Effects of continuum of care in maternal health services on adverse birth outcome in Northwest Ethiopia: a prospective follow up study


 Background

Globally, around 4 million of babies were died within the first month and more than 3 million were stillbirth. Of them, 99% of newborn deaths and 98% of stillbirth occur in developing countries. Despite give a priority for maternal health services, adverse birth outcomes are still major public health problems and big challenge in Ethiopia particularly in study area. Hence, to overcome those challenges, a continuum of care (COC) is core key strategies. Despite that study conduct on effectiveness of COC in maternal health services was scare in developing countries and not done in study area. We aimed to assess effectiveness of continuity of care on reduction of adverse events and determinants of adverse birth outcomes.
Methods

Community and health facility linked prospective follow-up study design was employed from March 2020 to January 2021 in Northwest Ethiopia. A multistage sampling technique was used to recruit 2,198 pregnant women. Data were collected by using semi-structured and pretested interviewer questionnaire. Collected data were coded, entered, cleaned and analyzed by using STATA 14. Mixed-effect model was used to identify community and individual level factors of adverse birth outcomes. Finally, propensity score matching (PSM) was applied to determine effectiveness of COC in maternal health services.
Results

Magnitude of adverse birth outcome was 26.8%: stillbirth (2.7%), neonatal mortality rate (3.1%) and neonatal morbidity (6.8%). Risk factors significantly associated were being poor household wealth index (AOR = 3.3), pregnant related complication during pregnancy (AOR = 3.29), labour and childbirth (AOR = 6.08), immediately after childbirth (AOR = 5.24), offensive odder amniotic fluid (AOR = 3.04) and history of stillbirth (AOR = 4.2). Whereas, women receive IFA during pregnancy (AOR = 0.44), initiate BF within 1hour (AOR = 0.22) and immunizing newborn within postnatal period (AOR = 0.33) were protective factors. As treatment effect, completion of COC in MHS via time dimension (β = -0.03), completion of key services of MHS (β = -0.04) and completion of COC in MHS via space dimension (β = -0.02) were associated with a significant reduction in the likelihood of perinatal death
Conclusions

The results suggested that neonatal and perinatal mortality rate were high as compared with national and international target. Completion of COC in MHS is an effective strategy in reduction perinatal death. Efforts should be made to strengthening COC of MHS and also encourage IFA supplementation, immunizing and early initiation of BF for newborn were strongly recommended.

Efforts should be made to strengthening COC of MHS and also encourage IFA supplementation, immunizing and early initiation of BF for newborn were strongly recommended.

Background
Women play a signi cant role in rearing of children and managing family affairs but their loss from maternity-related causes and neonatal mortality are a signi cant social, economic and personal disaster (1). Globally, around 4 million of babies were died within the rst month and more than 3 million of births were stillbirth. Hence, 99% of newborn deaths occur in developing countries (2). It is a big challenge for the world population and still public health problem particularly in the developing and lowincome countries (3). In Africa, annually 1million stillbirths were occurred, of them 300,000 cases were occurred during labour and also 1million babies were died within rst month, of them half million died within 1st day. In Sub-Sahara Africa, under ve child mortality rate was 16 times more likely to die than in high income countries due to poor access to quality of services and discontinuation of care in maternal health services (2,4). Having this in mind, those deaths were peak during the times of birth, the rst day and the rst week of life.
The health need of mother and newborn are inseparably linked together. In the past decay: maternal, newborn and child health policy and programs address their health need separately; this gap and challenge in care provision especially newborn health are subjected to disadvantaged. Even though, to overcome those challenges, a continuum of care (CoC) in maternal health services is one of the core key strategies. Thus, it reaches mothers and babies at a crucial time and place to reduce the risk of death, illness or disability of neonatal (5). Health of newborn relies on good linkage of maternal health intervention because newborn health is a sensitive indicator of continuity of maternal health services (5). Despite these, almost half of African women were not receiving skilled care during childbirth and postnatal care services after delivery (4).
While utilization rate of essential maternal health services reach 90%, neonatal death reduced by 37 -67% (430,000 -800,000) of deaths (4). Hence, continuum of care in maternal health services via time dimension could reduce neonatal mortality: completion of continuity of Antenatal (ANC) services could reduce by 7 -14%; continuity of care at child birth could reduce by 19 -34% and continuity of postnatal (PNC) services could reduce 10 -27% of neonatal deaths. Similarly, space dimension continuity of care could avert neonatal death: continuity of care at family/community level interventions (14 -32%); outpatient/outreach interventions (7 -14%) and clinical setting interventions (26 -51%) of neonatal death could avert (5). A variety of evidence revealed that completion of continuity of care in maternal health services via time and space dimension were more likely reduce adverse birth outcomes (stillbirth, neonatal mortality and morbidity) (4,(6)(7)(8)(9)(10)(11)(12).
Despite give a priority for maternal health services, adverse birth outcome (stillbirth, neonatal death and illness) are still major public health problems in Ethiopia particularly in Benishangul Gumuz Region. However, majority of literature suggested evidences based on cross sectional study design and facility based survey which couldn't show a cause and effect relationship and also using traditional binary logistic and linear regression couldn't explore the treatment effect of completion of continuity of maternal health services on adverse birth outcome. Moreover, relies on cross sectional and facility survey, it is di cult to organize clear epidemiological evidence on treatment effect. In general, clear epidemiological data on adverse birth outcome (stillbirth, neonatal death and morbidity) were generated using prospective follow up study design, which found to be an important evidence for implementation of maternal health services in Ethiopia.
Therefore, the objective of this study was to investigate effect of continuum of care in maternal health services on perinatal death and also determinant factors that hindering adverse birth outcome using propensity score matching (PSM) and multilevel regression model respectively.

Study design and setting
Community and health facility linked prospective follow-up study design was conducted in Benishangul-Gumuz Regional State (BGRS) from March 2020 to January 2021. The region is one of the eleven states of Ethiopia's Federal Democratic Republic of Ethiopia. Assosa town is the capital city of the region, located 670kms West of Addis Ababa, the capital city of Ethiopia. The region has three zones, three town administrative cities, twenty one districts/Woredas, one special district/Woreda's and 475 clusters/ Kebele's (439 rural and 36 urban clusters/ Kebele's). The region represents around 4.6% of the total land area of Ethiopia and most of the people in the region are sparsely populated (13).

Source population and study participants
All pregnant women and birth that registered as live birth as well as stillbirth at the time of birth within the follow up period were considered as source population. Whereas, the study population was newborns that registered as "live birth" or "stillbirth" (which is declared by women, birth attendants or health workers) at the time of birth and selected by simple sampling techniques.

Sample size and sampling technique
The sample size for this study was calculated by using STATA/MP 13.0 software by considering two population proportion formulas based on the following assumption. The outcome variable adverse birth outcome (stillbirth, neonatal death and any illness within neonatal period), and the predictor variable was continuum of care in maternal health service. In Ethiopia, literatures that support to determine sample size for this objective are not available. Like as Ethiopia particularly in Benishangul Gumuz Region; in rural India, most births take place in the home, where high-risk practices are common (10). Unfortunately, neonatal mortality in Uttar Pradesh, India (29 deaths per 1,000live birth: Urban 20 verses Rural 38 deaths per 1,000live birth)(14) is almost similar with Ethiopia neonatal mortality (29 per 1,000live birth) (15), so that evidence stated in Uttar Pradesh, India is used for sample size determination. Study done in Uttar Pradesh, India was found that proportion of adverse birth outcome, "neonatal death", among whose mother use complete continuum of care in maternal health services is 4.29% ((P 1 = 0.0429) and proportion of adverse birth outcome, "neonatal death", among whose mother never use ANC service is 8.43% (P 2 = 0.0843) (10). A 95% con dence level and 80% power were used to detect 9.8% difference or 54.4% increment among exposed and non-exposed group. Hence, r=ratio of exposure to non-exposure pregnant women equal to 1:1 for the population allocation ratio; P (pooled population proportion) was calculated (P = 0.0636); considering a design effect 2 and a non-response rate of 10%. As result, the nal sample size was 2,402 pregnant women had been follow up to look the effect of continuity of maternal health services on perinatal death, used for as nal sample size for this study.
Since this research work was carried out at a regional level, the study subjects (pregnant women) were chosen using a multistage sampling technique. The sampling procedure used for this study was as follow: primarily two zones and one town administration were chosen by simple random sampling (SRS).
Following that, four districts/"woredas" from the Assosa Zone, two districts/"woredas" from the Metekel Zone, and two districts/"woredas" from the Assosa town administration were chosen by simple random sampling (SRS) technique. Thirdly, from each selected district/"woreda", seven kebeles from each district (except Assosa district/"woreda": 10 kebeles were selected) and ve ketenes from each district/"woreda" of town administration were selected and included in the study. Then after, among the selected kebeles/ketenas (7 kebeles from each districts/"woredas", 10 kebeles from Assosa districts/"woredas" and ve ketenes from each districts/"woredas" of town administration), pregnant women were enumerated by using house-to house visit and all obtained and registered pregnant women were included in the study. All women who claimed an 8 weeks or longer pregnancy, as determined by the loss of two consecutive menses, were considered eligible and joined in the study, which followed for 11 months.
Assuming that each household with pregnant women had at least one pregnant woman, households with pregnant women and neonates were selected as the nal sampling unit (FSU). Similarly, all health facilities found within the catchment areas were list and considered as a candidate for health facility survey. Therefore, 46 health facilities (3 hospitals, 12 health centers and 31 health posts) were found within the catchment areas and included for health facility survey.

Data collection process
Data collection was conducted using semi-structured questionnaires and registration format adopted from EDHS 2016 (15), National Technical Guidance for MPDSR 2017(16), MCH Program Indictor Survey 2013 (17), Survey tools conduct in Jimma Zone, Southwest Ethiopia(18), Survey tools conducted in Rural South Ethiopia (19) and other relevant different literatures. The instrument was prepared in English and translated into local language ("Amharic") and then back translated to English to ensure validity of the instrument. Following that, training was offered for data collectors and supervisors for 3days and also pre-test was carried out on 35 individuals, located outside of the study areas/cluster. During actual data collection, principal investigator and supervisors were frequently supervised and checked the work of data collectors and also clari cation and direction was forwarded to those who had doubts.

Data management and analysis
Data was coded and entered into Epi Info version 7.2.2.6 to develop skipping patterns and avoid logical mistakes. The data was then cleaned, edited, and analyzed using STATA Software version 14. All variables were computed for descriptive statistics. Analysis with only one variable were performed, the crude odds ratio and 95% con dence interval were used to select candidate variables for multivariate analysis (p < 0.25). At the level of signi cance (p < 0.05), a maximum likelihood estimate of the independent effect on the outcome variable was calculated. The household wealth index was calculated and categorized by using Principal Component Analysis (PCA). Before running the full model, effect modi cation or interaction effect at p < 0.1 and multi-collinarity effect between independent variables using variance in ation factors (VIF > 10%) were assessed. All independent variables included had VIP < 10 and multi-collinearity effect of each variables were p < 0.1. Hence, there was no signi cant interaction and multi-collinearity effect was detected. Since, the sampling procedure for this study was multistage sampling procedure; due to cluster variability mixed effect multilevel regression model was applied to detect determinants factors of adverse birth outcomes (still birth, neonatal death and any neonatal illness). Thus, for this study 'Kebeles/Ketenas' were categorized as clusters and cluster level variables were place of residence, access to health center and household wealth index were taken as level -2 factors. Women who gave birth during follow study were nested within their households and therefor in the community. As a result women individual level variables were socio-demographic, obstetric, information, maternal health services and newborn health services were taken as level -1 factor. Log likelihood ratio (LR) test was to con rm goodness of t of the multilevel model that found to be statistical signi cant such as data t the model. Finally, the effect of continuity of maternal health services perinatal death was estimated by propensity score matching. The effect was measured by β 95%CI at p < 0.05.

Newborn Characteristics and Health Related Issues
The characteristics of newborns born: 1102 (53.7%) were male and 359 (17.4%) were amniotic member break before labour started and time of premature rupture of membrane was 1 -12 hours before labour started 1132 (55.5%). More than three fourth 1585 (76.8%) of women reported that the color of amniotic uid was clear, whereas 305 (14.8%) of women stated that amniotic uid had bad odor. Majority, 1896 (91.8%) was the baby ever breast fed within 28 days and 1851 (84.2%) of the newborns were crying immediately after birth (Table 1). and anemia 5 (13.0%) ( Table 2).  (Table 3). Factors affecting Adverse birth outcome The determinant factors of adverse birth outcome were discovered using the mixed effect multilevel logistic regression model. Before running the full model, ICC (ρ) was calculated in the empty model for the outcome to make decision whether the data t to a mixed effects multilevel model or not. Then, ICC (ρ) was calculated as a full model for the outcome to detect the variability attributed to clusters after controlling the individual level.
Likewise, the adverse birth outcomes, ICC (ρ) was calculated in the empty model and it was found to be 0.32, indicating that 32% of the variation was contributed by cluster variations. The test of the preference of log-likelihood versus logistic regression was statistically signi cant (P < 0.0001). Then, the full model was run by including both the cluster level and individual level variables and the ICC (ρ) was increased to 0.53. This again indicated that 53% of the variation was attributed to cluster level variables. The preferences of log-likelihood verses logistic regression was determined indicate that statistically signi cant (p < 0.0001). Hence, this is suggested that using multilevel analysis is crucial rather than logistic regression model (Table 4). After adjusting for confounding effects, a multivariable mixed effects logistic regression model was performed. Then, the determinant factors of adverse birth outcome identi ed at cluster and individual level factors such that at the cluster level (level -2) factors: house hold wealth index and at individual level (level -1) factors: women educational status, history of abortion, time of 1 st ANC initiation, iron folic acid supplementation, initiate BF within one hours, pregnant related problems during labour and immediately after childbirth, presence of offensive amniotic uid and immunizing newborn within postnatal period were found to have a statistical signi cant association with adverse birth outcomes.
The odds of having adverse birth outcomes among women resided in poor household wealth index (AOR = 3.3; 95%CI: 1.07, 10.23) were three times higher than among women resided in rich household wealth index. Similarly, the odds of having adverse birth outcomes among women who attended high school (AOR = 6.12; 95%CI: 1.87, 20.1) were six times higher than among women who had no formal education.

Effect of Continuity of Maternal Health Services on Perinatal death
To reduce the potential of confounding effects, we utilized propensity score matching (PSM) approach to compare women who had complete continuum of care in maternal health services with their main interventions to those who had terminated/discontinued the services. Among the three distinct techniques of propensity score matching (PSM), one to one matching was utilized to quantify the effect of intervention on perinatal death. After matching treated and control individual, the effects of completion of continuum of care in maternal health services via time and space dimension on perinatal death were determined.
Hence, the result found to be received 1 st (Table 6).

Status of Adverse Birth Outcome
Adverse birth outcomes are the most essential vital data used to evaluate maternal and child health programs and plan evidence-based interventions. They serve as a gauge for the quality and effectiveness of maternal health services that provided to mothers and children. The basic indicators used to measure adverse birth outcomes were stillbirth, neonatal death and any illness during postnatal period. Thus, this study found that 12.4% women had adverse birth outcomes which was lower than evidence from Northern Ethiopia 27.5% (20). This is because the variation of study time and design. Due to the time variation, a community based intervention that implemented by HEW will reduce the magnitude of adverse birth outcomes in the study area.
According to the global target, every country will have 12 or fewer stillbirths per 1,000 births by 2030. Even though developed countries and upper middle income countries were meet their target, developing and low income countries will have to more than double present progress to reach this target (21). Beside these, magnitude of stillbirth was 27 stillbirths per 1,000 pregnancies in the study area. It was much higher than global target (12 per 1,000 pregnancies), Mumbai Slums 8.68 per 1,000 births (22) (27) and Tigrary region (2.3%) (28). The reason may be reduction of stillbirth is much slower than other countries due to poor utilization of recommended ANC services via visit based and content based continuity of care and also the quality of maternal health services were poor as compared with other countries.
Neonatal mortality is the probability of dying within the rst month of life which is the most pertinent national indicator in maternal and child health programs. Hence, this study found that neonatal mortality rate was 31 per 1,000LBs. This nding is almost similar with the evidence from EDHS 2019 ( (7) and Tigrary 4.6% (28). This implies that neonatal mortality is still high in developing countries and Ethiopia particularly in the study area and there is no any improvement in the reduction of neonatal mortality to achieve global and national targeted goal.
In this study, we found that perinatal mortality rate was 48 per 1,000 pregnancies. This nding is consistent with the pooled estimate of perinatal mortality rate in Ethiopia was 51.3 per total births (34).
However, this nding was higher than study in Hawassa town HF 2.7% (35) and rural northern Ghana 39 per 1,000 births(36). Whereas, it was lower than evidence from low income countries 55 per 1,000 births (37), Rural Pakistan 69.95 per birth (24) and Rural Congo 61 per 1,000 births(38). Among newborns died within postnatal periods, 59.5% of the newborns were died within 48 hours and early neonatal deaths contribute 72.9% of neonatal deaths. This nding is consistent with different study done in the countries and outside the countries (32,33,37,39).
Worldwide, the major causes of neonatal death are infections (sepsis/pneumonia, tetanus and diarrhea), prematurity/preterm birth and birth asphyxia. In line with these, we found that the three major neonatal characteristics that causes neonatal death: birth asphyxia (66.0%), sepsis/infection (50.5%) and pneumonia (42.0%). This nding was supported by different studies (18,22,(36)(37)(38)(39)(40). However, some variation between countries may be sought because of the variability of quality, availability and utilization rate of maternal and newborn health services.
Maternal complications during pregnancy and childbirth are predisposing factors for neonatal death. As result, safe motherhood is a proven initiatives to prevent newborn death by overcoming pregnant related complications (39). Hence, this study found that obstructed labour (56.4%), obstetric sepsis/maternal infection (54.7%), and APH (33.3%) were common pregnant related problems that cause neonatal death.
This nding was consistent with evidence from low and middle income countries which found that obstructed labor and placental abruption cause birth asphyxia, and also maternal infections cause neonatal infections that cause neonatal death (41). This is due maternal and neonatal health condition are interlinked each other's.
In this study, the possible causes of stillbirth were maternal infection or obstetric sepsis, maternal malnutrition and unnecessary medication. This nding is consistent with evidence from low-income countries (37) and rural northwest Bangladesh (42). Moreover, before the occurrence of stillbirth, women were frequently suffered with pregnant related problems (severe abdominal pain, excessive vaginal bleeding and blurred vision and any illness during pregnancy (febrile illness, malnutrition and anemia). This nding is supported by other study (21). This is because pregnant related problem and maternal illness during pregnancy cause poor placental function, either with fetal growth restriction or preterm labor, or both, is a common cause of stillbirths. As a result, different evidence argued that better prenatal care is critical to preventing 1.3 million stillbirths and ending unnecessary neonatal deaths (21).

Determinant factors of Adverse Birth Outcome
Among community level factors, household wealth index is predictor of adverse birth outcome. Beside, women resided in poor household wealth index were three times more likely to have adverse birth outcome. This is because level of household index are directly related to access of information and receiving health services, which enable them to prevent themselves and her newborn from any illness and adverse event.
Educational status is a determinant factor of adverse birth outcomes, in which women who attended high school were six times more likely to had adverse birth outcome as compared with women who had no formal education. This may be educated people have better knowledge and information on the impact of adverse birth outcomes, as result they are early diagnosis the problems and report properly for further treatment and management as compared with people with no formal education. Moreover, people lack education did not report any adverse event and believing that the event as a usual life rather than considering as health problems. So, the magnitude of adverse birth outcomes was high among educated people.
In this study, initiating 1st ANC visit between 4 -6 months of gestational age reduced adverse birth outcomes by 62%. This study is consistent with previous study in Gondar University Hospital (26) and Ghana (23). This could be the fact that during ANC visits, relevant health issues can be identi ed and addressed early. This is only feasible if pregnant women present themselves to a suitable health facility at an early stage of labor and pregnancy. As a result, the probability of occurring adverse birth outcomes will be remission.
According to WHO directives, IFA supplementation during pregnancy, early initiating BF within 1hour for the newborn and immunization increased the survival of newborn by overcoming the adverse event. In line these, we found that women received IFA supplementation during pregnancy and initiate BF within 1hours for her newborn and immunized newborn within postnatal period were less likely to have adverse birth outcomes. This nding is concurrent with previous studies within the country and abroad (6, 10, 18, 32).
The occurrence of pregnant related problems during pregnancy, childbirth and immediately after childbirth decreases the survival of newborn. Beside, we depicted that women who had pregnant-related complications during pregnancy, labour and childbirth and immediately after childbirth were more likely to have adverse birth outcomes. This nding consistent with evidence from low-and middle-income countries (43), Developing countries (44), Ghana(23), Indian State of Bihar(31), Jimma Zone(18) and Gondar University Hospital(26). This is the fact that women who have pregnant related complications were prone for infection and birth asphyxia which are the leading cause of stillbirth, neonatal death and neonatal illness. This is also explained by the health condition of newborn relies on maternal health status before, during and after conception.
As principle, amniotic uid should be odorless or have a slightly sweet odor. However, if it is noticed that smells bad, it could be a sign and symptom of uterine infection. Thus, this study revealed that women who had an offensive odder of amniotic uid were three times more likely to have adverse birth outcome. This is explained by offensive odder of amniotic uid is an indication of infections, which is the leading cause of stillbirth birth, neonatal death and illness during postnatal period.
Effect of continuum of care in maternal health services on perinatal death Maternal health interventions during pregnancy, childbirth and after childbirth have a bene cial health impact on maternal and newborn health outcomes (45). This is because early detection of pregnancyrelated disorders throughout pregnancy, management of labor and delivery, care of any adverse event, and treatment of serious infectious infections and acute malnutrition are all effective interventions of maternal health services packages, which could reduce 77% of perinatal mortality(46).
In line with these, we found that women received 1st ANC visit; completed 4th ANC visit; skilled attendant of ANC services; completed continuity of care for both 4th ANC and skilled delivery; completion of continuum of care in maternal health services; completed key services of ANC package, completion of key services of PNC package, complete whole key service maternal health service and completed continuum of care via space dimension were signi cantly reduce perinatal deaths. This nding is consistent with previous studies conducted in the country and abroad (6, 7, 18, 23, 24, 28, 38, 40, 47-60). This is because more than one third of all child deaths occur in the rst month of life, providing skilled care to mothers during pregnancy, as well as during and after birth, greatly contributes to child survival, with two-thirds of newborn deaths preventable if known, effective health measures are provided at birth and during the rst week of life.

Conclusions
In this study, we concluded that neonatal and perinatal mortality rate were high as compared with national and international targets. Risk factors for adverse birth outcomes were poor household wealth index, pregnant related complication during pregnancy, childbirth and immediately after childbirth, bad odder amniotic uid and history of stillbirth. Whereas, the protective interventions were receiving IFA supplementation during pregnancy, initiating BF within 1hour for newborn, immunizing newborn and completion of continuity of care in maternal health services via time and space dimension. Efforts should be made to strengthening supplementation of IFA during, immunizing newborn and early initiation of BF for newborn as well as early detection and treatment of risk factors for poor birth outcomes were strongly recommended. Moreover, increased attention to newborns, as well as, in the context of the continuum of care, to women before and after giving birth, is required to effectively reduce adverse birth outcomes, perinatal and neonatal mortality. This must be accomplished through reaching out to disadvantaged populations' particularly poor population and providing them with the necessary health care. Necessary permission letters were obtained from Regional Health Bureau and respective all local districts. Con dentiality was maintained by avoiding any identities from the questionnaire. Before starting actual data collection, written and verbal consent was obtained from each study subjects.

Consent to publish Not Applicable
Availability of data and materials The d atasets used and analyzed during the current study are available from the corresponding author on reasonable request.