NHS Digital Data Release Register - reformatted

Roehampton University projects

13 data files in total were disseminated unsafely (information about files used safely is missing for TRE/"system access" projects).


Understanding diversity in childbirth and London’s ethnic gap in maternal morbidity — DARS-NIC-350331-P9L1M

Type of data: information not disclosed for TRE projects

Opt outs honoured: Anonymised - ICO Code Compliant, Identifiable, No (Does not include the flow of confidential data)

Legal basis: Health and Social Care Act 2012 - s261(5)(d)

Purposes: No (Academic)

Sensitive: Sensitive, and Non-Sensitive

When:DSA runs 2022-10-17 — 2023-10-16 2023.04 — 2023.05.

Access method: One-Off

Data-controller type: ROEHAMPTON UNIVERSITY

Sublicensing allowed: No

Datasets:

  1. Civil Registration (Deaths) - Secondary Care Cut
  2. HES:Civil Registration (Deaths) bridge
  3. Hospital Episode Statistics Admitted Patient Care
  4. MSDS (Maternity Services Data Set) v1.5
  5. Civil Registrations of Death - Secondary Care Cut
  6. Hospital Episode Statistics Admitted Patient Care (HES APC)
  7. Maternity Services Data Set (MSDS) v1.5

Objectives:

AIM/BACKGROUND:
Black women in Britain are five times more likely to die in childbirth than white women, and women of Asian background have a two-fold risk of maternal death (MBRRACE-UK 2018; https://www.npeu.ox.ac.uk/assets/downloads/mbrrace-uk/reports/MBRRACE-UK%20Maternal%20Report%202018%20-%20Web%20Version.pdf). Substantial ethnic gaps in maternal morbidity have also been described, with white women having the lowest rate of severe morbidity and black women the highest. While unconscious bias and institutional racism have been discussed as potential reasons for this disparity, one factor that has not been explored to date is the fact that obstetric care has been modelled on white women’s physiology: obstetric and midwifery training, and calculations that define what counts as a normal progression of labour are based on studies of childbirth in women of largely white ethnicity. Yet, there is evidence of ethnic differences in the length of labour, in birth weight, and in the most common presentation of the baby at birth, the latter probably due to ethnic differences in the shape of the pelvis. Standardizing the birth process on the basis of a model that applies to only a relatively small part of the world’s population ignores the diversity of women’s bodies and risks subjecting those who do not conform to the standards of ‘normal’ labour to unnecessary interventions, which bear significantly higher risks of adverse outcomes.

This pilot study intends to explore the extent to which the parameter of childbirth and the rate of birth interventions differ among women of different ethnic groups in London, to evaluate:
1) any ethnic gap in maternal mortality and morbidity within the same geographic areas and care providers (selected areas of London and selected hospitals);
2) whether key parameters of childbirth—such as the length of the two main stages of labour, duration of pregnancy, birth weight, baby’s presentation at birth—differ in white, Asian, East Asian and black women in London;
3) whether specific childbirth interventions—such as hormonal induction, forceps application, C-sections—are used more frequently in some ethnic groups;
4) whether any identified ethnic difference in the rate of interventions correlates with differences in the process of labour (e.g., length of different stages of labour), potentially providing an explanation for different incidence of labour interventions;
5) Whether any identified ethnic difference in the rate of certain interventions could contribute to the ethnic morbidity and mortality gaps.

The study is in the public interest as it aims to investigate several factors that could contribute to the ethnic gap in maternal mortality and morbidity observed in the UK, and understanding the factors involved will help reduce the gap in maternal outcomes. Roehampton University are a research institution as such access to the data requested though this application is therefore justified under Article 6(1)(e) and Article 9(2)(J) of the GDPR.

NHS DIGITAL DATA REQUEST:
Aggregated national data are difficult to interpret, as Black, Asian and Minority Ethnic (BAME) women are mostly cared for in a minority of hospitals in particularly highly diverse areas of the country, while most hospitals in the country serve a largely white demographic. It is hoped this study will allow to compare mortality and morbidity rates for white and BAME women within the same settings, while correcting for known risk factors (e.g., maternal age, weight, smoking…). To correct for the effects of individual risk factors, record-level data are needed, as well as access to some sensitive data (e.g., smoking status, some socio-economic factors, pre-existing conditions). Focusing on a small number of hospitals in diverse areas will also permit to test whether white and non-white women incur different rates of certain childbirth interventions (aim 3). Again, the national data suffer from the problem that preference for certain interventions vary among hospitals and hospital trusts, and it might vary in a non-random way between hospitals that serve a diverse or a majority-white community.

NHS Digital data hopes to also allow:
- to test whether the national ethnic gap in maternal mortality and morbidity is also present within highly diverse areas in London, in hospitals that care for a large portion of Black, Asian and minority ethnic (BAME) women (aim 1).
- to test whether there are significant differences between ethnic groups in key aspects of labour, such as the duration of different stages and foetal position (aim 2).
- to evaluate whether the current model of “normal” childbirth, based on white women’s physiology, is representative of normal variation in other ethnic groups. If black women on average tend to have a longer first stage of labour, for example, they might incur a higher rate of interventions simply because their normal variation is not well represented by the current white-centred model of childbirth. This project intends to test whether ethnic diversity in labour characteristics could explain ethnic differences in rates of interventions (aim 4).
- to test whether higher rates of interventions in BAME women could explain differences in maternal mortality and morbidity, while taking into account other risk factors (aim 5).

While this pilot project cannot provide evidence of causality, it is hoped will highlight whether ethnic differences in labour should be explored further, as potentially contributing to the higher maternal mortality and morbidity in BAME women (potentially via higher rates of intervention).

This project is specifically shaped to tackle a problem that has recently been highlighted in the news. During summer 2020, a petition to the UK parliament launched by activists and titled “Improve Maternal Mortality Rates and Health Care for Black Women in the U.K.” collected over 180,000 signatures, please see - https://petition.parliament.uk/petitions/301079. In response to the petition, the Joint Committee on Human Rights, a group of MPs and Members of the House of Lords, heard about the maternity ethnic gap in mortality from expert witnesses and in November 2020 made the recommendation that a target be set by the Government to reduce this mortality gap. The petition and the recommendations were discussed in parliament in 2021. This research project aims to collect evidence that can be fed into the parliamentary discussion of the petition and that can inform policymakers and help shape future policy changes to reduce the ethnic gap in maternal mortality and morbidity. The results of the analyses of the pilot data requested through this application is hoped will form the basis of a report to the Maternity All-Party Parliamentary Group. The Maternity All-Party Parliamentary Group (APPG) is an "informal, cross-party groups formed by MPs and Members of the House of Lords" who share a common interest in maternity policy. Its aim is to "improve the quality of maternity services, raise awareness of the contribution maternity care can make to public health, and improve support during the first 1000 days of parenthood". Details on the Maternity APPG can be found here - https://www.parallelparliament.co.uk/APPG/maternity.

A larger follow-up project (with a multidisciplinary team not directly involved in the analyses of these pilot data) will aim to better define ethnic differences in labour and the physiological and social causes behind them, and how any such ethnic differences are, or could be, taken into account by maternal care providers. The overarching aim is to build a more inclusive model of childbirth that is based on normal variation in all women, and not biased towards white women, therefore improving maternal care for BAME women in the UK. The data collected for the pilot study will not be needed for the follow-up study, which will involve recruitment of volunteers instead of analysis of existing data.

This pilot study is funded through a Roehampton University internal Impact Seed Funding award, and it emerges from the PI’s long-term investigation of physiological differences in the birth canal in women of different ethnicity and the potential effects on labour and obstetric care (e.g., Betti L and Manica A 2018. Human variation in the shape of the birth canal is significant and geographically structured. Proc. Roy. Soc. B 285: 20181807; Betti L 2021. Shaping birth: variation in the birth canal and the importance of inclusive obstetric care. Phil. Trans. Roy. Soc. B. 376: 20200024).

Roehampton University is the sole Data Controller who also process the data for the purposes described in this Agreement.

To answer the research questions for this pilot project, it is necessary to collect data that inform of:
1) maternal morbidity and mortality;
2) individual risk factors;
3) characteristics of labour;
4) the care received by women during childbirth, including timing and occurrence of interventions;
5) outcome of labour for the mother and child, including maternal complications linked to the birth.

To test for a relationship between mortality/morbidity and different maternal and healthcare factors, it is necessary to use record-level data. The individual records will include all women that gave birth in 11 London hospitals for the period 2015-2019 (i.e., the whole period for which data are available). To collect enough data to explore rare occurrences (such as maternal death or rare complications), all four years of available data for these hospitals are being requested.

To gain essential information on maternal morbidity and mortality, it will be necessary to link the MSDS (Maternity Services Data Set) dataset with the Hospital Episode Statistics dataset and the Civil Registration (Death – secondary care cut) dataset; this will allow to include in the analyses maternal morbidity and death that recorded after the first hospital discharge.

The data will be pseudonymised by NHS Digital and provided without an identification key.

To avoid accessing identifiable data (which in this case would be the date of birth of the baby and other related fields that could act as proxy for it), the study will include only derivations, computed by NHS Digital, of identifiable fields and of fields that could allow the reconstruction of identifiable data. As the data are pseudonymised and provided without any identification key, with no identifiable information provided to the researcher, re-identification is extremely unlikely and the data can be considered as pseudonymised. As such, the risk of harm for the individuals whose data will be accessed is very small. An important ethical concern regards the potential misrepresentation of any results as supportive of ideas of the biological reality of racial categories (which are scientifically unsupported) and even racial superiority. This is a risk that needs to be considered whenever a study addresses ethnic differences. To address this risk, extreme care will be taken on how the results will be presented to avoid misrepresentation by the media and the public. There will also be no attempts made by the study team to re-identify any individual.

Access to sensitive data in the form of ethnic origin and health information, for this study, is necessary “for the purposes of identifying and keeping under review the existence or absence of equality of treatment between groups of people specified in relation to that category with a view to enabling such equality to be promoted or maintained” (Data Protection Act 2018, part 2 [8]). Using available data is the least intrusive way to obtain the information required for this project.

DATA MINIMISATION:
1) Including only data for 11 hospitals in London.

2) Including only fields that are likely to be important for the study. The chosen fields include either variables that have already been identified as related to the risk of maternal mortality and morbidity (e.g., maternal age, height, weight, smoking status, socio-economic background, ethnicity…), or variables that can highlight ethnic diversity in the process of labour (e.g., length of different stages of labour, foetal presentation), or variables that refer to aspects of maternal care that could affect the process of labour and the risk of maternal morbidity (e.g., use of anaesthesia, forceps, waiting time before emergency c-section, genital tract trauma…). In addition, a small number of fields related to fetal health have been selected as potentially relevant to the process of labour, labour interventions, and potential ethnic differences in outcome. As this is a pilot project that explores new potential predictors of ethnic differences in maternal mortality and morbidity, reducing the number of fields or cases further could increase the likelihood of missing important information and limit how informative the project will be.

3) Duration of second stage of labour = *BabyBirthDateTime - *LabourOnsetSecondStageDateTime (this calculation is essential to calculate the length of the second stage of labour, for which there is some evidence of ethnic variation. However, this would only be possible if it is possible to use the time of birth for the calculation. The time appears to have been subjected to pseudonymisation, if it is not will be possible to reverse it for the calculation. Researchers will use a less effective measure as proxy for this calculation, which is listed as derivation number four).

Although the selection of requested fields includes fields that could be identifiable (LabourThirdStageEndDateTime, LabourOnsetSecondStageDateTime, BabyBirthDateTime), the identifiable content of these fields will not be needed, only some derivations (as agreed with NHS Digital).
To avoid the need to access identifiable data (i.e., the fields that can reveal the date and time of birth), the fields listed below and highlighted by an asterisk will not be required in their original form, but as the results of the following calculations:
1) Duration of first stage of labour = *LabourOnsetSecondStageDateTime - *LabourOnsetDateTime
2) Time between admission and second stage of labour = *LabourOnsetSecondStageDateTime - *StartDateTimeMotherDeliveryHPS
3) Duration of second stage of labour = *BabyBirthDateTime - *LabourOnsetSecondStageDateTime (this calculation is essential to calculate the length of the second stage of labour, for which there is some evidence of ethnic variation. However, this would only be possible if it is possible to use the time of birth for the calculation. The time appears to have been subjected to pseudonymisation, but it may be possible to reverse it for the calculation. If not possible, a less effective measure will be used as proxy for this calculation, which is listed as derivation number five).
4) Approximate duration of second stage of labour = *LabourThirdStageEndDateTime - *LabourOnsetSecondStageDateTime
5) Time between admission and rupture of membranes = *ROMDateTime - *StartDateTimeMotherDeliveryHPS
6) Time between admission and administration of oxytocin = *OxytocinAdministeredDateTime - *StartDateTimeMotherDeliveryHPS
7) Time between admission and decision to deliver = *DecisionToDeliverDateTime - *StartDateTimeMotherDeliveryHPS
8) Time between admission and Caesarean section = *CaesarianDateTime - *StartDateTimeMotherDeliveryHPS
9) Time between labour onset and admission = *StartDateTimeMotherDeliveryHPS - *LabourOnsetDateTime

The speculated cohort size is 200,000.

Yielded Benefits:

This is a new study and therefore there are no yielded benefits to report

Expected Benefits:

The aim of this pilot research is to evaluate whether the current model of “normal” childbirth used in midwifery and obstetric training, which is based on studies of white women’s physiology, could be discriminatory against non-white women and contribute to the large maternal mortality and morbidity gap in the UK.

The data is expected to allow to calculate the ethnic gap in maternal mortality and morbidity within highly diverse areas of London, in hospitals that care for a large portion of BAME women. The project is then expected to test whether there are significant differences between ethnic groups in key aspects of labour, such as the duration of different stages and foetal position. Its is hoped, these data will allow to evaluate whether the current model of “normal” childbirth, based on white women’s physiology, is representative of normal variation in other ethnic groups. If black women on average tend to have a longer first stage of labour, for example, they might incur a higher rate of interventions simply because their normal variation is not well represented by the current white-centred model of childbirth. Indeed, this project aims test whether ethnic diversity in labour characteristics could explain ethnic differences in rates of interventions. Finally, it is hoped the data will allow to test whether higher rates of interventions in BAME women could explain differences in maternal mortality and morbidity, while taking into account other risk factors.

Through this set of analyses, the project aims to explore a potentially important link between a white-biased model of normal childbirth, which includes expectations of labour progression based on white women’s physiology, and increased maternal morbidity/mortality in BAME women. While this pilot project cannot provide evidence of causality, it may highlight whether ethnic differences in labour should be explored further, as potentially contributing to the higher maternal mortality and morbidity in BAME women (possibly via higher rates of interventions).

The results of this study could lead to the development of a more inclusive set of expectations for the normal progression of labour in women of all ethnicities, and to the update of obstetric and midwifery textbooks to better represent women’s variation. The long-term aim is to improve maternal case for BAME women, and reduce the maternal health ethnic gap.

As the project is only at the pilot stage, it is impossible to give a clear timeline for these achievements. The plan is to start a larger follow-up study in 2023, with the aim to provide specific recommendations on how to make maternal care more inclusive and to produce new criteria for the definition of normal labour, which encompass women’s variation across ethnic groups.

This pilot study hopes to explore the effects of different factors on the ethnic gap in maternal mortality and morbidity. In particular, to investigate the link between a white-biased model of normal childbirth and increased maternal morbidity/mortality in BAME women. While this pilot project cannot provide evidence of causality, it may highlight ethnic differences in labour and in maternal care provisions that need to be explored further, as potentially contributing to the higher maternal mortality and morbidity in BAME women.

The study is in the public interest as an understanding of the factors involved may help reduce the gap in maternal outcomes. It is anticipated the benefits of this study will be achieved through a larger follow-up study, planned for 2022, which may provide:
- recommendations for a more inclusive set of expectations for the normal progression of labour in women of all ethnicities, based on a large and diverse sample of women
- recommendations for the update of obstetric and midwifery textbooks, to better represent women’s variation for maternal care textbooks

These recommendations may have a positive impact on:
1) mothers, especially BAME women
A better understanding of women’s variation in childbirth may help improve maternal care provisions for BAME women, moving away from a white-centred model. This research has also the potential to reduce the ethnic gap in maternal morbidity and mortality, by identifying some of the factors involved.

2) maternal care teams
By providing new information on normal variation in labour across diverse ethnic groups, the project hopes to provide useful information for maternal care providers. Specific recommendations for the update of maternal care guidelines may be produced during the larger follow-up study.

3) maternal care educators
Similarly, an increased understanding of women’s variation in labour and of the potential consequences of this unacknowledged variation on maternal care and maternal health may help update maternal care textbooks and syllabi. It is hoped, new evidence stemming from this research will be disseminated in several ways, including in formats that can be easily integrated in textbooks and lecture/workshops presentations.

Outputs:

All published results will be at the aggregate level and only used to draw group-level inferences in line with the HES analysis guide.

Study findings will be disseminated in the following ways:
- peer-reviewed academic journals (e.g., BMJ, Lancet Public Health),
- lay articles (e.g., for The Conversation)
- lay summaries though social media.

Relevant findings will be shared with:
- policy makers (e.g., through a report to the Maternity APPG, planned for later in 2022),
- health professional bodies (e.g., Royal College of Midwives and Royal College of Obstetricians and Gynaecologists),
- educators
- other stakeholders, including activist groups (such as FiveXMore). Roehampton University are in touch with the activist group FiveXMore, with the aim of getting FiveXMore involved in both the design of the next step of the project as well as planning of the dissemination of outputs to a wider public. It is early stages, though, as the NHS data analysis is effectively a pilot study to see if it is worth planning a wider project. Roehampton University do not expect direct benefits at this stage, only after the larger part of the study.

Depending on the relevance of the results, the results might be presented at academic conferences and outreach events. It is anticipated that communication of the results through these channels will occur in 2023.

The results are expected to form the basis for the development of a larger follow-up project, with a planned funding application in 2023. Any follow-up projects will be subject to separate DARS Agreement.

Processing:

There will be no flow of data from Roehampton University to NHS Digital. Roehampton University is requesting NHS Digital to provide a dataset that includes linked pseudonymised data from the MSDS, HES and Civil Registration (Death – secondary care cut) datasets. NHS Digital will transfer the pseudonymised data to Roehampton University without a re-identification key. No other organisations will be involved in the flow of data.

The analyses will be based on specific minimisation criteria:
- only MSDS data from 11 London hospitals will be included;
- HES and Death data will only be included in the analyses for entries that relate to hospital episodes and deaths that occurred within 6 weeks from a birth event included in the MSDS dataset;
- To avoid accessing identifiable data, the study will include only derivations, computed by NHS Digital, of identifiable fields and of fields that could allow the reconstruction of identifiable data. As the data are pseudonymised and provided without any identification key, with no identifiable information would be provided to Roehampton University. No identifiable data will flow from NHS Digital, instead, specific derivations will be calculated for such fields (as listed in the application).

Data obtained in pseudonymised form (without a key) will be stored on Roehampton University's secure computer network. It will be accessed only by the PI and for the purpose of this research project. The PI is substantively employed by Roehampton University and has received data protection and confidentiality training, as well as information security training.

The data will be encrypted and stored in a single machine, where all analyses will be performed by the PI; the machine will be password-protected and located in a secured room with very limited access to a small number of staff, and it will not be connected to the wider university network. Only results and graphs produced at the aggregate level with small numbers suppressed will be exported in line with the HES Analysis Guide to be used for publication and shared with collaborators and stakeholders.