Menu
Menu
Menu
Menu
Objective: Although substantial progress has been made in increasing access to care during childbirth, reductions in maternal and neonatal mortality have been slower. Poor-quality care may be to blame. In this study, we measure the quality of labour and delivery services in Kenya and Malawi using data from observations of deliveries and explore factors associated with levels of competent and respectful care. Methods: We used data from nationally representative health facility assessment surveys. A total of 1100 deliveries in 392 facilities across Kenya and Malawi were observed and quality was assessed using two indices: the quality of the process of intrapartum and immediate postpartum care (QoPIIPC) index and a previously validated index of respectful maternity care. Data from standardised observations of care were analysed using descriptive statistics and multivariable random-intercept regression models to examine factors associated with variation in quality of care. We also quantified the variance in quality explained by each domain of covariates (patient-, provider- and facility-level and subnational divisions). Results: Only 61–66% of basic elements of competent and respectful care were performed. In adjusted models, better-staffed facilities, private hospitals and morning deliveries were associated with higher levels of competent and respectful care. In Malawi, younger, primipara and HIV-positive women received higher-quality care. Quality also differed substantially across regions in Kenya, with a 25 percentage-point gap between Nairobi and the Coast region. Quality was also higher in higher-volume facilities and those with caesarean section capacity. Most of the explained variance in quality was due to regions in Kenya and to facility, and patient-level characteristics in Malawi. Conclusions: Our findings suggest considerable scope for improvement in quality. Increasing staffing and shifting births to higher-volume facilities – along with promotion of respectful care in these facilities – should be considered in sub-Saharan Africa to improve outcomes for mothers and newborns.
We used data from the Service Provision Assessment (SPA) surveys conducted in Kenya in 2010 and in Malawi in 2013–2014. The SPA surveys, developed by the Demographic and Health Surveys Program, have been conducted in several LMICs since the 1990s. The goal of the SPA survey is to provide a comprehensive overview of health service delivery in a country. The SPA surveys include four instruments: a facility audit, interviews with health providers, direct observations of care (family planning consultations, antenatal care, labour and delivery care and sick child care) and exit interviews with patients. We included Kenya and Malawi because they were the only two countries where the SPA survey conducted observations of labour and delivery services. In Kenya, the SPA survey used a randomly selected nationally representative sample of all health facilities. All three national referral hospitals and all eight provincial hospitals in Kenya were included 8. In Malawi, the survey was based on a census of all public health facilities and large private facilities, and on a representative sample of small private facilities 9. In each facility, delivery clients were selected for observation based on the number of women present on the day of the survey. The rule was to observe a maximum of five delivery clients for each provider, with a maximum of 15 deliveries per facility. We used two previously validated indices of competent and respectful care to measure quality of labour and delivery care services. Tripathi and colleagues 10 developed an index to assess the quality of the process of intrapartum and immediate postpartum care (QoPIIPC) in facility deliveries. This index is based on 20 process of care indicators available in the SPA related to the initial assessment and examination of the patient, the management of the first, second and third stages of labour and immediate newborn and postpartum care (Figure S1). In this study, we estimate care competence using the QoPIIPC index. We also used the respectful maternity care index developed and validated by the Maternal and Child Health Integrated Program (MCHIP) 11. This index was based on nine indicators of provider–client interactions reflecting actions the provider should take to ensure the client is informed and able to make choices about her care, and that her dignity and privacy are respected (Figure S1). In both countries, the labour and delivery observation checklist was divided into four sections: (i) initial client assessment, (ii) care during first stage of labour, (iii) care during birth and (iv) immediate newborn and postpartum care. For several women, the full labour and delivery process could not be observed because the observation began when labour had already started, or because the woman was referred to another facility during the first stage of labour (most of them for caesareans). In both countries, 8–9% of babies needed resuscitation at birth so routine postpartum care was not observed. This systematic missingness precluded the use of multiple imputation. Rather, we decided to calculate the quality index based on the sections observed for each woman. The denominator therefore varied across women. Nonetheless, 73% of women (400 in each country) were observed for all four sections of the labour and delivery process and had no missing data for the 29 quality indicators. As a sensitivity test, we repeated the analyses in this subsample. We explored potential determinants of competent and respectful care at the patient, provider and facility levels. The covariates considered for inclusion were identified based on prior research suggesting that they may influence quality of care and provider behaviour 1, 12. Availability of these covariates differed between the two countries. In Malawi, several characteristics of the delivering women were available including her age, time of delivery, whether she was HIV positive and whether she was giving birth for the first time. In Kenya, only the time of day during which the delivery took place was available at the patient level. In both countries, provider‐level covariates included gender and cadre and facility‐level covariates included the facility type, whether the facility had the capacity to perform caesarean sections, the ratio of full‐time clinical health professionals (medical and nursing) per maternity bed and the annual volume of deliveries. An indicator for urban location was also available in Malawi. Finally, in both countries, we also included indicator variables for subnational divisions as defined by the SPA surveys: eight regions in Kenya and five zones in Malawi. All covariates were included as binary or categorical variables for better interpretability. Health worker cadres were grouped into two categories based on years of training for country‐specific cadres. Higher cadres included MDs, clinical technicians, medical assistants, BScN and registered nurses and BScN and registered midwives. Lower cadres included enrolled nurses, enrolled midwives, community health nurses and nurse aides. The thresholds for categories of annual volume of deliveries (1500) were selected to reflect international thresholds for high and low volume facilities. In Kenya, annual volume of deliveries was reported in the survey. Because this variable was not available in the Malawi survey, we estimated annual delivery volume by multiplying the number of delivery clients present in the facility on the day of the survey by 365. Finally, the ratio of clinical staff per maternity bed was divided into quintiles and included in the analysis as a binary indicator comparing the top quintile to all other facilities. We first explored differences in quality across levels of the covariates by performing pairwise comparisons of means, using the Bonferroni method to adjust for multiple comparisons for categorical indicators. We then constructed multivariable two‐level random‐intercept regression models, with patients nested within providers, and standard errors clustered by facility. All covariates were included in the multivariable models for the exception of caesarean section capacity and annual delivery volumes which were strongly collinear with facility types. To quantify the variance explained by each domain of covariates (patient, provider, facility and subnational divisions), we progressively added blocks of variables to the multilevel models. We calculated the percentage of variation in quality explained by the group of covariates as the difference in variance between the adjusted model and the null model divided by the null model variance. All regression analyses were performed separately in each country and were not adjusted for sampling weights. The SPA survey used the same methods for observations of care in Kenya and Malawi. The quality indices were therefore measured identically in both countries. However, other questionnaires differed and certain characteristics of women and facilities were only available in one of the two countries. We therefore opted to conduct regression analyses separately by country. However, as a sensitivity analysis, we repeated the regression by pooling data and including covariates available in both countries. We conducted two additional sensitivity analyses. First, we conducted the analyses in the subsample of 800 women with complete data. Second, we performed the regression analyses using patient‐level sampling weights. All statistical analyses were performed using Stata version 14.2 (Stata Corp, College Station, United States of America). This study was funded by the Bill and Melinda Gates foundation. The Harvard T.H. Chan School of Public Health institutional review board approved this study as exempt from full review. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
