The efficient use of the maternity workforce and the implications for safety and quality in maternity care: a population-based, cross-sectional study

Article history

The research reported in this issue of the journal was funded by the HS&DR programme or one of its proceeding programmes as project number 10/1011/94. The contractual start date was in March 2012. The final report began editorial review in October 2013 and was accepted for publication in March 2014. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HS&DR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

none

Copyright statement

© Queen’s Printer and Controller of HMSO 2014. This work was produced by Sandall et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Introduction

The chapter commences with an overview of drivers for high-quality maternity care, followed by a discussion of evidence relevant to defining and measuring quality and safety in maternity care, use of routine data, maternity health-care workforce, quality and safety indicators and health-care workforce and efficiency, and other literature to inform study aims and objectives.

Several bibliographic databases, including PubMed, The Cochrane Library, the Cumulative Index to Nursing and Allied Health Literature (CINAHL), EconLit and that of the National Institute for Health and Care Excellence (NICE), were searched for relevant primary and secondary studies and guidelines. Search terms included ‘maternity care’, ‘safety’, ‘quality’, ‘outcomes’, ‘midwifery care’, ‘obstetric care’, ‘acute hospital settings’, ‘maternity workforce’, ‘cost effectiveness’, ‘efficiency’, ‘production function’ and ‘stochastic frontier’. No date limits were used and only studies published in English were considered. Relevant policy reports were identified from searches of relevant websites of government and professional organisations.

As a systematic review was not conducted, a formal search strategy was not developed; however, priority was given to evidence relevant to UK-based maternity care, including studies undertaken in high-income countries with similar maternity systems (i.e. where midwives and obstetricians were responsible for providing care) which considered quality and safety of maternity care. Evidence relevant to health-care workforce issues that may or may not have been conducted in UK maternity settings was also considered.

High-quality maternity care

All NHS providers have a mandate to enhance the quality of patient care, and the performance of maternity services has been viewed as a window into whether or not quality health services in general are being delivered. 1 The Department of Health’s National Service Framework for Children, Young People and Maternity Services,2 with its 10-year time frame for implementation, and Maternity Matters3 were consistent in the commitment to deliver a choice of safe, accessible, high-quality maternity care which was women focused and family centred. Underpinning principles included the view that pregnancy and birth are normal life events, maximising the opportunity for all women, regardless of risk profile, to have as physiological and positive a birth experience as possible. The coalition government has not developed or published a cohesive maternity policy but has published indications of its commitment to women’s choice of maternity care,4,5 continuity, and improved outcomes. 6

The professional commitment to improving the quality of maternity care, and sustaining the workforce to achieve this, has been led by the relevant Royal Colleges, including the Royal College of Midwives (RCM) and Royal College of Obstetricians and Gynaecologists (RCOG), often with joint College publications. In 2007, a joint report was published by the RCOG, the RCM, the Royal College of Anaesthetists and the Royal College of Paediatrics and Child Health (RCPCH) to provide guidance to develop equitable, high-quality standards for UK maternity care. 7 Standard 30 (out of 30) focused on provision of a high-quality workforce and promotion of appropriate leadership, skill mix and competencies in midwifery, obstetrics, anaesthetics and paediatrics. A 2009 position statement from the RCM on staffing standards in midwifery services to assist commissioners and providers, endorsed by the RCOG and RCPCH, recommended a minimum ratio of 1 midwife per 28 births per year. Falling below this ratio would be a strong indication that a service should undertake a thorough workforce review. 8 The RCOG report High Quality Women’s Health Care9 emphasised how high-quality care could promote health over the life course for women and their infants. It proposed changes that focused ‘on the needs of the woman and her baby by providing the right care, at the right time, in the right place, provided by the right person and which enhances the woman’s experience’ (Foreword, page iv), and the fundamental role that midwives have in delivering high-quality health care also continues to be recognised. 10

Defining quality

One of the key issues to address in any health system is how to derive robust, appropriate and usable measures of the quality and safety of care and measures of outcomes of care which resulted in harm. The ideal measure should be easy to define and observe, should reflect priority outcomes for patients, clinicians and service providers and should identify those areas where quality of care/outcome could be improved.

The Institute of Medicine11 identified six dimensions of quality: namely that health care should be safe, effective, patient-centred, timely, efficient and equitable. Some of these have been widely adopted within the UK. 12–14 The current UK government has taken the approach that process indicators or targets are unnecessarily bureaucratic and distract from the important objectives of improving safety, reducing morbidity and improving patient experience more broadly, and has therefore focused quality measurement on evidence-based clinical outcomes. 15 The resulting NHS Outcomes Framework provides a national overview of the quality of NHS care, provides accountability and acts as a catalyst for driving quality improvement involving five domains:16

1. preventing people from dying prematurely

2. enhancing quality of life for people with long-term conditions

3. helping people to recover from episodes of ill health or following injury

4. ensuring that people have a positive experience of care

5. treating and caring for people in a safe environment and protecting them from avoidable harm.

The NHS Outcomes Framework for 2013/14 introduced new outcome measurements. 6 Improvement areas in maternity care and their relevant indicators include:

• reducing deaths in mothers (or at least maintaining the low level) (domain 1)

• reducing deaths in babies – neonatal deaths and stillbirths (domain 1)

• helping women to recover from ill health or injury following birth (domain 3)

• improving women and their families’ experience of maternity services – women’s experience of maternity services (domain 4)

• treating and caring for people in a safe environment and protecting them from avoidable harm (domain 5).

The maternity indicators identified by the Outcome Framework, whilst important in determining quality of care, cover only a small aspect of maternity services, and other bodies have also outlined key indicators. For example, the UK-wide Midwifery 2020 recommended in its report Delivering Expectations13 a number of additional indicators to measure the quality of midwifery care. These included reducing perineal trauma; uninterrupted skin-to-skin contact between mother and baby following the birth; continuity of midwife care; and increasing the normal birth rate, using the definition of normal birth published by the Maternity Care Working Party in its consensus statement. 17 The Commissioning for Quality and Innovation (CQUIN) framework was introduced in 2009 to enable commissioners of health services to reward excellence by giving financial incentives to local health-care providers to deliver nationally agreed quality improvements and better outcomes for patients. There are four National CQUIN Scheme goals in 2013/14. 18 These are the Family and Friends Test (introduced for maternity in October 2013), dementia care, venous thromboembolism and the NHS Safety Thermometer,19 of which there is a maternity version currently in development by the NHS Quality, Efficiency and Support Team (NHS QUEST). Local CQUIN schemes agreed with local commissioners should be in place in late 2013, and collaboration is encouraged where contracts are made with several commissioners. Commissioners are encouraged to use appropriate existing indicators such as the Indicators for Quality Improvement (IQI),20 Advancing Quality Alliance,21 the NHS Atlas of Variation and NICE Quality Standards. 22 Maternity specific indicators used for CQUIN include IQI indicators for smoking cessation during pregnancy, prevalence of breastfeeding at 6–8 weeks and access to maternity services by 12 weeks + 6 days. 20 Indicators of quality, therefore, should measure a balance of aspects of harm or adverse outcomes, care which promotes health and patient-derived measures of the experience of care.

Measuring quality

Measuring quality of health care is not a new concept, and Donabedian described his approach in a paper published in the Journal of the American Medical Association. 23 It was based on three components: structure, process and outcome, which were viewed as inter-related. Structure referred to the conditions under which care was provided, including staffing levels and mix, facilities and equipment, and organisational characteristics such as supervision and performance review. Process referred to activities carried out and care provided, such as diagnosis, treatment and education. These activities are often carried out by staff. Outcomes referred to changes – good or bad – in individuals which could be attributed to the health care received. Donabedian differentiated these measures from actual aspects of quality, instead considering them to be alternative types of information which could be used to infer or indicate good quality. He highlighted the necessity of using these only when there was a relationship of cause and effect between the three components.

The King’s Fund report Getting the Measure of Quality24 still considered this approach to be useful in developing quality indicators, although the RCOG has pointed out that some obstetric indicators are hard to classify in this way25 – for example, a caesarean may be a process initiated by a clinician or an outcome following another intervention and may have a good or bad effect (or both) on the health of two individuals, mother and baby.

Structural characteristics – the way a health-care system is organised – may have an important impact on the quality of care provided. The Francis Report26 has highlighted how failings in management and governance caused serious failings in the processes and outcomes of care, but the direct relationship between these can be difficult to assess.

Where it is known that particular processes, such as continuity of care, are signifiers of quality of care, for example because they impact on some aspect of quality such as safety or patient experience, the measurement of the extent of that process can be used as an indicator of quality. 23 Process measures have the advantage of directly measuring care that is received by patients and potentially increasing the detection of poor care. They are capable of being measured contemporaneously, giving a more immediate assessment of quality. However processes can also be hard to measure and data may not be available. They can be subject to manipulation or ‘gaming’, particularly when performance is being assessed externally or financial incentives are at stake.

Clinical outcome measures have advantages over process measures in that they can assess the health outcomes (favourable or unfavourable) of patients who have received care. Outcomes are often routinely collected, making the data more readily available for analysis. They can be less subject to manipulation, but have the disadvantage of being affected by factors other than care, such as coding accuracy, disease severity, comorbidities and other independent characteristics or demographics. As it is not always possible to directly attribute outcomes to specific processes of care, in reality it may be more difficult than assumed to use outcome indicators to improve the quality of care in this way.

Measuring quality, safety and harm in maternity care

Although maternity care is the commonest reason for hospital admission among women aged 15–59 years in the UK,27 there is lack of agreement on what measures should be used. Safety measures developed for general populations often do not include measures appropriate for pregnant or postnatal women. 28 There are also methodological issues to address, for example population, context of care, data quality, variation in outcomes within and between units, and whether or not risk adjustment was used to address confounding factors.

It is increasingly recognised that measures traditionally used in maternity care such as maternal and neonatal mortality are essential, but also that other measures of quality of care are needed. 29 Studies to date which have reported the development of a quality measure for maternity care have tended to focus on patient-specific indicators (PSIs), such as primary caesarean delivery rate,30 or obstetric trauma in caesarean, instrumental and unassisted deliveries [used by the Agency for Healthcare Research and Quality (AHRQ)31] or aspects of patient satisfaction. 32

Several studies have considered the risk of interventions in maternity care (e.g. caesarean birth), which could inform their use as a potential measure of quality. Paranjothy et al. 33 from the NICE Collaborating Centre for Women and Children examined the variation in caesarean rates between maternity units looking at case-mix differences in a national prospective cross-sectional study. Data were collated from 216 maternity units in England and Wales on women who gave birth between May and July 2000. The relationship between case-mix characteristics and odds of a caesarean birth before or during labour was investigated using logistic regression models. Overall caesarean rates standardised for case mix were then calculated for each maternity unit. Heterogeneity between units was examined using random-effects meta-analysis. Adjustment for case-mix differences explained 34% of the variance in caesarean rates. The odds of having a caesarean birth before and during labour increased with maternal age. Women from ethic minority groups had lower odds of caesarean birth before labour and increased odds in labour. Women who had a previous vaginal birth had lower odds of caesarean, although the magnitude of this for caesarean before and during labour was markedly different. Findings showed that the variation in organisation of services, women’s preferences for mode of birth, staffing levels and clinician attitudes were all important factors to consider when quality and appropriateness of maternity care are evaluated.

Using routine data

Hospital Episode Statistics (HES) provides information on care provided by NHS hospitals and for NHS hospital patients treated elsewhere in England. It is the data source for a wide range of health-care analysis for the NHS, government and many other organisations such as the private benchmarking services CASPE Healthcare Knowledge Systems and Dr Foster.

Every episode of hospital inpatient care generates a patient record which includes demographic information along with details of the episode of care, diagnoses using International Classification of Diseases version 10 (ICD-10) codes34 and procedures such as operations using Office of Population Censuses and Surveys Classification of Interventions and Procedures (OPCS-4) codes. 35 Where a woman’s record additionally includes the delivery of a baby, limited further information on the birth is collected in a ‘maternity tail’ to the mother’s record. Individual HES records for the same person can be linked across time and providers to enable long-term patterns of care to be studied. The individual records of mother and baby are not linked.

Some researchers have compared outcomes of the US AHRQ PSIs, including obstetric trauma indicators, with routinely collected HES inpatient data, in an attempt to generate quality indicators for England. Raleigh et al. 36 compared UK and US data for nine PSIs using a case–control analysis of HES data for 2003–4, 2004–5 and 2005–6 for all English trusts. Length of stay and mortality between cases (patients experiencing the particular safety event measured by an indicator) and controls were matched for age, sex, health resource group (standard groupings of clinically similar treatments that use similar levels of health-care resource), main specialty and trust. They found some consistency in national rates for the nine indicators and, for all but one indicator, hospital stay and mortality were longer. The authors concluded that internationally comparative indicators could be derived from English data, although further validation was needed, and recording needed to be improved.

Bottle and Aylin37 also compared outcomes for nine AHRQ indicators (10 were originally selected but one – iatrogenic pneumothorax – was dropped for lack of equivalent ICD-10 code) for use in English routine data in relation to established measures of negative outcome including mortality. Using case-mix adjustment they found wide variations between trusts which were potentially due to inadequate adjustment, differences in coding definitions between trusts, or poor quality of coding. They concluded that the derivation of patient safety indicators from HES data were potentially useful for prospective evaluation of data quality.

Studies using HES data show that despite data quality issues, analyses using judicious cleaning and case-mix adjustment can be useful in identifying variations in patterns of maternity care. Some of the results showing the importance of case-mix adjustment using HES data have been validated using data from the Millennium Cohort Study. 38 This study showed that, for women having their first baby, operative birth rose with increasing maternal age, and that for all women mode of birth differed significantly by ethnicity.

Obstetric outcomes derived from HES inpatient maternity data from 146 English NHS trusts were analysed by Bragg et al. 39 in a cross-sectional analysis to ascertain if variation in unadjusted rates of caesarean births could be explained by maternal characteristics and clinical risk factors. The main outcome measure was rate of caesarean birth per 100 births (live or stillborn). The population included women aged 15–44 years who had a singleton birth between 1 January and 31 December 2008. The likelihood of women having a caesarean birth given their age, ethnicity, parity, socioeconomic status, deprivation status and clinical risk factors (including previous caesarean birth, breech presentation and fetal distress) were entered into a multiple logistic regression model.

A total of 147,726 (23.8%) women had a caesarean birth, which was more likely if they had a previous caesarean or a breech presentation. Elective and emergency caesarean rates by NHS trust were adjusted for maternal characteristics and clinical risk factors, using funnel plots to show significant variation between trusts. Maternal age, ethnicity and parity (particularly in multiparous women with a previous caesarean) were all significant factors in determining the likelihood of a caesarean birth. A number of clinical risk factors, including diabetes, hypertension and placental problems, also predisposed women to having a caesarean. However, adjusting for maternal characteristics and clinical risk factors did not greatly reduce the variation between individual trusts, with the observed variation in caesarean rates being 14.9–32.1%. This variation was largely due to emergency caesarean rates, rather than elective caesarean rates, which showed much less variation. The authors concluded that case-mix adjustment was necessary in order to compare caesarean rates between trusts, and that the remaining variation may be due to differences in clinical practice regarding emergency caesarean between trusts.

This work was furthered by the RCOG in its Clinical Indicators Project,25 which identified 11 potential performance indicators derived from HES data, the basis for selection being validity (reflecting quality of care), fairness, sufficient statistical power and ability to technically code the outcome adequately. In addition, the suite of indicators had to cover various dimensions of care to give a balanced picture of the service. The RCOG identified key issues with HES maternity data quality, particularly duplicate records, records not relating to deliveries and incomplete or inconsistent recording of data items. Despite extensive data cleaning to remove duplicates and records which did not relate to a delivery episode, identifying units with inconsistent or missing data and adjusting for case mix, this limited the use of some potential indicators. Results were stratified between nulliparous and multiparous women and adjusted for maternal age, ethnicity, social deprivation and a number of clinical risk factors such as previous caesarean, diabetes, hypertension, gestational age and birthweight. The results were shown in funnel plots and demonstrated a large variation in intrapartum processes and outcomes which could not be explained by random fluctuations. For example, among women giving birth for the first time, there was a twofold difference between hospitals with the highest and lowest rates of emergency caesarean after induction of labour (20% compared with 40%) and of instrumental birth (16% compared with 32%). The report concluded that further understanding of the unexplained variation is important in order to compare performance across trusts.

Organisational factors, such as trust configuration, size, models of care, staffing levels, skill mix, staff deployment and safety culture, remain unknown factors in the understanding of important influences of the quality and safety of maternity care, some of which are considered in the next section.

The maternity health-care workforce and quality and safety indicators

‘The performance of maternity services is a touchstone of whether we are delivering quality based on patient safety, effectiveness of care and patient experience.’40 This statement from the NHS Chief Executive, David Nicholson, followed the review of maternity services carried out by the Healthcare Commission [now the Care Quality Commission (CQC)] in 2007. This review raised key concerns that, in some trusts, levels of staffing were well below average and may have been inadequate. It recognised that staffing was a contentious issue, as it underpins the quality of the service but at the same time is the most costly element of providing that service. 41

In 2008 the report of the independent inquiry commissioned by The King’s Fund, Safe Births: Everybody’s Business,42 and the Healthcare Commission’s review of maternity services, Towards Better Births,41 identified areas in need of improvement, including staffing, training and communication. Staffing has been identified in numerous reports as being a critical component of safe, effective, patient-centred care. Staffing levels contributed to 3.5% of all reported safety incidents across the NHS43 with workforce factors likely to contribute to a far higher total proportion. This has resulted in the dilemma of maintaining, and ideally improving, the quality and safety of care in a maternity service facing greater demand and increasing complexity in the health of childbearing women.

Research from a number of sources, including studies of other health-care professionals, points towards better-quality care, improved outcomes or fewer adverse events being associated with higher levels of registered nurse staffing. 44–49 However, while reduced complication levels may be associated with reduced length of inpatient stay (improved productivity), the association between higher registered nurse staffing and reduction in stay is not universally supported. Kane’s systematic review found evidence to support such a relationship in intensive care and outcomes for surgical patients but not for medical patients, and further highlights complexity and the challenges of attributing cause. 50

There is also a body of literature on the optimal level of staffing for doctors from the USA51,52 and some evidence of a medical staffing outcome relationship from both the USA and the UK. 44,53 Other evidence that considers medical staffing suggests complex inter-relationships between workload, efficiency and quality. 54–58 However, this literature is more limited in extent than that on nurses, and there are also significant concerns with drawing causal inference from the extant literature to the UK maternity workforce.

The complementary and substitutability of nurses/midwives and doctors is even less well documented in large studies exploring routine practice (as opposed to experimental implementations). Outcomes may be sensitive to ratios between nurses and medical staff. For example, in the UK, a higher total of clinically qualified staffing (doctors + nurses) per bed and a higher number of doctors relative to the number of nurses were both associated with lower mortality-based failure to rescue in the fully adjusted analysis. 57

In maternity care, a survey of health-care professionals showed that many believed that low staffing levels have a direct impact on safety of maternity services as a result of increased error rates, burnout, tiredness and less direct care. 58 Respondents were of the opinion that higher midwifery staffing levels would allow all women to have one-to-one care in labour, reduce intervention rates, reduce postnatal hospital stays and release money to reinvest in services.

However, few studies have investigated the link between obstetric and midwifery staffing and outcomes. 42 Joyce et al. 59 drew on cross-sectional data from all 65 maternity units in the Thames region between 1994 and 1996, covering a total of 540,834 live births and stillbirths. After adjustment for birthweight, perinatal units with a more ‘interventionist’ approach (defined by higher rates of caesareans, epidurals and instrumental births) and higher levels of consultant obstetric staff were found to be associated with lower stillbirth rates; and this effect persisted after adjustment for other possible predictive and confounding factors. An analysis using HES 2008 data matched with staffing variables from the Maternity Matters Benchmarking Dataset found a relationship between higher levels of full-time equivalent (FTE) midwifery staffing and a lower chance of readmission at 28 days; however, risk adjustment was limited. 60 However, observational studies have limited capacity to identify causal pathways.

The NHS Operating Framework 2010/11 identified the need to help local managers to identify optimum skill mix for quality and productivity. 1 Birthrate Plus (BR+) is widely used in the UK to calculate the number of midwives required in a NHS maternity unit. 61 Despite the widespread use and recommended use of BR+, it is not known whether ratios or staffing establishment numbers reflect ‘the ideal’ or ‘what is current’ and how these are related to providing a high-quality and safe maternity service.

Strategic approaches to maternity support worker development are under way at a national level in Scotland, Wales and Northern Ireland. However, there is limited and inconclusive evidence that changing workforce skill mix or substitution of roles in maternity care and other acute or primary care settings is associated with improved health outcomes or a reduction in costs. Few, if any, studies have considered the potential trade-offs between staff groups to optimise quality and efficiency, nor have they attempted to explore differential effects on different outcomes simultaneously.

Health-care workforce and efficiency

The majority of the literature on the relationships between the health-care workforce and outcomes including efficiency and effectiveness is based within acute secondary care. Very little relates specifically to maternity services, although there may be lessons to learn. Work examined so far points to a relatively simple gradient of improving outcomes with more registered nurses, and improvements in both outcomes and cost-effectiveness with richer skill mix.

Moving beyond the nursing workforce, economic evaluations of nurse for doctor substitution (which could be construed, in part, as involving a dilution of skill mix) also suggest that such substitution can be cost-effective or lead to a net cost reduction. 62 The optimal use of scarce and expensive labour resources will depend upon whether or not they are complements or substitutes. There are two common approaches to this question used in the production economics literature: p-complementarity and q-complementarity. 63,64 Traditionally, p-complementarity is evaluated from a cost function, but in health-care applications cost data are not often available for all inputs. However, q-complementarity can be investigated via the production function, but is not often addressed. A rare example of this approach to health care is by Thurston and Libby,65 who estimated the staffing relationships for primary physician services in the USA. They found that nurses are q-complements for physicians, while technicians and unqualified nurse aides are q-substitutes for nurses, in the production of primary care visits. Economically, very little is known about the complementarity or substitutability of staff groups (skill mix) within the NHS, despite there being critical changes in the composition of the workforce over recent years. We have not found examples which address this important question from within any acute care settings.

Some of the economic models above also point to two conceptually distinct ‘outcomes’ for a given health-care team: quality (represented primarily by patient safety in the existing literature) and productivity (represented by volume of cases treated or length of stay). This is also embodied in the current NHS Quality, Innovation, Productivity and Prevention (QIPP) programme, which seeks to improve quality and productivity simultaneously, although it is not clear whether improvements in both are separate, linked or traded off. While there is considerable evidence on the benefits of investment in improved patient safety, very little is known about the impact on a health-care provider’s efficiency and output of diverting resources to this cause. 66

Cost-effectiveness and effective use of fixed resources involving alteration in the composition of the clinical team is clearly dependent upon wage differentials. Replication and extension of US findings in other health economies is clearly warranted. It also seems clear from the existing evidence that there is unlikely to be a general relationship between skill mix and quality/productivity that generalises across care settings. Furthermore, all the above-cited economic models are limited because the staffing variation observed in cross-sectional observational studies is assumed to be causing the differences that are observed. The effect of variation associated with nurse staffing is assumed to be accurately determined by parameters derived from regression equations, even though it is clear that neither costs nor outcomes are the result of a deterministic process.

In relation to economic evaluations of skill mix change and outcomes, this research in general is limited. 62 Jones et al. 57 noted that, while there are a few hospitals that have relatively low staffing levels but appear to produce good outcomes, there are hospitals with high staffing levels that appear to produce poor outcomes. This suggests that high staffing levels may be merely indicative of aspects of care, and existing economic models have simply presumed that the relationships observed are causal. However, it is unknown if reductions in staffing levels and mix would produce a corresponding reduction in outcome. Improving outcomes through staffing changes is not costless for health-care providers and standard microeconomic theory would suggest that they are subject to diminishing marginal returns. This notwithstanding, little is really known about the impact of variations of workforce and skills mix either positively or negatively in relation to health-care providers’ operational efficiency or the potential to substitute one grade of staff for another (e.g. nurses for doctors, health-care support workers for nurses, clinicians for managers, let alone midwives for obstetricians) and its impact upon outputs.

Some studies have considered the costs of providing maternity care and how costs vary between hospitals. Laudicella et al. 67 undertook an analysis using patient level data comparing obstetric departments between hospitals. They examined the effect of patient characteristics on costs and considered factors to explain differences in costs between hospitals. Using HES record data, they mapped costs to individual patients and found that costs were driven by women’s characteristics to a greater extent than was explained by the type of birth they had. Costs were higher for women who lived in an area of greater social deprivation or had a number of obstetric risk factors. Even after adjusting for maternal characteristics and the type of birth as identified by the Health Reference Group (HRG) code, they found large variations in costs of obstetric care. They proposed that these might arise from differences in coding practice, differences in how costs were apportioned within accounting systems or differences in efficiency.

Further work done at the Centre for Health Economics, York,68 considered cost and length of stay for women having babies. They found that older women, those with more risk factors, those from poorer areas and those having more complex births with interventions had a longer stay and higher costs. These factors have been rising for several years and continue to increase.

Task-shifting offers another possible route to cost savings. There is no robust evidence about the cost-effectiveness of maternity support workers. Several studies have sought to compare the costs of midwife-led care with consultant/medically led care. The studies use a variety of methods in their costing calculations and some include elements of ante- and postnatal care in addition to the intrapartum period. This makes it difficult to draw conclusions.

A comparative analysis of normal hospital birth in nine European countries confirmed the importance of labour costs and skill mix as determinants of total delivery costs. 69 While medical tests and drugs accounted for only 1–10% of these costs for all countries, staffing accounted for as much as 74% of total costs in Germany and 63% in Spain, although the equivalent figures were only 25% in Italy, 28% in Denmark, 34% in France and 42% in England. Denmark, France and England are identified as examples of countries that primarily use midwives to provide support before, during and after birth, while Germany and Spain almost always have an obstetrician present during birth, which accounts for their additional staff costs. The researchers conclude that higher nurse-to-physician ratios reduce costs because midwives and nurses are able to take on many medical tasks that would otherwise be performed by doctors.

Five studies in a Cochrane review that compared continuity of midwife-led with shared or medical-led care in 13 trials involving 16,242 women at low and mixed risk included cost data, using different economic evaluation methods. All found savings associated with midwife-led intrapartum care. Although the studies were inconsistent in their approach to estimating maternity care costs, it seems there is potential for cost-saving with midwife-led care. 70 Based on scant existing evidence, there appears to be a trend towards a cost-saving effect for midwife-led continuity care compared with other care models. 71 The estimated mean cost saving for each eligible maternity episode is £12.38. This translates to an aggregate saving of £1.16M per year, if half of all eligible women avail themselves of midwife-led care at booking. This equates to an aggregate gain of 37.5 quality-adjusted life-years (QALYs) when expressed in terms of health gain using a NICE cost-effectiveness threshold of £30,000 per QALY. The uptake of midwife-led maternity services affects results on two levels: first by its role in determining caseload per midwife and thus mean cost per maternity episode; second at the aggregate level by determining the total number of women who start in midwife-led services nationally. 72

Research questions

In order to fulfil the aims and objectives, our research question asked to what degree are maternal quality and safety outcomes explained by characteristics of NHS trusts, staffing levels and skill mix, after adjusting for mothers’ characteristics including clinical risk and sociodemographic factors. We wished to know: How important are staffing levels and skill mix in determining outcomes for women and babies? What is the relationship between maternity workforce staffing levels, quality and safety outcomes and health-care output? To what extent is there an optimal staffing mix? What is the implication of efficiency savings for key quality and safety indicators?

Study design

A cross-sectional analysis was undertaken using multilevel logistic regression to investigate the relationship between a number of quality and safety outcomes and the workforce configuration adjusting for confounding characteristics. A cost analysis using a framework where both outcomes and cost are taken into account when measuring efficiency was also completed.

Patient and public involvement

In order to ensure that service user views were included in the proposed research, a trained childbirth educator, with experience of facilitating learning and discussion with pregnant women and their partners (MD), was involved in the development of the proposal and the design of the research, and named as a joint applicant on the grant proposal. She is also an experienced maternity service user representative, having represented the views of women on a number of maternity projects and committees, including Midwifery 2020, the Maternity Care Working Party and the RCOG Women’s Network. She has been concerned with the experience of NHS patients generally and has served as a lay member on the NICE Guidance Development Group for Patient Experience in Adult NHS Services. Her interest in maternity research is reflected in her role as voluntary research networker for the National Childbirth Trust (NCT).

In addition, she has an interest in how maternity data can be used to inform women making decisions and choices about childbirth. Together with the data analyst on this project (RG), she represented BirthChoiceUK, a voluntary organisation which has, for the last 14 years, helped women to make choices about their maternity care through information provided by the website www.BirthChoiceUK.com. 74 It provided information and maternity statistics for each maternity unit in the UK in an accessible format for parents to help them know what questions to ask locally and to decide where to plan to have their baby.

As joint applicant, and a user researcher, MD contributed to the development of the research proposal. This led to the consideration of wider quality outcomes for maternity services rather than a focus primarily on safety, and a more woman-focused approach to the concepts of productivity and efficiency.

It was originally intended that she would only attend and contribute to co-investigator meetings. However, it became clear that her knowledge of HES data and coding, quality metrics and women’s experience of maternity care – all derived from previous service user representative activities and driven from a woman-centred point of view – would be essential and valuable to the project and she therefore joined the project team. In doing so, she was able to have a hands-on role in the project, contributing on a regular basis to the direction of the study, including the choice of indicators to represent quality of maternity care and undertaking part of the research herself, developing a way of identifying women who had clinical risk factors relevant to birth.

She has already had, and will continue to have, a role in dissemination, having given both oral and poster presentations on the project and its interim findings at a number of conferences. Findings from this project will also inform future developments of BirthChoiceUK in its aim to provide high-quality information for parents. For example, HES data, cleaned and stratified by clinical risk and parity using techniques developed in this project, are displayed on the Which? Birth Choice website (www.which.co.uk/birth-choice75), created in partnership with BirthChoiceUK. These data will help women understand differences in outcomes between maternity units in England and contribute to their decision about where to give birth.

Further patient and public involvement (PPI) was contributed by the members of the advisory group (see Appendix 1). The purpose of the advisory group was specifically to advise the co-investigator group on the study questions, analysis and outputs. The advisory group included representatives from the user organisations NCT, the Stillbirth and Neonatal Death Charity (SANDS) and the National Maternity Support Foundation (Jake’s Charity), which were able to comment on the progress and findings of the study at our two advisory group meetings. Their help will be sought for dissemination of the findings.

Data sources

We used the following data sets:

• HES from 143 NHS trusts in England, admitted patients (including the ‘maternity and baby tail’) for the NHS year 2010–11. Specifically, the population is women who delivered in an obstetric or maternity unit based in a NHS trust in England 2010–11.

• NHS Workforce Statistics, England: 2010–11 from the Health and Social Care Information Centre, which includes staff in post including bank staff.

• CQC Maternity Survey of Maternity Provider Trusts 2007 and survey of women’s experiences, 2007 and 2010, from the UK Data Archive.

• Office for National Statistics (ONS) Number of Maternities by Establishment, 2001/02 to 2010/11.

• The BirthChoiceUK database, held by Rod Gibson Associates Ltd and containing information on NHS trusts and maternity units by location and type (e.g. OU, FMU, AMU).

• Reference cost data by NHS trust – NHS reference costs 2010/11 were available under Open Government Licence v2.0 (URL: www.nationalarchives.gov.uk/doc/open-government-licence/version/2/). 76

Data storage, governance and ethics

The data were stored in a MySQL database [MySQL version 5.5. This is an open source program (www.mysql.com) that is overseen by Oracle Corporation, Redwood Shores, CA, USA]. The flat HES data files were reorganised into a relational database to facilitate faster processing. To further speed processing, the mothers’ delivery records were separated from the general inpatient records as were mothers’ non-delivery inpatient records. A similar split was carried out for babies’ birth records. The loading, organisation and cleaning of data as well as the calculation of indicators were all performed by routines written in Python, thus ensuring reproducibility. The number of deliveries by communal establishment code for 2010/11 was obtained from the ONS under Open Government Licence v1.0 (URL: www.nationalarchives.gov.uk/doc/open-government-licence/). These were matched with communal establishment place using a list obtained from NHS Connecting for Health. Communal establishment place was matched to NHS trust using the BirthChoiceUK database. Advice regarding ethical approval was sought from King’s College Research Ethics Committee, which advised that ethics approval was not required because the team planned to undertake secondary analysis of existing anonymised data. Careful measures were taken to safeguard the source data, in line with College research governance policies and procedures.

The HES data set available to the study consisted of:

• HES Inpatient Records 2000/01 to 2010/11

• HES Outpatient records 2003/04 to 2010/11

• HES accident and emergency (A&E) records 2007/08 to 2010/11

• ONS-HES Linked Mortality Statistics (including neonatal deaths) 2000/01 to 2010/11.

In summary, a mother has a single HES delivery record with fields available to record a limited amount of information on nine babies born in a single episode. Each baby has its own HES birth record, but this is not linked to the mother’s delivery record. Only the year 2010/11 was of direct relevance to the project. However, using anonymised unique patient identifiers in the HES records, women delivering babies could be linked to previous inpatient and previous delivery records for years back to 2000/01. This allowed a more complete picture of a woman’s obstetric history to be built; for example events such as stillbirth or a caesarean in a previous pregnancy which were considered risk factors for the delivery recorded in 2010/11.

Data cleaning and quality

The quality of HES data is improving. Murray et al. 77 examined the range and completeness of birth information recorded in HES and tested an approach for minimising the effect of hospital-level variations by selecting hospitals with high completeness of recording (90%) for key fields. They found that the proportion of missing data in key birth record fields has been decreasing annually, such as gestational age and birthweight (from 46.2% and 43.9% in 2005/06 to 18.1% and 16.9% in 2009/10 respectively). They compared the important characteristics such as size and access to specialist neonatal care between 71 high-coding and 85 low-coding hospitals and found no significant differences, suggesting hospitals with high birth record completeness may be generalisable and representative of all hospitals. Knight et al. 78 found that analyses using HES data were affected by the completeness and consistency of data. They found that different analysis rules had a small effect on the statistics at a national level but the effect could be substantial for individual NHS trusts.

Thus, we have drawn upon the guidelines developed by Sinha et al. 79 for reporting on data cleaning and quality. Appendix 2 provides an account of full data cleaning conducted for the study, prior to data analysis. The study restricted the records to NHS hospital deliveries resulting in a registrable birth. Duplicate delivery records were removed from the mother’s records. The babies’ birth records also contained duplicate records. These duplicates were not removed, as the majority of the project’s work concentrated on the mother’s delivery record and the resources were not available to clean both. We consulted the CQC regarding its Maternity Data Quality – Indicator specifications for maternity-related measures included within its surveillance programme, and developed a scoring system that was used to select records with the largest amount of most useful and relevant data to the project.

Study size and bias

We used the full census of 656,969 women’s deliveries in HES, so there was no bias caused by non-response. Any biases would therefore be caused by missing data, poorly recorded data or omitted variables from the risk adjustment model. Sample sizes for some indicators were reduced by the choice of denominator to create the indicator.

For example, healthy mother and healthy baby indicators (n = 518,698; 79%) were limited to records for which there was known gestational age, birth status, birthweight and postnatal duration. Normal birth (n = 548,272; 83%) was limited to records for which there was known onset of labour, mode of delivery and anaesthetic, and intact perineum (n = 493,449; 75%) was limited to women who did not have a caesarean. The final five indicators (delivery with bodily integrity, spontaneous vaginal delivery, elective caesarean, emergency caesarean and all caesareans) all had high levels of completeness (n = 656,135; 99.9%). Missing postcode information meant that a further 5462 delivery records (1%) were lost from the analysis, as were 68,482 records (10%) with mother’s age missing.

A pre-specified decision was taken not to include any trust where fewer than 80% of women could be coded for a particular indicator. Combining this with missing postcode data meant that the percentages of delivery records used in the multilevel models were as follows: healthy mother and healthy baby indicators (n = 431,391; 66%), normal birth (n = 467,022; 71%), intact perineum (n = 439,730; 67%, or 89% of women who had a vaginal delivery), and delivery with bodily integrity, spontaneous vaginal delivery, elective caesarean, emergency caesarean and all caesareans (n = 584,435; 89%).

Derivation of process and outcome indicators

The Institute of Medicine’s Crossing the Quality Chasm report11 identified large gaps between health care that is, and that could and should be delivered in the United States. The report identified six aims for quality improvement that have been widely adopted internationally; thus health care should be safe, effective, patient centred, timely, efficient and equitable. Some outcomes are also influenced by factors outside health care, for example by social and structural determinants of health. Some neonatal outcomes are influenced by quality of care in the neonatal sector. Boxes 1 and 2 detail the result of a review of quality indicators used in maternity care. Sources include AHRQ,30 UK Policy and guidelines,6,80 CQC,81 CQUIN,18,19 ONS,82 NHS Operating Framework 2012/13,83 RCOG,25 Midwifery 2020,13 Safer Childbirth: Minimum Standards for the Organisation and Delivery of Care in Labour,7 Information Centre Indicators for quality improvement 201320 and the Australian Council on Healthcare. 84

Indicators used in analysis

The selection of indicators was decided in consultation with the advisory group and informed by needing to have a balance of positive and negative indicators, the importance to women, costs, and the availability and quality of coding within the HES data set. Three indicators were derived to indicate a healthy mother and healthy baby, thus reflecting a concept of harm-free care, avoidance of longer-term morbidity and a positive outcome. The mode of birth indicators were chosen to compare important processes and outcomes across trusts and with other studies. Ten final indicators (some were composites) that measured maternal and infant outcomes were developed and are outlined in Table 1.

The subcomponents for each composite indicator are shown in Tables 2–5. A full description of each composite outcome is also provided; please see Chapter 3, Outcome indicators.

Independent variables

We included mother’s characteristics measured at the individual level that are known to affect the outcomes of interest. These included age, parity, clinical risk at the end of pregnancy as measured by the NICE guideline for intrapartum care,80 ethnicity, area socioeconomic deprivation as measured by the Index of Multiple Deprivation (IMD),85 geographical location (urban/rural) and region.

Trust-level characteristics included size measured by number of deliveries, teaching status, maternity configuration drawing on the Birthplace in England86 typology of whether or not AMUs and FMUs are part of trust provision and staffing variables. These included staffing levels (FTE obstetric medical staff, midwives and maternity support staff per 100 maternities, FTE all staff per 100 maternities) and skill mix (doctor/midwife and midwife/support worker ratio).

Statistical methods

Multilevel logistic regression models, where mothers (deliveries) were nested within trust, were fitted to the 10 dichotomous maternity indicators using the Statistical Analysis System (SAS) version 9.3 (SAS Institute Inc., Cary, NC, USA) Generalised Mixed Models procedure (GLIMMIX). An unstructured variance–covariance matrix parameterised via the matrix’s Cholesky root was used. The root is squared to give the residual variance (σ²). We present the residual σ in findings. This parameterisation has good computational and statistical properties (it guarantees that the variance–covariance matrix is at least positive semidefinite), and is often preferred to an unstructured variance–covariance matrix. Other simper forms of the variance–covariance matrix are available but they require certain assumptions to be made, so a decision was taken not to impose a predetermined structure. The independent variables used in the model are shown in Table 11.

Two models were fitted to the data using the mothers’ characteristics, sociodemographics and trust-level variables with either set 1 or set 2 of the staffing variables. Note we use the abbreviation FTE for full-time equivalent staff. Some of the categories do not appear in the statistical models because they were treated as missing data. Often these are related to the fact that the woman’s postcode was missing (e.g. for rural/urban classification, if a woman’s postcode was missing then she would not have a code for IMD).

The results from the 10 multilevel models were summarised in three tables (see Chapter 3, Tables 23–25).

Specifically for this study, symbols have been used to indicate the magnitude of each model effect using the relative chi-squared (chi-squared/degrees of freedom) value in a heuristic way to help summarise the results and provide a sense of which variables were important in predicting particular outcomes. The relative chi-squared values for all statistically significant effects were categorised as follows: minor (1–99), moderate (100–999), strong (1000–9999) or dominant (≥ 10,000).

The ↑ has been used to indicate a positive monotonic relationship (i.e. rises upwards in a linear or non-linear way) between a continuous predictor and the outcome, ↓ to indicate a negative monotonic relationship, * for non-monotonic relationships (e.g. increases with parity and then declines from parity 2 onwards) and ↔ for relationships involving categorical variables [e.g. ethnicity, Strategic Health Authority (SHA)]. So, for positive continuous relationships the symbols ↑, ↑↑, ↑↑↑, ↑↑↑↑ were used for minor, moderate, strong and dominant effects respectively. Where a variable was close to statistical significance (i.e. p > 0.05 but ≤ 0.10), this has been placed in parentheses in the report table (see Tables 23–25). Note the relative chi-squared value for each of the variables is presented in a separate table in the results section (see Table 22).

Sometimes outcomes were achieved more often as deprivation increased. This was represented by an upwards arrow ↑. Conversely, when they happened less often this was indicated by a downwards arrow ↓.

Independent variables were added in blocks starting with mother-level variables, then sociodemographics, trust-level and finally staffing variables. There were two different staffing variable models: the first used FTE per 100 maternities for the three staffing groups (doctors, midwives and support staff) and the second model used FTE (all) staff per 100 maternities alongside two ratio variables – doctor-to-midwife ratio and support worker-to-midwife ratio – to test the effect of substitution of one staff group with another.

The intercept was the only parameter allowed to vary between trusts, to ensure that clustering of mothers and babies within trusts was properly accounted for in the estimation of the parameter estimate standard errors (SEs). Funnel plots of the unadjusted and adjusted proportions (y-axis) against the number of known deliveries for each indicator (x-axis) were plotted. In Appendix 3, the funnel plots for the intercept-only model (unadjusted proportions) and the model that used set 1 of the staffing variables (adjusted proportions) are presented, noting that the plots for the two staffing models were very similar.

We did not attempt to adjust the funnel plot limits for overdispersion,89 which would happen if more trusts fell outside the control limits than would normally be expected. The main purpose of the plots was to judge what impact the model independent variables had upon the variation in the indicator rates across trusts rather than to attempt to identify outlier trusts, which is best achieved using adjusted funnel plots. Typically, model independent variables reduce the amount of variation between units of analysis (e.g. trusts) although it is possible for the variation to increase: for example, after the addition of the variables representing mothers’ characteristics into the model the variation increased for 8 of the 10 indicators.

Three sensitivity analyses were performed by fitting the statistical models to (1) trusts which only had a single OU (n = 50), (2) replacing the dichotomous clinical risk grouping with a three-category variable (NICE lower risk, individual assessment, NICE higher risk) and (3) testing the following interactions (multiplicative effects):

• parity × FTE doctors per 100 maternities

• parity × FTE midwives per 100 maternities

• parity × FTE support workers per 100 maternities

• clinical risk × FTE doctors per 100 maternities

• clinical risk × FTE midwives per 100 maternities

• clinical risk × FTE support workers per 100 maternities.

This was performed in order to find out if the effect of increasing/decreasing staffing levels upon outcomes was the same across different parities and clinical risk groups.

Costing analysis

The cost analysis attempted to model maternity services to identify the relationship between inputs and outputs, and more specifically for this study the relationships between the different labour inputs. In particular, we wanted to establish the extent that the different roles are substitutes or complements, i.e. are competing or aiding inputs.

Methodology to explore midwifery staffing levels, outcomes and the cost of providing maternity services for NHS trusts

Data on medical staffing were not detailed enough to ascertain the split between obstetric and gynaecological responsibilities in a trust. Therefore the costing analysis investigated the relationships between midwifery staffing levels, midwifery-related outcome measures and the cost of providing maternity services for NHS trusts in England and draws on an analysis for the RCM. For each NHS trust and PCT providing maternity services, the following were obtained:

• the average maternal age at time of delivery, determined using the HES data field MATAGE

• the percentage of nulliparous women delivering in the trust, determined using the data field NUMPREG and linkage with previous delivery records

• the average ranking of the overall IMD, determined using the data field IMD04RK

• the percentage of women who were deemed to be at increased risk of complications at onset of labour, determined using the risk allocation methodology described earlier.

These were used to create a demographic profile for each NHS trust.

Exclusion of data for quality and bias

A number of trusts were excluded from the analysis:

• Two PCTs and two NHS trusts which provided only midwifery services. These have different patterns of care and costs from trusts providing obstetric services.

• Seven NHS trusts which had atypical or inconsistent data, or data which were overly influential in regression analyses. Inclusion of these trusts could lead to spurious conclusions about relationships between variables.

As a result, analyses were performed using data from 134 English NHS trusts and 2 PCTs.

Economic modelling methodology

In economics, a production function describes the mechanism for converting a vector of inputs (x) into output (y). After selecting the appropriate functional form, econometric estimation of the function’s parameters allows the output elasticities to be calculated and returns to scale to be found. In the absence of data on input prices at the maternity services level of analysis, we adopted a production (i.e. quantity) function approach. Many health-care studies using production functions (as opposed to cost functions) have adopted Reinhardt’s90 specification of the production function, which was the first to include multiple labour inputs (registered nurses, technicians, administrative staff and doctors). However, this function assumes all inputs to be substitutes (solely because of the absence of cross-products). The advance in production function analysis of the 1970s gave rise to two flexible econometric specifications that allow us to relax this overly strict assumption. Berndt and Christensen91 introduced the transcendental-logarithmic (translog) production function and Diewert92 introduced the generalised linear production function (also known as the Allen, McFadden and Samuelson production function).

Using either of these functions would have allowed us to estimate the relationship between the labour inputs because the regression coefficient on the cross-products (interaction effects) can be simply used to calculate the Hicks64 elasticity of complementarity (see Sato and Koizumi93 or Syrquin and Hollender94 for an explanation). However, an advantage of the Diewert92 specification is that it allows zero quantities for some inputs, which may be a more realistic assumption when labour inputs are disaggregated, as they are in our study. This modelling enabled us to examine the output contribution of the different staff inputs (output elasticities) and their influence upon the productivity of other staff inputs (i.e. whether they are complements or substitutes). With these results available, we were able to investigate the input substitution possibilities available to hospitals under different scenarios.

Following Diewert92 we adopted a generalised linear production function defined as:

where in our study K = 4, X = {consultants, other doctors, midwives and support staff} and Y = Q, corresponding to the number of deliveries. This model was estimated using ordinary least squares in Stata (Stata 12, StataCorp LP, College Station).

To examine the q-complementarity (and therefore to answer the question relating to skill mix), we calculated the Hicks elasticity of complementarity,64 η^H defined for any two staffing inputs i, j (i, ≠ j):

where

The elasticities were computed at the means and the SEs using the delta method.

Profile of women who gave birth in 2011

Two-thirds of women who gave birth in 2011 were aged between 20 and 34 years (67%); most were nulliparous (43%) or had one previous live birth (32%) (Table 14). Women were more likely to be classified as higher risk than lower risk (55% vs. 45%) according to the definition of women at increased risk of complications based on the NICE intrapartum care guidelines. 80 Included in the 55% were 4% of women who required individual assessment to determine if they were at increased risk of complications. About two-thirds of women were categorised as white British, a further 9% were from another white background, 4% were Pakistani, 3% were African, 3% were Indian and 2% were from other Asian backgrounds. A higher proportion of women were living in a deprived area based on the IMD;78 28% versus 15% from the least deprived quintile. Most women lived in denser urban areas (86%).

Profile of NHS trusts in 2011

Tables 15 and 16 show that most trusts operated their maternity service either through one or more OUs only (42%) or through an OU with an AMU (OU/AMU, 32%). Nearly 30% of trusts were attached to a university and London SHA had the greatest number of trusts (24, 17%). The average number of births per trust in 2011 was 4620 (range 1214 to 10,678). There were 4.80 FTE staff for every 100 births, of which 0.82 FTE were doctors (0.21 to 1.65) (one trust had a particularly low FTE, the next lowest had 0.44 FTE and all other trusts had 0.50 FTE and above), 3.08 FTE were midwives (1.11 to 4.71) and 0.90 FTE were support workers (0.05 to 2.88). The ratio of doctors to midwives averaged 0.27 (0.07 to 0.50) and support workers to midwives 0.30 (0.02 to 0.85).

Outcome indicators

The analysis that follows focuses on 10 indicators described in Chapter 2, Table 1 and listed in Table 17. Five of these were composites of three or more other indicators. The 10 indicators have been placed into three groups. The healthy mother and healthy baby indicators form a natural group, as do the mode of birth and caesarean indicators. Healthy outcomes comprised (1) healthy mother (delivery with bodily integrity, mother returned home within 2 days, not readmitted within 28 days and without instrumental delivery, maternal sepsis or anaesthetic complication), (2) healthy baby (baby’s weight 2.5–4.5 kg, gestational age 37–42 weeks, live baby) and (3) healthy mother/healthy baby dyad (cases in which both the mother and the baby were healthy). Mode of birth comprised (4) delivery with bodily integrity (without caesarean, uterine damage, second-/third-/fourth-degree tear, sutures and episiotomy), (5) normal birth (without induction, instrumental and caesarean birth, episiotomy, general and/or regional anaesthetic), (6) spontaneous vaginal delivery and (7) intact perineum. Caesarean indicators comprised (8) elective caesarean, (9) emergency caesarean and (10) all caesareans. The full reports of the models informing the results are in Appendix 4.

How each indicator varies across trusts is shown in Table 17 and variation by the variables listed in Tables 15 and 16 is shown graphically in Appendix 3. Intact perineum was the indicator that had the greatest variation between trusts and elective caesarean the least.

Multilevel modelling

Given the detailed nature of the tables reporting the results of the multilevel models, we have placed them in Appendix 4 and have reproduced the results in summary form to aid the reader (Tables 22–26). Those mothers whose age and IMD were not known were excluded from the analysis. Note there were two different staffing models. The first model adds three variables to the model that contains mothers’ characteristics, sociodemographic and trust-level variables: FTE doctors per 100 maternities, FTE midwives per 100 maternities and FTE support workers per 100 maternities. The second model replaces the three previous staffing variables with FTE all staff per 100 maternities, doctor-to-midwife ratio and support worker-to-midwife ratio.

The residual variance for the 10 intercept-only models (i.e. the model without any independent variables) ranged from 0.203 to 0.283. Based on these estimates, approximately 1–2% of the total variation in the outcome indicators is attributable to differences between trusts whereas 98–99% of the variation is attributable to differences between mothers within trusts. We are not able to say how much of the variation could be due to unknown characteristics that are predictive of outcome, or to variations in the quality of care and/or different models of care received by different women in the same trust.

The area under the curve (AUC) statistics discussed below provide some indication of how well these models fit the data. An AUC of 0.5 is no better than tossing a coin whereas an AUC of 1 implies perfect prediction.

Most of the variation observed for each individual indicator was explained by mothers’ characteristics. The relative chi-squared values for the effects of mother’s age, parity and clinical risk were often of a different order of magnitude from most of the other independent variables entered into the models (see Table 22). Age, parity and clinical risk were the only variables where the relative chi-squared value exceeded 1000 on one or more occasions (four, nine and nine occasions respectively). Parity and clinical risk were the only two variables where the relative chi-squared value exceeded 10,000 (three and nine occasions respectively). Ethnicity and the IMD were the only other two variables where the relative chi-squared value managed to exceed 100 on one or more occasions (two and four occasions respectively).

The effect sizes of the staffing variables were very small by comparison. For example, the relative chi-squared value for clinical risk was 15,841 for the healthy mother indicator whereas the largest effect size for a staffing variable for this indicator was only 3. In Tables 22–25 effect sizes have been summarised using symbols (see key beneath each table). A monotonic relationship is one that increases consistently upwards (or downwards) in either a linear or a curvilinear fashion.

There was marginal improvement in a model’s capacity to predict outcomes following the addition of sociodemographic, trust-level and staffing variables. The largest improvement was for intact perineum (AUC 0.722 to 0.732). The improvement in the AUC brought about by adding the staffing variables to the model was negligible with hardly any change in the AUC. Based on the AUC, most models meet the criteria for fair (0.70 to 0.80) to good (0.80 to 0.90) prediction. The model for elective caesarean provided the best predictions (AUC 0.814) and the model for emergency caesareans was least able to predict this outcome accurately (AUC 0.698). Potentially there is capacity to improve the fit of these models by adding further variables to raise the AUC to 0.9 and above, although a number of the variables that might help in this respect were inadequate for use in the analysis (e.g. smoking, body mass index) because they were either poorly or not consistently recorded.

The addition of variables representing mothers’ characteristics into the intercept-only models (i.e. without any independent variables) more often than not resulted in an increase in the residual sigma (between trusts). The only indicators where this did not occur were normal birth (falling from 0.270 to 0.225) and spontaneous vaginal delivery (falling from 0.226 to 0.200). Conversely, sociodemographic variables had the reverse effect with a marked reduction in the variation for four of the indicators when these variables were added to the mother level model: healthy mother (falling from 0.308 to 0.256), healthy mother/healthy baby dyad (falling from 0.306 to 0.257), delivery with bodily integrity (falling from 0.297 to 0.244) and intact perineum (falling from 0.334 to 0.280). This suggests that when studying variability between trusts it is important to include in the model both mothers’ characteristics, which increase variation between trusts, and sociodemographic variables, which decrease variation, otherwise our assessment of the variation will be biased. This provides a better indication of unexplained variation between trusts that adjusts for trust differences in women’s risk profile.

Staffing, outcome and costs

The final objective was to examine the relationship between maternity workforce staffing levels, quality and safety outcomes and cost. Data on medical staffing were not detailed enough to ascertain the split between obstetric and gynaecological responsibilities in a trust. Therefore, this analysis investigated the relationships between midwifery staffing levels, where data quality was very good, midwifery-related outcome measures and the cost of providing maternity services for NHS trusts in England. This section draws on analyses conducted for the RCM. 95

For this analysis we used trust-level data to investigate relationships between outcome measures, midwifery staffing levels and the cost of providing maternity services for NHS trusts in England.

Initial analyses

Number of deliveries

Maternities for each site derived from ONS data matched well with the number of deliveries calculated using reference cost data. Two NHS trusts had discrepancies, with > 70% more deliveries according to ONS data than recorded in the activity from reference costs. These were excluded from further analyses. For further analyses, the number of deliveries derived from reference costs was used. In the following text the expression y ∼ x ⊕ z is used for the schematic relationships of ‘y depends on x and z’ being tested at each stage, although the functional relationship may be more complicated if transformations or link functions have been used.

Trust profiles and average costs

The average and range of variables used for the analyses are shown in Table 31.

TABLE 31 - Variables used in cost analysis

Based on number of deliveries derived from reference cost data.

Variables	Average of trusts	Lowest trust	Highest trust	SD
Average mother’s age (years)	29	27	33	1
Nulliparous (%)	44	26	61	8
Increased risk (%)	47	28	69	7
IMD ranking	14,422	4657	25,453	4783
Operative delivery rate (%)	37	27	53	4
Normal birth rate (%)	41	26	59	5
Intact perineum rate (%)	43	26	66	7
Total cost per delivery (£)a	4128	2606	6429	648
Proportion antenatal cost (%)	27	5	54	10
Delivery cost per delivery (£)a	1974	1108	3479	462
Deliveries per midwifea	34	21	53	5

Hypothesis 1: investing in providing good antenatal services results in reduced operative delivery rates and lower delivery costs

Operative delivery rate ∼ antenatal cost per delivery (⊕ risk ⊕ parity ⊕ age ⊕ IMD)

When the case-mix adjustments were included in the model there was no relationship with antenatal spend. Including the reference costs category ‘attendance’ (either face to face or not face to face, which is largely thought to be antenatal contact time) in the definition of antenatal expenditure reduced the dependence further.

Delivery cost per delivery ∼ antenatal cost per delivery ⊕ midwives per delivery ⊕ trust size ⊕ (age ⊕ % nulliparous ⊕ % risk ⊕ IMD)

An initial analysis showed that delivery costs have a significant dependence on the antenatal spend, with increased antenatal costs associated with a decrease in delivery costs. However, taken at face value, the extra amount incurred on antenatal care exceeded the reduction in delivery costs.

Figure 1 shows that there are around a dozen trusts with very high antenatal costs and correspondingly low delivery costs which are responsible for the regression relationship. When the analysis was repeated excluding trusts with antenatal costs in excess of £1500 there was no relationship between antenatal cost and delivery cost per delivery.

FIGURE 1.

Relationship between antenatal cost/delivery and delivery cost/delivery (adjusted for Market Forces Factor).

Hypothesis 2: trusts with a high antenatal expenditure as a proportion of all maternity expenditure will have lower operative delivery rates

Operative delivery rates ∼ antenatal/total costs (⊕ risk ⊕ parity ⊕ age ⊕ IMD) ⊕ trust size

No relationship was found between the proportion of maternity expenditure spent on antenatal care and operative delivery rates after adjusting for maternal characteristics and trust size.

Hypothesis 3: there is a relationship between delivery cost per delivery and midwifery staffing levels

Delivery cost per delivery ∼ midwives per delivery ⊕ trust size (⊕ risk ⊕ parity ⊕ age ⊕ IMD)

Higher midwifery staffing levels were associated with higher costs of each delivery, although the relationship was not strong. This relationship strengthened when antenatal expenditure was included as an explanatory variable in the model.

Only around 17% of the variation between trusts’ delivery costs could be accounted for by variables included in this model (midwives per delivery, trust size and case mix), rising to 23% when antenatal expenditure was taken into account (hypothesis 2b). The remaining variation in the average cost of a delivery was not accounted for by maternal characteristics, size of trust, number of FTE registered midwives employed or antenatal spend and must be due to other factors not included in the analysis.

Hypothesis 4: outcomes measures are correlated with the delivery cost per delivery

Delivery cost per delivery ∼ operative delivery rate (⊕ risk ⊕ parity ⊕ age ⊕ IMD)

The analysis found that higher operative delivery rates were not significantly associated with higher delivery costs after adjusting for maternal characteristics. Variations in costs between trusts were not related to the numbers of women having operative deliveries.

Delivery cost per delivery ∼ normal birth rate (⊕ risk ⊕ parity ⊕ age ⊕ IMD)

There was no association between delivery cost per delivery and the normal birth rate.

Delivery cost per delivery ∼ intact perineum rate (⊕ risk ⊕ parity ⊕ age ⊕ IMD)

There was no association between delivery cost per delivery and the intact perineum rate. As the denominator for intact perineum was vaginal deliveries, the analysis was repeated used delivery cost per vaginal delivery but this also showed no association.

Delivery cost per delivery ∼ healthy mother and healthy baby outcomes (⊕ risk ⊕ parity ⊕ age ⊕ IMD)

There was no association between delivery cost per delivery and any of the three healthy mother and healthy baby indicators (healthy mother, healthy baby and healthy mother/healthy baby dyad).

CQC scores ∼ delivery cost per delivery

No relationship could be established between delivery cost per delivery and women’s experience of maternity care as measured by the average of the CQC scores.

CQC scores ∼ antenatal cost per delivery

No relationship could be established between antenatal cost per delivery and women’s experiences of antenatal care as measured by the antenatal component of the CQC scores.

Hypothesis 5: trusts with high operative delivery rates have a high postnatal cost as a proportion of all maternity expenditure

Postnatal costs per delivery/total costs per delivery ∼ operative delivery rates (⊕ risk ⊕ parity ⊕ age ⊕ IMD)

A relationship could not be found that explained postnatal costs in terms of variations in operative delivery rates once adjustments were made.

Hypothesis 6: a trust’s total expenditure on maternity services is related to the characteristics of the women who use the trust

Total costs per delivery ∼ trust size ⊕ risk ⊕ parity ⊕ age

Having a higher proportion of women at increased risk of complications was associated with more expensive maternity care. Level of social deprivation approached significance (p = 0.052) with deliveries in trusts with a higher proportion of women living in deprived areas having more costly deliveries. However, in this analysis trust size and case mix accounted for only 22% of the variation in total delivery costs.

Some of the previous analyses also showed relationships between costs and maternal characteristics. For trusts with a larger proportion of women at increased risk of complications, delivery costs were higher and they spent a higher proportion of their total maternity costs on postnatal care, and trusts with a higher proportion of nulliparous women had higher costs per delivery.

Hypothesis 7: the efficiency of a trust measured in terms of the expenditure per delivery is not correlated with the size of the trust measured by number of deliveries

A number of previously described analyses included trust size as an explanatory variable. Larger trusts, measured by the numbers of deliveries, appeared to offer maternity services which cost less than those in smaller trusts. Repeating the analysis for the 50 trusts whose maternity services are delivered by a single OU, the significant effect of size increased, suggesting that the economies of scale were greater in a single site.

The size of the trust had no relationship to the cost of a delivery only (i.e. when costs of antenatal or postnatal care were excluded). Deliveries were not cheaper in larger units, once adjustments have been made for maternal characteristics and numbers of midwives. This lack of relationship held when repeating the analysis for the 50 trusts that delivered maternity services through only a single OU.

Outcome measures and trust size

Larger trusts were associated with lower CQC scores of women’s experience of maternity care. When broken down into components, women’s experience of antenatal care and women’s experience of labour and birth was poorer in larger trusts. No significant relationships were found between trust size and staff during labour and birth, care in hospital after the birth or feeding the baby during the first few days, although there was a trend close to significance for larger trusts to have lower scores. Overall, the analysis of CQC was very sensitive to individual data points, with significant results being lost once outliers were removed. There was no relationship between trust size and the other outcome measures: normal birth, intact perinea and number of complaints.

How do organisational factors affect variability in maternal interventions and maternal and perinatal outcomes?

Approximately 1–2% of the total variation in the outcome indicators was attributable to differences between trusts, whereas 98–99% of the variation was attributable to differences between mothers within trusts. There was marginal improvement in a model’s capacity to predict outcomes following the addition of sociodemographic, trust-level and staffing variables. Based on the AUC, most models meet the criteria for fair to good prediction. The key factors are summarised in the next section.

What is the relationship between maternity staffing, skill mix and maternal and perinatal outcomes?

Overall, the linear effects of the staffing variables were not statistically significant for eight indicators. However, all women benefited from an increase in midwifery staffing, in terms of retaining an intact perineum and bodily integrity.

Looking at mode of birth indicators, a higher number of midwives (FTE per 100 maternities) and higher levels of overall staffing were associated with improved chance of delivery with bodily integrity and an intact perineum. There was a reduction in the between trust variation across all four mode of birth indicators, particularly for normal birth, when mothers’ characteristics, sociodemographics, trust-level and staffing variables were added to the intercept-only model (i.e. the model that contains no independent variables). Staffing variables had a non-significant effect upon the chances of a caesarean.

We then investigated whether the effect of staffing levels upon outcomes could vary according to either a woman’s parity or her clinical risk. An analysis of the multiplicative effects of parity and clinical risk with the staffing variables was more revealing.

What is the relationship between maternity staffing, cost and outcomes?

We examined the relationship between maternity workforce staffing levels, quality and safety outcomes and cost. Data on medical staffing were not detailed enough to ascertain the split between obstetric and gynaecological responsibilities in a trust. Therefore, this analysis investigated the relationships between midwifery staffing levels, where data quality was very good. For this analysis we used trust-level data to investigate relationships between outcome measures, midwifery staffing levels and the cost of providing maternity services for NHS trusts in England.

When the case-mix adjustments were included in the model there was no relationship between antenatal spend and reduced operative delivery rates and lower delivery costs. No relationship was found between the proportion of maternity expenditure spent on antenatal care and operative delivery rates after adjusting for maternal characteristics and trust size.

Higher midwifery staffing levels were associated with higher costs of each delivery, although the relationship was not strong. Only around 17% of the variation between trusts’ delivery costs could be accounted for by variables included in this model. The remaining variation in the average cost of a delivery was not accounted for by maternal characteristics, size of trust, number of FTE registered midwives employed or antenatal spend and must be due to other factors not included in the analysis.

The analysis found that higher operative delivery rates were not significantly associated with higher delivery costs after adjusting for maternal characteristics. Variations in costs between trusts were not related to the numbers of women having operative deliveries.

There was no association between delivery cost per delivery and the normal birth rate. There was no association between delivery cost per delivery and the intact perineum rate, or any of the three healthy mother and healthy baby indicators, and women’s experience of maternity care as measured by the average of the CQC scores. A relationship could not be found that explained postnatal costs in terms of variations in operative delivery rates once adjustments were made.

Having a higher proportion of women at increased risk of complications was associated with more expensive maternity care. Level of social deprivation approached significance with deliveries in trusts, with a higher proportion of women living in deprived areas having more costly deliveries. However, in this analysis trust size and case mix accounted for only 22% of the variation in total delivery costs. Some of the previous analyses also showed relationships between costs and maternal characteristics. For trusts with a larger proportion of women at increased risk of complications, delivery costs were higher and they spent a higher proportion of their total maternity costs on postnatal care, and trusts with a higher proportion of nulliparous women had higher costs per delivery.

Larger trusts, measured by the numbers of deliveries, appeared to offer maternity services that cost less than those in smaller trusts. Repeating the analysis for the 50 trusts that only consist of a single OU, the significant effect of size increased, suggesting that the economies of scale were greater in a single site. The size of the trust had no relationship with the cost of a delivery only (i.e. when costs of antenatal or postnatal care were excluded). Deliveries were not cheaper in larger units once adjustments had been made for maternal characteristics and numbers of midwives. This lack of relationship held when repeating the analysis for the 50 trusts that delivered maternity services through only a single OU each.

Larger trusts were associated with lower CQC scores for women’s experience of maternity care. There was no relationship between trust size and the other outcome measures: normal birth, intact perinea and number of complaints.

From this study, the increased investment in staff did not necessarily have an effect on the outcome and experience measures chosen, where there was in general no relationship with midwifery staffing levels. However, there was a higher intact perineum rate in trusts with higher levels of midwifery staffing. Although this validates the result to some extent, any trusts submitting erroneous data will be correlated between years. It would be interesting to see if the relationship holds good if the data are analysed at individual record level. Maternity units with higher levels of midwifery staffing may find it easier to provide continuity of carer and one-to-one midwifery care. This finding would then be consistent with research evidence that one-to-one midwifery care can result in a significant reduction in perineal trauma. 70 This is an important outcome for women which impacts on the quality of their life with a new baby and could impact on future decisions about mode of birth. 96

Although there was no significant improvement in women’s experiences of care, as measured by CQC scores, as a result of higher staffing, there was a trend in that direction. These scores were not all directly related to midwifery staffing and covered a wide range of women’s experience of care.

While overall levels of midwifery staffing are important, other factors will also affect outcomes and experience of care for women. The deployment of those staff within the maternity service, their attitude towards the women they care for, their skills and the culture within which they work will all play a part in women’s care.

Despite the relationship of higher levels of midwifery staffing with improved intact perineum rates, and the higher costs associated with providing that staffing, when tested directly there was no relationship detected between delivery costs and improved perineal outcomes. In fact, few patterns connected the cost of providing maternity services with differences in the populations they serve or the complexity of births. Higher operative delivery rates were not associated with higher costs of intrapartum care (after adjusting for case mix) and this emphasises the inherent variability found in the reference costs, as found by Laudicella et al. 67 Excess bed-days were not included with delivery costs in these analyses (instead being included in postnatal costs). It may be that including the extra duration of stay following an operative birth would have an impact on this relationship,68 although this is unlikely, as the cost of excess bed-days was very small in relation to total cost. Trusts with higher numbers of women at increased risk of complications do appear to be associated with higher delivery costs, as found by Gaughan et al. ,68 and also with higher postnatal costs. This pattern is reflected in the new maternity pathway system, where higher payments are made for higher-risk women, but interventions themselves are not rewarded.

Providing more costly antenatal care alone did not appear to have a direct impact on operative delivery rates and was not recouped by a commensurate reduction in delivery costs, nor did it result in a better experience of antenatal care as measured by the CQC antenatal summary score. Some trusts did appear to be able to increase antenatal expenditure and reduce delivery costs. However, these trusts may be attributing costs differently between antenatal spend and delivery cost, which would create the illusion of a trade-off. From the data it is not possible to decide if these trusts have a winning formula.

The analyses generally adjusted for differences in size of trust as measured by the annual number of deliveries. The biggest effect of trust size was seen with total costs, particularly when restricted to trusts which had a single site only. Costs per childbearing woman were lower for larger units after taking into account case mix, which may have been a result of economies of scale. As no relationship could be seen between trust size and delivery cost alone, it could be deduced that the differences are related to antenatal or postnatal costs. However, the wide variations in costs suggest that trusts may be allocating costs differently along the maternity pathway.

Women seemed to be slightly less satisfied with their experience of care, as measured by CQC scores, in larger trusts than in smaller ones. Again, because of the complexities of trust configurations, it may be that size of unit is more important than size of trust, but CQC scores are available at trust level only. Although larger trusts appear to spend less overall, and may offer a lower quality of service from the point of view of women’s experience, this type of analysis does not show causality and therefore it is not possible to assume that spending less on maternity care results in lower-quality care.

Relationships between staffing, costs and outcomes are complex and it is perhaps unsurprising that there was often no clear correlation between these different variables. However, these results do not contradict the idea that quality of care is not directly related to the costs of providing the services and that quality and cost do not need to require a trade-off. We explore this in the economic modelling.

Economic modelling

The economic analysis was frustrated by the data limitations described in the next section, mainly with respect to the availability of detailed staffing data. Despite this, a number of interesting findings emerged from the economic modelling, which were consistent across the two definitions of maternity output: the total number of deliveries per trust and the cost-weighted deliveries per trust. The latter definition was an attempt to control for the greater complexity and resource use involved in performing caesarean births, which is accounted for in the higher reimbursement received by trusts.

To accommodate this, the cost-weighted deliveries combined these two broad types of deliveries based on their relative cost. However, the results were very similar across both models and we can therefore discuss them together. The Hick’s elasticities of complementarity were estimated. This measures the extent to which staff groups are complements or substitutes, i.e. whether they should be used in combination or can be used in place of each other in terms of the number of deliveries a provider can handle each year. This measure indicated that midwives were complements with consultants and other doctors in the production of deliveries, that is, midwives and doctors and midwives and consultants should be used in combination to maximise the number of deliveries a provider handles. Consultants and other doctors were found to be substitutes for each other, as were midwives and support workers. In that sense, there are some tasks that a consultant and more junior doctor can both do, as there are tasks that both midwives and support workers can perform. This echoes findings from research by Goryakin et al. 62 that looks at the relationship between registered nurses and health-care assistants. The elasticities in Table 34 indicate the degree to which staff groups are either complements or substitutes and none of them imply that you simply trade, for example, one FTE midwife for one support worker.

As described in the literature review, this is the first study that we are aware of to look at this issue and, therefore, there are no appropriate studies with which to compare our findings. Thurston and Libby65 employed a similar methodology but applied to primary care physicians in America. They found that all staff groups were complements except for technicians, who were substitutes for both nurses and administrators. Our findings are similar in that midwives and support staff are also found to be substitutes, while most other groups are found to be complements. We also find that consultants and other doctors are substitutes for each other, which was not an issue considered by Thurston and Libby65 because they considered primary care practices and the grade or experience of the physician was not modelled.

The marginal products of all staff groups were positive, indicating that adding an additional worker of any type would increase the total number of deliveries a provider would be able to cope with. Purely in terms of the number of additional deliveries and not considering the cost of adding an additional worker, doctors had the highest marginal productivity followed by consultants, then midwives and finally support workers. Adding an additional FTE worker in each of these categories would allow a hospital to handle an additional 43, 32, 18 and 10 deliveries per annum, respectively. Conversely, reducing staffing by one member of staff in each group would reduce a hospital’s output by an equivalent amount. Given that the national average number of deliveries per FTE in these staff groups respectively is 180, 384, 32.5 and 100, the marginal productivities support the results that most staff groups are complementary, i.e. they should be used in combination. Adding additional workers of one category (e.g. midwives) will not be as productive as adding a combination of all the staff groups: the right skill mix is therefore critical to the efficient operating of a maternity service. However, substituting support workers for midwives may also have an impact on some aspects of the quality of care depending on the groups of women involved or the care setting.

Limitations

Secondary analysis is dependent on the quality of data. We used the full census of women’s deliveries in HES (656,969 delivery records) so there was no bias caused by non-response. Any biases would therefore be caused by missing data, poorly recorded data or omitted variables from the risk adjustment model. A scoring system was used to select records with the largest amount of most useful and relevant data of greatest relevance to the project. Extensive data cleaning was conducted to remove duplicates and records which did not relate to a delivery episode, identifying units with inconsistent or missing data. A decision was taken not to include any trust where fewer than 80% of women could be coded for a particular indicator. This limited the use of some potential indicators.

A common problem when analysing routinely collected data is that it is only possible to work with the available data that are of a sufficient quality to use. For example, we could not include body mass index or an indicator of smoking status in our models because of data quality issues, although they are known to be important risk factors. There may well be other measures of clinical well-being and lifestyle that go beyond the NICE risk classification that would help to reduce variability.

Only a limited set of trust-level organisational variables were used. We were not able to include measures that tell us something about the organisation (e.g. organisational climate, local climate) and models of care that could be predictive of outcome. Our models may also have omitted other variables, either known or unknown, that are predictive of outcome.

Staffing data were available only at trust level so we could not explore the effects of staffing at the unit level. The data for trusts that have multiple units could not be disaggregated. Aggregated trust-level data makes the assumption that unit-level effects within a trust are similar, which may not be true. The staffing data are taken from a census undertaken every September. This single-point estimate will hide any fluctuations that may occur over time. We analysed data that were aggregated over a period of a year. These data will therefore miss those occasions when the service is placed under stress, or reaching a critical point, because of excess deliveries, low staffing levels or other factors.

This study, like many others in the literature, relies on a single cross-section of data that makes causal inference problematic if not impossible. At best we can claim an association between our independent variables and the outcomes. An obvious concern is omitted variable bias. Some potentially omitted variables have been listed above, including inter alia smoking status (for mothers) and a number of trust level variables. It is possible that trusts with higher staffing levels also have higher levels of other inputs that affect organisational performance and the quality of care, such as advanced medical equipment, high performance management teams or a culture of patient safety. A major problem involves the potential endogeneity between staffing levels and the outcomes we have used. In effect, trusts make decisions about factors that have an impact on the quality of care, for example staffing, subject to a set of constraints such as regulation, limited budget and case mix.

Adding additional years of data would allow for some control over unobserved variables that vary across providers but do not vary over time. However, only in an experiment that deliberately (or fortuitously as the result of a policy design) allows for the manipulation and randomisation of staffing levels could researchers make casual claims about the relationship between staffing and outcomes. Therefore, this concern is not unique to this study.

Public and patient involvement

This project has had PPI at every stage of the research project, from the development of the proposal, through undertaking some of the research, providing an advisory role, drafting and commenting on the report, to dissemination of the findings.

As a result the findings will be relevant not only to clinicians, policy-makers and NHS managers, but to pregnant women and their families. A particular impact has been on the choice of indicators used as measures of the quality of maternity care. These include a number of positive measures, rather than just a series of interventions or harms. These have included existing measures such as normal birth. This used a consensus definition agreed by the Maternity Care Working Party,17 but based on a definition originally proposed by the user organisation BirthChoiceUK. Further innovative positive measures which indicate an absence of physical harm are those of intact perineum and delivery with bodily integrity. Being without damage or sutures in the early postnatal days enables women to recover more quickly from the birth and to feel more physically comfortable as they begin the tiring work of looking after a new baby. A trio of further measures reflect the health of the mother, the baby and the mother and baby dyad. These combine the concepts of safety, clinical effectiveness and women’s experience to view how the well-being of both the mother and baby as they emerge from the birth process, whatever that may have been.

Our lay collaborator was also able to undertake part of the research herself, taking on responsibility with the obstetrician (SB) for identifying codes in women’s HES records which might indicate that a women would be considered to be at increased risk of complications for birth. This proved to be one of the most important characteristics in determining outcomes for women, according to the multilevel model, along with parity. This work has implications for future analyses, where outcomes can be stratified by risk and parity, and potentially provide women with more personalised information about their likelihood of particular outcomes.

One of the challenges of involving recent service users in this study has been that it has largely been a paper exercise, analysing routinely collected data and understanding the results of multilevel logistic regression, together with the selection of quality indicators. For someone familiar with quality metrics and HES maternity data and with a basic statistical understanding, this has not been as challenging as it might have been to a less expert user.

Although joining the project team and undertaking some of the research herself has brought benefits to the project, it has required a time commitment that might have been difficult for other service user representatives. In this case, the project was not able to provide full funding for this (as the time commitment and contribution were recognised only once the project had started). However, it is pertinent to consider for the future how such time should be costed in for ‘expert patients’. They may have little academic record but a wealth of experience and generally go unfunded. Universities must consider how to support payment to such self-employed individuals within an increasingly tightly regulated environment.

There also appears to be a trend for those reviewing grant proposals to discount the PPI contribution of experienced service users, and user-researchers who work independently. While having recent service users with a variety of experiences and sociodemographic backgrounds is clearly important to many projects, experienced service users can bring a wider, less personal perspective, bringing together views of many consumers, and often with a knowledge of the research literature and policy from a patient perspective also. It is vital that this not be undervalued by reviewers of grant proposals.

Implications for health care

Much of the variation in outcomes that was measured at trust level was explained by clinical risk and parity. The powerful impact of changing age, parity and clinical risk factors (even without the information about obesity and smoking in this study) have important implications for the ‘fitness’ of the reproducing population in terms of costs as well as outcomes. This study draws attention to the need to consider all these factors in terms of public health policy.

This study is the most comprehensive analysis of all women giving birth in England, which has adjusted for all the clinical risk factors present at the end of pregnancy and prior to labour as outlined in the NICE intrapartum care guidelines. It provides useful data for trusts, clinicians, women and their families in relation to variations in outcome. Knowing the extent to which rates of normal birth, for example, are affected by parity and risk provides a useful basis for planning maternity care for a given population, and for discussing likelihood of different outcomes with women. It also highlights the importance of adjusting for age, parity and clinical risk when comparing outcomes between trusts.

There is some indication that staffing levels have positive and negative effects on some outcomes, and deployment of doctors and midwives where they have most beneficial impact is important. Approximately 1–2% of the total variation in the outcome indicators was attributable to differences between trusts, whereas 98–99% of the variation was attributable to differences between mothers within trusts. There was less variation in staffing level, possibly because of existing staffing standards for the midwifery and obstetric workforce. Levels of midwifery staffing were associated with only 2 of the 10 indicators, delivery with bodily integrity and intact perineum. The models estimated population effects, which may not be a true reflection for all trusts particularly those with more diverse characteristics. Certain multiplicative effects revealed themselves and showed that the effect of staffing upon outcomes sometimes varied according to mother’s parity and clinical risk. However, there is potential here for reverse causation where units with higher proportions of high-risk women increase their staff to meet that demand.

Our findings are based upon current staffing regimes. Any extrapolations based on staffing levels not often encountered in this sample would have to be interpreted with a high degree of caution. The measure for medical staffing was problematic because it was not possible to disaggregate this into contributions made to obstetric care (over the course of a whole pregnancy) and maternal delivery (on one day) as opposed to gynaecology care. So, while the findings are accurate, they were not strictly ‘fit for purpose’ and could have led to bias in estimated effects. The preferred measure to full-time equivalence for all staff groups would have been hours committed to maternal delivery care, or FTE for all groups over a whole pregnancy, birth and postnatal pathway.

The staffing measures related to only one time point. Therefore, our results do not reflect the temporal fluctuations that would be experienced on a daily or weekly basis, where both number of maternities and staff availability would vary. Those critical points when the service is most under strain would remain hidden. A measure of bank and agency staff usage would also have added to our overall understanding, but good-quality data on bank and agency use in NHS maternity care were not available to this study. The results, therefore, assume that what occurs at the aggregated level (year) would have been replicated had staffing data been available over shorter time periods (e.g. a month or a week).

In the era of ‘Big Data’ it is essential that more detailed information on staffing be made available and disaggregated over shorter periods so that questions similar to those pursued in this research can be better answered. As each maternity unit will have its own unique set of circumstances, the acquisition of additional contextual data is an imperative, for example measures of the local and organisational environment that reflect the way the unit is run.

In sum, it is very hard to say what a ‘staffing’ ratio or model is, when the most cost-effective care is continuity of midwifery care antenatally/intrapartum/postnatally where midwives provide care for women either in the community or in an acute trust facility, which could be an obstetric or midwife-led setting. Thus, staffing measures that focus on ratios on a ward will miss this, especially when the care takes place over 6 months both in and out of hospital.

What these data add to what is already known is some indication that midwifery staffing levels do make a difference for a small number of indicators, although we cannot quantify accurately how this could be reliably deployed into workload models.

Managers may wish to exercise caution in increasing the number of support workers who care for higher-risk women. Careful attention needs to be paid to the relationships between staff groups and the potential and limits of skill mix and role substitution. Most staff groups were complements rather than substitutes, suggesting that substituting staff groups may actually harm productivity. Collecting more detailed workforce data at the level of the individual unit and greater detail over the time spent on different activities (e.g. obstetrics vs. gynaecology) would enable a number of the limitations of this research to be overcome.

Data on medical staffing were not detailed enough to ascertain the split between obstetric and gynaecological responsibilities in a trust. Therefore, this analysis investigated the relationships between midwifery staffing levels, where data quality was very good. For this analysis we used trust-level data to investigate relationships between outcome measures, midwifery staffing levels and the cost of providing maternity services for NHS trusts in England.

Higher midwifery staffing levels were associated with higher costs of each delivery, although the relationship was not strong. Only around 17% of the variation between trusts’ delivery costs could be accounted for by variables included in this model. The remaining variation in the average cost of a delivery was not accounted for by maternal characteristics, size of trust, number of FTE registered midwives employed or antenatal spend and must be due to other factors not included in the analysis.

After adjusting for maternal characteristics and trust size, no relationship was found between the proportion of expenditure spent on antenatal care and operative delivery rates, or between higher operative delivery rates and higher delivery costs. Variations in costs between trusts were not related to the numbers of women having operative deliveries.

There was no association between cost per delivery and the normal birth rate, intact perineum rate, or any of the three healthy mother and healthy baby indicators, and women’s experience of maternity care as measured by the average of the CQC scores. A relationship could not be found that explained postnatal costs in terms of variations in operative delivery rates once adjustments were made.

Having a higher proportion of women at increased clinical risk was associated with more expensive maternity care, and area deprivation approached statistical significance. These factors are currently reflected in the maternity pathway payment system. There appeared to be economies of scale across the total maternity episode (antenatal, intrapartum and postnatal care) by trust size, which were increased in trusts on a single site. However, larger trusts were associated with lower scores for women’s experience, although there was no relationship found between trust size and any clinical outcome.

From this study, the increased investment in staff did not necessarily have an effect on the outcome and experience measures chosen, where there was in general no relationship with midwifery staffing levels, apart from a higher intact perineum rate, and delivery with bodily integrity in trusts with higher levels of midwifery staffing.

Data quality in HES data is improving but a significant number of trusts have poor-quality data, and some important variables such as smoking and body mass index are poorly recorded. Measures that aim to influence improvement in data quality, such as the RCOG benchmarking project, are to be encouraged.

Current benchmarking indicators that are derived from HES data do not cover a range of outcomes of importance to women. However, this study has developed new indicators that can be derived from HES data and provide an indication of a healthy outcome such as healthy mother, healthy baby and the healthy mother/healthy baby dyad. These should be considered for inclusion in future work. These indicators provide a wider range with which to assess quality of care in maternity care where the overall aim is to promote health and well-being and prevent harm.

The relative impact of maternity workforce staffing is a key policy topic. There is a paucity of evidence regarding the effects of the number and grade of medical and midwifery staffing, and the impact of skill mix with unqualified staff. The evidence underpinning current workforce models such as BR+ is slight. This study has highlighted some effects, but also shown that the workforce needs to be assessed as a whole rather than separately.

The development of appropriate quality indicators for maternity care is a global initiative and has often focused on measurement of interventions as a proxy. The development of harm-free indicators and indicators of a healthy mother and healthy baby are an important development. In addition, there need to be valid indicators of women’s experience. Our analysis of the CQC women’s experience survey indicates that this may be a fruitful avenue to pursue, in providing another source of data to measure quality.

Recommendations for research

To examine the impact of policy, studies comparing different approaches in similar industrialised nations (e.g. across Europe), should be performed to understand, and potentially modify, current adverse trends for reproduction.

There are wide variations in a range of outcomes that remain after adjustment for sociodemographic and background risk. Further research is required on what may be influencing unexplained variation such as organisational climate and culture, use of NICE guidelines in practice and variation of models of care within trusts.

Further research is required on the validation and use of the new indicators developed to assess quality of care in future years as HES data improve.

The longer-term impact on the baby needs to be assessed by linking the mother and baby records within HES.

Further research is also needed on how to measure more accurately which factors in levels of consultant obstetrician, midwife and support worker staffing impact on outcomes and quality, in addition to productivity.

Further research is required on developing indicators of quality derived from women’s experience and these should be informed by what is important to women and their families.

Contributions of authors

Jane Sandall (Professor of Women’s Health) supervised analysis and drafted report.

Trevor Murrells (statistician/research data manager) developed and tested the risk model, conducted multilevel analysis and contributed to writing the report.

Miranda Dodwell (user representative, BirthChoiceUK and visiting research associate, King’s College London) provided a user perspective, developed and tested the risk classification, conducted costing analysis and contributed to writing the report.

Rod Gibson (data analyst) cleaned and validated the data, developed algorithms, developed and tested the risk classification, conducted costing analysis and contributed to writing the report.

Susan Bewley (Professor of Complex Obstetrics) provided a clinical perspective to development of the risk model and contributed to writing the report.

Kirstie Coxon (research associate) provided project and report management and contributed to writing the report.

Debra Bick (Professor of Evidence Based Midwifery Practice) provided a clinical perspective and contributed to writing the report.

Graham Cookson (Professor of Economic and Public Policy) provided an economic perspective and contributed to writing the report.

Cathy Warwick (chief executive officer, Royal College of Midwives) provided a midwife and support worker workforce perspective and contributed to writing the report.

Diana Hamilton-Fairley (Director of Education and Quality, Health Education South London) provided an obstetric workforce perspective.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health.

Hospital Episode Statistics data

The data set available to the study consisted of:

• HES inpatient records 2000/01 to 2010/11

• HES outpatient records 2003/04 to 2010/11

• HES A&E records 2007/08 to 2010/11

• ONS-HES linked mortality statistics (including neonatal deaths) 2000/01 to 2010/11.

Further details of these data sets are provided below.

Hospital Episode Statistics data cleaning

The information which follows provides an account of full data cleaning conducted for the study, prior to data analysis. The project restricted the records to NHS hospital deliveries resulting in a registrable birth. Duplicate delivery records were removed from the mother’s records.

The babies’ birth records also contained duplicate records. These duplicates were not removed, as the majority of the project’s work concentrated on the mother’s delivery record and the resources were not available to clean both.

Derivation of maternal and infant indicators used in the analysis

The indicators discussed here contributed to study outcomes. Further cleaning was undertaken on the following fields to ensure that each was as accurate as possible before the analysis took place.

Healthy mother: multilevel model