A facility-based cross-sectional study was conducted from March to April 2019, at Central zone Tigray regional state of Ethiopia. The central zone is located 1024km north of the Ethiopian capital city of Addis Ababa, 241 Km east of the Tigray capital city Mekelle. The central zone is divided into 12 weredas and the total population was 1,245,223 from those 631,972 were females and 613, 251 were males. The total people living with HIV were 4841, The reproductive-age women who are HIV positive were 2028 and HIV positive mothers who have an infant less than or equal to 24 months were 485. There were 72 health centers (20 of them have ART service), 6 primary hospitals, 3general hospitals, and 1 referral hospital.
All HIV positive mothers having infants less than or equal 24 month years visit in the selected public health facility central zone were taken as the study population. All HIV positive mothers who gave birth within 24 months before the study period are included, while mothers who unable to communicate, who were not mentally competent or who had any psychiatric disorders and critically ill during data collection were excluded.
Sample Size Determination
By taking into consideration 88.1% of knowledge exclusive breastfeeding among HIV positive mothers (12). 95% confidence level, 5% tolerable margin of error, possible non-response rate of 5%, a total sample of 239 was taken. To check for the adequacy of the sample size, Epi info was used by considering factors associated with knowledge exclusive breastfeeding. By comparing sample size based on a single population proportion formula and Epi-info, the sample size determined by the single population proportion was greater than that of the Epi-info. Therefore the largest sample size of 239 was taken.
All public health facilities in the central zone Tigray region was considered then the sample size was allocated to each health facility proportionally based on their expected number of HIV positive mothers having infants less than or equal 24 months visited all public health facilities during the preceding month before data collection. The study participant was selected by a systematic random sampling method every 2nd HIV positive mothers having infants ageless or equal to 24month.
Knowledge Exclusive Brest feeding
Socio-demographic and Reproductive characteristics, Institutional and source of information related factor
Good knowledge – mothers answer (Score>60% ) on the overall knowledge question(18).
Participants' knowledge was assessed by 13 questions in which each question had a group of 1-3 options. Those participants who answered more than8 questions correctly categorized as "Good knowledge" and those who scored less were categorized as “Poor knowledge” (18).
Data collection tools and techniques
Data was collected by using an interviewer-administered and structured questionnaire adapted from different similar research (15,18,19) with modification according to the context of the study area. To establish face validity and translation quality the questionnaire was tested on 12 women in exclusive breastfeeding outside the study area by data collectors and supervisors during training. A few questions, language clarity and information were revised and the questionnaire was finalized for the study. The questionnaire includes socio-demographic, Reproductive history, Institutional, and source of information related factors.
Data quality assurance and control
Five health professional data collectors and two supervisors were recruited from the Health Center and they were given training for one day. The supervisors followed the process of data collection daily, checked the data completeness consistency and communicate with principal investigators daily.
Data Processing and Analysis
Data were coded, cleaned, recorded and entered Epi info 7and finally export to SPSS version 22.00 for analysis. Simple descriptive summary statistics were done. Tables, statements, charts, and graphs were used to present the result of the analyzed data. Associations between independent and dependent variables were analyzed first using bivariate logistic regression analysis. Variables that had p<0.2 on bivariate analysis were entered multivariable logistic regression analysis. After checking using chi-square test variables with small cell size were merged into related categories. Appropriate model diagnostics and goodness of fit tests were done. Multicollinearity was checked to test correlation among predictor variables and Hosmer and Lemeshow test P-value (>0.2) were conducted to see model fitness. The statistical association between the different independent variables about dependent was measured using OR, AOR, 95% CI and P-values <0.05 was considered statistically significant.