The project reported here was a special component of a large multicenter randomized controlled trial, to evaluate a new ANC program . The primary hypotheses tested were that a new ANC model, consisting only of actions scientifically proven to improve maternal and newborn outcomes, was as effective as the traditional model with regard to specified maternal and perinatal end-points among singleton pregnancies, was not more expensive, and was accepted by women and providers . Conducted by WHO and collaborating organizations, fifty-three randomly allocated (cluster randomization) ANC clinics in Argentina, Cuba, Saudi Arabia and Thailand participated in the study, providing either the new program or the traditional program in use. There were 12,568 women randomized to the new model and 11,958 to the standard ANC model .
The model in the control clinics was the antenatal care currently offered, following guidelines formally recommended by the local health authorities, based on the traditional western model. In general, women made visits once a month during the first six months, one every 2–3 weeks for the next 2 months, and one every week until delivery. Clinical activities, urinary tests, syphilis screening, hemoglobin measurement, and blood group typing were done routinely.
In the intervention clinics, women judged not to need further assessment or special care at the time of the first visit according to predefined risk criteria were assigned to the new ANC model, which required fewer visits (usually four) with longer spacing between them than the standard ANC model recommended. Activities in the new program included: 1) screening for health conditions likely to increase the risk of adverse pregnancy outcomes; 2) providing interventions known to positively impact pregnancy outcomes, and excluding other common interventions that have not proved beneficial to pregnant women (e.g., maternal weight was measured only during the first visit; subsequent measurement was limited to patients with low weight); and 3) alerting pregnant women to potential health problems, especially emergencies, and instructing them on appropriate responses (e.g., recommendations for emergencies were provided in each visit; at the third visit instructions for delivery and suggestions for breastfeeding and contraception were included) . Results of the general effectiveness of the new ANC model (measured by examining low birth weight for the fetal condition and rates of preeclampsia/eclampsia, severe postpartum anemia, and treated urinary tract infection or pyelonephritis for the maternal conditions) have been published elsewhere . The assessment of women and provider satisfaction with the standard model and the new model is described here.
The assessment of women and providers' perception of the quality of both models of antenatal care was organized in two stages. First, we used an ethnographic approach, including focus group discussions and in-depth interviews with women and health personnel to assess the culture-related values in each country. During the qualitative stage we addressed general topics on health care provision and prenatal programs to gain initial understanding of the way health care was perceived in each specific cultural context . The findings of this in-depth study  were incorporated in the second stage (quantitative), which used standardized, closed-ended questionnaires that were prepared based on the most relevant categories obtained at the qualitative stage and the aspects of antenatal care that were expected to change as a result of the intervention (i.e., number of visits, spacing, time with provider, and information provision regarding maternal and perinatal health and complications). Both instruments (one for women, one for providers) were developed in English, translated into Spanish and Arabic, and piloted in each country. Changes suggested in each site were incorporated into the final version of the instruments by the investigators responsible of this component of the trial and then reviewed and approved by a WHO special technical review group. These final English versions (see Additional File 1 [Women's questionnaire] and Additional File 2 [Providers' questionnaire]) were again translated into local languages.
The questionnaire for women consisted of 24 questions addressing patients' preferences about the number of antenatal care visits; time spent in the waiting room and with the caregiver; and the amount and appropriateness of the information received during the visits. Women were also asked about their worries concerning their health status and their babies', and the reassurance they received from the provider. Because of the known limited validity of questions that include the word "satisfaction", we decided to include only one direct question ("Are you satisfied with the antenatal care you have received in this clinic so far?"), adapted from a previous antenatal care trial  to facilitate their meta-analysis, and two indirect ones ("Would you come back to this clinic?" and "Would you recommend this clinic to a relative or friend?"), which were developed based on information gathered from the focus groups (see Additional File 1 [Women's questionnaire]). We expected the "satisfaction" variable to synthesize women's overall perceptions of the quality of antenatal care .
The questionnaire for providers included 15 questions, probing the same issues as the patients': number and spacing of antenatal visits, time spent with the woman, information provided, perception of the quality of antenatal care, and recognition of women's satisfaction. (see Additional File 2 [Providers' questionnaire])
While some questions were worded per the terminology used in each country, their meaning was retained in the four settings. Both questionnaires were piloted in the four participating areas and adjusted accordingly .
Sample size and sampling strategy
The sample size was estimated to detect a minimum difference between a dissatisfaction rate of 5% in one arm and 10% in the other, with a two-sided test at a significance level of 5%, and with 80% power. The sample size obtained with standard formulas to compare two proportions for individually randomized design was multiplied by a design effect of 1.7 to account for the decrease in efficiency of the cluster randomized design. A sample of 1600 women (800 per arm of the study, 400 per country) was deemed necessary. A design effect of 1.5 had been previously calculated for the outcome of low birth weight . Since this design effect does not necessarily apply to a different outcome, and there was no information regarding design effects or intraclass correlation coefficients from other studies, we arbitrarily increased it to 1.7. The survey used clinics as strata, and women were sampled proportionally to each clinic's number of women per year.
The interviewers started surveying all eligible women on a randomly selected day and continued during working days until enrolling the estimated sample size. Because women needed to be sufficiently exposed to ANC in order to form their own opinion on the quality of care they had received, we administered the questionnaire only to patients that were at 32 weeks of gestational age and had attended the health care facility for their second or subsequent antenatal visit. The women were surveyed in a private environment, in approximately 15 minutes. We did not request an individual separate informed consent for this component of the trial but we did have special institutional consent. Therefore, the questionnaire was administered to all women that met the inclusion criteria (i.e., 32 weeks of gestational age and two or more antenatal care visits) in both clinics' groups until completion of the sample size.
We asked all 174 ANC providers from both intervention and control clinics to complete a self-administered questionnaire that took approximately 10 minutes. We recruited 92 caregivers from the experimental institutions (57 physicians, 33 nurses, and 2 midwives) and 82 from the routine care arm (54 physicians, 25 nurses, and 3 midwives). No provider refused to fill the instrument.
Regarding the women's survey, affirmative answers to questions about satisfaction measured overall satisfaction (primary outcome). Other satisfaction dichotomous outcomes were satisfaction with number of visits, spacing between visits, waiting time, and time spent with provider.
Additionally, the following four summary indexes were constructed as outcomes for the women's survey: 1) Information received, defined as the number of 'as much as you wanted' answers provided by a woman to the six questions on information received about looking after own health, tests during pregnancy, any treatment that might be needed during pregnancy, labor, breastfeeding and family planning; 2) Information about how to recognize problems, defined as the number of 'yes' answers provided by a woman to the six questions on: whether she was told how to recognize the following pregnancy-related problems: rupture of membranes, hemorrhage, premature contractions, dizziness and fainting, fever, and other; 3) Information about what to do in the presence of the above-described problems; and 4) Information about how to recognize and handle these problems. For every woman, each index summarized six questions of the survey, thus the numerical value could vary from 0 to 6.
For the providers' questionnaire, information given was measured through an index defined as the number of 'yes' answers to the six questions about health, tests and treatments during pregnancy, labor and delivery, breastfeeding and family planning. This index also ranged from 0 to 6.
Percentages or mean and standard deviations, as appropriate, were computed by group for baseline variables for the women interviewed, by arm, and checked for imbalance between groups. Baseline statistics for the sub-sample of women interviewed were compared with those for all participants to confirm that they were representative of the main trial population.
For the women's survey, the average values of the event rates of satisfaction outcomes were compared between arms, using a rate difference and a t-test at the cluster level, obtaining the standard errors for the difference from a variance analysis adjusted for strata. The indexes were analyzed as numeric outcomes using a random model approach, with clinic and subject as random factors, and arm and strata as fixed factors. Outcomes were adjusted for baseline variables showing a prognostically important imbalance to detect a possible confounding effect.
Since all the clinics' antenatal caregivers were interviewed, and as they constituted a fixed population, the providers' questionnaire was analyzed descriptively by computing percentages or mean and standard deviations, as appropriate.
This component of the study was reviewed and approved as part of the overall ethical review of the WHO trial, which was approved by the Scientific and Ethical Review Group of the UNDP/UNFPA/WHO/World Bank Special Programme on Research, Development and Research Training in Human Reproduction, the WHO Secretariat Committee for Research into Human Subjects, the Institutional Review Boards of the individual participating centres, and corresponding health authorities of the regions where the trial was implemented.