Risk Factors for Hospital Readmission in Patients with Chronic Obstructive Pulmonary DiseaseAlmagro P.a · Barreiro B.b · Ochoa de Echagüen A.a · Quintana S.c · Rodríguez Carballeira M.a · Heredia J.L.b · Garau J.a
aDepartment of Internal Medicine, and bRespiratory and cIntensive Care Units, Hospital Mútua de Terrassa, University of Barcelona, Barcelona, Spain
Background: Hospital readmissions for acute exacerbation of chronic obstructive pulmonary disease (COPD) are one of the leading causes of healthcare expenditures worldwide. Objectives: To identify risk factors for hospital readmission in COPD patients. Methods:We prospectively evaluated 129 consecutive patients hospitalized for acute exacerbation of COPD. Clinical, spirometric and arterial blood gas variables were measured during hospitalization. Socioeconomic characteristics, comorbidity, dyspnea, functional dependence, depression, social support and quality of life were also analyzed. Readmission was defined as one or more hospitalizations in the following year. Results:During the follow-up period, 75 (58.5%) patients were readmitted. In bivariate analysis, readmission was associated with previous hospitalization for COPD in the past year, dyspnea scale, PaCO2 at discharge, depression, cor pulmonale, chronic domiciliary oxygen and quality of life measured by the St. George’s Respiratory Questionnaire. In multivariate analysis, the best predictor of readmission was the combination of hospitalization for COPD in the previous year (odds ratio, OR: 4.27; 95% confidence interval, CI: 1.5–12), the total score of the St. George’s Respiratory Questionnaire ≥50 points (OR: 2.36; 95% CI: 1.03–5.04) and PaCO2 at discharge ≥45 mm Hg (OR: 2.18; 95% CI: 0.84–5.06). With this model, the probability of readmission for patients without any of these variables was 7%, while it was 70% for the patients with all three variables present. Conclusion: The combination of quality of life, hospitalization for COPD in the previous year and hypercapnia at discharge are useful predictors of readmission at 1 year.
Copyright © 2006 S. Karger AG, Basel
Hospitalization for acute exacerbation usually occurs in the later stages of chronic obstructive pulmonary disease (COPD) and represents more than 70% of all COPD-related medical care cost [1, 2]. An important part of these hospital stays and healthcare expenditures is due to hospital readmissions. The frequency of readmission varies from11.6% (48 h after discharge from the emergency room) [3 ] to 63% (1 year after admission to a general hospital) [4,5,6,7]. In patients with hypercapnia on admission, half of them will need hospital readmission and 7% will be readmitted 3 or more times within the ensuing 6 months . In a study in hospitalized COPD patients older than 65 years carried out in the US between 1984 and 1991, only 14% did not require rehospitalization, while 48% were admitted on five or more occasions .
Few studies have evaluated risk factors for hospital readmission in COPD patients. These investigations, which were usually based on retrospective analysis, have shown that social and psychological variables, lowest FEV1, poor performance status or previous admission for COPD are associated with a greater risk of readmission [4, 6, 10]. Prospective studies have found that a worse quality of life measured by the St. George’s Respiratory Questionnaire (SGRQ) [7, 11], a lower than usual physical activity  or FEV1 stratified according to the European Respiratory Society and Global Obstructive Lung Disease  scales were related to a higher risk of readmission.
The purpose of the present study was to evaluate prospectively a large series of potential risk factors for hospital readmission in a well-defined cohort of COPD patients hospitalized for acute exacerbation. Therefore, psychological, functional and social variables as well as quality of life, for example, were considered.
Patients and Methods
Between October 1996 and May 1997, we prospectively studied all consecutive patientsadmitted for acute exacerbation of COPD to any medical ward at our institution, an acute-care teaching referral center, in the province of Barcelona, Spain.
Details of the study protocol have been described previously . Briefly, inclusion criteria were: (1) a clinical diagnosis of COPD and (2) forced spirometry at discharge showing FEV1 ≤70% of their reference value and FEV1/FVC <0.7 . Reasons for hospital admission due to exacerbation were defined as worsening of breathlessness or change in the mental status due to hypercapnia . Exclusion criteria included a history of asthma, pneumonia or pulmonary edema on admission, hospitalization for causes other than COPD exacerbation or patient refusal to participate in the study. The study protocol was approved by the Hospital Research and Ethics Committee.
Arterial blood gases were assessed at rest during room air breathing. Serum albumin, body mass index, presence of cor pulmonale (signs of right-sided heart failure secondary to COPD or diuretic use for prior documented diagnosis), use of chronic oxygen therapy, number of previous admissions for COPD in the past year and a history of smoking were recorded on admission. At discharge, lung function tests were performed (forced spirometry and bronchodilator tests), and arterial blood gas determinations were repeated. The day before discharge, the patients were interviewed about a series of potential risk factors for readmission using the following specific questionnaires. (a) A questionnaire on regular treatment was used: the number of drugs and corresponding daily doses taken by the patient at home prior to admission were recorded. (b) A health-related quality-of-life questionnaire, the validated Spanish version of the SGRQ, a specific questionnaire for COPD patients, was administered. It consists of 50 items with 76 weighted responses and three component scores: symptoms, activities and impact. A total score is calculated from all three components, with zero indicating no health impairment and 100 representing maximum impairment. (c) Dyspnea was evaluated on a visual analog scale and the dyspnea scale of the American Thoracic Society. (d) Comorbidity was measured using the Charlson Index, a standard scale with 15 chronic diseases graded for severity of disease. (e) Cognitive impairment was assessed through the Pfeiffer Questionnaire; scores run from 0, representing no deterioration, to 10, indicating severe deterioration. (f) Functional dependence 2 weeks prior to admission was interrogated with the Katz Activities of Daily Living Scale, which reflects dependence in six primary social-biological functions. Scores range from 0 (independence) to 6 (high dependency). (g) Information about family composition, personal relationships and the availability of care was collected in a structured interview using the Social Resources Scale of the Older Americans Research and Service Center. The scores range from 1 (excellent social resources) to 6 (social disability). Marital status was dichotomized into two groups, married or unmarried. Unmarried patients included those who were divorced, separated, widowed and never married. (h) At the socioeconomic level, information about the educational level and work performed throughout his or her lifetime was collected. (i) The presence of depression was detected by the Yesavage Scale in its shortened version, which has 15 yes/no questions.
Variables were selected based on previous studies on readmission for COPD or other chronic diseases. Specific questionnaires were selected giving priority to those validated in Spanish, being concise and of common use. The different questionnaires used have been reported in detail elsewhere .
After hospital discharge, patients were monitored by their physicians, receiving no intervention provided by the investigators. In all cases, the need for readmission was decided by physicians in the Emergency Room unaware of the goals of the present study.
Patient readmission and evolution were assessed 1 year after hospitalization. Phone calls were used to determine the vital status in the year after discharge. Readmissions were assessed through revision of clinical records, and linkage was made with the database CMBD (Minimum Basic Dataset). The CMBD is a national administrative database of the ‘Servei Català de la Salut’, where all discharges of public hospitals of Catalonia are recorded. Fields used for linkage were full name, sex and date of birth. Readmission was defined as one or more hospitalizations for COPD in the year following hospital discharge. If patients had been hospitalized in the Emergency Room for less than 24 h, they were not considered as readmissions. Data on subjects who died without readmission were considered as missing data.
Analyses were carried out using the SPSS package version 10.0. The results are shown as means ± SD for quantitative variables. For qualitative variables, either a =χ2 test or Fisher’s exact test was used. Quantitative variables were analyzed using Student’s t test or corresponding non-parametric tests (Mann-Whitney or Kruskal-Wallis test) when required by the distribution of the data. Two-tailed testing was used, and a p value <0.05 was considered as statistically significant.
For multivariate analysis, two logistic regression models were constructed with hospital readmission in the following year as a dependent variable. In the first model, a subanalysis was carried out to evaluate quality of life, the three SGRQ subscales (impact, activity and symptoms) and the total score adjusted consecutively for age, sex, FEV1, comorbidity and social support. In the final model, independent variables were eligible for entry if they were associated with readmission at a p value ≤0.15 in the bivariate analysis. In this analysis, only total SGRQ scores were considered. We categorized continuous variables at clinically relevant cutoffs. Potential confounding variables were tested and eliminated if their exclusion did not reduce the model likelihood unless when ≥15% of the coefficients of the remaining variables were modified. Interactions between the independent variables included in the model were tested. To assess the goodness of fit, the Hosmer-Lemeshow test was used .
Overall, 156 patients were screened during hospitalization: 12 did not meet the inclusion criteria and 3 were unavailable for follow-up information and were therefore excluded from the analysis. There were no differences in age, duration of hospitalization and in the number of admissions in the previous year between cases excluded from the study and the remainder of the study cohort. Another 12 patients died within a year without requiring a new hospitalization, who were considered as missing data. These patients were significantly older that the rest of the group: 78.1 ± 9 versus 71.7 ± 9 years (p < 0.02), although FEV1 and comorbidity were not different. Thus, 129 patients were evaluated. Mean hospitalization of the study patientswas 12.7 ± 7.3 days, ranging from 2 to 40 days. Sixty-four patients (49.6%) were admitted in respiratory failure (basal PaO2 ≤60 mm Hg) and 47 (36.4%) were hypercapnic (basal PaCO2 ≥45 mm Hg). Eight patients required mechanical ventilation. Sociodemographic and functional characteristics are shown in tables 1 and 2, respectively.
|Table 1. Sociodemographic characteristics of the study population (n = 129)|
|Table 2. Parameters of respiratory function in the 129 patients|
Twenty-one of the 129 patients (16.3%) were readmitted within 1 month, and 45 (34.9%), 53 (41,1%) and 75 (58.1%) at 3, 6 and 12 months, respectively (fig. 1). The mean number of hospital readmissions in the following year was 2.6 ± 2 (fig. 2), and the mean length of hospitalization for these readmitted patients was 30.2 ± 24.2 days.
|Fig. 1. Percentage of patients readmitted 30, 90, 180 and 365 days following discharge.|
|Fig. 2. Number of readmissions.|
In the bivariate analysis, readmission was significantly more frequent in patients with higher PaCO2 values at discharge (43.5 ± 9.4 vs. 39.1 ± 6.3 mm Hg; p < 0.005), poorer scores in the dyspnea scale of the American Thoracic Society (2.6 ± 1.3 vs. 2.1 ± 1.3; p < 0.02) and a higher number of admissions due to exacerbation of COPD in the past year (1.4 ± 1.6 vs. 0.6 ± 0.9; p < 0.001). Similarly, readmitted patients showed higher values on the Yesavage Depression Scale (5 ± 3.4 vs. 3.7 ± 3.1; p < 0.05) and they were taking more drugs at home (5.9 ± 2.2 vs. 4.5 ± 2.7; p < 0.002) and higher number of doses per day (13 ± 5.2 vs. 9.3 ± 6.4; p < 0.001); furthermore, they used more frequently oxygen at home (p < 0.05) and were more likely to have cor pulmonale (p < 0.05). In the SGRQ scale, readmitted patients scored significantly worse on all three SGRQ subscales (impact: 46.7 ± 22 vs. 35.9 ± 20.6; p < 0.007; activity: 69.7 ± 19 vs. 59 ± 22.4; p < 0.003, and symptoms: 56.3 ± 19.5 vs. 45.9 ± 24.3; p < 0.01) and also had poorer total scores (54.8 ± 17.7 vs. 44.3 ± 18.9; p < 0.002) than patients who were not readmitted (fig. 3).
|Fig. 3. Quality of life measured with SGRQ and readmission. R = Readmitted patients; NR = non-readmitted patients.|
No significant differences were found between readmitted and non-readmitted patients for social variables (2.3 ± 1.3 vs.1.9 ± 1.1), functional status (5.4 ± 1.1 vs. 5.6 ± 0.9), comorbidity (2.3 ± 0.9 vs. 2.1 ± 1.1), body mass index (27 ± 5.2 vs. 25.6 ± 4.1), socioeconomic status, cognitive impairment, serum albumin, emergency room visits, length of current hospital stay or FEV1.
Two logistic regression models were performed with hospital readmission as a dependent variable. In the first of them, carried out to evaluate quality of life, the three SGRQ subscales and the total score were consecutively included and adjusted for age, sex, FEV1, comorbidity and social support. The readmitted patients reached higher scores (indicating a decrease in the quality of life) than non-readmitted patients (table 3).
|Table 3. SGRQ and readmission|
In our final multivariate model, cor pulmonale, hospitalization for COPD in the past year, long-term oxygen therapy, hypercapnia at discharge (PaCO2 ≥45 mm Hg), number of daily drugs ≥5, worse quality of life (categorized at 50 points, the median value of the SGRQ total) and depression (≥5 points on the Yesavage Scale, considered the cutoff of depression) were entered as independent variables.
Readmission was significantly related with previous hospitalization (p < 0.005; odds ratio, OR: 4.27; 95% confidence interval, CI: 1.51–12.04), hypercapnia at discharge (p = 0.1; OR: 2.18; 95% CI: 0.84–5.63) and poorer quality of life (p < 0.05; OR: 2.18; 95% CI: 1.03–5.41). Hypercapnia was retained in the final model because its exclusion changed the estimated coefficient of the two remaining variables >15%, an increase that was considered clinically significant. The Hosmer-Lemeshow Goodness-of-Fit Test indicated that the model is well calibrated (p = 0.85). In this test, a high p value indicates that the model is performing well, which means that no large discrepancy was found between observed and expected rates of readmission. In our final model, the probability of readmission at 1 year would be given by the formula: Pr = 1/(1 + eβ) where β = –2.24 + 1.45 (in case of previous hospitalization) + 0.86 (if SGRQ ≥50) + 0.78 (if PaCO2 ≥45 mm Hg). Using this model, the probability of readmission for patients without any of these characteristics was 7%, being 70% for patients showing all three variables. The probabilities of readmission associated with the different combinations, e.g. previous hospitalization, hypercapnia at discharge and poorer quality of life, are shown in table 4.
|Table 4. Probability of readmission at 1 year|
In this study, we assessed prospectively the frequency of and risk factors for hospital readmission in COPD patients during their hospital stay. The main findings were that admission for COPD in the past year, hypercapnia at discharge and quality of life assessed by SGRQ are independent predictors of readmission, and all three factors combined correctly predict the risk of unplanned readmission in the year after hospitalization.
Few studies have reported variables associated with hospital readmission in COPD patients, and they have been usually performed retrospectively [4, 6]. To our knowledge, only in four prospective studies, risk factors for rehospitalization of COPD patients have been evaluated [5, 7, 11, 12]. These studies have found that quality of life [7, 11] and a greater physical activity  predict readmission within 12 months after an index hospitalization. Finally, in a recent prospective study of 67 patients hospitalized for COPD, Tsoumakidou et al.  showed that lung function tests performed during stable condition and stratified in accordance with the European Respiratory Society and Global Obstructive Lung Diseases Scales correlated with readmissions in the ensuing 18 months.
In our study, health-related quality of life was measured with the SGRQ, a questionnaire specifically designed for respiratory diseases. Readmitted patients scored significantly worse in all three SGRQ components: disturbance of physical activity (activity), overall impact on daily life and well-being (impact) and distress due to respiratory symptoms (symptoms). All three subscales and the total score were independent predictors of readmission even after adjustment for age, sex, FEV1, comorbidity and social support. The association between poor quality of life, measured using the SGRQ, and hospital readmission extends the results of previous studies that have addressed this issue [7, 11, 16]. These studies, however, did not evaluate other variables such as arterial blood gases, comorbidity, social support or previous admissions, which are all thought to affect readmission rates for this population. To our knowledge, our study is the first to demonstrate that a poorer quality of life in COPD patients, measured with a specific questionnaire, poses a substantial risk of readmission independently of physiological, functional and social variables.
Previous hospitalization for COPD exacerbation is generally considered to be a hallmark of advanced disease. In our study, admissions in the past year were one of the best predictors of readmission, in agreement with previous retrospective studies . A plausible explanation is that these patients had more severe disease, as suggested by the higher frequency of cor pulmonale, the use of chronic oxygen therapy and a greater number of medications taken, even in the presence of similar derangements of pulmonary function and comorbidity, measured with FEV1 and the Charlson Index, respectively.
In our study, PaCO2 at hospital discharge was a predictor of subsequent readmissions. This finding is in agreement with the experience of Kessler et al. , who showed that in patients with stable disease, only PaCO2 and mean pulmonary artery pressure were independently related to the risk of hospitalization. Although the presence of hypercapnia during exacerbation has been considered an indicator of a poor prognosis , some patients develop hypercapnia only during acute exacerbation and revert to normocapnia during recovery; prognosis in these patients with reversible hypercapnia was similar to normocapnic patients [18, 19]. Hypercapnia at discharge was retained in the final multivariate model because its exclusion changed the OR of SGRQ and hospitalization in the previous year >15%, an increase that was considered clinically significant. The combination of these three variables constituted the most harmonious and well-calibrated model.
The significance of FEV1 in predicting hospital admission for COPD is well recognized [20, 21]. Although intuitively it would be expected that patients with poor respiratory function assessed by the percentage of predicted FEV1 would be rehospitalized more frequently, few data are available on the usefulness of spirometric assessment during hospitalization. In a retrospective study performed in 1,400 patients, Roberts et al.  found a higher rate of readmissions at 3 months in patients with the lowest tertiles of FEV1. Conversely, a prospective study on 266 patients  indicated non-significant differences in readmission rates at 1 year in patients with the lowest FEV1. Unfortunately, both studies had a high rate of missing values (50 and 37%, respectively) so that a selection bias cannot be ruled out. Finally, in studies by Garcia- Aymerich et al. [5 ] and Tsoumakidou et al. , spirometry and arterial gases were measured during a clinically stable period, at least 2 months after hospitalization, and lower FEV1 values were associated with a higher risk of readmission. In our work, spirometry measured at discharge was part of the inclusion criteria, and we could not confirm a relationship between FEV1 and subsequent readmission. Based on our results, we believe that spirometry data measured during hospitalization are a poor predictor of readmission in these patients.
Finally, depression, a common characteristic of COPD patients with advanced disease [22, 23], has been shown to be associated with a higher rate of readmission at 6 months in elderly patients, independently of the functional status, comorbidity and living situation . We also found significant differences in bivariate analysis between readmission and greater depression scores. However, the depression scale was not found to be an independent predictor of readmission. This could be explained by the association between SGRQ and depression [25, 26], and confirms that depressive symptoms are important components of SGRQ scores [27, 28].
One limitation of our study is that it was performed in only one center; however, FEV1 values and median age of our patients were similar those of previous studies performed in hospitalized patients [4, 5, 7, 11], suggesting that our population is representative of the patients hospitalized for COPD exacerbation.
Our results enabled us to identify additional variables associated with readmission in COPD patients: hypercapnia at discharge, poor health-related quality of life assessed using a specific questionnaire and incidence of previous admission. Our findings should help to delineate a subset of COPD patients that could benefit best from an active interventional program [29,30,31,32] with the aim of lowering hospital readmission rates and costs. A prospective interventional study incorporating these premises is warranted.
We thank Dr. M. Bustins from the ‘Servei Català de la Salut’ for data management, Dr. Manuel Alvarez del Castillo for her careful review of the manuscript and Joan Vila from the ‘Institut Municipal d’Investigacio Medica’ for his assistance in the statistical analysis.
Pedro Almagro, MD
Departamento de Medicina, Universitat Autònoma de Barcelona
Hospital Mútua de Terrassa, Plaza Dr. Robert 5
ES–08221 Terrassa (Spain)
Tel. +34 937365050, Fax +34 937365059, E-Mail firstname.lastname@example.org
Received: September 30, 2004
Accepted after revision: March 30, 2005
Published online: September 6, 2005
Number of Print Pages : 7
Number of Figures : 3, Number of Tables : 4, Number of References : 32
Respiration (International Journal of Thoracic Medicine)
Vol. 73, No. 3, Year 2006 (Cover Date: May 2006)
Journal Editor: Bolliger, C.T. (Cape Town)
ISSN: 0025–7931 (print), 1423–0356 (Online)
For additional information: http://www.karger.com/RES