Study protocol for evaluation of aid to diagnosis for developmental dysplasia of the hip in general practice: controlled trial randomised by practice

Introduction In the UK, a compulsory ‘6-week hip check’ is performed in primary care for the detection of developmental dysplasia of the hip (DDH). However, missed diagnoses and infants incorrectly labelled with DDH remain a problem, potentially leading to adverse consequences for infants, their families and the National Health Service. National policy states that infants should be referred to hospital if the 6-week check suggests DDH, though there is no available tool to aid examination or offer guidelines for referral. We developed standardised diagnostic criteria for DDH, based on international Delphi consensus, and a 9-item checklist that has the potential to enable non-experts to diagnose DDH in a manner approaching that of experts. Methods and analysis We will conduct a controlled trial randomised by practice that will compare a diagnostic aid against standard care for the hip check. The primary objective is to determine whether an aid to the diagnosis of DDH reduces the number of clinically insignificant referrals from primary care to hospital and the number of late diagnosed DDH. The trial will include a qualitative process evaluation, an assessment of professional behavioural change and a full health economic evaluation. We will recruit 152 general practitioner practices in England. These will be randomised to conduct the hip checks with use of the study diagnostic aid and/or as per usual practice. The total number of infants seen during a 15-month recruitment period will be 110 per practice. Two years after the 6-week hip check, we will measure the number of referred infants that are (1) clinically insignificant for DDH and (2) those that constitute appropriate referrals. Ethics and dissemination This study has approval from the Health Research Authority (16/1/2020) and the Confidentiality Advisory Group (18/2/2020). Results will be published in peer-reviewed academic journals, disseminated to patient organisations and the media. Trial registration number NCT04101903; Pre-results.


INTRODUCTION
Developmental dysplasia of the hip (DDH) is characterized by varying displacement of the proximal femur from the acetabulum with associated acetabular dysplasia. Dislocation occurs in 1-2/1000 infants per year but milder forms occur in 40-60/1000. 1 Early recognition of disease is associated with better outcomes. It is national policy 2 to examine all infants for the presence of DDH at birth and at age 6-8 weeks in primary care (6-week hip check). If diagnosed within the first 6-8 weeks, splinting of the hips is successful in 85% of cases. 3 Later diagnoses usually will require invasive treatment, with many years of continued monitoring. 4 Late diagnosed DDH is a common cause of medical negligence claims, with increased suffering for affected patients, as well as multi-million costs to the NHS. Timely diagnoses of DDH remain a challenge despite the compulsory 6-week check. 5 In one study, the median age at diagnosis was 14 months, with only 40% of infants diagnosed during routine examinations and 60% presenting owing to parental concerns. 6 In another study, 30% of infants were not diagnosed by 12 weeks. 7 Because there is no further compulsory check after that at 6-weeks, it is vital that the 6-week check is effective. Also infants incorrectly identified with "DDH" in primary care and referred to hospital remain a challenge for both families and NHS. This group does not require treatment and reassurance provided in primary care would avoid unnecessary anxiety and inefficient use of hospital resources. Of 1,918 infants referred to hospital from primary care for DDH, only 64 (3%) had DDH 7 but 1,270 (66%) were identified as 'DDH' based on inappropriate criteria, 8 e.g. "crease asymmetry" (n=234) or "click" (n=648). If GPs were able to discriminate better between benign abnormalities requiring reassurance (perhaps with a follow-up in primary care) and findings requiring referrals, outcomes of infants with and without DDH would improve.
In prior research we suggested that too many diagnostic criteria 8 and variability in the use of diagnostic criteria among clinicians 9 complicate the task of diagnosing DDH. We developed standardised diagnostic criteria to reduce the variability in assessment and management decisions in infants examined for DDH. 10 The weighted criteria demonstrated validity. 10 In a study of 44 patients referred from GPs to hospital, the weighted criteria demonstrated a positive predictive value of 89% (95% confidence interval 70, 97) and negative predictive value of 76% (95% confidence interval, 50-96). 11 We refined these binary criteria, in form of a checklist with 9 items, for use by GPs during the 6-week hip check and developed a training video, featuring a GP. This checklist and video were refined in a feasibility study which used qualitative methods to explore the acceptability of the format, style and delivery of the intervention (report available on request). The video explains the meaning the diagnostic criteria; it also demonstrates how precisely to elicit the diagnostic criteria. For example, the

Rationale
Current referral patterns suggest there could be considerable health gains from improved diagnostic and referral decisions at the 6-week check. 5 GPs, in a preparatory focus group, identified the need for a 'comprehensive, structured guide' for the 6-week check. Diagnostic aids enable physicians to overcome barriers in diagnostic reasoning 12 by shifting intuitive to analytical aspects of diagnostic reasoning. Decisions made under these circumstances approach normative reasoning and rationality more closely, and are more reliable and safer. 13 Building on our earlier research, we propose to facilitate GP's diagnostic reasoning at the 6-week check by structuring current practice with use of a previously developed diagnostic aid.
A diagnostic aid of this kind provides a structured approach to the assessment of infants and offers guidance about referrals.

Objectives
 To determine whether an aid to the diagnosis of DDH reduces the number of clinically insignificant referrals from primary care to hospital, and the number of late diagnosed DDH.
 To determine the cost-effectiveness of this intervention.
 To conduct an integrated qualitative and quantitative process evaluation in order to understand all participants' experience in the trial and with the intervention; study how the intervention is implemented; investigate contextual factors that affect the intervention.

METHODS AND ANALYSIS
The SPIRIT reporting guidelines were used in the preparation of this clinical trial protocol. This is a cluster randomised controlled trial of GP practices in England, with randomisation and intervention at practice level but the primary endpoint measured at patient level ( Figure   1). An internal pilot will be done in the first 4 months to ascertain accrual rates -we will progress to the main trial if we succeed in recruiting ≥12 practices per month (expected recruitment is ≥21 practices/month for the main trial). Incorporated in the trial are (i) a process evaluation investigating determinants of GPs' referral behaviour and the implementation of the intervention in practice; and (ii) a health economic evaluation in a child's lifetime.

Eligibility and recruitment
GP practices registered in England that carry out 6-week hip checks (at age 42-76 days) and agree (i) to being randomised and (ii) to hospitals releasing data concerning infants they examine in the trial. Ineligible are practices planning to close with 12 months of trial start.
We anticipate that ≥110 infants per year will undergo the 6-week hip check in each of the recruited practices. Each practice will recruit infants for a period of 15 months and infants will be followed for 2 years. This trial is planned therefore to run for 44 months across 152 GP practices.

Retention
The proposed trial is not onerous on the practice staff and practices will not be challenged by excessive study-related manoeuvres. Because the primary outcome will be collected using practice and hospital databases only, it is likely that we will attain high levels of follow up.
We will send periodic newsletters to GPs and foster regular contacts.

Randomisation
Practices will submit an eligibility form to the clinical trials unit, which will register and randomly allocate (concealed) practices within 48 hours. An independent statistician will coordinate randomisation (allocation ratio 1:1, stratified by practice size).

Interventions
Control intervention: GPs will assess infant hips following best practice principles; they will be provide a leaflet about the national best practice recommendations. 2 Experimental intervention: This is a complex intervention comprising a video developed specifically for this trial and a diagnostic aid in form of a 9-item binary checklist. At the time when informed consent is taken, GPs will watch the video. They will then examine all infants in the 6-week check according to best practice principles, but with the addition of the diagnostic aid (implemented electronically to each practice's computer system).
According to national guidelines, the 6-week check shall capture infants at age 6-10 weeks and infants in whom a diagnosis of DDH is considered should see a specialist within 2 weeks. 2 Participating GPs of both intervention arms will be asked to adhere to this policy.

Definitions for the purpose of this trial
'Appropriate referral' denotes a referred hip deemed 'clinically significant', i.e. is treated or monitored (at least one follow-up) by a specialist surgeon. Any ambiguous cases will be reviewed by our Expert Advisory Panel who will assign an ultimate diagnosis in consensus.
'Clinically insignificant' are referred hips resulting in reassurance and discharge from surgeons' clinics (i.e. no treatment or monitoring). 'Late diagnoses' are cases of DDH diagnosed by a specialist surgeon at age 3-24 months. 15;16 Sample size calculation We based its calculation on Poisson regression and audit data from the Nottingham area (partially presented in Price et al 7

) and data of 3 years' practice at University College
Hospital: a conservative estimate of the number of infants referred as a consequence of the 6-week check within the recruitment period of one year is, on average, 3 per practice. Based on the same data, we estimate that about 1 in 3 (or, on average, 1 per practice) of such referrals is correctly made ('appropriate' referrals, defined above as 'clinically significant', are a subset of these 'correct' referrals). Thus, on average, we expect 2 'incorrect' referrals per practice. These are the ones we seek to reduce, and the intervention should serve to cut them by 50%. Thus we aim to detect a reduction, on average, from 3 to 2 referrals per practice in the intervention arm. To account for correlation within practices, which can also be thought of as overdispersion relative to the underlying Poisson variation, we assumed a betweenpractice component of variance of 20% of the average Poisson 'counting' variance per practice. For 90% statistical power and testing 2-sided at p=0.05, we need 76 practices per group (15% of this total comprises a safety margin to allow for potential challenges).
Average referral rates from the Nottingham (2.8%) and University College Hospital data (2.7%) are very consistent, and suggest a target of about 110 children per practice per year to undergo the 6-week check.

Primary trial endpoint
Number of referred infants that are considered 'clinically insignificant' -this is a measure of the clinical importance of referrals in the 2-week hip pathway. We chose this outcome because it is extremely relevant clinically and one for which it is possible to power the trial. 2 An intervention could be successful at achieving this outcome whilst missing infants who do have DDH; we thus specified a principal secondary endpoint. False negatives are also important but, because of their rarity, would not be an efficient endpoint on which to power the trial (and would thus be outside ICH E9, 17 which recommends that the primary outcome should be both clinically and statistically convincing). A research assistant will retrieve these data from respective hospital electronic systems by conducting site visits to all secondary care facilities to which GPs refer infants, starting 6 weeks after the last patient enters the trial. • Number of appropriate referrals per practice (principal secondary endpoint): for all infants referred as a consequence of the 6-week check, a researcher will collect, using unique NHS patient numbers, and categorise the appropriateness of referrals by a practice with respective hospital databases using a standardised taxonomy, blind to practice random allocation. S/he will conduct site visits to all such hospitals starting 6 weeks after the last patient entered the trial.

Secondary trial endpoints (
• Number of late diagnoses: We will employ deterministic methods of data linkage. Using unique NHS numbers of any recruited infant, we will obtain from Health & Social Care Information Centre the corresponding Health Episodes Statistic (HES) identifiers. HES is a data warehouse that includes all hospital admissions and outpatients' visits occurring in all hospitals in England. In HES we will establish whether any infant in the trial was admitted to a UK hospital as a result of DDH using relevant ICD-10 codes and OPCS-4 codes. 18 We will extract, for the whole trial period, the full hospital history for all infants in the trial (in and out patient episodes), collate all episodes into a combined exploratory analysis view, collate interactions, and allocate outcomes to practice groups to facilitate outcome analyses. 19 These data will be compared with the data collected by the researcher (see above); the use of these 2 strategies will enhance the robustness of data.
• Consequences of late diagnoses: Using data collected above we will record nature and frequency of such consequences (nature of treatment, length of hospital stay, frequency of secondary care contacts).
• Health-related quality of life (parental proxy report): Child-Health-Utility-9D. 20 Process level • Volume of referrals: Total number of patients referred to secondary care during the trial period. This variable will be collected prospectively at practice level by the participating GP and forwarded monthly to the clinical trial unit. • Timeliness of referrals: In infants referred to secondary care, we will measure the days from referral issued to hospital appointment (collected in the same way as primary endpoint). This process measure will inform about target wait times 2 achieved.

Clinician level
• Confidence and attitudes of GP and secondary care clinicians towards the diagnostic aid: will be assessed 12 weeks after trial set-up and at trial completion using a modified measure based on the Theory of Planned Behaviour. 21 • Implementation issues and acceptability of diagnostic aid among GPs and secondary care providers: we will conduct qualitative research at the trial end to elicit this information.
• Use of diagnostic aid and acceptability of intervention: a self-administered questionnaire to evaluate the use of the aid will be posted to all the GPs in the intervention group; also direct observations and interviews will ascertain this outcome.

Parent/carer level (collected from one in ten parents/carers)
• General worry: State-Trait Anxiety Inventory 6-items short form. 22 • DDH-related worry: Infant Hip Worries Inventory. 23 • Satisfaction with trial: dimensions of care items from EUROPEP. 24 F o r p e e r r e v i e w o n l y

Statistical analysis
Primary endpoint: We will compare the randomised practices using Poisson mixed models, accounting for extra Poissonian variability by including random intercept terms for practices. The response variable will be the number of clinically insignificant referrals from each practice with an offset in the linear model of the log(e) total number of children checked in each practice, to account for differences in practice size and constitution. The random effect at the practice level will account for overdispersion.
Principal secondary endpoint: This will be analysed as above; however, counts here are expected to be lower. Should they be insufficient to support the full model, the model will be simplified by (e.g. use of additive variance component term rather than a generalised random intercept term to address overdispersion). We will describe the planned statistical analysis and its adaptation where there are difficulties in achieving convergence in a pre-specified statistical analysis plan.
Late diagnoses: As we anticipate small numbers of events in each randomised comparison we may not be in a position to account for practices using random intercept terms. Where this is the case we will report the total numbers over a minimum 2-year observation and compare the overall group scores using Fishers exact test.
Quantitative outcome measures: We will summarize scores by instrument, accounting for practices and report difference in group means for each treatment arm, using appropriate transformation where necessary. We expect mean scores to be lower in the intervention arm for State-Trait Anxiety Inventory and Infant Hip Worries Inventory, but higher or equal for EUROPEP.
Missing data: For the primary endpoint and principal secondary endpoint the data collection methods should identify qualifying episodes. Because of the nature of these data the conventional concept of missingness does not directly apply (eg we will not have individually randomised subjects who cannot be followed up). However if a practice withdraws from the trial we will explore the consequences of this action by assuming a poor outcome among that practice if in the intervention group and a good outcome if in the control condition, to identify the potential consequences of their withdrawal. Complete case analyses will be conducted for secondary outcomes. If there is a mismatch between practices in the two treatment conditions we will consider undertaking joint models examining simultaneously the binomial of missingness and the outcome measure of interest.

Cost analysis
We will analyse the cost associated with the intervention compared with usual practice for the entire trial period, and examine costs from the perspective of the NHS and of families. The cost components included in main analysis are: cost of 6-week in both intervention arms; any subsequent referrals, diagnostic tests and treatment. We will collect costs about GP time, which we will multiply by unit costs from routine sources. 25 Within-trial cost-effectiveness analysis With the costs described above we will produce a dataset of patient-level within-trial costs and outcomes. We will calculate the incremental cost per clinically insignificant referral avoided and the incremental cost per late diagnosis avoided. Using bootstrapping of the mean cost and outcomes differences, we will estimate confidence intervals around the incremental cost-effectiveness ratios. 26 With the bootstrap replications we will construct a cost-effectiveness acceptability curve to show the probability that the aid is cost-effective for different values of NHS willingness to pay for outcomes.
We will perform deterministic sensitivity analyses.

Long-run cost utility analysis
We will use several measures to evaluate the lifetime cost-effectiveness of the intervention. We will ask 20 carers of infants aged 2-4 years to complete (1) on behalf of their children the Child-Healthutility-9D 20 and (2) for themselves the EQ-5D-5L, 27 both measure health-related quality of life. With data from the trial about the impact of the intervention on appropriate referrals, we will calculate the monetary value that parents place on the intervention using willingness-to-pay methodology. 28 This will provide an estimate of the monetary value of the additional benefits (positive/negative) of the intervention. We will calculate the net benefit of the intervention by subtracting the incremental cost of the aid, as calculated above from the trial data, from the monetary value of the additional benefit. Following recruitment of the last infant in the trial, we will recruit 200 carers of infants undergoing the 6-week check from trial-participating practices. These will be 100 carers whose infants will be referred to secondary care as a consequence of the 6-week check and 100 who will not. They will complete a self-report questionnaire that utilised several techniques 28 to elicit willingness-to-pay values. We will calculate willingness-to-pay values for the whole sample and test for variations based on socio-demographic groups and referral to secondary care.

Integrated qualitative and quantitative process evaluation
This workstream will explore the implementation, adherence to protocol, receipt and setting of the intervention. We will examine the views of all groups of participants on the intervention; study how the intervention is implemented; investigate contextual factors that affect the intervention; and study how effects vary in subgroups of GPs. These data will help in understanding how, for whom and why the trial had effects and the extent to which outcomes result from issues of trial fidelity and implementation. We will collect process data from all 152 sites including clinician and carer outcomes. We will conduct alongside the trial non-participant observation and semi-structured interviews. We will include a purposive sample of 10 practices for the qualitative study, interviewing 4-5 participants in each (e.g. GP, carers, hospital consultant), resulting in 40-50 interviews. This sample will include a small number of practices in the control arm (for comparative purposes), and a range of intervention practices to include different locations, practice sizes and types. We will analyse process data before outcome data to avoid bias in interpretation. 29 Interviews will be audio- recorded and transcribed, data from observations will be recorded contemporaneously using a template. Data from process outcomes, interviews (transcripts) and observations will be analysed from the perspective of both behaviour change theory 30 and normalization process theory. 31

Strategies to mitigate potential bias
Since this effectiveness trial will test whether the intervention can work under usual circumstances, we will rely on paediatric orthopaedic surgeons in determining the ultimate diagnosis of DDH.
Variations in the surgeons' diagnostic accuracy are inevitable hence the need for a randomised study.
We will perform analyses by surgeon (or hospital) to quantify this variation. Blinding of GPs, practice staff, carers is impossible; however, most such outcomes will be assessed with validated questionnaires. Primary and principal secondary endpoints will be collected by an independent researcher blinded to treatment allocation. In case an infant is referred to hip ultrasound without an orthopaedic consultation, a trial-appointed advisory panel shall review the scan blinded and according to standard methods 32 to avoid reporting bias. There is a risk for verification bias -while our trial includes a 2-year follow up to capture late presenting DDH, we cannot rule out that some infants with DDH will remain undiagnosed within this period, thus underestimating the number of late diagnosed DDH. However, the 2-year mark has previously been found to be a robust outcome. 33

Patient and public involvement
We developed this protocol with carers of children with DDH and the founding director of 'Steps', a charity supporting patients with lower limb disorders. We discussed the need for the trial and trial procedures and conduct with staff members of GP practices. Our established patient and public involvement group has reviewed and commented on this protocol and will periodically review, support and advise on the conduct of the trial.

Ethics and dissemination
Protocol version 2.0 (12/11/2019) received approval from the Leicester Central Research Ethics Committee (19/EM/0317) and Health Research Authority Confidentiality Advisory Group (19/CAG/0198). It is registered at clinicaltrials.gov (NCT04101903). The results will be published in peer-reviewed academic journals, disseminated to patient organisations and the media.

Discussion
This randomised clinical trial is part of a programme of research to improve the diagnosis of DDH: consensus-based diagnostic criteria were established in prior research, tailored for use in primary care in form of a 'checklist', supplemented by a video specifically designed to explain to GPs how to use the tool at bedside. There has only been one randomised clinical trial on the topic of DDH in the UK 33 but it explored the use of ultrasound screening -our trial focusses on the compulsory clinical assessment conducted by GPs. Because the intervention tested in this trial is based on consensus of clinical experts, there is a risk that the opinions of experts change as clinical knowledge evolves.

Funder
The funding body had or has no involvement in study design; collection, management, analysis and interpretation of data; or the decision to submit for publication. The funding body will be informed of any planned publications, and documentation provided.

Sponsor
The sponsor for this trial is Great Ormond Street Hospital for Children. The sponsor is responsible for providing the investigator with the necessary information to conduct the clinical trial, to ensure proper monitoring of the trial and ensuring compliance to ethical bodies and legislation. The sponsor works to the UK Policy Framework for Health and Social Care Research. The sponsor is not involved in aspects of study design, report writing or data analysis. They are the data controller and all data shall return to the sponsor at the end of the trial. A collaboration agreement is in place with all organisations of the co-investigators. Data processing agreements are in place for situations where data will be collected and processed outside of Great Ormond Street Hospital. The sponsor can be contacted by email (research.governance@gosh.nhs.uk) or telephone 0207 905 2249.

Coordinating centre
PRIMENT Clinical Trials Unit is coordinating this trial. A trial management group has been set up within PRIMENT for the monthly monitoring of the trial conduct. It includes the chief investigator, director of the trials unit, programme manager and trial manager. PRIMENT are responsible for overseeing the conduct and progress of the trial.

Steering committee
The steering committee includes an independent chair, two further independent members (one is a biostatistician), sponsor representative, funder representative, chief investigator and two further members of the research team including. The committee will provide overall

Data monitoring committee
The steering committee will take on the role of the data monitoring committee.

Data Access
All identifiable data will be stored on encrypted servers, UCL Data Safe Haven. Only restricted members of the research team will be able to access this data. De-identified data will be shared with the wider members of the research team.

Instructions to authors
Complete this checklist by entering the page numbers from your manuscript where readers will find each of the items listed below.
Your article may not currently address all the items on the checklist. Please modify your text to include the missing information. If you are certain that an item does not apply, please write "n/a" and provide a short explanation.
Upload your completed checklist as an extra file when you submit to a journal.
In your methods section, say that you used the SPIRITreporting guidelines, and cite them as: international Delphi consensus, and a 9-item checklist that has the potential to enable non-experts to diagnose DDH in a manner approaching that of experts.

Methods and Analysis:
We will conduct a controlled trial randomised by practice that will compare a diagnostic aid against standard care for the hip check. The primary objective is to determine whether an aid to the diagnosis of DDH reduces the number of clinically insignificant referrals from primary care to hospital, and the number of late diagnosed DDH. The trial will include a qualitative process evaluation, an assessment of professional behaviour change and a full health economic evaluation. We will recruit 152 GP practices in England. These will be randomised to conduct the hip checks with use of the study diagnostic aid and/or as per usual practice. The total number of infants seen during a 15-month recruitment period will be 110 per practice. Two years after the 6week hip check we will measure the number of referred infants that are (1) clinically insignificant for DDH and (2) those that constitute appropriate referrals.

Ethics and Dissemination:
This study has approval from the Health Research Authority and the Confidentiality Advisory Group. Results will be published in peer-reviewed academic journals, disseminated to patient organisations and the media.

STRENGTHS AND LIMITATIONS OF THIS STUDY
 This is the first trial to evaluate a diagnostic aid for the 6-week check with reference to evaluating both missed and unnecessary referrals to hospital.
 Implementable on existing clinical software used by GPs, the proposed aid will be easy to use.
 A comprehensive process evaluation, using qualitative methods and behaviour change frameworks, will be done alongside to the trial; plus a full health economic evaluation.
 The reliance on existing practice staff to report monthly updates is essential but could pose a risk to timely and complete data collection.
 The collection of identifiable data through site visits across 152 practices will be challenging.

INTRODUCTION
Developmental dysplasia of the hip (DDH) is characterized by varying displacement of the proximal femur from the acetabulum with associated acetabular dysplasia. Dislocation occurs in 1-2/1000 infants per year but milder forms occur in 40-60/1000. 1 Early recognition of disease is associated with better outcomes. It is national policy 2 to examine all infants for the presence of DDH at birth and at age 6-8 weeks in primary care (6-week hip check). If diagnosed within the first 6-8 weeks, splinting of the hips is successful in 85% of cases. 3 Later diagnoses usually will require invasive treatment, with many years of continued monitoring. 4 Late diagnosed DDH is a common cause of medical negligence claims, with increased suffering for affected patients.
Timely diagnoses of DDH remain a challenge despite the compulsory 6-week check. 5 In one study, the median age at diagnosis was 14 months, with only 40% of infants diagnosed during routine examinations and 60% presenting owing to parental concerns. 6 In another study, 30% of infants were not diagnosed by 12 weeks. 7 Because there is no further compulsory check after that at 6-weeks, it is vital that the 6-week check is effective. Also infants incorrectly identified with "DDH" in primary care and referred to hospital remain a challenge for both families and NHS. This group does not  8 e.g. "crease asymmetry" (n=234) or "click" (n=648). If GPs were able to discriminate better between benign abnormalities requiring reassurance (perhaps with a follow-up in primary care) and findings requiring referrals, outcomes of infants with and without DDH would improve.
In prior research we suggested that too many diagnostic criteria 8 and variability in the use of diagnostic criteria among clinicians 9 complicate the task of diagnosing DDH. We developed standardised diagnostic criteria to reduce the variability in assessment and management decisions in infants examined for DDH. 10 The weighted criteria demonstrated validity. 10 In a study of 44 patients referred from GPs to hospital, the weighted criteria demonstrated a positive predictive value of 89% (95% confidence interval 70, 97) and negative predictive value of 76% (95% confidence interval, 50-96). 11 We refined these binary criteria, in form of a checklist with 9 items, for use by GPs during the 6-week hip check and developed a training video, featuring a GP. This checklist and video were refined in a feasibility study which used qualitative methods to explore the acceptability of the format, style and delivery of the intervention (report available on request). The video explains the meaning of the diagnostic criteria; it also demonstrates how precisely to elicit the diagnostic criteria.
For example, the video explains the difference between the Barlow and Ortolani manoeuvres, how to test for a leg length inequality, or how to identify limitations in hip abduction.

Rationale
Current referral patterns suggest there could be considerable health gains from improved diagnostic and referral decisions at the 6-week check. 5 GPs, in a preparatory focus group, identified the need for a 'comprehensive, structured guide' for the 6-week check. Diagnostic aids enable physicians to overcome barriers in diagnostic reasoning 12 by shifting intuitive to analytical aspects of diagnostic reasoning. Decisions made under these circumstances approach normative reasoning and rationality  To determine the cost-effectiveness of this intervention.
 To conduct an integrated qualitative and quantitative process evaluation in order to understand all participants' experience in the trial and with the intervention; study how the intervention is implemented; investigate contextual factors that affect the intervention.

METHODS AND ANALYSIS
The SPIRIT reporting guidelines were used in the preparation of this clinical trial protocol. 14

Design
This is a cluster randomised controlled trial of GP practices in England, with randomisation and intervention at practice level but the primary endpoint measured at patient level ( Figure 1). An internal pilot will be done in the first 4 months to ascertain accrual rates -we will progress to the main trial if we succeed in recruiting ≥12 practices per month (expected recruitment is ≥21 practices/month for the main trial). Incorporated in the trial are (i) a process evaluation investigating determinants of GPs' referral behaviour and the implementation of the intervention in practice; and (ii) a health economic evaluation in a child's lifetime. to being randomised and (ii) to hospitals releasing data concerning infants they examine in the trial.

Eligibility and recruitment
Ineligible are practices planning to close with 12 months of trial start.
We anticipate that ≥110 infants per year will undergo the 6-week hip check in each of the recruited practices. Each practice will recruit infants for a period of 15 months and infants will be followed for 2 years. This trial is planned therefore to run for 44 months across 152 GP practices.

Retention
The proposed trial is not onerous on the practice staff and practices will not be challenged by excessive study-related manoeuvres. Because the primary outcome will be collected using practice and hospital databases only, it is likely that we will attain high levels of follow up. We will send periodic newsletters to GPs and foster regular contacts.

Randomisation
Practices will submit an eligibility form to the clinical trials unit, which will register and randomly allocate (concealed) practices within 48 hours. An independent statistician will coordinate randomisation with an allocation ratio of 1:1 and stratified by practice size (based on observed sizes obtained from expressions of interest -we expect 2 or 3 strata).

Interventions
Control intervention: GPs will assess infant hips following best practice principles; they will be provide a leaflet about the national best practice recommendations. 2 Experimental intervention: This is a complex intervention comprising a video developed specifically for this trial and a diagnostic aid in form of a 9-item binary checklist. At the time when informed consent is taken, GPs will watch the video. They will then examine all infants in the 6-week check according to best practice principles, but with the addition of the diagnostic aid (implemented electronically to each practice's computer system). relevant ICD-10 codes and OPCS-4 codes. 18 We will extract, for the whole trial period, the full hospital history for all infants in the trial (in and out patient episodes), collate all episodes into a combined exploratory analysis view, collate interactions, and allocate outcomes to practice groups to facilitate outcome analyses. 19 These data will be compared with the data collected by the researcher (see above); the use of these 2 strategies will enhance the robustness of data.
• Consequences of late diagnoses: Using data collected above we will record nature and frequency of such consequences (nature of treatment, length of hospital stay, frequency of secondary care contacts).
• Health-related quality of life (parental proxy report): Child-Health-Utility-9D. 20 Process level • Volume of referrals: Total number of patients referred to secondary care during the trial period.
This variable will be collected prospectively at practice level by the participating GP and forwarded monthly to the clinical trial unit.
• Timeliness of referrals: In infants referred to secondary care, we will measure the days from referral issued to hospital appointment (collected in the same way as primary endpoint). This process measure will inform about target wait times 2 achieved.

Clinician level
• Confidence and attitudes of GP and secondary care clinicians towards the diagnostic aid: will be assessed 12 weeks after trial set-up and at trial completion using a modified measure based on the Theory of Planned Behaviour. 21 • Implementation issues and acceptability of diagnostic aid among GPs and secondary care providers: we will conduct qualitative research at the trial end to elicit this information.
• Use of diagnostic aid and acceptability of intervention: a self-administered questionnaire to evaluate the use of the aid will be posted to all the GPs in the intervention group; also direct observations and interviews will ascertain this outcome.

Statistical analysis
Primary endpoint: We will compare the randomised practices using Poisson mixed models, accounting for extra Poissonian variability by including random intercept terms for practices. The response variable will be the number of clinically insignificant referrals from each practice with an offset in the linear model of the log(e) total number of children checked in each practice, to account for differences in practice size and constitution. The random effect at the practice level will account for overdispersion.
Principal secondary endpoint: This will be analysed as above; however, counts here are expected to be lower. Should they be insufficient to support the full model, the model will be simplified by (e.g. use of additive variance component term rather than a generalised random intercept term to address overdispersion). We will describe the planned statistical analysis and its adaptation where there are difficulties in achieving convergence in a pre-specified statistical analysis plan.
Late diagnoses: As we anticipate small numbers of events in each randomised comparison we may not be in a position to account for practices using random intercept terms. Where this is the case we will report the total numbers over a minimum 2-year observation and compare the overall group scores using Fishers exact test.
Quantitative outcome measures: We will summarize scores by instrument, accounting for practices and report difference in group means for each treatment arm. We expect mean scores to be lower in the intervention arm for State-Trait Anxiety Inventory and Infant Hip Worries Inventory, but higher or equal for EUROPEP.
Missing data: For the primary endpoint and principal secondary endpoint the data collection methods should identify qualifying episodes. Because of the nature of these data the conventional concept of missingness does not directly apply (eg we will not have individually randomised subjects who cannot be followed up). However if a practice withdraws from the trial we will explore the consequences of this action by assuming a poor outcome among that practice if in the intervention group and a good outcome if in the control condition, to identify the potential consequences of their withdrawal. Complete case analyses will be conducted for secondary outcomes. If there is a mismatch between practices in the two treatment conditions we will consider undertaking joint models examining simultaneously the binomial of missingness and the outcome measure of interest.

Cost analysis
We will analyse the cost associated with the intervention compared with usual practice for the entire trial period, and examine costs from the perspective of the NHS and of families. The cost components included in main analysis are: cost of 6-week in both intervention arms; any subsequent referrals, diagnostic tests and treatment. We will collect costs about GP time, which we will multiply by unit costs from routine sources. 25

Within-trial cost-effectiveness analysis
With the costs described above we will produce a dataset of patient-level within-trial costs and outcomes. We will calculate the incremental cost per clinically insignificant referral avoided and the incremental cost per late diagnosis avoided. Using bootstrapping of the mean cost and outcomes differences, we will estimate confidence intervals around the incremental cost-effectiveness ratios. 26 With the bootstrap replications we will construct a cost-effectiveness acceptability curve to show the probability that the aid is cost-effective for different values of NHS willingness to pay for outcomes.
We will perform deterministic sensitivity analyses.

Long-run cost utility analysis
We will use several measures to evaluate the lifetime cost-effectiveness of the intervention. We will ask 20 carers of infants aged 2-4 years to complete (1) on behalf of their children the Child-Healthutility-9D 20 and (2) for themselves the EQ-5D-5L, 27 both measure health-related quality of life. With data from the trial about the impact of the intervention on appropriate referrals, we will calculate the monetary value that parents place on the intervention using willingness-to-pay methodology. 28 This will provide an estimate of the monetary value of the additional benefits (positive/negative) of the intervention. We will calculate the net benefit of the intervention by subtracting the incremental cost of the aid, as calculated above from the trial data, from the monetary value of the additional benefit. Following recruitment of the last infant in the trial, we will recruit 200 carers of infants undergoing the 6-week check from trial-participating practices. These will be 100 carers whose infants will be referred to secondary care as a consequence of the 6-week check and 100 who will not. They will complete a self-report questionnaire that utilised several techniques 28 to elicit willingness-to-pay values. We will calculate willingness-to-pay values for the whole sample and test for variations based on socio-demographic groups and referral to secondary care.

Integrated qualitative and quantitative process evaluation
This workstream will explore the implementation, adherence to protocol, receipt and setting of the intervention. We will examine the views of all groups of participants on the intervention; study how the intervention is implemented; investigate contextual factors that affect the intervention; and study how effects vary in subgroups of GPs. These data will help in understanding how, for whom and why the trial had effects and the extent to which outcomes result from issues of trial fidelity and implementation. We will collect process data from all 152 sites including clinician and carer outcomes. We will conduct alongside the trial non-participant observation and semi-structured interviews. We will include a purposive sample of 10 practices for the qualitative study, interviewing 4-5 participants in each (e.g. GP, carers, hospital consultant), resulting in 40-50 interviews. This sample will include a small number of practices in the control arm (for comparative purposes), and a range of intervention practices to include different locations, practice sizes and types. We will analyse process data before outcome data to avoid bias in interpretation. 29 Interviews will be audio- recorded and transcribed, data from observations will be recorded contemporaneously using a template. Data from process outcomes, interviews (transcripts) and observations will be analysed from the perspective of both behaviour change theory 30 and normalization process theory. 31

Strategies to mitigate potential bias
Since this effectiveness trial will test whether the intervention can work under usual circumstances, we will rely on paediatric orthopaedic surgeons in determining the ultimate diagnosis of DDH.
Variations in the surgeons' diagnostic accuracy are inevitable hence the need for a randomised study.
We will perform analyses by surgeon (or hospital) to quantify this variation. Blinding of GPs, practice staff, carers is impossible; however, most such outcomes will be assessed with validated questionnaires. Primary and principal secondary endpoints will be collected by an independent researcher blinded to treatment allocation. In case an infant is referred to hip ultrasound without an orthopaedic consultation, a trial-appointed advisory panel shall review the scan blinded and according to standard methods 32 to avoid reporting bias. There is a risk for verification bias -while our trial includes a 2-year follow up to capture late presenting DDH, we cannot rule out that some infants with DDH will remain undiagnosed within this period, thus underestimating the number of late diagnosed DDH. However, the 2-year mark has previously been found to be a robust outcome. 33

Patient and public involvement
We developed this protocol with carers of children with DDH and the founding director of 'Steps', a charity supporting patients with lower limb disorders. We discussed the need for the trial and trial procedures and conduct with staff members of GP practices. Our established patient and public involvement group has reviewed and commented on this protocol and will periodically review, support and advise on the conduct of the trial.

Trial and data management
The trial will be run through PRIMENT Clinical Trials Unit and conducted in accordance with established quality management systems and standardised operating procedures (Appendix 1). All 16 data will be handled in accordance with the UK Data Protection Act 2018. All analyses will be conducted blinded to allocation groups.

Ethics and dissemination
Leicester Central Research Ethics Committee (19/EM/0317) and Health Research Authority Confidentiality Advisory Group (19/CAG/0198) approved protocol version 2.0 (12/11/2020). The latter granted 'Section 251' approval as operating a consented model was unfeasible within the clinical setting (written informed consent will be obtained at cluster level from lead GPs). It is registered at clinicaltrials.gov (NCT04101903). The results will be published in peer-reviewed academic journals, disseminated to patient organisations and the media.

Discussion
This trial is part of a programme of research to improve the diagnosis of DDH: consensus-based diagnostic criteria were established in prior research, tailored for use in primary care, supplemented by a video designed for GPs. There has only been one randomised trial on the topic of DDH in the UK 33 but it explored the use of ultrasound screening -our trial focusses on the compulsory '6-week check'. Because the intervention tested in this trial is based on consensus of clinical experts, there is a risk that the opinions of experts change as clinical knowledge evolves. However, the criteria of the diagnostic aid have been in use for decades and will likely not loose relevance in the foreseeable future. The collection of outcome data from various hospitals connected to GP practices will be challenging; use of national health services databases should mitigate this challenge. While our trial includes a 2-year followup period to capture late presenting DDH, we cannot rule out that some infants with DDH will remain undiagnosed within this period. This trial has the potential to improve the compulsory 6-week hip check with use of a relatively simple intervention. It will also provide an understanding of the cost effectiveness of the intervention in a whole lifetime horizon of a 6-week old. If successful, the intervention can be rolled out to clinical services relatively easily and at low costs.  (2) Public Health England. Newborn and infant physical examination (NIPE) sceening handbook. www.gov.uk/government/publications/newborn-and-infant-physical-examinationprogramme-handbook/newborn-and-infant-physical-examination-screening-programmehandbook . 27-8-2019.

Funder
The funding body had or has no involvement in study design; collection, management, analysis and interpretation of data; or the decision to submit for publication. The funding body will be informed of any planned publications, and documentation provided.

Sponsor
The sponsor for this trial is Great Ormond Street Hospital for Children. The sponsor is

Steering committee
The steering committee includes an independent chair, two further independent members (one is a biostatistician), sponsor representative, funder representative, chief investigator and two further members of the research team including. The committee will provide overall

Data monitoring committee
The steering committee will take on the role of the data monitoring committee.

Data Access
All identifiable data will be stored on encrypted servers, UCL Data Safe Haven. Only restricted members of the research team will be able to access this data. De-identified data will be shared with the wider members of the research team.

Instructions to authors
Complete this checklist by entering the page numbers from your manuscript where readers will find each of the items listed below.
Your article may not currently address all the items on the checklist. Please modify your text to include the missing information. If you are certain that an item does not apply, please write "n/a" and provide a short explanation.
Upload your completed checklist as an extra file when you submit to a journal.

Methods and Analysis:
We will conduct a controlled trial randomised by practice that will compare a diagnostic aid against standard care for the hip check. The primary objective is to determine whether an aid to the diagnosis of DDH reduces the number of clinically insignificant referrals from primary care to hospital, and the number of late diagnosed DDH. The trial will include a qualitative process evaluation, an assessment of professional behaviour change and a full health economic evaluation. We will recruit 152 GP practices in England. These will be randomised to conduct the hip checks with use of the study diagnostic aid and/or as per usual practice. The total number of infants seen during a 15-month recruitment period will be 110 per practice. Two years after the 6week hip check we will measure the number of referred infants that are (1) clinically insignificant for DDH and (2) those that constitute appropriate referrals.

Ethics and Dissemination:
This study has approval from the Health Research Authority (16/1/2020) and the Confidentiality Advisory Group (18/2/2020). Results will be published in peer-reviewed academic journals, disseminated to patient organisations and the media.

STRENGTHS AND LIMITATIONS OF THIS STUDY
 To our knowledge this is the first trial to evaluate a diagnostic aid for the 6-week check with reference to evaluating both missed and unnecessary referrals to hospital.
 Implementable on existing clinical software used by GPs, the proposed aid will be easy to use.
 A comprehensive process evaluation, using qualitative methods and behaviour change frameworks, will be done alongside to the trial; plus a full health economic evaluation.
 The reliance on existing practice staff to report monthly updates is essential but could pose a risk to timely and complete data collection.
 The collection of identifiable data through site visits across 152 practices will be challenging.

INTRODUCTION
Developmental dysplasia of the hip (DDH) is characterized by varying displacement of the proximal femur from the acetabulum with associated acetabular dysplasia. Dislocation occurs in 1-2/1000 infants per year but milder forms occur in 40-60/1000. 1 Early recognition of disease is associated with better outcomes. It is national policy 2 to examine all infants for the presence of DDH at birth and at age 6-8 weeks in primary care (6-week hip check). If diagnosed within the first 6-8 weeks, splinting of the hips is successful in 85% of cases. 3 Later diagnoses usually will require invasive treatment, with many years of continued monitoring. 4 Late diagnosed DDH is a common cause of medical negligence claims, with increased suffering for affected patients.
Timely diagnoses of DDH remain a challenge despite the compulsory 6-week check. 5 In one study, the median age at diagnosis was 14 months, with only 40% of infants diagnosed during routine examinations and 60% presenting owing to parental concerns. 6 In another study, 30% of infants were not diagnosed by 12 weeks. 7 Because there is no further compulsory check after that at 6-weeks, it is vital that the 6-week check is effective. Also infants incorrectly identified with "DDH" in primary care and referred to hospital remain a challenge for both families and NHS. This group does not  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   4 require treatment and reassurance provided in primary care would avoid unnecessary anxiety and inefficient use of hospital resources. Of 1,918 infants referred to hospital from primary care for DDH, only 64 (3%) had DDH 7 but 1,270 (66%) were identified as 'DDH' based on inappropriate criteria, 8 e.g. "crease asymmetry" (n=234) or "click" (n=648). If GPs were able to discriminate better between benign abnormalities requiring reassurance (perhaps with a follow-up in primary care) and findings requiring referrals, outcomes of infants with and without DDH would improve.
In prior research we suggested that too many diagnostic criteria 8 and variability in the use of diagnostic criteria among clinicians 9 complicate the task of diagnosing DDH. We developed standardised diagnostic criteria to reduce the variability in assessment and management decisions in infants examined for DDH. 10 The weighted criteria demonstrated validity. 10 In a study of 44 patients referred from GPs to hospital, the weighted criteria demonstrated a positive predictive value of 89% (95% confidence interval 70, 97) and negative predictive value of 76% (95% confidence interval, 50-96). 11 We refined these binary criteria, in form of a checklist with 9 items, for use by GPs during the 6-week hip check and developed a training video, featuring a GP. This checklist and video were refined in a feasibility study which used qualitative methods to explore the acceptability of the format, style and delivery of the intervention (report available on request). The video explains the meaning of the diagnostic criteria; it also demonstrates how precisely to elicit the diagnostic criteria.
For example, the video explains the difference between the Barlow and Ortolani manoeuvres, how to test for a leg length inequality, or how to identify limitations in hip abduction.

Rationale
Current referral patterns suggest there could be considerable health gains from improved diagnostic and referral decisions at the 6-week check. 5 GPs, in a preparatory focus group, identified the need for a 'comprehensive, structured guide' for the 6-week check. Diagnostic aids enable physicians to overcome barriers in diagnostic reasoning 12 by shifting intuitive to analytical aspects of diagnostic reasoning. Decisions made under these circumstances approach normative reasoning and rationality  To determine the cost-effectiveness of this intervention.
 To conduct an integrated qualitative and quantitative process evaluation in order to understand all participants' experience in the trial and with the intervention; study how the intervention is implemented; investigate contextual factors that affect the intervention.

METHODS AND ANALYSIS
The SPIRIT reporting guidelines were used in the preparation of this clinical trial protocol. 14

Design
This is a cluster randomised controlled trial of GP practices in England, with randomisation and intervention at practice level but the primary endpoint measured at patient level ( Figure 1). An internal pilot will be done in the first 4 months to ascertain accrual rates -we will progress to the main trial if we succeed in recruiting ≥12 practices per month (expected recruitment is ≥21 practices/month for the main trial). Incorporated in the trial are (i) a process evaluation investigating determinants of GPs' referral behaviour and the implementation of the intervention in practice; and (ii) a health economic evaluation in a child's lifetime.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   6 GP practices registered in England that carry out 6-week hip checks (at age 42-76 days) and agree (i) to being randomised and (ii) to hospitals releasing data concerning infants they examine in the trial.

Eligibility and recruitment
Ineligible are practices planning to close with 12 months of trial start.
We anticipate that ≥110 infants per year will undergo the 6-week hip check in each of the recruited practices. Each practice will recruit infants for a period of 15 months and infants will be followed for 2 years. This trial is planned therefore to run for 44 months across 152 GP practices.

Retention
The proposed trial is not onerous on the practice staff and practices will not be challenged by excessive study-related manoeuvres. Because the primary outcome will be collected using practice and hospital databases only, it is likely that we will attain high levels of follow up. We will send periodic newsletters to GPs and foster regular contacts.

Randomisation
Practices will submit an eligibility form to the clinical trials unit, which will register and randomly allocate (concealed) practices within 48 hours. An independent statistician will coordinate randomisation with an allocation ratio of 1:1 and stratified by practice size (based on observed sizes obtained from expressions of interest -we expect 2 or 3 strata).

Interventions
Control intervention: GPs will assess infant hips following best practice principles; they will be provide a leaflet about the national best practice recommendations. 2 Experimental intervention: This is a complex intervention comprising a video developed specifically for this trial and a diagnostic aid in form of a 9-item binary checklist. At the time when informed consent is taken, GPs will watch the video. They will then examine all infants in the 6-week check according to best practice principles, but with the addition of the diagnostic aid (implemented electronically to each practice's computer system). According to national guidelines, the 6-week check shall capture infants at age 6-10 weeks and infants in whom a diagnosis of DDH is considered should see a specialist within 2 weeks. 2 Participating GPs of both intervention arms will be asked to adhere to this policy.

Definitions for the purpose of this trial
'Appropriate referral' denotes a referred hip deemed 'clinically significant', i.e. is treated or monitored (at least one follow-up) by a specialist surgeon. Any ambiguous cases will be reviewed by our Expert relevant ICD-10 codes and OPCS-4 codes. 18 We will extract, for the whole trial period, the full hospital history for all infants in the trial (in and out patient episodes), collate all episodes into a combined exploratory analysis view, collate interactions, and allocate outcomes to practice groups to facilitate outcome analyses. 19 These data will be compared with the data collected by the researcher (see above); the use of these 2 strategies will enhance the robustness of data.
• Consequences of late diagnoses: Using data collected above we will record nature and frequency of such consequences (nature of treatment, length of hospital stay, frequency of secondary care contacts).
• Health-related quality of life (parental proxy report): Child-Health-Utility-9D. 20 Process level • Volume of referrals: Total number of patients referred to secondary care during the trial period.
This variable will be collected prospectively at practice level by the participating GP and forwarded monthly to the clinical trial unit.
• Timeliness of referrals: In infants referred to secondary care, we will measure the days from referral issued to hospital appointment (collected in the same way as primary endpoint). This process measure will inform about target wait times 2 achieved.

Clinician level
• Confidence and attitudes of GP and secondary care clinicians towards the diagnostic aid: will be assessed 12 weeks after trial set-up and at trial completion using a modified measure based on the Theory of Planned Behaviour. 21 • Implementation issues and acceptability of diagnostic aid among GPs and secondary care providers: we will conduct qualitative research at the trial end to elicit this information.
• Use of diagnostic aid and acceptability of intervention: a self-administered questionnaire to evaluate the use of the aid will be posted to all the GPs in the intervention group; also direct observations and interviews will ascertain this outcome.

Statistical analysis
Primary endpoint: We will compare the randomised practices using Poisson mixed models, accounting for extra Poissonian variability by including random intercept terms for practices. The response variable will be the number of clinically insignificant referrals from each practice with an offset in the linear model of the log(e) total number of children checked in each practice, to account for differences in practice size and constitution. The random effect at the practice level will account for overdispersion.
Principal secondary endpoint: This will be analysed as above; however, counts here are expected to be lower. Should they be insufficient to support the full model, the model will be simplified by (e.g. use of additive variance component term rather than a generalised random intercept term to address overdispersion). We will describe the planned statistical analysis and its adaptation where there are difficulties in achieving convergence in a pre-specified statistical analysis plan. No interim analyses are planned.
Late diagnoses: As we anticipate small numbers of events in each randomised comparison we may not be in a position to account for practices using random intercept terms. Where this is the case we will report the total numbers over a minimum 2-year observation and compare the overall group scores using Fishers exact test.
Quantitative outcome measures: We will summarize scores by instrument, accounting for practices and report difference in group means for each treatment arm. We expect mean scores to be lower in the intervention arm for State-Trait Anxiety Inventory and Infant Hip Worries Inventory, but higher or equal for EUROPEP.
Missing data: For the primary endpoint and principal secondary endpoint the data collection methods should identify qualifying episodes. Because of the nature of these data the conventional concept of missingness does not directly apply (eg we will not have individually randomised subjects who cannot be followed up). However if a practice withdraws from the trial we will explore the group and a good outcome if in the control condition, to identify the potential consequences of their withdrawal. Complete case analyses will be conducted for secondary outcomes. If there is a mismatch between practices in the two treatment conditions we will consider undertaking joint models examining simultaneously the binomial of missingness and the outcome measure of interest.

Cost analysis
We will analyse the cost associated with the intervention compared with usual practice for the entire trial period, and examine costs from the perspective of the NHS and of families. The cost components included in main analysis are: cost of 6-week in both intervention arms; any subsequent referrals, diagnostic tests and treatment. We will collect costs about GP time, which we will multiply by unit costs from routine sources. 25

Within-trial cost-effectiveness analysis
With the costs described above we will produce a dataset of patient-level within-trial costs and outcomes. We will calculate the incremental cost per clinically insignificant referral avoided and the incremental cost per late diagnosis avoided. Using bootstrapping of the mean cost and outcomes differences, we will estimate confidence intervals around the incremental cost-effectiveness ratios. 26 With the bootstrap replications we will construct a cost-effectiveness acceptability curve to show the probability that the aid is cost-effective for different values of NHS willingness to pay for outcomes.
We will perform deterministic sensitivity analyses.

Long-run cost utility analysis
We will use several measures to evaluate the lifetime cost-effectiveness of the intervention. We will ask 20 carers of infants aged 2-4 years to complete (1) on behalf of their children the Child-Healthutility-9D 20 and (2) for themselves the EQ-5D-5L, 27 both measure health-related quality of life. With data from the trial about the impact of the intervention on appropriate referrals, we will calculate the monetary value that parents place on the intervention using willingness-to-pay methodology. 28 This will provide an estimate of the monetary value of the additional benefits (positive/negative) of the intervention. We will calculate the net benefit of the intervention by subtracting the incremental cost of the aid, as calculated above from the trial data, from the monetary value of the additional benefit. Following recruitment of the last infant in the trial, we will recruit 200 carers of infants undergoing the 6-week check from trial-participating practices. These will be 100 carers whose infants will be referred to secondary care as a consequence of the 6-week check and 100 who will not. They will complete a self-report questionnaire that utilised several techniques 28 to elicit willingness-to-pay values. We will calculate willingness-to-pay values for the whole sample and test for variations based on socio-demographic groups and referral to secondary care.

Integrated qualitative and quantitative process evaluation
This workstream will explore the implementation, adherence to protocol, receipt and setting of the intervention. We will examine the views of all groups of participants on the intervention; study how the intervention is implemented; investigate contextual factors that affect the intervention; and study how effects vary in subgroups of GPs. These data will help in understanding how, for whom and why the trial had effects and the extent to which outcomes result from issues of trial fidelity and implementation. We will collect process data from all 152 sites including clinician and carer outcomes. We will conduct alongside the trial non-participant observation and semi-structured interviews. We will include a purposive sample of 10 practices for the qualitative study, interviewing 4-5 participants in each (e.g. GP, carers, hospital consultant), resulting in 40-50 interviews. This sample will include a small number of practices in the control arm (for comparative purposes), and a range of intervention practices to include different locations, practice sizes and types. We will analyse process data before outcome data to avoid bias in interpretation. 29 Interviews will be audio-  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   15 recorded and transcribed, data from observations will be recorded contemporaneously using a template. Data from process outcomes, interviews (transcripts) and observations will be analysed from the perspective of both behaviour change theory 30 and normalization process theory. 31

Strategies to mitigate potential bias
Since this effectiveness trial will test whether the intervention can work under usual circumstances, we will rely on paediatric orthopaedic surgeons in determining the ultimate diagnosis of DDH.
Variations in the surgeons' diagnostic accuracy are inevitable hence the need for a randomised study.
We will perform analyses by surgeon (or hospital) to quantify this variation. Blinding of GPs, practice staff, carers is impossible; however, most such outcomes will be assessed with validated questionnaires. Primary and principal secondary endpoints will be collected by an independent researcher blinded to treatment allocation. In case an infant is referred to hip ultrasound without an orthopaedic consultation, a trial-appointed advisory panel shall review the scan blinded and according to standard methods 32 to avoid reporting bias. There is a risk for verification bias -while our trial includes a 2-year follow up to capture late presenting DDH, we cannot rule out that some infants with DDH will remain undiagnosed within this period, thus underestimating the number of late diagnosed DDH. However, the 2-year mark has previously been found to be a robust outcome. 33

Patient and public involvement
We developed this protocol with carers of children with DDH and the founding director of 'Steps', a charity supporting patients with lower limb disorders. We discussed the need for the trial and trial procedures and conduct with staff members of GP practices. Our established patient and public involvement group has reviewed and commented on this protocol and will periodically review, support and advise on the conduct of the trial.
'Section 251' approval was obtained, which omits the need for written informed consent from parents/carers; consent will be obtained at cluster level from the lead GPs. This trial is registered at clinicaltrials.gov (NCT04101903). We will publish results in peer-reviewed journals and disseminate results to patient organisations and the media.

Funder
The funding body had or has no involvement in study design; collection, management, analysis and interpretation of data; or the decision to submit for publication. The funding body will be informed of any planned publications, and documentation provided.

Sponsor
The sponsor for this trial is Great Ormond Street Hospital for Children. The sponsor is responsible for providing the investigator with the necessary information to conduct the clinical trial, to ensure proper monitoring of the trial and ensuring compliance to ethical bodies and legislation. The sponsor works to the UK Policy Framework for Health and Social Care Research. The sponsor is not involved in aspects of study design, report writing or data analysis. They are the data controller and all data shall return to the sponsor at the end of the trial. A collaboration agreement is in place with all organisations of the co-investigators. Data processing agreements are in place for situations where data will be collected and processed outside of Great Ormond Street Hospital. The sponsor can be contacted by email (research.governance@gosh.nhs.uk) or telephone 0207 905 2249.

Coordinating centre
PRIMENT Clinical Trials Unit is coordinating this trial. A trial management group has been set up within PRIMENT for the monthly monitoring of the trial conduct. It includes the chief investigator, director of the trials unit, programme manager and trial manager. PRIMENT are responsible for overseeing the conduct and progress of the trial.

Data monitoring committee
The steering committee will take on the role of the data monitoring committee.

Instructions to authors
Complete this checklist by entering the page numbers from your manuscript where readers will find each of the items listed below.
Your article may not currently address all the items on the checklist. Please modify your text to include the missing information. If you are certain that an item does not apply, please write "n/a" and provide a short explanation.
Upload your completed checklist as an extra file when you submit to a journal.