was elevated, it was no longer significant for persons diagnosed with chronic lung disease for more than 2 years (ie, in the last >2-4 years or the last >4-8 years). On the other hand, persons with heart disease had a significantly higher hazard for reporting the first episode of significant depressive symptoms even more than 2 to 4 years (HR, 1.74; 95% CI, 1.22-2.50) and more than 4 to 8 years (HR, 1.54; 95% CI, 1.01-2.34) after diagnosis. Persons with a recent diagnosis of stroke, arthritis, diabetes, and hypertension showed no increases in risk for developing significant depressive symptoms in the regression analyses. However, the persons with arthritis in whom significant depressive symptoms did not develop within 0 to 2 years of diagnosis had a significantly higher hazard for depressive symptoms more than 2 to 4 years after the diagnosis (HR, 1.46; 95% CI, 1.11-1.92).
This longitudinal study finds that a new diagnosis of certain medical conditions is associated with the risk of having significant depressive symptoms develop among adults in late middle age. Persons with a diagnosis of cancer, arthritis, chronic lung disease, or heart disease were at risk for significant depressive symptoms, whereas those with a diagnosis of stroke, hypertension, or diabetes faced no apparent excess burden of depressive symptoms during the 8-year study period.
We find that the relative risk for development of significant depressive symptoms and the pattern of risk over time after a new diagnosis depend on the type of medical condition. The risk for early onset (within >0-2 years of diagnosis) of significant depressive symptoms was the highest among persons with a diagnosis of cancer, followed by those with chronic lung disease. On the other hand, patients with a diagnosis of arthritis developed significant depressive symptoms only more than 2 to 4 years after receiving the diagnosis, suggesting a relationship between disease progression and onset of depressive symptoms. Finally, persons with a diagnosis of heart disease had a persistent risk for developing significant depressive symptoms during an 8-year period after disease diagnosis.
Although no study has assessed changes in patterns during an 8-year time horizon, the broad finding that persons with a diagnosis of cancer, arthritis, chronic lung disease, or heart disease were at risk for significant depressive symptoms is similar to the findings observed in the literature.18- 21 The lack of significant relationships in our study of stroke and diabetes with depressive symptoms is the notable exception.9,17,35,36 This may be a result of several methodological aspects of this study. The first relates to the generalizability of our findings. Most previous investigations have been based on small cross-sectional clinical or community samples, and few have specifically examined this preretirement age group. Cross-sectional designs or designs that do not control for prior depression are subject to confounding. This may explain our lack of findings for stroke and diabetes because individuals with new cases of these diseases had high-risk baseline characteristics in our study. Finally, unlike the numerous disease-specific studies in the literature, we measured the risk for depressive symptoms across multiple chronic conditions, thus controlling for confounding due to coexisting illnesses and permitting a comparison of the risk for depressive symptoms between conditions.
Our study has some limitations that deserve mention. First, subjects may have had episodes of significant depressive symptoms between the biennial interviews that were also resolved in the period. The CES-D measures how the respondents felt about each of the 8 items only in the past week and hence captures persons who may have had a brief duration of depressive symptoms. Therefore, some patients may have experienced an immediate onset of depressive symptoms (within weeks or months) after receiving a diagnosis of a new medical condition that may not have persisted until the next 2-year interview. These unmeasured episodes of depressive symptoms may have resulted in a bias toward the null for capturing all depressive episodes, regardless of duration. However, the biennial interviews are more likely to identify enduring depressive episodes. Also, we used a more stringent cutoff score of 5 or more on the 8-item version of the CES-D than has been used previously in literature to define patients with significant depressive symptoms. However, post hoc sensitivity analysis with the traditionally used cutoff score of 3 or more did not significantly change our results. Our study is also limited in that our measure of medical conditions was self-reported and our assessment was restricted to only the 7 condition measures available in the survey.
Despite the limitations, our prospective study design, which followed up subjects free of depressive symptoms at baseline for a period of up to 8 years, combined with the use of time-varying covariates strengthens our findings on the association between new diagnosis of medical conditions and a subsequent episode of significant depressive symptoms. There may be multiple mechanisms through which chronic medical conditions can cause depressive symptoms or depression. First, depressive symptoms or depression may be a psychological reaction after the diagnosis of a new medical condition. This may be the primary explanation for our findings of the high risk for early onset of depressive symptoms after a new diagnosis of cancer, chronic lung disease, or heart disease. Second, the chronic medical illness may have a direct or indirect biological or pathophysiological effect on the brain. For example, disorders such as coronary artery disease that cause vascular damage can increase the vulnerability to depression.37 Third, depressive symptoms or depression may occur secondary to the adverse effects of medications prescribed for certain chronic conditions.38 Corticosteroids and heart medications like certain calcium channel blockers and digoxin have been associated with depression.38 Fourth, functional impairment and aversive symptoms such as chronic pain associated with certain medical conditions may also cause depressive symptoms or depression.15 Also, certain medical illnesses lead to an increased prevalence of depressive symptoms through their impact on job loss related to disability. Thus, our findings related to arthritis wherein the risk of depressive symptoms rises only several years after its diagnosis may be due to a progressive increase in functional disability and pain over time. Limitations in functional activities are also known to mediate the association of heart disease with major depression.39 Hence, our findings on the high and continued vulnerability for depressive symptoms during an 8-year period among patients with a diagnosis of heart disease may be jointly explained by each of the above mechanisms.
Regardless of the mechanism, our findings on the increased risk of depressive symptoms after diagnosis of a new chronic condition have important implications for patients, physicians, and payers. Patients with comorbid depressive symptoms or depression are reported to have lower rates of treatment adherence and higher rates of adverse health behaviors such as sedentary lifestyle and smoking.40- 42 Hence, physicians may find difficulty in managing the chronic condition among patients with comorbid depression, whereas the patient may face increases in severity or complications of the chronic disease that in turn may aggravate depressive symptoms.43 Besides increased morbidity, there is some evidence that patients with depressive symptoms are at a higher risk of mortality due to chronic illnesses.15 Furthermore, recent studies suggest that such patients have significantly higher use of medical services and costs than those without depressive conditions.44- 47 Hence, health care plans and third-party payers, including Medicare (which eventually pays for health care in this preretirement population), also have direct financial incentives to ensure appropriate medical management of such patients.
Our study findings on the unique patterns of depression risk with timing of diagnosis of specific chronic conditions should enable clinicians to identify high-risk groups for screening, prevention, and treatment programs for depressive symptoms and depression. The various patterns of risk for depressive symptoms over time since medical diagnosis suggest that physicians must screen patients with certain chronic illness for depressive symptoms on a regular basis rather than 1 time. Besides primary care physicians, specialists also need to pay attention to this issue because several patients may seek specialty care for their chronic illness. Also, health care plans should consider integrating depression screening and management with chronic disease care into their specific disease management programs. Overall, a consolidated effort on the part of all stakeholders will be needed to improve the physical and psychological outcomes in these vulnerable late middle age patients facing new medical illnesses.
Correspondence: Daniel Polsky, PhD, 1212 Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104 (firstname.lastname@example.org).
Accepted for Publication: December 3, 2004.
Financial Disclosure: None.
Funding/Support: This study was supported by The Robert Wood Johnson Foundation, Princeton, NJ.