Article Text
PDF
Abstract
Objective To explore if there are differences in the design and/or conduct of studies that have tested the STarTBack treatment approach for the management of low back pain (LBP), potentially explaining differences in study results.
Design A literature review.
Data sources MEDLINE, CINAHL and EMBASE were searched from inception to 26 July 2023.
Eligibility criteria We included studies that included (1) participants with LBP and/or leg pain, (2) randomised controlled trials, controlled clinical trials and interrupted time series designs, (3) used the STarTBack Tool to stratify participants into subgroups and (4) studies provided matched treatments according to participants STarTBack score.
Data extraction and synthesis Two review authors independently reviewed the search results and extracted data into the data extraction form. Due to the exploratory nature of this study, results are presented descriptively.
Results 11 studies conducted across 5 countries were included. There were substantial differences in the proportion of participants allocated to the different risk groups; low-risk group (range: 19%–58%), medium risk group (range: 31%–52%) and high-risk group (range: 6%–38%). There were large differences between studies in the implementation of the STarTBack approach. The original STarTBack trial (Hill et al, 2011) had a more explanatory design while in many subsequent studies, the design was more pragmatic/real world. Only the two original studies provided clear evidence that the implementation of the STarTBack tool led to a higher proportion of participants receiving matched treatment. In the other studies, there was no evidence of a difference, or it was unclear. In two studies, a researcher made the decision about which matched treatment participants received based on the STartTback Tool, while in nine studies, this was done by a clinician. Most studies recommended the same matched treatment for each risk group as per the original study except for a small number of studies. Only three studies reported whether the clinician delivering matched treatment followed the recommended treatment as per the tool. There was substantial variability in the training clinicians received.
Conclusions Reporting of important study-level factors (eg, differences in study design, whether clinicians were trained and how the tool was used in each study) in how the STarTBack approach was implemented was unclear. There is some suggestion that key factors may include the individual who implemented the STarTBack tool, whether the recommendations of the tool were followed, the amount of training the clinician delivering the matched treatment received, and whether clinicians actually delivered the matched treatment.
- back pain
- prognosis
- spine
Data availability statement
All data relevant to the study are included in the article or uploaded as online supplemental information.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
Strengths of this study include using a sensitive search strategy and following a strict prespecified protocol.
We searched various databases and included several different study designs,16 which allowed us to summarise the available literature on the study-level factors that may explain the differences in findings for studies using the STarTBack approach.
A limitation of our review is that many studies did not clearly report on the study-level factors we aimed to explore.
Due to the qualitative nature of the study-level factors explored, we aimed to mitigate misinterpretation using two independent reviewers and a third reviewer where necessary.
Introduction
Participants with low back pain often seek treatment in primary care from clinicians such as a general practitioner (GP) or physiotherapist.1 There is currently limited evidence to guide clinicians to shift from generic treatment and move towards precision or personalised medicine approaches.1 One such approach that has been investigated in a number of studies is risk-based stratified care.
Risk-based stratified care involves selecting treatments according to a participants’ risk of persistent disabling pain, to maximise treatment benefit and reduce potential harm and wasted resources associated with unnecessary interventions.2 The STarTBack approach is a widely known and investigated example of a risk-based stratified care approach, introduced by Hill et al.3 The STarTBack approach was originally developed with the intention to match participants to specific treatment approaches based on prognosis or risk of poor clinical outcomes.4 The initial randomised controlled trial (RCT) by Hill et al 5 (n=851) and the follow-up pre–post design study by Foster et al 6 (n=922) assessing the effectiveness of the STarTBack approach compared with non-stratified therapist-led care, demonstrated initial promise. The Hill et al 5 and Foster et al 6 studies reported that the STarTBacK approach resulted in slightly better participant outcomes (effect on 0–100 scale: 5.5 points (95% CI 2.3 to 8.6) at 4 months follow-up, and more efficient use of resources in primary care participants with low back pain.
Since the initial studies from Hill et al 5 and Foster et al,6 subsequent trials1 7–15 investigating stratified care using the STarTBack Tool have failed to replicate the initial positive results. For example, Morsø et al performed a randomised trial and found no difference in pain or disability outcomes between low back pain participants who received stratified care compared with non-stratified care.13 This may explained by study-level factors, for example, differences in study design, how the STarTBack Tool was implemented, and whether clinicians in studies were trained.
The failure of subsequent studies1 7–15 to replicate the positive results of the initial STarTBack studies5 6 may potentially be explained by differences in the design and/or conduct of subsequent studies. For example, the Hill et al trial5 was an explanatory trial testing the effectiveness of the STarTBack approach under ideal conditions and was less reflective of routine clinical care (although still an important study design). This could largely explain the high fidelity, and therefore, may explain the small but significant outcome in favour of a stratified approach in this study.5 However, in subsequent studies such as Delitto et al,10 the STarTBack tool was used to generate recommended treatments, however, this was used as an opportunity for clinicians and participants to discuss the most appropriate treatment option. Intervention fidelity was not recorded in this study,10 however, the way the STarTBack tool was used may be more reflective of what occurs in clinical practice and the challenges associated with implementing stratified approaches. Exploring the differences between studies will help clinicians and researchers understand how best to implement the STarTBack Tool (if at all) and may inform the development and testing of other complex models of care. Given the conflicting study results regarding the effectiveness of the STarTBack approach and no studies have explored factors potentially influencing these differences, we conducted this literature review to qualitatively explore any differences.
Objective
Primary objective
To explore differences in study-level factors in studies that have tested the STarTBack treatment approach for the management of low back pain, and to investigate whether they may explain differences in study results.
Methods
A review protocol was specified in advance and registered on OSF Registries: https://doi.org/10.17605/OSF.IO/QJZXY.
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 study.
Search strategy
MEDLINE, CINAHL and EMBASE were searched from inception to 26 July 2023. The search strategy included a combination of several terms for each of three domains: low back pain, stratification and RCTs. The full search strategy is presented in online supplemental appendix S1. We also screened reference lists of included studies for additional studies. Forward citation searching was also performed. We included studies in any language.
Supplemental material
Inclusion criteria
Eligible studies met the following criteria: (1) RCTs, controlled clinical trials and interrupted time series16 designs
Used the STarTBack Tool to stratify participants into subgroups.
Studies provided matched treatments according to participants STarTBack score (even if a different matched treatment to that originally proposed by Hill et al 5 was used).
One review author (CSH) excluded clearly irrelevant titles. Two review authors (CJ and CSH) independently screened abstracts to exclude irrelevant studies. Two reviewer authors (CJ and CSH) then independently reviewed full texts for eligibility based on the inclusion criteria. Any disagreements were resolved through discussion and a third reviewer (CM or MH).
Study-level factors that were qualitatively explored
The following study-level factors were selected to be explored as they reflect the variables that would most likely explain the differences in study results in studies that used the STarTBack approach.
Who administered the STarTBack Tool to the participant?
Did the implementation of the STarTBack Tool lead to changes in the recommended or received interventions?
Who made the decision about which matched treatment participants received based on STarTBack score?
Which clinician provided matched treatments?
Did the clinician delivering matched treatment follow the recommendations for treatment based on the tool?
Was the clinician implementing the matched treatment trained to do so?
Treatment received in the control group.
Deviations from the protocol
We intended to explain the differences in study results by extracting pain and disability outcomes at all available time points, however, this was not performed due to the exploratory nature of the study, small number of included studies and did not seem appropriate.
Data extraction and analysis
Two review authors independently reviewed the search results and extracted data into the data extraction form. A form was used to record characteristics of the study (eg, study design, healthcare setting, participant age, follow-up duration, duration of pain) and study-level factors as listed above.
Due to the exploratory nature of this study, results are presented descriptively. Data relevant to (but not limited to) study design used, the healthcare setting the of each study (including country) and characteristics of the participants (eg, age, sex, duration of pain) were extracted.
We defined community clinicians (eg, GP, physiotherapist) as clinicians who practise in the community as part of their usual care. We defined a research clinician (eg, GP, physiotherapist) as clinicians working as part of a research team solely for the purpose of the research project.
Results
Results of the search
Our search retrieved 5399 records. After removing duplicates (n=190) 5209 records were identified. Of these, 4866 were clearly irrelevant based on title screening. A further 323 records were excluded based on title and abstract screening. Following full-text review by two independent authors, 10 studies were included (figure 1). One additional study was found by contacting experts in the field, resulting in a total of 11 included studies. Excluded studies and primary reasons for exclusion are presented in online supplemental appendix S2.
Study selection.
Characteristics of included studies
11 studies were included.5–15 Sample sizes ranged from 109 to 2971 (median: 550), studies were conducted in five different countries (three in the UK, five in the USA, one in Denmark, one in the Netherlands and one in Ireland) and were all conducted in primary care. Participant population data were reported in all 11 studies and are presented in table 1.5–15 Nine studies included participants with low back pain and one study included participants with sciatica and/or low back pain.11 One study included participants with neck and low back pain.9 Follow-up ranged from 4 weeks to 12 months. There was substantial between-study variability in the proportion of participants allocated to the different risk groups. The proportion of participants in the low-risk group ranged from 19% to 58%, in the medium risk group from 31% to 52% and in the high-risk group from 6% to 38%. One study9 did not report the results of the proportion of participants in each risk group. Study-level variables are presented in table 2 and further details are presented in online supplemental appendix S3.
Characteristics of included studies
Study-level implementation variables
Study population and protocol for matched treatment
Participant population data were reported in all 11 studies and are presented in table 1.5–15 Most studies recruited similar participants to the original studies,5 6 including participants with low back pain with or without leg symptoms. However, in one study,11 only participants with sciatica and/or low back pain were recruited which resulted in altered matched treatment received in this study compared with the other included studies. In another study, the medium-risk and high-risk participant groups both received the same matched treatment and participants were combined and reported in total,9 therefore, did not follow the matched treatment as per the original STarTBack approach. One study10 only included participants who scored as high-risk on the STarTBack Tool, however, the matched treatment provided followed the STarTBack approach. All other studies5–8 11–15 used the same STarTBack approach to matched treatment as the original studies by Hill et al 5 and Foster et al.6 A flow diagram of how the STarTBack Tool is implemented and where the main differences between studies in the tool’s implementation is presented in figure 2.
Study design and healthcare setting
Four studies5 11 13 15 were RCTs and five studies7–10 12 were cluster RCTs. One study14 was a non-RCT and one study6 was a sequential comparison.
Who administered the STarTBack tool to the participant
Data on who administered the STarTBack tool to the participant were available in 8 of the 11 studies.5–8 11 13–15 In the original Hill et al 5 study, a research clinician (ie, a research nurse) administered the tool, and in three other studies8 11 13 a member of the research team (ie, research specialist, project secretary, research clinician) administered the tool. The community GP administered the tool in one study by Foster et al 6 and a community physiotherapist administered the tool in one study.14 The tool was self-administered by the participant in two studies7 15 and it was unclear in three studies9 10 12 who administered the tool.
Who made the decision about which matched treatment participants received based on STarTBack score?
Data on who made decisions about the matched treatment based on the use of the tool were available in 8 of the 11 studies.5–8 10 11 13 15 In two studies5 11 including the original Hill et al 5 study, a member of the research team (ie, research physiotherapist) made the decision regarding the participants treatment, based on the tool. The community GP made decisions about treatment using the tool in one study by Foster et al,6 and a community physiotherapist made the decision about participant treatment in one study.7 In one study, a community clinician (ie, physicians, nurse practitioners or physicians’ assistant) made decisions about treatment using the tool.15 In three studies,8 10 13 participants were stratified through a variation of electronic health database record systems based on results of the tool, however, it was unclear who made the final decision regarding the participants treatment. It was unclear in three studies9 12 14 who made the decision about the participants treatment based on the tool results.
Did the implementation of the STarTBack tool lead to changes in the clinicians recommended interventions?
Data on whether clinician behaviour (recommended treatment) changed with the use of the STarTBack tool were reported in only 3 of the 11 studies.5 6 10 In the original Hill et al study,5 implementation of the STarTBack tool led to clinicians following the recommended matched treatment based on the risk-matched recommendations from the tool in 97% of participants and led to higher referral to physiotherapy in the intervention/stratified group (75%) compared with the control group (58%). In the Foster et al study,6 implementation of the STarTBack tool led to clinicians following the recommended matched treatment based on the risk-matched recommendations from the tool in 71% of participants and led to higher referral for physiotherapy in the medium-risk and high-risk groups in the intervention/stratified group (72%) compared with the control phase (40%). In another study,10 the implementation of the tool led to higher referral for physiotherapy in the high-risk group in the intervention/stratified group (57%) compared with the control phase (30%). However, in this study,10 it was not reported if clinicians followed the recommended treatment based on the risk-matched recommendations from the tool. In eight studies,7–9 11–15 it was unclear or not reported as to whether the recommended treatment matched the tool’s recommendation, or if the implementation of the STarTBack tool led to changes in the treatment received.
Which clinician provided matched interventions?
Data on who provided the matched treatments based on their STarTBack score were available in 10 of the 11 studies.5–7 9–15 In the original Hill et al study,5 a research physiotherapist provided matched treatment for one session (for low-risk treatment) at baseline, and then medium-risk and high-risk participants were referred for further community physiotherapy.
For participants stratified to low-risk treatment, a research physiotherapist provided matched treatment in one study,11 a community GP provided matched treatment in one study by Foster et al,6 a community physiotherapist provided matched treatment in five studies,7 12–15 and in one study,9 a community physiotherapist and ‘spine coach’ (professionally trained in motivational interviewing) provided matched treatment. It was unclear in one study8 which clinician provided matched treatment (eg, either a GP, nurse, physiotherapist) and in another study10 only results for high-risk participants were presented, which are discussed below.
For participants stratified as medium risk, community physiotherapists provided matched treatment in six studies,6 7 12–15 a research physiotherapist provided matched treatment in two studies5 11 and in one study9 a community physiotherapist and ‘spine coach’ provided matched treatment. It was unclear in one study8 which clinician provided matched treatment (eg, a GP, nurse or physiotherapist). One study10 only presented results for high-risk participants, where a community physiotherapist provided matched treatment.
For participants stratified as high risk, a community physiotherapist provided matched treatment in seven studies,6 7 10 12–15 a research physiotherapist provided matched treatment in one study (Hill et al),5 and in another study,9 a community physiotherapist and ‘spine coach’ provided matched treatment. In one study,11 including only participants with sciatica and/or low back pain, high-risk participants were referred immediately for MRI and to a Specialist, differing to the original (Hill et al)5 STarTBack approach. It was unclear in one study8 which clinician provided matched treatment (eg, either a GP, nurse or physiotherapist).
Was the clinician implementing the matched treatment trained to do so?
Data on training of the clinician providing the matched treatment were available in 9 of the 11 studies.5–8 10 12–15 In the original Hill et al 5 study, clinicians providing medium-risk treatment received 4 days of training and 10 days of training for high-risk treatment. One study8 trained clinicians (eg, physicians, nurses) in delivering matched treatment over six sessions (1 hour each) over a 6-month period and physiotherapists received 5 days of training for matched treatments for medium-risk to high-risk groups. One study15 trained community physiotherapists over a 2-day course (hours not specified) to provide matched treatments for all risk groups. One study10 trained clinicians for a total of 10.5 hours in delivering matched interventions only for high-risk groups. In one study,13 physiotherapists received 5 days of training, additional booster sessions and access to consulting physiotherapists (trained in STarTBack approach) to provide matched treatment for all risk groups. In another study,7 physiotherapists were trained (up to 12 hours over 4 weeks) in the implementation of the matched treatments for all risk groups. Two studies6 14 reported that physiotherapists were trained, however, did not report the amount of training provided. In two studies,9 11 it was not reported whether training was provided to clinicians on risk-matched treatment, and in another study,12 it was unclear if the training provided was specific for providing matched treatment.
Did the clinician delivering matched treatment follow the recommendations for treatment based on the tool?
Data on whether treating clinicians delivering matched treatment followed the recommended treatment (as per each study protocol for matched treatment) were available in 3 of the 11 studies.5 6 11 In the original Hill et al 5 study, among the participants referred for physiotherapy, initial treatment attendance was reported as 93% in the intervention group (ie, STarTBack group). No data were reported based on each risk group in either of the two studies.5 6 In one study,11 clinicians followed the recommended treatment in the low-risk group in almost all participants (except one participant), in the medium-risk group recommended treatment was followed in 82.5% of participants, and in the high-risk group recommended treatment was followed in 81.3% of participants. In the three studies5 6 11 reported above, the reasons for participants not receiving the recommended matched treatment based on the tool were not clear or not reported. In eight studies,6–8 10 12–15 it was unclear or not reported whether clinicians followed the recommended matched treatment based on the tool. In one study,9 participants in the medium and high-risk groups received the same treatment and data on whether clinicians followed the recommended treatment based on the tool was not reported.
Treatment received in the control groups
Data on the intervention received in the control group were available in 9 of the 11 studies.5–7 9 11–15 In all 11 studies, the STarTBack score was collected, however, all control groups did not use the STarTBack Tool to determine treatment or the results of the tool were not available to the clinicians providing treatment. In the original Hill et al study,5 decisions about referral in the control group were made based on the physiotherapists’ clinical judgement, without knowledge of a participants STarT Back Tool classification. In four studies,5 7 12 14 control participants received some form of intervention by a community or research physiotherapist (eg, treatment based on guidelines). In four studies,6 11 13 15 control participants received different forms of interventions from a community GP or primary care clinician (ie, physicians, nurse practitioners and physicians’ assistant) and they were advised to provide treatment as usual (eg, referral for physio, medication or guideline-based treatment). In one study,9 there were two control groups, one group received an individualised exercise programme (not based on risk score) and the other group received no treatment. In two studies,8 10 it was unclear what treatment was received in the control group.
Discussion
Principal findings
This review found that reporting of implementation of the STarTBack approach was often unclear for many important study-level factors. However, there were important differences between studies in the implementation of the STarTBack approach, which may explain differences in results between studies. The original STarTBack RCT (Hill et al 5) had a more explanatory design while in many of the subsequent studies, the design was more pragmatic/real world. For example, in the original STarTBack RCT (Hill et al 20115), the tool was administered by a researcher, while in two studies,6 14 the tool was administered by a community clinician (which is more reflective of clinical practice). In most studies, it was unclear, or not reported, if the implementation of the STarTBack tool resulted in (between group) differences in the recommended or received interventions. Only three studies5 6 10 (including the original Hill et al study) provided clear evidence that implementation of the STarTBack tool led to a higher proportion of participants receiving matched treatment based on the tool. In the other studies, there was either no evidence of a difference or it was not clear. Only three studies5 6 11 reported whether the clinician responsible for delivering matched treatment based on the tool followed the recommended treatment. Most studies reported that training was provided to clinician, however, there was a large range in the amount of training provided (range: 6 hours to 10 days). The amount of training was reported in only half of the included studies. In all studies, the intervention received in the control group either did not include the use of the STarTBack Tool or the risk profile based on the tool was not available to treating clinicians.
Comparison with previous studies
To our knowledge, this is the first review exploring the potential study-level factors that may explain the differences in study outcomes. Two systematic reviews17 18 have previously investigated the effectiveness of stratified care compared with non-stratified care (including the STarTBack approach). One review (n=8 studies) included four RCTs and four non-RCTs17 and another review (n=24 studies)18 included the STarTBack approach as part of other stratification systems. Neither study17 18 directly answers our research question and no study to date has explored potential factors that may contribute to the differing study results. Some of the included studies may have reported on the study-level factors explored in our study, however, this was not the main objective of the original studies. To our knowledge, this is the first study with the main objective to explore the study-level factors that may explain the differences between study results. However, some individual studies have highlighted some factors, similar to our results. For example, Delitto et al 10 found that GP’s largely did not follow the recommended treatments as per the STarTBack Tool. Cherkin et al 8 also did not encourage clinician adherence to matching treatments to patient subgroups. Although some individual studies may have explored these factors in their discussion, many studies did not provide enough data or any at all.
A recent Journal of Physiotherapy editorial discussed potential reasons behind the different outcomes between studies investigating the effectiveness of the STarTBack approach.19 The editorial considered one particular explanation for the lack of consistency in study results compared with the original studies: intervention fidelity. The editorial highlighted the importance of the two key components of stratified care (ie, subgrouping participants based on risk scores and providing targeted matched treatments based on the tool).19 All studies in our review subgrouped participants based on their risk scores, however, not all participants went on to receive matched interventions based on the tool. Our review highlighted that it was often not clearly reported whether participants received matched treatments (eg, only 3 of 11 studies provided data on whether participants received matched care based on the tool), limiting our ability to explore this further. Reporting on whether participants received matched care based on the tool is critical to interpreting the results of studies as it likely contributes to the different outcomes in studies investigating the implementation of the STarTBack Tool. The effectiveness of the STarTBack approach is also likely dependent on the amount of training clinicians delivering matched care based on the tool received, and therefore, had the clinical skills to provide the matched treatments based on the tool. There was substantial variability in the amount of training provided to clinicians (range: 6 hours to 10 days) and in half the studies it was not clear the amount of training clinicians received or if training of clinicians was even provided.
Strengths and weaknesses of the study
Strengths of this study include using a sensitive search strategy and following a strict prespecified protocol. We searched various databases and included several different study designs.16 This allowed us to summarise the available literature on the study-level factors that may explain the differences in findings for studies using the STarTBack approach. We were able to identify key differences between studies that may be useful for future research and how clinicians may implement this tool in clinical practice. A limitation of our review is that many studies did not clearly report on the study-level factors that we aimed to explore. We focused our results on factors most commonly reported across studies. It is likely that in some study-level factors it was not possible to extract adequate data, or authors did not consider extracting such data, contributed to between study differences. Due to the qualitative nature of the study-level factors explored, we aimed to mitigate misinterpretation using two independent reviewers and a third reviewer where necessary. The purpose of this scoping review did not include assessment of study risk of bias. We did not assess risk of bias for any of the included studies in our review and did not approach this review with a quantitative approach as commonly performed in systematic reviews, as we anticipated heterogeneity in the availability and reporting of data.
Implications for future research
It may be worthwhile for future research investigating stratified care (ie, STarTBack approach) to use relatively explanatory research designs like the original Hill et al 5 study to determine if there is merit in continuing with subsequent pragmatic studies. Following the initial Hill et al study,5 very few,6 if any, explanatory studies have been completed. Important nuances on how the STarTBack approach is implemented and how feasible the approach is may have been missed by performing follow-up studies based on a lack of available research. If future studies demonstrate feasibility, are able to appropriately subgroup participants based on risk scores and provide targeted matched care based on the tool, larger and more pragmatic RCTs should be conducted to assess the clinical effectiveness of the STarTBack approach. Future studies should also focus on how best to implement the STarTBack approach and strategies to monitor and improve the implementation of the tool should also be considered. Future studies will benefit from monitoring how the STarTBack Tool is implemented and ensuring the STarTBack approach is followed as intended by the original authors Hill et al 5 and Foster et al.6 A better understanding of the key variables likely influencing the implementation of matched interventions based on the tool may help inform clinicians of the best approach to implement stratified care in clinical practice. A related issue to implementing the STarTBack approach is whether participants received the matched intervention they were matched to by the STarTBack tool and whether clinicians implementing matched care based on the tool had the appropriate training and skills to do so. Future research should better monitor whether participants go on to receive the intervention they are matched too based on the tool, and more importantly, investigate the potential reasons for why participants did not receive matched treatment based on the tool for example, cost, access, clinician training.
Conclusions
We found that in most studies important study-level factors in relation to the implementation of the STarTBack approach (use of the tool and matched care) were not clearly reported. As a result, it was difficult to determine which factors may have impacted study outcomes. There is some suggestion that the key factors might be the individual who implemented the STarTBack tool, whether the recommendations of the tool were followed, the amount of training the clinician delivering the matched treatment received, and whether they actually delivered the matched treatment. Future research on the STarTBack approach must clearly report on the key implementation element of the approach and be designed to investigate the most important elements of implementation to produce better participant outcomes. Qualitative research may be useful to understand the clinicians experience when implementing the tool in real-life clinical practice.
Data availability statement
All data relevant to the study are included in the article or uploaded as online supplemental information.
Ethics statements
Patient consent for publication
Ethics approval
Not applicable.
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
X @C_Han_1
Contributors CSH: conceptualisation;data curation; formal analysis; investigation; methodology; project administration; resources; validation; visualisation; roles/writing–original draft; writing–review and editing. MH: conceptualisation; data curation; formal analysis; investigation; methodology; validation; visualisation; writing–review and editing. CJ: data curation; investigation; methodology; validation; visualisation; writing–review and editing. CM: conceptualisation; data curation; formal analysis; investigation; methodology; validation; visualisation; writing–review and editing.
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 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.