Ethical Approval

Section 251 approval was granted by the Health Research Authority Confidentiality Advisory Group to process patient identifiable information without consent for the purposes of service evaluation. (CAG 2-06(a)/2013).

Data Source

We used data extracted from the electronic maternity information systems (MIS) of 19 obstetric units across the UK that participated in the RCOG MIS Pilot Project. This project aimed to assess the feasibility of creating a national dataset using electronic patient records. The units were selected following a national call for participation from the RCOG. Ninety units responded positively and 25 were shortlisted on the basis of their size, geographic location, and type of MIS. HEK conducted follow-up telephone calls with the clinical director and data midwife at each unit to determine their ability to supply the required data item and their willingness to participate in the pilot. Following these telephone calls, 19 of the 25 hospitals confirmed that they were able to participate. Each hospital supplied a retrospective 12-mo extract of patient-level MIS data in accordance with a pre-defined specification (S1 Text). The extracts were pooled to create a single database comprising 112,458 infants born between 1 April 2012 and 31 March 2013, representing approximately 15% of the total number of births in the UK during this period.

The participating hospitals ranged in size from 1,800 to 9,800 deliveries per year. Two were large specialist women’s hospitals, 15 were teaching/university hospitals, and two were district general hospitals. Fifteen of the hospitals were located in England, one in Scotland, one in Wales, and two in Northern Ireland. All had an obstetric unit able to provide the full spectrum of obstetric care.

Study Population

The records of women who had a singleton birth were extracted from the database, excluding deliveries before 28 completed weeks of gestation. We also excluded women who had a CS prior to the onset of labour because these are predominantly planned in advance and performed during normal working hours, carrying a low risk of neonatal death [16,17]. Hence, inclusion of such cases could lead to an over-estimate of the relative risk of adverse perinatal outcomes for “out-of-hours” deliveries. We could not assess the impact of consultant presence on perinatal mortality due to the rarity of the outcome and the size of the cohort. Moreover, we were not able to distinguish between antepartum and intrapartum stillbirths in the dataset. Antepartum deaths account for six in seven stillbirths [18] and in most cases occur some days prior to the delivery of the baby. The cohort was therefore restricted to livebirths (Fig 1). We did not exclude women with other co-morbidities or complicating risk factors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Selection of the cohort.

https://doi.org/10.1371/journal.pmed.1002000.g001

The dataset included information on various maternal characteristics, including age, ethnicity, body mass index (BMI), and smoking status at booking. Obstetric risk factors included baby’s birthweight, gestational age at delivery, parity, previous CS, fetal presentation, baby’s sex, induction of labour, and mode of delivery.

“Consultant presence at time of birth” was defined as a binary variable by combining data on the time of birth with data from the consultant rotas for the period 2012–13 at each participating hospital. The rotas were obtained from the clinical directors of each maternity unit. Consultant presence refers to dedicated on-site labour ward cover, without other commitments such as antenatal or gynaecology clinics or theatre lists. For each day of the week, the time (in hours and minutes) that scheduled consultant presence on the labour ward began and ended was compared to the date and time of birth to define whether a delivery occurred within of a period of consultant presence (“in-hours”) or outside it (“out-of-hours”). The pattern of consultant presence across the week among the 19 hospitals is summarised in S1 Fig.

Outcomes

Outcomes were selected to reflect different aspects of perinatal morbidity. Neonatal outcomes were measured using Apgar score < 7 at 5 min, umbilical cord pH < 7.1, and admission to neonatal care. Maternal outcomes were described by third or fourth degree perineal tear and PPH > 1,500ml, and the rates of intrapartum CS and instrumental delivery were used as indicators of obstetric activity. All hospitals collected Apgar score. Information on perineal tears, cord pH < 7.1, admission to neonatal care, and PPH was supplied by 18, 17, 16, and 15 hospitals, respectively.

Statistical Analyses

We did not publish or pre-register a plan for this analysis. The analysis plan is described below, with any deviations noted in S2 Text.

We used proportions and medians to summarise the distribution of patient characteristics and the chi-square test and Kruskall-Wallis test for comparisons of dichotomous and continuous variables, respectively.

Multilevel multivariable logistic regression was used to estimate the crude and adjusted effects of consultant presence on the various outcomes, with the hospital of delivery modelled as a random-intercept. The potential confounding variables controlled for in all models were: maternal age (years), ethnicity (white, Asian, black, other, unknown), BMI (kg/m²), smoking status (smoker, non-smoker/ex-smoker, unknown), parity (0, ≥1), previous CS (yes, no), gestational age (completed weeks), fetal presentation (cephalic, non-cephalic), baby’s sex (male, female) and birthweight (g). For the continuous variables (maternal age, gestational age, and birthweight), quadratic terms were included in the models because there is clinical evidence that the relationship between these risk factors and the outcomes of interest is non-linear. Parity and previous CS were combined into one variable as these are not independent variables.

The completeness of data for the explanatory variables was generally good. No records were missing time of birth. Parity, mode of delivery, onset of labour, gestational age, birthweight, baby’s sex, and birth status were more than 99% complete. Patients missing one or more of these variables were dropped from the cohort (Fig 1). Ethnicity, BMI, and smoking status were missing in more than 1% of records, and we assigned missing values to a category of “unknown.” Apgar score was over 95% complete for all hospitals.

A sensitivity analyses limited to births at term (≥37 wk of gestation) was conducted to explore the possible risk of confounding due to preterm birth. Prematurity typically accounts for a significant proportion of adverse neonatal outcomes, and the inclusion of preterm deliveries could mask any out-of-hours effect observed among term deliveries. For example, a 30-wk infant will be admitted to neonatal care irrespective of the time of delivery or the care provided.

All statistical tests were two-sided and the level of significance was set at p < 0.05. All analyses were performed in Stata version 13 (StataCorp, College Station, TX, United States).

Comparison with Other Studies

Given the background of confidential enquiry reports highlighting that many cases of poor neonatal or maternal outcomes are linked to the failure to recognise and act on problems arising in labour [22–24], it might be expected that our study would show differences in outcomes during periods of time with and without consultant presence. Previous studies examining whether perinatal mortality and morbidity rates vary according to time of birth and have produced inconsistent findings. Some studies reported no difference [25–31], whereas others reported increased risks of mortality for births during the weekend [4,32–36], and/or the night [4,19–21,37,38]. A recent study reported some evidence of a “weekend effect” in perinatal mortality in England [36], although it was criticised for failing to exclude antepartum stillbirths from the outcome measure, leading to “unjustified extrapolations of what the results mean in terms of avoidable harm” [39].

That these differences are not apparent in this study may be due to various factors. Our study is smaller than those that are based on national data or use cohorts spanning several years and is therefore less likely to detect statistically significant differences in outcomes. Our study is also unique in that it uses a more nuanced exposure variable, which is likely to be a more accurate proxy of senior input than time of birth alone. As far as we are aware, no other multicentre studies have examined the extent to which variation in consultant presence on the labour ward is associated with maternal and neonatal outcomes.

Second, the results from several previous studies describe patterns of care among births that occurred during the 1980s and 1990s [4,19–21]. There have been considerable changes in the obstetric evidence base, diagnostic technology, and clinical governance since that time, which have improved the safety of NHS maternity services.

Third, since the mid-2000s, it has been recommended that UK obstetric units with over 2,500 deliveries annually have at least 40 h of consultant presence per week, and that larger units with over 5,000 annual deliveries have at least 60 h of consultant presence [11]. In their 2008 report, the Healthcare Commission reported that only 68% of English NHS trusts met the recommended standard. They also reported that roughly one in five midwives and one in four doctors thought more consultant obstetrician presence was needed on their labour ward [6]. The latest RCOG Workforce Census reported that by 2013, the mean number of hours of consultant presence on the labour ward in UK obstetric units had increased to 63.5 [12]. The 19 obstetric units that participated in this study had a mean of 75 h of consultant presence per week, which may be greater than in previous studies.

Finally, it is now expected that a consultant obstetrician should be available within 30 min outside the hours of consultant presence, and any risks associated with on-call cover may have changed [40].

Methodological Considerations

A strength of this study is that it is based on a large, multicentre dataset of over 87,000 deliveries that occurred in 2012–13. It therefore provides a description of recent practice in the UK across a range of obstetric units. We excluded planned CS deliveries, as these carry a low risk of neonatal mortality and are predominantly carried out during “office hours.” Their inclusion could have led to an overestimate the risks of out-of-hours deliveries.

The study was also focused on severe morbidity, which has been suggested as a better indicator of the quality of intrapartum care in high-income countries than mortality [41], both for mothers [42–44] and neonates [45]. Mortality is now a rare complication of childbirth in high-income countries, and perinatal mortality rates in the UK are at their lowest recorded levels, at 6.0 per 1,000 live births [18]. Moreover, national data suggests that 86% of stillbirths involve death of the baby prior to the onset of labour, and the majority of neonatal deaths are due to anomalies and preterm birth, with intrapartum complications being an uncommon cause [18]. Low Apgar score has been shown to be strongly associated with the risk of neonatal and infant death ascribed to intrapartum hypoxia [46]. Cord pH < 7.1 is also an objective measure of fetal acidosis.

A weakness of observational studies is the potential for “bias by indication.” We minimised this risk by excluding multiple and very preterm infants with a high probability of an adverse outcome and, in that way, including a relatively homogenous group of deliveries. Furthermore, we also carried out a sensitivity analysis confined to term deliveries ≥37 wk. Finally, we risk-adjusted outcomes according to relevant maternal and obstetric risk factors.

Another limitation of the study was that we only had access to consultant rotas for each of the hospitals, which allowed us to calculate a hospital-specific out-of-hours classification of routine consultant presence. Consequently, the results are likely to be an underestimate of the effect of delivering out-of-hours compared to a study that would have information about consultant presence for each delivery. Serious complications arise for a minority of women and babies, and to detect the potentially small differences in outcomes that occur in childbirth, it might be necessary to have data on consultant presence at an individual-patient level. Other aspects of the organisation and delivery of maternity care, such as the availability and grade of other staff such as midwives, trainees, and healthcare assistants; patient triage protocols; or deviations from the rotas due to staff absence or vacancies may be important but were not investigated in this study. Some neonatal outcomes may also be related to non-obstetric medical staffing, such as availability of senior paediatricians. A related issue is that we only studied the association between selected outcomes and consultant presence at the time of birth. It is important to note that outcomes may also be influenced by whether or not a consultant is present earlier during labour, when crucial decisions are being made.

Finally, the hospitals in our study may not capture the full variation in obstetric care and outcomes in UK hospitals. The hospitals cover all of the geographic regions of the UK and include three district general hospitals with fewer than 5,000 deliveries per year. Nonetheless, most participating units were teaching or university hospitals with more than 5,000 deliveries per year, and all were selected from a list of hospitals who were able to provide electronic MIS data.

Taken together, the available evidence provides some reassurance that the organisation of maternity care in the UK allows for good planning and risk management. This suggests that politically driven efforts to target resources at increasing senior obstetricians attendance out-of-hours may not, in fact, lead to improved clinical outcomes for women and babies. However, there is a need for more robust national evidence on the quality of care delivered at all times of the week by maternity units employing different models of labour ward staffing. Ideally, studies should also consider longer-term outcomes, including cerebral palsy and school attainment.