Article Text

Download PDFPDF

Original research
Who dies from venous thromboembolism after hospitalisation for other reasons in England?: a national retrospective cohort study
  1. Frances Healey1,
  2. Janine Gower1,
  3. Lara Roberts2,
  4. Roopen Arya2,
  5. Matthew Beresford1,
  6. Aidan Fowler1,
  7. Graeme Kirkpatrick1,
  8. Ethel Oldfield1,
  9. Rachel Weaver1
  1. 1National Patient Safety Team, NHS England, London, UK
  2. 2King's Thrombosis Centre, King's College Hospital NHS Foundation Trust, London, UK
  1. Correspondence to Dr Frances Healey; frances.healey{at}nhs.net

Abstract

Objectives Venous thromboembolism (VTE) is a major cause of morbidity and mortality globally, with hospital-associated thrombosis (HAT) accounting for at least half of VTE. We set out to understand more about deaths from HAT in England, to focus improvement efforts where they are needed most.

Design A retrospective cohort combining death certification and hospital activity data to identify people with an inpatient or day case hospitalisation where no VTE diagnosis was recorded, and who died from VTE in a hospital or within 90 days of discharge, between April 2017 and March 2020.

Setting All deaths occurring in England and all National Health Service-funded hospital care in England.

Participants After 0.1% of cases were excluded due to duplicate but conflicting records, a cohort of 13 995 deaths remained; 54% were women, and 26% were aged under 70 years.

Outcome measures Analysis of age, gender, primary diagnosis, type of admission, specialties and (for day cases) procedure types were preplanned.

Results Only 5% of these deaths followed planned inpatient admissions. Day case admissions preceded 7% of VTE deaths. Emergency inpatient admissions, medical specialties and infection-related primary diagnoses predominated in people who died from VTE after hospitalisation where no VTE diagnosis was recorded. Most deaths occurred in a hospital or within 30 days of discharge.

Conclusions International efforts to reduce HAT historically focused on planned inpatient admissions. Further initiatives and research to prevent deaths from VTE after hospitalisation should focus on the emergency care pathway where most deaths occurred, with people undergoing day case procedures an important additional focus.

  • Thromboembolism
  • Safety
  • Risk Factors

Data availability statement

Data may be obtained from a third party and are not publicly available. Access to the data used in this study is through role-based access granted by Secondary Uses Service (SUS) - NHS England Digital.

http://creativecommons.org/licenses/by-nc/4.0/

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

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Strengths and limitations of this study

  • The study is the first comprehensive description on a national scale of the healthcare pathway, demographics and diagnoses of individuals who died from venous thromboembolism (VTE) after hospitalisation for other reasons.

  • Strengths of the study include the comprehensiveness of the data sets used, which represent all certified deaths from VTE and all National Health Service-funded hospital activity in England.

  • A limitation was that the data set did not consistently include people whose inpatient care was within mental health units.

  • The data sets used have been previously assessed as having high levels of the capture of VTE diagnosis during hospitalisation, although in a country where autopsies are rare, levels of VTE missed in death certification are difficult to ascertain.

  • We focused on death from VTE after a hospital admission without any VTE diagnosis; this is not an exact equivalent to hospital-associated thrombosis but is the closest equivalent possible without extensive clinical review of locally held case records.

Introduction

Venous thromboembolism (VTE), including deep vein thrombosis and pulmonary embolism (PE) is a major cause of morbidity and mortality internationally.1 2 It impacts all ages, genders and ethnicities, although not equally1

Hospital-acquired venous thromboembolism—also known as hospital-associated thrombosis (HAT)—is defined as VTE occurring in a hospital or within the subsequent 90 days/3 months1 2 and is thought to account for more than 50% of all VTE seen.1 2

Despite the widespread use of thromboprophylaxis, supported by a range of quality improvement initiatives, and clinical guidelines, deaths from VTE continue to be recorded in substantial numbers in the USA2 and in England.3 In the USA, this has resulted in the ‘Call to Action to prevent VTE in Hospitalized Patients’ from the American Heart Association.2 In England, this has led to the National Patient Safety Committee4 undertaking a stocktake of existing initiatives and resources to review and refresh the national support offer on HAT. The analysis presented in this paper was undertaken to support the Committee’s work.

Objectives and methods

This study aimed to provide information on the demographics, diagnoses and hospital pathway of people who died of VTE after hospitalisation for other reasons, to help target national improvement strategies and research.

This study used death certification data held by the Office for National Statistics (ONS) linked at the individual pseudonymised patient level to hospital records in the Secondary Uses Service (SUS) database. SUS data is considered to have high levels of coverage and accuracy in the physical healthcare sector for National Health Service (NHS)-funded care in England5 but does not fully encompass admissions to mental health services, where the main data source is the Mental Health Services Dataset. SUS captures a range of demographic, diagnostic and major procedure-related information for hospital admissions5 but does not capture prescribed medication or minor procedures. If a clinical diagnosis of VTE has been reached during a hospital admission, even without confirmatory test results, this is recorded in the SUS data set either as a primary diagnosis or among multiple secondary diagnoses.

All deaths attributed to VTE are captured in mortality data held by the ONS, which we linked to the SUS data set using pseudoanonymisation at an individual patient level. An established annual national publication of VTE deaths provided the technical specification6 for our study. Our cohort encompassed deaths in part one of the death certificate that occurred in the hospital or within 90 days of the last day of an inpatient or day case admission after combining consultant episodes into single admission spells, where the admission did not record any VTE diagnosis (whether primary or secondary diagnosis). This study encompassed patients who died between 1 April 2017 and 31 March 2020, including 2016 data for any hospital admission preceding the death. To reduce analytical complexity given the date of admission preceding the date of death, a threshold of 19 years or over at death was adopted.

For this defined study population, we analysed demographics, admission type and specialty, all captured within SUS. We also examined the time interval between the discharge date closest to death and the date of death. We analysed the primary diagnosis field for the admission closest to death, which is coded using the International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10).7 We used a combination of primary diagnosis codes, specialty codes and procedural codes to identify VTE-related deaths in subgroups of patients specified in National Institute for Health and Care Excellence (NICE) guidance on reducing the risk of HAT2 (eg, admissions with stroke were identified by primary diagnosis codes, obstetrical admissions were identified by specialty codes and hip and knee replacement by procedural codes). We used the set of ICD-10 codes established by the previous national publications to define VTE.6 These were:

  • I26.0 Pulmonary embolism with mention of corpulmonale.

  • I26.9 Pulmonary embolism without mention of corpulmonale.

  • I80.1 Phlebitis and thrombophlebitis of femoral vein.

  • I80.2 Phlebitis/thrombophlebitis of femoral vein of other deep vessels of lower extremities.

  • O22.3 Deep thrombophlebitis in pregnancy.

  • O88.2 Obstetrical blood-clot embolism—covers obstetrical/puerperal (pulmonary) embolism NOS.

Patient and public involvement

Patient/public representatives were involved in the broad design of this research, gave feedback on the initial presentation of findings and advised on dissemination through their membership of the National Patient Safety Board, whose improvement work this study was undertaken to inform.

Findings

Data completeness

20 individuals with duplicate but conflicting records were excluded from the analysis. Our remaining study population comprised 13 995 unique individuals who died between 01 April 2017 and 31 March 2020 with VTE in part one of the death certificate, where the death occurred in a hospital or within 90 days of the last day of an inpatient or day case admission, and where the admission closest to death did not record any VTE diagnosis.

Patient demographics, admission routes, specialties and timing of death

These findings are presented in table 1. Most deaths (86%) in the study cohort occurred in people admitted as an emergency, with only 5% of deaths occurring in people with planned inpatient admission, although these types of admissions account for around half of hospital admissions. Approximately 7% of deaths occurred in people whose last admission was a day case. While the greatest proportion of deaths was recorded in the group aged 80–89 years at death, deaths occurred in every age group and 26% of the deaths were people aged under 70 years. We found that 54% of the cohort were women, which is a similar proportion to hospital admissions overall.

Table 1

Patient demographics and admission specialties

General medical services and the core medical subspecialty services of elderly medicine, oncology, respiratory medicine and gastroenterology predominated (72% when combined). Trauma and orthopaedic services had treated 7% of the people who died, general surgical services had treated 5% and urology services had treated 2%. Other services (including smaller medical and surgical subspecialty) in combination treated 15% of the people who died.

Around half of these deaths (48%) occurred before leaving the hospital with the remaining deaths showing a reducing frequency with time elapsed; 30% of deaths occurred between 1 and 30 days after leaving the hospital, 11% between 31 and 60 days and 6% between 61 and 90 days. We found that 5% of the deaths had dates of death that preceded the date of the end of their hospital admission, usually by a single day; these were likely to be deaths that occurred in hospitals where bed management systems did not record an episode as ended until after last offices are completed, or where human error affected date entry.

Primary diagnoses (subgroups and types)

The people who died had a wide range of primary diagnosis subgroups at their last hospital admission before death. Infections, including ‘influenza and pneumonia’ and ‘other bacterial diseases’, were the most common diagnostic subgroups, although there was a wide range of other primary diagnoses subgroups (figure 1).

Figure 1

Deaths from VTE in a hospital or within 90 days of the last day of an inpatient or day case admission, where the admission closest to death did not record any VTE diagnosis, April 2017 to March 2020, by top 10 ICD-10 primary diagnosis SUBGROUPS. ICD-10, International Statistical Classification of Diseases and Related Health Problems 10th Revision; VTE, venous thromboembolism.

Specific ICD10 primary diagnosis types were recorded for 12 035 people, with lobar pneumonia, sepsis and unspecified pneumonia being the most common (figure 2).

Figure 2

Deaths from VTE in a hospital or within 90 days of the last day of an inpatient or day case admission, where the admission closest to death did not record any VTE diagnosis, April 2017 to March 2020, by top 10 primary diagnosis TYPE (percentages of 12 035 deaths where this was recorded). VTE, venous thromboembolism.

Day case procedures

We examined the type of procedures undertaken for the subgroup of 938 patients whose admission closest to death was a day case and where procedure type was recorded. After grouping clear subsets of procedures (eg, grouping treatment bands of chemotherapy) the most frequent procedure types are shown in table 2 and included chemotherapy (20.4%), upper gastrointestinal tract endoscopy (10.4%), prosthetic lens insertion (7.2%), lower gastrointestinal tract endoscopy (6.8%) and transfusion (5.8%). A wide range of other procedures together accounted for 32.7% of the day case procedures; 154 different procedures preceded 307 of the deaths.

Table 2

Deaths from VTE in a hospital or within 90 days of a day case admission, where the admission closest to death did not record any VTE diagnosis, April 2017 to March 2020, by top 10 procedure groups (percentages of 938 deaths where procedure type was recorded)

HAT prevention intervention groups specified in national guidance

Reasonable matches in the SUS data set were found for some patient groups or subgroups specified in English national guidance for HAT prevention.2 Within the overall sample of 13 995 deaths in a 3-year period, we found:

  • 673 VTE deaths after admission with a primary diagnosis of cancer.

  • 93 VTE deaths after admission with a primary diagnosis of stroke.

  • 49 VTE deaths after admission with a primary diagnosis of acute coronary syndrome.

  • 45 VTE deaths where hip or knee replacement surgery had taken place during the admission.

  • 6 VTE deaths where spinal surgery had taken place during the admission.

  • No VTE deaths after obstetrical or miscarriage inpatient admission

Discussion

To our knowledge, this study provides the first comprehensive description of individuals who died from VTE after hospitalisation for other causes that uses routinely collected hospital data on a national scale. Its findings are intended as a ‘pareto rule’ for quality improvement planning and research prioritisation, to highlight the services, admission types, patient demographics, primary diagnoses and day case procedures where most of these deaths occur.

Our first important finding is the dominance of emergency admissions and medical specialties as the major future improvement challenge area for the prevention of HAT; 86% of the deaths in our cohort were in people admitted as emergencies. The core medical specialties were over-represented compared with hospital admission activity, and they represented almost three-quarters of the deaths, even before the addition of smaller medical subspecialties. Guidance in England and the USA recommends thromboprophylaxis for medical patients considered at high risk of VTE1 2 but notes that while studies have identified the risk factors that can predict VTE in medical patients, there is much more limited evidence on whether, even in the highest risk categories of medical patients, thromboprophylaxis reduces the occurrence of symptomatic VTE.2 8 Further, studies of thromboprophylaxis are typically underpowered to identify any significant impact on VTE-related deaths.9 10 Our findings underline the need for future studies exploring if and how VTE risk in medical patients can be mitigated, including studies that are powered to detect reductions in deaths.

Although planned inpatient admissions make up more than half of NHS admissions, only 5% of deaths from VTE in our cohort occurred in this group. In elective surgical patients, where the prevention journey began, there are now several established strategies that reduce VTE risk, including anticoagulant and/or mechanical prophylaxis; minimally invasive surgery; low-risk anaesthetic techniques; and interventions to promote early mobilisation, including effective pain relief, rehabilitation and early discharge.1 2

We found only ~15 deaths from VTE per year after an admission with a hip or knee replacement procedure, in a period when around 170 000 hip and knee replacements took place annually.11 Trauma and orthopaedics as a specialty were however over-represented compared with general hospital admission activity in the group of people who died from VTE. This appeared likely to derive more from emergency trauma admissions than planned orthopaedic procedures, as fractured neck of femur and ‘tendency to falling’ appeared in the top 10 primary diagnoses in our cohort. This suggests potential for further gains by combining falls prevention, frailty and VTE prevention initiatives, especially in relation to improving mobility and addressing underlying cardiovascular causes of falls.12 13

Another important finding was that almost half of the individuals in our cohort died from VTE before leaving the hospital. This finding is consistent with similar proportions in local case record audit14; the audit suggested that confidence to record a VTE diagnosis at death, but not previously, arose either because a fatal PE was the first obvious manifestation, or because a suspicion of VTE in life was confirmed by clinical presentation at the point of death. These findings suggest that alongside thromboprophylaxis, a quality improvement focus on timely diagnosis and treatment of VTE is needed to prevent deaths.

We found adults of all ages were affected, with 26% of deaths in people aged under 70 years, confirming that HAT is important not only in terms of the numbers of lives lost, but years of life lost. Although 45% of deaths were in people aged over 80 years, previous broad case record reviews suggest preventable deaths due to problems in care in older people often take months or years away from the time people would otherwise be expected to live.15 Although we did not have access to data on ethnicity, existing studies confirm differences in the lifetime risk of VTE, with black people at the highest risk.16–18 England, 12% of hospital admissions overall are people identifying as black and ethnic minorities,19 and therefore will be an important focus in any future improvement efforts.

Our analysis of primary diagnosis in people who died of VTE used ICD-10 coding, and because the highest level of ICD-10 grouping is by body system,7 we did not have the capacity to identify the overall proportion of primary diagnoses that might be considered infection-related within the many groups and types of diagnosis related to infection in different body systems. We found that infection-related diagnostic groups of ‘influenza and pneumonia’ and ‘other bacterial infections’ were the most common diagnostic groups in people who died from VTE, with lobar pneumonia and unspecified sepsis the most common primary diagnosis types, and with many other categories of infectious illness outside these groups also apparent. Although it is plausible that some of the patients recorded as admissions for pneumonia were showing symptoms of VTE misdiagnosed as these causes,20 infectious illness would appear to be a major area for future VTE prevention efforts, including consideration of implications for patients with infections who might once have been hospitalised but are now managed via ‘virtual wards’ or ‘hospital at home’ care.21

We found that ~225 people per year who died from VTE after admissions without any VTE diagnosis had a primary diagnosis of cancer recorded in the last hospital admission, and many more people who died were likely to have cancer diagnoses recorded among multiple secondary diagnoses. Several oncology guidelines now support VTE risk assessment and the provision of ambulatory prophylaxis for those at high risk of VTE.22 This practice has not yet been widely implemented in England, with NICE suggesting prophylaxis for ambulatory patients with pancreatic cancer only.23

We found that 7% of the deaths occurred in people whose last admission was a planned or booked day case; this exceeded the number and proportion of deaths from VTE in people with planned or booked inpatient admissions and might therefore also be an important focus area for future improvement support. The types of day case procedures preceding death from VTE represented a wide range. The relatively high frequency of chemotherapy, gastroscopy, lens insertion, colonoscopy and blood transfusion we found among people who died of VTE may simply reflect the volume of these procedures24 and the underlying vulnerability to VTE associated with age and comorbid disease in those people most likely to need those procedures. However, they may also be consistent with our understanding of the triad of VTE causation; gastroscopy, colonoscopy and transfusion are often undertaken in response to suspected or confirmed internal haemorrhage, and chemotherapy is known to have physiological impacts on the circulatory system over and above the tendency to VTE caused by cancer.22 The type of anaesthesia is often taken as a proxy for impact on mobility, but procedures such as cataract surgery can have a substantial short-term impact on mobility despite usually being performed under local anaesthetic.25

Our findings took place in a context where a range of national initiatives and incentives focused on reducing HAT had been in place for more than a decade, including national VTE prevention guidelines2 26; a National Venous Thromboembolism Prevention Programme27; VTE exemplar centres28; and financial and contractual incentives for VTE risk assessment.29 These initiatives resulted in sustained levels of 90%–95% of NHS-funded patients having VTE risk assessments between 2011 and 201930 and widespread adoption of thromboprophylaxis; in 2018 around 70 million doses of low molecular weight heparin in the preparations typically used for thromboprophylaxis were supplied to NHS hospitals,31 and almost four million pairs of anti-embolism stockings were purchased via the NHS’ central supply chain.32 Audits of acute hospital services in England in 2019/2020 identified high but imperfect levels of compliance with national VTE prevention guidelines, with 88% of sampled admissions receiving the recommended thromboprophylaxis, although 8% experienced missed doses of anticoagulant prophylaxis, with gaps in provision of patient information also observed.33 The generalisability of our study’s findings to other countries would be dependent on a similar context of interventions to reduce the risk of VTE, as well as comparable populations; the total population of England at the time of our study was around 56 million, of whom 12 million were aged over 65 years.34

Strengths of the study include the comprehensiveness of the data sets used, which represent all certified deaths and all NHS-funded hospital activity in England. As only 0.1% of the sample had to be excluded due to conflicting data, there was a low risk of bias. The data sets have previously been assessed for generally good quality and completeness of the data fields.5 A previous multicentre study comparing the identification of HAT diagnosis from clinical case record review to SUS data found a false-negative rate of 1% for VTE diagnosis within SUS,35 indicating our cohort of cases without a VTE diagnosis recorded during hospitalisation is reasonably complete. HAT-related deaths may have been missed because they remained undiagnosed even after death, particularly given the low rate of autopsies in England where they are conducted for less than 1% of hospital deaths.36 The rigour applied by the medical examiner system, where deaths are reviewed by a senior independent doctor, should reduce missed cases in hospital deaths,37 but during the study period the medical examiner system had not been extended to routinely encompass deaths that occurred outside hospitals.

Our case definition replicated the approach used in established national statistics6 but has some limitations in terms of reliably identifying all patients dying from HAT. The population used in this study, of people who died from VTE in a hospital or within 90 days of a hospitalisation where no VTE diagnosis was recorded, is a group that would primarily be expected to consist of people with HAT, alongside some patients whose VTE occurred in the community but was not diagnosed before they died from VTE. Our cohort did not include patients where HAT occurred and was diagnosed before they left that hospital, as the national data sets we used do not separate these from patients whose community-acquired VTE required hospital treatment. Clinical review of locally held case records may identify HAT more comprehensively, but many VTE occurrences, including VTE deaths, are not clinically reviewed.33 The limitations our study experienced in isolating HAT are not unique but reflect the limitations of routinely collected data and the complexities of HAT attribution.

Although our study was the first on this scale focused on fatal VTE after hospitalisation, other studies allow some comparisons. Our findings for inpatients and day cases are consistent with a case record audit of over 4000 occurrences of HAT (of which 2.9% were fatal) whose data collection period overlapped with this study; the audit found that most HAT occurred in medical patients (58%), and numerically HAT was most frequently associated with admissions for pneumonia, malignancy, hip fracture surgery, falls and non-bariatric gastrointestinal surgery.33 A study of medical patients with HAT within the USA (predominantly non-fatal) found HAT a significant issue within this population, impacting 1.2% of admissions, and found active cancer but not acute infection as a significant risk factor in multivariate analysis.38 In the UK, there were seven VTE deaths in pregnancy or within a year of delivery between 2017 and 201939 but these would not be expected to be captured in our analysis if they occurred in Scotland, Wales or Northern Ireland, or occurred antenatally, after home births, or more than 90 days after delivery; the zero figure we found is therefore not inconsistent with this source.

Our data set did not encompass information on who had received thromboprophylaxis, but a contemporaneous case record audit33 found that most patients with HAT had received optimal thromboprophylaxis, with 13% of HAT considered potentially preventable. A separate review of HAT reported as incidents found that local investigations typically described complex patients for whom expert advice or specially tailored VTE prophylaxis was required, rather than ‘standardised’ VTE prophylaxis.40

Research needs related to the prevention of HAT have previously been listed in national guidance.2 In addition to the key need for additional research on effective prevention strategies for medical patients admitted as emergencies that our study underlines, research needs include a better understanding of the ethnicity and social deprivation of people who die from VTE, and analysis of who dies from VTE without prior hospital admission. The data sets we used did not provide comprehensive information on people whose inpatient care was within mental health units, and further research is needed to identify who dies from VTE after a mental health unit admission. This study focused on death from VTE, but there is the potential to use SUS to understand who is experiencing the greatest morbidity from VTE, and the subsequent cost to the NHS and wider society.

Conclusion

The ~4700 deaths per year from VTE after hospitalisation without any VTE diagnosis are no longer an abstract number, as we have begun to build a picture of their age, gender, their hospital pathway and the diagnoses that made them vulnerable to fatal VTE. The findings suggest we need to build on the achievements of earlier years of improvement efforts for people with planned surgical admissions, with the major future improvement and research focus on emergency inpatient admissions and medical specialties, with day case admissions an important additional focus.

While this study could not identify who among the people who died received evidence-based preventive care, other contemporaneous studies suggest most but not all people with HAT had received the NICE-recommended thromboprophylaxis, with failure to provide optimal thromboprophylaxis especially likely to affect complex patients for whom standardised thromboprophylaxis might not be appropriate. Reducing deaths from HAT is therefore both a safety and an effectiveness challenge. Our findings suggest we need to build on past successes in elective surgery, where HAT-related death is now rare, and pivot future improvement and research work to patients admitted as emergencies, under medical specialties, and in patients undergoing day case procedures. Future patient safety initiatives should aim to improve compliance with existing guidance and timely diagnosis, while future research should explore new prevention strategies in patients at high risk of VTE for whom existing prophylaxis strategies prove inadequate.

The deaths we studied are only one-third of overall deaths from VTE in England, and patient safety efforts in reliably delivering the existing evidence base for HAT prevention will need to be complemented by improvements in the evidence base for effective prevention, timely diagnosis and treatment of VTE in community settings, incorporating healthcare and public health initiatives.

Data availability statement

Data may be obtained from a third party and are not publicly available. Access to the data used in this study is through role-based access granted by Secondary Uses Service (SUS) - NHS England Digital.

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.

Acknowledgments

Grateful acknowledgements to Julia Abernethy, Chris Mainey, Hassan Ismail, Sarah Jennings and Clare O’Brien for advice and suggestions related to specific areas of analysis.

References

Footnotes

  • X @FrancesHealey

  • Contributors FH is the guarantor and accepts full responsibility for the conduct of the study, had access to the data, and controlled the decision to publish. FH developed the original concept for a deep dive into Secondary Uses Service/Office for National Statistics data, designed the key areas for analysis to encompass, provided expert input into adaptations of the analytical approaches to increase clinical and patient safety relevance and drafted the paper. JG planned and conducted the analysis, set analytical quality thresholds and adapted the analytical approach and presentation of findings and figures in response to initial findings and clinical feedback. FH, JG, LR, RA, MB, AF, GK, EO and RW made substantial contributions to the interpretation of data, revising the paper critically for important intellectual content, gave final approval of the version to be published and agreed to be accountable for all aspects of the work.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests RA has received speaker fees from Cardinal Health and Sanofi. The other authors have no interests to declare

  • Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

  • Provenance and peer review Not commissioned; externally peer reviewed.