care with standard maternity services ... Objective To review randomised controlled trials of alternative maternity services characterised by ..... midwife managed.
British Journal of Obstetrics and Gynaecology November 1998, Vol. 105,pp. 1160-1 170
A systematic review comparing continuity of midwifery care with standard maternity services *Ulla Waldenstrom Professor, ?Deborah Turnbull Senior Lecturer (Epidemiology) *Department of Nursing, Karolinska Institute. Stockholm, Sweden; f Department of Public Health, University of Adelaide, Australia
Objective To review randomised controlled trials of alternative maternity services characterised by continuity of midwifery care. Methods A systematic review of randomised controlled trials, analysed on an intention to treat basis, in which the study intervention was characterised by a midwife or small group of midwives providing care from early pregnancy to the postnatal period (defined as that provided on the postnatal ward); and the controls by standard maternity care as practised in the place where the trial was conducted. The seven trials identified included 9148 women. Main outcome measures were interventions during labour, maternal outcomes and infant outcomes. Results The alternative models with continuity of midwifery care were associated with less use of obstetric interventions during labour (eg, induction, augmentation of labour, electronic fetal monitoring, obstetric analgesia, instrumental vaginal delivery and episiotomy). However, the caesarean section rate did not differ statistically between the trial groups (OR 0.91; 95% CI 0.78 to 1.05). The lower episiotomy rate in the alternative models of care (OR 0.69; 95% CI 0.61 to 0.77) was associated with a significantly higher rate of perineal tears in the pooled alternative groups (OR 1.15;95% CI 1.05 to 1.26). The percentage of intact perineums was very similar for the two groups (OR 1.11; 95% CI 1.00 to 1.24). There was no maternal death, and rates of maternal complications based on unpooled estimates did not show any statistically significant differences.The proportion of babies with an Apgar score < 7 at five minutes after the birth was approximately the same in the pooled alternative groups as in the control groups (OR 1-1395% CI 0.69 to 1.84). Admission to intensive care or special care baby unit was similar (OR 0.86; 95% CI 0.71 to 1.04). The difference in perinatal deaths was bordering on statistical significance (OR 1.60; 95% CI 0.99 to 2.59).
Conclusion Continuity of midwifery care is associated with lower intervention rates than standard maternity care. No statistically significant differences were observed in maternal and infant outcomes. However, more research is necessary to make definite conclusions about safety, for the infant as well as for the mother. This review illustrates the variation in the different models of alternative and standard maternity care, and thus the problems associated with pooling data from different trials.
INTRODUCTION Over the last 35 years various alternatives to standard maternity care have such as freestanding and in-hospital birth centres, home-like within Ordinary wards, and birthing teams of midwives providing comprehensive care throughout the childbirth year within standard hospita1 settings. These models have usually evolved as a reaction to what has been seen as a ‘medicalisation’ of normal birth, including high levels of obstetric interventions and insufficient attention to women’s emotional and social needs. Co-ponden~: Professor U. Waldenstrijm, Department of Nursing, Karolinska Institute, Box 286, 171 77 Stockholm, Sweden.
1160
In recent years the alternative care models have focussed on continuity of care, and particularly that provided by midwives, as a key concept in endeavours to increase women’s satisfaction with care. The definition of continuity of care has been debated in the literature, with research describing the concept in a range of different ways such as knowing the delivseeing the Same doctor or midwife ering each time or visit2,3, and a desire to be cared for by people who are known and trusted by the labouring WOman3.
Outcomes of the new have been reported mainly in descriptive and nonrandomised comparative studies, but Over the last few years randomi controlled trials have also been published. The principal outcomes in all these studies have been women’s 0 RCOG 1998 British Journal
of
Obstetrics and Cynaecology
STANDARD CARE VERSUS MIDWIFERY CARE
satisfaction with care, use of obstetric interventions, and maternal and infant morbidity and mortality. The results have suggested that the alternative models are associated with greater satisfaction and fewer interventions, but no single study has been large enough to make conclusions about safety. This has prompted a request for meta-analysis. While the study of perinatal mortality seems justified by metaanalysis, the total number of participants in published trials is still too small to provide sufficient power for the study of maternal mortality4. Thus the aim of this systematic review was to examine the evidence for alternative models of maternity care characterised by continuity of midwifery care.
METHODS Two essential characteristics for inclusion were defined before the review commenced. In order to be included a study had to be a randomised controlled trial and be analysed on an intention to treat basis. The third criterion rested on the definition of continuity of midwifery care. General consensus about this definition was reached before commencing the review. As the review progressed, however, the complexity of the different characteristics of the models of care became apparent, for example, with respect to features such as the involvement of medical staff. For this reason, the definition of continuity of midwifery care was of necessity decided during an ongoing collaborativeprocess. We defined continuity of midwifery care as that provided by a midwife or small group of midwives from early pregnancy to the postnatal period (as defined by that provided on the postnatal ward). In order to be included it had to be clear that the midwife's role in the experimental group was different in terms of responsibilities and spheres of practice. That is, the midwife had to be practising clinically across the three stages of pregnancy and childbirth. The following features could vary from trial to trial: 1. Birth setting. 2. Whether the care programme included routine input from a medical practitioner. 3. Continuity of carer (the extent to which the woman saw the same midwife). 4. Whether explicit criteria existed for transferring women to the obstetric team in the event of deviation from normal. 5 . Whether an explicit policy existed with respect to restricting medication and technology. 6. Whether the midwives involved in the alternative model provided postnatal care in the home. Further, the specific elements used to define high 0 RCOG 1998 Br J Obstet GynaecollO5, 1160-1170
1161
and low risk seemed to overlap in the trials. For example, one of the included trials-the Stockholm Birth Centre Tria15-while designed for women deemed to be at low risk, included women that many of the other trials excluded such as those with a previous caesarean section or perinatal death. By comparison, one of the trials included from Australia6 included both high and low risk women, but used a list of risk factors designed for birth centre care7.For this reason we decided to select trials regardless of the risk status at booking. Using our definition, continuity of midwifery care (the intervention) was compared with the standard pattern of maternity care as practised in the place where the trial was conducted. Outcome variables were defined in terms of obstetric interventions and maternal and neonatal outcomes. This compares with systematic reviews which place different emphasis on birth setting (home-like versus conventional hospital)8, care provider (predominantly midwives versus medicahhared care)g, and continuity of care (same team of carers during the antenatal, intrapartum and postpartum period versus different carers)lO.We chose our definition because policy initiatives such as the Changing Childbirth report' have concentrated on this issue, and similar models of care are being increasingly adopted in the UK and other western countries such as The Netherlands, New Zealand and Australia'*. Trials were identified by computerised literature searches using Medline (1966-1997), Cinahl (1982-1997), Current Contents (Life Sciences/ Clinical Medicine, 1993-1 997) and Sociofile (19741997). Search terms were: midwife, care, pregnancy, alternative, birth, service, models of care, clinical-trial, and trial. Studies were also identified by scrutiny of congress proceedings (International Congress of Midwifery) and through contacts with investigators. The first investigator checked each study for inclusiodexclusion as per the criteria listed above. These were then checked by the second author. In order to make decisions about whether a study met our three criteria, we checked study titles and abstracts and when necessary, the published reports. The final decision about whether a trial should be included was made collaboratively by the two authors. All data were extracted by the first author and then checked by the second author. When data were not available in the published papers the principal investigators were contacted for additional information. The following data were extracted from the studies: strategy for allocation concealment, number of women raidomised, proportion of prkiparae,
'
1162 u .
WALDENSTROM & D. T U R N B U L L
loss to follow up, entry criteria and description of models of care in intervention and control groups. Data on the following variables were extracted: obstetric interventions such as induction, augmentation of labour, electronic fetal monitoring, obstetric analgesia (epidural and pethidine or other narcotics), and operative delivery (caesarean section, instrumental vaginal delivery); maternal outcomes such as episiotomy, perineal status, duration of labour, morbidity, mortality and satisfaction with care; infant outcomes, such as admission to intensive care or special care baby unit, Apgar score c 7 at 5 minutes and perinatal mortality. Some figures were re-calculated in order to make consistent comparisons based on comparable denominators. The rates of caesarean sections, instrumental vaginal deliveries and inductions were based on the total number of deliveries; augmentation of labour, electronic fetal monitoring and obstetric analgesia on all deliveries except elective caesareans; episiotomies on vaginal deliveries, and infant outcomes on the total number of babies including twins.
Statistical analyses Information on interventions and infant outcomes were entered into the Cochrane Review ManagerI3. Odds ratios with 95% confidence intervals were calculated for each trial and jointly for all trials. The Pet0 method was used for pooling the data14,and the inverse of the variance was used to weight the odds ratios for each study. A x2 test for homogeneity was calculated on each outcome jointly for all trials. While there is debate about whether random or fixed effects models are suitable when there is significant homogeneity, we used the fixed effects models irrespective of homogeneity as suggested by the Cochrane Collab~ration'~ and in keeping with the notes of caution about random effects models provided by others16.Data on duration of labour, maternal morbidity, and satisfaction with care were not consistently measured between the trials and could therefore not be pooled together by meta-analysis. These outcomes are described only in the text.
RESULTS Altogether 13 randomised controlled trials of alternative maternity care with midwives as primary caregivers were identified from 1969 to 1997. Of these, the two earliest studies were excluded because analysis was not by intention to treatl7.I8. Four studies were excluded because the intervention included intrapartum care ~ n l y ' ~ -or* ~antenatal care only22,
and thus did not meet our definition of continuity of midwifery care. This left a total of seven trials which met our three criteria. Table 1 shows the characteristics of the trials, from which the reader can judge the quality of the trials. Table 2 describes the two intervention groups, thus providing an explanation about why each study was included. The total number of women included in the review was 9148. It was not always possible to assess the representativeness of the different samples, but there seemed to be considerable differences between the trials, ranging from predominantly women from low socioeconomic background23 to highly educated, middle-class women5.In several trials the proportion of primiparae was larger than in the general population of childbearing women. All studies except t ~ oincluded ~ . ~only ~ women classified as low risk at booking. The alternative models were distinguished by continuity of midwifery care as defined for the review, but they differed in several features such as whether routine visits by the doctor were included and the setting for intrapartum care. The models also differed with respect to the number of midwives involved and the structure of the care, which could be either teambased or based on a named midwife for the planned episodes of care (Table 2). In each trial standard care was comparable in that there was no continuity of midwifery care as defined for this review. However, there were also important differences between the trials. For example, the caregivers varied from predominantly doctors in Canada25,to a greater mix of doctors and midwives in the UK and A ~ s t r a l i a ~to% ~ ~ , ~ ~ predominantly midwives in Swedens.
Interventions Rates of induction and augmentation of labour were significantly lower in the alternative models of care (OR 0.76, 95% CI 0.66 to 0.86; OR 0.78, 95% CI 0.70 to 0.87, respectively) (Table 3). There was significant heterogeneity between the studies regarding augmentation, but not induction. Electronic fetal monitoring was used less in the alternative groups than in the control groups, (OR 0.19, 95% CI 0.17 to 0.21), but heterogeneity was highly significant. The studies of team midwifery within ordinary hospital settings did not report this outcome, probably because there was no difference in the protocol requirements for CTG and that normal practice therefore was expected by both groups. Pharmacological pain relief was used less frequently in the alternative care models (epidural: OR 0.76, 95% CI 0.68 to 0.85; narcotics: OR 0.69,95% CI 0.63 to 0.77) (Table 4). Significant heterogeneity 0 RCOG 1998 Br J Obstet Gynaecol 105, 1160-1170
9
0
4
-
w
-f
5b
0
%
Inclusion criteria
Risk status at booking
Primiparae,% (Intervention/ Control)
Participants, n c/o consent)
Low risk Over 152 cm, no serious medical problems, no previous uterine surgery, no specified past obstetric history; no Rhesus antibodies in current pregnancy
57/58
No previous caesarean section or difficult vaginal delivery, no complicating medical condition, no previous stillbirth or neonatal death or previous small for gestational age baby, no Rhesus antibodies or raised alpha-feto protein (two occasions)
Low risk
45/46
3510 (92)
Numbered opaque envelopes; 2: 1 in favour of new scheme; consent for new scheme obtained post randomisation
Envelopes; consent for new scheme obtained post randomisation
1001 (91)
MacVicar et a1.=’ (England)
Flint et al.26 (England)
Table 1. Characteristics of included trials.
Living within hospital catchment area, public patient, no drug abuse problems, booking within 16 weeks
Low and high risk
43/43
446 (85.4)
Numbered envelopes; consent preceded randomisation
Kenny et al.24 (Australia)
Cumulative score of 2 points on 91 point score sheet (based on maternal medical history, obstetric history, clinical findings and problems in previous infant)
Low and high risk
48/49
814 (95.3)
Computer generated assignment; consent preceded randomisation
Rowley et al. (Australia)
Low risk Living within hospital catchment area, booking within 16 weeks; absence of medical or obstetric complications
54/55
1299 (81.9)
Telephone; consent preceded randomisation
Tumbull et aLZ3 (Scotland)
N o previous caesarean section,not a primigravida under 17 years or over 37 years, booking within 20 weeks of gestation, at low risk for medical complications according to the Alberta perinatal risk scoring system
Low risk
55/47
218 (consent not reported)
Consecutively numbered sealed opaque envelopes; consent preceded randomisation
Harvey el a1.l’ (Canada)
Willingness to participate in research project, resident of Greater Stockholm, one partner in couple Swedish speaking, no complicating general conditions, no drug abuse problems, non- smokers, last delivery was vaginal (irrespective of previous caesarean sections), had attended for at least one antenatal visit
Low risk
59/56
1860 (consent not reported)
Opaque envelopes chosen by women from box; consent preceded randomisation
Waldenstrom et aL5 (Sweden)
2
+
E
e,
01
0
L
B.
% 8
Q
Bz
0
4
5L
00
8-
No
Restriction of medication and technology
Control group
Hospital antenatal, intrapartum and postnatal care provided by a variety of doctors and midwives
Postnatal home visits Yes provided by midwives participating in the alternative care model
Traditional hospital No
Setting for intrapartum care System for maternal transfer to obstetric team Maternal transfer to obstetric team in midwife managed group (YO)
Number of midwived Team of 4 structure of model of care Routine antenatal At booking, 36 visits by doctor and 41 weeks
Intervention group
Flint et a1.% (England)
Traditional hospital No
3 birthing rooms Yes
Hospital antenatal, intrapartum, and postpartum care provided by a variety of doctors and midwives, doctor at each antenatal visit
Yes
No
Antenatal care shared by GPs and community midwives, delivery within specialist unit by hospital staff
No
Yes (CTG and epidural)
23 antenatally, 18 in 1st stage, 4 in 2nd stage and after
At 26,36, and 41 weeks
Team of 8 (6.8 full time equivalents) At booking, 32 and 40weeks
Kenny et a[.” (Australia)
2 midwifery sisters + 8 midwives
MacVicar et aL2’ (England)
Table 2. Characteristics of the two interventions.
Hospital antenatal, intrapartum and postpartum care provided by a variety of doctors and midiwives, doctor at each antenatal visit
No
No
Yes
No
Shared care between midwives, hospital doctors and GPs
Yes
No
33
3 birthing rooms
20:1 named midwife and up to 3 associates On medical indication only
Tumbull ef (Scotland)
Traditional hospital
At 12-16,36 and 41 weeks
Team of 6
Rowley et aL6 (Australia)
Family physician or obstetrician selected by woman, in-hospital care in all city hospitals
Yes
No
1 antenatally 26 intrapartum
Yes
At booking and 36 weeks 1 birthing room
Team of 7
Harvey et al.25 (Canada)
E
F? tr
On medical indication
Antenatal care provided by community midwives, 1 to 2 routine visits by doctor, hospital intrapartum and postpartum care provided by a variety of midwives in close collaboration with obstetricians
Yes
Yes (CTG, epidural, pethidine, N20, oxytocin)
13 antenatally 19 intrapartum 2 postpartum
Yes
Birth centre
OdY
L
0
+
m cl
z
m
U
r
>
Team of 10 (8.5 full time equivalents)
Waldenstrom et al. (Sweden)
c
E
c c
STANDARD CARE VERSUS MIDWIFERY CARE
1165
Table 3. Labour interventions. EXP = experimental group (ie, continuity of midwifery care model); CTL = control group (ie, standard pattern of maternity care as practised in the place where the trial was conducted).
Induction Study Flint et ~ 1 . ~ MacVicar et ~ 1 . ’ ~ Kenny et al. 24 Rowley et al. Turnbull et al.23 Harvey et u L ’ ~ Waldenstrom et al.
~
TOTAL x2 = 7.00, df = 6, P = 0.3
EXP
CTL
OR
95% CI
511479 21812210 401194 581385 1461611 81101 251912
601473 13111134 41/21 1 681386 1991597 14193 4219 16
0.82 0.83 1.08 0.83 0.63 0.49 0.59
0.55-1.22 0,661.06 0.661.75 0.57-1.22 0.494.8 1 0.20-1.20 0.3W.97
54614892
55513810
0.76
0.6W.86
Augmentation Study
EXP
CTL
OR
95% CI
801451 27012178 301184 1181364 2641593 141100 1401895
1141443 19211099 301198 1041360 237158 1 19185 2231894
0.62 0.66 1.09 1.18 1.16 0.57 0.56
0.45-0.86 0,534.81 0.63-1.89 0,861.62 0.92-1.47 0.27-1.20 0.45471
91614765
91913660
0.78
0.70-0.87
~~
Flint et ~ 1 . ~ MacVicar et aL2’ Kenny et al. 24 Rowley et al. Turnbull et uI.’~ Harvey et al.25 Waldenstrom et nl.
~
TOTAL x2 = 30.20, df = 6, P = 0.00004
Electronic fetal monitoring Study Flint et ~ 1 . ’ ~ MacVicar et al.?7 Kenny et al. 24 Rowley et aL6 Turnbull et al. 23 Harvey et d Z 5 Waldenstrom et 121.~ TOTAL x’ = 119.65,df = 2, P < 0.000001
EXP
CTL
OR
95% CI
107412178
9821 1099
0.18
0.161.21
4401 593
4691 58 1
0.69
0.52490
2751 895
7591894
0.1 1
0.09-0.13
178913666
221012574
0.19
0.17-0.21
was observed between the trials, especially regarding the use of narcotics. The caesarean section rate in the alternative and control groups, did not differ statistically (OR 0.91; 95% CI 0.78 to 1.05). However, the difference in instrumental vaginal deliveries was statistically significant (OR 0.82; 95% CI 0.70 to 0.95) There was no significant heterogeneity between the trials (Table 5). Figure 1 shows a summary of the interventions.
Maternal outcomes Table 6 shows that the use of episiotomies was less frequent in the alternative models (OR 0.69; 95% CI 0.61 0 RCOG 1998 Br J Obstet Gynaecol 105, 116&1170
to 0.77). Heterogeneity between the trials was statistically significant. The lower episiotomy rate in the alternative models of care was associated with a significantly higher rate of perineal tears in the pooled alternative groups (OR 1.15; 95% CI 1.05 to 1.26) (x2= 11.83, df = 6, P = 0.06). The percentage of intact perineums was very s d a r for the two groups (OR 1.11; 95% CI 1.00 to 1.24) k’=5.86, df = 3, P=O.I). Data on duration of labour were available for six trials6-23*24-2”28, but the different measures applied by each of these trials made meta-analysis of conclusions impossible. The first stage of labour was longer in the alternative groups in five of the studies, but the differences reached statistical significance in only
1166 U.
W A L D E N S T R O M & D. T U R N B U L L
Table 4. Analgesia in labour. EXP = experimental group (ie, continuity of midwifery care model); CTL = control group (ie, standard pattern of maternity care as practised in the place where the trial was conducted). ~
Epidural Study
~
Narcotics
CTL
OR
95%CI
EXP
CTL
OR
95% CI
Flint et ~ 1 . ~ ~ 88/451 MacVicar et ub27 32612178 Kenny et al.24 521184 Rowley et a/. 691364 Turnbull et 1941593 Harvey et ~ 1 . ~ ~ 131100 Waldenstrom et al.' 1081895
1431443 20811099 641198 1981581 1981581 22185 1351894
0.51 0.75 0.83 0.94 0.94 0.43 0.77
0.38-0.69 06-0. 91 0'53-1'28 0.74-1.33 0.741.20 0.2 1-0.91 0'59-1.01
1501451 81212178 451184 2531364 2531593 161100 331895
1721443 47711099 401198 1271360 2621581 17/85 1201894
0.79 0.77 1.28 0.33 0.91 0.76 1.69
0.6Ckl.03 0.67491 0.79-2.07 0.240.46 0.72-1.14 0.3fx *62 0.634.77
TOTAL
84313660
0.76
0.68-0.85
136Y4765
121513660
0.69
0.634.77
EXP
85014765
x2 = 12.97, df = 6, P = 0.04
x2 = 59.91, df = 6, P < 000001
Table 5. Operative delivery. EXP = experimental group (ie, continuity of midwifery care model); CTL = control group (ie, standard pattern of maternity care as practised in the place where the trial was conducted). Caesarean section
Instrumental vaginal delivery
EXP
CTL
OR
95%CI
EXP
CTL
OR
95% CI
Flint et ul. 26 MacVicar et Kenny et al.24 Rowley et a1.6 Turnbull et Harvey et al.25 Waldenstrom et d 5
371479 14412210 241194 521385 7916 12 41101 6519 12
351473 7811 134 27121 1 591386 711597 14193 821916
1.05 0.94 0.96 0.87
561479 18712210 12/194 291385 831612 61101 361912
661473 11411134 29/21 1 371386 861597 7193 411916
0.82 0.82
0.27 0.78
0.65-1.69 0.71-1.26 0.54-1.73 0.58-1.29 0'78-1.55 0.10-0.71 0.561.09
0.77 0.93 0.78 0.88
0.561.19 0.64-1.06 0.23-0.83 0.461.27 0.67-1.29 0.25-2.39 0.561.38
TOTAL
40514893
36613810
0.9 1
0.78-1 '05
40914893
38013810
0.82
0.70-0.95
Study
1.10
x2 = 8.49, df = 6, P = 0.2
x2 = 4.41, df = 6, P = 0.6
Table 6. Episiotomy. EXP = experimental group (ie, continuity of midwifery care model); CTL = control group (ie, standard pattern of maternity care as practised in the place where the trial was conducted). Episiotomy Study Flint et al.z6 MacVicar et ~ 1 Kenny et u/.'~ Rowley et ul. Turnbull et a/.23 Harvey et al.25 Waldenstrom et TOTAL
EXP
CTL
OR
95% CI
1521443 201170 461333 1471525 15/97 661847
1851438 32611056 551184 561327 1731509 26179 691834
0-72 0.66 0.34 0.78 0.76 0.38 0.92
0.55-0.94 0.56-0.78 0.20-0.56 0.51-1.18 0.58-0.98 0.19-0.76 0.661.33
92114481
89013427
0.69
06-0. 77
.47512066 ~ ~
x2 = 14.26, df = 6, P = 0.03
0.44
~ w o ~One ~ ,study ~ ~ reported . a nonsignificant shorter first stage in the alternative group23.There was no obvious pattern regarding the duration of the second stage of labour. In one study it was significantly longer in the intervention group28.
There were no maternal deaths. Other maternal outcomes were not reported by all studies, but generally there was no major difference in maternal morbidity. When reported, no statistically significant differences were observed in postpartum haemorrhage5,23-25*27, manual removal of the antenatal admissions to hospita16,23,24~27~28, postnatal complications or readmissions to h o ~ p i t a l ~ ~ The -~~,'~. length of postpartum stay was usually the same in the alternative and control groups, except in two trials where women went home significantly earlier, by approximately one day, in the alternative group^^,^^. Satisfaction with care was reported by all studies except Despite the various measures and the different approaches to randomisation which aimed to take into account women's preferences (Table l), the findings were consistent. Women in the alternative groups were more satisfied with care during all phases of pregnancy (antenatal, intrapartum, postpartum), especially the psychological aspects, and the differences were statistically significant for each study separately. No findings were in favour of the control groups. Women in the alternative groups with 0 RCOG 1998 Br J Obstet GynaecollO5, 1160-1 170
STANDARD C A R E VERSUS M I D W I F E R Y CARE
Y
Induction Augmentation of labour Electronic 1-1
1167
monitoring
Epidural analgesia
Narcotic. Caesarean aectlon Instrumental vaginal dellvery Epiaiotomy
1.o
0.1
10.0
Odds ratio Eelter
wome
Fig. 1. Obstetric interventionsin continuity of standardcare, compared with midwifery care.
continuity of midwifery care were more pleased with information-giving and communication with the caregiver^^,^^,^^.^^. They were more involved in decision-making and felt more in They appreciated more the relationship with the caregivers than women in the control groups, and found them more ~ a r i n g ~ skilled24 ~ ’ ~ ~ and ’ ~ ~interested , in them as a person by providing individualised care6,23,29. Shorter waiting times in antenatal clinics was appreciated more in the alternative
groups concluded that there was no evidence of substandard care or management. In two trial^^,^^ the independent assessors could not exclude possible avoidable factors in two cases in each trial’s alternative group. One case in each study had to do with inadequate advice to a woman with reduced fetal movements during late pregnancy. In both cases referral to hospital for CTG recording would have been preferred. In the other two cases, also one in each trial, CTG at the onset of labour might have altered the course of events.
Infant outcomes The proportion of babies with 5 minute Apgar scores < 7 was not statistically significant between the two groups (OR 1* 13; 95% CI 0.69 to 1-84),as was admission to intensive care or special care baby unit (OR 0.86; 95% CI 0.71 to 1.04) (Table 7). No significant heterogeneity was found between the trials. The difference in perinatal mortality was bordering on statistical significance (OR 1-60;95% CI 0.99 to 2.59) (Table 8). Relative measures for the trial with no events25cannot be estimated at all, and it was dropped from the analysis by the statistical package. Dropping this trial, plus the other one with events missing from one arm24made little difference to the estimates (OR 1.52; 95% CI 0.93 to 2.49; x2 = 6.70, df = 4, P = 0.1). The difference in the stillbirth rate (OR 1.72; 95Y0 CI 0.90 to 3-32) was more pronounced than the neonatal death rate (OR 1-27;95% CI 0.49 to 3.34). Data from some of the trials did not allow differentiation of the stillbirths into deaths before the onset of labour and intrapartum deaths. The authors of three6,24,26 of the five trials that reported a larger number of deaths in the alternative 0 RCOG 1998 Br J Obstet Gynarcol 105, I 16k1170
DISCUSSION This systematic review aimed to assess the ef€icacy of alternative models of maternity care characterised by continuity of midwifery care. Such models have been at the centre of policy debate in the UK in recent years” and have been implemented across a range of developed countries such as Sweden, Australia, New Zealand and Canada. The review included a total of 9148 women in seven trials published over an eightyear period across five industrialised countries. While no quality assessment scale was used, the trials were generally of satisfactory quality with adequate concealment of allocation (with one possible exception, where the only details provided were ‘envelopes’26),and low rates of loss to follow up (generally in the magnitude of about 40/0, but as high as 11% for the control group for one The consent rates were generally high, between 82% and 95%; however, this should not be viewed as reflecting the likely uptake rates for these alternative models in a nontrial or steady state situation. Rather, these will be dependent on factors such as the number of
1168
u.
WALDENSTROM & D. T U R N B U L L
Table 7. Apgar scores and admission to intensive care or special care baby unit. EXP = experimental group (ie, continuity of midwifery care model); CTL = control group (ie standard pattern of maternity care as practised in the place where the trial was conducted). Admission to intensive care or special care baby unit
Apgar score < 7 at 5 minutes Study
EXP
Flint et al.26 MacVicar et al.21 Kenny et al.24 Rowley et nL6 Turnbull et d2’ Harvey er aL2* Waldenstrom et al. TOTAL
x2 = 5.28, df = 4, P = 0.3
CTL
OR
95YoCI
71197 61390 716 13 41101 111917
11214 71394 91603 4193 101920
5.02 0.86 0.76 0.92 1 10
1.24-20.33 0.29-2.59 0.28-2.05 0.22-3.77 0.47-2.6 1
3512218
3112224
1.13
0.69-1.84
Perinatal mortality
EXP
81478 Flint et a1.26 MacVicar et 18/2206 Kenny et al.24 21197 Rowley et al. 61390 41613 Turnbull et ~ 1 . ’ ~ 01101 Harvey et al. 25 Waldenstrom et d 5 81917
CTL
OR
95% CI
1.93 0.624.04 41471 5/1133 1.73 0 . 7 3 4 11 01214 8.10 0.50-130.19 31394 1.98 0.53-7.37 045 0.15-1.35 91603 0193 Undefined 2/920 3.35 0.97-1 1.61
TOTAL 46/4902 2313828 = 8.04, df = 5, P = 0.1
x’
CTL
OR
95Yn CI
231475 3 112206 151197 171390 5616 13
1.09 0.79 0.47 0.85 0.94 0.38
761911
211470 2011133 331214 201394 581603 18193 711920
1.08
0’59-1.99 04-1.41 0.26-0.86 0.441.65 0.64-1.39 0.16486 0‘77-1‘51
22614899
24113827
0.86
0.71-1.04
81101
X’ = 10.41, df = 6, P = 0.1
Table 8. Perinatal mortality. EXP = experimental group (ie, continuity of midwifery care model); CTL = control group (ie, standard pattern of maternity care as practised in the place where the trial was conducted).
Study
EXP
1.60
0.99-2.59
different options provided to women. For example, while the consent rate for one of the Scottish trials when continuity of midwifery care was was 82?40*~, routinely offered at the study hospital alongside four other options, 36% of women chose the continuity of midwifery model as examined in this review30. The review highlights the fact that evaluating individual models of maternity care using techniques of systematic review is far from straightforward, and multiple comparisons are especially problematic”. A particular problem relates to sources of variation in complex packages of care12.Unlike discrete interventions such as new drug regimens, the interventions examined in this review do not represent neatly comparable entities. This is reflected in the fact that we were not able to make all decisions about the review, for example inclusiodexclusion criteria on a strictly apriori basis. The standard care in the control groups was dissimilar in many respects, reflecting variation
in policies and the organisation of the maternity services between the countries in which the studies took place. The alternative models had important features in common, such as continuity of midwifery care, but varied in other ways such as content and philosophy of care and selection of participants. An additional potentially important source of variation lies with provider skillsI2. At an international level there are large differences in midwifery training in length, content and required number of deliveries. For example, the training in Sweden is approximately double that for Australia, with more than double the required number of deliveries. Another issue requiring consideration is that of compliance with the allocated trial group. At the broadest level this might be considered in terms of the proportion of women who did not require to be transferred to the obstetric team for complications, and thus received the complete programme of care. This was reported in four of the trials and ranged from 550/02’ to 73Y0’~.At a more specific level, compliance might be assessed in relation to outcomes like the proportion of women who saw the same midwife throughout their antenatal care, or were attended during labour by someone they knew. Again, this was not measured consistently because the models of care were somewhat different in their focus, with some aiming for continuity within periods of care, such as the study by Turnbull et al.23, and others being concerned with continuity across phases of care, such as Kenny et al.24.Furthermore, while outcome variables were measured fairly consistently, there were some important exceptions. For example, electronic fetal monitoring could be anything from a ‘door test’ at the onset of labour to continuous monitoring, and neonatal transfers could include those to special care and intensive care. In view of these limitations a cautious interpretation 0 RCOG 1998 Br J Obsret Gynaecol 105, 1160-1 170
STANDARD CARE VERSUS MIDWIFERY C A R E
is warranted. It appears that the benefits to women in the alternative models are a reduction in labour interventions and increased levels of satisfaction. Significant heterogeneity was observed in some of the obstetric interventions. It was most pronounced for less invasive procedures, such as augmentation of labour, electronic fetal monitoring, and pharmacological pain relief, while no statistical heterogeneity was observed in the rates of induction and operative deliveries. This seems logical considering that indications for the use of less invasive methods may be wider and less specific, and consequently the variation between services greater. Decisions about induction of labour and operative delivery on the other hand are usually based on more specific criteria, and taken by doctors only. Despite the observed heterogeneity, most of the findings were consistent in the sense that differences pointed in the same direction, such as a general reduction in the use of obstetric interventions in the alternative models of care, and all measures of satisfaction favouring the alternative groups In order for the alternative models to be regarded as worthwhile, these benefits need to be achieved without a concomitant reduction in safety, as measured via clinical outcomes. The number of women included in the review was too small to allow conclusions about maternal safety as measured by maternal mortalitf, and for many of the maternal outcomes we had to rely on unpooled surrogate endpoints. These outcomes (eg, postpartum haemorrhage, manual removal of placenta, antenatal admissions and postnatal readmissions) were consistent and showed no significant differences between the trial groups. Two other outcomes which warrant discussion are perineal status and duration of labour. The review does not suggest a clear benefit with respect to perineal status. The lower episiotomy rate in the alternative models of care was associated with a significantly higher rate of perineal tears and a similar percentage of intact perineums. It has been demonstrated that social support may reduce the duration of labour3’, and it would be reasonable to assume that continuity of midwifery care could have a similar effect by reducing anxiety associated with unknown caregivers. This assumption, however, was not supported by this review. The duration of labour was longer in the alternative groups than in the control groups in most of the trials. The other important component in considering safety is that of perinatal outcomes. The proportion of infants with an Apgar score < 7 at 5 minutes after the birth was approximately the same, as was admission to intensive care or the special care baby unit (with no statistical heterogeneity). The difference in 0 RCOG 1998 Br J Obstet Gynaecol 105, 1160-1 170
I169
perinatal mortality needs further investigation. The result, bordering on statistical significance, may be due to lack of power. Perinatal mortality was defined from different gestational weeks, such as week 2826327 week 2423,week P, and week 206.24.The perinatal deaths occurred during pregnancy, labour or within the first postpartum week in all studies but t ~ 0 ~in~ 7 ~ ~ , which follow up was 28 days after the birth. It was not possible for us to apply a standard definition, and these inconsistencies may limit useful interpretation. There are several issues which should be addressed in future research. Not only are consistent definitions crucial, but it is important to be able to make the distinction between stillbirths before the onset of labour and intrapartum deaths. This would allow a critical appraisal of antenatal versus intrapartum care. In addition, the presentation of case histories (as has been done in some trials5) as well as information about the cause of death may be useful for midwifery and obstetric clinicians. The aim of this review was to investigate the effects of ‘continuity of midwifery care’, not the effects of professional carer (midwifery led care versus medicalkhared care) or continuity of care per se. This distinction as well as the possibility that other factors associated with continuity of midwifery care may have influenced the outcomes of the services must be considered when interpreting the findings of this review. It is not justified to suggest that midwifery care is associated with an increase in perinatal mortality, as some have suggested32. While this paper should not be interpreted as providing a link between increased perinatal mortality and new models of midwifery care, the converse is also true-the paper does not suggest that there is definitely no association. As has been suggested to us, perhaps one of the ways forward may be to increase the number of trials in the systematic review by stratifying according to whether the care was antenatal and/or intrapartum and/or postpartum. This would have enabled all 13 trials to contribute to the analysis, thus increasing the power, rather than limiting it to seven trials. This approach however, would address a slightly different question from that to do with the efficacy of models of care encompassing continuity of midwifery care across the childbirth year. This review did not aim to examine the issue of costs, although this is one of the criteria listed as necessary for informed decision making about maternity service provision12.Three of the seven trials included C O S ~ S but ~ , ~only ~,~ one3O ~ , of these, considered the issue in relation to not only the health service but also the women. The two Australian trials found the continuity of midwifery models to be associated with
1170 U . W A L D E N S T R O M
&
D. T U R N B U L L
reduced costs. The third trial, from Scotland, found similarity in costs for the antenatal and intrapartum period, but increased costs for the postnatal period. Personal costs to women were reduced slightly. It was concluded that from an economics perspective, the alternative model of care was more efficient for the antenatal and intrapartum periods30. Before drawing definite conclusions more work is needed in synthesising results across trials, if possible in a way which takes into account costs in the light of benefits. To conclude, this review illustrates that challenge for modern maternity care is to optimise safety as well as maternal satisfaction. It also has implications for the individual obstetrician and midwife in providing women with informed choice about models of care, especially given that individualsjudge risks and benefits differently. The task is to convey these findings to women in order for them to weigh up the benefits and costs, while being clear about the limitations of current knowledge. References 1 Melia RJ, Morgan M, Wolfe CDA, Swan AV Consumers’views of the maternity services: implications for change and quality assurance. JPublic Health Med 1991; 13: 120-126. 2 Reid M, McIlwaine G. Consumer opinion of a hospital antenatal clinic. Soc Scr Med 1984; 14 A: 363-368. 3 Brown S, Lumley J, Small R, Astbury J. Missing voices. The Experience of Motherhood Oxford University Press,Melbourne, 1994. 4 Lilford RJ. Clinical experimentation in obstetrics. BMJ 1987; 297: 1298-1300. 5 Waldenstrom U, Nilsson CA, Winbladh B. The Stockholm birth centre trial-maternal and infant outcome. Br J Obstet Gynaecol 1997; 104:410-418. 6 Rowley MJ, Hensley MJ, Brinsmead MW, Wlodarczyk JH. Continuity of care by a midwife team versus routine care during pregnancy and birth a randomised trial. Med J Aust 1995; 163: 289-293. 7 Campbell J, Hudson H, Lumley J, Morris N, Rao J, Spensley J. Birth centre confinement at the Queen Victoria Medical Centre. I. Obstetric and neonatal outcome. Med JAust 1981; 2: 347-350. 8 Hodnett ED. Alternative versus conventional delivery settings. In: Enkin MW, Keirse MJNC, Renfrew MJ, Neilson JP, editors. Pregnancy and Childbirth Module of The Cochrane Database of Systematic Reviews, (updated 06 September 1996). Available in The Cochrane Library (database on disk and CDROM). The Cochrane Collaboration; Issue 3. Oxford: Update Software; 1996. Updated quarterly. London: BMJPublishing Group. 9 Renfrew MJ. Midwife versus medicahhared care. In: Enkin MW, Keirse MJNC, Renfrew MJ, Neilson JP, editors. Pregnancy and Childbirth Module. Cochrane database of systematic reviews: review no 03295, Aug 12, 1992; disk issue 1. Cochrane updates on disk. Oxford: Update Software, 1995. 10 Hodnett ED. Continuity of caregivers during pregnancy and childbirth. In: Enkin MW, Keirse MJNC, Renfrew MJ, Neilson JP, editors. Pregnancy and Childbirth Module of The Cochrane Database of Systematic Reviews, (updated 06 September 1996). Available in The Cochrane Library (database on disk and CDROM). The Cochrane Collaboration; Issue 3. Oxford: Update Software; 1996. Updated quarterly. Available from: BMJ Publishing Group, London.
11 Department of Health. Changing Childbirth. Part 1: Report of the Expert Muternity Group. London: HMSO, 1993. 12 Graham W. The Chief Scientists Reports. Devolving Maternity Services-Recommendations for Research and Development. Health Bulletin 1997; 55: 265-275. 13 RevMan (Review Manager). Version 3.0.1 for Windows [Software]; Oxford: Cochrane Collaboration, 1996. 14 Petitti DB. Meta-analysis, decision analysis, and cost-effectiveness analysis. Monographs in Epidemiology and Biostatistics, Vol. 24. Oxford: Oxford University Press 1994. 15 Mulrow CD, Oxman A ( editors). Cochrane Collaboration Handbook (updated 21 October 1996). Available in the Cochrane Library (database on disk and CDROM). The Cochrane Collaboration; Issue 3. Oxford: Update Software, 1996. Updated quarterly. London: BMJ Publishing Group. 16 Rothman KJ, Greenland S. Modern Epidemiology. Second edition. Philadelphia: Lippincott- Raven Publishers, 1998. 17 Runnerstrom L. The effectiveness of nurse-midwifery in a supervised hospital environment. Bul Cull Nurse-Midwives 1969; 14: 4G-52. 18 Slome C, Wetherbee H, Daly M, Christensen K, Meglen M, Thiede H. Effectiveness of certified nurse-midwives. A prospective evaluation study. Am J Ohstet Gynecoll976; 124: 177-182. 19 Chapman MG, Jones M, Spring JE, De Swiet M, Chamberlain GVP. The use of a birthing room: a randomised controlled trial comparing delivery with that of the labour ward. Br J Obstet Gynaecoll986; 93: 182-187. 20 Chambliss LR, Daly C, Medearis AL, Ames M, Kayne M, Paul R. The role of selection bias comparing caesarean birth rates between physician and midwifery management. Obstet Gynecol 1992; 80: 161-165. 21 Hundley VA, Cruickshank FM, Lang GD et al. Midwife managed delivery unit: a randomised controlled comparison with consultant led care. BMJ 1994; 309:1400-1404. 22 Giles W, Collins J, Ong F, MacDonald R. Antenatal care of low risk obstetric patients by midwives. A randomised controlled trial. MedJAust 1992; 157 158-161. 23 Turnbull D, Holmes A, Shields N et al. Randomised, controlled trial of efficacy of midwife-managed care. Lancet 1996; 348: 2 13-2 18. 24 Kenny P, Brodie P, Eckerman S, Hall J. Westmead Hospital team midwifery project evaluation. Westmead, New South Wales, Australia: Centre for Health Economics Research and Evaluation, Westmead Hospital, 1994. 25 Harvey S, Jarrell J, Brant R, Stainton C, Rach D. A randomized, controlled trial of nurse-midwifery care. Birth 1996; 23: 128-135. 26 Flint C, Poulengeris P. Grant A. The ‘Know Your Midwife’ scheme-a randomised trial of continuity of care by a team of midwives. Midwifery 1989; 5 11-16. 27 MacVicar J, Dobbie G, Owen-Johnstone L, Jagger C, Hopkins M, Kennedy J. Simulated home delivery in hospital: a randomised controlled trial. Br JObstet Gynaecull993; 100: 316-323. 28 Waldenstrom U, Nilsson CA. A randomised controlled study of birth centre care versus standard care: effects of women’s health. Birth 1997; 24: 17-26. 29 Waldenstrom U, Nilsson CA. Women’s satisfaction with birth center care: a randomized, controlled study. Birth 1993; 20: 3-13. 30 MDU Research and Midwifery Teams. The establishment of a Midwifery Development Unit based at Glasgow Royal Maternity Hospital. Glasgow: Glasgow Royal Maternity Hospital, 1995. 31 Hodnett ED. Support from caregivers during childbirth. The Cochrane Library, Issue 2. Oxford: Update Software: 1998. 32 Ind T. The Stockholm Birth Centre Trial: maternal and infant outcome. Letter to the Editor. Br J Obstef Gynaecoll997; 104: 1099.
Received I3 October 1997 Returnedjor revision 31 March I998 Accepted 7 July 1998
0 RCOG 1998 Br J Obstet Gynaecol 105, 1160-1 170
