Mortality Differences in Children of different Sexs' Admitted to UK Critical Care Units

This research is part of a Doctoral Research Fellowship at the Population Policy and Practice programme, at the University College London (UCL) Great Ormond Street Institute of Child Health (ICH). It uses information collected from administrative sources (such as admissions to critical care, hospital episode statistics, and maternity records) to conduct research in order to understand the impact of critical illness on children.

Every year, more than 20,000 children are admitted to Paediatric Intensive Care Units (PICUs) in the UK. Previous small studies have showed that baby girls may have higher mortality rates than baby boys in PICU. In 2017, an analysis was completed of all babies (0-12 months old) who were admitted to PICUs over an 11-year period. Anonymous records from PICANet of 86,000 babies were obtained and the rates of death between girls and boys during their admission to PICU were compared. The records were anonymous as no personal identifiers (e.g. DOB, postcode) were contained in these records. The data did contain the variable ‘sex’. It was shown that girls had higher death rates than boys. This is different to what is seen in the general population where boys have higher death rates than girls for children of all ages. Careful examination were carried out of whether this difference could be due to differences in age, disease severity, infections, and a number of other factors. None of the factors could explain why girls died more than boys in PICU. The purpose of this research project now is to examine these findings in greater detail as this could have implications for the care of critically ill children generally.

The lawful basis for using information collected routinely for administrative purposes for research is the ‘public task’. This is part of the University’s commitment to integrate research and innovation for the long-term benefit of humanity. The public task basis may be found in Article 6(1)(e) of the General Data Protection Regulation, which states:

“Processing is necessary for the performance of a task carried out in the public interest or in the exercise of official authority vested in the controller”

The processing also falls under Article 9(2)(j), which states:

“processing is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical Purposes”

Aims and Research Questions:

1. Why do infant girls have a higher mortality rate than infant boys in PICU?

2. Are there similar sex differentials in mortality rate at older ages for children admitted to PICUs in England?

3. After discharge from PICU, does the mortality for girls remain higher than boys?

4. Why do infant boys have a higher admission rate to PICU than infant girls?

5. Should the Paediatric Index of Mortality currently used by all PICUs in England be sex-specific?

6. Is there a better way of modelling the mortality data for PICU to get more accurate mortality estimates?

7. How should we analyse data on length of stay in PICU?

This project achieves two overall goals:

- answer a clinical question of why the mortality for females within PICU is the reverse of what is seen in the general population.

- build research capacity within the NHS and train future research leaders. Funding has been secured from NIHR and the study has been peer reviewed by the funding body.

The outcomes expected from this study are:

Primary:

- Death occurring at PICU with 7, 30 and 90 days of admission

- Death after discharge overall (any deaths occurring after discharge from PICU within 30 days of discharge, within 5 years of discharge).

Secondary:

Time to discharge from PICU (Length of Stay)

Child and infant mortality is an important indicator of child health and overall development in countries. The September 2010 United Nations Summit set the target to reduce mortality rate of children under five by two thirds. Similarly, the United Nations 2011 report titled “Sex Differentials in Childhood Mortality” investigated if recent improvements in child survival linked to the fourth millennium development goal had benefited the sexes equally, examining sex differentials in childhood mortality for 149 countries and regions from the 1970s to the first decade of the 2000s. The following link provides further insight:

(https://www.un.org/en/development/desa/population/publications/mortality/sexdifferentials.asp)

Under circumstances where boys and girls have the same access to resources such as food and medical care, boys have higher mortality rates than girls during childhood, showing that as overall childhood and infant mortality declines, the survival advantage of females over males is maintained. Thus, in developed countries, one expects (and sees) a relative survival advantage for females over males in infants, and children up to the age of five.

Although females have an overall biological advantage in childhood survival, some research suggests that the two sexes have variable immune responses to infections as mediated by sex hormones. This motivated a preliminary analysis of UK admissions due to severe infections to PICUs. The results showed greater male infant admissions to PICU over a two year period, but a greater female infant mortality over the same period, with the mortality difference not reaching statistical significance. A follow up to this analysis was carried out over a five-year period showed and similar results.

This motivated a larger study to find out if the conclusions from these preliminary analyses were also seen across all infants admitted to PICU regardless of diagnosis. In 2017 and after gaining ethical approval, UCL used data from the Paediatric Intensive Care Audit network (PICANet), which collects demographic and basic clinical data on all admissions to PICU in the UK and Ireland and analysed the entire cohort of infants admitted to all UK PICUs from January-2005 to December-2015 (final sample included over 70,000 infants within 35 PICUs). Here, UCL analysed all 35 PICUS, including Ireland, however, the researchers could not identify which unit the babies were from, as only an indicator was provided. The results of the survival analysis confirmed a higher hazard ratio (HR) of death for infant girls over infant boys, with strong evidence for a higher female mortality than male and this challenges conventional orthodoxy. A number of selected confounders were selected, none of which explained this HR. Thus, this study will be the first to address the questions of: what mechanisms drive the sex disparity in PICU mortality; what mechanisms drive the differential PICU admission rates for males and females, and what are the long term life trajectories and survival outcomes for children after an admission to PICU.

Data Sets Required:

Under a previous iteration of this Agreement, the datasets requested from NHS England were:

1) Reuse of an existing HES - civil registration cohort for children and their mothers (where mothers and children’s data are already linked using probabilistic linkage by UCL), held by UCL (DARS-NIC-393510-D6H1D), which is an existing and approved Data Sharing Agreement also held by UCL.

2) Maternity services data set (MSDS) for the available 5 years (2015 – 2019) from NHS England

3) Reuse of civil Registration mortality data for children (those identified by PICANet cohort only) held by UCL (DARS-NIC-393510-D6H1D)

Section 5b (Processing Activities) outlines the details of dataset linkage.

These datasets will provide the longitudinal cohort to study the course of the patients before, during and after admission to PICU. After linkage, this longitudinal cohort will provide the required data on the children admitted to PICANet; i.e. data prior to their admission and data after their discharge form PICU. Pre and post PICANet data is necessary to build a complete picture of all the factors that could be involved in the mechanisms leading to admission to PICU, discharge from PICU, and long term survival after PICU. The PICANet cohort data only provides a snapshot of what happens inside PICU and cannot offer mechanistic explanations for the disparity in sex admission rates mortality, hence the need to link it to the above datasets.

The linked data will provide a longitudinal cohort that starts before admission to PICU (mother and baby HES datasets), during admission to PICU (PICANet dataset), and follows the children up to five years post discharge from PICU (Civil Registration (deaths). There will be no requirement nor attempt to re-identify individuals from the data.

The required data sent to UCL's data safe haven is patient-level data. Some identifiers will be retained for analysis (date of birth [DOB], date of death, city, sex, occupation, ethnicity. These have been detailed and justified in the CAG application form and have received full approval (19-CAG0164 CAG approval). The identifiers are used only for the benefit of the PICANet cohort and the HES/Civil Registration data under NIC-393510. DOB and Date of death will be used to calculate a number of days of life as the main outcome is mortality estimated using survival analysis. Cause of death is also one of the objectives of the analysis.

Occupation as a field is required as certain occupations can lead to different health/lifestyle outcomes, which may have an impact on families and their children. For example, exposure of a mother to occupational factors during pregnancy (such as night shifts or chemicals) could contribute to the study's understanding of outcomes in children.

Ethnicity is one of the main factors which has shown to be of great importance in explaining differences in mortality, both in UCL’s previous analysis and in the literature in general. Ethnicity will also be in five broad categories to minimise the sensitive nature of the variable.

These sensitive fields have been detailed and justified in the CAG application form and have received full approval. There is no requirement at all to re-identify any subject as a result of this analysis, thus no attempt will be made for any re-identification of subjects.

The requested number of years is the minimum to achieve the sample size required (see below in cohort details) The calculated sample size ensures that the study have precise and meaningful answers. To conduct this research without the required sample size would not offer a meaningful contribution to the body of evidence and a waste of resources. The number of years requested (2010-2019) will allow a sufficient geographical spread from the variety of PICUs involved and this means the research will be far reaching and have higher levels of generalisability.

Data Minimisation, as follows:

- restricted to 10 years only (the minimum required to achieve the calculated sample size)

- restricted to children with an admission to PICU within the study period (2010 – 2019)

- MSDS data restricted to women, of child bearing age (15 to 50 years of age), who had children during the study period (2010 – 2019). MSDS is currently only available from 2015.

- restricted geographically to England.

- postcode will not be requested, instead Lower Layer Super Output Area (LSOA) and Index of Multiple Deprivation (IMD) will be used.

Recruitment and Cohort Details

There is no active recruitment for research participants. UCL plan to create a retrospective cohort study using data from PICANet covering the periods 2010-2019 and ages 0 to <18 years. With 20,000 admissions per year over the whole of the UK (16,000 from English sites), this will give a sample of approximately 160,000 cases from English sites alone over a ten-year period. Allowing for any exclusions, a sample size of at least 140,000 is expected. All children 0 to <18 years of age who had an admission to a PICU in England between January 2010 and December 2019 are eligible for inclusion. Currently there are no plans to have any additional phases to this project. Sufficient data will be requested to ensure the research questions are answered. The sample size has also been calculated to give a meaningful answer.

Seeking consent for linkage from approximately 140,000 children admitted to PICU would not be feasible without further disclosure (a need to obtain up to date address details). Further, particularly relevant for the outcome of death within PICU, it may be deemed insensitive to contact parents of non-survivors. The outcome is a rare event and is needed to capture data on all participants to carry out meaningful analysis. Due to the long time elapsed since the outcomes of discharge from PICU or death in PICU, families may have moved away or are un-contactable, which could introduce substantial bias into analyses due to a potentially high non-response rate.

Storage/Security Assurance Details

The data will be released to a safe haven at University College London. The Data Safe Haven has been certified to the information security standard (ISO27001) and conforms to NHS England's Information Governance Toolkit.

Storage of data on the UCL Data Safe Haven, with access restricted to authorised users, who are required to have certified training in the use of the safe haven and in data governance. Authorised users require a personal PIN, a dual identification token device, and a personal password to access the safe haven. Access to the project data set will be restricted to three members from the research team.

Standard operating procedures are followed for the destruction of data at the end of the study.

The sole controller is UCL. Great Ormond Street Hospital ICH are a processor.

Amazon Web Services (AWS) is a processor acting under the instructions of UCL. AWS’ role is limited to secure back-up of data stored in UCL’s Data Safe Haven.

UCL uses offsite data centre services provided by VIRTUS data centre. VIRTUS does not have access to the data.

Role of PICANet and other collaborators

Clarity on the role of GOSH- Great Ormond Street Hospital and UCL Institute of Child Health share the same R&D office. GOSH hold the budget for this project, while UCL are the sponsor. GOSH does not have access to the data, and only holds the budget to pay for the DARS requests.

PICANet are based at the University of Leeds and use data from the national project audit data. This audit is commissioned by the Healthcare Quality Improvement Partnership (HQIP) and University of Leeds is a data processor for the audit data. HQIP approval has been obtained for the purpose of this research (reference HQIP332). PICANet only hold the cohort of data to be linked and sent into NHS England, and will not have any influence over the means by which the data NHS England releases. PICANet have Health Research Authority Confidentiality Advisory Group (CAG) and ethical approvals granted by the Trent Medical Research Ethics Committee, reference 18/EM/0267. CAG approval in place (19CAG0164) to enable identifiable PICANet data to be released to NHS England for linkage with the data by NHS England. PICANet have ethics approval to collect and use patient data for the purpose of research. Collection of personally identifiable data has been approved by the Patient Information Advisory Group. .

Thus, PICANets involvement in this project will be to provide the cohort to NHS England and to co-supervise the PhD candidate. The supervisor is the Principle Investigator for the PICANet project (from the University of Leeds). PICANet will not have access to the final linked dataset once released to UCL from NHSD.

The PhD supervisors at Imperial NHS Trust (St Mary’s Hospital) and University of Leeds will not have access to the data at UCL and will be acting in clinical and statistical advisory roles. They will not be involved in decisions on data control or data processing.

This project is funded by the National Institute for Health Research. The funding award is a Clinical Doctoral Research Fellowship for the lead applicant.

As a PhD student at UCL, the main researcher has access to this DSA. After 31/12/2024, the individual will no longer have student status at UCL, but will be an honorary staff member at the same department to continue having access to the data via this DSA. Until this occurs, the individual remains a substantive GOSH employee.

Under a previous iteration of this Agreement, all admissions to PICU between 01/01/2010 and 31/12/2019 were identified through the PICANet project principal investigator. The PICU cohort was provided to NHS England for further linkage.

There is HQIP approval for the use of PICANet data in place (HQIP332).

There is CAG approval in place (19CAG0164) to enable identifiable PICANet data to be released to NHS England for linkage with Hospital Episode Statistics and civil registrations mortality data by NHS England (referred to as Set A – details below).

The identifiable PICANet data needed for linkage were:

Date of birth [DOB], date of death, sex and postcode (unit level). These have been detailed and justified in the CAG application form and have received full approval (19-CAG0164 CAG approval). NHS number and postcode were used for linkage by NHS England, but not released to UCL. Date of Birth and Date of death to derive secondary variables such as time from admission to death, and age at admission to HES or PICU.

Using the PICANet data identifiers, NHSD linked individuals' records from the PICU set to the already linked HES APC and Civil Registration (deaths) HESIDS records under NIC-393510 and NHSD provided the HESID (via a bridge file) to enable UCL to link the PICANet data to their existing HES cohort. This formed data set A. NIC-393510 already contains linked children and mothers and the PICANet set identified a subset of NIC-393510.

The PICU cohort will was also linked to the mothers’ data of the individuals within PICANet. NHSD identified the Maternity Services Data Set (MSDS) from 01/01/2015 to 31/12/2019 and linked the MSDS to HES and civil registrations mortality (mortality for children only, not mothers) and transferred the linked, MSDS data plus a link key so that UCL can link MSDS with their existing HES cohort. MSDS is only available from 2015. This formed set B.

Set A (relating to PICANet individuals HES and civil registration mortality) and Set B (relating to PICANet individual’s mothers HES and MSDS data) were housed at the data safe haven at UCL, where these data were merged with the already linked HES/Civil Registration extract held under NIC-393510. This formed set C.

Data Flows

To generate set A:

1. NHSD receive PICANet identifiers (identifiable data)

2. NHSD match the PICANet cohort to existing HES APC and civil registrations mortality (HES: Civil Registration Bridge held under NIC-393510),

3. NHSD flow back ‘set A’ which is PICANet study ID, linked to HESID of NIC-393510, (HES: Civil Registration)

4. UCL will merge the PICANet-HESID linkage key from step 3, with the HES data held by UCL under NIC-393510, to identify data for the children in the PICANet cohort that are already held by UCL (includes a number of CAG agreed variables: sex, DOB, date of death, occupation of the mothers).

To generate set B:

1. NHSD receive PICANet identifiers (identifiable data).

2. NHSD use NHS number to link mothers in MSDS to the children in PICANet.

3. For mothers in MSDS from step 2, NHSD add the HESID from NIC-393510.

4. NHSD produce set B, which is PICANet children (from step 1) linked with MSDS mothers (from step 2) and mother’s HESID from NIC-393510 (step 3)

5. NHSD flow back everything in step 4 above plus the data held in MSDS which is linked to PICANet (includes a number of CAG agreed variables: sex, DOB, date of death, occupation)

To generate the final dataset; set C

1. Set C will be generated at UCL

2. HESID from the pseudonymised NIC-393510 is the key to generating set C

3. The NIC-393510 HES data extract already contains details for the mothers and children from the PICANet Cohort . In NIC-393510 mothers and children have already been linked using probabilistic linkage.

4. UCL have permission to use data from NIC-393510.

5. UCL will merge set A to the data held by UCL using HESID NIC-393510. This will use just the relevant subset of NIC-393510 that match PICANet children.

6. UCL will use HESID NIC-393510 to merge set B to step 4 above.

7. UCL will obtain PICANet data from HQIP and merge with step 5 using study ID number.

Set C will be a longitudinal cohort following children from pre-birth (to account for parental factors), through to PICU admission, and up to five year post discharge. This will allow UCL to study the reasons for different admission rates for males and females, and the survival trajectory of children after discharge from PICU.

There will be no requirement nor attempt to re-identify individuals from the data at any of the above stages of data flow and linkage.

The data analysis will be primarily done by the lead applicant currently registered as a PhD student at UCL and has a substantive contract with Great Ormond Street Hospital. Thus, GOSH have been named as a processor to incorporate this arrangement. The Primary supervisor and secondary supervisors who are substantive employees of UCL will have access to the data. The data will only primarily be accessed by these people, but other members of the research team who are also substantive employees of UCL can also access the data. All UCL staff have undertaken training on Information Governance both at UCL and the NHS, and training on GDPR.

PICANet’s involvement in this project will be to provide the cohort to NHS England and to co-supervise the PhD candidate. The supervisor is the Principle Investigator for the PICANet project (from the University of Leeds). PICANet will not have access to the final linked dataset once released to UCL from NHSD.

The PhD supervisors at Imperial NHS Trust (St Mary’s Hospital) and University of Leeds will not have access to the data at UCL and will be acting in clinical and statistical advisory roles. They will not be involved in decisions on data control or data processing.

Amazon Web Services provides cloud hosting services to UCL and will store the data as contracted by UCL.

Storage of data on the UCL Data Safe Haven, with access restricted to authorised users, who are required to have certified training in the use of the safe haven and in data governance. Authorised users require a personal PIN, a dual identification token device, and a personal password to access the safe haven. Access to the project data set will be restricted to only members from the research team.

Version 0 (Sept ’20)

There will be a PhD thesis to be completed and submitted to the UCL examination board by May 2022. This is in fulfilment of the lead applicant’s doctoral training. Other reports will also be prepared for dissemination to members of the public. Specifically, the groups involved with this project: Children of St Mary’s Intensive Care (COSMIC) and useMYdata. The project has a PAG (project advisory group) which are two members of the public, one of whom is a member of the national group UseMyData. Additionally, YPAG (young persons advisory group) and PCAG (parent and carer advisory group) have regular meetings and input into the project. Some dissemination has taken place via COSMIC charity (Children of St Mary’s intensive care, https://cosmiccharity.org.uk/). Members of PIC Families, who are part of PICANet will also be involved in sharing these findings (https://www.picanet.org.uk/about/people/pic-families/). Representatives from these groups will assist in the format and content reports for dissemination of these results to the correct public forums, such as ESPNIC (European Society of Neonatal and Paediatric Intensive Care) and PICs (Paediatric Intensive Care society). The results will also be shared with the NIHR who are the funding body and during any NIHR events.

UCL also aim to produce submissions to peer reviewed journals. The findings from the analyses (see research questions 1 to 7) will be published in peer-reviewed journals such as Lancet adolescent and child health, or British Medical JournalArchives of Disease in Childhood (BMJ ADC). There is also the aim to submit abstracts for presentations at conferences such as ESPNIC and ADR (administrative Data Research Conference). There will also be presentations at UCL, Imperial NHS Healthcare, Great Ormond Street Hospital and other relevant forums. Publications in peer reviewed journals will follow the thesis submission and are expected to be no later than 2024 (should there be any queries resulting from the analysis that need to be addressed).

Recommendations to inform any clinical guidelines arising out of the research will be published and disseminated to professional societies concerned with the care of children presenting with acute illness, including PICS and the Royal College of Paediatrics and Child Health. One such recommendation would be findings from the reassessment of the Paediatric Index of Mortality (PIM) score, before the end of 2023

All publications, abstracts and presentations will be in summary forms and will not have any individual-level or identifiable data. The identity of the PICUs will also be kept confidential and not published. Only summary statistics and aggregate data with small numbers supressed will be included in the outputs, in line with HES analysis guidelines.

Dissemination and communication approach

The findings of the research will be shared within the critical care community via multiple platforms:

- Scientific: this will be through PICANet and PICS, peer-reviewed journals (funding available for open access) , and other scientific platforms such as ESPNIC and PICs. Any change in practice or change in policy will be communicated via the PICs community.

- Public engagement: in addition to COSMIC and useMYdata, PICANet also engage with the public by having lay members on their PICs families’ forum. These findings will be shared and discussed within the forum where the lay members can input further onto plans for disseminations. There is ongoing PPI/E activity throughout the duration of this project and members of the public will be able to share their thoughts at any time during the project.

- Other: This project is publicised through the UCL and PICANet websites and any findings will also be communicated through these websites.

October 2022 v1 update:

Due to the delays in receiving the data, the funders (NIHR) have extended the duration of the fellowship to the end of February 2023 (9-month extension). As a result of these delays, the content of the thesis has been re-evaluated. Therefore, some of the project outputs will now need to be delivered to the funder after the submission of the thesis. To replace the thesis content, a new piece of work which is a systematic review was completed and is in the process of peer review with a journal. This piece of work remains within the scope of the existing project and all outputs and analyses are anonymous.

So far, two oral presentations are booked at the European Academies of Paediatric Societies conference in late 2022. One is to present the results of the meta-analysis, and the other is for the preliminary results of the difference in PICU mortality between males and females.

Since the main researcher will be returning to full-time clinical work after the PhD, the 3-year duration is required to work on the remaining outputs. These are:

1. recalculation of the PIM score

2. the methodology part of the project where joint modelling of the length of stay and mortality outcomes are planned

3. to complete the publications expected from the project

Dissemination and communication approach

Publications in peer-reviewed journals will follow the thesis submission and are expected to be no later than August of 2026 (should there be any queries resulting from the analysis that need to be addressed).

Recommendations to inform any clinical guidelines arising out of the research will be published and disseminated to professional societies concerned with the care of children presenting with acute illness, including PICS and the Royal College of Paediatrics and Child Health. One such recommendation would be findings from the reassessment of the Paediatric Index of Mortality (PIM) score, before August 2026.

As a result of this research, the benefits to health and social care will include:

1. Improved patient care. This work will identify to clinicians and commissioners individuals who are at greatest risk mortality after PICU admission; this will enable follow-up practices to be tailored to patient needs, help identify potential health problems early and intervene so that patient wellbeing is maximized and NHS burden minimised. An example of tailoring practices to patients’ needs is where the research identifies the characteristics of children who have the highest risk of mortality after discharge from PICU, thus enabling practitioners and funders to target their needs. This can achieve maximum benefits for the patients and their families whilst ensuring that NHS resources are used efficiently.

For example, those individuals identified from the characteristics (such as certain aetiologies, or for example maternity illnesses during pregnancy) that lead to:

- higher risk of PICU admission, therefore either preventing admission altogether or avoiding emergency admissions which are known to have a higher risk of mortality

- Higher mortality in PICU, and focusing on interventions that are appropriate for the level of risk for each individual

- Higher mortality post discharge from PICU, therefore targeting them for appropriate care and support as explained above

2. Evaluation of treatments to identify best practice and guidance. This research project will work towards understanding the reasons for PICU admissions and in particular the higher rate of admission for males, so researchers can identify whether certain treatments are associated with an increased risk of PICU admission and disseminate this information through scientific journal articles. Critical illness of a child is a time of great anxiety for parents and families. Understanding factors that contribute to admission to PICU can help address and alleviate some of the clinical and family needs.

3. Evaluation of service provision. The research will highlight any inequalities in access to specialist critical care services, particularly in various socioeconomic areas, so that all patients have an equal chance of obtaining the best care irrespective of their personal circumstances and thereby having the best chance of treatment. The work will be written up in the form of reports to journal articles so that clinicians and commissioners can use this information in order to make any necessary changes to service delivery with the help and involvement of Paediatric Intensive Care Society (PICS) and the patient groups such as PICS families, UseMyData, and COSMIC.

As a results of the outputs, there will likely be discussions around how to improve on the current risk of mortality score, and if there should be an ongoing score thought the PICU admission to ensure better outcomes for both sexs', and to reduce the apparent inequality.

A revision of the index of mortality score will have an impact on an average of 20,000 PICU admissions per year UK wide. This means ensuring that patients are treated according to the severity of their disease not just on admission but through their PICU stay, which will result in more focused management strategies for patients.

The benefits will entirely be for patients and their families all over the UK. And if adopted externally to the UK, then further benefit to PICU patients beyond the UK may also be achieved in the future. UCL anticipate the benefits will be achieved once any management strategies have changed as a result of the available evidence from this project. Typically such changes (UK wide) will take between 3 to 5 years from availability of results to imbed and show benefit. Furthermore, to measure any benefits of this research, further mortality benchmarking can be made through the routine data collection of the PICANet project.

This research is in support of a PhD doctoral fellowship project .

So far there have been 1 publication but that is the systematic review which is in the process of peer review. This systematic review will also be presented in the European Academy of Paediatric Societies conference in late 2022.

However for the most part, no yielded benefit has yet been achieved as the data still continues to be processed with the 1st outcomes only being established recently.