Time to death and its predictors among neonates who were admitted to the neonatal intensive care unit at tertiary hospital, Addis Ababa, Ethiopia: Retrospective follow up study

Backgrounds Neonatal death is the major problem in developing world. Burden and predictors of neonatal mortality vary across countries and even among regions of a country, so understanding the problem concerning these factors is essential to overcome the problem. Therefore, this study aimed to determine time to death and its predictors of neonatal mortality among neonates who were admitted to the neonatal intensive care unit of Tertiary Hospital, Addis Ababa, Ethiopia. Methods A hospital-based retrospective cohort study was employed among 434 neonates admitted in Tertiary hospital, Addis Ababa, Ethiopia. A Kaplan Meier curve and a log-rank test were used to estimate the survival time and compare survival curves between variables. The cox proportional hazard model was also fitted to identify predictors. Results A total of 434 neonates included in the study, 11.1% of which were died, and the incidence rate was 19.2 per 1000 live births. The time to death of neonates was 17 days. Independent predictors of neonatal mortality were incomplete maternal antenatal follow up[AHR: 3.7 (95% CI:1.86,7.60)], low(Appearance, Pulse, Grimily, Activity, and Respiration(APGAR)score[AHR:5.0 (95%CI:1.51–15.04)], perinatal asphyxia [AHR:5.2 (95%CI:1.92–14.30)], preterm 4.2 (95%CI: 1.32–8.83)]. Moreover, small for gestational age [AHR:4.8 (95%CI:2.33–9.72)], respiratory distress[AHR: 2.5 (95%CI: 1.24–5.09)], sepsis [AHR: 3.4 (95%CI: 1.71–4.01)], low birth weight[AHR: 7.3 (95%CI:2.69,1.91)], and tracheoesophageal fistula [AHR: 2.2 (95%CI: 1.13–4.32)]. Conclusion The overall incidence rate was 19.2 deaths per 1,000 live births. Emphasis should be given to incomplete Antenatal care follow up, small for gestation, preterm, low birth weight, low 5th min APGAR score, neonatal sepsis, respiratory distress, perinatal asphyxia, and tracheoesophageal fistula.


Introduction
Globally, it is estimated that ∼6,500 neonatal deaths occurred every day. Of those, about a third of all neonatal deaths occurred within the first day after birth, and three-quarters occurring within the first week of life (1). Besides, the neonatal mortality rate was 37% in 1990, 31% in 2000 and 18% in 2017 (2). Despite, there is a decline in neonatal mortality from 1990 to 2017, it is slower than post-neonatal and under-5 mortalities (2)(3)(4)(5). Neonatal death is unevenly distributed based on age, sociodemographic population groups, and other factors (2,6). The 2022 global report bared that the highest neonatal mortality reported in Sub-Saharan Countries which was 27 deaths per 1,000 live births with 43% of global newborn deaths, followed by central and southern Asia which was 23 deaths per 1,000 live births, with 36% of global newborn deaths. Moreover, a child born from Sub-Saharan Countries is 10 times more likely to die than child born from high income countries (7).
The three major causes of neonatal mortality in developing countries include prematurity, infections, and perinatal asphyxia (2,8,9). Newborns that need intensive medical attention are often admitted into a special area of the hospital called the neonatal intensive care unit [NICU]. A new initiative called the Global Strategy for Women's, Children's, and Adolescent's Health [2015][2016][2017][2018][2019][2020][2021][2022][2023][2024][2025][2026][2027][2028][2029][2030] and the third Sustainable Development Goal [SDG] has been established to ensure healthy lives and promote wellbeing for all ages (2). The SDG goal 3 geared toward preventing deaths of newborns specifies that all countries should aim to reduce neonatal mortality to at least as low as 12 deaths per 1,000 live births by 2030, but without strengthened commitment to newborn survival, many countries couldn't meet the SDG goal to end preventable neonatal deaths (2).
In Ethiopia, a 2020 report estimated that neonatal mortality rate for Ethiopia was 27 deaths per 1,000 live births. The rate of neonatal mortality decreased from 47.8 deaths per 1,000 live births to 27 deaths per 1,000 live births in 2001 and 2020, respectively (10). Over two-thirds of neonatal deaths are mainly due to infections and neonatal medical conditions (11)(12)(13) In particular, under-five mortality rates declined from 166 to 67 deaths per 1,000 live births in 2016. Similarly, infant mortality decreased from 97 to 48 deaths per 1,000 live births in the same period, but the neonatal mortality decreased from 54 to 29 per 1,000 live births (14-17). This indicates the trend of neonatal mortality reduction in Ethiopia is moving at a slow step compared to the reduction of infant and under-five mortality over the same period. National analysis data showed that multiple birth neonates, neonates born to mothers who did not utilize ANC and neonates from rural area were predictors of neonatal mortality (18).
Despite different initiatives and implementations that have been implemented to prevent neonatal death, it is still high and not reduced as expected in developing countries including Ethiopia. Although several studies were made in different countries on the survival status and predictors of mortality among neonatal intensive care unit [NICU] admitted neonates, but only a few studies were done in Ethiopia. Moreover, studies regarding neonatal mortality have focused on rate and little has been done on the time to death and predictors.
Therefore, this study aimed to determine the time to death and its predictors of neonatal mortality among neonatal intensive care unit admitted neonates at Tertiary Hospital Addis Ababa, Ethiopia.

Eligibility criteria
All live-born neonates who were admitted to NICU of Tikur Anbessa Specialized Hospital from 1 January 2015 to 30 December 2018, Addis Ababa, Ethiopia were included. Whereas, Neonates of incompletely registered NICU logbooks and lost records at the time of data reviewing were excluded from the study.

Sample size determination and sampling procedure
The sample size was determined by using double population proportion in a STATA statistical program considering the following assumptions: 95% CI, power 80%, P1 = 10 % P2 = 20% (22), and 10 % incomplete data, yielding a total sample size of 434.
A sampling frame was prepared from the medical record of neonates that were admitted in NICU from 1 January 2015 -30 December 2018 [N = 12,888], then for each year, proportional allocation was taken, Randomly, the 2 nd MRN was selected in the same fashion for the rest 3 years, and then systematic sampling was employed every 30 th (Supplementary 1).

Study variables
Time to death was an outcome variable. Independent variables includes:-socio-demographic factors (maternal age, neonatal age, and sex); maternal health factors [(Antenatal care) ANC follow up, RH (Rhesus) factor, blood group; maternal medical disorders like hypertension, diabetic mellitus, HIV/AIDS, and others, and gynecologic-obstetric related factors like gravidity, parity, mode of delivery, multiple pregnancies, premature rupture of membrane, abortion, prolonged labor); neonatal factors (Age at admission, sex, the weight of neonate, date of admission and discharge. APGAR score, gestational age, maturity, neonatal medical conditions: asphyxia, congenital malformation, preterm, sepsis, hypothermia, RDS (respiratory distress syndrome), MAS(Meconium aspiration syndrome), HMD (Hyaline membrane disease), jaundice, and TEF (trachoesophageal fistula)]; health service-related factors (duration of hospital stay, access to emergency obstetric care, resuscitation efforts).

Operational definitions
Censored: didn't know survival time exactly due to study ends, loss to follow-up, withdrawal or being survived from study.
Low APGAR score: a score < 7. Event: death after NICU admission. Asphyxia: a neonate with an APGAR score of <7. Time to death: it is the time from admission at NICU to the occurrence of the outcome/event. Time scale: days from admission of a neonate to the occurrence of an outcome. Time origin: date of admission or start time of the cohort. Complete ANC follow up: ANC visit≥4 times at current pregnancy period.

Data collection tools and procedures
Data was collected by using a checklist developed from neonatal intensive care unit registration book format, which was prepared nationally as an "Ethiopian NICU form. The checklist consisted of new-born information recorded at admission and records of maternal information which were collected using a uniform extraction format developed by taking in to account all the relevant variables in the standard NICU registration book. Then all medical records that fulfill the inclusion criteria were reviewed retrospectively.
The checklist was prepared in English. A pretest was done on 5% of the sample size in Yekatit 12 hospital before the time of the actual data collection. The checklist was also examined by experts to check for content validity.
The data reviewers and supervisors were 8 BSc and 3 MSc Nurses. Before starting the actual work, one-day training and orientation were provided to chart reviewers on the objectives of the study, data reviewing techniques, how to keep confidentiality of information, how to fill the data abstraction checklist and data quality management. Close supervision was carried out by supervisors and investigators to ensure data quality. Then, the collected data was checked for its completeness and consistency before analysis.

Data processing and analysis
Data was coded, cleaned, edited and entered using Epi-data version 4.21 and exported to STATA SE version 14 for analysis. The event of interest was death. The time to death was calculated in days between the date of admission and the date of death. The log-rank test employed to illustrate whether considered covariates in the Kaplan Meier curves are statistically significant and in line with the Kaplan Meier curves. The bivariable cox regression was done and those variables having p-value < 0.25 in the bivariable analysis were included in the multivariable coxproportional hazards regression model. Then, those variables with P-value ≤ 0.05 at 95% confidence interval were considered as statistically significant.
Kaplan Meier survival curve together with log-rank test was fitted to test the survival time of neonates, and their survival curves didn't cross over each other. A goodness-of-fit (23) test was employed to assess the proportional hazard [PH] assumptions for each predictor variable. The result showed that variables included in the model satisfied the proportional hazard assumption [p-value was 0.7156 > 0.05], which indicates that the PH assumption has been met.

Results
Socio-demographic characteristics of mother and neonate

Maternal health conditions
Neonates born from mothers who had a complete antenatal follow up during their current.
Pregnancy, their death was 48/1,000 neonates, but neonates born from mothers of incomplete ANC was 360/1,000 neonates, and 615/1,000 for neonates born from mothers who did not have antenatal follow up were died.

Maternal obstetric conditions
About 15 [10.9%] of neonates born from mothers who had problems of obstetric problems died and 33 [11.1%] of neonates died born from mothers who did not have such problems (Table 3).

Maternal medical and related conditions
Neonates born from hypertensive mothers had 18% neonatal death risk than neonates born from non-hypertensive mothers 13.3%. HIV positive mothers had 29% hazard for neonatal death, but HIV negative mothers had 19% hazard of neonatal death (Table 4).  (Figure 1).

Survivorship function between groups of neonates
Those neonates who were diagnosed as small for their gestation had a longer survival time than those who were appropriate for their gestation and large for their gestation

Predictors of neonatal mortality
Results of the multivariable cox proportional hazard regression analysis revealed that neonates who were small for their gestational age were 5 times more likely to die as compared to those who were appropriate for their gestations [AHR:4.8 (95% CI: 2.33-9.72)].
Those neonates who had an APGAR score of less than 7 at 5 th min were about 5 times more likely to die as compared

Discussion
The study showed that the overall incidence rate was 19.2 deaths per 1,000 live births. Neonates were followed for 2,499 [95 CI; 2,326, 2,671] days, resulting in an overall incidence rate of 19.2 deaths per 1,000 live births. This finding was lower EDHS 2019 report [33 deaths per 1000 live births], respectively (24), Tigray region 62.5 per 1,000 live births (25), Wolayita Sodo teaching and university hospital 27 per 1,000 live births (26), and Sidama Zone in South nation 41 per 1,000 live births (27). This difference could be due to sample size taken, study design employed, study period, and also might be related to the variation of the health service coverage. Besides, this tertiary hospital gives service to those neonates who were referred from different areas of the country for better management. This made the referral system linked to treat to those client who couldn't' treated in other hospitals. Hence, the report from this hospital is lower than EDHS report. Because, EDHS data included reports from different institution from district hospital to tertiary hospital.
The result of this study showed that about 11.1% newborns were died, which was less than studies done in Cameroon 15.7% (28). The difference could be sources of data; this study reviewed data from only NICU admitted neonates, and might be also related to the quality of health care service given.
In this study, about 70% neonatal deaths were registered in the early neonatal period [0-6 days], and 30 % of were died during the late neonatal period which was inconsistent with study done in Kara Mara, in which 96 % were early . /fped. . . /fped. . The finding of this study had a lesser hazard of death than studies done in Ghana (32), and a pooled analysis was done on low income and middle-income countries (33), this variation might be due to methodological difference, and presence of less maternal predisposing factors for preterm birth in this study area.
Another significant predictor of neonatal death was perinatal asphyxia. Asphyxiated neonates were more likely to die than neonates without asphyxia, which makes it similar to a study done in NICU of Komfo Anokye Teaching Hospital [KATH] in Ghana (34), and NICU of Gondar university hospital (22). This might be related to asphyxiated neonates failed to establish breathing at birth than non-asphyxiated neonates, but this study finding was different from a study done in Wolayta Sodo about 2 times (26). This might be due to early diagnosis and treatment in Wolayeta Sodo before complications occur.
In this study, an APGAR score of <3 was found to be a significant factor, which was similar with the previous studies done in Cameroon (28), EasternEthiopia (35), Kenya (36), China (32,37), Swedn (38) and Arba Minch (30) and Eastern. This might be related to 5 th min APGAR score which is a useful index of the response to resuscitation efforts, and low APGAR score neonates are risky to be oxygen-deprived than an APGAR of >7. This discrepancy might be explained by the sample size, and obstetrical care, and neonatal resuscitation efforts. Low birth weight neonates were more likely to die than normal birth weight neonates, this was in line with a study done in NICU of Kath in Eastern Ethiopia (35), and in Northern Ethiopia (25). This might be due to a baby with low birth weight is at increased risk for complications of low birth asphyxia, hypothermia, feeding problem, infections, and prematurity, which might increase their risk of death than normal birth weight babies.
Those neonates who were small for their gestational age were more likely to die than those neonates who had appropriate gestational age, which was in line with a study done in Canada (34). Those who had neonatal sepsis at the time of admission were more hazardous to die than those who did not have sepsis, a consistent finding was observed in a study conducted at NICU of Gondar referral hospital (22), Tigray Hospitals (39) and Kenya (36). Neonates admitted with a diagnosis of respiratory distress were more prone to death than those who were not in distress, and was consistent with a study done in Wolayta Sodo (26), but this study had a lesser hazard than a study done in Ghana (32). This study illustrated that about 29% of neonatal death was related to respiratory distress, which was more hazardous than a study done in Kara Mara, Somali region of Ethiopia 24.4% (29). However, it was consistent with studies in Tigray (39), and Gondar (22).
Trachea esophageal fistula (TEF) was the other predictor of neonatal death. Neonates diagnosed as having TEF were more hazardous than those neonates who did not have. In this study, about 24.6% of neonates were died due to diagnosed TEF, which is consistent with previous study (40). Neonates with this condition potentially develop aspiration and faced feeding difficulties resulting in low calories and immunity which affect their growth and development. Neonates could easily dehydrate due to decreased fluid intake, inadequate breastfeeding.

Limitations and strength of the study
Strength of this study includes:-This is an approach to clearly indicate the temporal sequence of the predictor and outcome variable (death); Data reviewing was conducted by trained nurses, and standardized data abstraction tool was used and It enables to estimate the survival time of neonates, and identify predictors for future trend estimation and other related prevention and intervention programs.
Using secondary sources may result in loss of potential predictors, which may not be addressed in this study design like; maternal nutritional status, educational level, and family-related issues. As the incomplete charts were excluded from the study there may be selection bias. This study was done retrospectively which might cause recall bias due to failure of the caregiver to remember what was happened previously.

Conclusion
The overall incidence rate during the 2,499 person-time was 19.2 deaths per 1,000 person-day observation with a time to death of 17 days.
The independent predictors were incomplete or no maternal antenatal follow up, being small for gestation, preterm, low birth weight, low 5 th min Apgar score, neonatal sepsis, respiratory distress, perinatal asphyxia, and TEF. This recommended that maternal health during pregnancy, including quality of antenatal care utilization should be further strengthened, and risky pregnant mothers should be identified antenatally. Initiating different strategies to work on early identification, .

Author contributions
MM, GM, WK, FT, and BA was responsible to conception, design, acquisition of data, analysis and interpretation of data, and was responsible in the writing the original research and preparation of the manuscript for publication. BG and KD were responsible for design, supervision, and reviewing it for accuracy and integrity. All authors play roles in this research article, read, and approved the final manuscript.