Background Identification of risk factors for lower extremity (LE) injury in sport and military/first-responder occupations is required to inform injury prevention strategies.
Objective To determine if poor movement quality is associated with LE injury in sport and military/first-responder occupations.
Materials and methods 5 electronic databases were systematically searched. Studies selected included original data; analytic design; movement quality outcome (qualitative rating of functional compensation, asymmetry, impairment or efficiency of movement control); LE injury sustained with sport or military/first-responder occupation. The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines were followed. 2 independent authors assessed the quality (Downs and Black (DB) criteria) and level of evidence (Oxford Centre of Evidence-Based Medicine model).
Results Of 4361 potential studies, 17 were included. The majority were low-quality cohort studies (level 4 evidence). Median DB score was 11/33 (range 3–15). Heterogeneity in methodology and injury definition precluded meta-analyses. The Functional Movement Screen was the most common outcome investigated (15/17 studies). 4 studies considered inter-relationships between risk factors, 7 reported diagnostic accuracy and none tested an intervention programme targeting individuals identified as high risk. There is inconsistent evidence that poor movement quality is associated with increased risk of LE injury in sport and military/first-responder occupations.
Conclusions Future research should focus on high-quality cohort studies to identify the most relevant movement quality outcomes for predicting injury risk followed by developing and evaluating preparticipation screening and LE injury prevention programmes through high-quality randomised controlled trials targeting individuals at greater risk of injury based on screening tests with validated test properties.
- Lower extremity
- Functional movement screen
- Risk factor
Statistics from Altmetric.com
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.
Owing to the increasing prevalence and cost of treating chronic musculoskeletal (MSK) conditions such as hip, knee and ankle osteoarthritis, there has been a call for scientific inquiry focused on shifting the approach taken to manage these conditions away from treatment and towards prevention.1 From an epidemiological perspective, prevention of chronic MSK conditions may include strategies aimed at reducing MSK injuries in susceptible populations (primary prevention) and/or strategies aimed at slowing down or halting the onset of the chronic MSK disorder after a MSK injury has occurred (secondary prevention). Susceptible populations for MSK injury include those that participate in sport and recreation, or have a service-related occupation (eg, military and first responders such as police officers, fire fighters and paramedics).2 ,3 For example, epidemiological surveys have shown that the risk of injury is 1.5–2.0 times greater among individuals who participate in a variety of sporting and physical activities,2 and that MSK injury is the leading cause of disability in the military.4 In both of these at-risk populations, the most common MSK injuries are those that involve the lower extremities (LEs).3
van Mechelen et al5 proposed a four-step model for injury prevention. This model involves establishing the extent of the specific injury burden of interest, followed by identifying injury risk factors and causal mechanisms through prospective analysis. The first two steps inform the development and introduction of preventative strategies (step 4), which should then be evaluated to determine their impact on injury burden (step 5). Finch6 expanded on this model emphasising the importance of evaluating the effectiveness of preventative strategies in real-world implementation contexts, and Meeuwisse et al7 emphasised the importance of acknowledging that injury is a consequence of complex interactions of multiple risk factors and inciting events. Consequently, studies aimed at identifying risk factors for LE injuries and accompanying preventative strategies should engage end-users, use a prospective design and ensure an adequate sample size to facilitate biostatistical methods that consider the inter-relationships between various risk factors.8 Further, to establish the value of injury risk screening on injury burden, it is crucial that there is an accumulation of high-quality evidence indicating that an intervention programme targeting those at high risk of injury based on a screening programme is more beneficial than a non-targeted intervention.9
Concomitant to injury prevention models is the development of approaches to identify ‘high-risk’ individuals. Identification of these individuals enables prevention programmes to be individually targeted, improving their effectiveness and public health impact (eg, healthcare cost reduction). One method that is widely used to identify individuals at high risk of injury are movement screening tests. The value of these tests is that they can be administered on-field or in clinical settings and are less costly than tests that require specialised equipment or highly trained personnel (eg, laboratory tools such as three-dimensional motion analysis). Additional advantages of movement screening tests are that they can be administered to a large number of individuals, are easily adapted to various sporting or occupation environments and provide almost immediate results. Further, as movement is modifiable, these tests provide information that can directly inform a prevention strategy and possibly assist in return to activity decisions.
Movement screening can involve the assessment of a single movement task or a composite battery of movement tasks. Further, participants can be assessed on their physical performance and/or the quality with which they move. Assessment of physical performance would consider a quantifiable outcome(s) of sport or occupational strength, power, balance, agility, etc, often through multijoint movements (eg, Triple Single Leg Hop, Y Balance Test).10 ,11 Conversely, assessment of movement quality involves qualitative identification and rating of functional compensations, asymmetries, impairments or efficiency of movement control through transitional (eg, squats, sit-to-stand, lunge) or dynamic movement (eg, hopping, walking, running, landing, cutting) tasks. Both physical performance and movement quality assessments would ideally align with the sport-specific or occupation-specific context. Although there is consensus and several recent high-quality summaries of the clinimetric properties (eg, validity, reliability and diagnostic accuracy) and evidence related to predicting injury risk and successful return to sport for physical performance outcomes,10 ,11 the same cannot be said for movement quality outcomes.
As identification of risk factors and casual mechanisms are precursors to the development of effective prevention strategies, the lack of consensus related to movement quality risk factors for LE injury in sport and service occupations has likely hindered the process of developing and evaluating injury prevention strategies. The primary objective of this systematic review is to determine whether screening movement quality (a qualitative rating of functional compensation, asymmetry, impairment or efficiency of movement control either with individual movements tasks or a composite battery of movement tasks) can predict LE injury in sport and/or occupational (eg, military, first responders) populations of all ages. A secondary objective is to summarise the clinimetric properties of the movement quality screening tests in the identified literature to inform clinicians and future research aimed at the development and use of movement quality screening.
This review was registered in the PROSPERO database (CRD42015026958) and conducted according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines.12
Data sources and search
Relevant studies were identified by searching five online databases, selected based on their relevance to the research topics, from inception to January 2016. These databases included MEDLINE, EMBASE (Excerpta medical databases), CINAHL (Cumulative Index to Nursing and Allied Health Literature), Sport Discus and SCOPUS. The combination of medical subject headings (MeSH) and text words that were used to execute each search was developed in consultation with a health sciences librarian scientist (LD). Online supplementary appendix 1 outlines the search terms used for population, injury type, screening type, screening quality, measurement as well as limits and exclusions, along with combinations of search terms that formed the final search strategy. The search strategy (specifically filters 3 and 4) from Kroman et al13 was heavily used as a source of search terms for screening quality and measurement concepts. Limits included English language; human and MSK condition studies published in peer-reviewed journal. Articles were organised using the reference management software package, EndNote V.7.1 (Thomson Reuters, 2013). The number of references obtained from each search strategy for each database was recorded and a running total constructed. After accounting for duplication, the titles and corresponding abstracts of all returned records were independently reviewed by two of the authors blinded to record author(s) and journal title using a Microsoft Excel workbook designed specifically for screening.14 Data were compiled and consensus (first between the two reviewers and if required by the lead author) regarding potentially relevant studies was reached on items in which there was disagreement. Prior to title and abstract review, all authors independently screened a random sample of 120 titles and abstracts in which they were blinded to authors and journal title and reached strong agreement with the lead author (agreement ranging from 81% to 97%, κ ranging from 0.13 to 0.49) using an Excel workbook designed specifically for this purpose.14 ,15 Finally, two authors independently reviewed the full text of all potentially relevant studies to determine final study selection.
Studies were included if they investigated the prospective association between a movement quality outcome (defined as a qualitative rating of functional compensation, asymmetry, impairment or efficiency of movement control during either an individual movement task or battery of movement tasks) and MSK LE injury (defined as an injury involving the hip joint or distal). Additional inclusion criteria included primary research with original data, analytic or intervention design, an outcome measure of LE injury sustained during sport or military/first-responder occupation participation and an objective exposure measure of one or more potential movement quality risk factor for LE injury. Studies were excluded if they were not written in English or involved animal models or cadavers. Further, conference proceedings or abstracts, editorials, commentaries, opinion-based papers review articles (systematic and narrative), case series, case studies, or studies in which screening did not take place prior to injury onset (eg, cross-sectional) were excluded.
Data extraction and study rating process
Data extracted from each study included study year; design; study location and population (eg, sport, military/first-responder occupation, age, sample size); injury outcome (eg, definition) and how it was ascertained; injury estimates (eg, incidence proportion, incidence rate, prevalence); risk factors; and results (eg, significant and non-significant) including measures of reliability, measures of risk (eg, difference in means, correlations, OR, incidence rate ratios (IRR) and risk ratio (RR)) and diagnostic accuracy (eg, sensitivity, specificity, negative predictive value, positive predictive value, positive or negative likelihood ratios). Two authors independently assessed the quality and level of evidence of each study. Quality of evidence was evaluated based on criteria for internal validity (study design, quality of reporting, presence of selection and misclassification bias, potential confounding) and external validity (generalisability) using the Downs and Black (DB) quality assessment tool which assigns an individual score calculated out of 32 total points for each study (11 points for reporting, 3 points for external validity, 7 points for bias, 6 points for confounding and 5 for power: see online supplementary appendix 2).16 The level of evidence represented by each study was categorised based on the Oxford Centre of Evidence Based Medicine (OCEBM) 2009 model (see online supplementary appendix 3).17 As per study exclusion criteria, levels 1a, 2a, 3a (systematic reviews), 4 (case series) and 5 (opinion-based papers) were not included. Discrepancies in DB scoring or OCEBM categorisation were resolved first by consensus between the two reviewers who rated the study and if required, by the lead author (JLW).
Extracted data, quality and level of evidence were summarised for each study. The quantity, quality and level of evidence for the most commonly investigated movement quality risk factors for LE injury in sport military/first-responder occupation were collated.
Identification of studies
An overview of the study identification process is provided in figure 1. The initial search yielded 8219 articles, 3858 duplicates were removed leaving 4361 potentially relevant articles. Following the removal of studies not meeting inclusion criteria based on abstract review (eg, not human studies, ineligible study design, not sport or military/first-responder occupation, no LE injury, no movement quality risk factor, no association between a movement quality risk factor and LE injury) this was reduced to 119. Subsequent to full article evaluation by the two independent reviewers, 102 were excluded leaving 17 studies deemed appropriate for inclusion to the systematic review. Owing to inconsistent methodology and injury definition, and heterogeneity of the risk factors examined, meta-analyses were precluded (table 1).
Characteristics of the 17 included studies are summarised in online supplementary appendix 4. Sixteen of the 17 studies were cohort studies, representing four countries (13 from the USA,18–30 1 each from Canada,31 Iran32 and Japan)33 published between 2007 and 2015. Thirteen18 ,19 ,22–25 ,28–34 of the studies investigated the value of movement quality screening for athletes (including three18 ,24 ,25 involving professional athletes and four22 ,23 ,29 ,30 involving National Collegiate Athletic Association athletes;22 ,23 ,29 ,30 2128 total participants; 1159 males and 817 females), two20 ,27 in the military (total male participants 3350) and two21 ,26 in first-responder trainees (total participants 1153; unable to distinguish by sex). Fourteen18–20 ,24–34 of the studies are believed to have included male participants, while seven19 ,22 ,26 ,28 ,29 ,32 ,34 of the studies included female participants. Five21 ,23–25 ,30 of the studies did not specify participant sex; however, based on the sport or military group investigated in three24 ,25 ,30 of these, it is likely the participants were male. The age range of the athletes was 11–25 years, military members 18–57 years and first-responder trainees 11–22 years. One of the studies involving first-responder trainees21 and three18 ,24 ,25 with athletes did not report age range. Among the 17 studies, 320 ,26 ,27 had a sample size >500, 520 ,23 ,25 ,29 ,30 had at least 50 injury cases (range 7–916 with 3 not reporting the number of injured participants) and 420 ,27 ,29 ,33 used a multivariable statistical approach to identify if movement quality outcomes could identify injury risk. Five22 ,24 ,25 ,28 ,31 of the 17 studies included a metric of diagnostic accuracy, and no studies were identified that assessed the value of screening for movement quality on reducing the burden of LE injury.
Fifteen18–27 ,29–33 of the 17 studies (88%) employed the Functional Movement Screen (FMS) to assess movement quality, while two28 ,34 used the Lower Extremity Scoring System (LESS). Of those using the FMS, only three22 ,23 ,33 investigated the reliability of their measurement system while one34 of the investigations employing the LESS embedded an assessment of reliability into the study design. The most common reliability statistic estimated was an intraclass correlation coefficient with 229 ,33 of the 17 studies including 95% CIs. One33 study reported estimates of measurement precision.
Descriptions of injury estimates (incidence proportion, incidence rate, prevalence), effect estimates (IRR, RR, OR) and significant and non-significant movement quality outcomes are presented in online supplementary appendix 4.
Quality and level of evidence
The highest level of evidence demonstrated by all reviewed studies was level 2b (cohort study) with the majority (13/17) of studies classified as level 4 which corresponds to low-quality cohort study (n<500, injury sample <50, lack of multivariable analyses).
The median methodological quality for all 21 studies, based on the DB criteria, was 11/33 (range 3–15) with only 9/17 scoring >10. The aim of the DB criteria is to assess scientific study methodological quality (inclusive of randomised and non-randomised intervention as well as observational studies). As all of the included studies were observational in nature, seven items (4, 8, 14, 19, 23, 24 and 27; totalling 10 points) on the DB checklist were not applicable. Areas in which the included studies were consistently limited included incomplete description of how the sample was representative of the population of interest (eg, insufficient description of participant characteristics such as sex, history of previous injury, training exposure); limited description of the characteristics of those lost to follow-up; insufficient reporting of how participants were lost to follow-up and differing length of follow-up were accounted for in statistical analyses; inadequate sample size; and lack of adjustment for potential modification and confounding by factors, such as exposure and previous injury. Of further note is the fact that two of the studies reported significant findings even though the 95% CI of the statistical estimate included a null value and that 11/17 studies were published in non-indexed journals or in journals with an impact factor <2.
Synthesis of results
The quantity, quality and level of evidence for the most commonly investigated movement quality outcomes are summarised in table 1. The most common risk factors investigated included age, FMS total score, FMS total score ≤14, FMS total score ≤12, FMS hurdle step, FMS in-line lunge, FMS deep squat, LESS total score and LESS total score ≥5. Based on this synthesis, there is inconsistent evidence that poor movement quality is associated with increased risk of LE injury in sport and military/first-responder occupation.
To the best of our knowledge, this is the first systematic review examining movement quality risk factors for LE injury in sport and military/first-responder occupations that incorporates both a formal evaluation of study quality and level of evidence. Overall, there is inconsistent low-level evidence that poor movement quality is a risk factor for LE injury or to support widespread adoption of movement quality screening programmes for predicting LE injury in sport and military/first-responder occupation populations. Further, as the identification of risk factors is the first step in the injury prevention process, it remains unknown if movement quality screening has a role in reducing the burden of LE injury in these populations.
It is important to highlight that the findings of this review are based on a synthesis and evaluation of existing literature, and as such they are limited by the inadequacies of studies included. Overall, there was a lack of consistent high-quality evidence to support nominating any particular movement quality outcome as a LE injury risk factor due to inadequate reporting of concepts essential to establishing internal (how well an experiment was carried out) and external (can the results be applied to people and situations beyond the experiment) validity. The biggest threats to internal validity were related to the possibility of selection bias, and the reporting of, and adjustment for, potential influence of factors such as sex, injury history and training exposure. Specifically, due to the lack of participant characteristic reporting, it was often difficult to determine if the individuals selected for a study differed systematically from those in the source population (selection bias). Equally important was the consistent omission of the characteristics of those lost to follow-up, which made it impossible to determine if participants lost to follow-up were systematically different from those retained in a study. The inability to determine selection bias not only questions the internal validity of several included studies, it impacts the degree to which the findings of these studies can be generalised to the larger population from which the samples were drawn (external validity).
As indicated earlier, it is highly unlikely that a LE injury is a result of a single risk factor or aberrant movement pattern, but rather the consequence of complex interactions between multiple risk factors and inciting events.7 Multivariable biostatistical techniques can explore these complex interactions given an adequate sample size. Bahr and Holme8 estimated that 50 injury cases are needed to detect a moderate-to-strong association between a risk factor and injury. Of the 17 studies included in this review, only 420 ,27 ,29 ,33 employed multivariable biostatistical techniques, of which only 220 ,29 had 50 or more injury cases (with 127 not reporting the number of injury cases) and were able to assess the influence of additional covariates (eg, body mass index, smoking status, muscular and cardiovascular fitness, battalion, previous injury history, sex, age and sport). As sex, previous injury and exposure to training are known to influence the incidence of MSK LE injury, the lack of reporting and assessment of the impact of these factors on the association between a preseason movement quality deficit and injury incidence with adequate sample size and biostatistical techniques brings into question the value of only assessing movement quality to establish injury risk.
It is important to consider that the true value of being able to identify risk factors for future injury is dependent on it actually leading to strategies that result in injury reduction in real-world contexts. This empirical validation requires an accumulation of high-quality evidence indicating that an intervention programme targeting those at high risk of injury based on a screening test is more beneficial than a non-targeted intervention.9 However, before such a hypothesis can be tested, there needs to be an accumulation of high-quality evidence demonstrating a strong relationship between the risk factor, which was assessed with a valid and reliable screening test employing a specific cut-off value, and injury. Further, the predictive ability of the specific screening test cut-off value must be validated in multiple populations.9
To date, movement quality tests lack the foundation of rigorous development and validation (eg, psychometrics) common in other fields,35 and there is a lack of high-quality evidence demonstrating a strong relationship between any single movement quality outcome and injury. As movement quality tests were not specifically developed as diagnostic tools, but rather to identify deficits that inform clinical interventions from a mechanistic perspective, this is perhaps not surprising.9 ,36 Although several cut-off points for high LE injury risk have been proposed (eg, FMS total score ≤14,24 FMS total score ≤1230 and LESS total score ≥528), none appear to have sufficient diagnostic accuracy to be useful in real-world contexts. For example, the sensitivity and specificity of a FMS total score ≤14 has been shown to range between 0.54–0.83 and 0.61–0.91, respectively.22 ,24 ,25 ,31 This suggests that almost half of individuals that go on to suffer a injury may not have a FMS total score ≤14 and over half of those that do not go on to suffer an injury may not have a FMS total score ≤14.
The ability to establish a link between poor movement quality and injury risk holds great potential for identifying modifiable causal mechanisms for injury, which can be addressed with a targeted intervention. For example, the LESS aims to identify movement quality errors during jump-landing28 such as decreased hip flexion and knee valgus that have been associated with ACL injury. In doing so provides a starting point for targeted interventions aimed at improving jump-landing mechanics and reducing an individual's future ACL injury risk. However, the link between other movement quality outcomes that have an association with LE injury, such as reduced shoulder mobility,19 other FMS components or total FMS score, is not as intuitive. Without a theoretical basis linking the ‘non-optimal movement’ to the injury, it would be difficult to know how to use the finding of an ‘abnormal’ movement to guide an intervention aimed at reducing injury risk. This is a limitation of some movement quality outcomes that will have to be addressed prior to widespread application.
A final consideration is that this review was unable to identify any investigation that had assessed the value of screening for movement quality for reducing the burden of LE injury through targeted interventions in sport or military/first-responder occupation populations. With that said, there are several examples of attempts to do this in the field of sport injury prevention using physical performance outcomes that can provide valuable guidance.37 ,38 Specifically, these studies highlight the importance of developing an implementation strategy in conjunction with the intervention and then tracking and accounting for adherence to the prevention programmes in the analysis.
Meta-analyses were not possible due to the fact that the assumptions for meta-analyses were not satisfied by the included studies. In particular, there was considerable inconsistency in methodology (eg, reporting and controlling for confounding) and heterogeneity of injury definition. For example, the injury definition covered the span of ‘a MSK injury resulting from organised intercollegiate sport practice or competition that required medical attention or advice from a certified athletic trainer, athletic training student or physician’22 to ‘a MSK injury that occurred during participation in track and field practice or competition that prevented participation for 4 weeks’.33 This inconsistency in injury definition led to injury estimates ranging from 0.8% to 85% of participants across the included studies. Further, despite a comprehensive search strategy and rigorous approach to study selection, it is important to acknowledge the possibility of omitting a relevant study and inclusion of only English language articles as additional potential limitations. Finally, as the findings of this review are based on a synthesis and evaluation of existing literature, it is important to point out that the current evidence base of studies that have assessed the prospective relationship between poor movement quality and LE injury may not have considered all possible movement quality screening tests (eg, Nine Battery Test,39 Performance Matrix,40 Single Leg Squat,41–43 Tuck Jump Assessment and Star Excursion Balance Test44).
Both cohort and intervention study designs can play an important role in identifying potential risk factors and reducing the burden of LE injury in sport and military/first-responder occupations.8 ,9 While cohort studies are critical for establishing temporality between a risk factor and subsequent injury, randomised controlled trials (RCTs) provide the strongest evidence for the causal nature of a risk factor and the effectiveness of modifying that factor on injury burden. Based on the studies reviewed, it is recommended that future research focus on high-quality cohort studies aimed at identifying the most relevant movement quality outcomes for predicting injury followed by establishing the diagnostic accuracy of the movement quality screening tests used to assess these risk factors in relevant populations. Given the challenges and high cost of undertaking high-quality cohort studies, an alternative approach may be to simultaneously develop and evaluate preparticipation screening and LE injury prevention programmes through high-quality RCTs targeting athletes or workers at greater risk of injury based on previous injury. Further recommendations include ensuring consistency in injury definition among studies attempting to determine the relationship between a movement quality outcome and subsequent injury that aligns with international consensus45 and the development of movement quality screening tools according to psychometric principles.35 Implementation of these recommendations will assist in the advancement of injury prediction and prevention.
Overall, there is inconsistent level 4 evidence that poor movement quality is a risk factor for LE injury or to support widespread adoption of movement quality screening programmes for predicting LE injury in sport and military/first-responder occupation populations. It is recommended that future research focus on high-quality cohort studies to identify the most relevant movement quality outcomes for predicting injury. This should be followed by development and evaluation of preparticipation screening and LE injury prevention programmes through high-quality RCTs targeting athletes or workers at greater risk of injury based on psychometrically sound movement screening tests.
What are the new findings?
There is conflicting level 4 evidence that movement quality outcomes are risk factors for lower extremity injury in sport or to support widespread adoption of movement quality screening programmes for predicting lower extremity (LE) injury and military/first-responder occupation populations.
There is a need for consistency in injury definition among studies attempting to determine the relationship between a movement quality outcome and subsequent injury.
Based on the work carried out in the field, it is recommended that investigators focus on high-quality cohort studies to identify the most relevant movement quality outcomes for predicting injury risk that account for the multifactorial nature of injury by ensuring adequate sample size and employing relevant biostatistical techniques.
All of the authors are members of the International Movement Screening and Interventions Group of the Arthritis Research UK Centre for Sport, Exercise and Osteoarthritis http://www.sportsarthritisresearchuk.org/seoa/international-movement-screening-and-interventions-group-imsig/imsig.aspx. The authors would like to acknowledge the assistance of Matt Attwood (PhD student), Department for Health, University of Bath, Bath UK; Connor Power (PhD student), Faculty of Health Sciences, University of Southampton, Southampton, UK; and Tim Gribbin (Research Coordinator), Uniformed Services University of the Health Sciences, Bethesda, Maryland, USA for their help in screening titles and abstracts, as well as Joanne Bartram, Arthritis Research UK Centre for Sport, Exercise and Osteoarthritis Administrator for her administrative support.
Twitter Follow Jackie Whittaker @jwhittak_physio, Carly McKay @Dr_CMcKay, Carolyn Emery @CarolynAEmery, Darin Padua @DarinPadua, Sarah de la Motte @sarahdlm
Contributors JLW, CAE, NB and DW were responsible for the conception of the study while all authors were involved in study design. LD designed and executed the extensive search strategy. All authors (except LD) independently reviewed the literature, participated in rating the literature and extracted data. JLW was the primary author in preparing the manuscript; however, all authors contributed to the interpretation of the findings, critical revision of the manuscript and reviewed the document prior to submission.
Funding NB is funded by the National Institute for Health Research (grant number: CDRF-2014-05-021); DW and MW are funded by Arthritis Research UK (grant number: 20194); CLL is supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health (award number K23 AR063235). CAE is funded through a Chair in Pediatric Rehabilitation (Alberta Children's Hospital Foundation).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.