The PPA project: study design, sample size, inclusion criteria and data collection
Initially, 23 private hospitals were invited to participate in the PPA project in Brazil. A cross-sectional study was carried out in a convenience sample of 12 hospitals among these 23 hospitals to analyze the project's outcomes. For the sample selection, we considered three criteria that could have affected the degree of implementation of the PPA: hospital location according to geographic macro-region; type of hospital (hospitals owned or not owned by health insurance companies); hospital performance (hospitals that reported good and bad results in achieving the PPA c-section goals, according to administrative data provided by the PPA coordination board). The profile of these hospitals is presented in the Additional file 1. The study was conducted from March to August 2017, 6 to 8 months after the implementation of the PPA. It was eligible all women admitted for the birth of a live newborn (of any gestational age and birth weight) or a stillbirth (with gestational age ≥ 22 weeks and/or birth weight ≥ 500 g). Exclusion criteria were women who gave birth before admission to the hospital; women with extreme communicating difficulty, such as foreigners who could not understand Portuguese; deaf-mute women; women with mental or neurological diseases with severe cognitive impairment; and women who legally interrupted pregnancy. In each hospital, women were invited to participate in the study consecutively, until reaching the planned sample in each hospital.
Face-to-face interview with women at least 6 h after vaginal birth and 12 h after caesarean section was realized, using a structured questionnaire containing maternal identification, socioeconomic condition, previous obstetric history, maternal anthropometric data, prenatal care, illnesses and medication during gestation, labour, and birth, and assessment of care received by the woman and newborn. Also, data from medical records of the women and neonates following their discharge from the hospital, including prenatal cards and ultrasound exams was extracted. Trained interviewers by the study coordination applied all the questionnaires of the research. More information about PPA project can be seen in Torres et al. [6].
Birth experience study among woman of PPA Project: sample size, inclusion criteria and data analysis
To evaluate the association between exposition to PPA and the assessment by women of the birth experience we conducted an analysis in a sub-sample of PPA project. For this analysis, only women who desired vaginal delivery at the ending of pregnancy were included, totalizing 2348 participants out of 4798 of the total sample. This information was collected during the interview through the answer to two questions:
-
1.
“At the beginning of the pregnancy, what type of delivery did you want to have?” with options of response: vaginal or caesarean.
-
2.
“During the pregnancy, has your preference regarding the type of delivery changed? with options of response: Yes or No.
It was considered “women who desire vaginal birth at the end of pregnancy”: women who want vaginal birth at the beginning and not change their option or women who wish caesarean at the beginning of pregnancy and changed their opinion. Women who opted for scheduled caesarean section (c-section before labour) at the end of pregnancy were excluded (2530 women). This strategy aims to select women with a chance to be exposed to the PPA project.
Independent variable
Adequate Childbirth Project (PPA): It was considered exposed to PPA all women exposed to the four drives of PPA that stimulated vaginal birth, increasing the proportion of spontaneous or induced labour of full-term birth (≥ 39 weeks) and filled the target population criteria defined by each hospital. In two hospitals, the target population was composed by all primiparous women, in two hospitals by women in Robson’s groups 1 to 4, and in eight hospitals by women admitted by the hospital’s on-call staff (one of which was limited to women in Robson’s group 1 to 4 and another to women without anterior uterine scarring). These criteria were based on the higher probability of vaginal birth. The women that did not fill these criteria received the “Usual Care” service. All these pieces of information were collected in the interview with the women.
Outcome variable
Birth experience—The latent variable is composed by four indicators:
-
1.
Respectful treatment assessed through the question: “When you are in the hospital for delivery, how do you assess the respect of the professionals when receiving you and speaking to you?” with five options of response: Excellent, Good, Fair, Bad, Terrible.
-
2.
Participation in decisions assessed through the question: “When you are in the hospital for delivery, how do you assess the possibility of participating together with the health team in decisions about your labour and delivery?” with five options of response: Excellent, Good, Fair, Bad, Terrible.
-
3.
Assessment of childbirth care through the question: “In your opinion, the attendance to your delivery was…” with five options of response: Excellent, Good, Fair, Bad, Terrible.
-
4.
Assessment of baby care through the question: “In your opinion, baby care at the maternity ward where he/she was born was…” with five options of response: Excellent, Good, Fair, Bad, Terrible.
All indicators were used in statistical analysis in two categories Excellent/Good and; Fair/Bad/Terrible.
Other variables of the model
Education (complete high school and complete or incomplete higher education).
Age (14 to 19, 20 to 34, and 35 or more years old).
The economic status of the women was identified using the Brazilian Economic Classification Criteria that encompasses information about the level of education of the household's main breadwinner, the possession of selected appliances and durable assets, and whether there is a domestic employee at home, divided in six categories: A, B1, B2, C1, C2, and D [14]. In the descriptive analysis, the variable was grouped in three categories: A (represented by the richest women), B (represented by intermediate economic status), and C/D (represented the poorest women in the sample).
Parity, divided into two categories, primiparous or multiparous.
Planned pregnancy was evaluated by the question: “When you get pregnant you…” with the following answer options: wanted to become pregnant at that time, wanted to be pregnant later, or did not want to be pregnant. In this analysis, the variable was grouped in “yes” if wanted to become pregnant at that time; and “No” if wanted to be pregnant later or did not want to be pregnant.
Preference of type of birth in the early pregnancy was evaluated by the question: “At the beginning of the pregnancy, what type of delivery did you want to have?” categorized as vaginal or caesarean.
Pregnancy complication indicator was constructed by reference to at least one of the following manifestations during the pregnancy which could influence the health team and the woman to choose a caesarean section: hypertensive syndrome, gestational diabetes, infections, placenta previa, placental abruption, oligodramnia, polydramnia, and restricted uterine growth, analyzed as a dichotomic variable “Yes or No”.
Type of birth categorized as a vaginal birth (including forceps and vacuum-assisted deliveries) or a caesarean section.
Access to information was obtained asking for the woman if she was informed during the prenatal care about how labour begins, risk signs in pregnancy that should make her seek a health service, things she could do during the labour to facilitate the birth of the baby, not cutting the umbilical cord immediately after birth, having skin-to-skin contact with the baby in the delivery, and about breastfeeding in the first hour of life. The variable was a sum of the six items described above varying from 0 to 6. In the descriptive analysis, the variable was aggregate into two categories: less than three pieces of information and four or more information.
Oriented to look for this hospital/maternity because of PPA, with the option “Yes or No”.
Theoretical model
Figure 1 shows the theoretical model examining the association between exposition to Adequate Childbirth Project (PPA) (independent variable) and Birth experience (main outcome). The birth experience was defined as a latent variable composed by four indicators: respectful treatment, participation in decisions, assessment of childbirth, and satisfaction with baby care. Other variables which compose the model were considered confounding and mediating based on the existing literature. Observed variables are represented by rectangles, while ellipses represent a latent variable. The theoretical model took into account the temporality of the information obtained in the questionnaire, therefore time runs from left to right. All variables are connected by arrows forming a causal network of information.
Statistical analysis
Descriptive and bivariate analyses were conducted comparing the “Exposed to PPA” group versus “Usual Care” separately for women who had a vaginal delivery and caesarean section. To compare frequencies between groups, the Chi-square test was used, considering a confidence interval of 95%.
To account for the different number of births per year in each hospital, it was applying a weight to control this disparity. Hospitals with more births per year had a larger weight in sample size. Besides that, each hospital was considered as a stratum. After applying the calibration process described above was performed analysis using structural equation models. To construct the latent variable was considered a factor loading greater than 0.4, with a p-value of less than 0.05, indicative of a good correlation between the observed variable and the construct of interest [15]. Also, multigroup modelling to assess the differences in causal paths between vaginal birth and caesarean section was carried out.
The weighted least squares estimator adjusted by the mean and variance (Weighted Least Squares Mean and Variance adjusted—WLSMV) with probit link and theta parameterization to estimate the coefficients of the model was used. The full information method was used to leading with loss of information in some variables.
To assess suggestions for changes in the initial hypotheses were calculated the modification indices, using the MODINDICES command. When the proposed modifications (modification rates greater than 10) were considered plausible from a theoretical point of view, a new model was developed. In all analyses, the path was significant when the p-value was less than or equal to 0.05 [15, 16].
To assess the model adjustments, three criteria were adopted: The Root Mean Square Error of Approximation (RMSEA), the Comparative Fit Index (CFI) and the Tucker–Lewis Index (TLI). To indicate good fit model, were considered values less than 0.06 to RMSEA and above 0.95 to CFI and TLI [17, 18]. Also, for the RMSEA, the 90% confidence interval (CI) was calculated, and a lower limit close to 0 and the upper limit below 0.08 was considered appropriate [18]. Both the CFI and the RMSEA are sensitive to the lack of model specification and are affected only slightly by the sample size [15, 16].
For data analysis, R version 4.0.3 software (The R Foundation, Vienna, Austria), Mplus 8 software [19] and Stata 13 were used.
This study followed the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) recommendations for the reporting of cross-sectional research.