Kaplan-Meier survival estimates of mortality (A) and readmission after discharge (B) by depression status at baseline.
Ng T, Niti M, Tan W, Cao Z, Ong K, Eng P. Depressive Symptoms and Chronic Obstructive Pulmonary DiseaseEffect on Mortality, Hospital Readmission, Symptom Burden, Functional Status, and Quality of Life. Arch Intern Med. 2007;167(1):60-67. doi:10.1001/archinte.167.1.60
Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
Depressive symptoms are common among patients with chronic obstructive pulmonary disease (COPD), but depression’s impact on COPD outcomes has not been fully investigated. We evaluated the impact of comorbid depression on mortality, hospital readmission, smoking behavior, respiratory symptom burden, and physical and social functioning in patients with COPD.
In this prospective cohort study, 376 consecutive patients with COPD hospitalized for acute exacerbation were followed up for 1 year. The independent associations of baseline comorbid depression (designated as a Hospital Anxiety and Depression Scale score of ≥8) with mortality, hospital readmission, length of stay, persistent smoking, and quality of life (determined by responses to the St George Respiratory Questionnaire) were evaluated after adjusting for potential confounders.
The prevalence of depression at admission was 44.4%. The median follow-up duration was 369 days, during which 57 patients (15.2%) died, and 202 (53.7%) were readmitted at least once. Multivariate analyses showed that depression was significantly associated with mortality (hazard ratio, 1.93; 95% confidence interval, 1.04-3.58), longer index stay (mean, 1.1 more days; P = .02) and total stay (mean, 3.0 more days; P = .047), persistent smoking at 6 months (odds ratio, 2.30; 95% confidence interval, 1.17-4.52), and 12% to 37% worse symptoms, activities, and impact subscale scores and total score on the St George Respiratory Questionnaire at the index hospitalization and 1 year later, even after controlling for chronicity and severity of COPD, comorbidities, and behavioral, psychosocial, and socioeconomic variables.
Comorbid depressive symptoms in patients with COPD are associated with poorer survival, longer hospitalization stay, persistent smoking, increased symptom burden, and poorer physical and social functioning. Interventions that reduce depressive symptoms may potentially affect COPD outcomes.
Chronic obstructive pulmonary disease (COPD) is a major cause of disability and death worldwide1 and the trends are rising.2 Because medical treatments have made limited inroads, patients with COPD experience recurring acute exacerbations, frequent hospital admissions, poor survival, and impaired physical functioning and quality of life (QOL).3- 5 Significant depressive symptoms are reported in 16% to 74% of patients with COPD6- 9 and are often unrecognized and untreated in primary and specialty care.10
In studies of patients with coronary artery disease, diabetes mellitus, and hypertension, depression has been documented to have major adverse effects on survival, symptom burden, the accomplishment of self-management tasks, functional impairment, and health care utilization.11- 14 In contrast, the effect of depression on COPD outcomes has not been systematically investigated; hence, the data are sparse and incomplete.
Although numerous studies have investigated risk factors for COPD mortality and hospitalization, to our knowledge, only a few studies have investigated depression specifically as a risk factor for mortality,15- 17 with contrary findings. Although anxiety was found to be related to increased risk of relapse18 and hospital readmission,19 contrary findings were reported for depression with readmission.18,20 However, relatively more studies have demonstrated that depressive symptoms are associated with impaired functional health or health-related QOL.21- 25
In this prospective cohort study of patients hospitalized for acute exacerbation of COPD and followed up for 1 year, we performed baseline assessments of the presence of depressive symptoms and other clinical and psychosocial variables. We then evaluated their associations with length of stay (LOS) and QOL for the index hospitalization and at 1 year after discharge, persistent smoking at 6 months, and mortality and readmission at 1 year.
The study was conducted in 2 large acute-care tertiary hospitals, which admitted more than 80% of the patients with COPD in Singapore. Research nurses identified patients daily from the hospital computerized admission database system, which automatically flagged patients admitted for acute exacerbation of COPD, and verified the diagnoses from the medical records and spirometric testing results. Inclusion criteria were (1) having a physician record of a principal diagnosis of COPD; (2) having a ratio of the forced expiratory volume in 1 second (FEV1) to the forced vital capacity (FVC) of less than 70% with or without chronic cough and sputum, with acute exacerbation defined as an increase in dyspnea, sputum production, or sputum purulence; and (3) being a current or an ex-smoker with a history of smoking equivalent to at least 20 pack-years. The exclusion criteria were (1) coexisting active pulmonary tuberculosis, pulmonary fibrosis, pneumothorax, or lung cancer; (2) death during hospital stay; (3) inability to perform spirometry or being too physically ill or mentally incapacitated to participate; and (4) residence outside Singapore. Patients who satisfied the eligibility criteria and signed informed consent for participation were recruited into the study. The ethics committees of the 2 participating hospitals approved the study protocol.
Baseline interviews and assessments were performed when the patients were in stable condition during their hospital stay or 2 to 4 weeks after discharge. Baseline measurements assessed during the hospitalization included socioeconomic data, COPD duration, comorbidities, degree of dyspnea, chronic mucus hypersecretion, gasometric indices, body mass index, smoking, hospitalizations for acute exacerbation of COPD in the preceding year, participation in pulmonary rehabilitation, and home oxygen therapy. Measurements assessed at 2 to 4 weeks after discharge included spirometry, Hospital Anxiety and Depression Scale (HAD), St George Respiratory Questionnaire (SGRQ), and level of caregiver support. During follow-up home interviews at 6 and 12 months after discharge, a research nurse ascertained information from the patient and caregiver on vital status, rehospitalization (including number of episodes and LOS), persistent smoking (at 6 months), HAD score, and SGRQ score. For each rehospitalization episode, the research nurse confirmed whether it was also for acute exacerbation of COPD.
The HAD26 was used to assess anxiety and depressive symptoms at baseline and at 6 months and 1 year after discharge. The HAD is a validated and reliable psychological measure widely used in medically ill populations, particularly in hospital settings. This 14-item self-report questionnaire has two 7-item subscales for anxiety (HAD-A) and depression (HAD-D). Scores ranged from 0 to 21 for each subscale, and a score of 8 or higher on either subscale is conventionally used to define anxiety and depression. At this cutoff, the HAD has shown high sensitivity (80%) and specificity (90%)27 in Chinese depressed patients. The original HAD was translated without any modification using standard (forward and backward) linguistic and semantic validation procedures before use. In our patient population, the HAD demonstrated good internal consistency (Cronbach α, 0.86 for HAD-A and 0.87 for HAD-D) and test-retest reliability (intraclass correlation coefficient r, 0.86 for HAD-A and 0.85 for HAD-D).
Symptom burden and physical and social functional status were measured using the SGRQ.23 The SGRQ is a self-administered, disease-specific, health-related QOL questionnaire consisting of 50 items with 76 weighted responses and 3 subscales (symptoms, activities, and impact). The questionnaire assesses the patient's experience of symptoms, the amount of distress caused by symptoms, and the extent to which symptoms like breathlessness have limited activities of daily living in the past 12 months. The effect of respiratory symptoms on QOL was measured for 3 domains: the symptoms score relates to the frequency and severity of respiratory symptom, the activities score relates to activities that cause or are limited by breathlessness, and the impact score relates to social functioning and psychological disturbances resulting from respiratory disease. Total and subscale scores were calculated from the sums of individually weighted item scores and converted to percentage scores (ranging from 0 to 100), with higher scores denoting worse health-related QOL. A difference or change of more than 4% (or 4 U) is considered clinically significant. The SGRQ has been previously validated for use in Chinese patients.28 Local validation studies of the translated version of the SGRQ in our patient population showed good internal consistency (Cronbach α, 0.78, 0.85, and 0.76 for the symptoms, activities, and impact subscales, respectively) and test-retest reliability (intraclass correlation coefficient r, 0.86, 0.92, and 0.87 for the symptoms, activities, and impact subscales, respectively).
Spirometry was performed using an electronic microspirometer (Micro Medical Ltd, Rochester, England) when the patient's condition was stable before the planned discharge and was repeated 2 to 4 weeks after discharge; we used the better of the 2 sets of measurements. Measurements followed American Thoracic Society criteria for spirometric standardization and procedures29 and were based on the best of 4 readings, with less than 5% variation in the 2 best readings. The FEV1 was expressed as the percentage of predicted values from local norms.30
Dyspnea was assessed by a graded scale (New York Heart Association classification)31 from 0 to 5, in which 0 indicates not at all breathless; 1, breathlessness on heavy exercise, such as climbing 2 or 3 floors; 2, breathlessness on moderate exertion, such as climbing 1 floor or walking quickly; 3, breathlessness on mild exertion, such as walking at normal speed; 4, breathlessness on minimal exertion, such as slow walking; and 5, breathlessness on limited exertion, such as showering, bathing, or washing. Chronic mucus hypersecretion was determined by cough and production of sputum lasting at least 3 months during the past year and was based on responses to the British Medical Research Council respiratory questionnaire.32
Medical comorbidities, coded according to the International Statistical Classification of Diseases, 10th Revision (ICD-10),33 were extracted from the medical records. Arterial blood gas analyses were performed while the patients were breathing room air (42.0% of patients) or receiving continuous oxygen therapy (58.0% of patients), and saturated arterial oxygen pressure (SpO2) was measured by a microoximeter under the same conditions.
Caregiver and family support was assessed by a questionnaire that was administered to an informant who had known the patient and his illness, usually a family member and primary caregiver. The scale elicited responses on a 4-point Likert scale (in which 1 indicates “not at all”/“none of the time” and 4, “all of it”/“all the time”) for 6 items: (1) level of understanding of the patient's illness and treatment; (2) level of understanding of his or her tasks in relation to the patient's illness; (3) frequency of supervising the patient during administration of medicine; (4) frequency of supervising the patient's exercise at home; (5) frequency of taking the patient for regular check-ups; and (6) encouragement of the patient toward recovery. The scale demonstrated good internal consistency (Cronbach α, 0.92) and test-retest reliability (intraclass correlation coefficient r, 0.79).
Multivariate regression methods for time-to-event (mortality and rehospitalization), dichotomous (smoking), and continuous (LOS and SGRQ) dependent variables were used, with depression (HAD-D score, ≥8) as the primary independent variable of interest. The time to first COPD-related readmission or death following the index hospitalization was plotted in survival curves using Kaplan-Meier estimates (Figure) and analyzed as dependent variables in Cox proportional hazards regression models. Data were censored for subjects who died, were hospitalized, or dropped out before the event of interest or since the last date of observation. Crude and adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated, with significance at P<.05 based on the log-rank test. Because both the fully saturated model and the stepwise selection model of variables individually identified as significant variables from univariate analyses yielded consistent and robust results, the parsimonious selection model results are presented in this report. Collinearity was evaluated using the correlation coefficient or κ greater than 0.6 and/or a variance inflation factor of 5.0 or greater; no collinearity was found particularly for depression and SGRQ scores. All statistical analyses were performed using SPSS version 14.0 software (SPSS Inc, Chicago, Ill).
A total of 503 patients who were hospitalized for acute exacerbation of COPD between April 1, 2002, and October 31, 2004, satisfied the eligibility criteria for inclusion; we excluded 18 patients who were terminally ill and/or had died (n = 7) while in the hospital. Among the remaining patients, 99 refused participation in the study or were not able to provide data because they were mentally incapacitated and unable to participate in the interview and/or they were unable to perform spirometry satisfactorily owing to poor effort tolerance. We excluded 10 patients who did not complete the HAD questionnaire. The final sample comprised 376 patients, with a participation rate of 74.8% (376/503).
The 376 patients were followed up for a mean (SD) of 313 (31) days (median, 369 days). During this period, 57 patients died after discharge. Among the surviving patients, 44 patients (13.8%) were lost or refused further participation, and 275 surviving subjects provided information from 1-year follow-up interviews. A total of 982 hospitalization episodes and 4707 hospitalization-days were recorded. After the index hospitalization, 202 patients were readmitted at least once.
Among the 376 patients, 167 patients (44.4%) were found to be depressed (Table 1). The proportion of patients with anxiety was low, but more depressed (17.3%) than nondepressed (5.8%) patients had anxiety. Compared with nondepressed patients, more depressed patients to a statistically significant degree (P<.05) were older; lived in lower-end public (1-3 rooms) housing; were divorced, widowed, or single; lived alone; had a lower body mass index; had had COPD for 5 years or longer; had chronic mucus hypersecretion; received home oxygen therapy; and had a low caregiver support score.
There were significantly more deaths after discharge among depressed patients (35/167 [21.0%]) than among nondepressed patients (22/209 [10.5%]) (Table 2 and Figure). The HR of mortality risk for depressed patients was 1.93 (95% CI, 1.04-3.58), after controlling for COPD duration and severity indices that included FEV1 and arterial blood gas test results, smoking, index hospitalization LOS, and socioeconomic variables (Table 3).
The number of patients who were readmitted and the total number of readmissions did not differ significantly between depressed and nondepressed patients (Table 2). The likelihood of earlier first readmission was not significantly different (adjusted HR, 0.93; 95% CI, 0.68-1.28) (Tables 2 and 3 and Figure).
Depressed patients had a significantly (P = .003) greater LOS during their index hospitalization (5.4 days) than did nondepressed patients (3.9 days) (Table 2). The difference in LOS remained significant (P = .02) after controlling for potential confounding risk factors (Table 4).
The total number of days spent in the hospital in 1 year by each patient was significantly (P = .04) greater in depressed (14.2 days) than in nondepressed (11.0 days) patients (Table 2). The difference remained statistically significant (P = .047) after controlling for potential confounding by length of follow-up, COPD severity markers, and other risk factors (Table 4).
Depression was not associated with FEV1 percentage at baseline or at 1 year. The percentage change in FEV1 at 1 year from baseline was also not significantly greater in depressed patients (Table 2).
Depressed patients did not differ significantly from nondepressed patients on current smoking during their index hospitalization (28.1% vs 29.2%). However, at 6 months, nondepressed patients had a significantly lowered proportion of current smokers (16.3%), whereas depressed patients had almost the same proportion of current smokers (28.8%, P = .009), (Table 5).
The SGRQ symptoms score was significantly worse in depressed than in nondepressed patients, both at the index hospitalization (8.3 points difference; P = .003) and at 1 year (10.1 points difference; P = .006), (Table 6). Similar differences varying from 8.0 to 11.5 points were found for the activities and impact scores. These amounted overall to 12% to 37% differences in scores across the SGRQ domains.
We also analyzed the association of anxiety (HAD-A score, ≥8) with COPD outcomes. Given the small numbers and its comorbid association with depression, we found that anxiety was associated with mortality in univariate analysis, although with only marginal significance (HR, 1.97; 95% CI, 0.96-4.02), and was nonsignificant in multivariate analysis.
We found that depression was significantly associated with mortality, longer index hospitalization and total LOS, persistent smoking at 6 months, and worse physical and social functioning as assessed by the SGRQ at the index hospitalization and 1 year later, after controlling for chronicity and severity of COPD, comorbidities, and behavioral, psychosocial, and socioeconomic variables.
An important finding in this study was that an excess of depressive symptoms in hospitalized patients with COPD was associated with increased mortality risk, after controlling for disease severity and other risk factors. To our knowledge, only 1 earlier study15 has reported that depression independently predicted increased mortality after COPD admission. A second study17 found that depression was associated in univariate analysis with mortality, but was not significant in multivariate analysis. Another study16 gave contrary results by showing a lowered mortality risk associated with depression, but it studied a small sample of 49 mostly female (67%) outpatients and controlled for few factors that included only FEV1.
We observed no significant association of depression with early hospital readmission. A previous study20 has also reported no specific relationship between depression and hospitalization risk. On the other hand, we observed that depression was associated with a longer LOS during the index hospitalization, as well as with the total number of days hospitalized in 1 year. To our knowledge, this has not been reported in any previous study.
Studies have consistently found that QOL measures predict subsequent risk of hospital readmission and mortality independent of severity and risk markers.34- 38 However, because these studies have not included measures of depression, they could not establish a specific association of depressive symptoms with COPD outcomes. Indeed, we have also confirmed in this study that, in addition to depression, the SGRQ symptoms score was also independently associated with readmission risk in multivariate analysis. The SGRQ activities and impact scores in this study were individually associated with mortality risk in univariate analysis, but not in the presence of depression in multivariate analysis.
Our observation that depression was independently associated with mortality but not hospital readmission risk deserves comment. We believe that mortality and hospital readmission are competing outcomes. Because depressed patients commonly feel hopeless and helpless about changing their life circumstances, they lack the drive and motivation to seek help and hospital readmission, succumbing to early death instead. Approximately 20% of the initially nondepressed patients developed depressive symptoms during follow-up, whereas fewer of the surviving depressed patients (after attrition by death) remained depressed; together these were likely to bias the result toward the null difference.
We found that, in response to routine smoking cessation counseling in the hospitals, a significant proportion of the nondepressed patients had quit smoking at the 6-months follow-up, whereas most of the depressed patients persisted in smoking. This adverse effect of depression on smoking cessation behavior lends support to the role of psychotherapeutic interventions, including bupropion hydrochloride and nortriptyline hydrochloride therapy, for smoking cessation in patients with COPD.39
Our inability to demonstrate an association of depression with FEV1 during the index hospitalization and at 1 year is probably not surprising considering the published evidence. Although accepted as a reliable physiological marker of disease severity, FEV1 has been shown in many studies to be weakly correlated with QOL measures, suggesting that different dimensions of disease severity and outcomes are measured by physiological and patient-reported outcome measures.23,35 Indeed, a randomized, controlled trial of nortriptyline39 in patients with COPD showed that physiological measures of pulmonary insufficiency were generally unaffected by treatment.
Our results confirm the strong associations of depression with increased symptom burden and impaired physical and social functioning that have been reported in previous studies.21- 24 These findings support the role of antidepressant or psychological interventions in depressed patients with COPD to improve outcomes. Several studies have reported that pharmacological treatment in clinically depressed patients with COPD significantly reduces depression,40- 42 anxiety, respiratory symptoms, and day-to-day functioning.40 Psychosocial interventions in the context of a pulmonary rehabilitation program have also been shown to reduce depression and anxiety43- 45 and to improve QOL,45 but did not modify exercise performance.45 Thus far, no clinical trials have demonstrated whether psychotherapeutic interventions in depressed patients with COPD significantly lower the risk of mortality and hospitalization.
Our study has a number of limitations. The HAD is widely used to assess depressive symptoms, but the cases identified are not equivalent to diagnoses of major depressive disorder. Nevertheless, this may also imply that depressive symptoms, reflecting subsyndromal depression, could have an adverse effect similar to syndromal depression. Further research should include randomized controlled trials of nonpharmacological interventions for subsyndromal depression in improving COPD outcomes. Another limitation is the self-report of hospitalization episodes and LOS, which may be less reliable than desired; this may explain our weakly significant results. In addition, we were unable to obtain data on the specific causes of death among the patients. Finally, our exclusion criteria may limit broad generalization of the findings.
In conclusion, in patients with acute exacerbation of COPD, depression is associated with increased mortality risk, longer hospitalization LOS, persistent smoking behavior, increased symptom burden, poorer physical and social functioning, and reduced QOL. Trials of antidepressant and psychological interventions should provide conclusive evidence of improved survival, QOL, and self-management behavior and reduced health care utilization.
Correspondence: Tze-Pin Ng, MD, Department of Psychological Medicine, National University Hospital, 5 Lower Kent Ridge Rd, Singapore 119074 (email@example.com).
Accepted for Publication: September 22, 2006.
Author Contributions: Dr Ng had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Ng, Tan, and Cao. Acquisition of data: Ng, Niti, Tan, Ong, and Eng. Analysis and interpretation of data: Ng, Niti, Tan, Ong, and Eng. Drafting of the manuscript: Ng, Niti, Cao, Ong, and Eng. Critical revision of the manuscript for important intellectual content: Ng, Niti, and Tan. Statistical analysis: Ng and Niti. Obtained funding: Ng and Tan. Administrative, technical, and material support: Ng, Niti, Tan, and Cao. Study supervision: Ng and Tan.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by research grant 03/1/21/17/214 from the Biomedical Research Council, Agency for Science, Technology and Research (ASTAR), Singapore (Dr Ng), and by a donation for research from Aventis-Pasteur.
Role of the Sponsor: Aventis-Pasteur was not involved in the design, conduct, analysis, or reporting of the study.