A prospective, observational, longitudinal study was designed to follow up a cohort of pregnant women from the Biscay health area attending midwife offices of the public Basque Health Service/Osakidetza, in order to assess the effect of AE exposure on childbirth and breastfeeding. The project was approved by the Committee for Primary Care Research of the Basque Health Service/Osakidetza.
Thirty-four centres, 65% of those having primary care midwives within the Basque Health Service/Osakidetza, collaborated in the project. All four districts of the Biscay health area were represented.
AE is given in all collaborating primary care centres. This consists of at least 8 sessions where breathing techniques, pushing, and relaxation are practised, and different topics such as labour and delivery, postpartum period, newborn care, and breastfeeding are addressed. This study assumes that, in our area, AE sessions are relatively similar in terms of structure, duration, topics, and body work practised, while recognising that every healthcare professional may make use of his/her specific knowledge and pay special interest to any particular aspect.
The two hospitals of the public Basque Health Service having a delivery area (Hospital de Cruces and Hospital de Basurto) collaborated in the study.
All nulliparous pregnant women from the Biscay health area aged 18 to 42 years who attended the office of a primary care midwife from week 36 of pregnancy and who did not meet any exclusion criterion (Table ) were eligible for the study.
Exclusion criteria for participants in the EMA study
Recruitment lasted eight months, from September 2005 to May 2006. Primary care midwives invited a consecutive sample of nulliparous pregnant women eligible to participate to collaborate in the study. All women collaborating in the study signed an informed consent explaining the main study aspects and collaboration conditions.
Thirty primary care midwives collaborated in the study by proposing participation to all pregnant women eligible for inclusion, discussing with them the whole study process, providing informed consent, and collaborating in the initial measurements.
Hospital care midwives and resident midwives at both Hospital de Cruces and Hospital de Basurto detected pregnant women in the study and performed measurements during the labour, delivery and postpartum periods at the hospital.
AE exposure was measured by counting the AE sessions attended by each participant. This variable was measured during the postpartum period by hospital midwives and for incomplete records, by telephone interview conducted one and a half months after childbirth. Three exposure groups were formed based on the number of sessions attended by participants: (A) women attending no session, (B) women attending 1 to 4 sessions, and (B) women attending 5 or more sessions. Sixty percent of nulliparous pregnant women were expected to attend any AE session, and no AE exposure was expected in the remaining 40%.
Potential confounding variables recorded included age, sociodemographic characteristics such as social class, educational level and nationality [11
], and woman personality, that may influence both AE session attendance or otherwise and childbirth outcome and breastfeeding. The personality measurement instrument was the Battery of scales of generalized expectancies for control (BEEG-20) [12
Outcome variables related to objective childbirth aspects that were measured included: (1) number of visits for false labour (visits to the emergency room in a latent labour stage not leading to admission for labour), (2) duration of dilation stage in minutes, (3) duration of expulsion stage in minutes, (4) cervical dilation in centimetres at the time epidural anaesthesia is administered, (5) type of delivery (normal/instrumental/Caesarean section), (6) presence of episiotomy and/or tear and grade [13
], (7) sex of the newborn (male/female), (8) weight of the newborn in grams, and (9) Apgar test of the newborn. Objective childbirth variables were collected from the clinical records of pregnant women.
Subjective childbirth variables measured included: (10) degree of anxiety according to the Hospital Anxiety and Depression Scale (HAD) [14
], and (11) degree of satisfaction caused by this life experience, assessed based on the answers to two questions asked after childbirth: (a) How would you rate your childbirth experience in a scale where 0 would be very negative and 10 very positive? And (b) In relation to the expectations you had, would you say that your childbirth experience has been: Better than you expected/As you expected/Worse than you expected? The HAD anxiety scale was measured during the dilation period, always before the start of the expulsion period and satisfaction once delivery had ended.
Breastfeeding variables measured included: (12) early start (within 2 hours of childbirth) and (13) continuation of breastfeeding, either alone or combined with other feeding, at one and a half, three, six, nine, and 12 months of life. Early start of breastfeeding was recorded by hospital midwives and continuation of breastfeeding was recorded through telephone surveys conducted from the Primary Care Research Unit.
The estimated sample size was 657 women. Assuming that 40% of these women did not attend AE, this size provided a statistical power greater than 95% for detecting as significant 15% differences in false labour, anaesthesia in latent stage, instrumental deliveries and breastfeeding after one and a half months, and 20% differences in anxiety and episiotomy, and for detecting equivalence in the duration of the dilation and expulsion periods. Alpha error level is less than 5%.
The basic analysis will involve the calculation and comparison of proportion of dichotomous outcomes and averages of quantitative results among the cohorts exposed to different levels of AE. Both relative and absolute measures of association with 95% confidence intervals will be calculated and tested by chi-square and t tests The exposure groups will also be compared to detect unbalances with respect to potential confounding variables and characteristics that could be associated with the outcomes under study. To control for the eventual confounding effect of those characteristics, stratified analyses and multivariate statistical models will be used. These models will be linear for continuous outcomes an logistic for dichotomous ones. These same models will be extended to mixed effects models in which the midwife will be included as a random effect to take into consideration the hyerarchical structure of data (pregnant women clustered in the practice of each midwife) and to estimate the particular effect of each of them on the results. To analyse time to discontinuation of breastfeeding, Cox proportional hazards models will be adjusted. All analyses will be performed using EpiInfo (Centers for Disease Control and Prevention -CDC- V. 3.3.2 2005) and SAS (SAS Institute Inc., Cary NC, 2003) softwares.
The number of eligible women who refused to participate in the study was very low according to midwives in charge of recruitment, but was not recorded. As shown by the study flowchart (Figure ), 641 women eligible to participate in the study were recruited from September 2005 to May 2006. Among these 641 participants, 21 were withdrawn from the study because they met some initially undetected exclusion criterion. Finally, 620 pregnant women were included in the EMA study. Study losses included two women who did not have their babies in the reference hospitals and two women who refused to continue collaborating with the study. Thus, 616 women finally participated in the study. Breastfeeding measurements were performed in all women but two (99%).
Flowchart of the cohort study EMA.
Forty percent of pregnant women were expected not to have participated in any AE session. However, the proportion of pregnant women who did not attend AE did not exceed 7.3% of participants despite prolongation of the recruitment period from 4 to 8 months, while 92.6% of nulliparous pregnant women attended AE. This low exposure variability decreased statistical power of the study as compared to the initially planned power. Despite this, there was a greater than 80% power for detecting as significant differences between exposure groups, as shown in Table .
Hypothesised differences between antenatal education exposure groups for whose detection the EMA study has a statistical power greater than 80%
Mean age of participants was 31.3 years, virtually all of them are Spanish in nationality (94.3%), and a great majority have a secondary school or university educational level (n = 510). With regard to comparability of exposure groups, sociodemographic characteristics of pregnant women are seen to differ depending on grade of AE exposure. Thus, women attending more sessions (group C) are older (p < 0.01), have a higher educational level (p < 0.01) and socioeconomic level (p < 0.01), and are Spanish in a greater proportion. As to personality characteristics, pregnant women in all three groups (A, B, and C) are similar in all personality scales (Table ).
Sociodemographic and personality characteristics of participants in the ema study
During data collection in and after childbirth, two specially designated midwives, one at each hospital, monitored measurements to minimize potential interferences or losses.
The whole process, including recruitment and measurements, was supervised and controlled by the Primary Care Research Unit of Biscay. Indicators used for controlling such processes were as follows:
(a) weekly recruitment rate: 2 women per week each midwife
(b) percent losses during the process: less than 7%
(c) proportion of inadequately completed forms: less than 2%