Article Text
PDF
Abstract
Objectives We aimed to assess the healthcare costs and impact on the economy at large arising from emergency medical services (EMS) treated non-traumatic shock.
Design We conducted a population-based cohort study, where EMS-treated patients were individually linked to hospital-wide and state-wide administrative datasets. Direct healthcare costs (Australian dollars, AUD) were estimated for each element of care using a casemix funding method. The impact on productivity was assessed using a Markov state-transition model with a 3-year horizon.
Setting Patients older than 18 years of age with shock not related to trauma who received care by EMS (1 January 2015–30 June 2019) in Victoria, Australia were included in the analysis.
Primary and secondary outcome measures The primary outcome assessed was the total healthcare expenditure. Secondary outcomes included healthcare expenditure stratified by shock aetiology, years of life lived (YLL), productivity-adjusted life-years (PALYs) and productivity losses.
Results A total of 21 334 patients (mean age 65.9 (±19.1) years, and 9641 (45.2%) females were treated by EMS with non-traumatic shock with an average healthcare-related cost of $A11 031 per episode of care and total cost of $A280 million. Annual costs remained stable throughout the study period, but average costs per episode of care increased (Ptrend=0.05). Among patients who survived to hospital, the average cost per episode of care was stratified by aetiology with cardiogenic shock costing $A24 382, $A21 254 for septic shock, $A19 915 for hypovolaemic shock and $A28 057 for obstructive shock. Modelling demonstrated that over a 3-year horizon the cohort lost 24 355 YLLs and 5059 PALYs. Lost human capital due to premature mortality led to productivity-related losses of $A374 million. When extrapolated to the entire Australian population, productivity losses approached $A1.5 billion ($A326 million annually).
Conclusion The direct healthcare costs and indirect loss of productivity among patients with non-traumatic shock are high. Targeted public health measures that seek to reduce the incidence of shock and improve systems of care are needed to reduce the financial burden of this syndrome.
- health economics
- cardiac epidemiology
- adult intensive & critical care
Data availability statement
Data are available on reasonable request. The data that support the findings of this study are available from the corresponding author, JEB, on reasonable request.
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
Provides a population-wide estimate of the costs associated with emergency medical services treated non-traumatic shock.
Includes both prehospital and in-hospital direct healthcare-related costs.
The current study does not include costs associated with recurrent and ongoing direct healthcare expenditure, and thus is likely a conservative estimate of the healthcare costs arising from non-traumatic shock.
Background
Shock is a clinical syndrome that is characterised by cellular and tissue hypoxia arising due to inadequate oxygen delivery, increased oxygen demand or a combination of these processes.1 2 Despite improved systems of care for the treatment of patients with shock, effective resuscitative and supportive measures, early antibiotic administration and improved reperfusion therapies, it remains a clinical syndrome that continues to have a high risk of significant morbidity and mortality.3–6
The treatment of patients with shock is complex and resource-intensive, placing significant demands on all elements of the healthcare system.7–9 Furthermore, the significant morbidity and high rates of mortality have broader impacts on the economy at large.10 To date, there is a paucity of data assessing the total cost burden of non-traumatic shock. Using a large Australian emergency medical services (EMS) database, which has been individually linked to health service medical records and the death index, we aimed to characterise the direct healthcare-related costs and productivity losses of non-traumatic shock treated by EMS.
Methods
Study design and participants
This state-wide, population-based cohort study included consecutive adult (≥18 years) patients who received EMS care for non-traumatic shock between 1 January 2015 and 30 June 2019 in Victoria, Australia, a state of 6.7 million people, with a land area of 227 444 km2. Ambulance Victoria is the sole provider of EMS care in Victoria and is primarily funded through the Victorian State Government.11 The EMS service is two tiered, consisting of approximately 4500 advanced life support paramedics capable of laryngeal mask airway insertion and medication administration (eg, analgesics and bronchodilators), and 500 intensive care paramedics with a wider scope of practice that includes endotracheal intubation and the capacity to administer additional medications (including epinephrine infusions). During the study period, epinephrine was the only vasoactive medication carried by EMS and was protocolised for the treatment of non-traumatic hypoperfusion of any aetiology.
For this study, shock was defined as a prehospital systolic blood pressure <90 mm Hg sustained for more than 30 consecutive minutes, or having received EMS-led epinephrine (either as an intravenous bolus or infusion) to support perfusion. Exclusion criteria were age <18 years at presentation, traumatic aetiology of shock and patients transferred by EMS to a private hospital (online supplemental figure 1).10 At the conclusion of each case, paramedics complete an electronic patient care record (ePCR) using a computer tablet, which is synchronised with a digital data warehouse. Each patient record was individually linked to the Victorian Emergency Minimum Dataset (VEMD), the Victorian Admitted Episodes Dataset (VAED), Victorian Ambulance Cardiac Arrest Registry and the Victorian Death Index (VDI) (see online supplemental methods). The linkage process has been described in detail previously.2 10 12
Supplemental material
Participant demographics, pre-existing comorbid conditions, prehospital observations and paramedic-initiated interventions were obtained from ambulance ePCR data. Socioeconomic status was determined by the Index of Relative Scio-Economic Disadvantage Score (IRSD), a validated measure that ranks individual postcodes into deciles of relative disadvantage.13 The score is derived from Census data that includes household income, education level, employment status, occupation, housing ownership and non-English-speaking background.14 For this analysis, we divided the IRSD into quintiles.13 Geographical remoteness was determined using the residential area postcode of each event via the Accessibility and Remoteness Index of Australia—a geographical accessibility index that divides Australia into five classes of remoteness (‘major city’, ‘inner regional’, ‘outer regional’, ‘remote’ and ‘very remote’) to reflect relative access to services.15 Due to low numbers of patients from ‘remote’ or ‘very remote’ regions, these groups were combined with the ‘outer regional’ group for this study.
For patients who survived to hospital, primary discharge diagnosis (eg, ST-elevation myocardial infarction) and diagnosis group (eg, cardiogenic shock) were determined from either the VEMD (for those discharged from the emergency department (ED)) or VAED (for those discharged after inpatient admission) listed primary diagnosis. Both VEMD and VAED diagnosis definitions are consistent with the International Classification of Diseases, 10th Revision, Australian Modification (ICD-10). Further details pertaining to the cohort derivation are detailed in the supplemental methods (Supplementary online supplemental methods section 1 and online supplemental figure 1). Details of in-hospital care (eg, intensive care admission, length of stay) were obtained from the VAED and VEMD.
Cost calculations
Direct healthcare-related cost arising from index episode of care
Direct healthcare-related costs (in Australian dollars, AUD) arising from each episode of care were estimated using a ‘bottom-up’ approach for each element of care, with adjustment performed using the Australian Health Price Index to reflect the costs from the 2020–2021 financial year (online supplemental figure 2).16 Costs associated with the prehospital phase of care were determined from 2021 to 2022 estimates of cost per ambulance transport which take into consideration the event location, emergency status, transport mode (road or air ambulance) and whether attendance required transport to hospital or only treatment at the scene. ED costs were determined from the Victorian-specific National Hospital Cost Data Collection for 2018–2019, which provides the average cost of ED episodes of care for various scenarios that include whether the patient was admitted to hospital, was transferred to another hospital, discharged against medical advice or died within the ED (online supplemental table 1).17 This methodology has been previously used and has been described previously.18
Costs associated with admission to a public hospital are funded through a casemix model, which is determined according to hospital activities and patients treated. Traditional casemix funding models classify patients according to diagnosis-related groups (DRGs) with each DRG funded at the same rate. However, in the Victorian healthcare system, this has been modified to apply weighting for hospital length of stay, resulting in the development of the Weighted Inlier Equivalent Separation (WIES) model. Therefore, admission costs were determined by multiplying the WIES weights by the appropriate WIES unit price for the financial year in which the episode of care occurred19 (online supplemental table 2). For patients requiring interhospital transfer for further management, cost for the second admission was estimated using the overall average admission cost of a non-transferred patient for the same final hospital discharge diagnosis. To account for patients who were unlinked to VAED or VEMD, mean and total costs are calculated assuming patients had the same diagnoses and cost profile as those with successful linkage as described previously.20
Economic impact of prehospital shock
Cohort participation
As the employment status of each patient prior to their index presentation was unknown, preadmission employment was modelled using age and gender-defined data from the Australian Bureau of Statistics (ABS) Economic Security dataset (online supplemental table 3).21 These data stratify the Australian population by gender and 5-year age increments (ranging from 15 to 79). We assumed that the ABS population dataset was representative of the study population, as many of these presentations (eg, cardiac arrest or acute myocardial infarction (AMI)) are not associated with a prodrome of illness.22 For the purpose of this study, those aged 80 and above were assumed not to be in the paid workforce. Using the ePCR demographics, the dataset was stratified by gender and age to match that of the ABS employment data, and the expected full-time and part-time employment was calculated. Those in part-time employment were assumed to work 0.5 equivalent full-time (EFT) load. A full-time load comprised 38 hours of work per week.
Markov model development
A Markov state transition model was developed to simulate follow-up of patients following their index presentation and was limited to a 3-year time horizon. Three states were included in the model, ‘alive and working’, ‘alive and not working’ and ‘deceased’. At the commencement of the simulation, patients were classified as either ‘alive and working’ or ‘alive and not working’ based on their age and gender, and all experienced a healthcare encounter with shock. Vital status of patients was noted at 24 hours, 30 days, 1 year and 3 years following index presentation and was determined using linkage to the VDI dataset. To determine 3-year vital status, we only included patients enrolled in 2015, to ensure complete VDI follow-up and assumed that the remaining cohort would experience comparable rates of survival at 3 years following presentation. The model, therefore, consisted of four cycles: cycle 1 at 24 hours, cycle 2 at 30 days; cycle 3 at 335 days (365 minus 30 days) and cycle 4 at 3 years (3 minus 1 year). It was assumed that for the initial 30 days (cycle 1), survivors who were working prior to entering the model were unable to work. For cycles 2 and 3, it was assumed among patients working prior to their hospital presentation, following 30 days, only 50% would resume working if they survived to 30 days, 1 year and 3 years. We elected to use a conservative rate of 50% of survivors returning to work as prior observational data has shown that only 52% of patients with AMI-related cardiogenic shock, and 43% of those with septic shock, returned to work within 1 year of discharge.23–25 Years of life lived (YLL), productivity-adjusted life-years (PALYs) and costs incurred beyond the first year of the model were discounted at a rate of 5% per annum in accordance with Australian recommendations.26 Additional information relating to data inputs for the model is presented in online supplemental table 4.
A counterfactual situation was used for the comparison, in which patients were assumed to have not suffered from shock. Hence, the counterfactual patients remained in their baseline living health states (‘alive and working’ and ‘alive and not working’, depending on age and sex) in cycles 1 and 2 of the model. In cycles 3 and 4, transition probabilities were drawn from contemporary data provided by the Australian Institute of Health and Welfare regarding age-adjusted and sex-adjusted mortality in Australia.27 The comparator group comprised 21 329 patients of identical age and gender to those in the shock group.
Economic impact
The model employed a human capital approach, focusing on lost productivity arising due to premature death due to the development of shock. Productivity was measured in terms of PALYs.28 PALYs are calculated by multiplying YLL by a productivity index, which ranged between 0 (no productivity, eg, those who are deceased or alive and not working) and 1 (full productivity). The model assigned an economic value to each PALY lived, using the mean hourly wage per hour worked for the 2020–2021 financial year, which was $A40.6, and obtained from the Australian System of National Accounts.29 It was assumed that for 1.0 EFT, an Australian worker worked 38 hours per week for 48 weeks of the year, generating $A73 827 of wage-related productivity per year. This was the economic value applied to each PALY. Lost productivity was, therefore, the difference between PALY-derived productivity generated by the comparator group without shock and those with shock. We elected not to ascribe a gross domestic product (GDP) value per hour worked due to likelihood of this value overestimating true GDP losses, due to replacement of deceased individuals in the workforce.30
Statistical methods
Continuous variables are presented as median and interquartile range (IQR) or mean and standard deviation (SD), as appropriate. Categorical variables are presented as frequencies and percentages. All tests were two tailed and assessed at the 5% significance level. P values for trend across groups were calculated using the Cochran-Armitage test for categorical variables, linear regression for parametric continuous variables and Jonckheere-Terpstra test for non-parametric continuous data as appropriate. To estimate patient and prehospital factors which influence estimated healthcare costs, a linear regression was performed with cost as the dependent variable. Statistical analysis was performed by using Stata V.16.1 for Windows and Microsoft Excel (Redmond, Washington).
Patient and public involvement
Patients and the public were not involved in the design, conduct, reporting or dissemination plans of this research.
Results
Over the 4.5-year study period, a total of 2 485 311 ambulance attendances occurred with 21 334 patients meeting study inclusion with prehospital shock (online supplemental figure 1). Baseline characteristics of these patients are presented in table 1. The average age was 65.9 (±19.1) years and 9641 (45.2%) were female. Initial event location was in a metropolitan setting for 14 862 (71.6%) cases and there was a greater proportion of patients from the lowest socioeconomic quintile, compared with the highest quintile (5280 (28.1%) vs 2433 (13%), p<0.01). Of those attended by ambulance, 69.7% (14 871) were transported to hospital. Of those transported patients, successful VAED linkage occurred for 65.8% (9780) and 76.9% (11 435) were linked to VEMD. Linkage with either VAED or VEMD was successfully performed for 81.4% (12 104) of patients transferred to hospital by EMS.
Baseline characteristics of cohort
Healthcare-related costs
Over the study period, the average estimated direct healthcare costs for various clinical scenarios are presented in table 2. The overall average estimated cost for each episode of care was $A11 031 (±22 284) per patient and the total expenditure was calculated to be $A279 726 715 for the study period ($A62 161 492 per annum). The average estimated episode cost for prehospital care was $A1913 (±2661). There was variation in cost depending on the ED disposition scenario, overall hospital length of stay and event location. For patients transported to hospital for ongoing care, direct discharge from the ED was associated with lower estimated per episode costs (ED management and discharge $A2468 (±1605), discharge from ED against medical advice $A2602 (±3345) or died in the ED $A4169 (±1160)). The need for hospital admission from ED was associated with increased expenditure, with a further increase for those who required ED management and subsequent transfer to an alternate medical facility ($A20 760 (±28405) vs $A31 047 (±24746) per episode). Increased hospital length of stay was also associated with a stepwise increase in per episode cost (0–24 hours, $A7685 (±8527) vs 24–72 hours, $A15 343 (±11043) vs >72 hours, $A33 263 (±38215), p<0.001). Total estimated healthcare costs remained stable throughout the study period and are presented in figure 1A p-trend=0.33. However, there was an annual increase in the average estimated cost per episode of care, p-trend=0.05.
(A) Trend in total annual and mean episode direct healthcare costs ($AUD) for patients with emergency medical services (EMS) treated shock in Victoria, Australia (2015–2019). (B) Total ($millions, AUD) and mean episode ($thousand, AUD) costs stratified by shock aetiology among patients who survived to hospital.
Mean episode, total and annual direct healthcare costs for emergency medical services treated shock according to patient characteristics, admission characteristics and event location
Among patients who survived to hospital, the estimated costs were stratified by diagnosis and diagnostic group and are displayed in table 3. Obstructive aetiologies of shock were the most expensive with an average episode cost of $A28 057, but this was also the least common diagnostic group with only 120 patients over the study period. The second most expensive group was cardiogenic shock with an average episode cost of $A24 382, followed by septic shock ($A21 254), hypovolaemic shock ($A19 915) and non-specific ($A12 503) aetiologies (figure 1B and online supplemental figure 1B).
Diagnosis-specific costs for patients presenting to hospital by ambulance with shock
A linear regression model assessing prehospital and patient factors and their impact on total estimated healthcare costs is contained in the supplemental materials (online supplemental table 5). After adjustment, age over 65 years at presentation ($AA1490, 95% CI $A542 to $A2439, p<0.001), transfer from the ED to an alternate hospital ($A12 766, 95% CI $A10 702 to $A14 829, p<0.001), prehospital intubation ($A9482, 95% CI $A8035 to $A10 929, p<0.001) and intensive care admission ($A22 254, 95% CI $A21 196 to $A23 312, p<0.001) were independently associated with increased estimated healthcare costs.
Lost productivity associated with shock presentations
Modelled costings
From the total 21 334 patients in the cohort, 21 329 were included in the Markov model with 5 patients from the original cohort omitted due to gender data not being available in the ePCR. After matching with the ABS dataset, it was estimated that at the time of ambulance attendance for shock, 7613 (35.7%) were working in some capacity, equating to 6427 full-time jobs (6427 EFT). At 3 years following presentation with shock, 1940 (9.1%) were alive and working, 7496 (35.1%) were alive and not working and 11 892 (55.8%) had died.
Modelled estimates demonstrated that 28 992 YLL (discounted) by the shock cohort over a 3-year horizon, compared with 52 971 YLL (discounted) by the counterfactual comparator group without shock. This equates to 24 355 YLL lost (discounted). Among the shock cohort, 3098 PALYs (discounted) lived over the 3-year horizon, compared with 8157 PALYs (discounted) in the counterfactual cohort without shock, equating to 5059 PALYs (discounted) lost (table 4).
Lost productivity reported over a 3-year horizon
Productivity losses arising from premature mortality and resulting lost human capital, presented as lost productivity, over the 3-year horizon equated to $A373 500 267 (discounted). Lost productivity was attributed to shocked patients who survived and were working generating $A229 million (discounted), compared with $A602 million (discounted) in the counterfactual cohort without shock. This represents a 62.0% loss of productivity, equating to $A17 511 (discounted) per person. When extrapolated to the entire Australian population, productivity losses approached $A1.5 billion.
Discussion
In this population-based cohort study, we demonstrated that the direct healthcare costs and broader economic impacts of patients with non-traumatic shock in the prehospital setting are substantial. These data were produced through the unique linkage of four state-wide datasets (EMS, ED, hospital admissions and death index), allowing for a comprehensive assessment of the financial costs of this syndrome. The results from this study demonstrate that in Victoria, Australia the estimated direct healthcare costs over the study period equated to $A280 million (mean episode cost $A11 031), with an average cost for cardiogenic shock of $A24 382 and septic shock of $A21 554, per episode of care. Furthermore, modelling demonstrated that over a 3-year horizon, approximately 24 355 life-years and 5059 PALYs were lost. Lost productivity was also significant with $A374 million in lost productivity, which when extrapolated to the entire Australian population was $A1.5 billion in productivity-related losses. These modelled findings were derived in the context of a well-supported, conservative model.
The effective management of shock is complex with first medical contact frequently occurring in the prehospital setting. Prior data assessing the healthcare-related costs of patients with shock have derived their cohorts from inpatient administrative samples, using ICD-based definitions of shock.9 31 Due to the granular data available from the prehospital medical record, this study was able to define shock by haemodynamic criteria and included patients with sustained hypotension or the need for paramedic-initiated vasoactive medications to support perfusion. The definition we have used attempts to harmonise our patient cohort with that of contemporary trial and guideline definitions of shock, which may not be possible to achieve with an ICD-based definition.3 32–34 Furthermore, unlike previous data, the current study was able to capture the entire prehospital shock population (including 5664 patients who died prior to arrival to hospital). The use of a prehospital cohort, therefore, has two unique benefits; first, it allows for the inclusion of a significant proportion of patients who die prior to hospital arrival and would otherwise not be included in an in-hospital cohort. Second, it allows the direct costs of prehospital care to be accounted for in the analysis. In totality, by electing to use a prehospital patient cohort, this analysis presents a more accurate reflection of the total direct healthcare and economic costs of this clinical syndrome.
In the Australian context, there has been a temporal increase in the costs of delivering healthcare. Over the last decade, healthcare expenditure has increased by 3.4% per year in real terms, equating to 14.3% of total Australian Governmental spending.35 These increases have been observed across multiple components of healthcare delivery, including primary care, pharmaceuticals and hospital-based care, which has increased by 3.6% per year over the last decade (adjusted for inflation). Interestingly, the current study has shown increasing average costs per episode of care, although with stable total healthcare expenditure. These findings are likely attributable to an increase in advanced and complex therapies offered to patients during their acute hospital stay, with a concomitant overall reduction in the annual incidence of shock treated in the prehospital environment in our study setting.10 36 These findings highlight the need to adopt public health measures which continue to reduce the incidence of the underlying pathologies which lead to the development of shock and also seek to rationally administer cost-effective therapies.
Lost productivity arising from premature death in patients with shock is significant. However, the implementation of low-cost interventions can potentially improve mortality outcomes, thereby reducing the economic impacts of this condition. In the setting of cardiogenic shock, it has been shown that ambulance transport of patients to increasingly sophisticated hospitals reduces the probability of all-cause 30-day mortality.37 In this population-wide study, the adjusted probability of 30-day mortality was 34.6% among patients with cardiogenic shock transported by EMS directly to level 1 trauma centres, compared with 44.9% for those received by metropolitan hospitals with invasive cardiology services, but no on-site cardiac surgical capacity. The establishment of robust, regionalised systems of care that focus on EMS prioritising transport of patients with suspected cardiogenic shock to large, multidisciplinary centres may represent a meaningful way to improve survival outcomes, and therefore, attenuate the economic impacts arising secondary to premature mortality. When applied to our cohort, a 10% reduction in the incidence of 30-day mortality would equate to approximately $A150 million dollars in wage-related productivity gains. In addition to the potential improved outcomes through EMS transfer to larger hospitals, this study has shown that transfer to centres that can provide definitive therapy, negating the need for interhospital transfers, was associated with a $A12 766 (95% CI 10 702 to 14830) cost saving. Furthermore, for patients with sepsis, the introduction of standardised bundles of care within hospital has been shown to have meaningful impacts on clinical outcomes. In hospitals with high compliance with Surviving Sepsis Campaign performance bundles, there was a significant difference in rates of in-hospital mortality, compared with low compliance sites (29.0% vs 38.6%, p<0.001).38 Additionally, for every 10% increase in bundle compliance, there was a 4% decrease in ICU length of stay, potentially leading to significant reductions in healthcare costs. The implementation of regionalised systems of care, in addition to standardising hospital management of shock, may, therefore, have the capacity to reduce lost productivity and the direct healthcare costs arising from shock management.
The current findings should be interpreted with acknowledgement of several limitations. First, this cohort has been drawn from a population for which universal healthcare is the standard of care. While this represents the care delivered in the state of Victoria, it may not be generalisable to other states in Australia or international jurisdictions. Second, these data were derived from patients with prehospital shock treated by EMS. Therefore, individuals who self-present to hospital or develop shock in-hospital subsequent to their EMS care were not included. The final estimated direct healthcare-related costs presented are, therefore, likely to be a conservative representation of the financial impact of this condition. Finally, as indicated in our methodology, a substantial number of patients were excluded from analysis for a range of reasons including transport to private hospitals or interhospital transfers. The reported economic costs reported in this study are, therefore, by definition, a very conservative estimate of real-world costs.
Conclusion
The direct healthcare and indirect costs incurred due to lost productivity arising from the management of EMS treated non-traumatic shock are high. Improving both EMS and in-hospital systems of care and targeted public health interventions is likely to yield significant cost savings.
Data availability statement
Data are available on reasonable request. The data that support the findings of this study are available from the corresponding author, JEB, on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and ethics approval for the data linkage, in addition to this specific analysis, was granted by the Monash University Human Research Ethics Committee (approval number 11681). This study was conducted in accordance with the principles stated in the Declaration of Helsinki.
Acknowledgments
The authors would like to thank Ambulance Victoria’s paramedics and acknowledge the Victorian Department of Health as the source of VAED and VEMD data for this study, the Victorian Department of Justice and Community Safety as the source of Victorian Death Index data and the Centre for Victorian Data Linkage (Victorian Department of Health) for the provision of data linkage.
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
ZN and DS are joint senior authors.
X @pretzeldr
Contributors Concept and design of study: JEB, CM, DS and ZN; acquisition of data: JEB, EN, SC, ZN and DS; analysis of data: JEB, EN, SC and CM; drafting of the manuscript: JEB, DS, ZN and EN; revision of the manuscript: JEB, EN, EDP, LD, AJN, AE, AV, DA, SB, AB, AAU, DP, SC, WC, CM, DK, ZN and DS; approval of the final manuscript: JEB, EN, EDP, LD, AJN, AE, AV, DA, SB, AB, AAU, DP, SC, WC, CM, DK, ZN and DS. JEB and DS accepted full responsibility for the finished work and conduct of the study, had access to the data and controlled the decision to publish. JEB and DS acted as guarantors.
Funding DS is supported by National Heart Foundation (NHF) and National Health and Medical Research Council (NHMRC) Investigator Grant. JB and LD are supported by NHF and NHMRC Post Graduate Scholarships. ZN is supported by an NHF Fellowship. EDP is supported by a University of Melbourne Senior Research Fellowship. EN is supported by an NHMRC postgraduate scholarship. DK is supported by an NHMRC Investigator Grant.
Disclaimer The funding bodies had no role in the data collection, interpretation or production of this manuscript.
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.