Predictors of neonatal mortality in Assosa zone, Western Ethiopia: A matched case control study

Background: Benshangul Gumuze region is one of the regional states in Ethiopia, with highest rate of neonatal mortality rate. The trend increased at alarming rate from 42/1000 live birth in 2005 to 62/ 1000 live birth in 2011. Hence, identifying predictors of neonatal death and implement evidence based interventions at community level is crucial to reduce the mortality. Therefore, the purpose of this study was to identify predictors of neonatal mortality in Assosa zone, Western Ethiopia. Methods: A community based matched case control study was conducted from February 1, until December 30, 2013. The study included 114 cases who died during the first 28 completed days after birth from September 1, 2010 till September 1, 2013. For each case, one alive control matched approximately by the same date of birth (-/+ 2 days) was identified from the preliminary data collected. Finally, multivariate conditional logistic regression analysis was performed; and goodness of fit of the final model was tested using likely hood ratio test. All analysis was done using EPI Info version 7 and SPSS version 16 statistical softwares. Results: Model households in health extension packages [AmOR = 0.32; 95%CI:0.12-0.86], age at first pregnancy < 20 years old [AmOR = 4.3;95%CI: 1.13-16.27],pregnancy complication [AmOR = 4.59; 95%CI: 1.53-13.78], delivery complication [AmOR = 2.80; 95%CI: 1.06-7.39], antenatal care visit [AmOR = 0.34;95%CI: 0.12-0.94], primipara mothers [AmOR = 3.37; 95%CI:1.05-10.78], small size neonate at birth [AmOR = 3.40: 95%CI: 1.05-11.55], gestational age < 37 weeks [AmOR = 4.35;95%CI:1.16-16.28], and home delivery [AmOR = 2.84; 95%CI:1.07-7.55] were found statistically significantly associated with neonatal mortality. Conclusions: Model households in health extension package and antenatal care visit were associated with reducing risk of neonatal mortality. However, age at first pregnancy < 20 years old, primipara mothers, pregnancy complication, delivery complication, small size neonates, gestational age < 37 weeks, and home delivery were associated with increasing risk of neonatal death. Therefore, promotion of model household in health extension package, anti natal care visit, institutional delivery, family planning to prevent early age pregnancy; and improve access to basic emergency obstetric care and intensive newborn care centers are effective interventions to reduce risk of neonatal mortality at community level. A matched case control study was conducted in Assosa zone, Western Ethiopia from February 1 until December 30, 2013. Assosa zone is located 667 km to West of Addis Ababa. It has a population of 342,287; male 188,258 and female 154,029 (2007 census projected). The zone has 7 woredas (districts) and 72 kebeles (villages). The study population was sample of neonates who died during the first 28 completed days after birth and sample of neonates who survived the first 28 completed days after birth, from September 1, 2010 until September 1, 2013. Cases were neonates (index birth) who died during the first 28 completed days after birth and controls were neonates (index birth) who survived the first 28 completed days after birth and alive during data collection. Neonates, match the definition of case and control were eligible for the study. However, neonates born outside Assosa zone and neonates mothers who were sick or unable to communicate were excluded from the study. Sample size of the study was determined by PS software version 3.0.43 power and sample size calculation for matched case control study. Different predictors were considered to determine the sample size. Thus, preterm birth was chosen, as it gave large sample size using parameters of 95% confidence interval (CI), 80% power, proportion of preterm birth among cases (P1) was 48.5%, and proportion of preterm birth among controls (P0) was 8.2% [39]. The minimum detectable odds ratio was 3.5 and ratio of case to control (m) 1:1. Correlation coefficient (r) for exposure between matched case and control was unknown; hence, 0.2-phi coefficient was taken on the assumption of dependency [40]. Adding 5% contingency for non-response, the total sample size required for the study was 238(119 cases and 119 controls). With regard to sampling techniques, from the total seven woredas (districts) of the zone, four districts were selected first, then from each four districts, four kebeles (villages) were chosen. Simple random sampling method was used to choose both the districts and villages. As result, totally 16 villages were included in the study to get sufficient sample of cases (Fig. 2). Schematic presentation of sampling technique to study predicators of neonatal mortality in Assosa zone, Western Ethiopia, 2013 Preliminary data was collected from the 16 villages’ health posts child health registration book, ahead of the actual study. Information was collected on the following variables: date of birth, date of death, household identification, place of birth and type of birth (singleton or multiple). Then, all live birth neonates who survived the first 28 completed days after birth and all live birth neonates who died during the first 28 completed days after birth from September 1, 2010 till September 1, 2013 were indentified. From the preliminary data result, 139 singleton cases were identified and eight cases were excluded based on the exclusion criteria listed (six cases were born outside Assosa zone and two mothers of cases were unable to communicate). Afterward, 131 eligible cases were identified and sampling frame was prepared. However, due to constraint of resource, only the required samples of 119 cases were selected using simple random sampling method. Finally, for each identified cases, one alive control was matched approximately by the same date of birth (−/+ 2 days) from the preliminary data collected. Structured questionnaire was prepared in English by reviewing different literatures and adapting World Health Organization verbal autopsy to local context. Then, it was translated to Amharic and checked for consistency by back translation to English by different individuals. The questionnaire had five parts: socio demographic and economic characteristic, maternal biological and obstetric factors, neonatal factors, behavioral and psychosocial characteristics, and health service and delivery related factors. Model households in health extension package was defined when the households were attended training on 16 health packages (personal hygiene, control of insect and rodents, healthy home environment, food hygiene and safety, water supply and safety, solid and liquid waste, safe excreta disposal, TB and HIV control, malaria prevention and control, first aid and emergency measure, maternal and child health, family planning, immunization, adolescent reproductive health, nutrition and treatment of common childhood diseases like diarrhea, pneumonia, malaria and severe malnutrition) given by health extension workers. In line with this, the households must also implement at least 75% and above of the packages listed above; and certified by districts health office for their achievement. Then, when respondents comply with the above conditions, it was scored in to yes otherwise into no. Pregnancy complication was measured based the following medical conditions during pregnancy period: vaginal bleeding, abdominal pain, persistence of back pain, blurry vision, no fetal movement and swelling of hands or face. Delivery complication was measured based on the following medical conditions during delivery time: mal presentation, obstructed labor, meconium stained amniotic fluid, premature rapture of membrane (< 1 days), umbilical cord prolapsed, vaginal bleeding and retained placenta (> 30 min). For both delivery and pregnancy complication, the above medical conditions were predefined in the questionnaires. So mothers were responded from the predefined medical conditions; if they answered presence of at least one or more signs from the list, it was scored yes unless scored no. Neonate size at birth was proxy indicator of birth weight at birth and measured by perception of the mother. It was labeled in to small and average size neonate. Small size neonates was proxy indicator of neonates with low birth weight (< 2500 g) and average size neonates was proxy indicator of neonates with normal birth weight (2500–4200 g). Early initiation of breast-feeding was measured when the neonates start breast-feeding with in 1 h (= < 1 h) after birth and late initiation of breast-feeding when the neonates start breast-feeding > = 2 h after birth. Birth attendant was defined when mothers were assisted by health professionals (physicians, nurses, health officers, and health extensions trained in clean delivery) during delivery time, then it was labeled into skilled birth attendant otherwise labeled in to unskilled birth attendant. Structured interviewer administered questionnaire was used to collect information from each participants. Interview for the mothers were conducted face to face. Twelve diploma nurses, currently working on maternal and child health in health center and four bachelor sciences nurses currently working as health center –health post linkage focal person in health center; who speak and understand local language were recruited and trained as data collectors and supervisors respectively. Before data collection, the instrument was pretested in 5% of the total sample size. During data collection, the administered questionnaires were checked for completeness and consistency on daily basis by supervisors. After data collection completed, the principal investigator checked the data during data entry, cleaning, and analysis. Data was entered, processed, and analyzed using EPI Info version 7 and SPSS version 16 softwares (Additional file 1). Descriptive analysis was done, to check for outliers and inconsistencies. Then, univariate conditional logistic regression analysis was performed to identify candidates for multivariate conditional logistic regression. Strength of the association was measured using parameters of crude matched odd ratio (CmOR) with 95% confidence interval (CI) and significance test at P value < 0.05. An inclusion criterion for multivariate conditional logistic regression was P -value < 0.2. Backward elimination strategy was used to build the final model and measure of association of each predictors were determined using parameter of adjusted matched odds ratio (AmOR) with 95%CI. Statistical significance was tested using Wald statistical test at P value < 0.05. In line with this, biologically meaningful interactions were assessed for inclusion in the final model. Furthermore, multicollinearity was checked using variance inflation factor (VIF) parameter, with acceptable range of 1–10 coefficients. Accordingly, collinearity was found between parity and gravidity; and diagnostic was made by removing gravidity from the model. Finally, goodness of fit of the final model was tested using likelihood ratio statistics and it was found fit.