Introduction: maternal and neonatal health status indicators have steadily improved over time in Kenya. Significant challenges remain, including persistent inequities between population subgroups, and a health system that delivers variable quality care and inconsistent access to care. This paper highlights results of an ex-post evaluation to assess the impact of maternal and health systems strengthening intervention to improve newborn health outcomes in Bungoma County, Kenya, focusing on access to and quality of maternal and neonatal care. Methods: the study design was quasi-experimental, using household surveys to assess outcomes at baseline and end-line. Stratified cluster sampling was used to identify households, based on heath facility catchment areas. Inclusion criteria were women aged 18-49. Chi-square and fisher´s exact tests were used. Patched-up design was used to compare outcomes before and after the intervention and intervention and control sub-counties. Results: provision of transport vouchers decreased barriers to accessto health care, resulting in an increased number of deliveries in health facilities. Women in the end-line group were 95% more likely to deliver at a health facility compared to 74% at baseline. The intervention improved potential and effective access to antenatal care as well as deliveries in health facilities. This positively impacted quality of care provision in the sub-counties. Conclusion: key elements of health system strengthening included reducing cost barriers and enhancing the capacity of the health facilities to deliver high quality care. The intervention addressed commonly identified supply-and demand-side barriers, providing stronger evidence that addressing these hindrances would improve utilization of maternal and child health services.
Study setting: Bungoma County is located in Western Kenya and has a population of approximately 1.5 million. The county comprises of ten sub-counties. Its per capita gross domestic product (GDP) is half that of the national average [9]. Bungoma lagged behind the national trend in key maternal health indicators. For example, with a Skilled Birth Attendant (SBA) rate of 40.8%, the demographic health survey 2014/15, ranked Bungoma 43rd out of 47 counties in Kenya [3]. Women in Bungoma face significant social, economic and cultural challenges when accessing maternal and new-born health care. Most women stay at home to give birth, mostly as a result of lack of transport/money, lack of education, poor conditions at the health facility, and culture (i.e. importance of burying the placenta). High maternal and neonatal deaths were observed in the community, and poor health of newborns [8]. Bungoma presents a useful case for examining the effect of health systems strengthening on maternal and newborn health care given the poor maternal and child health indicators. Study population: study population consisted of health facilities in Bungoma County, mothers from the catchment populations of the health facilities included in the survey who had delivered a baby during 12 months prior to the survey. Surveys were carried out by a team of researchers that was independent from the team involved in the intervention. Study design: an ex-post evaluation was conducted when the program had ended. The design had to be adapted to the available data at hand. The classic pre-post control group design requires baseline and end-line data for both intervention and control groups; the implementation of MANI program did not allow for this to be used [10]. MANI programme was implemented in six sub-counties (Kabuchai, Kanduyi, Sirisia, Tongaren, Webuye East and Webuye West); the remaining four sub-counties (Bumula, Mt. Elgon, Kimilili and Cheptais) were controls (Figure 1). At this stage, the design could be a static group comparison design (also known as the non-equivalent control group post-test only design) [11]. As selection is a major threat to the validity of this design, we attempted to control for it by using matching, as described below. In order to generate a sampling for the end-line surveys, eight health facilities in intervention sub-counties were matched with eight health facilities in the control sub-counties based on ownership, level, type and service provision characteristics. The service areas of each health facility were mapped using census data, and stratified cluster sampling was employed to identify households for interview. Respondent inclusion criteria were having delivered a live infant, a stillbirth or a third trimester spontaneous abortion within 12 months of the survey. A total of 448 women aged 18 – 49 years were interviewed in each of the two groups. map of Bungoma County, highlighting health facilities and sub-counties supported by the MANI project and Save the Children International Availability of a pre-existing survey in the intervention areas allowed us to add a non-equivalent historical control for the intervention counties, which approximated a pre-test post-test design. Sample size for this baseline survey (conducted in 2015) was 478. Baseline survey questionnaire examined variables that assessed access utilization and quality of MNH services in intervention areas of the study. Questionnaires for both baseline and end-line surveys had been developed by adapting the demographic and health survey questionnaires to the indicators relevant to the MNH program and thus were comparable. In addition to demographics, questions focused on antenatal care attendance, delivery by skilled providers, access to care and perception of service quality and level of satisfaction with MNH services. Questionnaires were translated into Swahili and back translated into English for comparison with the original English version, to ensure accuracy of translation. The Swahili version was pretested in a population similar to the study population in a neighbouring sub county not involved in the survey. Final design is depicted in Table 2 and described as a “patched up” design (10-17): O1 is the historical control (i.e. the pre-test, or the 2015 baseline) in the intervention counties, O2 is the post-test in the intervention counties, and O3 is the non-equivalent control. The ‘X´ depicts the intervention, and non-random (NR) signifies non-randomization [12]. It is important to note that the three groups (O1, O2 and O3) represent different samples. As recommended by Campbell, analysis of this design is based on comparing outcomes in the three groups. If O2 > O3 and O2 > O1 and both differences are statistically significant, we can state that there is an association of the intervention with the outcome [13]. As per Habicht et al. such congruency (of expected trends) classifies this design as ‘plausible’ (in their scale of adequate, plausible and probabilistic designs) [14]. “patched up” study design NR: non-random assignment; X: intervention; row 3 and 4 represent the intervention counties; row 5 represents the control counties Variables: variables collected during surveys were grouped into five domains: a) mother´s potential access; b) mother´s actual access; c) baby´s actual access; d) quality of care received by the mother; and e) quality of care received by the baby. Potential access (mother): costs are a recognized barrier to health care access, and interventions at the national level (elimination of user fees under the Linda Mama Scheme) and local levels (provision of transport vouchers as part of the intervention) aimed to decrease, if not eliminate, this barrier. The voucher scheme covered return transportation to the health facilities for the mother and one companion, which was provided by boda-boda (motorcycle taxi) drivers affiliated to the programme, thus reducing financial barriers to care. We assessed expenditures for antenatal care (ANC) services, transport to health facility, and delivery services as binary variables (zero, some expenditure). Actual access (mother): the goal of interventions was to increase utilization of health services or improve the effective (i.e. realized) access. Maternal access to services was captured by asking women whether they had attended any ANC clinic during their last pregnancy, whether they had had at least 4 ANC contacts during the last pregnancy, whether an ANC contact had occurred in the first trimester, whether she had delivered in a health facility and whether she had received post-natal care (PNC) services within 48 hours of the delivery. We dichotomized all variables (no/yes). Actual access (baby): was evaluated with a binary variable (no/yes) whether PNC was provided within 48 hours of birth. Quality of care (mother): mothers were asked if they had delivered via C-section, and whether they were satisfied with the ANC care received. These variables were binary (no/yes). Quality of ANC was assessed by asking women about having received 14 clinical interventions or provider advice as part of service; ranging from measuring blood pressure to advice on nutrition. Thirteen or more positive answers were coded as high quality, 7 to 12 were coded as medium and six of less was coded as poor quality. Quality of PNC to the mother was assessed by whether they had received nine clinical interventions or provider advice as part of first PNC examination. These ranged from measuring blood pressure to advice about breastfeeding. Depending on the number of positive responses, this was coded as high (7 or more), medium (5-6) or low (4 or less). Quality of care (baby): was captured with two variables. A dichotomous variable (no/yes) captured whether breastfeeding was initiated in the first hour. Quality of PNC provided to the child was assessed by asking mothers whether the child had received six clinical interventions or provider advice as part of the first PNC examination, ranging from physical examination to information about signs and symptoms that suggested that the baby was unwell. Six positive answers were coded as high quality, four and five as medium quality and three or less as low quality. All variables were categorical, we used Pearson´s chi-square or in cases of small numbers, fisher´s exact tests. Study participants provided written assent to participate. Study protocols were reviewed and approved by Nairobi University and Kenyatta National Hospital Ethics Review Boards. Ethical approval certificate Ref KNH -ERC/A/27.
N/A