Setting and data collection
A retrospective cross-sectional population-based household study was conducted from 7th July to 15th August 2014 in the Northern Province and Kigali City, the capital and largest city in Rwanda. From a complete list of 4791 villages in the five districts of the Northern Province and three districts within Kigali City [15] a random selection was done to select 48 villages. Approximately 20% of Rwandan population lives in urban areas and this proportion of villages were selected from urban areas [3]. A proportionate number of households were selected from each village and community health workers in each village who keep records of pregnancies and childbirths helped identify study participants. Sample size was estimated to include 922 participants from a total population of 2,865,355 and is described elsewhere [16].
Participants
Women were eligible for inclusion if they had given birth 1 to 13 months earlier (gave birth between 31st May 2013 and 30th June 2014). In total, 922 women were asked to participate and only one woman declined. All women received verbal and written information and all participants gave written consent. The interviews were conducted in private, and only one woman was interviewed in each household for confidentiality purposes.
Data collection
Twelve female interviewers (10 nurses/midwives and 2 clinical psychologist) were recruited to interview eligible participants. Before the data collection the interviewers were trained for 5 days; 1 day of training focused on identifying eligible households and other listing procedures, 2 days were spent on questionnaire contents and ethical issues, 1 day of fieldwork to pre-test survey instruments and fieldwork procedures and 1 day of debriefing with feedback after the pre-test fieldwork.
Questionnaire
An interview questionnaire was made by the research team. It included background variables, questions about the women’s perceptions of care received during childbirth and a question where the women rated their overall childbirth experience. The questionnaire was developed in English and translated into Kinyarwanda by a medical doctor native in Kinyarwanda and skilled in English. The Kinyarwanda version of the questionnaire was checked during a working day with the data collectors and the research team and adjustments were done. Next a pilot study including 36 women from a neighbouring village was done to test face validity of questionnaire. All 36 women completed the test interviews and some minor changes of wording in a few questions were done but no major revision of the questionnaire was needed.
The outcome variable in this study was women’s rating of the overall childbirth experience. The question was formulated: “What was your overall experience of the childbirth?” and answered on an 11-point numeric rating scale ranging from 0 (Very bad) to 10 (Very good). In order to be used as the dependent variable in a logistic regression analysis, the response options were recoded to a dichotomous response, where 8–10 was defined as a good experience and 0–7 was considered being not a good experience (bad or mixed). The dichotomisation was based on the distribution of responses. The median value was 9 and 8–10 were the most common values with a clear decline in response rate from 8 (15.0%) to 7 (9.5%).
Used as independent variables were statements concerning perceptions of care for the woman to agree with or not; the women’s confidence in the medical staff, information received during childbirth, being treated with respect by staff, receiving necessary pain relief, getting support from the health care staff, getting help with initiating breastfeeding, and having the baby skin-to-skin after birth. These statements were answered on 4-point Likert scale with response options ranging from “Totally agree” to “Totally disagree”, except the question about having the baby skin-to-skin, which had dichotomous response options, with a “Yes” or “No”.
Statistical analyses
Descriptive analyses were computed for background variables; age, parity, education, marital status, number of people in the household, home province, place of childbirth, mode of delivery, uncomplicated pregnancy, maternal health status, age of baby and health status of the newborn baby 1 day after birth.
Univariable analyses were performed to test the association between each of the independent variables and the dependent variable to find predictors of childbirth experience. Distributions or response options and unadjusted odds ratios (OR) with 95% confidence intervals for each independent variable against the dependent variable were computed.
Variables from the univariate analyses with a p-value below 0.05 were included in a stepwise forward selection logistic regression model with the dichotomised childbirth experience as the outcome variable. Adjusted odds ratios (AOR) from the multivariable logistic model for childbirth experience were calculated with their 95% confidence intervals (CI) for each significant variable.
SPSS version 23 (SPSS Inc., Chicago, IL, USA) and version 9 of SAS System for Windows (Cary, NC, USA) was used for statistical analyses. All significance tests were two-tailed.