Article Text

Original research
The economic burden associated with unmet surgical needs in Liberia: a retrospective macroeconomic analysis based on a nationwide enumeration of surgical procedures
  1. Håvard A Adde1,2,3,
  2. Mark D Oghogho3,
  3. Alex J van Duinen1,3,4,5,
  4. Caris E Grimes6,7,
  5. Theophilus C Hampaye8,
  6. Matthew Quaife9,10,
  7. Håkon A Bolkan1,3,4
  1. 1Department of Public Health and Nursing, Faculty of Medicine and Health Sciences, NTNU - Norwegian University of Science and Technology, Trondheim, Norway
  2. 2Department of Surgery, Ålesund Hospital, Møre and Romsdal Hospital Trust, Ålesund, Norway
  3. 3CapaCare, Trondheim, Norway, & Tappita, Liberia
  4. 4Department of Surgery, St Olav’s Hospital, Trondheim University Hospital, Trondheim, Norway
  5. 5Department of Surgery, ELWA Hospital, Monrovia, Liberia
  6. 6King's Centre for Global Health and Health Partnerships, King's College, London, UK
  7. 7Medway NHS Foundation Trust, Gillingham, UK
  8. 8Physician Assistant Association, Monrovia, Liberia
  9. 9Department of Global Health and Development, London School of Hygiene and Tropical Medicine, London, UK
  10. 10Evidera Ltd, London, UK
  1. Correspondence to Dr Håvard A Adde; haavarad{at}


Objectives The economic consequences of untreated surgical disease are potentially large. The aim of this study was to estimate the economic burden associated with unmet surgical needs in Liberia.

Design A nationwide enumeration of surgical procedures and providers was conducted in Liberia in 2018. We estimated the number of disability-adjusted life years (DALYs) saved by operative activities and converted these into economic losses averted using gross national income per capita and value of a statistical life (VSL) approaches. The total, the met and the unmet needs for surgery were determined, and economic losses caused by unmet surgical needs were estimated. Finally, we valued the economic losses avoided by various surgical provider groups.

Results A total of 55 890 DALYs were averted by surgical activities in 2018; these activities prevented an economic loss of between US$35 and US$141 million. About half of these values were generated by the non-specialist physician workforce. Furthermore, a non-specialist physician working a full-time position for 1 year prevented an economic loss of US$717 069 using the VSL approach, while a specialist resident and a certified specialist saved US$726 606 and US$698 877, respectively. The burden of unmet surgical need was associated with productivity losses of between US$388 million and US$1.6 billion; these losses equate to 11% and 46% of the annual gross domestic product for Liberia.

Conclusion The economic burden of untreated surgical disease is large in Liberia. There is a need to strengthen the surgical system to reduce ongoing economic losses; a framework where specialist and non-specialist physicians collaborate may result in better economic return than a narrower focus on training specialists alone.

  • Health economics
  • Health Equity

Data availability statement

Data are available upon 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:

Statistics from

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.


  • This study is based on a comprehensive nationwide enumeration of operative numbers and surgical providers in Liberia; it includes surgical facilities from all parts of the country, at all levels of care and from both the governmental and the private sector.

  • Determining the burden of unmet surgical need in a population is complex and the present analysis is based on best estimates.

  • The classification of surgical procedures as either death or disability averting is an oversimplification of reality.

  • Most surgeries had an unknown status of urgency, and these procedures were all weighted as planned operations, pushing the estimate in a conservative direction.


Untreated surgical disease causes profound economic losses.1 Low-income and middle-income countries are projected to lose a cumulative US$12.3 trillion if surgical services are not scaled up between 2015 and 2030.2 Investing in surgery is highly cost-effective, with an impact comparable with other widely recognised health interventions.3 4 Failing to acknowledge the role of surgery as part of any well-functioning healthcare system will be expensive, not only in terms of human lives and suffering, but also in the form of impaired economic development.2

The Global Burden of Disease Study uses disability-adjusted life years (DALYs) to enumerate the extent of diseases and injuries worldwide5; the DALY is a composite measure combining both mortality and morbidity into one metric. The DALY methodology provides an estimate of disease burden, and it allows further quantification of economic gains and losses caused by treated and untreated surgical disease.6 7 Although a growing number of studies are urging stakeholders to invest in individual surgical procedures,8–10 and surgical delivery platforms such as the first level hospital,11 few reports on economic implications exist at national levels. A study from Sierra Leone estimated that the national economy suffered losses between US$1.1 and US$3.8 billion annually due to unmet surgical needs.12 Such data, highlighting the developmental implications of untreated surgical disease, can support policymakers who are reaching for economic development and prosperity.

Liberia is a small West-African country with a violent history of civil war and the devastating Ebola virus outbreak.13 It is ranked 178 out of 191 nations on the UN Human Development Index and has a low-income economy.14 The country is heavily depending on foreign aid and domestic challenges such as electricity constraints, a need for skills, high import dependency and an enclave growth sector is affecting the economy.15 In 2021, the gross domestic product (GDP) per capita was US$676,16 which is small compared with the world average of US$12 237.17 In the health sector, the financial constraints are demonstrated by the fact that 41% of the governmental health workers were not on the payroll in 2015.13 The surgical system is undersized with a density of 1.6 specialist surgical providers per 100 000 population18 and a surgical volume of 462 surgical procedures per 100 000 population,19 resulting in a high unmet need for surgery. Long travel times delay surgical treatment, and many patients lack coverage of essential surgical care.20 It is unknown how the limited coverage of surgery affects the national economy. The aim of this study was to estimate the burden of premature death and disability averted by surgical activities in Liberia and value the economic costs associated with unmet surgical needs.


Data collection

A nationwide survey of surgical facility activities and resources was conducted in Liberia in 2018. All healthcare facilities performing surgical procedures requiring general, regional or local anaesthesia within an operating theatre the year prior to the data collection were eligible for inclusion. All data were collected through on-site visits. Operative logs were reviewed to describe current surgical activities; a predefined 4-month sample of the operative case-mix was extracted and scaled up to represent a complete year. Furthermore, surgical workforce data were collected using the operative logs and structured interviews with clinical and administrative staff. Further details of the data collection are available elsewhere.19


A surgical provider was defined as any provider listed as the main operator in the 4-month logbook sample. A full-time equivalent (FTE) position was quantified for each surgical provider to reflect the length and size of their working position over the study period; a half-time position over 1 year and a full-time position over half a year both equalled 0.5 FTE.18

Total surgical need is the sum of ongoing surgical activity (met surgical need) and surgical procedures that are needed but not performed (unmet surgical need). The Lancet Commission on Global Surgery estimated that all nations should aim to perform at least 5000 operations per 100 000 population annually to cover population needs.2 This estimate has been used by previous studies to define the total need for surgery in a population and from there derive the unmet surgical need.12 21 We therefore defined a total need for surgery of 5000 operations per 100 000 population in Liberia, and the unmet surgical need was calculated by subtracting the met need from the total need.

Disability-adjusted life years

DALYs can be divided into (1) years of life lost (YLL) which describes the burden of premature death and (2) years lost due to disability (YLD) describing the burden of living with a disability. Previous studies have explained the calculation of DALYs for surgical conditions.8 9 Grimes et al12 estimated the economic consequences of met and unmet surgical needs in Sierra Leone, which is a neighbouring country to Liberia, and we applied the same methodology in the present study. Surgical procedures were either deemed death averting or disability averting. The number of YLLs saved from death averting procedures was estimated based on the number of deaths from a condition and the life expectancy at the age of death. Death averting procedures were weighted based on the risk of death and treatment efficiency, and the number of deaths averted was estimated adjusting for these weights. We applied the same weights as described by Grimes et al12 (online supplemental table 1). Emergency operations typically involve a more acutely ill patient compared with planned operations, and these procedures were therefore weighted with a higher risk of death. In cases where the status of urgency was unknown, the procedure was weighted as a planned operation to be conservative. The number of YLDs saved from disability averting procedures was estimated based on the incidence, the duration of illness and the disability weight. Disability weights were extracted from the Global Burden of Disease Study.22 We did not apply age weighting to the DALY calculations, which is consistent with the recent Global Burden of Disease studies.

All YLLs and YLDs were summarised to represent DALYs averted by surgical activities (met surgical need). The analysed data represented a met surgical need of 8.2% and an unmet need of 91.8%. A simple scaling factor was used to scale up the DALYs to represent 80%, 50% and 20% unmet surgical need, and economic losses averted were calculated for each of these scenarios.

Translating DALYs to dollars

Translating DALYs into economic values can provide a meaningful metric of economic burden. We used two different approaches to estimate economic benefits and losses caused by met and unmet surgical need. The first method, known as the human capital approach, was part of the WHO’s Commission on Macroeconomics and Health.23 The basic assumption is that the values of individuals to society is determined by their contributions to the economy. This methodology is likely to underestimate economic losses due to disease because it does not account for unpaid labour, such as lost productivity when a healthy individual needs to care for a sick family member.23 The human capital approach uses gross national income (GNI) per capita to convert DALYs to dollar values. The World Bank uses both the Atlas method and the purchasing power parity (PPP) method to calculate GNI per capita. The Atlas conversion factor reduces the impact of exchange rate fluctuations in the cross-country comparison of national incomes, while the PPP method eliminates the differences and changes in relative price levels. We used both the Atlas method and the PPP method to calculate GNI per capita using 2018 US$16 and simply multiplied GNI per capita by the number of DALYs averted to estimate the economic value of surgical activity.

The second approach, called the value of a statistical life (VSL), provides a more reliable measure from a human welfare point of view; it describes the amount an individual would be willing to pay to reduce or avert the risk of dying. Thus, it suggests that individuals’ worth can be larger than their salary. The VSL of US$39 800 as estimated by Robinson and Hammitt for Liberia24 was multiplied by a 2007–2018 inflation correction of 1.210, making a VSL of US$48 158.25 The VSL was converted into its annualised equivalent in line with the method described by Alkire et al.6 Age weighting was not applied to this value, assuming it remains constant throughout a lifetime. A discount rate of 5% was used on costs and benefits, which is in line with the previously mentioned analysis from Sierra Leone.12

Sensitivity analysis

A one-way sensitivity analysis was used to explore the effects of changes in model parameters on the estimated economic losses caused by unmet surgical needs. Various scenarios were run adjusting one model parameter at a time. The first scenario describes a change in discount rate to 3% as per guidelines. The second scenario doubles surgical volumes, while the third scenario only includes mortality DALYs in the model. The fourth scenario explores uncertainty related to GNI per capita and VSL; the model was run using Sierra Leonean values. Also, to investigate the effect of the inflation correction on the VSL, a scenario was run using the original VSL of US$39 800 as reported by Robinson and Hammitt.24 Another model parameter with the potential to influence results is the choice of disability weights for high-volume procedures. Planned hernia operations were all classified as a mild abdominopelvic problem with a disability weight of 0.012; however, many of these patients may, in fact, experience more symptoms. We therefore run a scenario where planned hernia operations were reclassified as a severe abdominopelvic problem with a disability weight of 0.326.22 Finally, the methodological choice to classify all procedures with unknown status of urgency as planned operations is likely to push the results in a conservative direction. In Sweden, a high-income country, approximately 6% of all hernia operations were emergency procedures.26 Due to a higher burden of untreated surgical disease, this proportion might be even higher in low-income countries. We explored a scenario where 15% of hernia operations were reclassified as emergency procedure.

Patient and public involvement

Patients were not involved in the development or conduct of this study. Representatives from the Ministry of Health were involved in the planning of the study and throughout the data collection. Furthermore, the study findings have been presented to relevant stakeholders in Liberia.


A total of 52 healthcare facilities were identified as surgical facilities and 51 of these were included in the data collection. There were 25 governmental, 14 private for-profit and 12 private non-profit facilities. The case-mix sample identified 6428 surgical procedures for the months of October, January, April and July. A complete breakdown of the case-mix has been presented elsewhere19; obstetrics/gynaecology surgery accounted for 54% of surgical procedures, while general surgery and orthopaedics accounted for 40% and 6% of the volume, respectively. A total of 524 (8.2%) procedures were excluded due to unknown procedure name or patient age. Of the included operations were 3590 (60.8%) categorised as death averting, while 2314 (39.2%) operations were primarily disability averting. We assumed that the 4-month sample was representative for the entire year and scaled it up to represent all 12 months. The included operations represented a met surgical need of 8.2% and an unmet need of 91.8%.

For the entire year, a total of 49 474 YLLs and 6416 YLDs were averted because of surgical activities, summarising to 55 890 DALYs. Based on this, we estimated that between US$35 (Atlas method) and US$141 million (VSL method) were saved due to surgical activities. Most DALYs were averted by obstetrics and gynaecology procedures (table 1), and non-specialist physicians contributed considerably to this category (online supplemental table 2). Large disparities were observed between geographic areas (online supplemental figure 1). The number of DALYs averted per 100 000 population can be seen in context with population access to surgery, as presented elsewhere.20 Caesarean sections, representing 43% of all operations, contributed with the highest number of DALYs averted (table 2). Furthermore, there were large differences in the number of DALYs averted per procedure for various operations. Especially, striking was the difference between elective and emergency hernia repair; emergency hernia repair saved between US$7240 and US$29 474 per procedure, while elective hernia repair saved between US$119 and US$484 per procedure.

Table 1

DALYs and economic losses averted by surgical activities

Table 2

Economic losses averted per surgical procedure performed

About half of all DALYs were averted by non-specialist physicians (table 1). A total of 286 individual surgical providers were registered in the operation theatre logbooks, and these providers worked 198 FTEs. Using the VSL method, economic losses averted per surgical provider FTE were estimated to be US$698 877 for specialists, US$726 606 for residents and US$717 069 for non-specialist physicians (figure 1). In 2018, annual costs associated with unmet surgical needs were estimated to lie between US$388 million and US$1.6 billion (figure 2); this equals between 11% and 46% of the annual GDP, respectively.

Figure 1

Economic losses averted (US$) by surgical provider full-time equivalent position. PPP, purchasing power parity; VSL, value of a statistical life.

Figure 2

Yearly economic losses (US$) associated with current unmet surgical needs and three scenarios where surgical activity is scaled up to 80%, 50% and 20% unmet need. PPP, purchasing power parity; VSL, value of a statistical life.

The one-way sensitivity analysis (figure 3) reveals that the model estimate is particularly sensitive to changes in VSL. Furthermore, the model is sensitive to the classification of high-volume procedures such as hernia operations; a reclassification of planned hernia operations as a severe abdominopelvic problem with a disability weight of 0.326 resulted in an estimate change of US$525 million.

Figure 3

One-way sensitivity analysis describing change in estimate of economic losses at current unmet needs. GNI, gross national income; PPP, purchasing power parity; VSL, value of a statistical life.


This study aimed to estimate the number of DALYs averted by surgical activities in Liberia and value the costs associated with unmet surgical needs. In 2018, surgical efforts averted an estimated 55 890 DALYs; this translated into an economic benefit of between US$35 and US$141 million. The non-specialist physician workforce contributed with half of all DALYs averted and thus prevented large economic losses. Furthermore, the economic cost averted per non-specialist physician was comparable to the costs averted per specialist. The high burden of untreated surgical disease imposes large economic losses on the national economy.

The economic benefits from investing in health systems are significant27 and surgery is no exception.2 A scale-up of surgical rates in low-income and middle-income countries is estimated to cost US$300–US$420 billion, or 4%–8% of total annual health expenditures in these countries.28 Although these numbers are high, they fade when compared with the estimated US$12.3 trillion of economic losses should surgical services be ignored.2 We estimated that the Liberian economy suffers economic losses of US$388 million to US$1.6 billion annually due to a high burden of unmet surgical need. Ongoing surgical activities are, on the other hand, preventing large productivity losses, and more than half of these losses are averted by the non-specialist physician workforce. Training specialists are crucial, as they are meant to supervise and deal with complicated cases. However, a training model focusing on specialists alone will be less effective than a model where specialists and non-specialists collaborate.2 In Liberia, there were small differences in DALYs averted per provider FTE between non-specialist physicians and specialists. Non-specialist physicians are generally less costly to train, and they have lower wages, making them a very effective part of the surgical workforce. Hence, policymakers should recognise their contribution to the national economy.

Although the literature supporting the economic benefits of surgery at the continental and global level is growing,1 6 7 21 few analyses exist at national levels. In Sierra Leone, the estimated economic loss caused by the unmet burden of surgical disease was US$3.8 billion using the VSL approach,12 whereas we estimated loss of US$1.6 billion in Liberia. Sierra Leone and Liberia are neighbouring countries and have many similarities historically and developmentally. The two countries have almost identical surgical volumes per 100 000 population.19 29 Furthermore, we replicated the methodological steps used to calculate economic losses; yet, the losses in Sierra Leone were more than twice the size of the losses in Liberia using the VSL approach. Little is known about people’s willingness to pay for health risk reduction in sub-Saharan Africa; in fact, the VSL values presented by Robinson and Hammitt24 are based on a VSL from the USA which are then extrapolated to African countries by adjusting for income differences. The logic is that people with lower wages will have less money to spend on reducing health risks. Hence, a lower GDP in Liberia, as compared with Sierra Leone, contributes to the difference in economic losses due to unmet surgical needs.

Another key factor with the potential to influence economic estimates is the composition of the surgical case-mix. Caesarean sections constituted 43% of the case-mix in Liberia compared with only 21% in Sierra Leone. Using the VSL approach, we estimated an economic return of US$9484 per procedure for caesarean sections, which is modest compared with the estimated US$30 552 economic return per appendicectomy performed. Appendicectomy constituted about 7% of the case-mix in Sierra Leone, compared with 2% in Liberia. This illustrates how the weighting of procedures and case-mix composition can influence economic estimates. Furthermore, there are some differences related to the completeness of analysed data. About 46% of the data from Sierra Leone were excluded due to missing variables, compared with 8% excluded data in the present analysis. There is no way to know how this affected the results. On the other hand, 87% of the operations performed in Liberia had an unknown status of urgency and these procedures were therefore categorised as planned operations. Only 7% of the included procedures were defined as emergency operations; this is likely to be very conservative. Logbook data from a surgical training programme in Sierra Leone indicate that as much as 69% of all operations might be emergency procedures.30 In the sensitivity analysis, we reclassified 15% of unknown hernia operations as emergency procedures, leading to a change in the estimate of economic loss associated with unmet surgical needs of US$170 million. Emergency procedures are weighted with a higher risk of death and typically avert more DALYs than planned operations; hence, the completeness of data can have a significant impact on economic estimates.

Calculating DALYs for patients undergoing surgery is complex as it requires a framework of weighted values. There is a need to quantify the risk of death, treatment efficacy and disability weights for a broad range of surgical conditions. There were large differences in economic loss averted for various procedures, especially striking for emergency and elective hernia repair. These differences should not be misinterpreted as an argument to invest in certain procedures, as most of the operations in the present analysis are considered essential and highly cost-effective4; rather, the differences should be seen as an example of how small changes in procedure weights, especially for high volume procedures, have the power to alter results. The exercise where planned hernia operations were reclassified as a severe abdominopelvic problem demonstrates how the weighting of procedures can change the estimates. Salomon et al22 defined specific disability weights for a wide range of conditions to improve the measurement of the global burden of disease. This framework makes the methodology transparent and easily reproducible. Such a process, defining standardised weights for threat to life and treatment efficacy for surgical procedures, could be a necessary next step to ensure comparability of DALY estimates across nations and over time.

The major strength of this study is its nationwide design with the inclusion of surgical facilities at all levels of care. Furthermore, facilities from both the governmental and the private sector were included. Some limitations, however, should be considered when interpreting the results. Although 5000 operations per 100 000 population have been used by previous authors to define the total need for surgery,12 21 such an approach has limitations. Disease burden is likely to vary between countries, as well as across important dimensions of equity; unfortunately, social and local considerations were not factored into the present analysis. Furthermore, the Lancet Commission2 benchmark does not offer a breakdown of population needs for individual procedures. A prevalence study would be better suited to inform unmet burden of disease and could also provide a more granular picture of the need for individual procedures. Such data is, unfortunately, not available from Liberia. Population-based studies have provided insight into the unmet burden of surgical disease in Sierra Leone,31 32 and establishing such data in Liberia should be a focus for future research. Another area of focus for future research should be to determine the implementation costs of scaling up surgical services, as such estimates would be highly valuable from a policy point of view.

When interpreting the findings of this study, the classification of procedures as either death or disability averting should be considered; this is an oversimplification of reality. Amputation of a limb, for example, is classified as a death averting procedure. A successful procedure, however, would lead to a permanent disability, and this is not accounted for in our analysis. We replicated the classification and procedure weights developed by Grimes et al,12 which improves the comparability of the findings; nevertheless, classifying procedures as either death or disability averting is likely to facilitate an underestimation of DALYs.

Another limitation is the small number of ophthalmic operations in the case-mix. The eye clinic in the largest teaching hospital in Monrovia is in a separate building on the hospital campus, and this building was missed during the data collection. A total of 180 ophthalmic operations were performed in the hospital in 2012,33 and it is likely that an even higher number was performed in 2018. Furthermore, we calculated the DALYs averted by current surgical activities and assumed that these figures were representative also for the unmet surgical need. This is, however, unlikely to be the case. The proportion of caesarean section and hernia surgery has been found to decrease with increasing GDP per capita.34 In Liberia, most surgical procedures are essential,18 and these are all highly cost-effective.4 As more specialised services develop, operative activities are likely to avert fewer DALYs per procedure performed. To assume that the case-mix is the same for both the met and the unmet surgical need may therefore lead to an overestimation of DALYs for the unmet need. On the other hand, as unmet surgical need decreases, barriers to surgical care will also decrease, and many emergency cases with high DALY averting potential are more likely to end up in the operation theatre. This is not adjusted for in our analysis, pushing the estimate towards an underestimation. Moreover, all procedures where the status of urgency was unknown were defined as planned procedures, contributing towards an underestimation. Finally, planned hernia operations, which is a high-volume procedure, were weighted as a mild abdominopelvic problem, pushing the estimate in a conservative direction again. Overall, we believe that our estimate lands on the conservative side.


A large burden of untreated surgical disease results in substantial productivity losses, leading to high economic costs. For countries like Liberia, such costs are heavy to bear. The major question for policymakers is how to best use existing resources. Training specialists are important; however, in terms of economic return per surgical provider, the non-specialist physician workforce should be recognised as an important economic engine. Creating a framework where specialist and non-specialist physicians are trained to collaborate, with clearly defined scopes of practice, is likely to give a better return on every dollar spent compared with a narrow approach focusing on training specialists alone. Enabling policymakers to observe the economic benefits generated by surgical activities over time may motivate to stable investments. To ensure that estimates are comparable over time, their methodological frameworks need to be consistent. Defining standardised weights to calculate DALYs averted by surgical procedures is a logical next step to ensure comparable estimates over time.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

The Regional Committee for Medical and Health Research Ethics in central Norway exempted this study from review (number 2018/1008). The Institution Review Board, University of Liberia, granted ethical clearance for the study. All administrative leaders of the included facilities consented on behalf of their facility to participate.


The authors thank Nelson Dunbar, Lekilay Tehmeh and Bangalie Kamara for their support during the data collection for this study. The authors are also grateful to all health facility leaders who made the collection of data possible.


Supplementary materials


  • Twitter @AskimAdde, @carisgrimes, @matthew_quaife

  • Contributors HAA, MDO, AJvD, CEG, MQ and HAB conceived the study and wrote the study protocol. HAA, MDO and TCH collected the data, supervised by HAB. HAA, AJvD, CEG and MQ analysed the data. HAA wrote the first manuscript draft. All authors reviewed the article and approved the final draft. HAB is responsible for the overall content of this work.

  • Funding The data collection was funded by Torunn and Oles Foundation, a Norwegian non-governmental organisation. The funding source had no role in any aspect of the study.

  • Map disclaimer The inclusion of any map (including the depiction of any boundaries therein), or of any geographic or locational reference, does not imply the expression of any opinion whatsoever on the part of BMJ concerning the legal status of any country, territory, jurisdiction or area or of its authorities. Any such expression remains solely that of the relevant source and is not endorsed by BMJ. Maps are provided without any warranty of any kind, either express or implied.

  • 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.