Mistreatment of newborns after childbirth in health facilities in Nepal: Results from a prospective cohort observational study

Background Patient experience of care reflects the quality of health care in health facilities. While there are multiple studies documenting abuse and disrespect to women during childbirth, there is limited evidence on the mistreatment of newborns immediately after childbirth. This paper addresses the evidence gap by assessing the prevalence and risk factors associated with mistreatment of newborns after childbirth in Nepal, based on a large-scale observational study. Methods and findings This is a prospective observational cohort study conducted over a period of 18 months in 4 public referral hospitals in Nepal. All newborns born at the facilities during the study period, who breathed spontaneously and were observed, were included. A set of indicators to measure mistreatment for newborns was analysed. Principal component analysis was used to construct a single newborn mistreatment index. Uni-variate, multi-variate, and multi-level analysis was done to measure the association between the newborn mistreatment index and demographic, obstetric, and neonatal characteristics. A total of 31,804 births of newborns who spontaneously breathed were included. Among the included newborns, 63.0% (95% CI, 62.5–63.5) received medical interventions without taking consent from the parents, 25.0% (95% CI, 24.5–25.5) were not treated with kindness and respect (roughly handled), and 21.4% (95% CI, 20.9–21.8) of them were suctioned with no medical need. Among the newborns, 71.7% (95% CI, 71.2–72.3) had the cord clamped within 1 minute and 77.6% (95% CI, 77.1–78.1) were not breast fed within 1 hour of birth. Only 3.5% (95% CI, 3.2–3.8) were kept in skin to skin contact in the delivery room after birth. The mistreatment index showed maximum variation in mistreatment among those infants born to women of relatively disadvantaged ethnic groups and infants born to women with 2 or previous births. After adjusting for hospital heterogeneity, infants born to women aged 30–34 years (β, -0.041; p value, 0.01) and infants born to women aged 35 years or more (β, -0.064; p value, 0.029) were less mistreated in reference to infants born to women aged 18 years or less. Infants born to women from the relatively disadvantaged (chhetri) ethnic groups (β, 0.077; p value, 0.000) were more likely to be mistreated than the infants born to relatively advantaged (brahmin) ethnic groups. Female newborns (β, 0.016; p value, 0.015) were more likely to be mistreated than male newborns. Conclusions The mistreatment of spontaneously breathing newborns is high in public hospitals in Nepal. Mistreatment varied by hospital, maternal ethnicity, maternal age, and sex of the newborn. Reducing mistreatment of newborns will require interventions at policy, health system, and individual level. Further, implementation studies will be required to identify effective interventions to reduce inequity and mistreatment of newborns at birth. The study has been reported as per the checklist for STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) [21]. We conducted a prospective cohort observational study in 4 public hospitals in Nepal for a period of 18 months (14 April 2017–17 October 2018). The study was nested within a larger study of 12 hospitals in Nepal that was evaluating the impact of a quality improvement package for neonatal resuscitation care on perinatal outcomes [22]. In 4 of these hospitals, observations of labour and delivery were done to measure provider performance of neonatal resuscitation and essential newborn care. The analysis of mistreatment of newborns after childbirth was not part of the original analysis; however, these observations also captured elements of mistreatment of newborns and are presented here. The protocol for the original prospective study is published elsewhere [21]. Elements of essential newborn care are measured against the “WHO standards for maternal and newborn care in health facilities” [16]. The 4 public hospitals were in Province 1, 3, 4 and 5 of the country. All the hospitals provided spontaneous vaginal, assisted vaginal, and caesarean section delivery. All 4 hospitals have specialized sick newborn care services. Each hospital had between 6,000 and 12,000 annual deliveries. Three of the hospitals were in the flat lands and one hospital was in a mountainous region. The labour unit in each hospital was led by skilled birth attendants and had access to neonatal resuscitation services at birth. The caesarean section percentage at the hospitals was between 18–27%. The number of nurses for labour and childbirth ranged from 6 to 8 in each hospital. During the study period, the intrapartum related mortality ratio ranged from 15 to 31 per 1000 births at the 4 hospitals. All vaginally-born newborns who were spontaneously breathing at birth and born to women who were admitted into the 4 hospitals for delivery were eligible to be included in the study. Women were approached at admission to the hospital for delivery. Those who consented to the study were enrolled, along with their newborns. Newborns who did not spontaneously breathe were excluded from the study, as they required different interventions and care. A data collection system was set up at each hospital and observations were conducted by independent clinical research nurses in each hospital’s admission, labour-delivery, and postnatal units. Each data collection team in each hospital had a research site coordinator on-site. Women who consented to be part of this study were tracked from admission until discharge. Consent to participate in the study was not related to consent for clinical care. Observation of labour, delivery, and immediate newborn care was assessed using an observation checklist. Socio-demographic information on women’s ethnicity, literacy, age, and previous obstetric history (parity) were collected using a semi-structured questionnaire administered at the time of discharge. After completing observations and semi-structured interviews with the participant, the forms were reviewed by the research site coordinator, on a daily basis. Any incomplete information and discrepancies were corrected by the site coordinator based on consultations with the data collector before being indexed. The data-entry assistant reviewed the forms from each hospital, coded the open-ended questions, and provided the data to the data-entry operator for entry. This is a nested study within a larger study whose aim was to evaluate a quality improvement package on perinatal care [23]. For the larger study, an estimated 80,000 women-infant pairs were required to assess the change in intrapartum related mortality [24]. No sample size calculations were required for the nested observational component. The indicators selected to assess mistreatment of newborns were based on the WHO’s 2016 “Standards for improving quality of maternal and newborn care in health facilities” quality of care statements and process of care [15]. Deviation in the process of care for newborns after childbirth as stated in the guideline was defined as mistreatment (S1 Table). Women’s ethnicity was categorized based on self-identified membership in groups according to the caste system [25] (i.e., relatively disadvantaged ethnic groups (chhetri, janjati, madeshi, dalit other disadvantaged) and relatively advantaged ethnic group (brahmin). Women’ age was categorized as 18 years or less, 19–24 years, 25–29 years, 30–34 years, and 35 years or more. Women’s literacy was categorized as those who were not literate and those were literate, based on self-report. Parity was categorized as women who had no previous births, at least one previous birth, and two or more previous births. Preterm birth was defined if the gestational age of birth was less than 37 completed weeks based on last menstrual period count. Sex of baby was defined as male or female. The distribution of the measures for mistreatment of newborns (Table 1) shows that the prevalence of some of the measures was more than 90% while for some it is less than 10%. * For the quantiles of the actual variable. Given these conditions, if the mistreatment is defined based on the condition that if any of these 9 items have “Yes” then the mistreatment prevalence will be overestimated. Also, combining variables to binary ones would lose important information. To overcome this issue, we constructed a continuous score to represent the mistreatment index using principal component analysis (PCA) of the 9 items. PCA is a dimension reduction technique used for combining many variables into one. The usual practice is to weight the variable according to the first principal component i.e. the component which has the highest variance, and thus the highest discriminatory power. This single index is similar to a single factor in factor analysis setting. The continuous score is more flexible to analyse and to model. We consider the first principal component as the proxy for the mistreatment index as it explains more than thirty percent of the total variation in the mistreatment measure. Among the potential covariates, we have missing values in the women’s literacy and sex of the newborn. We graphically checked that the items distributions are not uniform between newborns with missing observations and without missing observations (S1 and S2 Figs). The same is also found in the distribution of PCA1 (S3 Fig). The t-test for the mean PCA1 was also found to be significant for both of the variables with p-values <0.001. This is an indication that missingness is not completely random. Therefore, we have imputed the missing values in the women’s literacy and newborn sex utilizing the classification and regression tree (CART) method in the “mice” package in R. The association between the mistreatment index and the covariates was assessed using graphical tools and the uni-variate linear and non-linear regression model. Potential covariates were maternal age, literacy, ethnicity, and parity, sex of the newborn, and term or preterm status of the newborn. These are selected based on a literature review and the availability of the data. The plot is suggestive that the selected covariates have an association with the mistreatment index. A non-linear association between maternal age and the mistreatment index was observed using Loess-smoothing on a scatter plot (S4 Fig) and two knots association was seen at 21 and 24 years using the broken-stick regression model (S2 Table). To avoid this ambiguity, linear regression was done for age category to assess the association with the mistreatment index. For adjusting the effect of multiple covariates, we have fitted a multiple linear regression model including all the variables at the same time. After adjusting, the covariates still showed a significant association with the mistreatment index. This study was conducted in 4 hospitals and 24 trained data collectors observed the labour and delivery. Therefore, there is a possibility that the observations between the observers and between the hospitals are heterogeneous and those within observers and within hospitals were correlated. In this situation, the standard error could be underestimated which produces a narrow confidence interval of the estimates. This leads to a false decision against the null hypothesis. To adjust for potential heterogeneity, we have employed a random intercept model. First, we considered the hospital as random variable and employed a mixed model. Secondly, we considered observers as random variable. We also had a third model, where we considered observers nested within the hospital as random variable. In the third model, there were three different levels (birth, observer, and hospital), and we were unable to get the convergence. The outcome from the first and second model is presented in Table 3 and the outcome of the first model has been reported in the main text. All of the data analysis was conducted using R version 3.6.2 on the Linux operating system. The main study was approved by ethical committee at Nepal Health Research Council NHRC (reference number 26–2017). Written consents were obtained from all the participants included in the study prior to their participation. Written consent was obtained from a parent or legal guardian on behalf of the participants under the age of 18.