Menu
Menu
Menu
Menu
Background The burden of maternal and neonatal mortality remains persistently high in Nigeria. Sepsis contributes significantly to both maternal and newborn mortality, and safe delivery kits have long been promoted as a cost-effective intervention to ensure hygienic delivery practices and reduce sepsis. However, there is limited evidence on the effectiveness of home birth kit distribution by community health workers, and particularly the impact of this intervention on health outcomes. This paper reports a secondary analysis of data from a cluster randomized trial in rural northern Nigeria in which birth kits were distributed by community health workers to pregnant women in their homes, analyzing non-experimental variation in receipt and use of birth kits. More specifically, associations between pregnant women’s baseline characteristics and receipt and use of birth kits, and associations between birth kit use, care utilization and maternal and newborn outcomes were assessed. Methods and findings Baseline, post-birth and endline data related to 3,317 births observed over a period of three years in 72 intervention communities in Jigawa state, Nigeria, were analyzed using hierarchical, logistic regression models. In total, 140 women received birth kits, and 72 women used the kits. There were no associations between baseline demographic characteristics, health history, and knowledge and attitudes and receipt of a kit, suggesting that community health workers did not systematically target the distribution of birth kits. However, women who used the kit reported reduced odds of past pregnancy complications (OR = 0.44, 95% CI: 0.19-1.00) as well as significantly higher odds of feeling generally healthy at baseline (OR = 2.00, 95% CI: 1.06-3.76), of exposure to radio media (OR = 1.97, 95% CI: 1.21-3.22), and of perceiving themselves as having a low-risk pregnancy (OR = 3.05, 95% CI:1.39-6.68). While there were no significant associations between birth kit use and facility based delivery, skilled birth attendance or post-natal care, women who used a kit exhibited significantly lower odds of completing four or more ANC visits (adjusted OR = 0.39, 95% CI: 0.18-0.85) and significantly higher odds of reporting prolonged labor (adjusted OR = 4.75, 95% CI: 1.36-16.59), and post-partum bleeding (adjusted OR = 3.25, 95% CI: 1.11-9.52). Conclusions This evidence suggests that use of birth kits is low in a rural population characterized by minimal baseline utilization of maternal and neonatal health services, and the use of birth kits was not associated with reductions in maternal or neonatal morbidity. While further research is required to understand how the effectiveness of birth kits may be shaped by the mechanism through which women access and utilize the kits, our findings suggest that the provision of kits to women outside of the formal health system may be associated with increased risk of adverse outcomes.
This study utilizes data from a cRCT of community-based interventions designed to reduce maternal and newborn mortality conducted in Jigawa state, Nigeria. Jigawa is located in northwestern Nigeria and is characterized by extremely poor baseline health outcomes, particularly for women and neonates. The region has also been exposed to ongoing violence linked to the Boko Haram conflict. The cRCT was implemented by the Abdul Latif Jameel Poverty Action Lab (J-PAL) in partnership with the Planned Parenthood Federation of Nigeria (PPFN) and was designed to evaluate three interventions: training local women as Community Resource Persons (CoRPs) to provide health education to pregnant women and their families; deploying the CoRPs in conjunction with the distribution of safe birth kits to pregnant women; and deploying the CoRPs in conjunction with community dramas to change social norms around maternal health. This paper does not report the primary experimental effects of the interventions; these effects are reported in a separate, forthcoming paper. Ethical approvals for the cRCT analyzed here were provided by the Massachusetts Institute of Technology (MIT) and the Operational Research Committee (ORAC) of the Ministry of Health in Jigawa state, Nigeria. ORAC serves as the institutional review board for human subject research conducted within Jigawa state. The cRCT protocol is registered at ClinicalTrials.gov under the registration number {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01487707″,”term_id”:”NCT01487707″}}NCT01487707. All data collection was conducted in Jigawa, a region with low female literacy. All respondents provided written consent using paper consent forms, and the consent process was documented in the electronic data system by enumerators. In addition, consent was sought from the parents or guardians of any non-emancipated minors included in the sample. These consent processes were approved by the relevant Institutional Review Boards. Jigawa state had a population of 4.3 million during the 2006 census, and includes 27 local government areas (LGAs); 80% of the population lives in rural areas [22]. The state is characterized by generally low rates of maternal health services utilization and poor baseline health outcomes, including the third lowest rate of facility based delivery in the country (6.7% vs 35.5% nationally), and the fourth lowest percentage of fully vaccinated children 12-23 months of age (4% vs 25% nationally) [3]. In response to the observed low rates of facility delivery, the Nigerian government rolled out the Midwives Service Scheme (MSS) in 2009, recruiting midwives to be deployed to government primary health centers (PHCs) in rural communities to provide 24-hour maternity care [23]. In northern Nigeria, key stakeholders in the health sector hypothesized that the expansion of the MSS alone might not lead to enhancement of maternal and child health outcomes given the observed low rates of utilization of maternal health care ex ante. Accordingly, our partner organization, the Planned Parenthood Federation of Nigeria (PPFN), proposed to implement three community-based interventions involving health educators deemed community resource persons or CoRPs, designed to stimulate utilization of these newly available maternal health services and thus enhance maternal and neonatal health [24]. These interventions were evaluated using a cluster randomized controlled trial with four experimental arms. In the first experimental arm, CoRPs services were rolled out to sampled villages; CoRPs are local women between 20 and 45 years of age recruited by PPFN in conjunction with local ward health committees. They received a one-week training and were mandated to provide information on antenatal care, nutrition in pregnancy, identification of danger signs, birth preparedness, labor and postnatal care, breastfeeding, immunization, birth registration, and family planning over a series of six home visits to pregnant women. PPFN described CoRPs as serving as “a bridge crucial to foster trust, confidence and acceptance between the midwives and their clients and to ensure effective communication between the two groups”. The CORPs also received a small stipend (2000 naira a month, or approximately $5). In the second experimental arm, CoRPs were provided with birth kits to distribute to pregnant women in their third trimester, in conjunction with the provision of instructions around how to use the kit. The kits included a plastic sheet for the woman to lie on during delivery, surgical gloves for a birth attendant to utilize, a sterile razor and cord clamps to cut and tie the umbilical cord, methylated spirit to clean the umbilical stump, clean gauze, swabs and perineal pads to be used by the mother after birth, a gallipot, a mechanical suction tube to clear secretions from the baby’s airways, and a wrapper and diapers for the use of the mother and baby immediately after birth [25]. All materials were packaged in a single sterile unit. In the third experimental arm, in addition to the CoRPs intervention, PPFN also conducted community drama activities in order to promote the importance of safe motherhood at the community level. The fourth experimental arm served as the control arm, in which PPFN did not provide any new services. Women in control communities continued to have access to standard health services provided at MSS clinics. The RCT design entails randomization at the community level, given that all three interventions are subject to significant within-community spillovers. In total, 96 clusters in 24 local government areas (LGAs) were included. All clusters were located within 20 kilometers of a MSS PHC, and each cluster included 75 respondents enrolled at baseline. In addition, neonatal deaths were tracked for the full cluster population (approximately 3,000 individuals per cluster). Baseline power calculations estimated the cRCT had 90% power to detect a 24% decrease in neonatal mortality and a 23% decrease in maternal morbidity using a one-tailed test (α = .1, K = .2); this is assuming a baseline neonatal mortality rate of 47 per 1,000 and a baseline maternal morbidity rate of 35%, and attrition of less than 5% [26]. Attrition was assumed to be low given that data would be collected about pregnancies at two points in time: shortly after the birth, and again at endline. Given limited ex ante evidence on intracluster correlation for the outcome variables of interest in the Nigerian context, the power calculations drew on published estimates of k for perinatal indicators in other developing countries as well as methodological guidance for sample size calculation in cluster randomized trials [27, 28]. The main trial included three phases of data collection: baseline, continuous data collection, and endline. Data were collected by a team of trained female Hausa-speaking enumerators, recruited from Jigawa state; at baseline, the team also included some individuals recruited from neighboring states. Enumerators received a minimum of two weeks training focusing on survey administration, ethics, and techniques of anthropometric measurement. Enumerators were supervised by a team of four on-the-field supervisors who were responsible for overseeing data quality by directly observing enumerators and conducting backcheck surveys. Data were collected on smart phones using SurveyCTO, and surveys were administered in Hausa. The baseline survey (N = 7,069) was conducted between December 2011 and May 2012 using a 15% random sample of all enumerated households including a woman of reproductive age (between 15 and 49); if more than one woman of reproductive age was present in a sampled household, one woman was randomly selected utilizing an on-the-field randomization protocol. The baseline survey included information about household composition and socioeconomic characteristics, birth history, utilization of health services for pregnancies within the preceding 24 months, contraceptive utilization, the respondent’s health and the health of any infant children, and health knowledge and attitudes. In addition, anthropometric measurements were collected for the respondent’s children under the age of two (i.e., children born during the intervention period). Following the baseline, continuous monitoring of pregnancies via a RapidSMS surveillance system was initiated between November 2013 and November 2015. Female community members were recruited and trained to monitor pregnancies and vital events among baseline respondents, and were assigned to send a simple SMS to the survey team when a birth was observed in baseline households, and to report deaths of women and infants in any household. The SMS messages were then redirected to an enumerator who had the responsibility of identifying the household and administering questionnaires within 3 days and at 28 days after birth. For deaths of women of reproductive age, verbal autopsies were conducted to determine the cause of death. The three-day survey included questions about utilization of antenatal care, the mother’s health during pregnancy, the delivery itself, and the mother’s and infant’s health since birth; the infant’s weight and length were also measured. The 28-day survey included questions about attitudes toward utilization of maternal health care, maternal and neonatal morbidity in the first month, and infant health practices; again, infant weight, length and mid-upper arm circumference (MUAC) were measured. In addition, an audit of all baseline households was conducted at the halfway point of continuous data collection in which all baseline households were revisited. The enumerators were mandated to pose a brief series of questions about births in the household, and to collect additional information from any births that had previously been missed. An endline survey with the baseline sample was conducted between February 2016 and July 2016. Attrition in the endline was 9.8%. However, some women who were not surveyed at endline due to migration or divorce had been represented in earlier surveys (post-birth surveys or the audit). Accordingly, the rate of attrition from any follow-up data collection was only 7.8% of the original sample. This paper reports summary statistics on receipt and use of birth kits in the target population, and uses hierarchical logistic regression to analyze associations between baseline characteristics of respondents and receipt and use of birth kits, as well as associations between use of birth kits and outcomes during pregnancy and delivery. The analysis focuses on the sample “as treated” and is complementary to an intention-to-treat analysis analyzing experimental evidence around the primary effects of the intervention, reported in a separate, forthcoming paper. More specifically, an “as treated” analysis is utilized here given that an intention-to-treat analysis cannot provide evidence about the correlations between birth kit use and health outcomes. Receipt and use of birth kits varied at the individual level even within clusters assigned to receive birth kits. Accordingly, individual-level variation rather than cluster-level variation was employed, using data from women who reported a birth during the evaluation period in any of the three intervention arms; this was a sample of 3,317 women. For the analysis of birth kit use, the sample was restricted to women who reported receipt of a birth kit from a CoRP. Given that there is presumably correlation in delivery outcomes within clusters, the analysis accounted for this intra-cluster correlation by estimating logistic regressions including clustered standard errors. Data were analyzed using Stata 14.2 (Stata Corp, College Station, TX), and the Stata command “logit” was employed, using the “cl” option. More specifically, logit models were estimated iteratively employing maximum likelihood estimation, and sandwich standard errors were employed to conduct inference given an arbitrary structure of within-cluster correlation of errors [29]. Binary variables capturing birth kit receipt and use were constructed as follows: “birth kit receipt” was defined as equal to one if the respondent reported receipt of a birth kit from a PPFN CoRP. “Birth kit use” was defined as equal to one if the respondent reported using the birth kit during her delivery. The baseline characteristics assessed included demographic characteristics, health history, utilization of health services, health knowledge, knowledge of pregnancy-related complications, perceptions of risk in pregnancy and childbirth, and household dynamics, as described in more detail below. Demographic characteristics included variables capturing age, ethnicity, household structure, educational background, birth parity, and a wealth index. Birth parity was defined as the number of births (including stillbirths) reported by the respondent; it was converted to a categorical variable for parity zero, parity one, and parity 2+. The household wealth index was computed by principal component analysis from four binary variables capturing whether the household reports a solid roof, a solid floor, access to a latrine and a house constructed from solid materials. Variables capturing baseline health history and health care utilization included the respondent’s history of miscarriage, stillbirths, infant deaths, and pregnancy complications, and antenatal care utilization, skilled attendance at birth, and utilization of postnatal care as reported for her last birth. Additional variables captured baseline health knowledge, including radio exposure, coded as equal to one for women who report listening to the radio at least once a week, and awareness of tuberculosis, HIV/AIDS, and the importance of exclusive breastfeeding. Variables capturing baseline knowledge of pregnancy-specific complications included the number of labor and delivery complications, danger signs, and postpartum complications identified by the respondent. The respondent’s perceptions of risk in pregnancy and delivery were measured using the following variables: whether she believes maternal death to be preventable, whether she is confident she will not encounter challenges in delivery, and three binary variables representing whether the respondent correctly identified the riskier scenario in a series of questions posed about pregnancy, delivery, and the post-partum period. Based on the number of correct responses, women were assigned to one of three categories of risk knowledge (low, medium and high knowledge, corresponding to one, two, and three correct responses, respectively). Finally, baseline household dynamics were captured by the husband’s educational and occupational characteristics and household decision-making patterns. It is important to note that the sample for some regressions is more limited if certain variables are not reported at baseline; in particular, respondents who had not reported a birth in the 24 months prior to baseline do not report baseline care utilization. In the final part of the analysis, associations between birth kit use and pregnancy outcomes (health care utilization, health practices, neonatal and maternal health outcomes, and under-two anthropometric measurements) were analyzed. The variables of interest are defined in Table 1. Pregnancy outcomes were selected based on a review of the relevant literature on the impact of birth kits and characteristics predictive of maternal and neonatal care utilization. Care utilization variables included whether the respondent utilized antenatal care and additional questions describing her care utilization pattern (if more than four antenatal visits were conducted, if ANC was initiated in the first trimester, if the respondent received certain ANC services including a tetanus vaccine and iron folic pills, and if she received ANC at a PHC or hospital), whether the respondent delivered in a facility, whether she delivered at home (alone or accompanied), whether a skilled attendant was present at the birth, and whether she utilized postnatal care. Variables capturing health practices included whether the respondent developed a birth plan, whether the husband was present at ANC and/or delivery, and whether complementary feeding was initiated in the first three days. Variables capturing maternal and neonatal morbidity included a series of questions around whether the respondent experienced enumerated symptoms during pregnancy, delivery or post-partum, and whether the respondent reported symptoms of illness for the infant in the first 60 days of life. In addition, three binary variables were constructed capturing whether the respondent’s children under two at endline are underweight, stunted or characterized by low MUAC-for-age, defined as weight-for-age, height-for-age, or MUAC-for-age more than two standard deviations below the mean of the WHO reference population. For these variables, we preferentially utilized data reported in the surveys conducted three and 28 days after birth; if these surveys were missing, we drew on parallel data reported in the audit survey or the endline. Given the long follow-up period and the challenges posed by data collection in a remote and conflict-affected region, we benefited from utilizing these various complementary sources of data to obtain maximum information about our sample of interest. Again, this analysis was restricted to women who reported they received a birth kit. In these specifications, we also considered variables that may confound or modify these associations. The confounding variables included were dummy variables for age categories, Hausa ethnicity, marital status, polygamous status, ever attended school, literacy, birth parity status (parity zero, parity one, and parity two or higher), and assignment to the birth kits arm. The specifications examining maternal and neonatal morbidity and under-two anthropometrics also adjusted for variables capturing utilization of maternal health care: whether the respondent utilized antenatal care, whether she delivered in a facility, and whether the delivery was attended by a skilled provider. The confounding variables were chosen based on a review of the literature around demographic variables predictive of birth kit utilization and maternal health outcomes.
N/A
