Predictors of Depressive Symptoms in Patients with COPD and Health Impact

Abstract

While depression is a common co-morbid condition among patients with COPD, little is known about predictors or health impact of depression among these patients. To address these gaps in knowledge we conducted a cross-sectional survey of 207 patients with COPD cared for in a network of primary care clinics affiliated with an urban academic health center. A standardized questionnaire was used to measure demographic characteristics, smoking status, co-morbid medical conditions, current medications, self-efficacy, social support, illness intrusiveness, and self-reported health care utilization during the previous 6 months. Depressive symptoms were assessed using the Centers for Epidemiologic Studies-Depression scale. Overall, the prevalence of moderate to high levels of depressive symptoms was 60.4%. In a multivariate analysis independent predictors of depressive symptoms were being a former smoker (OR = 0.41 (95% CI 0.19–0.89)), higher self-efficacy (OR = 0.42 (0.28–0.64)), higher social support (OR = 0.72 (0.52–0.99)), and higher perceived illness intrusiveness (OR = 1.05 (1.02–1.08)). Depressive symptoms were associated with increased physician visits, emergency room visits, and hospitalizations for lung disease. In conclusion, depressive symptoms are common among patients with COPD and associated with an increase in healthcare utilization. These findings suggest that the identification of risk factors for depressive symptoms (e.g., continued smoking) may increase detection and improve management of depression and health outcomes among patients with COPD.

Key words: :

• Chronic obstructive pulmonary disease
• Depression
• Prevalence
• Risk factors
• Healthcare utilization

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is an increasing cause of morbidity and mortality worldwide ([1], [2]). Although pulmonary impairment is the target of therapeutic trials and clinical management ([1], [2], [3]), there is growing recognition that health outcomes among patients with COPD are determined by a number of factors apart from abnormal lung mechanics ([3], [4], [5]). However, many gaps remain in our knowledge about the risks for and health impact of these factors.

While depression is a common co-morbid condition among patients with COPD ([5], [6]), few investigations have been conducted to examine predictors ([4], [7], [8]), health impact ([9], [10], [11]), or treatment of depression ([6]) in these patients. The objectives of this investigation were to examine predictors of depressive symptoms and health impact of these symptoms among patients with COPD.

MATERIALS AND METHODS

Study design, subjects, and setting

This cross-sectional study was conducted as part of a randomized controlled trial of nurse-assisted home care for patients with COPD and the research methods have been described previously ([12]). Briefly, patients 45 years of age and older were enrolled from 17 primary care clinic sites, part of an urban academic health center, during the period September 2000–August 2001. The University of Florida Health Science Center/Jacksonville Institutional Review Board approved the project and all subjects provided informed consent.

Case definition

To be eligible for the study, patients had to fulfill three criteria based on smoking status, respiratory symptoms, and spirometric testing. Eligible persons had to report at least a 20 pack-year smoking history and have at least one respiratory symptom (cough, shortness of breath, or wheeze) during the past 12 months. Persons who fulfilled the smoking and respiratory symptom criteria also had to demonstrate airflow obstruction (FEV₁/FVC ratio < 70% and FEV₁ < 80%) by spirometric testing. Spirometry was performed in the subject's home by the study coordinator (BB), following American Thoracic Society (ATS) standards ([13]) using a MicroLoop spirometer (Micro Medical Limited, ML3535, UK), which meets ATS performance criteria ([13]).

Data collection

After obtaining informed consent from persons who fulfilled the case definition for COPD, data collection was conducted by the study coordinator. Measurements included a number of characteristics that have been associated with depression in other populations ([14]) including demographics, smoking, co-morbid medical conditions, current medications, self-efficacy ([15]), social support ([16]), and illness intrusiveness ([17]). The outcome measures were depressive symptoms measured with the Centers for Epidemiologic Studies-Depression (CES-D) scale ([18]) and self-reported health care utilization during the previous 6 months.

Potential psychosocial predictors of depressive symptoms included self-efficacy, social support, and illness intrusiveness. Self-efficacy was assessed using a 34-item scale ([15]) that was specifically developed for patients with COPD and that measures confidence in managing or avoiding breathing difficulty in the following domains: negative affect (e.g., feeling down or depressed); intense emotional arousal (e.g., becoming angry); physical exertion (e.g., going up stairs too fast); weather/environment (e.g., humidity); and behavioral risks (e.g., overeating). The 20-item Medical Outcome Study social support questionnaire ([16]) assesses patient reports of the availability of someone to provide support in the following four domains: tangible (e.g., takes patient to doctor); affectionate (e.g., shows love and affection); positive social interaction (e.g., has a good time with); and emotional/informational (e.g., listens). The illness intrusiveness questionnaire is a 13-item instrument ([17]) that measures patients' perceptions about how much their illness and/or treatments interferes with 13 domains of their life (i.e., health, diet, work, active recreation, passive recreation, financial situation, relationship with significant other, sex life, family relations, other social relations, self-expression/self-improvement, religious expression, and community and civic involvement).

The outcome measures included depressive symptoms and self-reported healthcare utilization. The CES-D scale is a 20-item instrument ([18]) used to assess the frequency of depressive symptoms. Self-reported healthcare utilization for the six months prior to the interview was determined separately for physician office visits, emergency room visits, and hospitalizations for lung disease and for other conditions.

Statistical analysis

The main objectives of the analysis were to determine predictors and impact of depressive symptoms on healthcare utilization. High levels of depressive symptoms were defined as a score of 16 or greater on the CES-D scale, which was the cut-off value used in a previous study of patients with COPD from the Netherlands ([7]). In addition to these factors we also examined the association of high levels of depressive symptoms with smoking status, severity of lung function impairment, social support, self-efficacy, and illness intrusiveness using stratified and multivariate logistic regression.

Self-reports of physician-diagnosed co-morbid medical conditions were used to further categorize patients into mutually exclusive groups of associated potentially life threatening or other chronic conditions. This classification was used in an investigation of risk factors for depression among older adults ([14]). Chronic diseases considered potentially life-threatening were coronary artery disease, congestive heart failure, and stroke. Other chronic conditions included hypertension, diabetes, arthritis, ulcer disease, liver disease, and renal disease.

For this analysis, spirometric impairment was categorized using the GOLD criteria ([19]), which uses percent predicted forced expiratory volume in 1 second (FEV₁) with stage I ≥ 80%, stage II ≥ 50% and < 80%, stage III ≥ 30% and < 50%, and stage IV (< 30%). Based on the entry criteria of a FEV₁ < 80%, all participants were either stage II, III, or IV.

Self-reported healthcare utilization for lung disease and for other conditions was determined for physician visits, emergency room (ER) visits, and hospitalizations. The impact of depressive symptoms on healthcare utilization as continuous variables was examined using Poisson regression and categorical variables using multiple logistic regression adjusting for gender, severity of lung function impairment, and smoking status. In the logistic regression models healthcare utilization was categorized as none or one or more visits. All analyses were conducted using Statistical Analysis System version 9.1.2 (Cary, NC).

RESULTS

Of the 207 patients, the majority was white, female, and elderly with a mean age of 69.2 years (±8.1) years (Table 1). Slightly fewer than 50% of the patients were married and overall the socioeconomic status of the participants was low. Because cigarette smoking was a component of the diagnostic criteria for COPD, all subjects were current or former smokers. While nearly two-thirds were former smokers, 35.7% continued to smoke despite having COPD.

Table 1 Distribution of categorical characteristics of 207 patients with COPD

Characteristic	Total %	CES-D ≥ 16%	Odds Ratio (95% CI)
Age (years)
45–64	22.7	66.0	Reference
65–74	53.1	62.7	0.87 (0.40, 1.88)
≥75	24.2	48.0	0.48 (0.19, 1.17)
Gender
Male	44.9	57.0	Reference
Female	55.1	63.2	1.29 (0.71, 2.35)
Race
White	87.9	61.3	Reference
Black	12.1	52.0	0.68 (0.27, 1.71)
Marital status
Widowed	28.0	46.6	Reference
Married	46.9	62.9	1.95 (1.00, 3.77)
Separated/divorced	16.9	71.4	2.87 (1.17, 7.04)
Single	8.2	70.6	2.76 (0.86, 8.82)
Education
≥High school	58.9	53.3	Reference
< High school	41.1	70.6	2.10 (1.17–3.78)
Income
< 10,000	41.1	62.4	Reference
10,001–20,000	29.5	60.7	0.93 (0.45, 1.93)
20,001–30,000	17.8	54.1	0.71 (0.30, 1.67)
> 30,000	11.6	62.5	1.01 (0.71, 1.42)
Employment
Unemployed	89.3	60.3	Reference
Employed	10.7	63.6	1.15 (0.46, 2.88)
Smoking status
Former	64.3	55.6	Reference
Current	35.7	68.9	1.77 (0.97–3.22)
GOLD stage
II	37.9	56.4	Reference
III	40.3	61.5	1.23 (0.63, 2.43)
IV	21.8	64.4	1.40 (0.62, 3.20)
Chronic Conditions
None	5.3	36.4	Reference
≥1 LT	58.9	53.3	2.00 (0.56, 7.17)
≥1 non-LT	35.8	75.7	5.44 (1.43, 20.76)
Depression/Anxiety Medications
None	65.7	47.1	Reference
≥1	34.2	84.5	6.14 (2.83, 13.58)

*LT = life threatening.

Spirometric impairment and self-reports of other chronic conditions were used to characterize the health status of the patients (Table 1). About 78% of the patients had mild-to-moderate impairment (stages II and III) of FEV₁ and 22% severe impairment (stage IV). Nearly 95% of patients had at least one other chronic illness, with a mean of 3.6 (±2.1) other conditions. Of the persons with chronic conditions (n = 196), 62.2% had at least one chronic illness classified as life threatening and 37.7% had at least one non-life threatening chronic illness with no life threatening illness.

Overall, 60.4% of patients reported a high level of depressive symptoms (CES-D ≥ 16), which varied with a number of factors including demographic, behavioral, social, and health-related (Table 1 and Table 2). The strongest unadjusted associations were with marital status, educational level, smoking status, and other chronic conditions. As expected, the prevalence of depressive symptoms was very high among persons taking at least one anti-depressant and/or anti-anxiety medication (84.5%). Although the occurrence of depressive symptoms increased as the FEV₁ level declined, the change was not statistically significant (Table 1).

Table 2 Mean values for characteristics of COPD patients (n = 207) with (CES-D ≥ 16) and without depressive symptoms

Characteristic	CES-D < 16	CES-D ≥ 16
	Mean (±SD)	Mean (±SD)
BMI	27.39 (6.03)	27.98 (6.23)
Self-efficacy
Negative affect	3.75 (1.01)	2.84 (0.87)^+
Emotional arousal	3.74 (0.99)	2.89 (0.92)^+
Physical exertion	2.75 (1.21)	1.99 (0.92)^+
Weather	3.25 (1.04)	2.50 (0.78)^+
Risk Factors	3.56 (1.18)	2.81 (0.98)^+
Social support
Tangible support	4.03 (1.09)	3.57 (1.12)^+
Affectionate	4.08 (1.18)	3.67 (1.27)^#
Social interaction	3.88 (1.11)	3.21 (1.23)^+
Emotional and informational	4.01 (1.03)	3.31 (1.17)^+
Illness intrusiveness	30.50 (14.41)	45.3 (17.16)^+

*BMI = body mass index (weight (kg)/height² (m)).

⁺ p < 0.01, ^# p < 0.05.

Other factors significantly associated the frequency of depressive symptoms were self-efficacy, social support, and illness intrusiveness (Table 2). On average, all components of self-efficacy and social support were inversely related to the level of depressive symptoms. In contrast, there was a direct relationship between illness intrusiveness and depressive symptoms.

Step-wise multivariate logistic regression was used to determine which factors were independently associated with high levels of depressive symptoms. Among the factors associated with depressive symptoms in the unadjusted analysis (Table 1 and Table 2), only four were statistically significant independent predictors in the multivariate analysis; smoking status, self-efficacy, social support, and perceived illness intrusiveness. Former smokers had a lower risk of depressive symptoms compared to current smokers (odds ratio OR = 0.41 (95% confidence interval (CI) 0.19–0.89)). Patients with higher confidence in managing or avoiding breathing difficulty associated with negative emotions were less likely to report depressive symptoms (OR = 0.42 (0.28–0.64)). Similarly, patients reporting higher frequency of availability of persons to provide emotional and informational support had lower levels of depressive symptoms (OR = 0.72 (0.52–0.99)). Finally, as illness and/or treatments were perceived to interfere with health, functioning, and relationships (i.e., illness intrusiveness) the risk of depressive symptoms increased (OR = 1.05 (1.02–1.08)).

The impact of high levels of depressive symptoms was assessed using self-reports of healthcare utilization for lung-and non-lung-related conditions during the previous six months. On average, high levels of depressive symptoms were consistently associated with increased lung disease-related physician visits, ER visits, and hospitalizations, but were only statistically significant for hospitalizations (p ≤ 0.05) (Table 3). Using multivariate Poisson regression, healthcare utilization was found to increase with depressive symptoms and was lower with less severe lung function impairment (Table 4). The associations between high levels of depressive symptoms and healthcare utilization were statistically significant (p ≤ 0.05) for doctor visits and hospitalizations, but not ER visits. Similar results were found with multivariate logistic regression models, adjusting for smoking, severity of lung function impairment, and illness intrusiveness (data not shown). Depressive symptoms were not consistently associated with non-lung-disease-related utilization (Table 3).

Table 3 Mean values for self-reported healthcare utilization during the previous 6 months among COPD patients with (CES-D ≥ 16) and without depressive symptoms

Characteristic	CES-D < 16	CES-D ≥ 16
	Mean (±SD)	Mean (±SD)
Healthcare Utilization for Lung Disease
Doctor visits	1.13 (1.67)	1.89 (3.65)
ER visits	0.17 (0.62)	0.28 (0.90)
Hospitalizations	0.06 (0.24)	0.27 (1.02)
Healthcare Utilization for Other Conditions
Doctor visits	3.24 (3.94)	3.38 (4.52)
ER visits	0.18 (0.50)	0.28 (0.59)
Hospitalizations	0.20 (0.64)	0.20 (0.47)

*p ≤ 0.05 (two sample t-tests were performed for each of the six comparisons).

Table 4 Incidence ratios (IR) for lung-related healthcare utilization during the previous 6 months among patients with COPD

Predictor	Doctor Visits	ER Visits	Hospitalizations
Characteristic	IR (95% CI)	IR (95% CI)	IR (95% CI)
CES-D ≥ 16	1.70 (1.02, 2.81)	1.76 (0.93, 3.34)	5.74 (2.02, 16.28)
Males	1.38 (0.87, 2.18)	1.35 (0.77, 2.40)	1.57 (0.82, 3.00)
Current Smoker	0.73 (0.43, 1.24)	0.98 (0.52, 1.84)	0.73 (0.35, 1.52)
Gold Stage
II	0.53 (0.29, 0.98)	0.22 (0.09, 0.54)	0.24 (0.10, 0.57)
III	0.79 (0.46, 1.35)	0.68 (0.37, 1.25)	0.32 (0.15, 0.68)
IV	Reference	Reference	Reference

*from Poisson Regression.

DISCUSSION

Among patients with COPD from a network of urban primary care clinics, we found that the majority reported depressive symptoms. Factors independently associated with these symptoms were current smoking, low self-efficacy and social support, and higher perceived illness intrusiveness. However, severity of lung function impairment was not an independent predictor of depressive symptoms. Finally, depressive symptoms had adverse health impacts with an increased risk of ER visits and hospitalizations for lung disease independent of smoking status or severity of lung function impairment.

Depression is common among the elderly with chronic medical conditions and complex biological mechanisms including genetic factors, and psychosocial characteristics probably contribute to the development of depression ([20]). While the prevalence of depression is high among patients with COPD ([5], [6]), limited data are available on predictors of depression in these patients ([4], [7], [8]). van Manen and co-workers ([7]) conducted a cross-sectional survey of 162 COPD patients and 359 controls from general practices in the Netherlands using the CES-D ≥ 16 to define depressive symptoms. Overall, they found that 21.6% of COPD patients had depressive symptoms compared to 17.5% among control patients. In a multiple logistic regression model independent risk factors for depressive symptoms included living alone, severe impairment of physical functioning, and lack of FEV₁ reversibility. Although these results are consistent with our findings (i.e., low social support, illness intrusiveness) they did not examine risk associated with self-efficacy or smoking status.

Other evidence linking psychosocial factors and depression among patients with COPD was found by McCathie and co-workers ([4]). In a cross-sectional survey of 92 Australian men to determine the relationship between coping styles, self-efficacy, and social support, these investigators found that lower levels of self-efficacy and social support were independently associated with higher levels of depression.

Among patients with COPD scant evidence is available on the association between smoking and depression ([8]). In a cross-sectional survey of 147 primary care patients with COPD from the Netherlands, Chavannes and co-workers ([8]) used the Beck Depression Inventory and found in univariate analyses that female gender, BMI, and smoking were associated with increased risk for depressive symptoms. However, in a multivariate logistic regression only female gender, BMI, and dyspnea were independently associated with depressive symptoms, and smoking was not an independent risk factor. Methodological differences including different measures of depressive symptoms, sample size, and model variables may explain the discrepant findings between the multivariate analyses of our study and Chavannes and co-workers ([8]).

Although limited evidence is available on the specific association between smoking and depression among patients with COPD, other sources of evidence provide support for a causal link ([6]). Epidemiological studies have demonstrated that depressed youths have an increased risk of smoking ([21]), and there is consistent evidence for an association between smoking and depression in the general population ([14], [22], [23]) and clinic populations ([24]). Moreover, depression has been associated with lower success with smoking cessation and recurrence of depression has been reported with smoking cessation ([25], [26], [27]).

Complex biological, social, and psychological mechanisms, which are poorly understood, likely contribute to the association between smoking and depression ([6]). Limited evidence suggests that smoking and depression may be mediated through the stimulation of dopaminergic pathways in the brain from nicotine and other constituents of cigarette smoke ([28]). In addition, non-nicotine components of cigarette smoke have been shown to cause anti-depressant action by increasing monoamine oxidase inhibition in the brain ([27]). However, the direction of cause-and-effect between smoking and depression has not been established.

Two possible explanations for the link are that smokers may smoke to self-medicate their depression or the long-term neuropharmacological effects of smoking may lead to depression ([29]). Furthermore, chronic hypoxemia in COPD may cause depression through microvascular changes in the brain ([6]). Regardless of the mechanisms, taken together the available evidence suggests that depression may contribute to onset of smoking and/or difficulty quitting that ultimately causes COPD.

Studies of the health impact of depression among patients with COPD have largely examined quality of life, functional disability, and survival with few studies of healthcare utilization ([5]). Our finding of increased healthcare utilization associated with high levels of depressive symptoms is consistent with a limited number of studies involving a small number of patients with COPD ([9], [10], [11]) and other populations ([30], [31], [33]). The poorer health outcomes associated with depression among patients with COPD have included adverse effects on health-related quality of life and healthcare utilization ([11]), on outcome of emergency treatment ([9]), and on mortality ([10]). These findings are consistent with results from studies of health impact of depression in the general population and among outpatients that depression and depressive symptoms adversely affect functioning and well-being ([30]), decreases adherence to prescribed treatments ([31]), increases healthcare utilization and costs ([32]), and increases mortality ([33]). Impaired self-management among depressed patients with COPD may partly contribute to the adverse health impact of depression ([34]).

Several factors need to be considered in the interpretation of our findings. The CES-D is validated and widely used in epidemiological studies as a screening instrument for clinical depression but it is not diagnostic of clinical depression ([18], [35]). Because women, who comprised the majority of our sample, more frequently report depressive symptoms ([20]), and the symptoms of depression are non-specific overlapping with the symptoms of COPD, our prevalence estimates likely over-estimate the true prevalence of clinical depression. Regardless, sub-clinical depression is associated with high morbidity among patients with COPD ([36]). Because of the cross-sectional study design causal inferences about the association between the risk factors we identified and depression must be made cautiously. While the validity of self-reported healthcare utilization may be questioned, it has been shown to be an acceptable method of measurement, and may actually result in an underestimation of utilization with more frequent use ([37]). Overall, the prevalence of ER visits and hospitalizations was low, which reduced power to detect statistically significant differences in ER visits between the two groups of patients.

In summary, the increased healthcare utilization and other adverse health outcomes associated with depressive symptoms provide a strong clinical rationale for the detection and treatment of depression among patients with COPD. Findings of this investigation further emphasize that biological (e.g., smoking), psychological (e.g., self-efficacy and illness intrusiveness), and social (e.g., social support) factors are operational in depression ([20], [38]) and need to be addressed in the management of depression. Growing evidence suggests that pulmonary rehabilitation ([39], [40]), anti-depressant medications ([6], [41], [42]), and counseling ([43], [44], [45]) may improve health outcomes of patients with COPD and depression. However, depression is frequently under-diagnosed ([46]), and the implication of our findings is that the identification of predictors of depressive symptoms (e.g., continued smoking, low self-efficacy and social support) may increase detection and ultimately improve the management of depression and health outcomes among patients with COPD.

Funding: Robert Wood Johnson Foundation.

The authors thank the staff and physicians of the University of Florida Jacksonville Physicians for enabling us to enroll their patients.