Hodnett ED, Lowe NK, Hannah ME, Willan AR, Stevens B, Weston JA, Ohlsson A, Gafni A, Muir HA, Myhr TL, Stremler R, for the Nursing Supportive Care in Labor Trial Group . Effectiveness of Nurses as Providers of Birth Labor Support in North American HospitalsA Randomized Controlled Trial. JAMA. 2002;288(11):1373-1381. doi:10.1001/jama.288.11.1373
Author Affiliations: Faculty of Nursing (Drs Hodnett and Stevens, and Mss Weston and Stremler), Department of Obstetrics and Gynaecology (Dr Hannah), Department of Paediatrics (Dr Ohlsson), and Centre for Research in Women's Health (Maternal-Child Nursing Research Unit [Dr Hodnett and Mss Weston and Stremler], Research Facilitation Office [Ms Myhr], and Maternal, Infant, and Reproductive Health Research Unit [Dr Hannah]), University of Toronto, Toronto, Ontario; Hospital for Sick Children (Dr Stevens), Toronto, Ontario; Department of Clinical Epidemiology and Biostatistics (Drs Willan and Gafni), McMaster University, Hamilton, Ontario; College of Nursing (Dr Lowe), Ohio State University, Columbus; and Department of Anesthesiology, Duke University, Durham, NC (Dr Muir). Dr Lowe is now with the School of Nursing, Oregon Health and Sciences University, Portland. Dr Willan is now with the Program in Population Health Sciences, Hospital for Sick Children, Toronto, Ontario.
Context North American cesarean delivery rates have risen dramatically since
the 1960s, without concomitant improvements in perinatal or maternal health.
A Cochrane Review concluded that continuous caregiver support during labor
has many benefits, including reduced likelihood of cesarean delivery.
Objective To evaluate the effectiveness of nurses as providers of labor support
in North American hospitals.
Design Randomized controlled trial with prognostic stratification by center
and parity. Women were enrolled during a 2-year period (May 1999 to May 2001)
and followed up until 6 to 8 postpartum weeks.
Setting Thirteen US and Canadian hospitals with annual cesarean delivery rates
of at least 15%.
Participants A total of 6915 women who had a live singleton fetus or twins, were
34 weeks' gestation or more, and were in established labor at randomization.
Intervention Patients were randomly assigned to receive usual care (n = 3461) or
continuous labor support by a specially trained nurse (n = 3454) during labor.
Main Outcome Measures The primary outcome measure was cesarean delivery rate. Other outcomes
included intrapartum events and indicators of maternal and neonatal morbidity,
both immediately after birth and in the first 6 to 8 postpartum weeks.
Results Data were received for all 6915 women and their infants (n = 6949).
The rates of cesarean delivery were almost identical in the 2 groups (12.5%
in the continuous labor support group and 12.6% in the usual care group; P = .44). There were no significant differences in other
maternal or neonatal events during labor, delivery, or the hospital stay.
There were no significant differences in women's perceived control during
childbirth or in depression, measured at 6 to 8 postpartum weeks. All comparisons
of women's likes and dislikes, and their future preference for amount of nursing
support, favored the continuous labor support group.
Conclusions In hospitals characterized by high rates of routine intrapartum interventions,
continuous labor support by nurses does not affect the likelihood of cesarean
delivery or other medical or psychosocial outcomes of labor and birth.
As a consequence of widespread recognition of the health, economic,
and social costs of the rising cesarean delivery rates in the United States,
both Healthy People 2000 and Healthy
People 2010 included targets of lower cesarean rates.1,2
Considerable research attention has been given to the influence of physician
behavior, institutional variations, patient characteristics, and types of
insurance coverage.3- 5
The potential impact of the quality of intrapartum nursing care on the cesarean
delivery rate has received little attention, although a large, retrospective,
observational study concluded that the nurse who provided intrapartum care
was an independent predictor of the risk of cesarean delivery.6
Labor support is a term used to describe the
presence of an empathic person who offers advice, information, comfort measures,
and other forms of tangible assistance to help a woman cope with the stress
of labor and birth.7 A Cochrane Review, first
published in 1995, currently includes 14 trials involving over 5000 women,
conducted in a wide range of settings in 10 countries.7
Support was provided by a variety of women: midwives, student midwives, nurses,
doulas, lay women, or female relatives. Continuous labor support was associated
with significant reductions in the likelihood of cesarean delivery (relative
risk [RR], 0.80; 95% confidence interval [CI], 0.68-0.93), operative vaginal
delivery (RR, 0.81; 95% CI, 0.72-0.92), use of intrapartum analgesia or anesthesia
(RR, 0.87; 95% CI, 0.83-0.92), and a 5-minute Apgar score of less than 7 (RR,
0.50; 95% CI, 0.29-0.89).
The results of the Cochrane Review have been widely disseminated in
practice guidelines in the United States, Canada, and the United Kingdom,
which recommend continuous caregiver support for all women during labor.8- 10 However, 3 issues were
barriers to implementation of the guidelines in North American hospitals.
The first issue was generalizability. The largest effect sizes were in the
trials of labor support in settings where family members were not permitted,
support was provided by a woman who was not part of the hospital staff, and
routine intrapartum medical intervention rates were low.7
In the United States and Canada, 1 or more family members are usually present
during labor, and interventions such as continuous electronic fetal monitoring,
oxytocin infusion, and epidural analgesia are often routine.
The second issue was the cost of staffing labor wards for 1:1 patient-nurse
ratios. In all of the trials in the Cochrane Review,7
the continuous support was provided by caregivers who were not part of the
existing staff complement. Adding additional staff, to implement a policy
of continuous support for all laboring women, would have considerable cost
implications under current conditions, where staffing is usually based on
an average patient-nurse ratio, despite the fact that the patient census is
highly variable and largely unpredictable.9
The third issue concerns the providers of labor support. Many nurses
have not received formal training in labor support techniques,11
and providing labor support was but one part of the labor nurse's role. Work
sampling studies have found that the average labor nurse spent 6% to 24% of
work time in supportive care activities.11- 13
A pragmatic trial was needed, to determine the effectiveness of support
by staff nurses with training in labor support, within a system of staffing
that was sufficiently flexible to allow continuous support without adding
The Nursing Supportive Care in Labor trial was a multicenter, randomized
controlled trial with prognostic stratification for parity and hospital. Prior
to conducting the study, nurses at each participating hospital who volunteered
were trained in labor support to ensure 24-hour availability throughout the
trial. The 2-day training program was conducted by an expert labor nurse and
doula trainer. Prior to and throughout the trial, the nurses had opportunities
to practice their skills, meet regularly with their colleagues for case reviews,
and communicate with the labor support trainer and with each other on an electronic
listserv. The principal investigator (E.D.H.) visited each center, for formal
and informal meetings with staff, at which she stressed the importance of
maintaining clear distinctions between the nursing care provided to the 2
study groups. Periodic videoconferences and teleconferences promoted compliance
and maintained morale. A consultant visited each hospital and recommended
changes that would allow more efficient staff deployment, thereby allowing
women to have continuous nursing support during labor, without jeopardizing
the safety of other patients or requiring additional nursing staff.
The study was approved by the research ethics committees at the University
of Toronto and all participating hospitals. Women gave informed consent before
being enrolled and were informed of the uncertainty about the benefits of
continuous nursing support compared with usual nursing care.
Participating hospitals had to have overall cesarean delivery rates
of at least 15%, a 24-hour epidural analgesia service, and a willingness to
introduce flexibility to their staffing systems in delivery suites. Women
were eligible for the study if they had a live singleton fetus or twins, no
contraindications to labor, were competent to give informed consent, and in
established labor but second stage was not imminent. Women were not eligible
if the gestational age was less than 34 weeks at labor onset or if they were
planning a cesarean delivery, were already enrolled in a labor/delivery management
study and the study protocols were incompatible, were expecting continuous
support from either midwives or doulas/labor coaches, or were at such high-risk
that a 1:1 patient-nurse ratio was deemed medically necessary.
Women who consented to participate were randomly assigned to either
continuous support by a trained labor support nurse (experimental group) or
usual care by a nurse who had not received the labor support training (control
group). Randomization was centrally controlled with the use of a computerized
randomization program at the data coordinating center, accessible by means
of a touch-tone telephone. For women in the experimental group, the nurse
was expected to provide continuous support to the woman for a minimum of 80%
of the time from randomization to delivery. The 80% minimum was chosen to
reflect the realities of everyday nursing practice, in which nurses need time
for meal and rest breaks, and unit emergencies may dictate temporary changes
to nurses' assignments. For women in the control group, the usual nursing
care was provided. Usual nursing care depended on many factors, including
the stage of labor, the condition of the mother and fetus, and the nurses'
workload. The nurses were part of the regular staffing complement on that
shift. When the shift ended, if the woman had not delivered, the appropriate
nurse (eg, with/without training in labor support) from the next shift took
over the woman's care. Only the amount and nature of nursing support varied
between groups; all other nursing and medical care was in accordance with
usual hospital practices and policies.
Study outcomes were derived from those reported in the Cochrane Review.7 The primary outcome was cesarean delivery. Secondary
outcomes were epidural analgesia, length of time from randomization to delivery,
spontaneous vaginal delivery, women's reports of personal control during childbirth,
and newborns with evidence of intrapartum asphyxia. Other outcomes included
indicators of physical and psychological maternal morbidity (in-hospital and
within 6-8 postpartum weeks), other perinatal morbidities up to 6 to 8 weeks,
women's evaluations of their experiences, and costs.
Information about medical outcomes was abstracted from the medical records
by the center research nurses. Because the data collection form included a
question about compliance (eg, whether a nurse with training in labor support
provided continuous support to the participant), research nurses may not have
been fully blinded during retrieval of medical record data. However, the primary
and other medical outcomes were objective data that were recorded in the medical
records as part of routine practice. Audits of selected medical records were
conducted at every site during the 2-year recruitment period to assess and
promote accuracy of data abstraction. Physicians were not explicitly informed
of their patients' group assignments, but they could not be blinded to the
continuous presence of a nurse.
Trial participants completed 2 questionnaires. The Labor Support Questionnaire
was completed before hospital discharge. Its purpose was to determine if the
2 study groups received different amounts and types of nursing support. The
Labor Support Questionnaire consisted of 23 items concerning physical comfort
measures (11 items), emotional support (6 items), information/advice (3 items),
and advocacy (3 items). Participants were asked, "How often did a nurse do
this for you or to you?" and item scores were 1 (never), 2 (occasionally),
and 3 (often). The second questionnaire, completed between 6 to 8 postpartum
weeks, assessed maternal physical and psychological morbidity, neonatal morbidity,
breastfeeding, and views of their birth experiences. Two validated, summated
measures, the Labour Agentry Scale and the Edinburgh Postnatal Depression
Scale, assessed women's sense of control during labor and postpartum depression,
respectively.14,15 An Edinburgh
Postnatal Depression score of more than 12 is strongly associated with postpartum
depression.15 A systematic review concluded
that a sense of personal control is an important psychological outcome of
the birth experience.16 Other questions, developed
and used in a prior multicenter trial,17 asked
women about their future preferences for labor support and their likes and
dislikes regarding participation in the study. Future preferences, to the
extent that they differ from one's experience, are indicators of dissatisfaction.16,18 The questions about likes and dislikes
regarding study participation included opportunities to indicate pleasure
or disappointment in their study group assignment.
Surveys of the participating hospitals during protocol development indicated
that their overall cesarean rates ranged from 18% to 25%, and the cesarean
rates for women who would meet trial eligibility criteria were 11% to 13%.
The sample size was calculated to provide 80% power to detect a reduction
of 2 percentage points in the cesarean rate for those who would meet trial
eligibility criteria. This reduction would represent a 15% relative change
in the proportion of cesarean to vaginal delivery, which is slightly more
conservative than the 20% reduction in RR found in the Cochrane meta-analysis.7 A 1-sided test of hypothesis was chosen for the primary
outcome, because there was neither any empirical evidence nor a theoretical
basis for considering support as potentially harmful, and type I error protection
was unnecessary since the health care implications of any result other than
benefit would be the same (no change in practice would be warranted). A second
1-sided test was used to rule out the possibility of harm. Tests of significance
for all other outcomes were 2-sided. The total required sample size was 6728.
The results were analyzed according to the intention-to-treat; all women
who were randomized were included in the main analyses. The groups were compared
via contingency table χ2 analyses for categorical and binary
variables. Due to potential skewness with continuous variables such as time
durations, Wilcoxon rank sum tests were used. P<.05
for the primary outcome, P<.01 for the secondary
outcomes, and P<.001 for the other outcomes were
considered to indicate statistical significance. Logistic regression explored
interaction effects between baseline variables and cesarean rate, using the
Wald χ2 statistic. All statistical procedures were performed
using SAS version 8.2 (SAS Institute, Cary, NC).
Thirteen hospitals in the United States and Canada participated in the
trial (9 were tertiary care and 4 were community hospitals). The annual birth
census at the hospitals ranged from 2200 to 8500. Two US hospitals were in
the East, 2 in the Midwest, and 1 in the Southwest. The 8 Canadian hospitals
were geographically distributed from coast to coast. We enrolled 6915 women
between May 19, 1999, and May 25, 2001. Data were received for all 6915 women,
of whom 3454 were assigned to the continuous labor support group and 3461
to the usual care group (Figure 1).
For reasons of cost and feasibility, we did not collect data on the approximately
60 000 eligible but nonparticipating women. However, center staff informed
us that women rarely refused to participate. The main reason for nonenrollment
of eligible women was logistical: either the unit staff were too busy or an
intervention nurse was unavailable and therefore a woman could not be randomized.
The mean (SD) age of the sample was 29.5 years (5.6), ranging from 15
to 48 years (Table 1). Although
the majority were white, the sample represented a variety of racial and ethnic
groups, including black, Asian, Hispanic, and Native American/Canadian. More
than 60% had some postsecondary education. Slightly more than 49% (n = 3395)
were giving birth for the first time. Labor onset was spontaneous for 73.5%
(n = 5082). At the time of enrollment, 7.0% were undergoing augmentation of
labor and 6.0% were receiving epidural analgesia. Thirty-four women (0.5%)
were known to have twin pregnancies and 4.9% had had a prior cesarean delivery.
The appropriate form of intrapartum nursing care was provided to 94.9%
(n = 3279) of the experimental group and 98.6% (n = 3413) of the control group.
The median length of time from admission to delivery was 8.1 hours (Table 2). The median length of time from
randomization to delivery was 6.6 hours. In 73.5% of the sample (n = 5080),
cervical dilation had been assessed in the hour prior to randomization; for
97% of these women, randomization occurred in early or early active labor.
More than 96% of the participants completed the Labor Support Questionnaire.
For each of the 23 nursing support activities, women in the experimental group
reported more nursing support than women in the control group (P<.001 for each comparison).
The rate of cesarean delivery was almost identical in the 2 groups:
432 (12.5%) in the continuous labor support group and 437 (12.6%) in the usual
care group (P = .44). Reasons for cesarean delivery
are listed in Table 2. Women in
the continuous labor support group were less likely to have continuous electronic
fetal monitoring, although the absolute difference was small (75.0% vs 79.2%
in the usual care group; P<.001). There were no
other significant differences in maternal events during labor, delivery, or
the in-hospital postpartum period that met our preset criteria for statistical
significance, including length of time from randomization to delivery, length
of time from initiation of epidural analgesia to delivery, spontaneous vaginal
delivery, perineal trauma, health problems during the postpartum hospital
stay, or length of postpartum hospital stay (Table 2). There were no significant differences in immediate neonatal
outcomes, including evidence of intrapartum asphyxia, need for higher level
of nursery care, or length of hospital stay (Table 3). All newborns were born alive; there were 3 neonatal deaths.
The 6 to 8 postpartum week questionnaire response rate was 81% (82.1%
in the continuous labor support group and 79.9% in the usual care group).
The median length of time from delivery to completion of the questionnaire
was 6.8 weeks; 90% were completed within less than 12 weeks.
A statistically significant difference was noted in mean scores on the
measure of experienced control during childbirth, but the actual difference
in mean scores was less than 1 point (mean [SD], 54.1 [9.7] for continuous
labor support vs 53.2 [9.9] for usual care; P<.001).
A comparison of the numbers of women who reported low levels of control, defined
a priori as 2 SDs or more from the mean, yielded nonsignificant differences
(3.4% of the continuous labor support vs 4.3% of the usual care; P = .10). Two hundred forty-five women (8.7%) in the continuous labor
support group had evidence of postpartum depression compared with 277 women
(10.1%) in the usual care group (P = .08).
There was no significant difference in the numbers of women reporting
they were not breastfeeding at 6 weeks' postpartum (1027 women in the continuous
labor support group and 937 women in the usual care group; P = .25). In the continuous labor support group, 738 women (26.0%)
visited their physicians for a health problem in the first 6 to 8 postpartum
weeks, as did 738 (26.7%) in the usual care group. Forty-two women (1.5%)
in the continuous labor support group and 39 (1.4%) in the usual care group
were admitted to the hospital, and 23 women in each group (0.8%) had surgery.
In the continuous labor support group, 120 infants (4.2%) were admitted to
the hospital vs 103 (3.7%) in the usual care group.
Indicators of participants' satisfaction with their birth experiences
favored the continuous labor support group (Table 4). Few women chose "I was not randomized to the group I wanted"
(0.4% of the continuous labor support vs 11.4% of the usual care), while 33.7%
of the continuous labor support group and 11.0% of the usual care group chose
"I was randomized to the group I wanted." Although responses to the question
"If you had it to do over again, would you participate in the Nursing Supportive
Care in Labor Trial?" favored the continuous labor support group, 91.3% of
the usual care group indicated they definitely or probably would participate.
While the majority of women (63.4%) in the continuous labor support group
would like continuous labor support in a future labor, nearly one half (46.6%)
of the group that was randomized to usual care also would prefer continuous
labor support the next time.
For the primary outcome of cesarean delivery, there were no significant
interactions between treatment group and country, center, race/ethnicity,
education, parity, pregnancy complications, cervical dilation, onset of labor
(spontaneous or induced), epidural analgesia, or use of oxytocin (Table 5).
Most of our results are contrary to the current Cochrane Review7; continuous labor support by nurses did not affect
cesarean delivery rates or other medical or psychosocial outcomes. The one
outcome that did differ, participants' evaluations of their experiences, may
be either a true treatment difference or a function of the fact that they
were not blinded to the intervention. However, in the usual care group, less
than 12% reported disappointment with their study group assignment and more
than 91% would definitely or probably participate in the trial again, suggesting
that disappointment did not exert a large effect on responses.
In trials of behavioral interventions, the inability to blind, the challenges
in ensuring that the intervention is applied appropriately, and the possibility
of contamination are serious potential threats to validity. Neither the research
nurses abstracting medical records nor the patients' physicians were fully
blinded. However, since the primary outcomes were objective, it is unlikely
that this affected the results. Also, unblinding tends to bias results away
from the null hypothesis, which was not the case here.
We took several steps to ensure that the intervention and control conditions
were applied appropriately, and we used several measures to assess whether
they were. To ensure a strong intervention, the nurses who were trained were
volunteers who wished to provide continuous labor support, the training was
from an expert, and we used a variety of approaches to reinforce training
throughout the trial. Compliance was high with more than 97% of the sample
receiving the appropriate form of care. The intervention period began at the
appropriate time, when the woman was in early or early-active labor.
It was equally important that the usual care group received usual nursing
care. Prior to beginning the trial, we held formal and informal meetings with
the nursing staff to explain the importance of retaining 2 distinct study
groups throughout the trial. Participants' reports indicated that each of
the nursing supportive activities was provided significantly more often to
women in the experimental group. As a further check to determine if control
nurses began to adopt the behaviors of the intervention nurses, we divided
the sample into those recruited in the first year of the trial and those enrolled
in the second year; the comparisons of participants' reports of supportive
activities yielded the same results.
Another reason for discounting contamination as a factor in the trial
results stems from a prior large multicenter trial of a strategy specifically
aimed at encouraging nurses to provide more labor support.19
Twenty hospitals participated in the trial. The intervention at the 10 hospitals
in the intervention group involved small groups of nurses who were selected
by their peers as opinion leaders and who worked to change their colleagues'
practice during a 1-year period. There were no significant changes in the
amount of time nurses spent in labor support and no changes in patient outcomes.
One of the trial's conclusions was that nursing care was powerfully influenced
by the organizational culture. Therefore, our results are unlikely to be a
function of a weak intervention or contamination.
Another possibility is that the 2 groups received the same amount of
support, because the usual care group received more support from husbands
or partners and family members. We did not measure the amount of support provided
by husbands or partners and other family members. It is possible that husbands
in the experimental group provided less support because a nurse was continuously
present, although the nurses' training program included specific content about
how to help partners to support their wives. A prior small Canadian trial
of labor support (n = 150) found that compared with the control group, women
in the experimental group reported more support from their husbands or partners
in all categories except information/advice.20
Their husbands or partners explained that they were encouraged by the caregiver,
who acted as a role model and showed them how to help their wives. There have
been no published trials to our knowledge that evaluated the effect of partner
support on cesarean delivery. Regardless, it does not change our conclusion
that continuous labor support by nurses had no effect on the cesarean rate
in our trial.
Another possibility for the discrepancy between our results and that
of the other labor support trials is that the results of the Cochrane Review7 are incorrect. The results of a large, well-controlled
trial may be more reliable than those of a meta-analysis, if the meta-analysis
is small or involves poor-quality trials.21
However, the trials in the Cochrane Review7
were methodologically sound, the test for heterogeneity was not significant,
and the sample was large (14 trials with more than 5000 women).
A more plausible explanation for our results is that the benefits of
continuous labor support are overpowered by the effects of birth environments
characterized by high rates of routine medical interventions. In our sample
of primarily low-risk healthy women, 62% of labors were induced or augmented
with oxytocics, 77% had continuous electronic fetal monitoring, and 75% had
regional analgesia. A smaller trial, evaluating continuous support by nurses
in a setting characterized by high intrapartum intervention rates, reported
similar results and conclusions.22
Our results call into question the usefulness of national practice guidelines
in the United States, Canada, and the United Kingdom,8- 10
which recommend that all women receive continuous 1:1 support from specially
trained caregivers during labor. The results of a recent study of Ontario
hospitals with below-average cesarean rates provide further evidence that
the effectiveness of labor support may depend on the setting in which it is
provided.23 The hospitals had annual overall
cesarean rates of 12% to 17% (well below the current provincial rate of 20%),
and in addition to a commitment to 1:1 labor support by nurses, the organizations
shared many characteristics, including regular audits of their practices,
shared decision-making by multidisciplinary teams, active consumer involvement
in policy setting, and a belief system based on labor and birth as healthy
life events.23 These organizational characteristics
are consistent with those of other birth environments with low cesarean rates,
such as in-hospital and freestanding birth centers.24- 26
Those who wish to improve women's evaluations of their care during labor
and birth may want to ensure that all women have continuous intrapartum nursing
support. However, reducing US and Canadian cesarean delivery rates cannot
be accomplished solely through a policy of continuous nursing support. Those
who wish to decrease cesarean rates and rates of other intrapartum interventions
would be advised to work with all of the stakeholders to implement comprehensive
changes to the routine care of women during labor and birth. Large well-controlled
evaluations of such complex changes would be helpful.