*: Ms K. Gottvall, Department of Nursing, Karolinska Institutet, 23300, 141 83 Huddinge, Sweden.
Objective To study perinatal mortality in women booked for birth centre care during pregnancy.
Design Retrospective cohort study.
Setting In-hospital birth centre and standard maternity care in Stockholm.
Population Two thousand and five hundred and thirty-four women (3256 pregnancies) admitted to an in-hospital birth centre over 10 years (1989–2000) and 126,818 women (180,380 pregnancies) who gave birth in standard care during the same period and who met the same medical inclusion criteria as in the birth centre. Multiple pregnancies were excluded.
Methods Data were collected from the Swedish Medical Birth Register. Information on all cases of perinatal death in the birth centre group was retrieved from the medical records.
Main outcome measure Perinatal mortality.
Results No statistically significant difference in the overall perinatal mortality rate was observed between the birth centre group and the standard care group (odds ratio [OR] 1.5, 95% CI 0.9–2.4), but infants of primiparas were at greater risk (OR 2.2, 95% CI 1.3–3.9). Infants of multiparas tended to be at lower risk, but this difference was not statistically significant (OR 0.7, 95% CI 0.3–1.9). These figures were adjusted for maternal age and gestation in multiple regression analyses.
Conclusion Birth centre care may be less safe for infants of first-time mothers.
Alternative models of care, such as birth centres, homelike birth environments and team midwifery models, have been developed to counter the medicalisation of childbirth. Compared with standard care, these models have a stronger focus on childbirth as being a natural family event and on personal continuity of midwifery care. Some of these models have been evaluated by means of randomised controlled trials, but we are only aware of one large trial of birth centre care. This was the Stockholm Birth Centre Trial, published in this journal in 1997.1 It comprised 1860 women who were randomly allocated to either birth centre care or standard care as provided by approximately 75 antenatal clinics and 6 maternity hospitals in the greater Stockholm area. That study showed that birth centre care was associated with fewer interventions, such as obstetric analgesia, induction, augmentation of labour, electronic fetal monitoring, but a higher rate of prolonged labour. However, no statistically significant differences were found in operative delivery rates or perinatal mortality, defined as intrauterine death after 22 weeks of gestation or infant death within 7 days of birth. Eight infants died in the birth centre group (0.9%) and 2 (0.2%) in the standard care group (odds ratio [OR] 4.0, 95% CI 95% CI 0.8–39.2). The higher OR for perinatal mortality in the birth centre group aroused concern that lack of statistical power might have accounted for the non-statistical difference between the two groups.
A systematic review was conducted in order to further explore the reliability of alternative forms of maternity care.2 Comparing such models was however far from straightforward as they differed extensively in care content. That review therefore focussed on forms of care characterised by continuity of midwifery care and included 7 trials and 9148 women. As in the Stockholm Birth Centre Trial, maternal satisfaction was greater and intervention rates were lower, but perinatal mortality was higher in the alternative groups, although not statistically significant (OR 1.6, 95% CI 0.99–2.95). A more recent review, by Hodnett,3 focussing on home-like but institutional birth environments versus conventional hospital care, included six trials, two of which were the same as in the abovementioned review. Again, a home-like setting was associated with greater maternal satisfaction and less intervention, but a non-statistically significant increase in perinatal mortality (OR 1.5, 95% CI 0.8–2.8). The author concluded “Just as an overenthusiastic focus on risk and intervention can lead to unnecessary interventions and avoidable complications for healthy childbearing women and their fetuses, an overemphasis on normality may lead to delayed recognition of or action regarding complications. Caregivers and their clients should be alert to the need for detection and prompt action in the event of unforeseen complications”.
Additional randomised controlled trials of birth centre care and similar forms of alternative birth care should be conducted to further explore the issue of safety. However, since these models have become an established option for maternity care in many countries, the possibility of conducting new trials has diminished. The purpose of this study was to investigate perinatal mortality at the Stockholm Birth Centre over a 10 year period by the best available methodology when randomisation is no longer possible.
This is an observational study comparing perinatal mortality in all women admitted to the Stockholm Birth Centre over a 10 year period versus other women who gave birth in the greater Stockholm region during the same period and who met similar inclusion criteria as for birth centre care.
At the Stockholm Birth Centre, women were cared for by the same team of midwives, starting from early pregnancy, throughout birth and up to the final visit two months after the birth. An obstetrician met the women at antenatal check-ups, but only when medically indicated. The midwives assisted women in labour without the attendance of a doctor and made their own decisions about transfer in labour according to the centre's medical guidelines. Electronic fetal monitors and pharmacological pain relief were not available at the centre, and induction and augmentation of labour were reasons for transfer. High blood pressure, diabetes or gestational diabetes, multiple pregnancy, medication for epilepsy, alcohol abuse, drug abuse and smoking precluded birth centre care. Women with a previous caesarean section were eligible if they had had a subsequent normal delivery. Labour before 37 + 0 weeks or after 42 + 6 weeks precluded birth centre care. From 42 + 0 weeks, women were checked by electronic fetal monitoring and ultrasound scans every second day in the same way as in standard care. In 1997, labour at the birth centre after gestational week 41 + 6 was contraindicated, and from 1998, all women were offered electronic fetal monitoring, but only as an admission test.
The birth centre data in this study comprised all women booked at the birth centre, usually early in pregnancy but in some cases later (mean: 20 weeks; median: 18 weeks), from October 1989 (when the birth centre was opened) to December 1999. Women were included even if they transferred to specialist care during pregnancy or if they moved to some other place in Sweden. The births took place during the period November 1989 to July 2000. Women who participated in the Stockholm Birth Centre trial (October 1989 to June 1993) and who were randomised to the ‘intervention group’ were part of the birth centre group in this study. The only exclusion criterion was multiple pregnancies.
Control group data were extracted from the Swedish Medical Birth Register and included all women with singleton pregnancies who met the same low risk medical criteria as for the birth centre group and who gave birth in greater Stockholm during the same period. As with the birth centre group, women in the control group could have given birth several times during the observation period and each pregnancy was in this context treated as ‘one woman’. Besides adjusting for singleton births, the following were excluded: smokers (15.4) and women with a history of caesarean section (5.8%), diabetes (0.3%), epilepsy (0.2%) or hypertension (0.3%).
The size of the two samples was defined by practical circumstances by including in the birth centre group all women admitted to the centre since its opening, and in the standard care group, all women who gave birth in the same catchments area and who met similar medical inclusion criteria. The study had 80% power (95% CI) to detect a difference in perinatal mortality from expected 5/1000 in the standard care group to 9/1000 (0.48–0.88) in the birth centre group or a rate ratio (RR) or odds ratio (OR) of 1.8.
Data were retrieved from the Swedish Medical Birth Register, which stores information on 99% of all births in Sweden.4 The register is based on information from the standardised antenatal, intrapartum, postpartum and paediatric records. All live births from 22 + 0 weeks and all stillbirths from 28 weeks are included in the register. Perinatal mortality was defined as fetal death from 28 weeks of gestation to infant death within 7 days after birth. Later, infant death was defined as death 7–27 days after birth. Any misclassification of perinatal mortality is likely to be extremely low as the register is linked to the Swedish Cause of Death Register kept by the National Board of Health and Welfare.
All case records of perinatal deaths in the birth centre group were examined by an external and independent obstetrician—a consultant in obstetrics and gynaecology at the National Board of Health and Welfare handling cases of malpractice in Sweden. These cases were analysed in order to establish if the particular model of care could possibly explain the outcome, and each case was classified by using the criteria ‘unavoidable’, ‘possibly avoidable’, and ‘avoidable’.
Comparisons between the birth centre and the standard care groups concerning sociodemographic and obstetric background were conducted by descriptive statistics and χ2 test.
Maternal age was determined at the time of birth. ‘Married or cohabiting’ included cohabitation with the infant's father or other family situation. Body mass index (BMI) (weight in kilograms divided by square of height in metres) was measured in early pregnancy and categorised according to recommendations by the World Health Organisation (underweight: BMI < 18.5; normal: BMI = 18.5–24.9; overweight: BMI = 25–29.9; obese: BMI ≥ 30). Mortality rates were estimated by rate ratios (RR) and 95% confidence in all women and stratified by parity. Logistic regression analyses were conducted in order to control for differences in maternal age and length of gestation.
The Ethical and Research Committee of Karolinska Institutet gave consent to the study (Dnr: 102/2000).
Figure 1 shows the number of bookings at the Stockholm Birth Centre during the 10 year period and the birth centre group, comprising 3256 pregnancies, after excluding miscarriages, multiple pregnancies and 44 pregnancies that could not be matched with information from the national Medical Birth Register. The medical records of these 44 pregnancies were studied, but no case of perinatal death was found. The figure also shows that women who booked for birth centre care but were transferred to standard care antenatally, intrapartum or postnatally nevertheless remained in the birth centre group for the purpose of the current study. The standard care group comprised 180,380 pregnancies in 126,818 women.
The proportion of primiparas was slightly lower in the birth centre group (45.4%) than in the standard care group (48.2%). Table 1 shows that women in the birth centre group were older, while of the primiparas, fewer were cohabiting with the infant's father. In the birth centre group, obesity was less common. Estimation of the date of delivery by ultrasonography was less common in the birth centre group. The duration of gestation also differed between the two groups, showing a lower rate of preterm delivery and a higher rate of postterm delivery in the birth centre group. Induction of labour, caesarean section and instrumental vaginal deliveries were less frequent in the birth centre group, as also was infant birthweight less than 2500 g.
Table 1. Sociodemographic background and obstetric data of women in the birth centre group (BCG) and standard care group (SCG). Figures indicate percentages.
BCG (n= 3256)
SCG (n= 180,380)
BCG (n= 1479)
SCG (n= 86,906)
BCG (n= 1777)
SCG (n= 93,474)
Married or cohabiting
Overweight or obese
Expected date of delivery by ultrasound
Length of gestation (weeks)
Instrumental vaginal delivery
Infant birthweight (grams)
Perinatal mortality was 5.5 per 1000 births in the birth centre group and 4.8 in the control group, a non-statistically significant difference (Table 2). However, analysis by parity revealed a statistically significant increase among primiparas in the birth centre group, 9.4/1000 compared with 5.2/1000 in the control group (RR 1.8, 95% CI 1.06–3.00), and a non-significant decrease in multiparas, 2.2/1000 compared with 4.5/1000 in the control group. Fetal deaths before the onset of labour and infant deaths after the birth did not differ statistically but the intrapartum death rate was higher in the birth centre group, this increase appearing in the group of primiparas. Table 2 also shows that the infant death rate at 7–27 days after the birth did not differ statistically between the two groups.
Table 2. Fetal and infant deaths in the birth centre group and standard care group. Values are given as n (%) and RR [95% CI].
BCG (n= 3256)
SCG (n= 180,380)
RR [95% CI]
BCG (n= 1479)
SCG (n= 86,906)
RR [95% CI]
BCG (n= 1777)
SCG (n= 93,474)
RR [95% CI]
Perinatal mortality: all deaths from 28 weeks of gestation to 7 days after birth.
Death before labour
Death postpartum 0–6 days
Infant death after birth 7–27 days
Table 3 shows that maternal age (>35 years), as well as preterm (<37 weeks) and postterm (≥42 weeks) delivery were associated with an increased risk of perinatal death, and controlling for these variables in the multivariate analysis increased the risk of perinatal death in the Birth centre group to RR 1.5 (95% CI 0.9–2.4) and in primiparas to RR 2.2 (95% CI 1.3–3.9).
Table 3. Perinatal mortality in the birth centre group and standard care group adjusted for gestational age and maternal age.
n (incidence of perinatal deaths)
Model of care
Birth centre care
Maternal age (years)
Table 4 provides details on all the 18 perinatal deaths in the birth centre group. The external reviewer found one case where intrapartum death might have been avoided if standard care had been applied after transfer to the ordinary labour ward 8 hours before delivery. This was a primiparous woman who gave birth in 41 + 6 weeks and who was monitored as a natural childbirth candidate even after transfer despite a very complicated postdate labour (case 18). In three other cases, the outcome might have been influenced by how care was provided. In one case of intrauterine death before labour, the care provider was criticised for not checking the fetal hearth rate by cardiotocography (CTG) soon enough. In two cases of intrapartum death, the outcome might have been prevented by CTG monitoring, although in one of these cases, the mother was opposed to its use.
Table 4. Details of all cases of perinatal death in the birth centre group.
Method of EDD
Gestation when booking at BC
Time of transfer
Indication for transfer and other comments
Gestation at birth (gwk + days)
Cause of death
Comments by external audit
During pregnancy (gwk + days)
In labour (hours from birth)
Perinatal mortality: all deaths from 28 weeks of gestation to 7 days after birth.
36 + 1
Suspected growth retardation
40 + 0
Intrauterine growth retardation
The woman was not transferred immediately when growth retardation (>22%) was diagnosed. Would however not have altered the outcome.
Moderately increased blood pressure (145/100 mmHg), no proteinuria. Back to birth centre, normal blood pressure rest of pregnancy. IUFD at check-up at 41 + 0 weeks
41 + 2
Umbilical cord complication. Cord five times around the body and twice around an arm.
Spontaneous rupture of membranes at gestation 24 + 4
25 + 1
Immaturity + intrauterine infection. Death after 39 minutes
Spontaneous rupture of membranes at gestation 22 + 1
22 + 5
Immaturity. Death after 40 minutes
Slow progress. Cervix: 5–6 cm. Established labour in 33 hours before birth (4 cm to birth in 33 hours)
41 + 6
Meconium aspiration. Death after 55 minutes
The woman was treated as a natural childbirth candidate even after transfer, in spite of a very complicated labour.
This study found no statistically significant difference in overall perinatal mortality between women booked for birth centre care during pregnancy and women in standard maternity care. However, when stratifying by parity, the perinatal mortality rate was higher in the primiparas and lower, although not statistically significant, among the multiparas in the birth centre group.
As a result of the adjustment for smoking, multiple pregnancy and low risk medical criteria, the standard care women were more low risk than the general population of women who gave birth in Stockholm during the same period. This was reflected in a lower perinatal mortality rate, 4.8/1000 in the standard care group compared with 5.6/1000 in the population from which the sample was drawn. Even if we succeeded in making the standard care group more low risk at baseline than the general population, the birth centre group was still probably even more low risk. Some risk factors for perinatal mortality were underrepresented in the birth centre group (e.g. maternal overweight,5,6 a history of infertility,7 preterm birth and low infant birthweight).8 We also know from a previous study that women who choose birth centre care differ from other women, not only by running a lower risk of medical complications but also by being older, better educated and by being more concerned with the psychological aspects of childbirth and by having a more critical attitude to medical interventions.9 In some cases, such attitudes may increase the risk of an adverse outcome by delaying necessary intervention or transfer to standard care. Taking all the differences in women's background and risk status into account, we conclude that the birth centre women did not run a higher risk of perinatal mortality, but quite the contrary.
In birth centre care as well as in standard care, a detailed analysis of the individual cases of perinatal mortality would probably reveal events that could have been avoided or managed differently. We decided to analyse cases of perinatal mortality only in the birth centre group. Apart from the difficulties associated with tracing and analysing all the 874 hospital records in the standard care group, the observed differences in mortality rates favoured the standard care group. Consequently, we tried to elicit additional information about these cases in birth centre group in order to explain the possible causes of the observed differences.
Analysis of the 18 perinatal deaths did not allow of any valid conclusions about possible unsafe practices. The external reviewer identified only one case where perinatal death might have been avoided and three cases where it could have possibly been avoided. However, the individual cases identified aspects of birth centre care that may have caused an increased risk, and these issues will be discussed further.
Transferring women during labour may be associated with some delay if the birth centre is not located near to a delivery ward. The staff on the delivery ward may be influenced by the probability that a woman transferred from a birth centre may favour ‘natural childbirth’ and therefore continue to manage labour as if the woman were still low risk when actually she has become high risk (see case 18).
Two of the intrapartum deaths could possibly have been avoided by CTG monitoring. One (case 12) occurred when CTG was not available at the Centre. This woman declined electronic fetal monitoring, which may have made the midwife less inclined to transfer when she recorded tachycardia by the stethoscope. In the other case (case 13), CTG was used as an admission test but not during the prolonged second stage. In case 14, the external reviewer would have recommended more careful monitoring after the fetus had reacted with deceleration to audit stimuli. Labour was monitored intermittently by CTG, and the trace was normal 1 hour prior to fetal death. The probable cause of death was infection with Group B streptococci, and possibly unavoidable according to the external reviewer. This was supported by the National Board of Health and Welfare to which this case was reported. These three cases illustrate that CTG may have made a difference, but even when it is available, it is not necessarily used and interpreted in an optimal way.
Another potential risk factor with birth centre care is the philosophy that emphasises a strong belief in the natural process. As stated by Hodnett3 in the initial quotation to this article, this may delay the recognition of or action regarding imminent complications. This may be a more important explanation than the issue of CTG in the cases of intrapartum death discussed above and may also explain a possible delay in case 10.
The higher rate of postterm pregnancy in the birth centre group (Table 1) may be, to some extent, related to the lower labour induction rate in this group. Other reasons, such as different methods of estimating the date of delivery, must be considered. An ultrasound-dating scan was made in 59% of the pregnancies in the birth centre group compared with 74% in the standard care group. Date of delivery was estimated by ultrasound in seven of the nine cases of postterm perinatal death and in three of the four cases deemed by the external reviewer to have been avoidable or possibly avoidable. Consequently, the method of estimating the date of delivery cannot explain the higher proportion of postterm pregnancies among the cases of perinatal death in the birth centre group. During pregnancy, the birth centre used the same routines as standard care for postterm pregnancies. From 42 + 0 weeks, CTG monitoring and ultrasound examinations were conducted every second day, and induction for postmaturity was initiated at 43 + 0 weeks. A difference between the two models of care was the management during labour. During the major part of the observation period of this study, women with normal test results could still labour and give birth at the centre until 42 + 6 weeks. As stated earlier, this policy was changed in 1997.
The three cases of intrapartum death ranged from 41 + 3 to 42 + 0 weeks (cases 12, 13 and 14) suggesting that prolonged pregnancy even before 42 weeks may require special attention. This conclusion is supported by studies by Ingemarsson and Källén,10 who showed a higher rate of intrauterine death in primiparas at 41 weeks and beyond, and by Divon et al.,11 who found a small but significant increase in fetal mortality in accurately dated pregnancies that extended beyond 41 weeks of gestation. Whether these findings are strong enough to motivate induction of labour earlier than the current policy for postterm pregnancy is a question beyond the scope of this report.
This study found a more favourable perinatal outcome in the multiparous women in the birth centre group than in the standard care group. Perinatal mortality and infant death 7–27 days after birth were both lower, but these differences were not statistically significant and may have occurred by chance.
This study showed that birth centre care was associated with an increased risk of perinatal mortality in primiparas, but not in multiparas. These findings raise questions about medical guidelines and the selection and care of women who wish to give birth in birth centres. Birth centres in different countries share the same philosophy of encouraging a natural childbirth, but guidelines and clinical practices may differ. Additional studies from other birth centres are needed in order to draw general conclusions about the safety of birth centres. This should be considered when new centres are opened and when there is a possibility to conduct a randomised controlled trial that controls for the selection of women to these birth options. When samples are too small to allow conclusions about perinatal mortality, infant morbidity should be studied. One such study, based on the sample of this study, is in progress.
The authors would like to thank Professor Ingemar Ingemarsson, senior obstetrician at the Department of Obstetrics and Gynaecology, Lund University, for reviewing the cases of perinatal death in the birth centre group. This study was funded by the Vårdal Foundation for Health Care Sciences and Allergy Research and supported by South Hospital, Stockholm.