Hospital use by babies enrolled in the PREVAIL randomised controlled trial

The University of Liverpool (UoL), the University of York (UoY) and University College London (UCL) jointly require HES data and demographic data (date of death) for use in the PREVAIL study.

The PREVAIL study is a randomised controlled trial to determine the clinical and cost-effectiveness of using antimicrobial and antifungal impregnated peripherally inserted central venous catheters (AM-PICC) in very preterm babies compared with standard-PICC. It is funded by the National Institute for Health Research - Health Technology Assessment (NIHR-HTA) programme

Recruitment started in August 2015 and finished on the 11th January 2017 following which 828 babies were recruited via their parent(s)/guardian(s) throughout England. The parents or guardians of the babies taking part in PREVAIL gave informed consent for the PREVAIL research team to collect information on their baby's routine records from the birth to up to 6 months following their inclusion in the study.

The trial has three Data Controllers. UCL, UoY and UoL collaborate in deciding the appropriate methods for analysis and on the interpretation of the results. The co-principal investigators are (respectively) an employee of UCL and an employee of Bradford Teaching Hospitals NHS Foundation Trust who is the honorary chair at the Hull York Medical School (a partnership between the University of York and the University of Hull). The UoL runs the PREVAIL trial and will conduct the effectiveness analysis. The UoY will conduct the health economic analysis. UCL, as co-PI, will oversee the study.

The UCL co-PI will only have access to results in the aggregate form, but will collaborate in the planning of the analyses. Given this involvement, UCL is a co-data controller although not data processor of the patient level data. The individual at UCL will only have access to aggregated results from the data processors within this agreement which may contain small numbers. Bradford Teaching Hospitals NHS Foundation Trust will have no input in the planning of the analyses but will assist the team in the interpretation of the results. The input of the Bradford Teaching Hospitals NHS Foundation Trust co-PI in the PREVAIL project is centred in the conceptualisation, supervision and design of the clinical trial, interpretation of the results and application to clinical practice. Bradford Teaching Hospitals NHS Foundation Trust will have no access to the data and will only have access to the aggregated results which are suppressed in line with the HES Analysis Guide. For these reasons, Bradford Teaching Hospitals NHS Foundation Trust is not a data controller nor a data processor.

The University of Liverpool (UoL) and University of York (UoY) are co-data processors because they will have access to the individual level data and conduct the analysis. The UoL will send the identifiers of the babies enrolled in PREVAIL to NHS Digital, since it runs the PREVAIL trial and stores the personal data. Neither UCL nor the UoY have access to this personal information on the PREVAIL babies.

The PREVAIL study has 3 major work streams:

1. Randomised controlled trial comparing AM-PICC with standard-PICC in reducing infections and associated complications during the stay at the neonatal care unit. This is led by the University of Liverpool (UoL).

2. Health economic analysis comparing AM-PICC with standard-PICC in improving health-related quality of life and life expectancy, and their impact on costs of health care. This is led by the University of York (UoY).

3. Generalisability analysis to understand the risk of infection across the various neonatal units. This is led by the University College London (UCL).

The PREVAIL study will use NHS Digital data to inform Workstream 1, namely about whether the type of PICC affects the risk of death, and Workstream 2, namely on the cost of hospital care and risk of death by type of PICC.

The UoL will evaluate whether AM-PICC avoids infections (primary outcome) and other adverse health outcomes, including risk of death and time to death, compared with standard-PICC. In order to compare risk and time to death over the 183 days of follow-up, the UoL requires information on the date of death over the 183 days of follow-up. It is necessary to identify the date of death to calculate the effect of AM-PICC vs standard-PICC on mortality using appropriate time to event methods. Comparing risk of death between groups using information on whether death occurred, without access to the exact date of death, will risk diluting any effect between the two groups, and increases the risk of bias.

The UoL will obtain the data on date of death from PDS and date of discharge (if method of discharge is death) from HES. Previous studies suggest that there are some inconsistencies in routine records. Having the two sources indicating date of death will allow the UoL to determine whether there are inconsistencies in records, and if inconsistencies are found, to present a sensitivity analysis on the effect of AM-PICC vs. S-PICC on the time to death under each data source. The purpose of this sensitivity analysis is to understand whether using a different data source has an impact on the effect of AM-PICC vs. S-PICC.

The UoY will conduct the health economics analysis. The health economic analysis will evaluate the costs to the NHS of using AM-PICC or standard-PICC, and compare those with the health outcomes, namely health-related quality of life and life expectancy, under the two options. The objective of the health economic analysis is to inform a decision of whether the NHS should invest in AM-PICC for all or some preterm babies.

Health economic evidence is important for the NHS so that clinicians can make an informed decision on whether a specific type of PICC is good value for money. The NHS runs a fixed budget, therefore if additional funds are invested in some interventions, other interventions cannot be funded. This means that an intervention is good value for money if it is better for health and cost saving, or it increases costs but this cost increase is offset by the gains in health, compared to other interventions that could have been funded.

The health economic analysis has five components: (a) comparison of health care costs between trial groups, (b) decision modelling to extrapolate the health outcomes and health care costs from the 183 days follow-up to the expected lifetime of the babies, (c) value of information analysis on the key sources of uncertainty, (d) value of implementing the cost-effective intervention in the NHS, and (e) estimate of the costs of a blood stream infection to the NHS. Component (a) and (b) are relevant to this application.

Component (a) cost comparison will identify the health care resource use recorded for each PREVAIL baby and cost it using the appropriate unit cost. It will indicate whether there are differences in cost between trial group over the 183 days of follow-up and the magnitude of any differences. Component (b) decision modelling will simulate the health outcomes (health-related quality of life and life expectancy) and health care costs of the PREVAIL babies to their expected lifetime. The decision model requires information on the costs of the PREVAIL babies by trial group and health status at follow-up; this includes whether babies had a blood stream infection, whether death occurred within the 183 days of follow-up, and its date, so that UoY can calculate survival by trial group and infection group. Hence, the UoY request for data on health care resource use held in HES, and date of death, between randomisation and 183 days post-randomisation, held in PDS.

The UoL holds the PREVAIL trial database. This includes identifying details of the participating babies plus details of blood stream infections, resistance, antibiotics and feeds. Other than the Date of Birth, no identifiers would be shared with the UoY or UCL, nor linked with the data to be supplied by NHS Digital.

The UoL will send to NHS Digital the babies’ identifiers (NHS number, date of birth, post code and the PREVAIL trial ID). The UoL will also send the time period for which the data is requested for each baby, which is between the date of randomisation and 183 days post randomisation.

NHS Digital will extract the relevant HES records for each baby and, if applicable, the date of death sourced from the Personal Demographic Service (PDS). NHS Digital will link the hospital records and date of death to the PREVAIL trial ID and strip the dataset of any personal identifiers (date of birth, NHS number, postcode).

NHS Digital will securely transfer the dataset to the Centre for Health Economics (CHE) at UoY. The data will be saved directly to the CHE data server via secure LAN connection. The data transfer across the LAN network is not encrypted but secured by the internal network security of the UoY (data is not copied outside of the internal network).

Additionally, the UoL receives data from the Paediatric Intensive Care Unit (PICU) linked to the PREVAIL trial ID from the Paediatric Intensive Care Audit Network (PICANet). This data contains HRG code and length of stay at the PICU plus date and type of discharge. UoL will provide pseudonymised data from PICANet linked to the PREVAIL trial ID to the UoY.

The UoY also receives pseudonymised Neonatal data linked to the PREVAIL trial ID from Public Heath England’s Neonatal Data Analysis Unit (NDAU). This data contains HRG code and length of stay at the neonatal unit plus date and type of discharge.

The UoL will also supply to the UoY a copy of the data collected during the PREVAIL clinical trial (PREVAIL dataset). This copy of the PREVAIL dataset will contain a minimum of personal information about the PREVAIL babies: trial ID, date of randomisation, date of birth, and relevant data that has been collected by the UoL from the recruiting hospitals during the trial. Date of randomisation is required to calculate the hospital length of stay during the PREVAIL trial follow-up period. Date of birth is required to ascertain whether there are differences in costs or health outcomes by the baby’s age at randomisation and trial enrolment.

From the HES and PDS data supplied by NHS Digital, the UoY will create a dataset containing the PREVIAL trial ID, date of death if death occurred, and date of discharge if method of discharge from hospital is death. This dataset will be encrypted and sent via the access controlled UoY drop-off service to the UoL. These are the only data fields sent from the UoY to the UoL.

Using the trial ID, the UoL will link this dataset to a copy of the PREVAIL trial dataset stripped of personal identifiers. The UoL will use these data to determine the effect of AM-PICC vs. standard-PICC on the risk of death and the time to death.

The UoY will use the NHS Digital dataset to calculate the cost of inpatient care, outpatient care and accident and emergency care for each PREVAIL baby.

The UoY will link the cost of inpatient care, outpatient care and accident and emergency care, and date of death to the data supplied by UoL (described above). The UoY will use the date of death to calculate survival over the follow-up period. This will inform the calculation of survival and quality-adjusted survival in the decision model.

The three cost elements (cost of inpatient care, the cost of outpatient care and the cost of accident and emergency care) do not constitute the full cost of hospital care for the babies enrolled in the PREVAIL trial (the PREVAIL babies). In order for the total cost of hospital care to be calculated, data is required on the hospital stays by the PREVAIL babies in the neonatal care unit and in the paediatric care unit. Data on hospital stays in the neonatal care unit is held in the National Neonatal Research Database (NNRD). Data on hospital stays in the paediatric care unit are held in the Paediatric Intensive Care Audit Network database (PICANet).

The PREVAIL babies have all had stays in the neonatal care unit, since it is the setting where the trial took place and the babies were recruited. Babies in the neonatal care unit may transfer to the inpatient wards if surgery takes place, such as surgery for retinopathy, which is common in preterm babies. Therefore, if some babies have had surgery, this additional resource use will be recorded in HES inpatient rather than in NNRD. Babies can be hospitalised in a paediatric unit, which is recorded in PICANet, if they were discharged home but have required more hospital care subsequently. This is because babies, if readmitted to hospital, are admitted to the paediatric unit, rather than the neonatal unit, due to concerns about infections. Babies may be seen at outpatient appointments for follow-up, and this data is stored in HES outpatients. Accident and emergency admissions, if they occurred, will be recorded in HES accident and emergency.

For these reasons, the UoY, the UoL and UCL have sought hospitalisations data from NNRD, PICANet and HES. In summary, for each PREVAIL baby, the UoY will calculate:

• The cost of neonatal care, from data held in NNRD.

• The cost of paediatric care, from data held in PICANet.

• The cost of inpatient care, from data held in HES inpatient.

• The cost of outpatient care, from data held in HES outpatient.

• The cost of accident and emergency care, from data held in HES accident and emergency.

Summed up together, these 5 cost elements constitute the cost of hospital care for each PREVAIL baby.

The UoY will link the 5 cost elements to the PREVAIL dataset. The objectives are:

• To determine whether having an AM-PICC or a standard-PICC has a direct impact on costs, whether infection has a direct impact on costs, and if so, in which cost element.

• To estimate the cost of hospital care by whether infection occurred and by whether death occurred, which will inform the decision model for the health economic analysis.

• To investigate clinical characteristics that may predict the cost of hospital care and their risk of infection. These characteristics will be used to determine whether there are babies for whom AM-PICCs or standard-PICCS are more beneficial and more cost-effective, and therefore should be prioritised for receiving them.

UCL will not have access to the individual level data, but it will collaborate in the development of the methods and interpretation of the results.

The UoY will store and archive the data used for the health economic analysis. Of the NHS Digital data, this is date of exit for reason of death, and hospitalisations recorded in HES inpatient, HES outpatient and HES accident and emergency. The NHS Digital data will be stored for the duration of the analyses. The derived data (e.g. hospital cost for each baby) will be held for minimum of 10 years, in line with data retention requirements stipulated by the funder (NIHR) and a maximum of 15 years, in line with contractual agreements with the study sponsor (UCL). These data will be encrypted and held on central IT system with backup provided by IT services.

The UoL will store and archive the data used for the effectiveness analysis. Of the NHS Digital data, this will consist of the the date of death (derived from the PDS date and/or HES discharge dates) of the PREVAIL babies who died within the follow-up period of the PREVAIL trial (183 days). The personal information on the PREVAIL babies is stored by the UoL on a networked filestore with access restricted (via ACL’s) to central IT staff who maintain the filestore and CTRC staff with a demonstrated need to access the files.

The persons at the UoL and at the UoY conducting the processing activities involving data from NHS Digital have no access to the personal information on the PREVAIL babies other than date of birth, date of randomisation, and, if applicable, ‘date of death’ (i.e. PDS date and/or HES discharge date). There will be no attempt to identify the PREVAIL babies from this data. The UoL holds the PREVAIL patient identifiers separately and unlinked, and will not link the identifiers to the data received from NHS Digital.

The individuals with access to the data are all direct employees of the UoL or UoY.

For UoY, only members of CHE who are authorised to access sensitive data by the CHE liaison officers and who require access to these data as part of their research activities will be allowed to use the system. All staff members are required to comply with the department’s and the university’s policies on the handling and usage of research and sensitive data.

For UoL, all staff members are required to comply with CTRC SOP’s and University Policies on the handling and usage of research and sensitive data. All CTRC staff members undertake data protection training on an annual basis.

The data requested will only be used to calculate the cost of hospital care, risk of death and time to death for each baby taking part in PREVAIL over the 183 days follow-up period as detailed. The data will not be used for any commercial or marketing purposes; the data will not be provided in record level form to any third party. The outputs of the analysis will be reported in aggregate form as average and standard error. All outputs will be aggregated with small numbers suppressed (in line with the HES Analysis Guide).

The outputs of this research are an understanding about whether AM-PICC compared with standard-PICC reduces the risk of infections and their complications, including death, and the consequences for NHS costs. The NHS makes decisions on which interventions to fund based on information on their effect on health and costs. Cost is important given that the NHS has limited funds; hence cost savings can release funds to be invested in other interventions that improve health. Conversely, interventions that are more costly may warrant funding if they improve health over and beyond the health forgone due to disinvesting in other interventions. In this context, hospitals should invest in AM-PICCs if AM-PICC prevent infections and are good value for money. In other words, AM-PICCs may be cost-saving, in that preventing infections may reduce the cost of hospital care; or AM-PICC may be more costly, but their additional costs are offset by the gain in health-related quality of life and/or life expectancy, from the prevention of complications due to infection.

The UoY, UoL and UCL will report the output of the study in a number of outlets; the outputs will be produced in an appropriate way depending on the audience:

• NIHR Health Technology Assessment monograph detailing the methods and the results of the PREVAIL study. This includes: the methods of the PREVAIL trial, the methods and results of the effectiveness analysis on the risk and short term consequences of infection, and the methods and results of the economic analysis on the short and long term consequences of infections in health-related quality of life, life expectancy and costs. The expected publication date is 2019, and it is open access to all readers.

• Peer-reviewed publications in high impact journals such as Paediatrics. PREVAIL envisage that the UoL will lead at least one peer-reviewed publication on the effectiveness analysis and that the UoY will lead at least one peer-reviewed publication on the economic analysis. These will be submitted by the end of 2018. Given the long lead times required for peer-reviewed papers, publication will likely be in 2019 but unfortunately this cannot be more specific. The publications will be deposited in each institution open access repositories or published in open access format.

• Presentation of the results at national and international conferences and neonatal network meetings, throughout 2018 and 2019.

• The results will be shared with the parents of the PREVAIL babies, through the PREVAIL website.

• The results will also be shared with user representatives such as via the Bliss parent newsletter, website and social media. Examples of this kind of summary can be seen here http://www.bliss.org.uk/achievements

The NIHR monograph, peer-reviewed publications, and presentations at conferences are targeted at clinicians, hospital managers, and researchers. Clinicians make decisions about which type of PICC to use, and therefore are ideally placed to use this research to directly inform the care provided to preterm babies. Clinicians and hospital managers are involved in purchasing decisions about which type of PICC to obtain. Researchers can take the results of this study forward and continue to improve the understanding about infections in preterm babies and how to prevent them. Health economists working in the area of neonatal care, for example, may be interested in PREVAIL’s methodological approach.

The outputs disseminated by Bliss are targeted at parents and aim to improve their understanding about the risks and consequences of infection and whether AM-PICC vs. standard-PICC can prevent them. This will help parents understand the evidence on which decisions affecting their baby are based on.

The output of the analysis will be reported only in aggregate format as average and standard deviation of the cost of hospital care. Small numbers will be suppressed in line with HES analysis guide.

The PREVAIL study will provide evidence on whether AM-PICCs reduce infections and are cost-effective in improving health of preterm babies. The question of the effectiveness and cost-effectiveness of AM-PICC for preterm babies is of high importance to the NHS. Survival in neonatal care has improved remarkably over time, but the rates of permanent neurological disability in preterm babies are high and have not improved. For example, more than 1 in 4 babies with gestations of less than 26 weeks have a serious disability at 2 years. Infections increase the risk of neurological disability. Consequently, interventions that reduce the risk of infections will reduce the burden of disability in premature babies.

AM-PICCs have been shown to be effective in reducing blood stream infections in children and adults. However, there is no evidence on their benefits in preterm babies, who are at a high risk of infection. Given the lack of evidence on benefit, and their increased costs compared to the standard PICC, NHS neonatal units currently use the standard PICC. The PREVAIL study will address this gap in knowledge by finding out whether AM-PICCs reduce the risk of infections in preterm babies, and whether this reduced risk justified their increased cost.

The findings of the study will help hospitals and policy makers to decide whether to implement AM-PICCs or not. To inform these decisions, PREVAIL will present the results to policy makers in a report to NIHR-HTA in which explicit research recommendations and implications for practice will be made. Findings and recommendations for practice will also be presented at clinical conferences and in peer reviewed journal publications. Findings will be fed back to neonatal units and regional neonatal networks through the Neonatal Data Analysis Unit.

It is expected that, thanks to the PREVAIL study, hospital managers and clinicians will have evidence about how good are AM-PICC vs. standard-PICC in preventing infections, improving health and reducing costs. This evidence will inform their decisions about which type of PICC to buy for neonatal care units. We expect this evidence to be used as soon as it is reported, in late 2018 – early 2019.

Parents will have a better understanding about the risks of infection and their consequences to the health related quality of life and life expectancy of their babies, and how the different types of PICC can prevent infections. PREVAIL will disseminate this information via Bliss, as soon as the results are available in late 2018 – early 2019.

PREVAIL cannot give a date for benefits to the NHS being realised as any benefits depend on many factors, including the results of the study (i.e. whether there are any clinical and cost benefits or harms of AM-PICCs and how large and certain these effects are) and on other barriers and facilitators of implementation. PREVAIL’s aim is to disseminate and explain the findings of the study to policy makers, service providers and the public so that decisions can be made about impregnated CVCs and any benefits to the NHS can be realised.

If results show an important benefit of AM-PICCs, PREVAIL will advocate for the findings to be incorporated into national guidance to neonatal units and to be reviewed by NICE. PREVAIL will also advocate for national monitoring of uptake of AM-PICCs (if findings are positive) through national audits.

Yielded Benefits is not a requirement for new applications.