Article Text
PDF
Abstract
Objectives The long-term clinical trajectory of chronic obstructive pulmonary disease (COPD) in terms of year-to-year hospital utilisation rates can be highly variable and is not well studied. We investigated year-to-year trends of hospitalisation or emergency department (ED) visits among patients with COPD over 3 years, identified distinct trajectories and examined associated predictive factors.
Design A retrospective cohort study.
Setting Data were extracted from the Changi General Hospital, Singapore COPD data warehouse.
Participants Patients with COPD aged ≥40 years with 3 years of follow-up data.
Primary and secondary outcome measures The yearly rates of hospitalisations or ED visits, stratified by COPD-related or all-cause, were described. Group-based trajectory modelling was used to identify clinically distinct trajectories year-by-year. Baseline predictive factors associated with different trajectories were examined.
Results In total, 396 patients were analysed (median age 70 years; 87% male). Four trajectories were generated for year-to-year trends in COPD-related hospitalisations/ED visits (C1–C4: consistently frequent, consistently infrequent, improving and worsening); post-bronchodilator forced expiratory volume in 1 second (FEV1) was a significant predictor of trajectory, with worse lung function being the main factor associated with less favourable trajectories. For all-cause hospitalisations/ED visits, four trajectories were identified (A1–A4: infrequent and stable, frequent and stable, frequent and decreasing, frequent and increasing); significant differences in age (p=0.041), sex (p=0.016) and ethnicity (p=0.005) were found between trajectories. Higher overall comorbidity burden was a key determinant in less favourable trajectories of all-cause hospitalisations/ED visits.
Conclusions Distinct trajectories were demonstrated for hospitalisations/ED visits related to COPD or all causes, with predictive associations between FEV1 and COPD trajectory and between comorbidities and all-cause trajectory. Trajectories carry nuanced prognostic information and may be useful for clinical risk stratification to identify high-risk individuals for preventative treatments.
- hospitalization
- pulmonary disease, chronic obstructive
- retrospective studies
- singapore
- prognosis
Data availability statement
Data are available upon reasonable request. The data sets supporting the results reported in this manuscript are the property of Changi General Hospital and are not publicly available. Access to the raw data may be granted on reasonable request to the corresponding author dependent on the intended use and subject to third-party agreements.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
This study used latent modelling as a novel approach to identify distinct year-to-year trends of chronic obstructive pulmonary disease (COPD) hospitalisation rates, providing more nuanced prognostic information than traditional analytic methods; for example, further insight into the relationship between forced expiratory volume in 1 second and exacerbations.
As COPD is a multisystem disease with a host of comorbidities, both COPD-related and all-cause hospitalisations were examined, with different predictive factors found for each.
The trajectories were derived using data from a single centre and may not be fully reproducible in other healthcare systems; longer follow-up periods (ie, >3 years) will be useful for understanding long-term prognosis.
The small size of some subgroups means that differences in characteristics between groups may have been overestimated/underestimated or significant statistical differences not detected.
Hospitalisations in the third year could have been impacted by the COVID-19 pandemic.
Introduction
Chronic obstructive pulmonary disease (COPD) is characterised by persistent respiratory symptoms and progressive and irreversible airflow limitation.1 The burden of COPD is substantial; it is the third leading cause of death worldwide and was responsible for an estimated 3.23 million deaths in 2019.2 Exacerbations of COPD, and hospitalisations associated with exacerbations, represent a major source of the healthcare burden and costs associated with the disease.3 In Singapore, the prevalence of COPD is estimated to be around 6%.4 Between 2018 and 2019, the average length of stay following hospitalisation due to COPD was 3.8 days, and the average median bill for unsubsidised patients was S$3270 (approximately US$2400).5 Frequent exacerbations of COPD lead to a decrease in quality of life, accelerated lung function decline and a higher risk of death.6–8 COPD is also associated with multiple comorbidities, which can result in further unscheduled healthcare utilisation.9 10
Assessing the risk of future hospitalisation and/or emergency department (ED) visits due to exacerbations is an important component of COPD management, and is aimed at identifying high-risk individuals who may benefit from preventative treatments to reduce recurrent exacerbations. Reducing hospital COPD readmission rates is also a key quality indicator in many healthcare systems.11 12 Therefore, it is important to understand the key predictors and drivers of hospitalisation for patients with COPD.
One such predictor/driver could be an intrinsic predisposition to recurrent exacerbations, in which future exacerbations are strongly predicted by history of previous exacerbation.13 As such, history of exacerbation in the past year (either ≥2 events or ≥1 requiring hospitalisation) is the current recommended strategy to identify patients at high risk of future events.1 However, other clinical trajectories—whether favourable or unfavourable—may also be possible, as evidenced by the high variability of exacerbation rates observed from patient-to-patient, as well as within-patient from year-to-year.13 14 This marked variability and apparent unpredictability of exacerbation rates among patients may be due to the complex interplay of disease factors, exacerbation precipitants, comorbidities, treatable traits and response to treatment, culminating in multiple and distinct clinical trajectories.
There is a paucity of studies on long-term clinical trajectories of COPD exacerbation rates. Most trajectory analyses have focused on changes in lung function and airflow limitation instead of exacerbation rates, and studies investigating exacerbation rates over several years have not been aimed at discovering latent trajectories of COPD exacerbation rates.15 Understanding the various long-term clinical trajectories of COPD allows for more nuanced prognostication and risk stratification and may help to guide treatment.
The aim of this study was to identify distinct latent trajectories of COPD in terms of year-to-year trends in the frequency of hospitalisations or ED visits, either COPD-related or all-cause, over a 3-year period in a cohort of patients with COPD in Singapore. In doing so, we can understand influencing factors that determine different clinical trajectories in patients with COPD.
Methods
Study design and data source
This was a retrospective analysis of data that was prospectively collected for routine clinical management, extracted from the Changi General Hospital (CGH) COPD data warehouse. To be included in the data warehouse, patients were identified using primary or secondary International Classification of Diseases, 10th revision, Australian Modification COPD diagnoses codes (ie, J440, J441, J448, J449) for either an inpatient or ED visit to CGH. The warehouse contains data starting from October 2017 and is updated prospectively on a rolling basis.
Study population
The study cohort used for modelling was formed from patients aged ≥40 years with a primary or secondary diagnosis of COPD for a hospitalisation or an ED visit on or after 1 October 2017 and having 3 years of follow-up (from 1 October 2017 to 30 September 2020) (online supplemental figure S1). Because complete visit data for 3 years were required for modelling, we excluded patients who died before 30 September 2020 (ie, before the end of the third year) and patients not on follow-up (defined as no inpatient/outpatient/ED visits whether routine or unscheduled, for any reason or medical condition, for at least 1 year during the study period). We also excluded those who did not have supportive evidence of COPD (defined as a record of spirometric evidence of airflow limitation (forced expiratory volume in 1 s (FEV1)/forced vital capacity <0.7), a prescription for a COPD medication or radiological evidence of emphysema).
Supplemental material
Outcome measures
The primary measured outcome of the study was the yearly rate of hospitalisations or ED visits, stratified according to COPD-related or all-cause for the 3-year study duration. In this study, COPD-related hospitalisations or ED visits were considered to represent moderate-to-severe exacerbations of COPD. Hospitalisations and ED visits were treated as discrete events; an ED visit leading to hospitalisation in the same encounter was counted as a single event. Treatment regimens were described yearly; for each patient, this was based on the first prescription filled each year. The dispensation rate of each maintenance therapy (based on all prescriptions throughout the year) was calculated as the ratio of the number of inhalers dispensed/the number of inhalers prescribed, providing a surrogate measure of adherence.
Data analysis
All variables collected and included in the data warehouse are shown in online supplemental table S1. Recorded FEV1% predicted values were used to assign a patient’s Global Initiative for Chronic Obstructive Lung Disease (GOLD) grade of airflow limitation (grade 1–4).1 Categorical variables were described using the number of observations and percentage or the median and IQR, and continuous variables were described using the mean and SD.
Study objectives
The primary objective of the study was to identify clinically distinct trajectories of hospitalisation/ED visit frequency (COPD-related and all-cause) from year-to-year. The secondary objective was to identify predictive factors associated with different trajectories of hospitalisation/ED visit rate, stratified by COPD-related or all-cause.
Group-based trajectory modelling
Distinct trajectories were identified using group-based trajectory modelling,16 17 which assumes that the population is composed of distinct groups, each with a different underlying trajectory of a variable measured repeatedly over time. The trajectories were numbered for moderate-to-severe COPD exacerbations (C1–C4) and all-cause admissions (A1–A4). Trajectories were then analysed separately for COPD-related hospitalisations/ED visits and all-cause hospitalisations/ED visits, and the Stata procedure ‘Proc Traj’ was used for this analysis. The frequency of hospitalisations or ED visits was modelled as a zero-inflated Poisson distribution; outliers that skewed the trajectory were omitted. Different models with varying numbers of groups and polynomial orders were compared to find the best-fit model; the number of latent trajectories was increased stepwise in different iterations of the model from 2–5 to find the best fitting number of trajectories, selected using the Bayesian information criteria. Patients were classified to a specific trajectory on the basis of the maximum estimated probability of assignment; a probability of 0.9 or higher for membership of a particular patient in a particular trajectory was considered an excellent fit, whereas a value of <0.7 was considered a poor fit.18 Multinomial logistic regression was used to evaluate selected baseline variables as predictors of trajectory.
Patient and public involvement
Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Results
In total, 1619 patients aged ≥40 years with a primary or secondary diagnosis of COPD were identified from the data warehouse (online supplemental figure S2). After exclusions, 396 patients were available for trajectory derivation. Outliers were excluded for the derivation of trajectories; two patients were excluded from the analysis of trajectories of COPD-related hospitalisations/ED visits, and seven patients were excluded from the all-cause trajectory analysis.
Baseline characteristics
Baseline characteristics (during the first year of the study) are shown in online supplemental table S2. The median (IQR) patient age was 70 (60–79) years and 87% were male. The predominant ethnicity was Chinese (55%) followed by Malay (24%). In total, 41% of patients were current smokers and 40% were former smokers. The prevalence of asthma was 18%. Of the patients with available data (74.7%), the median post-bronchodilator FEV1% predicted was 70% and the majority of patients were categorised as GOLD grade 1 (22%) or GOLD grade 2 (39%). The most common comorbidities were hypertension (51%) and ischaemic heart disease (41%). The average rate of moderate-to-severe COPD exacerbation was 0.99 per person-year during baseline. The average rate of all-cause hospitalisation/ED visit was 3.07 per person-year during baseline. The baseline characteristics of the excluded patients are shown in online supplemental table S3. Patients who died were older and had a higher overall comorbidity burden compared with the patients included in the analysis.
Year-by-year trend in the number of COPD-related hospitalisations/ED visits (representing moderate-to-severe exacerbations) is shown in figure 1. Overall, 100 of 396 (25.3%) patients consistently had zero hospitalisations/ED visits for COPD in all 3 years, and 25 of 396 (6.3%) patients consistently had ≥2 hospitalisations/ED visits per year for COPD in all 3 years. The remaining patients appeared to have inconsistent exacerbation status over time, characterised by variable number of exacerbations from year-to-year.
Variability in frequency of moderate-to-severe exacerbations over the 3-year study period. The width of the bands is proportional to the number of patients. Exac, exacerbation(s).
Latent trajectories of moderate-to-severe exacerbations
To discern latent trends, the same data set was subjected to group-based trajectory modelling (figure 2). For the year-to-year trends of COPD-related hospitalisations/ED visits, four trajectories were generated by the model. These were labelled C1 (consistently infrequent, 69.8% of patients), patients had stable disease with few and intermittent exacerbations (on average <1 exacerbation/year) across all 3 years; C2 (worsening, 16.4%), patients had few exacerbations initially (<1 exacerbation/year), but exacerbation rate increased over subsequent years (approximately two exacerbations/year at year 3); C3 (improving, 8.9%), patients had frequent exacerbations initially (>3 exacerbations/year), but exacerbation rate decreased over time (approximately one exacerbation/year at year 3); and C4 (consistently frequent, 4.9%), patients had very high rates of moderate-to-severe exacerbations (≥5 exacerbations/year) over all 3 years.
Group-based trajectory models for year-to-year trends of COPD-related hospitalisations/ED visits. Dots represent the observed rates of moderate-to-severe exacerbations. Dotted lines represent the 95% confidence intervals. COPD, chronic obstructive pulmonary disease; ED, emergency department.
There were no differences identified in demographics, smoking history, pattern of comorbidities or COPD Assessment Test score (based on patients with available data) between the four trajectories (table 1). Using trajectory C1 (consistently infrequent) as the reference, post-bronchodilator FEV1 values were a significant predictor for trajectory; increasing COPD GOLD grade was associated with significantly increased odds of belonging to trajectory C2 (worsening, OR=2.94, 95% CI: 1.81 to 4.77; p<0.001) or trajectory C4 (consistently frequent, OR=4.45, 95% CI: 2.16 to 9.14; p<0.001) (online supplemental figure S3).
Comparison of baseline characteristics between trajectory groups of COPD-related hospitalisations/ED visits
Treatment regimens in the first year included: no treatment (24.6% of patients), short-acting β2-agonist/short-acting muscarinic antagonist (17.3%), long-acting muscarinic antagonist (LAMA; 7.9%), LAMA+long-acting β2-agonist (LABA; 11.7%), inhaled corticosteroid (ICS)+LABA (12.9%), triple therapy (ICS+LAMA+ LABA; 20.3%) or other (5.3%). Treatment regimens by trajectory are shown in online supplemental figure S4. For patients on trajectory C1, ‘no treatment’ was the predominant regimen at baseline, declining across the 3 years. Conversely, the proportion of patients receiving LAMA+LABA therapy increased and was the predominant regimen at year 3. For patients on trajectory C2, triple therapy was the predominant treatment; LAMA+LABA therapy also became highly prescribed. For patients on trajectory C3 and C4, triple therapy was the predominant therapy across all 3 years. Across all trajectories, the median (IQR) dispensation rate was 1.00 (0.67–1.00) in the first year, 0.94 (0.60–1.00) in the second year and 0.92 (0.58–1.00) in the third year. There was no difference in dispensation rates between trajectory groups in any of the 3 study years.
Trajectories of all-cause hospitalisations/ED visits
For the year-to-year trends of all-cause hospitalisations/ED visits, four trajectories were generated by the model (figure 3). These were labelled A1 (63.4% of patients), patients had infrequent hospitalisations/ED visits with a mean rate of 1.68 per person per year across all 3 years; A2 (25.8%), patients had frequent hospitalisations/ED visits with a mean rate of 3.72 per person per year across all 3 years; A3 (9.0%), patients had frequent healthcare utilisation, but the rate of hospitalisations/ED visits decreased over time (7.69 per person per year to 4.22 per person per year); and A4 (1.8%), patients had a high rate of hospitalisations/ED visits, which increased over the subsequent years.
Group-based trajectory models for year-to-year trends of all-cause hospitalisations/ED visits. Dotted lines represent the 95% confidence intervals. ED, emergency department.
Significant differences in age (p=0.041), sex (p=0.016) and ethnicity (p=0.005) were found between trajectories (table 2). Compared with trajectory A1, trajectory A4 was associated with younger age (OR=0.92, 95% CI: 0.85 to 0.99; p=0.022), trajectory A2 was associated with higher prevalence of the male sex group (OR=2.64, 95% CI: 1.00 to 6.95; p=0.049) and trajectories A3 and A4 were both associated with lower prevalence of Chinese ethnicity (OR=0.18, 95% CI: 0.07 to 0.49; p=0.001 and OR=0.08, 95% CI: 0.01 to 0.93; p=0.043, respectively). Trajectory A1 was associated with a lower Elixhauser Comorbidity Index, compared with all other trajectories, even after adjustment for age, sex and ethnicity.
Comparison of baseline characteristics between trajectory groups of all-cause hospitalisations/ED visits
Discussion
The primary objective of this study was to identify distinct clinical trajectories of year-to-year trends of hospitalisations/ED visits for a cohort of patients with COPD in Singapore. We also examined changes in COPD treatment patterns over time for different trajectories. The results indicate that the marked variability and apparent unpredictability of hospital utilisation rates can be distilled into distinct and coherent clinical trajectories using longitudinal latent modelling, and that clinical variables, such as lung function and overall comorbidity burden, can be used to identify future clinical trajectory. We observed a general trend of escalating treatment over time. We found that the use of triple therapy was high among patients with frequent exacerbations (trajectories C2, C3 and C4), although triple therapy use decreased among patients with decreasing exacerbations (trajectory C3).
Previous studies have described a ‘frequent exacerbator’ phenotype of COPD, in which patients persistently experience recurrent events over the longer-term; however, more recent studies have reported a considerable variation in exacerbation rate from year-to-year, suggesting that future exacerbation risk cannot be accurately predicted based on exacerbation history alone.13 14 Our study has demonstrated that a seemingly unpredictable year-to-year variability in exacerbation rate can be distilled into distinct and coherent underlying trajectories. Our results confirm the presence of a small group of individuals with persistent frequent exacerbations every year, corresponding to the ‘frequent exacerbator’ phenotype. However, we also found other trajectories that are more prevalent among this cohort of patients; this includes individuals who remain stable with few exacerbations over time, those whose exacerbation rates improve over time and those whose exacerbation rates worsen over time. In clinical practice, these trajectory patterns are implicitly recognised, and using group-based trajectory modelling as a novel approach to analyse hospital utilisation rates, the present study has formally and quantitatively described these trajectories and evaluated causal associations of clinical trajectory.
In our study, we found that FEV1 at baseline was a significant predictor of disease course (incidence of moderate-to-severe exacerbations resulting in hospitalisation or ED visit) over a 3-year period. The significance of FEV1 as a predictor of exacerbation risk in patients with COPD is controversial. Airflow limitation has been reported to be dissociated from indices of airway inflammation,19 and some patients with COPD are susceptible to recurrent exacerbations, irrespective of spirometric assessment of airflow limitation.13 In addition, GOLD guidelines state that FEV1 alone should not be used clinically as a predictor of exacerbation risk in patients with COPD.1 Despite this, improvements in lung function have been found to be associated with a reduced exacerbation risk in therapeutic trials; a previous analysis of pooled data from three studies (n=3313 patients) found that positive changes in FEV1 were associated with improved patient-reported outcomes (including St George’s Respiratory Questionnaire, Transition Dyspnoea Index, exacerbation rate and rescue medication use), independent of the bronchodilator formulation used.20 However, it has been suggested that the effect of the reduced exacerbation risk may be small.21 Our results offer a nuanced insight into the relationship between FEV1 and exacerbations: whilst impaired FEV1 is associated with exacerbations overall, there seems to be an unravelling of this relationship when considering short-term exacerbation rates, and the effect of FEV1 is related more to whether clinical trajectory can improve over the longer-term. Those with poor lung function can exhibit fewer exacerbations initially, but poor lung function is associated with increasing exacerbations over time; similarly, those with better lung function may experience frequent exacerbations initially, but can improve over time.
Our results have implications for clinical practice as well as research. The current GOLD strategy relies on history of ≥2 exacerbations or ≥1 requiring hospitalisation in the past year to identify patients at increased risk of future events.1 This strategy will identify consistently frequent exacerbators (corresponding to the C4 trajectory), but may fail to identify other detrimental trajectories (eg, C2) and may wrongly classify those with initially frequent exacerbations but who later improve (C3) as high risk. Elucidating long-term trajectories thus represents more refined risk stratification. Other than risk stratification for preventative management, understanding the distinct clinical trajectories of COPD may also help to highlight divergent needs of patients on different trajectories and the associated care needs. For example, patients demonstrating frequent exacerbations every year despite maximal medical treatment may benefit more from multidisciplinary case management, and palliative and home/ambulatory services to reduce healthcare utilisation compared with stable patients who may only require infrequent routine care outpatient visits; this facilitates resource allocation. Reducing the high readmission rate is a key quality indicator in many healthcare systems. In the USA, financial penalties are incurred for 30-day all-cause hospital readmissions following hospitalisation for an exacerbation of COPD,11 and in Singapore public hospitals, the COPD readmission rate is a quality of care indicator.12 As COPD is associated with many comorbid conditions, strategies must also target reduction of readmissions for other causes, and so understanding the patterns of all-cause readmission following a COPD exacerbation is helpful. Identifying biomarkers that may predict future trajectory should also be a focus of future research.
The strengths of this study include the enrolment of patients whose COPD diagnosis is confirmed by multiple clinical criteria, including a COPD diagnostic code, post-bronchodilator airflow obstruction, radiological evidence of emphysema and prescription of COPD medications. As such, patients with COPD without a record of spirometry could be included in the study, capturing a proportion of patients often encountered in real-world practice. Also, the population of patients with COPD captured by the CGH database is likely representative of the patient population in other regions of Singapore, due to similar specialist inpatient and outpatient practice and resources.
The study does have some limitations to consider. As the trajectories were derived using data from a single centre, they may not be fully reproducible in other healthcare systems, and longer follow-up periods (ie, >3 years) may allow depiction of more accurate trajectories. The small size of some subgroups means that differences in characteristics between groups may have been overestimated/underestimated, with significant statistical differences not detected. Missing data is also a limitation common to studies using data collected in real-world practice. Hospitalisations and ED visits were considered as a composite outcome during this analysis; there may be variation in the characteristics of patients in each of these groups (ie, patients with frequent ED visits but few hospitalisations may have relatively good functional performance compared with patients with frequent hospitalisations, despite having a similar number of total exacerbations). Patients who did not regularly attend CGH throughout the study period (ie, less than one encounter in any one of the 3 study years) were excluded from the analysis, which may introduce selection bias. However, it could be that these patients were hospitalised for an exacerbation during the study period but were routinely managed at another centre. Therefore, the group of patients with infrequent exacerbations (ie, patients on trajectory C1) may have been underestimated. In addition, many patients died over the course of the study and were, therefore, excluded; this may potentially lead to a selection bias towards less severe patients, and as a result, the rate of hospitalisations/ED visits may have been underestimated. Another consideration is that treatment dispensation rate does not necessarily reflect treatment usage and adherence by the patient, and the model was unable to take into account any impact from treatment change. Finally, hospitalisations/ED visits in the third year could have been influenced by the COVID-19 pandemic.
Conclusions
Trajectories were identified to demonstrate distinct year-to-year trends of hospitalisations/ED visits for either COPD exacerbation or all-cause diagnoses. For COPD-related hospitalisations/ED visits, worse lung function or higher COPD GOLD grade was the main factor associated with less favourable trajectories. In contrast, a key determinant of less favourable all-cause hospitalisation/ED visit trajectory was higher overall comorbidity burden. Understanding the distinct clinical trajectories of COPD may help to highlight divergent needs of patients on different trajectories and the associated care needs, thus facilitating resource allocation.
Data availability statement
Data are available upon reasonable request. The data sets supporting the results reported in this manuscript are the property of Changi General Hospital and are not publicly available. Access to the raw data may be granted on reasonable request to the corresponding author dependent on the intended use and subject to third-party agreements.
Ethics statements
Patient consent for publication
Ethics approval
This study complied with all applicable laws regarding personal data protection. The study was granted exemption by the SingHealth Centralised Institutional Review Board (2018/2698). No direct subject contact or primary collection of individual human subject data occurred; therefore, informed consent was not required. In accordance with SingHealth personal data protection policies, the data were submitted to a trusted third party for de-identification and/or anonymisation prior to analysis.
Acknowledgments
Editorial support (in the form of writing assistance, including preparation of the draft manuscript under the direction and guidance of the authors, collating and incorporating authors’ comments for each draft, assembling tables and figures, grammatical editing and referencing) was provided by Rebecca Cunningham of Apollo, OPEN Health Communications, and was funded by GSK.
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
AY and XX contributed equally.
Contributors AY contributed to the study conception, acquisition of data and data analysis/interpretation. XX, CHL, PB, AANR, DM and AT contributed to the study conception and data analysis/interpretation. All authors made a significant contribution to drafting, revising and critically reviewing the manuscript; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. AY acts as guarantor and accepts full responsibility for the work and/or the conduct of the study, had access to the data and controlled the decision to publish.
Funding This study was funded by GSK (study number 215241). Award/grant number: N/A.
Disclaimer GSK-affiliated authors were involved in study conception and design, data analysis, data interpretation and the decision to submit the article for publication. GSK funded the article processing charges.
Competing interests AY and CHL have nothing to disclose. XX, PB, AANR and DM are employees of GSK and hold stocks/shares in GSK. AT was previously the co-chair of the Singapore Ministry of Health COPD Service Improvement Team.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.