Predictors of mortality among neonates hospitalized with neonatal sepsis: a case control study from southern Ethiopia

Institution-based unmatched case-control study was conducted among neonates admitted in.

NICU in Durame General Hospital (DGH) from March 8 to 30, 2020.

Durame General Hospital is one of the four hospitals in Kembata-Tambaro zone situated in South Nations Nationalities and Peoples’ regions, which is located 352 km from Addis Ababa, the capital. The hospital provides services for approximately 1.5 million people in its catchment area. The pediatrics department has 6 units that include outpatient and follow-up units, pediatrics emergency triage, assessment, and treatment unit (ETAT), neonatal intensive care unit (NICU), surgical unit, nutritional rehabilitation unit, and medical unit. The NICU is rendering service under critical newborn care unit, septic ward, kangaroo mother care (KMC), and mother side and receives 50 to 90 neonates monthly. Investigational modalities including electrocardiography (ECG), ultrasound, x-ray, and basic hematologic and chemistry tests were readily available.

In this study, cases were defined as neonates who were diagnosed with neonatal sepsis by the attending physician and died (registered as sepsis related death by attending physician). Also, in this study, hematological criteria along with the established IMNCI (Integrated Management of Neonatal and Childhood Illness), clinical features of neonatal sepsis were used to diagnose neonatal sepsis. Neonates in the presence of one or more of the established IMNCI clinical features [either of fever (> 38 °C) or hypothermia (< 36 °C), fast breathing (> 60 breath per minute), severe chest indrawing, poor feeding, the movement only when stimulated, convulsion, lethargic or unconscious] along with two of the hematological criteria; total leukocyte count(< 5000 or > 12,000 cells/m3, absolute Neutrophil count (< 1500 cells/mm3 or > 7500cells/mm3), erythrocyte sedimentation rate (ESR) (> 15/1 h) and platelet count (< 150 or > 440cells/mm3) were considered as having neonatal sepsis. Controls were defined as newborns admitted with a diagnosis of neonatal sepsis who were discharged alive (improved). The source population for both cases and controls were the whole neonates admitted to the Neonatal Intensive Care Unit (NICU) with the diagnosis of either early-onset or late-onset neonatal sepsis in the hospital. All neonates with features of sepsis (diagnosed as either LONS or EONS) who were admitted at the neonatal unit of Durame general hospital during the study period were included in the study. Neonates with incomplete records or chart didn’t available at the time of data collection and those neonates who were transferred to other hospitals or referred before the outcomes were assessed as well as neonates left the hospital against medical advises (withdrawn treatment) were excluded from the study.

The sample size was determined by using Epi info version 7 stat calc programs by considering the following assumptions: Power 80, 95% confidence interval, 1:3 case to control ratio and using abdominal distension predictor, percent of controls who were exposed were 44.1%, odds ratio of 2.7 and 10% for incomplete records. This yields a sample size of 219 (55 cases and 164 controls).

Cases were selected by taking the deaths of neonates consecutively among those newborn infants admitted with the diagnosis of neonatal sepsis in the neonatal intensive care unit of the hospital until the sample size was achieved. This retrospective sampling covers a period of 1 year extended from January 1st to December 31st, 2019. The next immediate three corresponding controls were selected by lottery method from the NICU case registration book.

The lists of sampled neonatal septicemia patients’ medical record numbers were retrieved from the neonatal intensive care unit case registration book. The patients’ medical records (charts) were then collected from the hospital registry and checked for inclusion criteria. The medical records of eligible patients were reviewed and information was transferred into the data collection form (checklist) by the data collectors. The available data on the patient chart, NICU case registration book is observed, and an appropriate data extraction checklist is prepared in English. The checklist was adapted from the national neonatal registration book and previous related studies [13, 15]. Data were collected by two data collectors who have experience in data collection (one bachelor’s degree holder nurse and one specialty in neonatology nurse) and one supervisor (general practitioner) using the structured checklist. The data collectors were trained for 2 days on the objectives of the study, the selection of study participants (card), how to keep confidentiality of information, the contents of the questionnaire, and how to fill the data collection format by the principal investigator. Intensive supervision was maintained during the whole period of data collection. The pretest was performed on a 5% random sample of the registration form by the principal investigator to confirm the reliability of the data before the actual data collection. Proper coding and categorization of data were maintained for the quality of the data to be analyzed. Double data entry was used to ensure data quality.

Data were checked for completeness and consistencies, and then it was coded and entered into EpiData version 3.1 and it was exported to SPSS windows version 20 for analysis. Descriptive statistical techniques were used to obtain summary values for cases and controls separately. Bivariate analysis was performed to identify the crude association between dependent and independent variables. Then variables that show association in the bivariate model (p < 0.25) were entered and analyzed in a multivariable logistic regression model by using a backward stepwise method to identify the predictors of sepsis-related neonatal mortality. Model fitness was evaluated through inspection of Hosmer–Lemeshow statistic test and provided (P = 0.514), which implies that the model’s estimates fit the data at an acceptable level. Odds ratio (OR) with a 95% confidence interval was used to assess the strength and direction of the association between factors associated with the occurrence of neonatal mortality. Statistical significance was declared at P < 0.05.

This study revealed that more than nine-tenths of mothers 52 (94.5%) of cases and 156 (95.1%) of controls ever had ANC service during their pregnancy of the current neonate. The proportion of mothers who got ANC service less than three times was higher in cases 10 (18.2%) than in controls 21 (12.8%). Similarly, the proportion of women who had multiple births (twin birth and above) was higher in cases 3 (5.5%) compared to controls 6 (3.7%). More than nine-tenths of women had given birth at a health facilities 51 (92.7%) of cases and 150 (91.5%) of controls. Also in this study 22 (40%) of the mothers among the cases and nearly three fourth among the controls 118 (72.0%) were in the gestational age group of 37–42 completed weeks (term) whereas the proportion of mothers with gestational age < 37 completed weeks was higher in cases 33 (60%) than controls 46 (28%). Regarding mode of delivery, about three fourth (73.2%) of controls and two-third (67.3%) of cases had a spontaneous vaginal delivery (Table 2).

In this study the proportion of neonates with first- and fifth-minute Apgar score below 7 (low Apgar score) was 27 (49.1%) and 16 (29.1%) in cases which were higher than controls 45 (27.4%) and 17 (10.4%) respectively. Among clinical features, most neonates 46 (83.6%) of cases and 106 (64.6%) of the controls had poor feeding at presentation. More than half of the cases 30 (54.5%) were presented with a hypothermic appearance of their body temperature whereas most of the controls 81 (49.4%) were presented with normal body temperature. Regarding associated comorbidities, perinatal asphyxia is more common among the cases 12 (54.5%) and controls 16 (38.1%) than the others (Table 3 and Fig. 1).

The multivariable logistic regression result showed that the odds of sepsis-related mortality among neonates who had history of poor feeding were about 4 times higher than those neonates who did not have history of poor feeding [AOR = 4.15; 95% CI (1.64, 10.49). This study showed that neonates who had respiratory distress had 2.7 times higher odds of sepsis-related mortality compared to those neonates who did not have respiratory distress [AOR =2.72; 95% CI (1.31, 5.61), p-value = 0.007]. The odds of sepsis-related mortality among the neonates with history of convulsion was three times higher than those neonates who did not have history of convulsion [AOR =3.13; 95% CI (1.12, 8.76)]. This study also showed that newborn infants delivered before 37 completed weeks of gestation (preterm babies) were at risk for sepsis-related neonatal mortality. Neonates who delivered before 37 completed weeks of gestation (preterm babies) had 5 times higher odds of mortality compared to neonates who were born after 37 completed weeks of gestation [AOR = 4.64; 95%CI (2.17, 9.91)] (Table 4).