Incidence and predictors of neonatal mortality among neonates admitted in Amhara regional state referral hospitals, Ethiopia: prospective follow up study

Neonatal Mortality Incidence Rate (NMIR) is a probability of dying during the first 28 days of life and expressed per 1000 admitted neonates [1, 2]. The neonatal period commences at birth and ends at 28 completed days after birth and this period is the most vulnerable time for a child’s survival [3‐5]. This period is also further classified as early (the first 7 days) and late (the remaining 21 days) neonatal period and majority neonatal mortality occur in the early neonatal period [4, 6, 7].

Neonatal mortality (NM) is a major global public health challenge [5, 8, 9]. Approximately 2.7 million neonates worldwide each year die in the first month of life and most of these deaths happen in developing countries [6, 10]. NM accounts for about 44% of under-five deaths and more than 99% of these deaths occur in middle or low-income countries [5, 6]. According to United Nations International Children’s Emergency Fund (UNICEF) 2018 report, Neonatal Mortality Rate (NMR) in the globe and the least developed countries was 19 and 26 per 1000 live birth respectively [1]. Accordingly, the problem is much higher in South Asia (SA), West and Central Africa (WCA), and sub-Saharan Africa (SSA) [1]. Similarly, NMR in Ethiopia is higher (i.e., reported to be 30 per 1000 live births) and the largest figure was recorded at Amhara national regional state where it was estimated to be about 47 per 1000 live birth [2, 11].

To alleviate these problems, efforts have been made globally as well as nationally. For instance, it was included in and remained ‘unfinished agenda” of Millennium Development Goals (MDGs) which was extended to Sustainable Development Goals (SDGs) [9, 12]. Hence, target 2 of SDG 3 is planned to end preventable deaths of newborns by all countries aiming to reduce neonatal mortality to at least as low as 12 per 1000 live births by 2030 [9]. In line with this, Ethiopia has planned to reduce NMR from 28 in 2015/16 to 10 per 1000 live births by 2019/2020 [13]. Consequently, the NMR has went down from 36 deaths per 1000 live births in 1990 to 19 in 2015, and the number of neonatal deaths declined from 5.1 million to 2.7 million globally [10]. Generally, the number of children dying before the age of five has declined from 9.9 million in 2000 to 5.6 million in 2016 [3]. However, the proportion of under-five deaths in the newborn period has increased from 41 to 46% between 2000 and 2016 globally [3]. Moreover, the decline in neonatal mortality (47%) from 1990 to 2015 has been slower than that of post-neonatal under-five mortality (58%) globally [3]. The scenario is similar in Ethiopia, where NMR has decreased from 49 in 2000 to 29 in 2016 as to EDHS reports [2]. However, the NMR decrement in Ethiopia is also much slower than that of post neonatal under-five mortality (41% compared with 85%) from 2000 to2016 [2]. Thus, identifying the causes of such challenges is one import part of the efforts. The previously identified causes of neonatal mortality such as hypothermia, neonatal sepsis, respiratory distress syndrome, and prenatal asphyxia may vary across settings and over time [12, 14‐19].

The high NMR reflects the poor quality of care service especially during the perinatal period in a given country or region [3]. Hence, its implication reminds the concerned bodies to evaluate their respective health care system and to develop effective strategies. To design such evidence-based and problem oriented strategies which may be effective - for instance, in reducing delays in accessing obstetric and neonatal health care services such as improving community awareness of neonatal death risk factors, cultivating practices towards birth preparedness and complication readiness plan, reducing transportation barriers, preventing financial barriers, strengthening referral mechanisms, reducing barriers to good health-seeking behaviors, and enhancing the level of health facility’s preparedness and readiness for neonatal obstetric emergencies, the role of accurate contemporize information is compulsory for combating the burden thereby for achieving the SDG. Thus, methodologically sound and representative investigations need to be done by far, more importantly in the settings where high NMR has been reported, and the Amhara national regional state is an example. Having this insight, this study was conducted to assess incidence and predictors of neonatal mortalities among neonates admitted in Amhara region referral hospitals, Northern Ethiopia, 2018.

A multicenter prospective follow up study was conducted in referral hospitals which are located in Amhara national regional state, Northern Ethiopia from July 01 to August 30/2018.

Amhara national regional state is the second-largest and populous region in Ethiopia [2]. The region has five referral hospitals namely University of Gondar comprehensive specialized hospital (UoGCSH), Felege-Hiwot comprehensive specialized hospital (FHCSH), Dessie referral hospital (DRH), Debre-Markos referral hospital (DMRH) and Debre-Birhan referral hospital (DBRH). Among these, Dessie, Felegehiwot and Debre-Markos referral hospitals were selected randomly.

Thereafter, we included all neonates who had been admitted in the corresponding neonatal intensive care unit (NICU) of each selected hospital.

We calculated the sample size for this study by using STATA version14 with the following assumptions: the probability of neonatal death-0.88 and the hazard ratio of the predictor breast feeding-7.5 as reported in a previous study done in Tigray, Northern Ethiopia [12]; power-80%; confidence level-95%; margin of error-4% and non-response rate-10%. Accordingly, the final sample size was obtained to be 631.

The number of sample size for each selected hospital was allocated proportionally based on the number of previous cases flow. All recruited study participants had been followed till the outcome of interest (i.e. either death or censure) appeared. Data were collected through face to face interview and chart review. The outcome variable of the study was admission outcome which was dichotomized as death (coded as ‘1’) or censored (coded as ‘0’) whereas, the explanatory variables included socio-demographic, obstetrics, medical, health service and neonatal related factors. These include maternal age, mother’s educational status, husband’s educational status, marital status, religion, maternal occupation, husband’s occupation, ethnicity, residence, distance from nearest health facility, premature rupture of membrane (PROM), delivery complication, mode of delivery, parity, Infant of Diabetic Mother (IDM), history of previous dead fetus, polygamous household, number of ANC visit, HIV status of mother, tetanus toxoid vaccine, multiple pregnancies, birth interval, skilled birth attendant, place of ANC visit, place of delivery, birth weight, congenital abnormality, fetal presentation, 1st minute APGAR score, 5th minute APGAR score, time of initiation of breastfeeding, hypothermia, neonatal sepsis, respiratory distress syndrome (RDS), jaundice, perinatal asphyxia (PNA), anemia, hypoglycemia, prematurity, neonatal sex, age on admission, neonatal complication, feeding option, cry at birth and weight at admission.

Neonatal mortality rate remains unacceptably high in most of resource limited countries including Ethiopia although under-five mortality has been declining substantially since 2000 [20]. Moreover, the proportion of neonatal deaths over under-five deaths is increasing time to time despite the governmental commitments. For example, in Ethiopia, the neonatal mortality rate was reported to be 29 per 1000 live births as of EDHS 2016 and was targeted to reduce to 10 per 1000 live births in 2019/20 in line with the target two of SDG 3 [9]. Paradoxically, the magnitude reported to be elevated (i.e. 30 per 1000 live births) according to mini EDHS 2019 [11]. Hence, it is among major current public health challenges which require appropriately designed strong studies to generate accurate and representative information so that stakeholders can apply evidence based interventions. Eventhough some studies had been conducted on neonatal mortality and predictors in Ethiopia, no large scale study including all referral hospitals of Amhara region was carried out, as to the investigators deep review. In addition, no prospective follow up studies had been done on neonatal death in the study area. Therefore, this study is aimed to address these gaps particularly to assess NMR, time to death and predictors among admitted neonates in referral hospitals in Amhara national regional state to provide basic information for any intervention aimed at improving neonatal health. .

The magnitude of NMR is higher than the national figure and the highest probability of neonatal death has been recorded in the early neonatal period. The risk of NM was higher among neonates whose mothers were HIV positive, uneducated, advance aged and gravid of multiple fetuses in the indexed pregnancy.

In the current study, NMR was 186 per 1000 admitted neonates. This finding is comparable with the study conducted in Wolaita Sodo referral hospital in southern Ethiopia (173) [21] and sub-urban hospital in Cameron (157) [22].

However, the current study findings demonstrated that the magnitude of NMR is higher than the result of other previous studies done in Iran- 91 [19], China- 12 [23], Suriname- 129 [24], Indonesia- 52 [25], Nigeria- 41 [7], Burkina Faso- 46.3 [26], Sudan- 30 [15], UNICEF 2018 report [1] and EDHS 2016 report- 29 [2]. It is also higher as compared to local studies conducted in various part of Ethiopia such as the Somali region of Ethiopia- 57 [27], Dilchora referral hospital in Dire Dawa- 114.4 [28], Jimma Zone- 35.5 [17], rural area of Eastern Ethiopia- 28.37 [29], Tigray, Ethiopia-62.5 [12] and Ethiopia-36.7 [10]. This discrepancy might be attributed to variation in study design. Unlike the current study, some previous studies were community-based [15, 17]. Hence, NM tends to be lower in the community-based studies where low-risk neonates were also included in the denominator, unlike the investigations conducted in the NICU. In the studies which have been undertaken at the NICU, like the current one, the denominators are all admitted neonates (i.e. the population at higher risk for death), and it is logical that the probability of neonatal death is likely to be higher among such population than the counterparts. Moreover, the disparity in magnitude could be accredited to variation in study settings. In other words, some previous studies were conducted in developed nations [24]. Empirical evidence supports that the overall health care system, as well as the survival rate of neonates, is much better in the developed countries than in resource-limited countries [1]. The implication is ascribed to the impact of the financial barriers for receiving basic obstetric and neonatal cares in economical poor countries including Ethiopia. In connection to this, one surprising point in the current study was that about 93.79% of respondents received at least one ANC. Similarly, more than 96% of participants gave birth at health facilities. This implies that the maternal health care service coverage was high and that was promising. However, about 52.78% of the mothers had experienced delivery complications. Moreover, nearly half (46.12%) of the neonates had poor 1st minute Apgar score. In addition, almost all reasons for neonatal admissions and neonatal deaths were found to be preventable. This questions the quality of health care services during the perinatal period. One most probable reason for such unacceptable poor quality of maternity care and poor neonatal health outcome is delays in accessing emergency obstetric and / or neonatal cares which include delays in decision to seek care (delays at the home), delays in reaching health facilities (delays at the journey) and delays in receiving appropriate care (delays at the health facilities). One known contributing factor for such delays is financial barrier which further seems true in the current study since about four-fifth (77.1%) of the respondents’ family had earned less than 139 US dollar monthly. The costs of essential maternal and child healthcare – whether formal or informal costs – are major barriers preventing the poorest and most excluded women and children getting the services they need. User fees have been shown to be highly regressive, with a particularly negative impact on the poor, reducing uptake of essential health services. Out-of-pocket (OOP) expenditure for healthcare during pregnancy and childbirth is still common in many low- and middle-income countries. Although the country, Ethiopia, endorsed the cost-free maternal and child health care services during pregnancy and child birth processes, most of the supplies and medications are unavailable in public health facilities at the time of request. Thus, women often have to pay formal or informal charges to receive care, and have to buy their own medicines and basic necessities, such as soap and clean sheets, when going into hospital to give birth. Thus, improving transportation, strengthening referral mechanisms, assuring ambulance distribution and maintenance policies, equipping health facilities with essential supplies, mobilizing resources for MNCH quality improvement, availing essential reproductive health commodities, facilitating inter-sectoral collaboration to address common maternal and newborn health problems and work for adequate budget allocation for MNCH would decrease the high magnitude of neonatal mortality. Furthermore, the discrepancy could be attributed to variation in the scope of the study. The former neonatal mortality reports were made based on the country’s aggregated findings [1, 2]. Thus, these averaged magnitudes are expected to be lower as compared to the death incidences from referral hospitals where near-missed-referred cases are more likely to be abundant. Also, the dissimilitude in mortality might be ascribed to variation in parent’s socio-demographic characteristics.

On the other hand, the magnitude of NMR in the current study is lower as compared to the findings of the studies done at University of Gondar referral hospital- 252 [30], Paul’s Hospital Millennium Medical College Addis Ababa- 232 [31], Pakistan Health Survey of 2014–680 [32] and University of Gondar comprehensive Specialized hospital (UoGCSH)- 288 [33]. The variation in the magnitude of NMR could be secondary to the difference in the scope of the study. It is observed that some other previous studies were conducted in a single centered institution which might have lacked representativeness unlike the current one [30, 31, 33]. The disparity also could be linked to discrepancy in sample size. In the former study conducted in Addis Ababa, the sample size was 216 which accounts only nearly one-third of the current one [31]. The difference in magnitude also might be related to respondent’s residency. In the previous study, about seven-tenths (70%) of the respondents were rural dwellers which is much greater than that of in the current study [32]. In this perspective, evidences exhibit that neonatal mortality tends to be higher among rural residents [34, 35]. This could be rationalized as rural residents usually could not get access to health-related information easily and health care services timely as equal as urban residents. Besides, the dissimilarity in NMR extent might be secondary to variation in the study population. Unlike the current study which recruited all admitted neonates disregard to their gestational age, some previous studies had entailed only preterm neonates [30, 33]. In connection to this, studies indicate that the risk of neonatal death is higher among preterm neonates than among term neonates [31, 36].

The current study exhibits that the overall cumulative survival probability of neonates at the end of the follow-up period was 71.11%. This result is lower as compared to the findings of a previous study done in Tigray, northern Ethiopia 93.96% [12]. This variation in survival probability of neonates might arise from the difference in the study population. In the former study, all live births (i.e. all alive neonates) had been recruited, whereas only admitted neonates (i.e. population at high risk) had involved in the current one. Thus, the survival probability is more likely to be higher among all live births than that of neonates admitted with some problems.

On the contrary, the survival probability of neonates in the current study is higher as compared to that of the study carried out at UoGCSH (57.14%) [33]. This dissimilarity might be accredited to variation in study population again. Accordingly, the study population in the previous study was premature neonates that pose a higher probability of death as compared to that of in the current study (i.e. all neonates irrespective of gestational age).

This study also shows that the cumulative survival probability of neonates at the end of the first day was 89.15%. This is lower than the previous study done at UoGCSH where the cumulative survival probability over 24 h was 96.71% [33]. As far as the result of these two studies compared, some sort of surprising findings are observed. In the first 24 h, the survival probability of neonates in the former study (96.71%) was higher than that of the current one (89.15%). Thereafter, the survival probability in the early neonatal period and overall cumulative period comes to be lower in the previous study (74.62 and 57.14% respectively) as compared to the current study (79.70 and 71.11% respectively). This implies that the probability of death among study population in a previous study (i.e. premature neonates) has highly increased as neonatal age advanced than that of the general admitted neonates.

In the current study, it was observed that the risk of death among neonates born to mothers with advanced age is more than two times higher as compared to their counterparts. This finding is also supported by other studies conducted in Sudan [15] and Suriname [24]. This could be related to the evidence-based facts that advanced maternal age is associated with certain pregnancy-related risks including poor neonatal outcome [37]. Therefore, efforts need to be collaborated to minimize the risky conditions such as pregnancy at advanced maternal age.

In many circumstances, maternal educational status matters obstetric and perinatal outcomes [10]. Studies indicated that the outcome of pregnancy among educated women is more likely to be better than among uneducated ones [38]. The current study had come up with findings which support this fact. Accordingly, the risk of death for neonates born to mothers who are unable to read and write was 1.40 times higher as compared to those neonates whose mothers can read and write. This aligns with the findings of previous studies done in different settings in Ethiopia [10, 34]. This could be explained by the role of education on the maternal level of apprehensive. Accordingly, educated mothers have a better awareness of the healthcare-related information and thus, might behave and act as per the standard of health care service utilization in a symphonious manner with their caregiver. Moreover, educated mothers are more likely to search health-related information in a variety of other ways such as by attending mass media, browsing internets and reading magazines. Consequently, they are likely to evolve with accurate health-related information enough to make appropriate decisions and to take corrective actions even in the emergency. Surprisingly, nearly two-fifth (38.72%) of the respondents in the current study had never attended any formal education at the time of interview. Meanwhile, being unable to read and write exhibited to be the independent predictor of neonatal mortality. The finding implies that all stakeholders need to work more on the gender inclusive education policy which promotes all females to be engaged in formal education. Similarly, about 40.81% of fathers of neonates had never been educated through formal education. Hence, the finding also entails the need of information, education, and communication to improve the community awareness of neonatal death risk factors and danger signs.

Also, neonates born to HIV positive mothers were 6.57 times more likely to die during the neonatal period as compared to those born to sero-negative mothers. This is following a systematic review study [8]. This is also in line with a study conducted at University of Gondar Referral Hospital, Ethiopia [30]. This might be related to the effect of HIV infection on the maternal overall health condition. And it is evidence-based fact that the fetal health condition is directly or indirectly influenced by maternal health status [39]. In other words, the maternal and fetal general condition is strongly interlinked to each other. Thus, if the maternal immunity and other conditions are compromised by HIV infection, the fetal health condition would be endangered too, which the jeopardized fetal health condition would more likely lead to neonatal death. The finding, herein, suggests that offering integrated PMTCT services to MNCH programs would decrease neonatal mortality rate. The result also further alerts the health care providers to be vigilant enough to screen, diagnose and treat maternal infections. According to the current study finding, the risk of neonatal mortality is more than three times higher in multiple pregnancies than single tones. This finding is supported by previous studies done in different settings [17, 19, 26, 30, 40]. This could be attributed to the effect of multiple pregnancies in life-threatening perinatal outcomes. Multiple pregnancies usually carry a higher risk of prematurity, premature rupture of membrane, abnormal presentation and amniotic fluid abnormality and IUGR which all may increase the probability of neonatal death [41‐43]. Hence, as multiple gestations is one of the high risk pregnancies and is an independent predictor of neonatal death, it demands close monitoring and care.

This study recruited all admitted neonates who were at higher risk of death. Thus, the NMR reported from this study could be overestimated as compared to other reports made by taking all live births as a denominator. In this study, the measurement of some variables might have been committed to recall bias. Of course, we had executed certain efforts to minimize such potential recall biases. First, the questionnaire had been well devised in such a way that it was structured from simple to complex and recent to past in chronological order. It was also translated to local language. In addition, we had evaluated the tool by conducting a pretest prior to the actual data collection. Second, we had delivered trainings for the data collectors and supervisors on the strategies of minimizing recall bias. Thus, the interviewers had been told to assure that each question is clearly understood by the respondents before the responses, to inform the local events that may trigger the respondents’ recalling capacity, to give participants sufficient time to recall long term memory, to verify with any record or any other potential source of information such as family members, intimate friends and any other persons who were around at the time of event, and to use multiple source of information like crosschecking the documented records and contacting the concerned health care providers for uncertainties in documentations.

The neonatal mortality incidence rate in the study setting was higher than the national figure, and the highest probability of neonatal deaths has been recorded in the early neonatal period. Advanced maternal age, uneducated maternal status, positive maternal HIV status, and multiple pregnancies were the independent predictors for neonatal mortality. Thus, stakeholders need to consider the distinguished risky time and the identified independent risk factors of neonatal death while they design strategies to reduce the high mortality rate.