Article Text
PDF
Abstract
Objective To assess the frequency and predictors of unplanned readmissions after hospitalisation for incident atrial fibrillation (AF) and the association of readmissions with mortality over 2 years.
Methods We performed a retrospective cohort study using Western Australian morbidity and mortality data to identify all patients, aged 25–94 years, who survived incident (first-ever) hospitalisation for AF (principal diagnosis), between 2001 and 2015. Ordinal logistic models determined the covariates independently associated with unplanned readmission(s), and Cox proportional hazards models with time-varying exposures determined the hazard ratios (HR) of one or more readmissions for mortality over 2 years after incident AF.
Results Of 22 956 patients, 57.7% male, mean age 67.9 (SD 13.8) years, 44.0% experienced 22 053 unplanned readmissions within 2 years, 50.6% being cardiovascular-related. All-cause death occurred in 8.0% of the cohort, and the multivariable-adjusted mortality HR of 1 (vs 0) readmission was 2.9 (95% CI 2.6 to 3.3), increasing to 5.6 (95% CI 5.0 to 6.5) for 3+ readmissions. First emergent readmission for AF, stroke, heart failure or myocardial infarction was independently associated with an increased hazard for mortality. Coexistent cardiovascular and other comorbidities were independently associated with increased readmission and mortality risk, whereas AF ablation was associated with reduced risk.
Conclusion This study highlights the large burden of unplanned all-cause and cardiovascular-specific readmissions within 2 years after being hospitalised for incident AF and their associated adverse impact on mortality. Concomitant comorbidities are independently associated with unplanned hospitalisations and mortality, which supports integrated multidisciplinary management of comorbidities, along with AF-targeted treatments, to improve long-term outcomes in patients with AF.
- Atrial Fibrillation
- Epidemiology
- Risk Factors
- Outcome Assessment, Health Care
- Quality of Health Care
Data availability statement
Data may be obtained from a third party and are not publicly available. Data may be obtained from a third party and are not publicly available. We will consider requests for data sharing on an individual basis, with the aim to share data whenever possible for appropriate research purposes. However, this research project uses data obtained from a third-party source under strict privacy and confidentiality agreements from the Western Australian Department of Health databases, which are governed by their ethics committee and data custodians. The data were provided after approval was granted from their standard application processes for access to the linked datasets. Therefore, any requests to share these data with other researchers will be subject to formal approval from the third-partyethics committees and data custodians. Researchers interested in these data should contact the Client Services Team at the Data Linkage Branch of the Western Australian Department of Health (http://www.datalinkage-wa.org.au/contact-us).
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Atrial fibrillation (AF) is associated with substantial morbidity and mortality, with hospitalisations posing the main burden on healthcare services and costs. However, there are limited population-level data on the pattern, predictors and associated mortality risk of unplanned (emergent) hospitalisations over several years after admission for incident AF.
WHAT THIS STUDY ADDS
Nearly half of all patients had one or more unplanned readmission within 2 years after hospitalisation for incident AF, and half of these readmissions were cardiovascular-related.
Occurrence of one to three or more (vs no) unplanned readmission after hospitalisation for incident AF was associated with a threefold to sixfold adjusted hazard of all-cause mortality over 2 years.
A first emergent readmission for AF, stroke, heart failure or myocardial infarction was each independently associated an increased hazard of all-cause mortality.
Concomitant cardiovascular and other comorbidities were independently associated with increased readmission and mortality risk, whereas AF ablation was associated with a reduced risk.
HOW MIGHT THIS STUDY AFFECT RESEARCH, PRACTICE OR POLICY
These findings underscore the importance of integrated multidisciplinary management of coexistent comorbidities, along with AF-targeted treatments, to potentially reduce the morbidity and mortality burden of AF.
Introduction
The economic and public health impact of atrial fibrillation (AF) in Western societies is substantial.1 2 The majority of healthcare burden associated with AF is related to hospitalisations and inpatient services, accounting for 50%–70% of all direct costs.3 4 In Western countries, hospitalisations for AF are escalating due to ageing populations, improved survival and advances in AF procedural treatments.5–9 Patients with AF are also at increased risk of non-cardiovascular hospitalisations associated with multimorbidity.6 10 Further, hospitalisation negatively impacts on quality of life11 and increases mortality risk in patients with AF.10 12–15
Previously, we reported that 1 in 10 patients will experience an unplanned readmission within 30 days after hospitalisation for incident (first-ever) AF, and this was associated with a twofold excess mortality risk over the subsequent year.15 However, contemporary population-level data on the pattern, predictors and mortality risk of unplanned cardiovascular and other emergent readmissions over several years after hospitalisations for incident AF are sparse. Therefore, we assessed the frequency, reasons and predictors of unplanned readmissions over 2 years in patients after hospitalisation for incident AF in Western Australia (WA) and determined the multivariable-adjusted hazard of readmission(s) for subsequent all-cause and cardiovascular mortality.
Methods
Data sources
Full hospitalisation and mortality records for patients with a cardiovascular-related hospitalisation between 1 January 1991 and 31 December 2017 were available from the WA Hospital Morbidity Data Collection (HMDC) and the Death Registry. Both data collections are linked via probabilistic matching and continuously audited for quality assurance.16 The HMDC includes all private/public hospitalisation records, demographic information, admission/discharge dates, principal/additional diagnoses and procedural fields, using the International Classification of Diseases, Ninth Revision (ICD-9) and International Classification of Diseases, 10th Revision–Australian modification (ICD-10-AM). The Death Registry includes date of death and underlying/associated causes of death within WA.
Cohort selection and follow-up
We identified all patients, 25–94 years, with a hospitalisation for incident AF or atrial flutter (principal diagnosis code ICD-9: 427/ICD-10-AM: I48) between 1 January 2001 and 31 December 2015, with no prior hospitalisation for AF on 10-year lookback. These AF discharge codes were demonstrated to have a high positive predictive accuracy within administrative data.7 Patients who died during the hospitalisation for incident AF and those with recurrent admissions for renal dialysis were excluded (figure 1). After discharge, patients were followed up for 2 years or until death, whichever came first.
Flow diagram showing the selection of cohort. AF, atrial fibrillation.
Baseline covariates
Demographic information including age, sex and length of stay (LOS) were identified from the incident admission. For each patient, Indigenous and Torres Strait Islander status (referred to as Indigenous) and comorbid conditions were identified using any diagnosis within hospitalisation records up to 10 years preceding and including the incident admission.7 17 18 Comorbidities identified included heart failure (HF), hypertension, ischaemic heart disease (IHD), myocardial infarction (MI), cerebrovascular disease, stroke, transient ischaemic attack (TIA), arterial embolism, peripheral arterial disease (PAD), valvular heart disease, syncope, chronic obstructive pulmonary disease (COPD), pneumonia, chronic kidney disease (CKD), cancer, diabetes mellitus, obesity, excessive alcohol use, anaemia and thyroid disease. Electrical cardioversion or catheter ablation was identified on the incident admission. Coding requirements for each condition/procedure are provided in online supplemental table 1. The sexless CHA2DS2-VA score was calculated per person using concurrent and 10-year comorbidity history.2 15
Supplemental material
Exposure and outcomes
The primary outcomes were the frequency and reasons of unplanned readmissions and all-cause mortality within 2 years after hospitalisation for incident AF. Readmissions were separated by elective or unplanned (emergent) status, then unplanned readmissions categorised by principal diagnosis according to cardiovascular and non-cardiovascular disease groupings (online supplemental table 2). Secondary exposures included the first emergent readmission for AF, HF, stroke/TIA/embolism or MI, and catheter ablation performed on any readmission. The secondary outcome was cardiovascular death determined by using the primary cause of death with an ICD-10-AM code between I00 and I99.
Statistical analysis
Baseline characteristics were summarised by mean and SD or median and IQR for continuous variables, and by number and proportion for categorical variables. Cumulative incidence functions estimated probability of a first all-cause, cardiovascular or non-cardiovascular unplanned readmission with all-cause death treated as a competing risk. The person-time mortality rates were calculated for patients according to their readmission category (0–3+).
Ordinal logistic regression models determined the covariates independently associated with unplanned readmission(s) within 2 years, and presented as OR and 95% CI. Multivariable Cox proportional hazards models determined the independent predictors of all-cause and cardiovascular mortality over 2 years, presented as HR and 95% CI. The hazards models also determined the multivariable-adjusted association between any unplanned readmission(s) treated as a time-updated exposure and all-cause or cardiovascular mortality. In addition, the association between first cardiovascular-specific readmission (for AF, stroke, HF, MI or catheter ablation) and all-cause/cardiovascular mortality was assessed. In all regression models, sex, age group (25–54, 55–64, 65–74, 75–84 and 85–94 years), Indigenous status, admission calendar year, LOS and all baseline comorbidities (table 1) univariately associated with the outcome (p<0.05) with no high levels of collinearity were included in the multivariable analyses. Results on Indigenous status other than proportion of cohort, are not presented due to data sovereignty. Interactions between age group, sex and admission year, were tested, but were non-significant and therefore not included in the multivariable analyses. The proportional hazards assumption was assessed by determining the interaction between the covariates and the log of time, which demonstrated no major violations. Additionally, the assumption of linearity was tested by including non-linear terms for the calendar year of admission, which were non-significant. Statistical significance was reached (p<0.05) where the 95% CI around an OR/HR did not cross unity. The statistical analyses were conducted in SAS V.9.4 for Windows.
Clinical characteristics of patients who survived an hospitalisation for incident AF between 2001 and 2015, stratified by frequency of unplanned readmission within 2 years of discharge
Patient and public involvement in study
Patient and public involvement was not applicable in this study.
Results
Of 35 286 patients with an index AF hospitalisation between 2001 and 2015, 11 524 had prior AF hospitalisation, 144 died during incident admission, and 662 were on renal dialysis, and therefore were excluded (figure 1). The final cohort comprised 22 956 patients, mean age 67.9±13.8 years, 57.7% male, 1.5% Indigenous, and with a median sexless CHA2DS2-VA score of 2 (IQR 0–3) (table 1). Almost half (44.0%) of the cohort experienced any unplanned readmission within 2 years, and patients in increasing readmission categories (vs none) were progressively older, more likely female, and had an increasing prevalence of cardiovascular and non-cardiovascular comorbidities, lifestyle risk factors (obesity, diabetes mellitus and excessive alcohol use), and higher CHA2DS2-VA scores (table 1). Conversely, readmitted patients were less likely to undergo electrical cardioversion or catheter ablation on incident admission.
A total of 10 102 patients experienced 22 053 unplanned readmissions (50.6% cardiovascular related) within 2 years, with 22.8%, 10.3% and 10.9% experiencing 1, 2 and 3+ readmissions, respectively (table 1). Among all readmissions, AF was the most frequent cause (21.3%), then MI/IHD/chest pain (11.0%), and HF (7.1%), whereas all stroke/TIA/embolism and major bleeding comprised 1.9% and 1.7%, respectively (table 2). Further, among all unplanned readmissions, 16.8% (n=3704) had AF recorded as an additional diagnosis, and among all cardiovascular-related readmissions, 58.4% (n=6526) had AF recorded as a principal or additional diagnosis.
Frequency and principal reasons for unplanned readmissions that occurred within 2 years after discharge from hospitalisation for incident AF
Of the cohort, 8.0% died within 2 years (n=1828), and cardiovascular disease was recorded as the primary cause in 41.2% of all deaths. The crude mortality rate per 100 person-years increased from 2.0 to 6.4, 10.7 and 21.1 for those with none vs 1, 2 and 3+ readmissions, respectively.
Figure 2 shows the cumulative incidence of first all-cause, cardiovascular or non-cardiovascular unplanned readmission with death treated as a competing risk. The cumulative mortality increased steadily over 2 years. The cumulative all-cause readmission probability was 23.4%, 32.7% and 45.1% at 180 days and 1 year and 2 years, respectively. The cumulative probability of first cardiovascular and non-cardiovascular unplanned readmission was 29.2% and 26.7%, respectively, at 2 years, with a similar temporal profile as all-cause readmissions, where half of readmissions occurred within 180 days.
Cumulative incidence (probability) of all-cause death and first all-cause, cardiovascular and non-cardiovascular unplanned readmission with death treated as a competing risk, after hospitalisations for incident atrial fibrillation.
Table 3 shows the univariate and multivariable-adjusted predictors of unplanned readmission as an ordinal variable. Advancing age, female sex, LOS, calendar admission year, and a prior history of most comorbidities and lifestyle risk factors (obesity, diabetes mellitus and excessive alcohol use) were independently associated with higher readmission odds, but electrical cardioversion and AF catheter ablation were negative associates. The OR for readmission was 1.43 (95% CI 1.41 to 1.46) per unit increase in CHA2DS2-VA score (p<0.001).
Ordinal logistic regression models showing ORs and 95% CIs of univariate and multivariable-adjusted predictors of unplanned readmission frequency (0, 1, 2 and 3+ vs 0) within 2 years after hospitalisation for incident AF
Table 4 shows the univariate and multivariable-adjusted predictors of all-cause and cardiovascular mortality over 2 years of follow-up. Advancing age, male sex, LOS, HF, MI, prior stroke/TIA/embolism, PAD, valvular heart disease, COPD, pneumonia, CKD, cancer, anaemia and excessive alcohol use were independently associated with increased all-cause mortality, while calendar admission year, electrical cardioversion and catheter ablation on incident admission were negative associates (table 4, all p<0.001). These same factors, excluding COPD, pneumonia, cancer, anaemia and catheter ablation, were also independent predictors of cardiovascular mortality. The HR for all-cause death increased by 1.58 (95% CI 1.54 to 1.62) per unit increase in CHA2DS2-VA score and with a similar HR for cardiovascular mortality (both p<0.001).
Cox proportional hazards models showing HRs and 95% CIs of univariate and multivariable-adjusted predictors of all-cause and cardiovascular mortality over 2 years after hospitalisation for incident AF
After adjusting for covariate predictors, the occurrence of 1, 2 and 3+ readmissions (vs 0) was associated with a mean 2.9, 4.6 and 5.6 adjusted HRs, respectively, for all-cause mortality and 2.0, 2.6, and 3.8 adjusted HR, respectively, for cardiovascular mortality (all p<0.001) (figure 3). First emergent readmission for stroke/TIA/embolism, HF or MI was each associated with a 2.6 to 3.6 adjusted HRs for all-cause mortality and even higher adjusted HRs for cardiovascular mortality compared with patients without these respective events (all p<0.001) (figure 3). First emergent AF readmission (vs none) was associated with a 1.2-fold and 1.4-fold adjusted HR for all-cause and cardiovascular mortality, respectively (p=0.014 and p=0.004) (figure 3). Patients who had catheter ablation (n=1566) on a subsequent admission were younger, more likely male and less comorbid than those without ablation (online supplemental table 3). Despite confounder adjustment, catheter ablation as a time-updated exposure was independently associated with a reduced HR for all-cause death (p=0.029) but not cardiovascular death (figure 3). A sensitivity analysis with the inclusion of an interaction term between prior HF and catheter ablation as a time-updated exposure was conducted and found not significant for all-cause mortality or cardiovascular mortality (both p>0.25).
Unadjusted and multivariable-adjusted HRs (95% CIs) for all-cause and cardiovascular mortality associated with any unplanned readmission (1, 2 and 3+ vs 0) and first readmission for AF, stroke (combined all stroke/transient ischaemic attack/embolism), heart failure or catheter ablation after hospitalisation for incident AF. AF, atrial fibrillation.
Discussion
In this current Australian population-based study, we determined that nearly half of all patients after hospitalisation for incident AF will experience one or more unplanned readmissions within 2 years, and one in five had multiple readmissions. Half of all unplanned readmissions were cardiovascular-related, most frequently for AF, HF or coronary events. All-cause and cardiovascular-specific readmissions were independently associated with a higher hazard of death within 2 years. Coexistent comorbidities were independently associated with readmission and mortality risk, whereas AF ablation was negatively associated. These findings highlight the high emergent hospitalisation burden in patients with AF and supports integrated and holistic management to reduce excessive hospitalisations and improve long-term prognosis (online supplemental figure 1).
Supplemental material
Long-term unplanned hospitalisation risk
Our study extends previous investigations of long-term hospitalisation risk in patients with AF but is unique in that we considered unplanned readmissions only in the primary outcome, as this more so reflects quality of care. The observed high long-term risk of cardiovascular hospitalisations is generally concordant with previous investigation in population or registry-based AF cohorts,6 10 12 19 and patients enrolled in oral anticoagulant (OAC) trials.13 20 However data are not strictly comparable as prior studies included all cardiovascular hospitalisations (including elective) over varying periods.
Excluding AF, the most common causes for cardiovascular hospitalisations were for HF and IHD events, highlighting the need to optimise treatment of these common comorbidities.12 17 20 Stroke and major bleeding events were each responsible for a minority (<2%) of all readmissions, which likely reflects advances in AF care particularly in stroke prevention and uptake of non-vitamin K-dependent OACs.1 2 Our study further indicates that patients with AF are as likely to experience an emergent non-cardiovascular as cardiovascular readmission and have a higher risk of readmissions from any cause within the first 6 months after hospitalisation for incident AF (figure 2). This finding emphasises the case complexity of many patients with AF and supports the evidence that multidisciplinary, team-based, transition-of-care interventions can improve patients’ capacity for self-care and reduce their risk of readmission.1 21
Impact of readmissions on mortality
We provide novel information with respect to the mortality impact of unplanned repeat hospitalisations as time-varying exposures after admission for incident AF. We demonstrated that the adjusted HRs for all-cause and cardiovascular mortality increased significantly with repeat readmissions, although noting that there was some overlap of the 95% CIs around the HRs for 2 and 3+ readmissions (vs 0) (figure 3). We also established that first emergent readmission for AF, stroke/TIA/embolism, HF or MI was each associated with excess hazard for both all-cause and cardiovascular death (figure 3). Previous studies in AF-affected patients have indicated that cardiovascular hospitalisation can be a surrogate marker for all-cause death.12–14 Our data further show that any unplanned hospitalisation, regardless of the reason, is a marker for worsening prognosis and may be used as a surrogate endpoint in studies of AF.
Readmission and mortality risk predictors
Women had a higher adjusted risk of readmission compared with men, whereas it was the converse for mortality. Unsurprisingly, we found that older age and chronic diseases and lifestyle risk factors, which can all exacerbate AF burden, are also drivers of recurrent hospitalisations8 10 and risk of cardiovascular and all-cause mortality.22–24 We found that prevalent cardiovascular conditions and risk factors like HF, atherosclerotic vascular diseases, valvular heart disease, diabetes mellitus, excessive alcohol use and CKD were all positively associated with readmission and all-cause/cardiovascular mortality risk, whereas other comorbidities like COPD, pneumonia, cancer and anaemia were more associated with non-cardiovascular mortality in addition to readmission risk (tables 3 and 4).22–24 Since the conventional CHA2DS2-VASc score comprises major cardiovascular risk factors, it may serve as a useful predictor of mortality in patients with AF.22 24 Here we document that the sexless CHA2DS2-VA score also provides a simple, though non-exhaustive, calculation of factors associated with readmission and mortality risk.
Our results suggest that the adjusted risk of unplanned readmissions has increased, while all-cause and cardiovascular mortality risk has declined over the same period, and notably, with advances in cardiac care, less than half of all deaths in AF-affected patients are due to cardiovascular causes.23 24 Our study reinforces the fact that all-cause mortality in AF is largely determined by comorbid conditions and level of multimorbidity22–24 and further supports the recommended Atrial Fibrillation Better Care (ABC) pathway, where the ‘C’ component represents cardiovascular and comorbidity risk optimisation.1 Notably, ABC-adherent management has been shown to significantly reduce the risk of cardiovascular events and all-cause death in patients with AF.25 However, a recent meta-analysis indicates that adherence to the ABC pathway remains suboptimal, being adopted in only one in every five patients.26
Conversely, electrical cardioversion and catheter ablation on the incident admission were independently associated with a lower adjusted odds of readmission and hazard of all-cause and cardiovascular mortality. This suggests that use of an active rhythm control strategy, including catheter ablation, can be associated with improved AF outcomes including risk of rehospitalisations and mortality.27 28 We also found that after covariate adjustment, subsequent AF ablation treated as a time-updated exposure was independently associated with a lower adjusted hazard of all-cause death. However, these observational results should be treated with caution, and a survival bias by indication in patients undergoing catheter ablation is a real possibility. Nonetheless, there is increasing evidence that an early rhythm control strategy including ablation in patients with AF may effectively reduce irreversible atrial remodelling and prevent AF-related deaths and cardiovascular events in high-risk patients.28 A case–control study using the US Nationwide Readmission Database (2016 to 2018) reported that catheter ablation for AF was associated with significantly lower AF and stroke-related admissions, but not HF or all-cause readmissions, although the follow-up was limited to around 6 months only.29 Further, a meta-analysis of randomised controlled trials of catheter ablation versus medical therapy found that catheter ablation reduces cardiovascular hospitalisations and recurrences of atrial arrhythmia for patients with AF and was also associated with all-cause mortality benefit that was mostly driven by patients with HF with reduced ejection fraction.30
Strengths and limitations
We are unable to establish a true cause-and-effect relationship between readmissions and mortality from an observational study. Although AF diagnosis within administrative data has a demonstrated high predictive accuracy,7 we are unable to ascertain AF pattern or duration, or to obtain patients’ symptom severity or haemodynamic status. We are also unable to adjust for socioeconomic, health system and medical treatment factors, including OAC use, which may impact on risk of readmission and/or mortality. However, inclusion of these data is unlikely to change our overall findings with respect to relative risk of readmission and mortality and the association of comorbidities with these outcomes. Despite confounder adjustment, we are unable to totally exclude survival bias by indication, particularly for AF ablation. Nevertheless, we are able to use a whole of WA population-based design to identify a ‘real-world’ hospitalised incident AF cohort, with virtually complete individual record linkage to assess for prior comorbidities/procedures over a 10-year lookback period. We are also able to detect all subsequent hospitalisations, with stratification by unplanned status, and all deaths where they occurred in our jurisdiction.
Conclusion
This study underlines the considerable burden of unplanned cardiovascular and other hospitalisations within 2 years after admission for incident AF and indicates that emergent hospitalisations, irrespective of cause, are associated with a high excess risk of death. Many concomitant cardiovascular and other comorbidities were independent associates of recurrent hospitalisations and mortality, whereas AF ablation was a negative associate. These findings strongly support integrated multidisciplinary management of multimorbidity as part of a comprehensive holistic care pathway, as well as AF-targeted treatments, to improve long-term outcomes in patients with AF.
Data availability statement
Data may be obtained from a third party and are not publicly available. Data may be obtained from a third party and are not publicly available. We will consider requests for data sharing on an individual basis, with the aim to share data whenever possible for appropriate research purposes. However, this research project uses data obtained from a third-party source under strict privacy and confidentiality agreements from the Western Australian Department of Health databases, which are governed by their ethics committee and data custodians. The data were provided after approval was granted from their standard application processes for access to the linked datasets. Therefore, any requests to share these data with other researchers will be subject to formal approval from the third-partyethics committees and data custodians. Researchers interested in these data should contact the Client Services Team at the Data Linkage Branch of the Western Australian Department of Health (http://www.datalinkage-wa.org.au/contact-us).
Ethics statements
Patient consent for publication
Ethics approval
The study conformed to the principles outlined in the Declaration of Helsinki and Australia’s National Statement on Human Research (2007, updated 2018). Ethical approval was obtained from the Western Australian Department of Health, Human Research Ethics committee (ethics number: 2014/55, 5 September 2016).
Acknowledgments
The authors wish to thank the Linkage, Data Outputs and Research Data Services Teams at the Western Australian Data Linkage Branch, in particular the WA Department of Health Hospital Morbidity Data Collection, and the Registrar General of the WA Department of the Attorney General for the provision of data.
References
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
Twitter @c_weber_cvd
Contributors CW, JH, SH, IL and TB conceived the study. CW, JH and TB contributed to the study design and methods. CW performed the data and statistical analyses with statistical advice from KM. CW drafted the manuscript. CW, JH, SH, KM, IL and TB interpreted the results, provided critical review and approved the manuscript for submission. CW responsible for the overall content and is the guarantor of the work.
Funding CW is currently a PhD student with the UWA School of Population and Global Health, and is the recipient of a Postgraduate Scholarship from the National Health and Medical Research Council of Australia, Centre of Research Excellence in Cardiovascular Outcomes Improvement (# 1111170).
Competing interests None declared.
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.