Increased adverse perinatal outcome of hospital delivery at night
Dr EAP Steegers, Division of Obstetrics and Prenatal Medicine, Department of Obstetrics and Gynaecology, Erasmus MC, University Medical Centre Rotterdam, PO Box 2040, 3000 CA, Rotterdam, the Netherlands. Email firstname.lastname@example.org
Please cite this paper as: de Graaf J, Ravelli A, Visser G, Hukkelhoven C, Tong W, Bonsel G, Steegers E. Increased adverse perinatal outcome of hospital delivery at night. BJOG 2010;117:1098–1107.
Objective To determine whether delivery in the evening or at night and some organisational features of maternity units are related to perinatal adverse outcome.
Design A 7-year national registry-based cohort study.
Setting All 99 Dutch hospitals.
Population From nontertiary hospitals (n = 88), 655 961 singleton deliveries from 32 gestational weeks onwards, and, from tertiary centres (n = 10), 108 445 singleton deliveries from 22 gestational weeks onwards.
Methods Multiple logistic regression analysis of national perinatal registration data over the period 2000–2006. In addition, multilevel analysis was applied to investigate whether the effects of time of delivery and other variables systematically vary across different hospitals.
Main outcome measures Delivery-related perinatal mortality (intrapartum or early neonatal mortality) and combined delivery-related perinatal adverse outcome (any of the following: intrapartum or early neonatal mortality, 5-minute Apgar score below 7, or admission to neonatal intensive care).
Results After case mix adjustment, relative to daytime, increased perinatal mortality was found in nontertiary hospitals during the evening (OR, 1.32; 95% CI, 1.15–1.52) and at night (OR, 1.47; 95% CI, 1.28–1.69) and, in tertiary centres, at night only (OR, 1.20; 95% CI, 1.06–1.37). Similar significant effects were observed using the combined perinatal adverse outcome measure. Multilevel analysis was unsuccessful; extending the initial analysis with nominal hospital effects and hospital–delivery time interaction effects confirmed the significant effect of night in nontertiary hospitals, whereas other organisational effects (nontertiary, tertiary) were taken up by the hospital terms.
Conclusion Hospital deliveries at night are associated with increased perinatal mortality and adverse perinatal outcome. The time of delivery and other organisational features representing experience (seniority of staff, volume) explain hospital-to-hospital variation.
Over 70% of Dutch women deliver at hospital.1 At the time of delivery, care is focused on risk surveillance and intervention, if indicated, including assisted delivery and neonatal intensive care. This requires the ready availability of experienced professionals and supportive facilities. High-care facilities and multiple expert competences cannot be represented at all hospitals on a 24-hour/7-day basis, however, because of issues of cost-effectiveness. In the Netherlands, 9% of deliveries are scheduled, but the majority of nonscheduled deliveries occur around the clock, with a biphasic pattern, including a peak – under natural conditions – in the early morning.2
Heterogeneity with respect to facility and personnel coverage around the clock is the rule rather than the exception for most clinical care, even in surgery and intensive care. Studies have shown moderate to strong associations between patient outcomes and organisational features, both with regard to the volume of care and care that is daytime dependent, such as physician staffing and the immediate availability of anaesthesiological services.3–7
In maternal and perinatal care, this evidence is not unequivocal. Different studies have demonstrated that high-risk newborns have better outcomes in high-volume hospitals,8–11 whereas controversy exists in the case of low- and moderate-risk newborns.12–16
Little is known about the interaction between fixed and time-dependent organisational characteristics. The time of delivery may be regarded as an indirect expression of organisational vulnerability, as conditions may be more suboptimal during the evening and night. Indeed, recent studies have suggested that perinatal outcomes are compromised during the weekend and at night,17–24 and a recent analysis in the Netherlands – without elaborating on the specific organisational features – has suggested a similar relation for off-hour deliveries with regard to intrapartum and neonatal mortality.25,26
The objective of this study was to evaluate the role of some organisational features (time of birth, volume of the maternity unit and physician staffing) in the performance of clinical maternity units of nontertiary hospitals and tertiary centres in the Netherlands, whilst adjusting for clinical risks. The organisational features of community midwifery care fall outside the aim of this study, as the organisation of more than 400 independent midwifery practices differs completely from that in hospitals, and possible differences in perinatal outcome in relation to the time of delivery may depend on other mechanisms. The scope of our evaluation was expanded from delivery-related perinatal mortality (about 2% in this population) to delivery-related perinatal adverse outcome, including intrapartum death, early neonatal death, a low Apgar score 5 minutes after birth and admission to a neonatal intensive care unit (altogether about 13% in this population), to enhance the sensitivity of the analysis.
Materials and methods
Patient data were obtained from the Netherlands Perinatal Registry, a linked professional database of all pregnancies, of 20 weeks and above, in the Netherlands, collected from (referring) midwives, obstetricians and paediatricians.27 The registry information consists of maternal demographic factors, pregnancy and labour characteristics, and neonatal outcomes. The study was executed with the explicit permission of the holder of the patient registration data (Netherlands Perinatal Registry), which consists of representatives of all professional caregivers involved in the registry. The statistical analysis presented here was part of the required data application. The permission was subject to the strict requirement of nondisclosure of the identity of any individual hospital, directly or indirectly. Patient data in the Netherlands Perinatal Registry are anonymised.
From 1 January 2000 to 31 December 2006, all deliveries of each delivery hospital and home births in the Netherlands were enrolled. The national coverage is near to complete as registration is compulsory (professional requirement to receive health insurance fees).
Multiple deliveries of two or more were excluded (3.9%, 50 577 of 1 297 017). Home and transferred home deliveries in a hospital under the responsibility of community midwives (35.9%, 447 408 of 1 246 440) were also excluded from the analysis. In the case of the fusion of two hospitals during the study period, the records of both hospitals were combined. Deliveries at hospitals which had not participated in the registry for 2 years or more were also excluded (0.8%, 6078 of 799 032). The final database contained 792 954 births in 99 hospitals.
For the purpose of this study, the homogeneity of setting required the separation of the 10 tertiary perinatal centres (109 858 births, 13.9%). From the nontertiary hospitals, fetal deaths during gestation (n = 2585, 0.4%), children with congenital malformations (n = 17 516, 2.6%) and deliveries with a gestational age before 32.0 weeks or unknown gestational age (n = 7034, 1.0%) were excluded from the analysis. Outside tertiary centres, these deliveries are unexpected and unintended as to their location. Therefore, 655 961 singleton births of nontertiary hospitals remained for further analysis. From the tertiary centres, fetal deaths during gestation and deliveries with a gestational age before 22.0 weeks or unknown gestational age (n = 1413, 1.3%) were excluded from the analysis. From the tertiary centres, 108 445 singleton births were used for further analysis.
The sizes of the nontertiary hospitals were stratified into six categories based on the yearly number of deliveries, yielding about equal-sized groups. Although this number occasionally showed some fluctuation, the categorisation was primarily based on the category that was most prevalent.
All 99 participating maternity units were evaluated by a standard questionnaire to collect information on organisational factors. The information collected referred to teaching hospital (yes or no), and the number of obstetricians, clinical midwives, residents in training and residents not in training. In the Netherlands, after completing medical school and becoming a registered doctor, an individual attempts to obtain a formal training post to become a GP or in one of the specialities, such as obstetrics and gynaecology. Once accepted, the entire scheme of training (for 5 or 6 years) in peripheral and tertiary university hospitals is scheduled up to the moment of registration as a gynaecologist (or other speciality). Not all doctors manage to obtain such training posts immediately. Those doctors take jobs as Obstetrics and Gynaecology Residents (or other specialities), who are not formally in training, in order to obtain clinical experience.
The survey was executed prior to any data analysis. At the time of the study, no information was available on any of the primary study relations, at either the aggregate or individual hospital level. The information providers usually had an administrative background.
Evening-time deliveries were defined as those taking place between 18.00 and 23.59 hours, and night-time deliveries as those taking place between 00.00 and 07.59 hours.
Based on the outcomes of the maternity unit enquiry, a senior index variable for nontertiary hospitals was constructed [(number of obstetricians + residents in training) divided by the total staff of a hospital (obstetricians + residents in training + residents not in training + clinical midwives)]. A higher score implies a better qualified staff. Several weights for the numerator were tried, but the outcomes of these analyses were similar. The primary outcome parameter was perinatal mortality, defined as intrapartum or early neonatal death (number of deaths within 7 days after live birth); perinatal adverse outcome was defined as the presence of any of the following outcomes: intrapartum or early neonatal death (number of deaths within 7 days after live birth), 5-minute Apgar score below 7, or transfer of the newborn to a neonatal intensive care unit after birth.
The statistical software package SAS version 9.1 (SAS Institute Inc., Cary, NC, USA) was used for data analysis. A multiple logistic regression analysis was used to determine whether the organisational factors, additional to pregnancy characteristics such as parity and gestational age, contributed independently to the outcome rates. For this purpose, we first estimated an optimal model including these pregnancy-related variables alone, and calendar year, as perinatal adverse outcomes show an annual trend. The resulting variable set was then treated as a forced case mix control to which organisational variables were added in two steps: first, time at delivery (daytime, evening, night and weekend or public holiday); second, organisational characteristics. To check for the general effect of the hospital on the predictors, we applied a multilevel analysis using the NLMIXED procedure of SAS. The first level was the hospital, which was reflected by a dummy. The second level was the remaining organisational and case mix variables listed previously. The a priori estimates of these variables, as required for the analysis, were drawn from the previous logistic regression analysis. When the model did not converge, we simplified the explanatory structure, re-estimated the basic logistic regression to obtain the priors, ran the analysis and increased the number of iterations. Special hardware was available to accommodate the memory requirements. If the analysis failed, we considered the introduction of a nominal hospital effect in the first analysis, together with an interaction term of hospital with time (evening, night; daytime was reference).
Odds ratios, with 95% confidence intervals, were calculated for each risk factor. A two-sided P value of <0.05 was considered to be statistically significant. Data were presented as frequencies and proportions (%), unless specified otherwise. The influence of maternity unit size (annual number of deliveries) and staff seniority was only analysed for the nontertiary hospitals, because there was little variety in the organisational features of the tertiary perinatal centres.
A total of 655 961 deliveries in nontertiary hospitals was analysed, including 11 118 adverse outcomes (1.7%) [1206 (0.19%) with perinatal mortality]; 108 445 deliveries were analysed in tertiary centres, including 12 705 adverse outcomes (11.7%) [1915 cases (1.8%) of perinatal mortality].
The characteristics of the included births are shown in Table 1. Fifty-three per cent of the total population were nulliparous, 20% were above 35 years and 17% were non-Western. The time at delivery and hospital characteristics are depicted in Table 2. About one-half of the deliveries occurred during the evening or night-time. More than 75% of women were delivered in nontertiary hospitals with a volume of more than 1000 births per year. Tables 3 and 4 show, after adjustment for case mix, the associations of time of delivery and organisational characteristics with perinatal outcome in nontertiary and tertiary centres, respectively.
Table 1. Study population characteristics
|Maternal parity and age group|
|Nulliparous and <25 years||62 058||9.5||9484||8.8|
|Nulliparous and 25–29 years||124 567||19.0||17 030||15.7|
|Nulliparous and 30–34 years||121 888||18.6||19 424||17.9|
|Nulliparous and ≥35 years||41 600||6.3||8707||8.0|
|Multiparous and <25 years||17 023||2.6||3195||3.0|
|Multiparous and 25–29 years||66 525||10.1||10 685||9.9|
|Multiparous and 30–34 years||135 135||20.6||22 039||20.3|
|Multiparous and 35–39 years||74 570||11.4||14 825||13.7|
|Multiparous and ≥40 years||12 595||1.9||3056||2.8|
|Non-western||104 111||15.9||21 593||19.9|
|Western||551 850||84.1||86 852||80.1|
|Gestational age at birth (weeks)|
|38||102 138||15.6||15 706||14.5|
|39||138 476||21.1||21 076||19.4|
|40||159 198||24.3||23 583||21.8|
|41||116 347||17.7||16 846||15.5|
|Mode of delivery|
|Spontaneous||403 019||61.4||66 726||61.5|
|Instrumental vaginal delivery||112 665||17.2||16 084||14.8|
|Elective caesarean section||60 609||9.2||13 497||12.5|
|Emergency caesarean section||79 668||12.2||12 138||11.2|
|Neonatal mortality <24 hours||300||0.05||944||0.87|
|Early neonatal mortality (1–7 days)||303||0.05||440||0.41|
|Apgar score after 5 minutes <7||7300||1.11||4533||4.18|
|Admission to NICU||2612||0.39||10 077||9.29|
|Total adverse outcome||11 118||1.69||12 705||11.7|
|No adverse outcome||644 843||98.3||95 740||88.3|
Table 2. Time at delivery and organisational characteristics
|Time at delivery (hours)|
|00.00–07.59||151 002||23.0||28 752||26.5|
|08.00–18.00||355 922||54.3||54 021||49.8|
|18.00–23.59||149 037||22.7||25 672||23.7|
|Day of the week|
|Saturday||79 101||12.1||13 086||12.1|
|Sunday and public holiday||77 263||11.8||12 965||12.0|
|Monday||97 617||14.8||15 691||14.5|
|Tuesday||103 092||15.7||17 227||15.9|
|Wednesday||103 487||15.8||16 918||15.6|
|Thursday||102 336||15.6||16 556||15.3|
|Friday||98 642||15.0||16 002||14.8|
|Maternity unit characteristics|
|Total number of staff (sum of gynaecologists, residents, both in training and not, and clinical midwives)|
|0–9||214 681||32.7 (20 units)||0||0|
|10–19||250 372||38.2 (48 units)||0||0|
|20–29||190 908||29.1 (21 units)||30 883||28.5|
|Yes||338 519||51.6 (42 units)||108 445||100|
|No||317 442||48.4 (57 units)||0||0|
|Annual number of deliveries|
|<750 deliveries||54 316||8.3 (15 units)||0||0|
|750–999||98 914||15.1 (19 units)||0||0|
|1000–1249||141 869||21.6 (21 units)||16 300||15.0 (2 units)|
|1250–1499||113 956||17.4 (13 units)||49 090||45.3 (5 units)|
|1500–1749||110 807||16.9 (11 units)||0||0|
|≥1750||136 099||20.7 (10 units)||43 055||39.7 (3 units)|
Table 3. Organisational characteristics of nontertiary hospitals. Perinatal mortality and perinatal adverse outcome adjusted for case mix (gestational age, maternal age, parity, mode of delivery, ethnicity and calendar year trend). Odds ratios with 95% confidence intervals
|Time of delivery (hours)|
|00.00–07.59||1.47 (1.28–1.69)||1.28 (1.22–1.34)|
|08.00–18.00||1 [Ref]||1 [Ref]|
|18.00–23.59||1.32 (1.15–1.52)||1.30 (1.24–1.36)|
|Day of delivery|
|Saturday||1.28 (1.03–1.59)||1.01 (0.93–1.08)|
|Sunday||1.13 (0.91–1.41)||1.03 (0.95–1.11)|
|Monday||1.03 (0.84–1.28)||1.00 (0.93–1.08)|
|Tuesday||1 [Ref]||1 [Ref]|
|Wednesday||0.99 (0.80–1.22)||1.08 (1.01–1.15)|
|Thursday||1.16 (0.95–1.43)||1.04 (0.97–1.11)|
|Friday||1.07 (0.87–1.32)||1.06 (0.99–1.14)|
|Annual numbers of deliveries|
|<750||1.08 (0.83–1.40)||1.09 (0.99–1.19)|
|750–999||0.89 (0.70–1.14)||1.16 (1.07–1.26)|
|1000–1249||0.99 (0.81–1.20)||1.05 (0.98–1.12)|
|1250–1499||1.01 (0.83–1.23)||1.00 (0.93–1.06)|
|1500–1749||0.97 (0.81–1.18)||1.06 (1.00–1.13)|
|≥1750||1 [Ref]||1 [Ref]|
|Continuous measure (0.0–1.0)||0.61 (0.32–1.15)||0.49 (0.39–0.61)|
|Yes||1 [Ref]||1 [Ref]|
|No||0.91 (0.77–1.07)||1.28 (1.21–1.36)|
|Pregnancy characteristics (case mix adjustment)|
|32.0–32.6||11.5 (7.9–16.7)||20.5 (18.3–23.0)|
|33.0–33.6||9.50 (6.94–13.0)||12.6 (11.4–14.0)|
|34.0–34.6||4.76 (3.40–6.66)||7.30 (6.62–8.05)|
|35.0–35.6||4.72 (3.58–6.23)||4.42 (4.02–4.86)|
|36.0–36.6||2.58 (1.96–3.39)||2.21 (2.01–2.44)|
|37.0–37.6||1.99 (1.57–2.52)||1.55 (1.42–1.68)|
|38.0–38.6||1.11 (0.89–1.38)||0.96 (0.89–1.03)|
|39.0–39.6||1.11 (0.91–1.35)||0.93 (0.87–0.99)|
|40.0–40.6||1 [Ref]||1 [Ref]|
|41.0–41.6||1.42 (1.17–1.72)||1.21 (1.13–1.29)|
|≥42||1.08 (0.83–1.41)||1.19 (1.10–1.29)|
|Maternal parity and age group|
|Nulliparous and <25 years||1.27 (1.00–1.60)||1.14 (1.06–1.23)|
|Nulliparous and 25–29 years||1 [Ref]||1 [Ref]|
|Nulliparous and 30–34 years||1.25 (1.03–1.52)||1.03 (0.97–1.09)|
|Nulliparous and ≥35 years||1.26 (0.97–1.64)||1.12 (1.03–1.21)|
|Multiparous and <25 years||1.52 (1.05–2.20)||1.01 (0.88–1.15)|
|Multiparous and 25–29 years||1.56 (1.24–1.96)||1.04 (0.96–1.13)|
|Multiparous and 30–34 years||1.60 (1.32–1.94)||1.04 (0.98–1.11)|
|Multiparous and 35–39 years||1.55 (1.24–1.94)||1.05 (0.97–1.13)|
|Multiparous and ≥40 years||1.46 (0.96–2.22)||1.24 (1.08–1.42)|
|Mode of delivery|
|Spontaneous||1 [Ref]||1 [Ref]|
|Instrumental vaginal delivery||1.48 (1.25–1.77)||2.18 (2.07–2.30)|
|Elective caesarean section||1.69 (1.39–2.05)||1.87 (1.75–1.99)|
|Emergency caesarean section||3.04 (2.64–3.51)||2.97 (2.83–3.13)|
|Non-western||1.43 (1.24–1.66)||1.28 (1.21–1.34)|
|Western||1 [Ref]||1 [Ref]|
|Year of registration||0.99 (0.96–1.01)||1.00 (0.99–1.00)|
Table 4. Organisational characteristics of tertiary hospitals. Perinatal mortality and perinatal adverse outcome adjusted for case mix (gestational age, maternal age, parity, mode of delivery, ethnicity and calendar year trend). Odds ratios with 95% confidence intervals
|Time of delivery (hours)|
|00.00–07.59||1.20 (1.06–1.37)||1.25 (1.17–1.34)|
|08.00–18.00||1 [Ref]||1 [Ref]|
|18.00–23.59||1.08 (0.95–1.24)||1.21 (1.13–1.30)|
|Day of delivery|
|Saturday||0.86 (0.70–1.06)||1.16 (1.05–1.30)|
|Sunday||0.84 (0.69–1.04)||1.02 (0.91–1.13)|
|Monday||0.89 (0.73–1.08)||1.06 (0.96–1.17)|
|Tuesday||1 [Ref]||1 [Ref]|
|Wednesday||0.99 (0.81–1.198)||1.07 (0.97–1.18)|
|Thursday||0.94 (0.78–1.14)||1.16 (1.05–1.28)|
|Friday||1.03 (0.85–1.24)||1.08 (0.98–1.20)|
|22.0–27.6||186 (148–232)||∞ (861–∞)*|
|28.0–31.6||13.8 (10.8–17.6)||250 (221–284)|
|32.0–36.6||5.6 (4.40–7.10)||7.63 (6.97–8.35)|
|37.0–38.6||1.57 (1.22–2.03)||1.12 (1.01–1.23)|
|39.0–39.6||0.73 (0.53–1.01)||0.80 (0.72–0.89)|
|40.0–40.6 + unknown||1 [Ref]||1 [Ref]|
|≥41||0.86 (0.64–1.16)||0.92 (0.83–1.02)|
|Maternal parity and age group|
|Nulliparous and <25 years||1.22 (1.00–1.51)||1.14 (1.02–1.28)|
|Nulliparous and 25–29 years||1 [Ref]||1 [Ref]|
|Nulliparous and 30–34 years||1.11 (0.91–1.34)||0.92 (0.83–1.01)|
|Nulliparous and ≥35 years||1.09 (0.85–1.40)||0.86 (0.76–0.97)|
|Multiparous and <25 years||1.15 (0.83–1.60)||1.07 (0.90–1.28)|
|Multiparous and 25–29 years||1.47 (1.18–1.82)||1.01 (0.90–1.13)|
|Multiparous and 30–34 years||1.45 (1.20–1.75)||0.99 (0.90–1.09)|
|Multiparous and 35–39 years||1.60 (1.31–1.96)||0.97 (0.88–1.08)|
|Multiparous and ≥40 years||1.28 (0.90–1.83)||0.98 (0.82–1.17)|
|Mode of delivery|
|Spontaneous||1 [Ref]||1 [Ref]|
|Instrumental vaginal delivery||0.95 (0.77–1.17)||1.51 (1.38–1.65)|
|Elective caesarean section||0.43 (0.37–0.51)||2.80 (2.60–3.01)|
|Emergency caesarean section||0.47 (0.38–0.59)||2.87 (2.65–3.10)|
|Non-western||0.97 (0.85–1.11)||1.07 (1.00–1.15)|
|Western||1 [Ref]||1 [Ref]|
|No||1 [Ref]||1 [Ref]|
|Yes||12.2 (10.8–13.8)||11.6 (10.8–12.5)|
|Year of registration||1.00 (0.98–1.03)||1.09 (1.07–1.10)|
In nontertiary hospitals, relative to daytime, increased perinatal mortality was found during the evenings (18.00–23.59 hours: OR, 1.32; 95% CI, 1.15–1.52) and nights (00.00–07.59 hours: OR,1.47; 95% CI, 1.28–1.69). In tertiary centres, increased perinatal mortality was found during the night only (OR,1.20; 95% CI, 1.06–1.37). Relative to weekdays, perinatal mortality was increased on Saturdays in nontertiary hospitals (OR, 1.28; 95% CI, 1.03–1.59). In both nontertiary hospitals and tertiary centres, perinatal adverse outcome was more frequent during the evenings (OR, 1.30; 95% CI, 1.24–1.36 and OR, 1.21; 95% CI, 1.13–1.30, respectively) and nights (OR, 1.28; 95% CI, 1.22–1.34 and OR, 1.25; 95% CI, 1.17–1.34, respectively). Perinatal adverse outcome in tertiary centres was higher on Saturdays (OR, 1.16; 95% CI, 1.05–1.30).
In nontertiary hospitals, the seniority of staff was inversely associated with both perinatal mortality and perinatal adverse outcome (OR, 0.61; 95% CI, 0.32–1.15 and OR, 0.49; 95% CI, 0.39–0.61, respectively). Moderately small maternity units of nontertiary hospitals with between 750 and 999 deliveries showed an increased risk of perinatal adverse outcome (OR, 1.16; 95%, CI, 1.07–1.26).
When multilevel analysis was applied, none of the four models converged satisfactorily. Partial results showed essentially the same coefficients (with larger confidence intervals; still significant in the case of time categories), but not all the case mix variables (e.g. the 22.0–27.6-week variable) could be estimated (too many so-called ‘empty cells’ in this hospital-specific estimation procedure). The addition of a nominal factor for each hospital and interaction terms for all hospital–time categories to the initial standard regression analysis revealed a significant effect of delivery at night-time in nontertiary hospitals (mortality, all adverse outcome); in tertiary hospitals, these interaction factors decreased the organisational effects to a nonsignificant level.
To our knowledge, this is the first Dutch population-based report to demonstrate a strong association between time of delivery and hospital experience, and adverse obstetric outcomes, with care having been taken to correct for case mix. Delivery at night in all analyses was associated with an increased risk of perinatal mortality and adverse outcome in nontertiary hospitals. Smaller size and a lower seniority index of nontertiary hospitals emerged as two experience-related factors, which are additive if present together. Multilevel analysis to correct for the potential statistical effect of the hospital as a grouping unit of the data generally failed for this purpose, because of the unfavourable characteristics of the data: despite its size, the dataset contained too many ‘empty cells’ for rare outcomes, i.e. cases of mortality or adverse outcome, as specified by the variables, which did not exist in one or more hospitals. The alternative approach of adding nominal effects for each hospital and interaction terms for all hospitals with the time of delivery categories showed delivery at night to be significant in nontertiary hospitals for both outcomes. This analysis, however, yielded a severe loss of power (in nontertiary hospitals almost 300 coefficients were introduced), with the result that experience variables (volume, seniority) were obscured by the nominal hospital categories.
Our study suggests that this association may be related to the lower availability of experienced caregivers. In almost all nontertiary Dutch hospitals, the senior, experienced obstetrician and neonatologist are not present during the evenings, nights and weekends, but on call from home. Therefore, they are less likely to be involved in the judgement of (pre)critical conditions and the initial management of high-risk situations. Studies from intensive care units support such an interpretation. Increased mortality during off-hours in the case of non-24-hour availability of intensivists has been reported,28 and the introduction of 24-hour availability of senior intensivists was associated with decreased intensive care unit complication rates, hospital length of stay29,30 and no differences in mortality between day, evening or night-time.31,32 The same results were found in paediatric intensive care units33 and maternity units.34 Joyce et al.35 observed a reverse correlation between stillbirth rate and consultant obstetric staffing levels. We assume that issues of 24-hour availability extend to the ready availability of anaesthesiologists and neonatologists as well.
Fatigue of the attending clinical midwife or resident may also be involved.36,37 Changes in work schedules from days to nights may be too rapid to allow the circadian system to adapt to the scheduled wakefulness at night, placing many providers in a permanent state of ‘jet lag’ as they attempt to remain awake and work, and subsequently sleep, at the incorrect internal circadian phase.38,39 Such circadian misalignment is responsible for the higher rates of accidents by night-shift industrial workers.40 Independent of the circadian system, acute continuous sleep deprivation has a profound impact on fatigue. After about 16–18 hours of wakefulness, alertness and performance decline rapidly.41 During the night, it is allowed for residents and midwives to take a nap. Furthermore, it has been shown that naps during the night induce post-nap impairment, where alertness and performance are particularly decreased during the 15–30 minutes after waking.36,42 A Dutch law on labour conditions in 2007 issued limitations on resident work hours in an attempt to reduce fatigue-related medical errors. This law was not in force during the study period, when the practice of working for 24 hours was quite normal.
Volume–outcome relationships have been studied frequently in surgery and surgery-related specialities, and show that high surgeon volume and specialisation are associated with improved patient outcome, but high hospital volume is of limited benefit.4,6,7
In our study, the moderately small maternity units with between 750 and 999 deliveries showed a higher risk of perinatal adverse outcome (15%) than the smallest units (7%). Further research may reveal whether this phenomenon reflects chance, selection uncovered by our case mix (e.g. referral to a tertiary perinatal centre at an earlier stage) or true performance.
One limitation of this observational study was our statistical approach of case mix adjustment. We deliberately included intervention during delivery as an adjustment factor. We were aware that, to some extent, this variable is not a true independent predictor of adverse outcome as it partially coincides with outcome. Nevertheless, it represents a useful and pragmatic coverage of risks at onset of delivery, yielding conservative estimates of the remaining factors. Using this procedure, we accepted a degree of overadjustment to arrive at nonexaggerated estimates of the organisational effects. Another limitation was the lack of information on actual staffing levels just before and during each delivery individually. Maternity ward staffing is typically a healthcare organisation issue that, despite its potential importance in clinical and economic outcomes, is not often studied.34 Our seniority index essentially reflects availability, rather than actual use, of senior competence, although the former will always be a precondition for the latter. A third limitation was the use of the hour of birth as a proxy for the time of delivery effect, which implicitly defines this phase to be the time window of highest vulnerability for organisational effects. One may argue that such a critical phase starts, for example, 1 hour before delivery. When we tested the effect of such a change in the time definition (all time period definitions minus 1 hour), the results were unaffected. A fourth limitation was the inability to subject the dataset to a formal multilevel analysis without removing the standard case mix variables. The dataset was less suitable for this approach. Possibly, nonlinear algorithms will enable such an approach in the future. The application of a second best approach to test a hospital effect maintained a significant role for the time of delivery, where this can be interpreted as a conservative estimate in view of the large number of additional parameters introduced by this second approach. A fifth limitation, by intent, was the scope of our study. We did not address specific aspects of transfer, nor the roles of referring community midwives, and the availability and specific roles of anaesthesiologists and neonatologists. Finally, insufficient data were available to allow for an informed estimate of the economic and professional consequences of a change towards institutional concentration and full on-call coverage of off-hours. We could not therefore derive a crude estimate of the costs of constant staffing levels around the clock, allowing for a crude cost-effectiveness analysis if it is assumed that outcome levels in that case would return to those observed in the daytime.
Hospital deliveries at night are associated with increased perinatal mortality and adverse perinatal outcome. The time of delivery and other organisational features representing experience (seniority of staff, volume) can explain the hospital-to-hospital variation.
Disclosure of interest
We declare that we have no conflict of interest.
Contribution to authorship
JPG initiated and developed the core idea, participated in the study design, interpretation of the results and writing of the manuscript, and was responsible for the provision of organisational data. ACJR participated in the study design and interpretation of the results, and was responsible for the data analysis. GHAV initiated and developed the core idea, and participated in the interpretation of the results and writing of the manuscript. CH provided suitable registry data, and participated in the interpretation of the results and writing of the manuscript. WHT participated in the provision of organisational factors, interpretation of the results and writing of the manuscript. GJB was responsible for the study design and co-responsible for the data analysis, and participated in the interpretation of the results and writing of the manuscript. EAPS initiated and developed the core idea and study design, and participated in the analyses, interpretation of the results and writing of the manuscript.
Details of ethics approval
This is not required for this type of study in the Netherlands.
This study was based on data from the Netherlands Perinatal Register. We acknowledge all midwives, obstetricians, paediatricians, nurses and residents who routinely collect the perinatal data for this register and we thank the obstetricians for their contribution of the additional dataset.