Study settings, period and design

The study was carried out in multicenter health facilities in Addis Ababa, the capital of Ethiopia, and the diplomatic center of Africa. So, it hosts a number of international organizations, such as the headquarters of African Union (AU) and the United Nations Economic Commission for Africa (UNECA). Due to its location and status, several people come to the city in search of employment opportunities and services. The city has three layers of administration, including city administration, eleven subcities called kifle-ketema and 116 woredas at the lowest administrative units (40). In the city, there are 12 Public heath hospitals (six federal hospitals and six regional hospitals) and 100 public health centers. Among these, 30% of these health facilities (n = 30) and 30% of public health hospitals (n = 4), namely Zewditu, Ghandi, St. Peter and ALERT Hospitals, were selected randomly.

A cross-sectional study design was employed among HIV-positive mothers who had less than 6 months of aged infants and attending PMTCT and/or on ART services at randomly selected 34 public health facilities in Addis Ababa city from August 1, 2022, to August 30, 2022.

Population

The source population was all HIV-positive mothers who had less than 6 months of aged infants and attending PMTCT and/or on ART services at all health facilities in Addis Ababa, whereas randomly selected HIV-positive mothers who had less than 6 months of aged infants and attended PMTCT and/or on ART services during the study period were study population.

Sample size determination and sampling procedures

The sample size was calculated using single population proportion formula by considering the assumptions of confidence interval Z score standard value with a confidence level at 95%, proportion (P) of recommended infant feeding practice in Gulele subcity (37.4%) and d2 (marginal of error) 5% [22]. With this assumption, the calculated sample was 359, and by considering 10% non-response rate, the final sample size become 413. The study subjects were selected from each health facility based on simple random sampling techniques following probability proportional to size of study subjects served. Accordingly, a total of 358 and 55 study participants were selected from 30 health centers and four public health hospitals [Zewditu (n = 18), Ghandi (n = 14), St. Peter (n = 10) and ALERT (n = 13)]. The study subjects were all HIV-positive mothers who had infants aged between 0 and 6 months, and fulfill the eligibility criteria were randomly selected during the study period.

Eligibility criteria

The study included HIV-positive mothers who had infants aged between 0 and 6 months lived in Addis Ababa for more than 6 months prior to the study period, volunteer to participant in the study were included, where as those HIV-positive mothers with mental health problem and severe ills and unable to respond or communicate were excluded from the study.

Study variables and definitions

Infant feeding practice was the dependent variable, and it was measured dichotomously as appropriate and non-appropriate infant feeding practices. The variable was measured as appropriate infant feeding practice if HIV-positive mothers and whose infants aged between 0 and 6 months feeding their infant either exclusive breast-feeding or exclusive replacement (formula) feeding, whereas it is inappropriate if the feed their infants in a mixed way before 6 months of infants age. Independent variables that contained socio-demographic and economic characteristics included the age, marital status, educational status of the mother, occupational status of both parents and their income; maternal health- and obstetrics-related factors such as antenatal care, number of ANC visits, mode of delivery, place of delivery, disclosure of HIV status to partner, postnatal care, maternal knowledge and maternal attitude toward infant feeding practices were measured; and infant-related factors such as timely initiation of infant breast-feeding and pre-lacteal feeding status were measured as independent factors. Most independent factors were measured as a dichotomous: presence (yes) or absence (no) of responses, whereas knowledge of mothers was measured as “good knowledge” when mothers who scored above the median value for knowledge questions, whereas mothers who scored below the median were considered to have “poor knowledge” [11]. Mothers’ attitude toward infant breast-feeding was measured as “appropriate attitude” and “inappropriate attitude” toward infant feeding practice. Accordingly, mothers who scored above the median for attitude questions were considered to have “favorable attitude,” whereas mothers who scored below the median were considered to have “unfavorable attitude” [6]. The data collection tool was organized after a rigorous review of related studies.

Data quality management and statistical analysis

The quality of the data was assured throughout the research phases from inception of the research tool development to the report phase. Prior to data collection, the data collection tool was validated by inculcating experts’ opinion, and then, pretest was done in 5% of the total sample size in health facilities different from the current study area. In addition, 3-day training was given for data collectors and supervisors on the research objective, methods, data collection instrument, data collection technique, data collection procedure and the relevant ethical issues. During data collection, completeness of the interviewer administered structured questionnaires was checked by the supervisors and the principal investigators. After data collection, data were coded, entered using Epi Info 7 and then transferred into SPSS version 24 statistical packages for analyses. A binary logistic regression was done to determine the strength of association between the independent and independent variables. So, variables that had a p value less than 0.2 in the bi-variable logistic regression analysis were entered into a multivariable logistic regression to adjust the effect of confounders on the outcome variables. Multivariable logistic regression models were fitted to determine the presence of an association between the dependent and independent variables at a p value of 0.05 and an AOR with a 95% confidence interval.

Ethical considerations

The researchers secured ethical approval from Menelik II Medical and Health Science College and Addis Ababa Health Bureau, Addis Ababa public health research and emergency management directorate. Permission letter was secured from the randomly selected public health facilities administrators. Informed written consent was obtained from each study participant before data collection.