Between April 1, 2009 and September 30, 2011, a prospective study was conducted in consecutive patients presenting AECOPD admitted to the general ward of the respiratory department of a tertiary hospital in Peking, China. For patients admitted more than once during the study period, only the first admission was included in the analysis. Only admissions of more than 24h duration were included. Inclusion criteria were: COPD diagnosis and presence of exacerbation according to the Global Initiative of Chronic Obstructive Lung Disease definition.14 The stable-stage post-bronchodilator spirometric values recorded in the patients’ medical history in the 6 months prior to study inclusion were used to diagnose COPD. Exclusion criteria were: asthma (excluded by bronchodilation test and history); bronchiectasis (excluded by high-resolution CT); pneumonia; cancer; sleep apnea syndrome; or other forms of active lung disease; hospitalization for reasons other than COPD exacerbation including acute coronary syndrome; congestive heart failure; need for intubation; length of stay (LOS) longer than 30 days; long-term oral corticosteroid (CS) therapy (more than 3 months treatment with 7.5mg per day of prednisone or equivalent); patients who had received systemic CS for their exacerbation for more than 48h before presentation; relapse within 14 days of initial presentation15; patients who died without being readmitted for an AECOPD during the follow-up period. Written informed consent was obtained from all patients, and the Ethics Committee of the hospital approved the study protocol.

Variables were collected at admission, and included demographic data, stable-state spirometric values, comorbid conditions, ongoing management before admission, smoking habits, exacerbation frequency in the previous year, and admission symptoms. Exacerbation frequency in the previous year was based on the number of exacerbations reported by the patient in the year prior to recruitment. Previous studies have shown a good correlation between the number of exacerbations recorded in the medical history and the number of exacerbations remembered by the patient over the same 1-year period.16 According to the definition of Hurst et al, frequent exacerbator was defined as exacerbation frequency ≥2 in the previous year, non-frequent exacerbator was defined as exacerbation frequency <2 in the previous year.17 On the day of admission, AECOPD Anthonisen type was determined according to the symptoms presented before starting treatment.7 At admission, at discharge and 8 weeks after discharge (stable status) patients completed the COPD Assessment Test (CAT) based on their symptoms experienced on that day. Classification based on the GOLD 2011 combined assessment of COPD was carried out using the spirometric parameters in stable status, exacerbation frequency, and CAT data obtained in stable status.18

Following GOLD guidelines, patients were treated with nebulized salbutamol, ipratropium bromide and budesonide, intravenous prednisolone at a dose of 30–40mg daily. Duration of intravenous therapy was 4 days. On day 4, patients were switched to an oral tapering schedule of prednisolone. Antibiotics were administered if bacterial infection was suspected (patient-reported sputum purulence) and adjusted according to antimicrobial susceptibilities, when available.

Mechanical ventilation was started by the attending physician for indications such as respiratory arrest, deterioration in level of consciousness, and increasing partial pressure of carbon dioxide (PaCO2), despite optimal pharmacological treatment. Non-invasive ventilation was used initially whenever possible and indicated. Decisions regarding admission or transfer to ICU were taken by the treating unit.

Patients were followed from the day of discharge until September 29, 2012 or the day of readmission, if earlier. The primary study outcome was readmission for AECOPD.

Patients were monitored monthly by telephone to document the occurrence of AECOPD hospitalization. At each telephone contact, a short questionnaire was administered to assess the change in respiratory symptoms and medical interventions in the past month. Patients were also encouraged to report to their attending physicians whenever they experienced symptom worsening. An event-based exacerbation was confirmed if patients experienced at least 1 key symptom worsening plus a change in at least 1 of 3 medications (antibiotics, corticosteroid, and bronchodilator). The end of an exacerbation episode was determined when patients’ symptoms improved and returned to their pre-episode status for at least 3 days, or when the symptoms improved and remained stable for at least 3 days. Readmission within 14 days of the previous discharge was not included in the final analysis, in order to distinguish relapse (symptom fluctuation in the same episode) from recurrence of exacerbations.19,20 The need for hospitalization was determined according to GOLD (Fig. 1).

Fig. 1.

Study flow chart. *A total of 173 eligible patients were included in this study, of whom 45 were withdrawn due to exclusion criteria, 33 died before any readmission, and 9 had relapse within 14 days of discharge. Accordingly, a total of 86 patients were included in the final analysis.

(0.12MB).

Patient sputum and plasma samples were collected on the day of admission, prior to initial treatment and, at discharge (day 10–14). Spontaneous sputum was used; when no spontaneous sputum was available, induced sputum was collected, according to a previous study on the equivalence of induced and spontaneous samples for assessment of airway inflammatory markers.19 Of the sputum samples taken, 97% were spontaneous. Samples was confirmed as valid if there were >25 polymorphonuclear leukocytes and <10 squamous cells present on Gram stain on low power (100×) magnification.21,22 Sputum samples were processed within 2h of collection and divided into two aliquots. One sample was processed with phosphate-buffered saline (PBS),14,23 cytospins were prepared, and cell-free supernatant was collected and stored in aliquots at −80°C pending analysis of soluble mediators. Differential cell counts were performed on May Grünwald Giemsa stained cytospins in a blinded fashion. Quantitative sputum cultures were performed using the second sample, according to previously described methods.12 Bacterial agents were classified as potentially pathogenic microorganisms (PPMs) or non-PPMs. Only PPM growth at thresholds higher than 105 colony-forming units for Streptococcus pneumoniae and 106 colony-forming units for other isolates was considered significant. Peripheral venous blood (7ml) was collected in a Vacutainer tube (BD Diagnostics, NJ, USA) and centrifuged at 6716 G for 10min at 4°C. Plasma was then separated and stored at −80°C until further analysis.

Statistical analysis was performed using SPSS version 11.5.0. Data were not normally distributed and nonparametric statistics were therefore used. Categorical variables are presented as n (%). Numerical variables are presented as median (interquartile ranges) for skewed numerical data. Wilcoxon's signed rank test was used for analysis of paired data, and the Mann–Whitney U test was used for paired group analysis. Categorical outcome variables were analyzed using Fisher probabilities. Statistically significance was set at P<.05.

Exploration of variables independently associated with readmission was conducted using a binary logistic model. Readmission was used as the dependent variable. Factors associated with readmission in the univariate analysis (P<.10) were used in a binary logistic regression predictive model. A P-value of less than .05 in the multivariate model was considered statistically significant.

A total of 173 eligible patients were included in the study. However, 45 patients were subsequently excluded due to pneumonia (n=6); congestive cardiac failure or left ventricular failure at admission (n=5); pulmonary embolism (n=3); need for intubation at admission (n=14); length of stay (LOS) longer than 30 days (n=5); having received systemic CS for their exacerbation for more than 48h (n=5). Seven patients had “long-term oral corticosteroid use” and were also excluded. Eleven of the 173 patients recruited did not survive the recruitment admission. Twenty-two patients died without having a readmission due to COPD exacerbation during the follow-up period and were excluded from the study of risk factors for a COPD readmission. Nine patients had relapse within 14 days of discharge. Therefore, a total of 86 patients were included in the final analysis; 47 patients (54.7%) were readmitted for AECOPD at least once during the 12 months following the index admission. Table 1 shows the characteristics of the groups. Compared with the no-readmission group, the readmission group comprised older patients; more patients with frequency of exacerbation in previous year ≥2; higher stable-status CAT value; more patients in GOLD group D; and more patients with diabetes.

Compared with patients without readmissions, patients with readmissions had higher hs-CRP value at discharge. Neutrophils, sputum IL-8 and MPO levels, and plasma IL-6 levels at both admission and discharge were also analyzed, although no significant differences were observed (Table 2).

The clinical indices at admission and discharge of the two groups are listed in Table 3. Compared with patients without readmissions, patients with readmissions had higher CAT values at admission and discharge.

Univariate comparison (Tables 1–3) demonstrated that age, frequency of exacerbation in previous year ≥2, stable-status CAT value, GOLD group, diabetes, cold symptoms at presentation, hs-CRP at discharge, CAT value at admission and at day 14, met entry criteria for logistic regression. The logistic regression model (Table 4) indicated that age, hs-CRP at day 14, and CAT value at day 14 were statistically significant independent variables for predicting readmission.

We used ROC curves to identify factors with diagnostic value for readmission for AECOPD. For hs-CRP, the best cutoff value for predicting readmission for AECOPD was ≥4.165mg/L at discharge (sensitivity 74.5%, specificity 81%). For CAT, the best cutoff value for predicting readmission for AECOPD was ≥14.08 at discharge (sensitivity 66.7%, specificity 66.7%). For age, the best cutoff value for predicting readmission for AECOPD was ≥68.5 (sensitivity 51%, specificity 73.8%).