Time to death and its predictors of mortality among neonates admitted to neonatal intensive care unit in public general hospitals of central zone of Tigray, Ethiopia 2024: a retrospective cohort study

Abstract

Background

Neonatal mortality, accounting for nearly half of global under-five deaths, declines slower than post-neonatal mortality. Preterm birth, perinatal asphyxia, and infections are leading causes in Ethiopia, where skilled birth attendance, Kangaroo Mother Care, and Antenatal Care are prioritized. The Ethiopia Demographic Health Survey noted a rise in neonatal mortality from 29 to 30 deaths per 1000 live births (2016–2019). This trend highlights critical gaps in prevention and healthcare delivery. Targeted research is essential to reduce preventable deaths and enhance child survival.

Objectives

To determine Time-to-Death and Predictors of Neonatal Mortality Among Neonatal Intensive Care Unit, Admissions in Public General Hospitals of Tigray’s Central Zone, Ethiopia: A 2024 Cohort Study.

Methods

A hospital-based retrospective cohort study involved 357 neonates selected via systematic random sampling. Data was coded and entered using Epi-data 4.6 and analyzed with STATA 14. Kaplan–Meier analysis estimated the median survival time. Log rank tests compared survival differences across explanatory variable categories. The Cox proportional hazard regression model assessed relationships between independent and outcome variables. Model fitness was evaluated graphically using the Cox Snell residual graph.

Result

Among 357 Neonates, 60(16.8%) died with an overall incidence rate of 17.7 (95%CI: 13.7, 22.8) deaths/1000 neonates-days total follow-up 3388 days. And the median survival time was 27 days. The predictors of neonatal mortality were rural residence (AHR = 3.44, 95% CI: 1.55, 7.66), lack of ANC (AHR = 2.80, 95% CI: 1.22, 6.51), HIV positive Mothers (AHR = 2.93; 95% CI: 1.27, 6.72) and delayed breast feeding (AHR = 2.12, 95% CI: 1.07, 4.20).

Conclusion

Neonatal mortality was 17.7 per 1000 neonate-days, with a median survival of 27 days. Rural residence, no ANC visits, maternal HIV positivity, and delayed breastfeeding were key predictors of neonatal death. Programs should focus on improving ANC access and community-based survival strategies, especially for rural mothers. Enhanced efforts are critical to reduce neonatal mortality.

Background

Neonatal mortality, defined as death within the first 28 days of life, represents the most critical period for child survival [1, 2]. In 2022, approximately 2.3 million newborns died globally, equating to one death every 14 s or 6,300, daily [1, 3]. These deaths comprised nearly half of under-five fatalities, with 75% occurring in the first week, including one million within the first 24 h [2]. The neonatal mortality rate (NMR), measured as deaths per 1,000 live births, declined by 54% from 37 in 1990 to 17.3 in 2022, reflecting a 12% reduction from 2015 [3]. However, progress has slowed since 2010, and 64 countries are projected to miss the Sustainable Development Goal (SDG) target of reducing NMR to 12 per 1,000 live births by 2030 without urgent action [4]. Leading causes include prematurity, birth asphyxia/trauma, neonatal infections, and congenital anomalies, accounting for 78% of deaths [2]. A prospective study in Vietnam identified infections (38%), cardio/respiratory disorders (27%), congenital abnormalities (17%), and neurological disorders (10%) as primary contributors [5].

Sub-Saharan Africa has the highest NMR globally, at 27 per 1,000 live births in 2022, down 11% from 30 in 2015 [1, 3]. Five countries—Nigeria, Democratic Republic of the Congo, Ethiopia, Tanzania, and Uganda—account for 50% of the region’s neonatal deaths [6]. Newborns in sub-Saharan Africa face a tenfold higher mortality risk compared to those in high-income countries, while South Asia’s NMR stands at 21 per 1,000 live births [2].

In Ethiopia, NMR increased from 29 in 2016 to 30 in 2019, despite a decline from 39 in 2005, as reported by the 2019 Ethiopia Mini Demographic and Health Survey [7]. Approximately three-quarters of neonatal deaths occur in the first week, with one-third on the day of birth, underscoring the need for interventions during this critical window [1, 2].

Studies in northern Ethiopia attribute 34% of deaths to prematurity and 31% to asphyxia, with early neonatal deaths driven by prematurity (37%) and late deaths by asphyxia (35%) [8]. Additional risk factors include low birth weight, respiratory distress syndrome, congenital malformations, meconium aspiration, low 5-minute Apgar scores, delayed breastfeeding, and hyperthermia [9, 10]. In Tigray, prematurity, low birth weight, asphyxia, infections, and congenital anomalies are predominant, worsened by war-related challenges like malnutrition and destroyed health facilities [11].

Effective interventions, such as kangaroo mother care, skilled birth attendance, essential newborn care, and high-quality antenatal care, are critical for reducing NMR by identifying and managing risks during pregnancy and delivery [1, 12, 13]. Ethiopia’s Health Sector Transformation Plan (HSTP) aimed for an NMR of 10 per 1,000 live births by 2020, but the current rate of 30 reflects a significant gap [14, 15]. High-impact, high-cost interventions, including care for small and sick newborns and emergency obstetric care, offer quadruple returns by reducing maternal and neonatal morbidity and mortality [1]. Despite progress toward Millennium Development Goal 4, which targeted a two-thirds reduction in child mortality by 2015, Ethiopia’s neonatal mortality remains high, requiring at least double the current reduction rate to meet SDG targets by 2030 [7, 16, 17]. The World Health Organization, in collaboration with Ethiopia’s Ministry of Health, is expanding quality services for small and sick newborns, including strengthened neonatal nursing, yet challenges persist [1].

Despite several studies conducted in the Tigray region of Ethiopia to identify predictors of neonatal mortality, significant gaps remain, underscoring the necessity for further research. Previous studies, such as those by Mengesha et al. 2016 [11], and Mengesha & Sahle 2017 [8], have primarily focused on socio-demographic and health service-related factors, often overlooking critical clinical variables like maternal HIV status, neonatal anemia, and congenital anomalies. Moreover, these studies were conducted prior to the onset of the conflict in Tigray, which has had profound impacts on healthcare infrastructure and service delivery.

The study by Hadgu et al. 2020 [18], at Ayder Comprehensive Specialized Hospital employed a cross-sectional design, limiting the ability to establish temporal relationships between risk factors and neonatal mortality. Additionally, this study did not consider variables such as antenatal care attendance, maternal comorbidities, or the timing of breastfeeding initiation. Similarly, Fisseha et al. 2025 [19], focused exclusively on preterm neonates, thereby excluding term neonates who also contribute significantly to neonatal mortality statistics.

Our study addresses these gaps by employing a retrospective cohort design to assess time-to-death and predictors of neonatal mortality among both term and preterm neonates admitted to NICUs in public general hospitals of Tigray’s Central Zone in 2024. This approach allows for a comprehensive analysis of a wide range of variables, including socio-demographic factors, obstetric and gynecological conditions, neonatal medical conditions, and maternal medical histories. By encompassing a broader population and a more extensive set of variables, our study aims to provide a more holistic understanding of the factors influencing neonatal mortality in this region.

In summary, our study fills a critical gap in the literature by providing a comprehensive analysis of neonatal mortality predictors in Tigray’s post-conflict context, encompassing a broader neonatal population and a more extensive set of variables than previous studies. The insights gained from this research are essential for developing effective strategies to improve neonatal outcomes and align with global and national health priorities.

Methodology

Study area, period, and design

A hospital-based retrospective 1 year cohort study (January 1, 2023, to December 31, 2023.) was conducted in public general hospitals in Tigray’s central zone, located 1,004 km north of Addis Ababa and 186 km from Mekelle. Aksum, the administrative center, has a population of 1,347,212 (686,161 females, 2017 zonal report). A religious and tourist hub, Aksum boosts the local economy. The zone’s three hospitals—Adwa, Aksum Saint Mary’s, and Abi Adi General Hospitals—have annual neonatal admissions of 660, 262, and 736, respectively. NICU capacities include: Adwa (10 NICU, 6 mother-side beds); Aksum Saint Mary’s (8 NICU, 7 KMC/mother-side beds); Abi Adi (7 inborn, 7 out born NICU, 5 KMC, 10 mother-side beds, adding 29). Data were collected from all hospitals from July 20 to August 20, 2024.

Populations

The source population comprised all neonates admitted to NICUs in public general hospitals in Tigray’s central zone. The study population included selected neonates admitted to NICUs at Adwa, Aksum Saint Mary’s, and AbiAdi General Hospitals from January 1, 2023, to December 31, 2023.

Eligibility criteria

The study included neonates admitted alive to NICUs at Adwa, Aksum Saint Mary’s, and Abi Adi General Hospitals from January 1, 2023, to December 31, 2023, with complete records. Neonates with incomplete records (missing admission date, time, or discharge date), lost records, or unrecorded treatment outcomes were excluded.

Sample size determination

The sample size was calculated using STATA Version 14, assuming 5% types I error, 80% power, and 0.5 standard deviation. Based on a prior study in Addis Ababa with a 0.11 event probability and an adjusted hazard ratio of 2.5 for respiratory distress, requiring 38 events, the sample size was 340 [20]. Other variables included not crying at birth, with a hazard ratio of 3.52, requiring 20 events and a sample size of 122 [21], and delayed breastfeeding initiation, with a hazard ratio of 7.5, requiring 8 events and a sample size of 135 [11]. The respiratory distress hazard ratio was selected due to its recent study in a nearby area, as it yielded the most conservative (largest) sample and would therefore provide adequate power for all our study objectives, including the analysis of associated factors. Adding a 5% non-response rate, the final sample size was 357.

Sampling technique and procedure

A systematic random sampling approach was used to select neonates from the neonatal registration logbook in all public general hospitals in Tigray’s central zone, using every fifth registration number interval, with the first sample chosen randomly. The sample size of 357 was proportionally allocated across Adwa General Hospital (660 annual admissions, 142 samples), Aksum Saint Mary’s General Hospital (262 annual admissions, 56 samples), and AbiAdi General Hospital (736 annual admissions, 159 samples), based on each hospital’s proportion of the total one-year admissions across all hospitals.

Study variables

The dependent variable was the time to death of neonates. Independent variables included socio-demographic factors (birth weight, sex, gestational age, maternal age, residence, neonatal age), obstetric and gynecological factors (mode of delivery, place of delivery, parity, prolonged premature rupture of membranes, prolonged labor, preeclampsia, placenta previa, antenatal care, type of pregnancy, obstructed labor, postpartum hemorrhage), neonatal medical conditions and related factors (perinatal asphyxia, respiratory distress syndrome or hyaline membrane disease, sepsis, first- and fifth-minute Apgar scores, resuscitation at birth, hypothermia, breastfeeding initiation time, neonatal anemia, congenital anomalies, neonatal jaundice, neonatal dehydration, crying at birth, small for gestational age), and maternal medical conditions (HIV/AIDS, hypertension, diabetes mellitus, anemia, hepatitis B virus, congestive heart failure, epilepsy, sexually transmitted infections).

Data collection tools and procedure

Data were collected using a standardized English-language data abstraction form, developed after reviewing relevant literature [11, 20–30], capturing socio-demographic characteristics of neonates and mothers, maternal medical conditions, obstetric and pregnancy-related factors, common neonatal medical disorders, admission date, length of stay, discharge date, time of neonatal death, and neonatal outcome. Neonatal chart numbers were obtained from NICU medical record books, and charts were reviewed for baseline and follow-up records. Records meeting inclusion criteria were selected, and data were extracted by six BSc nurses and two MSc nurse supervisors using a checklist, from January 1, 2023, to December 31, 2023. The follow-up period started at NICU admission and ended at death or censorship (discharge, transfer, self-discharge, or referral).

Data quality assurance

To ensure data quality, the data abstraction checklist was pretested on 17 neonates (5% of the sample) admitted to the NICU at Mekelle General Hospital one week before the study, with necessary amendments and modifications made to standardize and validate the tool. Six experienced BSc nurses working in NICUs and two MSc nurse supervisors, trained for three days on the study’s purpose, checklist, data collection methods, and ethical considerations, collected the data. Supervisors monitored the process, while the principal investigator and supervisors conducted spot-checks and reviewed completed checklists to ensure completeness and consistency, with random card selections examined for accuracy.

Data processing and analysis

Data were checked for completeness and consistency by the principal investigator and supervisors, entered using Epi Data 4.6, and analyzed with STATA 14 after cleaning, editing, and coding. Descriptive analyses provided percentages, medians, and standard deviations for dependent and independent variables, with mean ± standard deviation for normally distributed continuous covariates and median with interquartile range for skewed covariates. Findings were presented in text, tables, and figures using frequencies and summary statistics. Outcomes were dichotomized as censored or death, with survival time defined as days from admission to event or censor. Kaplan-Meier analysis estimated median survival time and cumulative survival probability, comparing differences across covariates, while the log-rank test assessed statistical survival differences between categories. Life tables estimated cumulative survival at different intervals. Bi-variable and multivariable Cox regression models evaluated associations between dependent and independent variables, with covariates having a p-value < 0.25 in bi-variable analysis included in multivariable analysis to control confounding. Multicollinearity was tested using the variance inflation factor (mean VIF = 1.67). Adjusted hazard ratios with 95% confidence intervals were computed, with statistical significance at p < 0.05. The Cox proportional hazard assumption was verified using the Schoenfeld residual test, and model fitness was assessed graphically with the Cox-Snell residual graph.

Operational definitions

Neonate; is an infant younger than 28 days of age (29).

Censored: Neonates who were discharged, transferred to other hospitals or other ward, or self-discharged were considered censored (22).

Survival time: time in days from the date of admission to NICU to the occurrence of outcome (event or censored) (24).

Time to death or event; is the time from admission at the NICU to the occurrence event (20).

Event: death after NICU admission (20).

Length of stay: The number of days the child stayed in hospital from admission until the child develop event of interest (death) or censored.

Low birth weight; defined as birth weight less than 2500 gm (30).

Recovered; were neonates that declared as improved or recovered by physician (31).

Perinatal asphyxia (PNA); was considered when the 5th APGAR score is < 7 or a neonate who did not cry or needed resuscitation (32).

Result

Of the 357 neonates studied, approximately two-thirds (228, 63.87%) were male, and three-quarters (258, 72.26%) were admitted within 24 h of birth. Over half (219, 61.34%) had a gestational age greater than 37 weeks, with gestational ages ranging from 28 to 43 weeks. Additionally, more than half (217, 60.78%) resided in rural areas, and 289 (80.95%) had mothers aged 20–34 years (Table 1).

Table 1.

Distribution of Socio-demographic characteristics of neonate and their mother in general public hospitals of central zone of Tigray, Ethiopia, 2024(n = 357)

Category	Variable	Status		Total N (%)
		Death (N, %)	Censored (N, %)
Sex	Male	36(60)	192(64.65)	228 (63.87)
	Female	24(40)	105(35.35)	129(36.13)
Age of the neonate	< 24 h	54(90)	204(68.69)	258(72.26)
	24 h − 7 day	4(6.67)	48(16.16) `	52(14.56)
	8–14 days	0	26(8.75)	26(7.2)
	> 15 days	2(3.33)	19(6.4)	21(5.88)
Birth weight in grams	< 2500	39(65)	146(49.16)	185(51.82)
	>=2500	21(35)	151(50.84)	172(48.18)
Gestational age in weeks	28–32	15(25)	42(14.14)	57(15.97)
	33–36	19(22.69)	62(20.88)	81(22.6)
	> 37	26(43.33)	193(64.98)	219(61.34)
Age of the mother in years	< 20	6(10)	29(9.76)	35(9.8)
	20–34	48(80)	241(81.14)	289(80.95)
	> 35	6(10)	27(9.09	33(9.24)
Residence	Rural	48(80)	169(56.9)	217(60.78)
	Urban	12(20)	128(43.1)	140(39.21)

Among the mothers enrolled in the study, 29 (8.12%) had HIV/AIDS, 14 (3.92%) had hypertension, 15 (4.2%) had hepatitis B virus (HBV), and 13 (3.64%) had anemia during pregnancy. Additionally, 14 (3.92%) had congestive heart failure (CHF), 13 (3.64%) had mental health issues, and 18 (5.04%) had sexually transmitted infections (STIs) (Table 2).

Table 2.

Distribution of maternal medical diagnosis in NICU in general public hospitals of central zone of Tigray, Ethiopia, 2024(n = 357)

Variables	Category	Status		Total (N,%)
		Died (%)	Censored
HIV positive mother	Yes	13(21.67)	16(5.39	29(8.12)
	No	47(78.33)	281(94.61)	328(91.88)
Hypertension	Yes	5(8.33)	9(3.03)	14(3.92)
	No	55(91.67)	287(96.93)	343(96.08)
Diabetes mellitus	Yes	3(5)	10(3.37)	13(3.64)
	No	57(95%)	287(96.63)	344(96.37
Anemia	Yes	2(3.3)	11(3.70)	13(3.64)
	No	58(96.67)	286(96.3)	344(96.37)
HBV	Yes	5(8.33)	10(3.37)	15(4.2)
	No	55(91.67)	287(96.63)	342(95,8)
CHF	Yes	0	14(4.71)	14(3.92)
	No	60(100%)	283(95.29)	343(96.08)
Mental health problem	Yes	2(3.33)	11(3.70)	13(3.64)
	No	58(96.67)	286(96.30)	344(96.37)
STI	Yes	2(3.33)	16(5.39)	18(5.04)
	No	58(96.67)	281(74.61)	339(94.96)

[Note=Diabetes mellitus, HIV=Human immune deficiency virus, HBV= Hepatitis B virus, CHF=Congestive heart failure, STI=sexually transmitted infection]

A total of 357 neonates that were admitted to NICU had been followed from 0 to 28 days. And the median survival time was 27 days. During the follow up period 60(16.8%) died, and 297(83.2%) were censored with 241(67.5%) discharged after improved, 25(7%) referred, 19(5.3%) transferred to Pediatric ward and12 (3.3%) were self-discharged or left against medical advice (Table 3).

Table 3.

Over all life with seven intervals for time to death and predictors of neonatal mortality among neonates admitted to NICU in general public hospitals of central zone of Tigray, Ethiopia, 2024(n = 357)

Interval in days	Beginning total	Deaths	Lost	Survival	Std error	95% CI
0–7	357	36	121	0.8786	0.0190	0.8357 ,0.9109
7–14	200	12	106	0.8069	0.0264	0.7487 ,0.8529
14–21	82	3	34	0.7696	0.0328	0.6975, 0.8267
21–28	51	9	36	0.5387	0.0684	0.3967 ,0.6612

The Kaplan-Meier survival curve being high and flatter suggests a better long-term survival probability for the neonates, indicating a lower risk of death during the 28-day study period. This pattern reflects a slower decline in survival probability over time, with the curve remaining elevated, particularly in the early days (e.g., 97.1% on day 1, 87.86% on day 7, and 53.87% on day 28), signifying a relatively lower incidence of the event (death) throughout the follow-up ( Fig. 1).

Fig. 1.

Overall Kaplan-Meier survival estimate of neonate admitted to NICU of public general hospitals of central zone of Tigray, Ethiopia, 2024(n = 357)

Survival analysis

Survival from birth to 28 days was estimated using the Kaplan–Meier method. The median survival time was 27 days, defined as the time point at which the estimated survival probability reached 50%. Although the estimated survival probability at day 28 was 53.87%, the median survival time was reached by interpolation between day 27 and day 28 based on the observed timing of deaths. This reflects the concentration of mortality events late in the neonatal period. We will clarify this point explicitly in the revised Results section to improve transparency and prevent misinterpretation.

The Kaplan-Meier survival analysis indicated that neonates born to mothers who received antenatal care (ANC) during pregnancy had a higher median survival time of 27 days compared to those whose mothers did not, with a median survival time of 9 days. The survival probability on the first day was 98.1% for the ANC group and 85.71% for the non-ANC group. This difference was statistically significant (p-value = 0.000) (Fig. 2).

Fig. 2.

The Kaplan-Meier survival curves compare survival time of neonate with groups of Antenata care (ANC) at NICU general public hospitals of central zone of Tigary, Ethiopia, 2024(n = 357)

In the study, predictors of neonatal mortality were identified using log-rank tests (p-value < 0.05) for bivariable Cox regression analysis, with variables having a p-value < 0.25 included in the multivariable analysis. The multivariable Cox regression revealed the following predictors of neonatal mortality: Neonates born to mothers residing in rural areas had a three-fold increased hazard of death (AHR = 3.44, 95% CI: 1.55, 7.66) compared to those from urban areas. Neonates born to mothers without antenatal care (ANC) follow-up during pregnancy faced a three-fold higher risk of death (AHR = 2.80, 95% CI: 12.20, 6.51) compared to those with ANC follow-up. Positive maternal HIV status was associated with a three-fold higher risk of neonatal death (AHR = 2.93, 95% CI: 1.27, 6.72) compared to negative maternal HIV status. Neonates who did not initiate breastfeeding within one hour of birth had a two-fold increased hazard of death (AHR = 2.12, 95% CI: 1.07, 4.20) compared to those who did (Table 4).

Table 4.

Bivariable and multivariable cox regression analysis among admitted neonates to NICU general public hospitals of central zone of Tigray, Ethiopia, 2024(n = 357)

Variables	Categories	Death (N, %)	Censored (N, %)	CHR (95%CI)	P-value	AHR (95%CI)	P-value
Residence	Rural	48(80)	169(56.9)	2.35(1.25,4.4)	0.008	3.44(1.55,7.66) *	0.002
	Urban	12(20)	128(43.1)	1
Antenatal care	Yes	46(76.67)	283(95.29)	1
	No	14(23.33)	14(4.71)	3.6(1.97,6.6)	0.000	2.80(1.22, 6.51) *	0.017
HIV positive mother	Yes	13(21.67)	16(5.39	3.35(1.8,6.2)	0.000	2.93(1.27,6.72) *	0.011
	No	47(78.33)	281(94.61)	1
Bag and mask done	Yes	28(46.67)	75(25.25)	2.28(1.36,3.80)	0.002	1.17(0.43,3.12)	0.751
	No	32(53.33)	222(74.75)	1
Crying	Yes	18(30)	213(71.74)	1
	No	42(70)	84(28.28)	4.08(2.34,710)	0.000	1.50(0.53,4.24)	0.77
Initiation of breast feeding	Yes	16(26.67)	214(72.05)	1
	No	44(73.33)	83(27.95)	4.27(2.40,7.60)	0.000	2.12(1.07,4.20) *	0.030
Fifth minute APGAR score	< 7	25(55.56)	49(26.92)	2.77(1.54,5)0.00	0.001	1.86(0.84,4.10)	0.121
	>=7	20(44.44)	133(73.08)	1
Hypothermia	Yes	50(83.33)	182(61.28)	2.54(1.29,5.02)	0.007	1.28(0.52,3.19)	0.582
	No	10(16.67)	115(38.72)	1

[Note: APGAR = HIV=Human immune deficiency virus Appearance, Pulse, Grimace, Activity, Respiration, CHR: Crude Hazard Ratio, AHR: Adjusted Hazard Ratio; CI: Confidence Interval]

NB: The asterisk (*) above showed significantly associated predictors of time to death

The Cox proportional hazard assumption was verified using the Schoenfeld global test for the full model, which was satisfied (p-value = 0.6279). All covariates met the proportional hazard assumption, as their p-values were greater than 0.05. The model’s goodness of fit was assessed using Cox-Snell residuals, confirming that the final model adequately fit the data, as illustrated in the hazard function following a 45-degree line in the figure below (Fig. 3).

Fig. 3.

Cox-Snell residual cumulative hazard graph on neonates admitted in NICU general public hospitals of central zone of Tigray, Ethiopia, 2024(n = 357)

Discussion

This retrospective follow-up study investigated the survival status and predictors of mortality among 357 neonates admitted to the NICU in public hospitals in the central zone of Tigray, Ethiopia, over a total of 3,388 neonatal days. In the current study, conducted in public general hospitals of Tigray’s Central Zone, 16.8% of neonates died, yielding an incidence rate of 17.7 deaths per 1,000 neonate-days (95% CI: 13.7–22.8) with a median survival time of 27 days. This mortality rate is comparable to a study from Eastern Ethiopia, which reported a facility-based neonatal mortality of 20% (95% CI: 16.7–23.8%) among NICU admissions [9]. The slight variation could be attributed to differences in the case mix, particularly in causes of admission such as prematurity and infection, which were leading causes in that setting. Compared to Jimma Medical Center [10], which reported a slightly lower mortality proportion of 13.3%, the difference may reflect better-equipped neonatal care units or greater availability of specialized personnel and interventions at tertiary-level centers compared to general hospitals. Additionally, differences in referral systems and care-seeking behaviors might contribute to improved outcomes in Jimma. However, the mortality rate in the current study is much lower than that reported in another study from Northern Tigray [11], where the neonatal mortality rate was 62.5 per 1,000 live births. Although this rate is based on live births and not NICU admissions, the higher proportion of early neonatal deaths (73.52% within 7 days) reflects delays in seeking care, resource limitations, or a higher prevalence of severe complications among the population. The median survival time of 27 days in this study suggests that most neonatal deaths occurred in the late neonatal period, differing from [11] where the majority of deaths occurred within the first 7 days. This might indicate better immediate postnatal care, or possibly that admissions to NICU occurred after the early neonatal period in our setting, thus shifting the survival pattern. However, Conditions such as delayed or inadequate breastfeeding, untreated infections, and neonatal jaundice may contribute to this pattern. Furthermore, reduced access to routine postnatal care and follow-up services during the conflict period in the study area may have limited early detection and timely management of these conditions, thereby increasing late-neonatal mortality.

The overall neonatal mortality rate was 16.8% (60 neonates), equivalent to 17.7 deaths per 1,000 admitted neonates (95% CI: 13.7, 22.8), which was approximately three times higher than a pre-war study in Tigray (6.25%) and exceeded rates reported in Addis Ababa tertiary hospitals (11.1%) [20], Bombe Primary Hospital (10.6%) [25], and Wollega University Referral Hospital (15.3%) [26]. The elevated mortality may be attributed to differences in methodology, study period, and the impact of the war in Tigray, which led to the destruction of health facilities, displacement of healthcare professionals, and limited access to medical equipment and trained personnel. The incidence of neonatal mortality was 17.7 per 1,000 neonatal days, higher than the 6.81 per 1,000 person-days reported at Wollega University Referral Hospital [26]. The median survival time was 27 days, consistent with findings from Wollega [26] but higher than the 17 days reported in Addis Ababa tertiary hospitals [20], likely due to shorter follow-up periods and the referral of critical cases from general hospitals in Gondar [21] and Addis Ababa [20]. The overall survival probability at 28 days was 53.78%, with survival probabilities of 97.1%, 87.86%, and 80.69% on days [1, 4, and 8] respectively. These figures align closely with a study from Arba Minch General Hospital [24], which reported survival probabilities of 96.1%, 75%, and 69.9% on days [1, 4, and 8] respectively, and an overall survival probability of 66.2%.

Multivariate Cox proportional regression analysis identified significant predictors of neonatal mortality: rural residence, lack of antenatal care (ANC), positive maternal HIV status, and delayed initiation of breastfeeding. Neonates from rural mothers had a three-fold higher hazard of death (AHR = 3.44, 95% CI: 1.55–7.66). This is consistent with findings from Jimma Medical Center [10], where neonates from outside Jimma city had a higher mortality risk (AOR = 1.89). This supports the hypothesis that geographic barriers, transportation delays, and limited maternal health service utilization in rural areas contribute to poor neonatal outcomes. In addition, greater distance from functional health facilities (especially critical for emergency obstetric and neonatal care), lower household income affecting nutrition and care-seeking, and potentially lower maternal literacy.

Lack of ANC follow-up was another significant predictor in our study (AHR = 2.80). This aligns with the general understanding that ANC visits provide critical opportunities to detect maternal infections, fetal complications, and educate mothers, thus reducing neonatal risks. In [11], maternal complications were protective (AHR = 0.37), likely because those who had complications were also more likely to have accessed ANC and received timely interventions. Neonates born to HIV-positive mothers were nearly three times more likely to die (AHR = 2.93). While not directly examined in the three comparative studies, this association is supported by global literature indicating that maternal HIV is linked to preterm labor, low birth weight, and increased neonatal infections, all of which increase the risk of mortality. The lack of PMTCT services or poor ART adherence might further exacerbate this risk in rural and resource-limited settings. Moreover, as interrupted maternal Antiretroviral Therapy (ART), lack of infant prophylaxis, and the potential for increased maternal morbidity, all of which could directly and indirectly elevate neonatal mortality risk in this crisis setting.

Delayed initiation of breastfeeding beyond 1 h after birth was associated with a 2.12-fold higher risk of death. This finding is strongly supported by [11], which found that not initiating exclusive breastfeeding led to an even higher hazard (AHR = 7.5). Early breastfeeding has been shown to reduce neonatal infections, promote thermoregulation, and improve survival particularly in low-resource settings. Therefore, this reinforces the need to strengthen early breastfeeding counseling and immediate newborn care practices. Compared to urban counterparts, consistent with findings from Wollega University Referral Hospital (AHR = 2.04, 95% CI: 1.14–3.66) [26]. This may reflect poor road infrastructure and limited access to healthcare in rural areas.

Neonates born to mothers without ANC follow-up faced a three-fold increased risk of death (AHR = 2. 80, 95% CI: 1.22–6.51). This finding aligns with studies from Addis Ababa [20] and Wollega [26], which also reported higher neonatal mortality associated with lack of ANC follow-up. Antenatal care facilitates early detection and management of maternal and obstetric complications, thereby reducing neonatal risks.

Strengths and limitations

The study demonstrates several strengths that enhance its reliability and utility. Data collection was handled by nurses specifically trained in neonatal intensive care unit (NICU) practices, which significantly improved the overall quality of the information gathered. Leveraging existing data streams streamlined the process, making it more efficient and straightforward. Furthermore, the research offers important insights that can inform and guide future prospective studies, while also establishing clear temporal links between neonatal death as the primary outcome and various predictor variables documented at the time of admission.

However, the study is not without its limitations, which should be considered when interpreting the results. Notably, it excluded key factors such as community and environmental influences that could play a substantial role in neonatal mortality outcomes. Additionally, the investigation was confined to the institutional setting of general hospitals, failing to address mortality rates at the community level or in lower-tier facilities like primary hospitals and health centers, which may lead to an underestimation of the true prevalence. The findings, derived from general hospitals in the central zone of Tigray, may not be broadly applicable to the entire region. Finally, the exclusion of incomplete or lost medical records could introduce selection bias, potentially resulting in either an underestimation or overestimation of the study’s conclusions. As neonates who were referred, transferred, or self-discharged were lost to follow-up. Their unknown outcomes beyond the NICU may introduce bias if their mortality risk differed systematically from those observed.

Conclusion

In this study, 60 (16.8%) of the 357 neonates died during the follow-up period, with an overall neonatal mortality incidence of 17.7 deaths per 1,000 neonate-day observations and a median survival time of 27 days. The high mortality rate may be attributed to the destruction of health systems and displacement of healthcare workers due to the war in Tigray, with a notable increase in deaths during the early neonatal period. The cumulative survival probability at the end of the 28-day study period was 53.78%. Independent predictors of neonatal mortality included rural residence (AHR = 3.44, 95% CI: 1.55, 7.66), lack of antenatal care (ANC) follow-up (AHR = 2.80, 95% CI: 1.22, 6.51), delayed initiation of breastfeeding within one hour of birth (AHR = 2.12, 95% CI: 1.07, 4.20), and positive maternal HIV status (AHR = 2.93, 95% CI: 1.27, 6.72).

Recommendation

Given the concerning neonatal mortality rate of 16.8% and key risk factors identified including rural residence, lack of antenatal care, delayed breastfeeding, and maternal HIV status we recommend focused action. This entails rebuilding health systems in conflict zones with mobile and community-based units, improving rural access and antenatal education, training providers to support timely breastfeeding, and scaling up integrated HIV care. Achieving sustainable impact requires policy support for further research and strong multi-sector collaboration to protect vulnerable newborns.

Biographies

Tsehaye Hailemariam Teklu

Tsehaye Hailemariam Teklu serves as a Lecturer at the School of Nursing, College of Health Sciences, Haramaya University, Ethiopia. With a Master of Science in Nursing, he brings expertise in nursing education and research, focusing on maternal and child health, clinical nursing practices, and healthcare education. His work is dedicated to advancing nursing education and enhancing healthcare services through research and teaching, contributing to improved health outcomes in Ethiopia.

Mekuria Kassa Nerea

Mekuria Kassa Nerea serves as an Assistant Professor at the School of Nursing, College of Health Sciences, Mekelle University, Ethiopia. Holding a Master of Science in Pediatric and Child Health Nursing, he brings over ten years of expertise in nursing education and research, medical volunteer. His scholarly interests focus on pediatric health, maternal and child health, and public health, with multiple publications in peer-reviewed journals. His research contributions include investigations into treatment failure among children on antiretroviral therapy and factors influencing neonatal mortality in intensive care settings.

Almaz Berhe Weldemicheal

Almaz Berhe Weldemicheal is an Assistant Professor at the School of Nursing, College of Health Sciences, Mekelle University, Ethiopia. With a Master of Science in Pediatrics and Child Health Nursing, she possesses extensive experience in nursing education and research. Her scholarly interests center on maternal and child health, evidenced by multiple publications in peer-reviewed journals on topics such as postnatal care services, antenatal care initiation, and factors influencing neonatal outcomes. She has notably contributed to research on the quality of postnatal care and adherence to prevention of mother-to-child transmission (PMTCT) programs in Tigray, Ethiopia.

Author contributions

TH conceptualized the research, drafted the manuscript, and conducted the analysis and interpretation. MK contributed to drafting the manuscript and critically revised the study design and analysis. AB participated in revising the study design and analysis. All authors reviewed and approved the final manuscript.

Funding

There was no funding source available for this research.

Data availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study adhered to ethical standards as outlined in the Declaration of Helsinki and was conducted in accordance with the ethical guidelines of Mekelle University. The research was approved by the Institutional Review Board (IRB) of Mekelle University, College of Health Sciences, under protocol number ERC 2279/2024. Ethical approval for this study was granted by the Institutional Review Board (IRB) of the College of Health Sciences, Mekelle University, Ethiopia, following a thorough review of the research protocol to ensure adherence to ethical standards for human subjects research. The university issued formal cooperation letters to the Tigray Regional Health Bureau, which authorized the study in three public general hospitals in the Central Zone of Tigray. Research activities began after obtaining official permissions from these hospitals. The study’s objectives were clearly communicated to hospital administrators, who provided consent for accessing neonatal medical records.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.