How women are treated during facility-based childbirth in four countries: a cross-sectional study with labour observations and community-based surveys

Background: Women across the world are mistreated during childbirth. We aimed to develop and implement evidence-informed, validated tools to measure mistreatment during childbirth, and report results from a cross-sectional study in four low-income and middle-income countries. Methods: We prospectively recruited women aged at least 15 years in twelve health facilities (three per country) in Ghana, Guinea, Myanmar, and Nigeria between Sept 19, 2016, and Jan 18, 2018. Continuous observations of labour and childbirth were done from admission up to 2 h post partum. Surveys were administered by interviewers in the community to women up to 8 weeks post partum. Labour observations were not done in Myanmar. Data were collected on sociodemographics, obstetric history, and experiences of mistreatment. Findings: 2016 labour observations and 2672 surveys were done. 838 (41·6%) of 2016 observed women and 945 (35·4%) of 2672 surveyed women experienced physical or verbal abuse, or stigma or discrimination. Physical and verbal abuse peaked 30 min before birth until 15 min after birth (observation). Many women did not consent for episiotomy (observation: 190 [75·1%] of 253; survey: 295 [56·1%] of 526) or caesarean section (observation: 35 [13·4%] of 261; survey: 52 [10·8%] of 483), despite receiving these procedures. 133 (5·0%) of 2672 women or their babies were detained in the facility because they were unable to pay the bill (survey). Younger age (15–19 years) and lack of education were the primary determinants of mistreatment (survey). For example, younger women with no education (odds ratio [OR] 3·6, 95% CI 1·6–8·0) and younger women with some education (OR 1·6, 1·1–2·3) were more likely to experience verbal abuse, compared with older women (≥30 years), adjusting for marital status and parity. Interpretation: More than a third of women experienced mistreatment and were particularly vulnerable around the time of birth. Women who were younger and less educated were most at risk, suggesting inequalities in how women are treated during childbirth. Understanding drivers and structural dimensions of mistreatment, including gender and social inequalities, is essential to ensure that interventions adequately account for the broader context. Funding: United States Agency for International Development and the UNDP/UNFPA/UNICEF/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction, Department of Reproductive Health and Research, WHO. Twelve health facilities (maternity hospitals and maternity units within general hospitals [eg, district or regional hospitals]; three per country, all in urban areas) were purposively selected (appendix p 2). Health facilities were included in the study if they were not included in the formative phase, were a secondary-level facility or higher, had at least 200 births per month, had a well defined community catchment area, and allowed non-clinicians to perform observations. Data collection took place in Nigeria from Sept 19, 2016, to Feb 26, 2017, in Ghana from Aug 1, 2017, to Jan 18, 2018, in Guinea from July 1 to Oct 30, 2017, and in Myanmar from Jun 26 to Sept 5, 2017. The labour observations were continuous, one-to-one observations of women by study researchers from admission, throughout labour and childbirth, until 2 h post partum. Labour observations were not done in Myanmar. The community-based survey was done with women up to 8 weeks post partum. Women were eligible for the labour observation if they were admitted for childbirth in early established or active labour (<6 cm cervical dilation), were aged at least 15 years, were willing and able to participate, and provided informed consent. Women were not eligible if they were admitted for reasons other than childbirth, immediately transferred or taken directly to theatre, a first-degree relation to a facility employee (mother, sister, cousin), or distressed or otherwise unable to reasonably consent. Pregnant women who were not admitted were eligible to participate if they returned and were admitted for childbirth. Women were eligible for the survey if they were admitted for childbirth, were aged at least 15 years, were willing and able to participate, resided in the catchment area, and provided consent. Women were not eligible if they were admitted for reasons other than childbirth, were a first-degree relation to a facility employee, were distressed or otherwise unable to reasonably provide consent, resided outside the catchment area, or were unable to provide sufficient contact information. All women provided written consent. Institutional permission for recruitment and observation was obtained from each site; consent was not sought from providers. This study was approved by the WHO Ethical Review Committee, WHO Review Panel on Research Projects, and in-country ethics committees. The country-specific ethical review committees that reviewed and approved this project were Le Comité National d'Ethique pour la Recherche en Santé (Guinea); Federal Capital Territory Health Research Ethics Committee (Nigeria); Research Ethical Review Committee, Oyo State (Nigeria); State Health Research Ethics Committee of Ondo State (Nigeria); Ethical Review Committee of the Ghana Health Service (Ghana); Ethical and Protocol Review Committee of the College of Health Sciences, University of Ghana (Ghana); and Ethics Review Committee, Department of Medical Research (Myanmar). Each study site had two or three data collectors per shift to manage recruitment and data collection. All data collectors were experienced women (aged ≥18 years) trained in research methods, and not providers or clinical trainees or students.14 All women admitted to the facility during the study period were assessed for eligibility. Data collectors approached women face to face and invited them to participate. Women meeting the eligibility criteria were provided with information about the study and those who agreed to participate consented and were enrolled. Women eligible to participate in both the labour observation and survey were asked to participate in both, and data were linked by the medical record number (results of linked data will be reported elsewhere). For the labour observations, every eligible woman could not be observed because of practical limitations around the number of data collectors required. To minimise selection bias, when a data collector completed an observation, she returned to the admissions area to enrol the next eligible woman. Each participant was assigned a number, used in all other data collection forms. Further communication between the data collector and the participant was discouraged. The timeframe of interest was from admission until 2 h post partum, facility discharge, or maternal death (whichever happened first). The data collector observed the participant continuously throughout labour, childbirth, and up to 2 h post partum, meaning that there was one data collector per woman, observing only one woman at a time throughout the period of interest. Data collection took place 24 h per day, 7 days per week to ensure no coverage gaps and minimise truncation bias (terminating the observation early because the woman had not given birth). A structured observation guide was used to record interactions between the woman and provider and her birth environment. Recruitment continued until the facility sample size was reached. There was no contact with participants after the observation, unless they were also enrolled in the survey. For the survey, women received a telephone call at 2–3 weeks post partum to schedule the survey at a time and place of their convenience. Contact was attempted up to three times over 2 weeks. Women who could not be contacted were recorded as lost to follow-up. Data collectors travelled to the interview location, reaffirmed consent, and administered the survey in a private place with no other people present. Recruitment continued until the facility sample size was reached; there was no contact after survey administration. An iterative mixed-methods approach was used for tool development, described in detail elsewhere.14 The typology of mistreatment provided the structure, domains, and items.5 Both tools are available in open access.14 Data were collected using digital, tablet-based tools (BLU Studio XL2, Android, BLU Products, Miami, FL, USA). The labour observation tool has three forms completed and submitted at different times: (1) admission form; (2) incident report form; and (3) childbirth, interventions, and discharge form.14 The admission form was completed once (immediately after enrolment) for all women, and included screening questions and sociodemographics. The incident report form was completed for the following events: physical or verbal abuse, stigma or discrimination, or vaginal examination, and could be submitted multiple times (repeating form for multiple events). For physical or verbal abuse and stigma or discrimination, the incident report included the timing and type of provider involved. For vaginal examinations, information was collected about consent, privacy, and confidentiality. The childbirth, interventions, and discharge form was completed once at the end of the observation for all women, and included pain relief, mobilisation, fluids, companionship, fees, neglect, privacy, health outcomes, and interventions. The survey tool had two forms completed and submitted at different times. The screening form assessed eligibility.14 The survey form was completed during survey administration,14 and included sociodemographics, birth experiences (including mistreatment, vaginal examinations, companionship, and pain relief), health outcomes, interventions, post-partum depression, and satisfaction with care. Few data exist estimating the prevalence of mistreatment of women during childbirth, complicating sample size calculation. For the labour observation, we prespecified sample size for the development sample (Nigeria) of 130 women per facility and 390 women in total.14 For the survey, we used the same calculation and assumed 30% loss to follow-up between recruitment and survey administration; the target sample size for Nigeria was 169 women per facility and 507 women in total. The prevalence of any type of physical abuse, verbal abuse, or stigma or discrimination in Nigeria was used as a proxy for a prevalence estimate for the other countries. For the validation sample, the required sample size was 209 women per facility, and 627 women per country, based on ±5% precision, 80% sensitivity, 5% type 1 error (two-tailed), and 30% prevalence. Data were submitted using a 3G cellular connection. Consistency checks of screening logs, recruitment, and data were done weekly by WHO and country research teams; inconsistencies were resolved during data collection. Data analysis was done with SAS version 9.4. For observation and survey data, sociodemographics, health outcomes, and interventions (categorical variables) were aggregated and presented as proportion of the total study population and by country (see appendix p 3 for description of variables). In the labour observation, specific acts of physical or verbal abuse, and stigma or discrimination were collected as recurring events. Observed events were aggregated and presented as proportion of participants with at least one occurrence for the total study population, stratified by country. We assessed temporal patterns of physical or verbal abuse among women with complete observations for at least 1 h before and after childbirth using two methods: (1) aggregating the total number of incidents in 15-min intervals and deriving a density measure of number of mistreatment events per 1000 women for each 15-min interval, and (2) generalised linear regression to determine odds of an incident in the 15-min interval, relative to 1 h before childbirth, adjusting for country and correlation due to repeated measures. For the community survey, specific acts of physical or verbal abuse, and stigma or discrimination were aggregated into dichotomous variables (yes or no), then aggregated into a single indicator (yes or no) for each domain. Multivariable logistic regression models were fitted to evaluate factors associated with mistreatment (age, education, marital status, number of previous births, use of curtains) across each domain, adjusting for country. Possible effect modification by woman's education was assessed for verbal abuse, and by single marital status for non-consented vaginal examination. Generalised linear models were fitted to estimate the odds of privacy during vaginal examinations and autonomy, adjusting for a facility clustering effect given potentially different policies across facilities. Outcomes of interest were selected items from each domain: physical abuse; verbal abuse; any physical or verbal abuse, or stigma or discrimination; non-consented vaginal examination; non-private vaginal examination; neglect; long wait times or delays; and autonomy or mobilisation (ie, upright and able to move freely around the room or ward). All the funders of the study were involved in developing the research question and investigator meetings, but had no other roles in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.