Continuous Emotional Support DuringLabor in a US HospitalA Randomized Controlled TrialJohn Kennell, MD; Marshall Klaus, MD; Susan McGrath, PhD; Steven Robertson, PhD; Clark Hinkley, MD

The continuous presence of a supportive companion (doula) during labor anddelivery in two studies in Guatemala shortened labor and reduced the need forcesarean section and other interventions. In a US hospital with modern obstetricpractices, 412 healthy nulliparous women in labor were randomly assigned to asupported group (n = 212) that received the continuous support of a doula or anobserved group (n = 200) that was monitored by an inconspicuous observer.Two hundred four women were assigned to a control group after delivery.Continuous labor support significantly reduced the rate of cesarean sectiondeliveries (supported group, 8%; observed group, 13%; and control group, 18%)and forceps deliveries. Epidural anesthesia for spontaneous vaginal deliveriesvaried across the three groups (supported group, 7.8%; observed group, 22.6%;and control group, 55.3%). Oxytocin use, duration of labor, prolonged infanthospitalization, and maternal fever followed a similar pattern. The beneficialeffects of labor support underscore the need for a review of current obstetricpractices.

(JAMA. 1991;265:2197-2201)

PROMISING results from two studiesin Guatemala demonstrated that thecontinuous presence of a supportive fe¬male companion during labor and deliv¬ery could significantly reduce the needfor cesarean section.IZ Additionally,providing a labor companion resulted infewer obstetric interventions, shorter

See also pp 2202 and 2236.

labors, and fewer perinatal problems inthe fetuses and neonates. The possibili¬ty that similar results could occur inthe technologically advanced obstetricunits in US hospitals is important be¬cause of the potential financial, physi¬cal, and emotional consequences. Inparticular, a low-cost, low-risk inter¬vention that reduces the extremely highcesarean section rates in US hospitals

could have a major impact on obstetricpractices.3'6

Compared with the obstetric prac¬tices in place at the time of the Guatema¬lan studies, current US obstetric prac¬tices provide considerably more controlover the process and progress of labor.Specifically, electronic fetal monitoringis routinely used to evaluate the condi¬tion of the fetus. Epidural anesthesia iswidely available for pain relief. Oxyto-cin and artificial rupture of membranesare used frequently to augment labor.Do these technological interventions di¬lute or override the influence ofcontinu¬ous support during labor? Ifeffects sim¬ilar to those in Guatemala could bedemonstrated in a US hospital, the im¬plications for the quality and cost ofperinatal health care could be substan¬tial. The present randomized controlledstudy examined the medical effects ofsupport during labor in a US teachinghospital that had full availability of con¬

temporary obstetric techniques andequipment.METHODSSetting

The study was conducted at JeffersonDavis Hospital in Houston, Tex, a pub-

lie hospital providing care for a low-in¬come Hispanic, black, and white popula¬tion, with care provided by theDepartment of Obstetrics and Gynecol-ogy, Baylor College of Medicine. All pa¬tients were in the staff service, withcare directed by English-speaking resi¬dents who followed established obstet¬ric protocol. At this obstetric teachingfacility, companions were not routinelypermitted to be with a woman duringlabor and delivery because most pa¬tients labored in a 12-bed ward that hadinsufficient privacy to allow visitors.Due to the volume of patients and thenumber of hospital personnel involved,patients were usually in the presence ofstrangers and did not receive continu¬ous support from the staff. Interpreterswere available if needed to translateconversation between Hispanic pa¬tients and non-Spanish-speaking medi¬cal personnel.Participants

Participants were nulliparous women

ranging in age from 13 to 34 years, withsingle-gestation, term, uncomplicatedpregnancies. These women were askedto participate in the study after theywere admitted to the hospital in activelabor. Because of the difficulty in pre¬cisely measuring cervical dilatation,women were admitted to the study withinitial cervical dilatation of either3 or 4 cm. Women with pregnancy-in¬duced hypertension, breech presenta¬tion, gestational diabetes, a history ofdrug or alcohol abuse, or other high riskconditions were not enrolled. The 412women who met the criteria and con¬sented to be in the study were randomlyassigned to a supported group (n = 212)or an observed group (n = 200). After255 participants were enrolled, a con¬trol group (n = 204) was added to thedesign to examine potential supportiveeffects of the observer on women in theobserved group. Following delivery,patients were asked to be in the control

From the Department of Pediatrics, Case WesternReserve University, Cleveland, Ohio (Drs Kennell andMcGrath); the Department of Pediatrics, Childrens Hos-pital, Oakland Calif (Dr Klaus); the Department of Hu-man Development and Family Studies, Cornell Univer-sity, Ithaca, NY (Dr Robertson); and the Department ofObstetrics and Gynecology, Baylor College of Medi-cine, Houston, Tex (Dr Hinkley).

Reprint requests to Department of Pediatrics, Rain-bow Babies and Childrens Hospital, 2103 Adelbert Rd,Cleveland, OH 44106 (Dr Kennell).

by guest on August 4, 2009 www.jama.comDownloaded from

group if they had been admitted to thehospital on days that doulas were al¬ready assigned to patients and if hospi¬tal chart information indicated that theymet the enrollment criteria.

ProceduresIndications for specific procedures

were standardized prior to the start ofthe study and were consistent with theroutine obstetric protocol at JeffersonDavis Hospital at that time. Patientswere confined to bed as soon as possibleafter admission to allow for electronicfetal monitoring. When admission bloodwork was completed, an intravenous in¬fusion was started for each patient. Ar¬tificial rupture of membranes was doneroutinely after 5 cm of cervical dilata¬tion so that internal monitoring could beobtained if needed. For the manage¬ment ofpain, butorphanol tartrate (Sta-dol) or epidural anesthesia was used atthe patient's request or if, in the judg¬ment of nursing and/or medical staff,the patient was unable to deal with herpain, as evidenced by vocalization, rest¬lessness, or lack of cooperation betweencontractions. Epidural anesthesia wasconsidered if the patient was not near

delivery. Epidural anesthesia was alsoused for operative deliveries. Oxytocinwas administered following failure toprogress in effacement or dilatation af¬ter 1 lk hours when uterine inertia was

thought to be the cause. Oxytocin wasalso given to patients with rupturedmembranes ifcervical dilatation had notreached 5 cm 12 hours after rupture.Patients did not have a labor coach, al¬though brief visits by family memberswere permitted if the labor area was nottoo busy. Due to the crowded conditionsin the postpartum units, mothers withan uncomplicated hospital course wenthome within 24 hours of delivery, andinfants were routinely discharged with¬in 48 hours of delivery.Doula Intervention

The Greek word doula refers to an

experienced woman who guides and as¬sists a new mother in her infant-caretasks. In the current study the term isused to describe the women who pro¬vided continuous labor support to theparticipants in the supported group.Eleven women served as doulas overthe course of the study, but the majority(82%) of the supported deliveries were

accompanied by one of four doulas. Thedoulas ranged in age from 22 to55 years, were fluent in both Spanishand English, had experienced a normallabor and vaginal delivery with a goodoutcome, and were comfortable dealingwith patients and medical staff from di-

verse backgrounds. All doulas wentthrough a 3-week training period, dur¬ing which they became familiar with thecourse of normal and abnormal labor,obstetric procedures, and hospital poli¬cies as well as a wide variety of support¬ive techniques. While doula supportwas intended to be individualized tomeet the needs ofeach patient, the dou¬las met with one another almost dailyand with the principal investigatorweekly throughout the study to ensurea degree ofconsistency in their methodsof supporting patients. Doulas were

paid from research funds on an hourlybasis for the time they were present inthe hospital, for an average cost of $200per patient supported.

Doula intervention was provided forwomen randomly assigned to the sup¬ported group after informed consentwas obtained. The doula met the studyparticipants for the first time after hos¬pital admission. The doula stayed at herassigned patient's bedside from admis¬sion through delivery, soothing andtouching her patient and giving encour¬

agement. In addition, the doula ex¬

plained to her patient what was occur¬

ring during labor and what was likely tohappen next. When necessary, shetranslated medical instructions for thepatient. The doula also kept a writtenrecord of staff contacts, interventions,and procedures. Women in the ob¬served group received routine hospitalcare, and an observer kept a record ofstaff contacts, interactions, and proce¬dures. The observer never spoke to thelaboring woman and attempted to re¬main as inconspicuous as possible in themidst of the large number of labor room

personnel.Data

Demographic data (race, annual fam¬ily income, marital status, mother's andfather's ages and education levels,amount of prenatal care, and atten¬dance at childbirth education classes)were obtained from interviews with theparticipants within 24 hours after deliv¬ery. Prenatal care was categorized as

none, minimal (first visit after 27 weeksor fewer than four visits), moderate(first visit between 13 and 27 weeks,with at least four visits), or optimal(first visit before 13 weeks, with at leastsix visits).

Intrapartum data (cervical dilatationat the time of admission and minutes ofattention from hospital staff for partici¬pants in the supported and observedgroups) were obtained from hospitalcharts or from the research labor rec¬ord. Newborn data (birth weight,length, and head circumference; gesta-tional age at delivery; and sex) were

obtained from the infants' hospitalcharts.

Intrapartum and newborn outcomedata were obtained from hospital chartsor the research labor record. The intra¬partum outcome data were defined asfollows: anesthesia (epidural, spinal,general, pudendal, local, or none), nar¬cotic analgesia (yes/no, number of 2-mgdoses), oxytocin (yes/no), duration of la¬bor (from time of admission at 3 or 4 cmof cervical dilatation to delivery), deliv¬ery type (cesarean section, forceps de¬livery, or spontaneous vaginal deliv¬ery), and maternal fever (temperature3=37.8°C at any time during labor). Thenewborn outcome data were 1- and 5-minute Apgar scores, hospital course

(normal was ^48 hours, abnormal was>48 hours), and the presence or absenceof neonatal problems (jaundice requir¬ing phototherapy, tachypnea, meconi-um aspiration, or sepsis evaluation).RESULTS

All women who agreed to participatein the study agreed to do so. However, atotal of 14 women who agreed to partici¬pate were not included in the study.(Four were transferred to the birth cen¬ter because of staffing limitations. Fivewere sent home because they were notin active labor. Three withdrew fromthe study. One had an undetectedbreech presentation. One had inter¬rupted observations.)

The supported, observed, and controlgroups did not differ in race, maritalstatus, mother's age, father's educa¬tion, annual family income, and prenatalcare (Table 1). Mother's education in thesupported group did not differ signifi¬cantly from that in the observed or con¬trol groups. However, mother's educa¬tion was slightly lower in the observedgroup than in the control group(P<.05). Ninety-three percent of thewomen in the study reported receiv¬ing some prenatal care. The women inthe control group had significantlymore prenatal childbirth education(P<.0006), although only 15% of thesample attended childbirth educationclasses, and 40% of these women hadattended only one class. The mean ges-tational age of the infants was 39.8weeks (SD, 0.96 weeks), with a meanbirth weight of 3268 g (SD, 423.5 g),mean length of 50 cm (SD, 3.9 cm), andmean head circumference of 34 cm (SD,2.7 cm). The three groups did not differin gestational age, neonatal sex, or birthmeasurements. There were no still¬births, neonatal deaths, or detectablemalformations in the sample. All partic¬ipants were enrolled with an initial cer¬vical dilatation between 3 and 4 cm. Theproportion of patients with an initial

by guest on August 4, 2009 www.jama.comDownloaded from

cervical dilatation of 4 cm differedamong the three groups (supportedgroup, 84%; observed group, 83%; andcontrol group, 75%; P = .04). Hospitalstaff interacted with patients for an av¬

erage of21% of the time patients were inlabor (defined as minutes ofcontact withstaff divided by total minutes in labor),and the amount of interaction did notdiffer between the supported and ob¬served groups. The mean proportion oftime a family member was present was

greater in the observed group than inthe supported group (5.6% vs 3.7% oftotal minutes in labor, P<.02). (Be¬cause the control group participantswere enrolled after delivery, minutes ofstaff and family time could not becollected.)Anesthesia/Analgesia

Because of the expected associationbetween delivery type and anesthesiause, forceps and cesarean deliverieswere excluded in this analysis to allow aclearer analysis of the effect of laborsupport on anesthesia use. Among par¬ticipants who had spontaneous vaginaldeliveries, epidural anesthesia wasused in 14 (7.8%) of 179 women in thesupported group, 31 (22.6%) of 137women in the observed group, and 68(55.3%) of 123 women in the controlgroup (x2[2] = 86.9, P<.0001). The useof pudendal anesthesia for spontaneousvaginal deliveries did not differ signifi¬cantly across the three groups (4% forthe supported group and 2% for both theobserved and control groups). In addi¬tion, regardless of the type of delivery,there were no significant differencesacross the three groups in the number ofwomen who received butorphanol tar-trate for pain (21.7% in the supportedgroup, 28.0% in the observed group,and 25.5% in the control group) or in themean number of 2-mg doses given(mean, 1.2 doses). No other narcotic ornonnarcotic analgesia was adminis¬tered.

Oxytocin UseThe use of oxytocin to augment labor

differed across the three groups whenall delivery types were considered (Ta¬ble 2). Among women who went on todeliver vaginally without forceps, oxy¬tocin use also differed across the threegroups. When forceps orcesarean deliv¬eries were considered alone, the differ¬ence in the percentage ofwomen receiv¬ing oxytocin across the three groups didnot achieve statistical significance.Duration of Labor

The mean duration of labor differed(P=.O001) across the three groups forall delivery types combined, with the

Table 1 —Demographic Characteristics

GroupSupported

(n = 212)Observed(n = 200)

Control(n = 204)

Race, No. (%)Hispanic 136 (64) 129 (65)* 116 (57)*Black 53 (25) 50 (25) 56 (27)White 21 (10) 21 (11) 29 (14)Asian 2 (1) 3 (1)

Marital status, No. (%)Single 105 (50) 109 (55)* 115 (56)Married 107 (50) 91 (46) 89 (44)

Income level, No. (%)<$8000 105 (50)* 115 (58)* 108 (53)$8000-$15 000 65 (31) 54 (27) 44 (22)>$15 000 15 (7) (5) 5 (2)Unreported 27 (13) 22 (11) 47 (23)

Prenatal care. No. (%)Optimal 43 (20) 40 (20)* 57 (28)Moderate 86 (41) 85 (43) 83 (41)Minimal 63 (30) 61 (31) 55 (27)None 18 (8) 12 (6) 7 (3)Unknown 2 (1) 2 (1) 2 (1)

Childbirth education, No. (%)tYes 19 (9) 30 (15) 46 (23)No 193 (91) 170 (85) 158 (77)

Mother's age, yMean (SD) 19.9 (3.5) 19.7 (3.6) 20.3 (3.8)Range 14-34 14-34 13-33

Father's age, yMean (SD) 23.6 (5.5) 22.9 (4.7) 23.6 (4.9)Range 15-50 16-40 15-42No data, No. of participants

Mother's education, ytMean (SD) 9.5 (3.1) 9.2 (2.8) 10.2 (2.8)No data, No. of participants

Father's education, yMean (SD) 9.9 (3.4) 10.0 (3.0) 10.5 (2.9)No data, No. of participants 23 23 19

Initial dilatation, No. (%)§3 cm 34 (16) 34 (17) 51 (25)4 cm 178 (84) 166 (83) 153 (75)

•Percentages do not add to 100% because of roundingtx!(2) = 14.8, P=.0006.*F(2, 608) = 3.7, P = .03.§X*(2) = 6.4,P=.04.

Table 2—Oxytocin Use

Type ofDelivery

No. ofParticipants

Oxytocin Use by Group, No. (%)Supported Observed Control

Total* 616 36/212 (17.0) 46/200 (23.0) 89/204 (43.6)Spontaneous vaginalt 25/179 (14.0) 18/137 (13.1) 46/123 (37.4)Forceps 97 6/16 (37.5) 12/37 (32.4) 20/44 (45.5)Cesarean section:): 80 5/17 (29.4) 16/26 (61.5) 23/37 (62.2)

*X2(2) = 40.2, P<.0001. P<.0001 for the supported group vs the control group and for the observed group vs thecontrol group.

txz(2) = 31.0, P<.0001. P<.0001 for the supported group vs the control group and for the observed group vs thecontrol group.

tx*(2) = 5.7,P=.06.

supported group having the shortestduration of labor (Table 3). When onlyvaginal deliveries without forceps wereconsidered, the mean duration of labordiffered across the three groups.Among women who had spontaneousvaginal deliveries without labor aug¬mentation by oxytocin, the mean dura-

tion of labor differed across the threegroups. There were no statistically sig¬nificant differences in mean duration oflabor across the three groups for un-

medicated, nonaugmented, unanesthe-tized, spontaneous vaginal deliveries.The pattern of results in the analyses ofduration of labor did not change when

by guest on August 4, 2009 www.jama.comDownloaded from

Table 3.—Duration of Labor

Duration of Labor by Group, h

Supported Observed Control

_Type of Delivery_Mean (SD) n Mean (SD) n Mean (SD)_nAll deliveries* 7.4 (3.8) 212 8.4 (4.2) 200 9.4 (4.2) 204Vaginal, no forcepst_6.8 (3.5) 179 7.3 (3.1) 137 8.2 (3.9) 123Spontaneous vaginal, no oxytocin): 6.1 (3.0) 154 7.1 (3.1) 119 6.8(3.3)_77Vaginal, no medications 5.5 (2.8) 116 6.3 (2.5) 62 5.3 (2.7) 25

*F(2, 613) = 12.8, P= .0001. P<.02 for the supported group vs the observed group, the supported group vs thecontrol group, and the observed group vs the control group.

tF(2, 436) = 5.5, P= .004. P= .002 for the supported group vs the control group. P= .03 for the observed groupvs the control group.$F(2, 347) = 3.3, P = .04. P = .01 for the supported group vs the observed group.

Table 4.—Neonatal Course

No. (%) by GroupSupported Observed Control

_(n =212)_(n =200)_(n = 204)Abnormal course (remained in the

hospital >48 h for medical reason)*_22 (10.4)_34 (17.0)_49 (24.0)Sepsis evaluations (as the reason

for abnormal course)t_9 (4.2)_19 (9.5)_30 (14.7)Maternal fever (as the indication

for sepsis evaluation)* 3 (1.4) 14 (7.0) 21 (10.3)

*X!(2) = 13.7, P = .001. P = .05 for the supported group vs the observed group. P = .001 for the supported groupvs the control group.

tx!(2) = 13.3, P = .001. P = .05 for the supported group vs the observed group. P = .0005 for the supported groupvs the control group.

tx!(2) = 14.5, P=.0007. P = .009 for the supported group vs the observed group. P = .002 for the supportedgroup vs the control group.

initial cervical dilatation (3 vs 4 cm) wasused as a covariate.

Type of DeliveryThe provision of social support also

had an effect on the type of delivery.Thirty-seven (18%) of the 204 patientsin the control group and 26 (13%) of the200 patients in the observed group re¬

quired a cesarean section for delivery,in contrast to 17 (8%) of the 212 patientsin the supported group (x2[2] = 9.4,P = .009; P=.004 for the supportedgroup vs the control group). The rateof forceps use in vaginal deliveriesalso differed across the three groups(P<.0001), with the supported grouphaving significantly fewer forceps de¬liveries (8.2%) than either the ob¬served (21.3%) or control (26.3%) group(P=.0OO6 for both pairwise com¬

parisons).Neonatal Outcome

The proportion of infants remainingin the hospital for more than 48 hoursbecause of neonatal problems differedacross the three groups, with fewer in¬fants born to mothers in the supportedgroup requiring a prolonged hospitalstay (Table 4). The reasons for the pro¬longed stays were distributed similarlyacross the three groups, with the excep¬tion of the number of infants kept in thehospital for a sepsis evaluation. Both

the observed and control groups re¬quired more sepsis evaluations than thesupported group. Most sepsis evalua¬tions in the observed (14 [74%] of 19) andcontrol (21 [70%]of 30) groups were per¬formed because of maternal fever, andmaternal fever was more common inthese two groups than in the supportedgroup. In the entire sample, maternalfever was associated with longer labor(11.6±4.3 vs 8.2±4.1 hours, P<.001)and more vaginal examinations (7 ± 2.1vs6±2.2, P= .002). Anesthesia use alsodiffered between febrile and afebrile pa¬tients, with significantly more womenin the maternal fever group receivingepidural anesthesia (55.3% vs 32.2%,P=.004).

Timing of InterventionsThe mean elapsed time between ad¬

mission to the study and administrationof epidural anesthesia for pain duringlabor differed across the three groups,with the control group receiving epidu¬ral anesthesia sooner than the sup¬ported group (control group, 6.2 hoursafter admission; supported group, 9.5hours after admission; P<.005). Oxyto¬cin was also administered sooner to thecontrol group, with the mean elapsedtime between admission and oxytocinadministration differing across all threegroups (supported group, 7.5 hours; ob¬served group, 6.0 hours; and control

group, 4.9 hours; P<.0001). Timing ofcesarean section differed only betweenthe supported and control groups (8.9 vs11.8 hours, P=.02). Indications for ce¬sarean section differed across the threegroups, with the control group havingmore cesarean deliveries for "failure toprogress" than either the supported orobserved group (supported group, fiveof 17; observed group, 10 of 26; and con¬trol group, 25 of 37). The three groupsdid not differ in the number of cesareandeliveries performed because of fetaldistress, cephalopelvic disproportion,or other indications.COMMENT

The provision of support during laborin controlled trials has resulted in a con¬sistent pattern of outcomes favorable tomothers and infants both in a modernUS hospital and in Guatemala a decadeago. Unlike the participants in thesestudies, however, most women in UShospitals are now accompanied duringlabor by their spouses or male part¬ners.7 Women rate their partners' pres¬ence during labor and delivery as ex¬tremely important and helpful, but thepartners' actual effect on the course oflabor has not been adequately as¬sessed.' While doulas and male part¬ners fulfill similar roles for laboringwomen, the question remains whetherthe male partner has the same positiveimpact on perinatal outcome.

Recent research by Bertsch et al8 in¬dicates that male partners and doulasdiffered in the support that they pro¬vided to laboring women (while the la¬boring women were experiencing dis¬comfort). For example, when all formsof touching (rubbing, stroking, clutch¬ing, and holding) by doulas and malepartners were considered, on averagethe doulas touched the laboring womenmore than 95% of the time comparedwith less than 20% by the male part¬ners. In addition, male partners choseto be present for less time during laborand to be close to the mother less oftenthan the doulas. Studies have shownthat under some circumstances thepresence of the male partner reducesthe pain medications given to the moth¬er, but no randomized study of the pres¬ence of male partners during labor hasreported any decrease in the duration oflabor, the use of forceps, the rate ofcesarean section, or the use ofoxytocin.7Because of his personal involvementand lack of experience with labor, it isuncertain whether a male partner canhave the same impact on perinatal out¬comes as a doula.

The mechanism by which support in¬fluences labor, delivery, and perinataloutcome is not fully understood. The

by guest on August 4, 2009 www.jama.comDownloaded from

fact that an observer who stayed in thelabor room at some distance from themother during the entire labor had a

significant effect on obstetrical outcomemeasures even though she did not speakwith the mother may provide a clue tothe needs of a woman during labor. Intrying to dissect the process by whichthe doula has an effect on labor, it isimpressive that part of her effect maybe solely her presence, without intimateinteractions such as talking and touch¬ing. Even in a busy, well-staffed laborarea, a mother may feel alone andneedy. Some of the positive influence ofthe doula and the attentive (but nonin-teractive) observer could result fromtheir effect on the hospital staff. Thepresence of a doula or observer clearlyindicated that the patient was a partici¬pant in a research project. As a result,the staff might have followed hospitalprotocol more closely for the use of ob¬stetric interventions such as anesthesiaand oxytocin, which could explain thedecreased use of these agents in thesupported and observed groups and theshorter elapsed time between admis¬sion and the administration of suchagents in the control group.9

The association between acute mater¬nal anxiety and disturbances in theprogress of labor is strongly suggestedby studies of human and animal moth¬ers.10"14 Circulating catecholamines maybe the mechanism by which anxiety in¬fluences the course of labor. For exam¬

ple, in humans, an increased level ofcatecholamines as a result of maternalanxiety has been shown to decreaseuterine contractility.10,11 In animal stud¬ies, uterine and placental blood flow re¬duction resulting in fetal distress hasalso been related to increased catechol-amine levels.1214 Thus, a supportivecompanion may reduce catecholaminelevels by reducing maternal anxiety andfacilitating uterine contractile activityand uterine blood flow. A doula maydecrease maternal anxiety by her inter¬actions with the laboring woman—herconstant presence, physical touch, reas¬

surance, explanations, and anticipatoryguidance. These aspects of doula sup¬port may make the laboring woman feelsafer and calmer, needing less obstetricintervention for labor to proceedsmoothly. While some obstetric inter¬ventions are necessary and have a posi¬tive impact, complications may arise.For example, obstetric pain relief dur¬ing labor in the form of epidural anes¬thesia has recently been linked with ma¬ternal fever.15 In the present study, thediffering rates of epidural anesthesiause across the three groups may explainthe group differences in maternal feverand associated sepsis evaluations. In

addition, differences in epidural anes¬thesia use across the three groups mayhave played a role in the proportion ofcesarean deliveries, as epidural anes¬thesia has been related to increased ce¬sarean section rates.16

The long-term effects of support dur¬ing labor on maternal-infant attach¬ment, maternal self-esteem, and post-partum depression are areas requiringfurther investigation. However, the im¬pact of labor support on maternal andinfant health is apparent. Providing la¬bor support decreases the need for ob¬stetric interventions, such as oxytocinaugmentation, epidural anesthesia, andcesarean deliveries, reducing the possi¬bility of complications. If the incidenceof maternal fever during labor is re¬

duced, fewer infants will be kept in thehospital for sepsis evaluations. Thus,fewer infants will be exposed to nosoco¬mial infections, fewer mother-infantpairs will be separated shortly afterbirth, and fewer mothers may discon¬tinue breast-feeding.

Finally, a reduction in medical costswould be realized by the addition of a

supportive companion for all laboringwomen.1718 Anesthesia costs would besignificantly lower as more suportedwomen labored and delivered with localor no anesthesia. Expenses would bereduced because fewer infants born tosupported women would require an ex¬tended hospital stay. The greatest re¬duction in medical costs would occurwith the decrease in cesarean deliveriesand the accompanying reduction in op¬erating room costs, decreased need forskilled delivery and postpartum staff,less use ofmedications, and shorter ma¬ternal hospital stays.

While there may be financial savings,the physical and emotional benefits ofdoula support for the welfare of moth¬ers and infants make a compelling casefor the review of current obstetric prac¬tices. The current study is the first dem¬onstration in a technologically sophisti¬cated US hospital of a behavioralintervention that significantly reducedcesarean section rates, forceps deliver¬ies, need for oxytocin, and use of epidu¬ral anesthesia. In addition, the presenceof a labor companion in this settingshortened labor and decreased the inci¬dence of maternal fever. In light of thehigh level of obstetric technology usedfor the women in the present study, theresults of this simple and noninvasiveintervention are impressive. Labor sup¬port is centuries old, but its advantageshave now been validated in three con¬trolled studies and its positive benefitsshould not be overlooked in the trendtoward more and increasingly complextechnology. For those who provide care

for mothers during labor, the challengeis to turn to obstetric technology onlywhen necessary, relying instead on thepractice of continuous labor support tohelp the birth process follow its natural,normal course.

This research was made possible in part by grantHD 16915 to assess Perinatal Effects of Support forCouples During Labor, awarded by the NationalInstitute ofChild Health and Human Development,Bethesda, Md; by the Arthur Vining Davis Founda¬tions, Jacksonville, Fla; and by the Pittway Corpo¬ration Charitable Foundations, Chicago, 111.References1. Sosa R, Kennell J, Klaus M, Robertson S, Urru-tia J. The effect of a supportive companion on peri-natal problems, length of labor, and mother-infantinteraction. N Engl J Med.1980;303:597-600.2. Klaus MH, Kennell JH, Robertson SS, Sosa R.Effects of social support during parturition on ma-ternal and infant morbidity. BMJ. 1986;293:585-587.3. Halpin GJ, Rose E, Shapiro E. Trends in cesare-an section rates. N Engl J Med. 1989;86:867-873.4. Stafford RS. Alternative strategies for control-ling rising cesarean section rates. JAMA. 1990;263:683-687.5. Notzon FC. International differences in the useof obstetric interventions. JAMA. 1990;263:3286\x=req-\3291.6. Shiono PH, Fielden JG, McNellis D, RhoadsGG, Pearse WH. Recent trends in cesarean birthand trial of labor rates in the United States. JAMA.1987;257:494-497.7. Keirse MJNC, Enkin M, Lumley J. Social andprofessional support during labor. In: Chalmers I,Enkin M, Kearse MJNC, eds. Effective Care inPregnancy and Child Birth. New York, NY: Ox-ford University Press; 1989;2:805-814.8. Bertsch TD, Nagashima-Whalen L, DykemanS, Kennell JH, McGrath S. Labor support by first\x=req-\time fathers: direct observations. J Psychosom Ob-stet Gynecol. 1990;11:251-260.9. Lomas J, Anderson TM, Domnick-Pierre K,Vayda E, Enkin MW, Hannah WJ. Do practiceguidelines guide practice? the effect of a consensusstatement on the practice of physicians. N Engl JMed. 1989;321:1306-1311.10. Lederman RP, Lederman E, Work BA Jr,McCann DS. Anxiety and epinephrine in multipa-rous women in labor: relationship to duration oflabor and fetal heart rate pattern. Am J ObstetGynecol. 1985;153:870-877.11. Zuspan FP, Cibils LA, Pose SV. Myometrialand cardiovascular responses to alterations in plas-ma epinephrine and norepinephrine. Am J ObstetGynecol. 1962;84:841-851.12. Barton MD, Killam AP, Meschia G. Responseof ovine uterine blood flow to epinephrine and nor-epinephrine. Proc Soc Exp Biol Med. 1974;145:996\x=req-\1003.13. Adamsons K, Mueller-Heubach E, Myers RE.Production of fetal asphyxia in the rhesus monkeyby administration of catecholamines to the mother.Am J Obstet Gynecol. 1971;109:248-262.14. Myers RE. Maternal psychological stress andfetal asphyxia: a study in the monkey. Am J ObstetGynecol. 1975;122:47-59.15. Fusi L, Maresh MJA, Steer PJ, Beard RW.Maternal pyrexia associated with the use of epidu-ral analgesia in labor. Lancet. 1989;1:1250-1252.16. Thorp JA, Parisi VM, Boylan PC, JohnstonDA. The effect of continuous epidural analgesia oncesarean section for dystocia in nulliparous women.Am J Obstet Gynecol. 1989;162:670-675.17. Stafford RS. Cesarean section use and sourceof payment: an analysis of California hospital dis-charge abstracts. Am J Public Health. 1990;80:313-315.18. Newton ER, Higgins CS. Factors associatedwith hospital-specific cesarean birth rates. J Re-prod Med. 1989;34:407-411.

by guest on August 4, 2009 www.jama.comDownloaded from