Introduction

Potentially avoidable brain injury during perinatal care can cause devastating consequences for babies and their families,1,2,3 along with significant lifetime costs for health and care services.4 In the UK, successful clinical negligence claims for perinatal brain injury can result in settlement costs exceeding £20 million per avoidable case of cerebral palsy in the National Health Service (NHS).5,6 Perinatal brain injury concerns brain damage occurring during or soon after birth, including conditions such as intraventricular haemorrhage, perinatal stroke, and hypoxic-ischaemic encephalopathy (HIE).7 While its definition continues to be debated,8,9 HIE commonly refers to a brain injury caused by a series of fetal and neonatal insults around the time of birth10 that may be avoidable under conditions of optimal antenatal and intrapartum care.5,6,11,12,13 With estimates of average incidence ranging from 1.5 to 2.0 per 1000 live births,7,14,15 HIE is one of the leading causes of neurodisability and mortality in near-term and term babies.1,2

High quality data are needed for effective clinical audit and service improvement,15,16,17,18,19,20,21,22,23,24 and to enable understanding of the epidemiology related to incidence and risk factors for avoidable perinatal brain injury associated to HIE.7,15,19,20,21,24,25 Such data are especially important in identifying unwarranted variation across settings and over time, establishing targets for improvement, learning from positive deviance, building feedback loops with healthcare units about their performance, and informing the design, development and testing of improvement efforts.16,17,18,21,22,23,25,26,27 While data on perinatal care are routinely collected in the UK,19,20,28,29 challenges remain in relation to reliability and integration,20,30,31 inclusion and presentation of data most relevant to families,26,32,33 and deployment in data-driven improvement efforts.21 These challenges are similar to those in other national and international efforts to improve reporting and use of routinely collected data on perinatal injury,22,24,34,35,36,37,38,39 including HIE.10,25

A range of national data sources (i.e. datasets and audits) captures data relevant to perinatal brain injury associated to HIE in the UK.19,20,28,29,30,31,40,41,42,43,44,45,46,47 However, it remains unclear which data items relevant to incidence or risk factors are currently available across the data sources, and whether the data definitions used across the various data dictionaries are comparable.48 It is also not clear how challenges in collecting, recording and reporting of data might best be addressed. Accordingly, we aimed to:

  1. 1.

    identify and compare data items and their definitions relevant to monitoring incidence or risk factors in data dictionaries of perinatal data sources in the UK,

  2. 2.

    elicit multi-professional views on data items and an avoidable brain injury definition most relevant for an integrated UK data source, and

  3. 3.

    explore multi-professional views on challenges and optimisation of capturing and using routinely collected data in the UK.

Methods

As part of a quality improvement effort (see Acknowledgements), we conducted a review of relevant data dictionaries and consulted a multi-professional group.

Review of data dictionaries

National data sources were selected based on their relevance and size, including the Maternity Services Data Set (MSDS),28,31,40 the Hospital Episode Statistics for Admitted Patient Care (HES APC),29,41,42 the National Neonatal Research Database (NNRD),19,43,44 and the National Maternity and Perinatal Audit (NMPA)20,30,45 (Table 1). The data dictionary of the National Neonatal Audit Programme (NNAP)46,47 was not separately reviewed as it uses data extracted from the same source as the NNRD (Table 1).

Table 1 Main data sources (i.e. dataset and audits) used in UK perinatal care settings in the National Health Service (NHS) that include routinely collected data relevant to clinical indicators or risk factors for avoidable perinatal brain injury associated to HIE.
Full size table

The origins of the reviewed data sources are variable (Table 1), including “warehouses” of routinely collected data managed by NHS England. The MSDS, for example, re-uses clinical and operational data from the time of booking appointment to discharge from maternity care, while the HES APC includes data on episodes of care for patients admitted to hospital. By contrast, the NNRD, which captures specific data items about the care of all babies admitted to NHS neonatal care across England, Scotland, Wales and the Isle of Man, is managed by a research group based in a university. Finally, the NMPA, led by a clinical consortium including the Royal College of Obstetricians and Gynaecologists, is a large-scale national clinical audit that assesses specific care processes and outcomes using linked maternity data from the MSDS, HES APC and/or NNRD.

We reviewed the data dictionaries of the MSDS, HES APC, NNRD and NMPA to:

  1. (i)

    identify which of their data items are relevant to clinical indicators or risk factors of avoidably perinatal brain injury associated to HIE, and

  2. (ii)

    assess homogeneity of definitions of similar data items across the data sources.

We manually searched the data sources’ dictionaries and associated International Classification of Diseases 10th revision (ICD-10) codes. We mapped the identified data items to a longlist of potentially pertinent clinical indicators and risk factors, which we generated based on authorial expertise, national guidelines, and literature review.6,7,24,49,50,51,52,53,54,55,56,57,58,59,60 This longlist included potential clinical indicators of injury (e.g. Apgar scores, resuscitation, therapeutic hypothermia), antenatal risk factors (e.g. tobacco smoking, suspected fetal growth restriction, pre-eclampsia), and intrapartum risk factors (e.g. gestational age, maternal pyrexia, mode of birth).6,7,49,50,51,52,53,54,55,56,57,58,59,60,61 Findings were summarised in colour-coded tables with clinical indicators (Table 2) or antenatal/intrapartum risk factors (Table 3).

Table 2 Clinical indicators of avoidable brain injury associated to hypoxic ischaemic encephalopathy (HIE): potentially relevant data items that were identified in one or more of the data sources.
Full size table
Table 3 Risk factors of avoidable brain injury associated to HIE: potentially relevant data items that were identified in one or more of the data sources.
Full size table

Multi-professional consultation

We conducted a consultation with a multi-professional group with expertise in a range of relevant fields, including neonatal care, maternity care, statistics, and clinical negligence. Using individual interviews, the goal was to explore views on relevant data items and definitions of avoidable injury as well as current challenges in capturing data (i.e. collecting, recording and reporting data), including:

  • definitions of avoidable perinatal brain injury

  • data items indicative of presence of avoidable injury

  • data items on antenatal and intrapartum risk factors relevant to avoidable injury,

  • feasibility of reliable data collection across healthcare units, electronic patient record platforms and data sources, and

  • use of data to support improvement in care.

We employed a purposive sampling technique, with the aim of achieving a diversity of professional perspectives.62 Discussion among the authorial team (consisting of a range of experts on perinatal brain injury and data collection) had identified neonatologists, obstetricians, midwives, solicitors, statisticians, and academics as having expertise most relevant for the topic of the consultation. Suitable participants were identified using the authorial team’s professional networks and knowledge of the field. The principle of information power was used to confirm that the consultation captured a sufficient range and depth of views.63

Trained interviewers with a clinical background in neonatal or maternity care (VK, KW, MS, LD, RW) used a prompt guide (Supplementary Material 1), which had been developed through group discussion and pilot testing among the authorial team. The interviews (all conducted online and lasting 22–90 min) were transcribed verbatim and anonymised. Analysis used a thematic framework based on the prompt guide and on cross-cutting themes generated through inspection of the transcripts (see details in Supplementary Material 2).

Review by an institutional review board/ethics committee was not required, as the consultation was an engagement exercise classified as a quality improvement activity,64 in which all participants were invited to join the authorial team or a contributor group if they wished to be named. All participants provided written informed consent, including permission to publish anonymised quotes and synthesis of their expressed views.

Results

Review of data dictionaries

Tables 2 and 3 show that none of the four reviewed data sources on its own captures the range of items potentially relevant to avoidable perinatal brain injury. For example, the two sources that capture data from maternity services (MSDS and HES APC) do not systematically include clinical indicators for brain injury (Table 2), since maternity settings are not typically where the diagnosis is made. In contrast, the NNRD does hold a range of clinical indicators of brain injury from all babies admitted to UK neonatal care (Table 2), but is limited in capturing antenatal and intrapartum data items (Table 3). Moreover, the NNRD does not capture data on babies with mild HIE (a condition that may result in impaired brain development in up to one in five babies)65 who are usually managed on postnatal or paediatric wards rather than being admitted to neonatal care. This likely results in under-reporting of mild HIE and associated risk factors.

Ongoing linkage between the data sources to enable continuous monitoring of integrated data is lacking. The NMPA has used local maternity records, HES APC data, and NNRD data to link a range of relevant data from mothers and babies across set periods (e.g. in 2015–2017 using local records and NNRD data and later in 2017–2019 using MSDS and HES APC, see Table 1). However, this practice is not routinised to ensure ongoing linkage of all available data of maternal and neonatal data. Further complicating the picture, operational definitions of similar data items differ across the data sources.

Further, most intrapartum risk factors for perinatal brain injury identified by National Institute for Health and Care Excellence (NICE)54 – including fetal heart rate features, delay in labour, meconium-stained liquor, vaginal bleeding, maternal pyrexia, and tachysystole49,50,51,52,53,54 – are not consistently captured within any of the data sources we reviewed (Table 3).

Multi-professional consultation

Twenty-seven participants took part in the interview consultation (Table 4). Analysis of their views generated seven themes relating to current challenges in using data to guide improvement efforts aimed at reducing risk of perinatal brain injury (Table 5).

Table 4 Professional background of the multi-professional group (N = 27) taking part in the consultation.
Full size table
Table 5 Multi-professional views (N = 27): synthesis relating to addressing current challenges to using data to guide improvement efforts aimed at reducing risk of perinatal brain injury.
Full size table

The participants viewed the current lack of a standardised, consensus-built definition of avoidable perinatal brain injury as an important problem in effectively monitoring incidence and risk factors (Table 5 – theme 1). Frequently suggested definitions focused on HIE (Table 6), which was seen as a potentially avoidable condition under ideal care conditions. Related or other frequently suggested definitions referred to causes of avoidable brain injury such as absent, poor, late or suboptimal clinical management during antenatal, intrapartum or perinatal care (Table 6). Participants noted that questions relating to avoidability should consider whether the outcome would have been different in a different healthcare unit, with different healthcare professionals, or using different clinical practices, while remaining cognisant of the challenges in determining avoidability.

Table 6 Participants’ recommendations when asked to characterise avoidable perinatal brain injury.
Full size table

Participants identified the need for agreement on a set of data items relating clinical indicators and risk factors to generate an integrated data source on avoidable brain injury (Table 5 – theme 2). They suggested a large range of clinical indicators and risk factors to consider incorporating in an integrated source (Table 7). Some of these are already captured in current data sources as identified in our data dictionary review (see Tables 2 and 3), while others – particularly some intrapartum risk factors as recommended by (NICE)54 – are not. Participants strongly recommended that follow-up of babies with potential brain injury should be considered, since injury or developmental challenges may present some time after birth, even if there is no early indication of injury. They acknowledged that long-term follow-up may be challenging, and recommended exploring the use of existing community services (e.g. health visitors) and parents’ reports.

Table 7 Participants’ recommendations when asked to suggest data items relevant to monitoring incidence or risk factors for avoidable perinatal brain injury.
Full size table

To enable and optimise an integrated data source, participants recommended addressing the duplication, inconsistency and subjectivity involved in capturing many data items relevant to brain injury (e.g. Apgar scores, therapeutic hypothermia criteria, fetal growth restriction, intrapartum risk factors) (Table 5 – theme 3). This would require a programme of work needed to operationalise and standardise operational definitions that are currently not harmonised across data sources and associated electronic patient record (EPR) platforms (Table 5 – theme 4). Further work would be needed to improve consistent, reliable and feasible data collection (Table 5 – theme 5), ranging from making the data collection system more user-friendly through to real-time capture of data rather than retrospective data inputting (Supplement 3), and training professionals involved in data capture (Table 5 – theme 6). Finally, participants identified the need for co-designing systems with healthcare professionals and families to improve use of data, including generating feedback loops that are meaningful to healthcare professionals and families based on sensible comparative methods (Table 5 – theme 7).

Discussion

This review of UK perinatal data source dictionaries and multi-professional consultation identified challenges and opportunities in capturing and using data relating to avoidable perinatal brain injury associated to HIE. It shows many limitations of current data sources: none of them on its own captures the range of items relevant to injury, key intrapartum risk factors are not consistently captured, operational definitions of similar data items differ across data sources, and a mechanism for ongoing linkage of data across data sources is absent. Our consultation suggests that a first step towards improvement is to develop consensus on a standardised definition of avoidable perinatal brain injury and the range of clinical indicators and risk factors that should be captured in an integrated UK data source. A second step is to address current inconsistencies in capturing data across different data sources and healthcare units, with a focus on enabling systematic and ongoing linkage and integration of available data. Third, systems for better use of data should be co-designed with both healthcare professionals and families.

The challenges we identified are likely linked to the definitional morass that characterises the international and national field both in clinical guidelines and the wider literature,10,39,52,54,61,66,67,68 absence of UK standardisation in data collection and reporting,21,30,69,70 and fragmentation of data across different UK data sources and care pathways.30,71 A consensus-built integrated data source, based on a standardised definition for avoidable perinatal brain injury that could be informed by clinical criteria for HIE,8,9,10,11,12,13,67 would provide the foundation for identifying and defining risk factors potentially amenable to practice improvement. An integrated data source would also help in assessing the extent to which clinical indicators and risk factors for injury might be patterned by structural factors such as socio-economic disadvantage and ethnicity.72,73 Consensus work on a standardised definition and a set of data items for an integrated data course could be informed by ongoing definitional efforts in neonatology74,75 and neonatal encelopathy.8,9 It should include engaging with healthcare professionals and families to generate a data source that includes data items that are meaningful to professionals and families.8,9,26,32,33,70,74,75,76,77,78 The process and outcomes of this UK-focused work may inform other national and international perinatal audit processes and quality improvement efforts that also face challenges of limited standardisation or data fragmentation.10,22,24,27,35,37,39,59,79

Our consultation showed strong support for better and continuous linkage and integration of data captured across various UK data sources, electronic patient record platforms, and healthcare units. This resonates with previous findings of the NMPA30 and the British Association of Perinatal Medicine (BAPM).46 Linkage between data sources currently occurs on an ad-hoc basis, highlighting the challenges in consistently linking and integrating data from different sources.30 Although awareness is growing, further action by commissioners and policy-makers is needed to address the problem of electronic patient record platforms with different styles and formats, which risk worsening heterogeneity in data capture across maternity or neonatal units.46 Better aligned national data sources might also help improve international harmonisation efforts on perinatal data.24,35,38

Essential to the formation of an integrated data source is exploring possibilities for modern technology to facilitate safe data input and transfer across different platform providers. This is in line with national ambitions to create a more digitised and learning-focused NHS, guided by the principles of user need, privacy and security, interoperability and openness, and inclusion.80 How data should be extracted for local, national and potential international purposes requires further consultation, including families, maternity advocates, healthcare professionals, audit specialists, quality improvement practitioners, software developers, academics, and others who might be tasked with collecting, recording or reporting data.

Another key consideration for design of an integrated data source relates to when, how and by whom the data will be collected. Currently, complex transitional care arrangements differ between UK neonatal units,71 leading to variation in how neonatal data are captured in one of the current datasets and on whom. For example, in some cases babies are admitted under neonatal care and therefore have a neonatal dataset record with details captured by the NNRD. In other cases, data for the baby may only be entered into the mother’s local electronic patient record, with selected data items about the birth (e.g. Apgar score, resuscitation, seizures) subsequently captured by the MSDS. The need to address the “who” of data collection is also relevant to limited use of parent-reported symptoms of perinatal brain injury, which is an area in need of further exploration.

The development of an integrated UK data source built on consultation or co-design would support a learning health systems approach to improvement.81 For example, availability of a national integrated data source, coupled with a co-designed approach to data feedback and visualisation at local and national level, could provide the means for iterative, data-driven improvements through learning from routinely collected data.81 Despite the significant potential of a learning system approach,81,82 including for perinatal care,83 it is yet to be implemented in national perinatal settings.82,83 Investments would be needed, such as in resources and personnel required to monitor local, regional and national trends in brain injury incidence. These investments may not only help with better data capture, but also build professionals’ capability to reshape care, including reducing variations in quality and safety that contribute to avoidable brain injury.3,5,6

Strengths and limitations

This paper offers insights into the potential for current UK perinatal data sources to support improvement in care relating to avoidable perinatal brain injury associated to HIE, including the generation of an integrated UK data source based on harmonised definitions. It also synthesises the views of diverse professionals regarding challenges and suggestions for developing such a data source. Its methods may be useful for other clinical areas in which national or international sources of routinely collected data could inform healthcare improvement activities, but this requires further evaluation.

The review is limited by its focus on national data sources (that is, not reviewing data captured in unit-level databases), and an inability to collect accurate information on the quality, completeness and sample size of data captured in the data sources. Although the multi-professional consultation included a varied sample with a multitude of views that resonate with recommendations by other expert groups,30,46 it is possible that wider consultation could further strengthen or expand the conclusions in this paper. Families and those affected by perinatal brain injury were not involved during this initial phase owing to its technical nature, but will be crucial to a next phase of work of a programme to develop a data source that addresses the priorities and views of families and professionals.

Conclusion

This study has identified challenges in monitoring incidence and risk factors of avoidable perinatal brain injury in the UK, including problems in standardisation and integration of routinely collected data across different sources. These challenges limit the value and utility of current data sources as a basis for improvement efforts. A strategy including consultation, consensus-building, and co-design with healthcare professionals, families, and other specialists is needed to harmonise data definitions and ensure reliable and feasible ways to capture data. This strategy should aim to ensure that all perinatal events are reported using the same definitions and according to appropriate quality standards. Work to develop a high-quality integrated source should be accompanied by co-design to enable use of data to understand problems, generate solutions, and test solutions as part of improvement efforts.