Neonatal care practice and factors affecting in Southwest Ethiopia: A mixed methods study

Background: A significant proportion of neonatal mortality can be prevented by the provision of the minimum neonatal care package. However, about 3 million neonates die each year globally because of lack of appropriate care. This situation is the worst in Ethiopia. Thus, the objective of this study was to determine the status of neonatal care and identify factors affecting. Methods: A mixed methods study involving both quantitative and qualitative methods was conducted from September 2012-December 2013 in Southwest Ethiopia. Randomly selected sample of 3463 mothers were interviewed to collect the quantitative data. Twelve in-depth interviews with purposively selected key informants and six focus-group discussions with purposively selected mothers were conducted for the qualitative data. Mixed-effects multilevel linear regression model was used to identify predictors of neonatal care practice by using STATA 13. Audio recording, transcription and thematic content analysis was done for the qualitative data. Results: The overall status of neonatal care practice was 59.5 % (95 % CI: 57.6 %, 61.3 %). Of the respondents, 53.8 % received tetanus toxoid, 23.8 % planed for birth, 41.9 % received at least one antenatal care and 43.0 % received adequate information during pregnancy. Only, 17.5 % received skilled care at birth and 95.0 % received social support. Of the neonates, 96.5 % received appropriate thermal care, 86.5 % received clean cord care, 64.1 % initiated breast-feeding within one hour, 91.5 % were on exclusive breast-feeding, 56.5 % received appropriate bathing and 8.1 % received vaccination on date of birth. Place of residence, maternal education, husband’s occupation, wealth quintiles, birth order and inter-birth interval were identified as predictors of neonatal care practice. Conclusions: The status of neonatal care practice was low in the study area. Skilled care at birth and receiving vaccination on date of birth were the worst practices. Factors affecting neonatal care existed both at cluster level and at the individual level and included socio demographic, economic and obstetric factors. Appropriate birth spacing, birth limiting and behaviour change communications on the importance of neonatal care are recommended. This community-based prospective follow up study, employing both quantitative and qualitative data collection methods, was conducted in Jimma Zone, Southwest Ethiopia from September 2012 to December 2013. Jimma Zone is one of the 17 Zones of the Oromia Regional State of Ethiopia. Administratively, the Zone is sub-divided in to 17 rural districts called ‘Woredas’ and two town administrations. According to the 2007 national population and housing census, the Zone has a total population of about 2.6 million, of whom 88.7 % are rural residents [19, 20]. Mothers who had given birth 28 days before the survey were the study populations for the quantitative method. The minimum required sample size for this study was determined by using Epi-Info V.3.5.1 based on the following assumptions. The status of neonatal care practice as determined by the mean score of composite variable (indices) was assumed to be 29 % (p = 0.29) (Gashaw A. Assessment of New Born Care Practices During the first week of life Among mothers in Addis Ababa, Ethiopia (Unpublished)). The allowed margin of error to be 3 % (d = 0.03) with 95 % level of confidence. In addition, as multistage-clustered sampling method was used, a design effect of 2 was considered. Finally, 10 % was added for non-responses and missed-to-follow up and the final sample size became 1934 mothers. This study was part of a bigger longitudinal study in which 3463 mothers were followed up. Therefore, to increase the precision of the estimates and power of the study, we included all the 3463 mothers in this study. A multistage-clustered sampling technique was used to identify the study participants. Initially, the Zone was stratified as town administration and rural districts called ‘Woredas’. Then, 5 districts were selected by simple random sampling from the 17 districts. At the second stage, 9 rural ‘Kebeles’ and 2 urban ‘Kebeles’ were selected from each selected district randomly. Jimma town administration and Agaro town administration have 13 and 5 ‘kebeles’, respectively and all were included purposefully. With this, in total, 73 clusters (‘Kebeles’) were included in the study from which 3682 pregnant women were enumerated and enrolled to the study at the baseline. All the enrolled pregnant women were followed till 28 days postpartum period and neonatal care practice was assessed at the end of neonatal period. To have in-depth understanding of neonatal care practices and contributing factors, 12 in-depth interviews (IDIs) and 6 Focus Group Discussions (FGDs) were conducted. The IDIs involved 4 service providers, 4 traditional birth attendants (TBAs) and 4 Health Extension Workers (HEWs) all of whom were selected purposively based on their close relation with mothers and neonates and assumed to be rich sources of information on the topic of the study. The FGDs involved purposively selected 8-10 mothers having post neonatal infants (1-6 months) each. The number of IDIs and FGDs were determined based on level of saturation of the required information. The data were collected by using pre-tested interviewer-administered structured questionnaires, which were adapted from related literatures. The indicators for the wealth index were adapted from Ethiopian Demographic and Health Survey (EDHS) [8]. Indicators for neonatal care practice were adapted from the World Health Organization (WHO) minimum neonatal care package [14]. The questionnaire was prepared in English, then translated to local languages ‘Afan Oromoo’ and Amharic and used to collect the data. The dependent variable for this study was neonatal care practice, which was a composite score (index) created from 12 items and treated as continuous variable. By taking ‘Kebeles’ as clusters, the independent variables were divided into two levels. Level-2 (higher level variables) included community or cluster-level variables such as place of residence, access to health centres and access to hospitals. Level-1 (lower- level variables) included individual and household characteristics such as: socio-demography, wealth quintiles and maternal obstetric factors. The detail description of each variable is given below (Tables 1 and ​and22). Description of variables and measurement for the study, Jimma Zone, Southwest Ethiopia, September 2012-December 2013 Description of variables and measurement for the study, Jimma Zone, Southwest Ethiopia, September 2012-April 2013 The collected data were coded and entered into Epidata V.3.1 to minimize logical errors and design skipping patterns. Then, the data were exported to SPSS for windows version 20.0 for cleaning, editing and analysis. Descriptive analysis was done by computing proportions and summary statistics. Wealth quintiles were determined by using Principal Component Analysis (PCA). Similarly, neonatal care practice, a continuous dependent variable, was created as a composite index (score) by using PCA. The index was created by including the 12 elements of the minimum neonatal care package described in Tables 1 and ​and22 above. Each variable were measured in terms of “Yes” or “No” response categories and later changed to dummy variables by assigning “1” for “Yes” responses and “0” for “No” responses for the PCA. While doing the PCA, colinearities between the independent variables were checked by producing correlation matrix. However, no correlation coefficient was 0.9 or above for a variable to be excluded. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was 0.74 (>0.50 is acceptable) and Bartlett’s Test of Sphericity was significant (p < 0.001). The importance of each variable for the model was checked by looking at the communalities and those variables having communalities < 0.5 were removed one by one from the model until all have ≥ 0.5. The eigenvalues was set to include values over 1. Based on this, receiving social support during labor, receiving vaccination on date of birth and BPCR were excluded. As a result, 9 variables remained in the final model creating 4 principal components with eigenvalues > 1.0 explaining 65.04 % of the total variance (>60 % is acceptable to use PCA). No variable was found to have complex structure or high loadings (> 0.4 in more than one component in the rotated component matrix). Inter-item consistencies for the variables making each component were checked by Cronbach’s alpha and all were > 0.7. The existence of outliers was checked by sorting each principal component by ascending order and all cases were within the range of ±3 factor scores. Finally, all the 4 components were added and an index (score), the continuous dependent variable, was created. The status of neonatal care practice was determined by dichotomizing the score based on the mean value of the score. As Jimma town administration and Agaro town administration were included purposefully, weighted analysis was done to avoid urban over representation and over estimation of the status of neonatal care practices. The weighted analysis was done based on the complex-sample survey procedure by considering the probability of exclusion at different stages and the non-responses. To identify factors affecting neonatal care practice, first, bivariate analysis was done to see associations between each independent variable and neonatal care practice. Then, all variables having p < 0.25 were considered as candidates for the multivariable model. Because of the different levels of factors, a mixed-effects multilevel linear regression model was used by using STATA 13 to identify factors having significant associations with neonatal care practice. This model was preferred in order to avoid the clustering effects that violets the assumption of independence among the study subjects. To evaluate the existence of sufficient variance at the cluster level in influencing neonatal care practice, intercept-only model was fitted and Interclass Correlation Coefficient (ICC) was determined. Besides, goodness-of-fit of the multilevel model was tested by the log likelihood ratio (LR) test. Multicollinearity between the independent variables was assessed by using variance inflation factors (VIF >10 were considered as suggestive of existence of multicollinearity). In addition, cross-level two-way interactions were checked. Beta (β) coefficients along with 95 % CI were used to show the strength of the associations and level of significance. The audio taped qualitative data were transcribed in to English language. Then, codes or terms were identified and tallied to come up with some categories, which later used to establish themes based on the objective of the study. Finally, thematic analysis was done and the findings were triangulated with the quantitative one. Ethical approval was obtained from the Institutional Review Board (IRB) of the College of Health Sciences of Addis Ababa University. In addition, written informed consent was obtained from each respondent before actual data collection.