A retrospective follow-up study was conducted from 7–30 March 2023 at Woldia Comprehensive Specialized Hospital (WCSH). It is located approximately 521 km and 360 km from Addis Ababa and Bahir Dar, respectively (20, 21). There were 10 nurses, two neonatal nurses, two general practitioners, and two pediatricians who provided services for admitted neonates. Annually, 750–1,000 neonates are admitted because of various health problems. Approximately 90% of neonates were admitted from the labor ward and approximately 10% were referred from nearby health institutions (22).

The source populations were neonates within 28 days of diagnosis of neonatal sepsis who had been admitted to the NICU of WCSH.

The study populations were neonates within 28 days of being diagnosed with neonatal sepsis who had been admitted to the NICU of WCSH between January 2021 and December 2022.

The study excluded neonates who died within 30 min of admission.

The sample size was calculated using STATA software version 14, a sample size for time-to-event data, considering a 95% confidence interval (CI), alpha (0.05), probability of recovery of 0.8098, a hazard ratio for the determinant factors, percentage of recovery, and power of 80% (0.80). We used three variables to estimate the sample size. The sample sizes for the three variables, namely birth weight, intrapartum fever, and time of infection onset, were 351, 313, and 121, respectively, at the Central Gondar Public Hospital (16). Thus, the final sample size was the largest among the 351 mother–newborn pairs.

A systematic random sampling technique was used to recruit the study participants. Using medical numbers of neonates admitted to NICU from January 2021 to December 2022 with neonatal sepsis, a sampling frame was prepared separately for each year. The final sample size was proportionally allocated to each year (124 and 227 neonates in 2021 and 2022, respectively). Lastly, a systematic random sampling technique was used (K = N/n = 774/351 = 2.205 ≈ 2) to select the study participants from the sampling frame.

The dependent variable was the time to recovery from neonatal sepsis, defined as the time from admission to discharge when the neonate had recovered from NS (1, 15, 16, 23, 24).

Socio-demographic variables: maternal age, residence, age of neonate at admission, and neonate sex.

Maternal-related variables: parity, gravidity, onset of labor, duration of labor, mode of delivery, place of delivery, ANC visits, multiple pregnancies, Pregnancy-Induced Hypertension (PIH), antepartum hemorrhage, intrapartum fever, duration of PROM, maternal infection history, and chronic illness.

Neonate-related variables: birth weight, gestation of neonate at birth, admission weight, temperature, 1-minute and 5-minute Apgar scores, congenital anomalies, resuscitation at birth, meconium aspiration syndrome, respiratory distress, and kept in kangaroo mother care within one hour of birth.

Clinical & medical care-related and investigation variables: jaundice, cyanosis, and white blood cell count in the complete blood profile, onset of infection, enteral feeding, critical conditions, and outcome status.

Healthcare service-related variables: the prompt initiation of treatment and the timing of seeking medical care after the neonate fell ill.

Recovery: if a neonate recovered from the infection after completing treatment, according to the physician's diagnosis.

Median time to recovery: the average duration it took for neonates to be declared as recovered by the attending physicians in the unit.

Neonatal Sepsis: if the neonate was diagnosed as having neonatal sepsis by the attending physician after taking a detailed medical history, physical examination, and laboratory tests based on the integrated management of neonatal and childhood illness criteria. The criteria include the presence of two or more persistent fevers (≥37.5°C) or persistent hypothermia (≤35.5°C) lasting for more than 1 h, rapid breathing (≥60 breaths per minute), severe chest indrawing, grunting, poor feeding, movement only when stimulated, a bulged fontanelle, convulsions, lethargy, or unconsciousness. In addition, two or more of the following hematological criteria were used: total leukocyte count (<4,000 or >12,000 cells/mm³), absolute neutrophil count (<1,500 cells/mm³ or >7,500 cells/mm³), platelet count (<150 or >450 cells/mm³), and random blood sugar (<40 mg/dl or >125 mg/dl) (4, 25, 26).

Defaulter: A neonate is considered a defaulter when they voluntarily stop or leave the treatment unit against medical advice or without completing the prescribed treatment.

Congenital anomalies: Congenital anomalies refer to structural or functional abnormalities that are present in a neonate at birth. These anomalies can involve various body systems, such as heart defects, neural tube defects, and genetic conditions like Down syndrome.

Death: In the context of this study, death refers to the unfortunate event of a neonate passing away either during the treatment period or while still admitted to the treatment unit due to neonatal sepsis.

Censored: A neonate is considered censored when they either default from the treatment, experience death, or are transferred out of the treatment unit. Censored cases are no longer actively followed up or included in the analysis of outcomes.

Length of stay: The length of stay refers to the duration in days that a neonate remains in the hospital from the time of admission until the occurrence of an event of interest (such as recovery) or until the neonate is censored (due to defaulter status, death, or transfer out).

Early initiation of treatment: Initiate broad-spectrum antimicrobials within the first hour of diagnosis.

Prolonged rupture of membrane: The time from membrane rupture to delivery >18 h (27).

Normal WBC range: the normal range of WBC is from 5,000–12,000 cells/µl.

Early initiation of treatment: initiate broad-spectrum antimicrobials within the first hour of diagnosis.

Supportive care: giving care or support with oxygen and feeding without drug management.

Critical conditions: when a neonate is hypoxic, hypoglycemic, or unable to maintain a normal body temperature.

The data for this study were collected using a checklist that was developed based on previous studies. To gather the necessary information, the medical records of both the mothers and neonates were thoroughly reviewed. The checklist encompassed various domains, including neonatal and maternal socio-demographic characteristics, maternal health-related factors, neonatal health-related factors, healthcare service-related factors, and clinical and medical care-related factors. These domains were carefully selected to provide a comprehensive understanding of the factors that may influence the outcomes of neonates with sepsis.

Prior to commencing the data collection process, a preliminary review was conducted on a subset of the sample, specifically 15 neonates, which accounted for 5% of the total sample size. This review served as a pilot study to ensure the effectiveness and feasibility of the data collection tool. The data collection tool was prepared in English, considering the language proficiency of the nursing professionals involved in the study. These professionals were recruited specifically for the purpose of extracting variables from the medical records of the neonate and mother. To ensure the accuracy and consistency of data collection, the investigators provided a comprehensive one-day training session to the nursing professionals. This training covered the purpose and objectives of the study, the techniques for data collection, and the ethical considerations involved in handling patient information.

Throughout the data collection period, the investigators closely monitored the completeness and consistency of the collected data. Daily checks were conducted to identify any discrepancies or missing information, and necessary corrections were made promptly. By conducting a preliminary review, providing appropriate training, and implementing rigorous quality control measures, the study aimed to maintain the integrity and reliability of the collected data.

After the data collection process, the collected data underwent a series of cleaning, coding, and entry procedures using Epi Data version 4.6. To determine the factors associated with recovery time from neonatal sepsis, Cox regression analysis was conducted. This statistical method allows for the examination of time-to-event data and was deemed appropriate for this study. Initially, variables with a p-value less than 0.25 were selected as candidates for the subsequent multivariate Cox regression analysis. In the multivariate analysis, variables with a p-value less than 0.05 were considered statistically significant, and their results were reported with 95% confidence intervals. This approach ensured that only the most relevant and impactful factors were considered in determining recovery time. Additionally, to address any potential issue of multicollinearity, the time required for recovery was taken into account during the analysis. Finally, the data were presented in figure, table, and narrative form.

A total of 124 out of 273 neonates and 227 out of 501 admitted neonates in the years 2021 and 2022, respectively, were included in the study (a total of 351 neonates diagnosed with sepsis). The majority of the mothers of these neonates were aged 25–29 (31.91%), followed by those aged 20–24 (29.06%). The mean age of the neonates upon admission was 42 h (1.75 days), and the mean age of the mothers was 26.9 years. Out of the total number of neonates in the study, 270 (76.29%) were only one day old, and 221 (62.96%) were male. More than half of the respondents were living in rural areas (52.71%) (Table 1).

Of the neonates admitted to the NICU with NS, 299 (85.19%) were delivered at the same facility. More than half of the mothers were primiparous (51.28%), and the majority had ANC visits. Furthermore, 28 (7.98%) mothers had a history of infection during their pregnancy, and 40 (11.4%) of the mothers gave birth to twins, with either one developing neonatal sepsis. Among the mothers of admitted neonates, 78 (22.22%) had prolonged labor (Table 2).

Among the neonates admitted with NS, 248 (70.66%) were at 37–42 weeks of gestational age and 134 (38.18%) weighed less than 2.5 kg at admission. In total, 127 (36.18%) of them had a 1-minute Apgar score of seven or above. Approximately 11 neonates with neonatal sepsis (3.13%) had a congenital abnormality (Table 3).

Among the neonates admitted with sepsis, 317 (90.31%) had early-onset neonatal sepsis and 267 (76.07%) initiated treatment within 1 h. Among those with CBC profiles, 170 (48.43%) were within the normal range of WBC counts (Table 4).

Among the 351 study participants, 276 (78.63%) successfully recovered from NS, 45 (12.82%) died, 11 (3.13%) were referred, and 19 (5.41%) defaulted/left against medical advice (Figure 1). The mean time to recovery was nearly 7 days, with IQR = 5–10 days, and the median time was 6 days. For preterm neonates included in the study, the mean time to recovery was nearly 8 days (7.71 days) and the median recovery time was 8 days, with IQR from 6 days to 10 days.

Kaplan–Meier survival proportional hazard graph was used to check the assumption of Cox proportional hazard. The Kaplan–Meier survival curve, performed on the time to recovery in neonates with sepsis based on resuscitation at birth, showed that recovery was delayed among neonates who were resuscitated at birth compared to those who were not resuscitated (Figure 2).

The Cox proportional hazards regression survival graph shows the time to recovery of neonates with sepsis, based on the onset of sepsis. Therefore, in neonates with sepsis, EONS has a higher probability of recovery than LONS (Figure 3). In addition, the global test was performed with a χ² of 5.92 and a p-value of 0.2056, which is greater than 0.05. The dependent variable was std. err of 2.668803 and 95% CI: (170.2511, 180.7489).

After testing, each variable in the study with bivariable analysis (p ≤ 0.25) was entered into the multivariable Cox regression analysis (Table 5). Multivariable Cox regression model analysis was performed to show the association between covariates and time to recovery from neonatal sepsis.

Labor Induction [AHR = 0.538, 95% CI: (0.369, 0.78)] and resuscitation at birth [AHR = 0.7, 95% CI: (0.51, 0.974)] have a significant association with the recovery time of neonates. Cox proportional hazard regression analysis showed that the chance of recovery decreased as the length of hospital stay increased.

The diagnosis of NS depended on clinical symptoms, which may result in an increase in false-positive cases, potentially affecting the results of the study.