Background: Several qualitative studies have described disrespectful, abusive, and neglectful treatment of women during facility-based childbirth, but few studies document the extent of person-centred maternity care (PCMC)—ie, responsive and respectful maternity care—in low-income and middle-income countries. In this Article, we present descriptive statistics on PCMC in four settings across three low-income and middle-income countries, and we examine key factors associated with PCMC in each setting. Methods: We examined data from four cross-sectional surveys with 3625 women aged 15–49 years who had recently given birth in Kenya, Ghana, and India (surveys were done from August, 2016, to October, 2017). The Kenya data were collected from a rural county (n=877) and from seven health facilities in two urban counties (n=530); the Ghana data were from five rural health facilities in the northern region (n=200); and the India data were from 40 health facilities in Uttar Pradesh (n=2018). The PCMC measure used was a previously validated scale with subscales for dignity and respect, communication and autonomy, and supportive care. We analysed the data using descriptive statistics and bivariate and multivariate regressions to examine predictors of PCMC. Findings: The highest mean PCMC score was found in urban Kenya (60·2 [SD 12·3] out of 90), and the lowest in rural Ghana (46·5 [6·9]). Across sites, the lowest scores were in communication and autonomy (from 8·3 [3.3] out of 27 in Ghana to 15·1 [5·9] in urban Kenya). 3280 (90%) of the total 3625 women across all countries reported that providers never introduced themselves, and 2076 (57%) women (1475 [73%] of 1980 in India) reported providers never asked permission before performing medical procedures. 120 (60%) of 200 women in Ghana and 1393 (69%) of 1980 women in India reported that providers did not explain the purpose of examinations or procedures, and 116 (58%) women in Ghana and 1162 (58%) in India reported they did not receive explanations on medications they were given; additionally, 104 (52%) women in Ghana did not feel able to ask questions. Overall, 576 (16%) women across all countries reported verbal abuse, and 108 (3%) reported physical abuse. PCMC varied by socioeconomic status and type of facility in three settings (ie, rural and urban Kenya, and India). Interpretation: Regardless of the setting, women are not getting adequate PCMC. Efforts are needed to improve the quality of facility-based maternity care. Funding: Bill & Melinda Gates Foundation, Marc and Lynne Benioff, and USAID Systems for Health.
The data for this analysis are from four different surveys: two in Kenya10 and one each in Ghana (unpublished) and India.33 These were independent studies with different study goals, in-country collaborating partners, data-collection teams, and study procedures. The University of California, San Francisco (UCSF), USA, was, however, a common collaborating partner across all four studies. Additionally, the questionnaire for each of the surveys included the PCMC scale and some questions on characteristics of the respondents and the facility they gave birth in. Respondents in all studies were postpartum women and girls aged 15–49 years and all gave written informed consent after receiving information about the research. Differences in the four data sources are summarised in table 1. Data sources UCSF=University of California, San Francisco. KEMRI=Kenya Medical Research Institute (Kenya). IPA=Innovations for Poverty Action (Kenya). NHRC=Navrongo Health Research Center (Ghana). CEL=Community Empowerment Lab (India). The proposal and study materials for the projects that provide data for this manuscript were reviewed and approved by the UCSF Committee for Human Subjects, the Kenya Medical Research Institute Scientific and Ethics Review Unit, the Navrongo Health Research Center in Ghana, and the Community Empowerment Lab in India. In Kenya, one survey was done in the Migori County—a predominantly rural county in western Kenya. The survey was part of a research study on community perceptions of quality of care during childbirth.10, 34, 35, 36 The interviews were in English, Swahili, and Luo and took place in private spaces at health facilities or in the homes of the respondents. Data were collected by use of the REDCap application on a tablet, with data uploaded directly online. 1052 women were interviewed, with 433 (41%) of the interviews held at a health facility. We analysed data from women who delivered in a health facility (n=894) and who provided complete information on the PCMC items (n=877). We refer to this sample as the rural Kenya group. The other survey in Kenya was done at seven government health facilities in the Nairobi and Kiambu Counties. Nairobi is the national capital of Kenya and is 100% urban. The Kiambu County, which is adjacent to Nairobi, is 60% urban, but the study facilities were located in the urban portions of the county.37 The survey was done to obtain baseline data for the evaluation of a project for the improvement of person-centred care quality. Interviews were in English or Swahili, or both, and took place in a private space at the facility. Data from the interviews were collected using the SurveyCTO platform on a tablet, with data uploaded to the server at the end of each day. 531 women were interviewed. We analysed data from women who provided complete information on all the PCMC items (n=530). We refer to this sample as the urban Kenya group. The sampling procedures for the two Kenya surveys are described in detail elsewhere.10 In Ghana, the survey was conducted in five health facilities in the East Mamprusi district—a rural district in northern Ghana. The survey was done to obtain baseline data for the evaluation of an intervention for the improvement of maternal and newborn quality of care. The interviews were held in Mampruli and Kokomba, in private spaces at the health facilities and in the women’s homes. Interviews were all paper based, and responses were subsequently entered into the REDCap portal on a computer. 268 women were interviewed. We analysed data from women who delivered in a health facility (n=227) and who provided complete information on the PCMC variables (n=200). We refer to this sample as the Ghana group. In India, the survey was conducted in 40 public health facilities in 20 predominantly rural districts of Uttar Pradesh, a state in northern India. The survey was done as part of a cross-sectional study on quality of maternity care in Uttar Pradesh. All interviews were in Hindi and took place at the health facility, most of them (2015 of 2018 interviews) in the postnatal ward at the patient’s bed. Interviews were held using the CommCare platform on tablets, with data uploaded to the server at the end of each day. 2018 women were interviewed, with roughly 50 women interviewed per facility. We refer to this sample as the India group. We measured PCMC on the PCMC scale, which was initially validated in the Kenya group and subsequently in the India group, and shown to have high content, construct, and criterion validity and to offer good internal-consistency reliability (described in detail elsewhere).10, 33 The scale includes 30 items that span three domains: dignity and respect, communication and autonomy, and supportive care. Each item has a four-point response scale—ie, 0 (“no, never”), 1 (“yes, a few times”), 2 (“yes, most of the time”), and 3 (“yes, all the time”). The process towards the development of the scale included literature and expert reviews to assess content validity, cognitive interviews with women to evaluate wording and appropriateness of the items, and psychometric analysis using survey data to assess construct and criterion validity and reliability. The validation of the PCMC scale was one of the objectives of the Kenya studies and that of a related study in India. The final scale is based on findings from expert reviews and cognitive interviews from both Kenya and India, with iterative translation from English to the local languages at each stage.10 This scale was used in Ghana with only minor modifications during pretesting. The full scale and subscales have good internal-consistency reliability in all the groups, with a Cronbach’s α value of over 0·8 for the full scale across all groups and ranging between 0·61 and 0·75 for the subscales. The overall PCMC score is a summative score from the responses to individual items in the 30-item PCMC scale (with negative items reverse coded—ie, questions that were framed negatively, such as the physical and verbal abuse questions, had to be recoded so that high numbers represent good care). The minimum possible score is 0 and the maximum possible is 90, with a low score indicating poor PCMC. In addition to presenting overall PCMC scores and domain scores, we examined individual items to highlight gaps in key dimensions of PCMC. We examined potential predictors of PCMC using variables that were captured similarly in all four groups. These included demographic variables such as age, parity, and marital status and measures of socioeconomic status (ie, education, employment, and household wealth). We also included variables to capture complications, antenatal attendance, and facility and provider characteristics. Facilities were characterised by the type of facility the woman delivered in, categorised as public or government hospital (high level), health centre (low level), or private or mission health facility (too few to group by levels). Provider characteristics were type (ie, skilled providers, including nurses or midwifes, clinical officers or medical assistants, and doctors; non-skilled providers, including support staff or traditional birth attendants; and more than one skilled provider) and gender of delivery providers. All data were retrieved from REDCap, SurveyCTO, and CommCare and imported into Stata 15 for cleaning and analysis. Analysis involved descriptive statistics for each of the groups and bivariate and multivariate analyses to examine associations between independent variables and the PCMC score. We first examined mean differences in PCMC by all the potential predictors using cross tabulations and unadjusted ordinary least-squares regressions, as the PCMC scale is normally distributed. We then built the multivariate models for each group by including all variables that were significantly associated with PCMC in the bivariate models for at least one group. We also included variables such as age, parity, and pregnancy complications that were not significant in any of our bivariate models, but which we believed were potential predictors of PCMC on the basis of previous research.13, 25, 27, 38 A p value of less than 0·05 was considered significant. We did not combine the datasets to test statistical significance between countries because of the described differences between groups. The sponsors of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the Article. The corresponding author had full access to all of the data in the study and had final responsibility for the decision to submit for publication.
N/A