Study area and period
The study was conducted in Addis Ababa city, Capital of Ethiopia. According to data obtained from Addis Ababa city health bureau, there were about 108 governmental health facilities including six regional hospitals, five federal hospitals and 97 health centers in Addis Ababa city and, all of them report to the city health bureau. All of them provide comprehensive maternal and child health care services, including ANC, PMTCT, safe delivery, postnatal care and counseling on infant feeding. There were 1673 HIV positive mothers who gave birth in those health facilities from January 1st, 2014- December 31st, 2015; and 67 of the children born to those mothers were found to have been positive for HIV within 24 months of follow-up period. This study was conducted from April to May, 2017 among 108 government health facilities, reporting to the city health bureau.
Study design and population
Unmatched Case–control study design was conducted among HIV positive mothers who were on PMTCT program in Addis Ababa with their children who had confirmed HIV test results at or before 24 months of age.
A case was defined as HIV positive mother who had been on the PMTCT program with her child tested confirmed HIV positive at or before 24 months. A Control was HIV positive mother who had been on the PMTCT program with her child tested definitive HIV negative at 24 months. The HIV statuses of the mother, the partner and the child were obtained from their records.
Exclusion criteria
Children accompanied by someone else other than their mothers or children whose mother were deceased or transferred-out to facilities outside of Addis Ababa were excluded from the study.
Sample size
Sample size was calculated using Epi Info™ Version 7 StatCalc function of Sample Size Calculation for Unmatched Case-Control Study at 95% confidence interval (CI) and power of 80%, assuming 7 % of HIV negative children were exposed to mixed feeding and odds ratio (OR) of 3.55 based on similar previous study [21], which gave the largest sample size, considering 1: 4 ratio of cases to controls; thus, a sample size of 220 child-mother pairs (44 cases and 176 controls) was estimated.
Sampling technique and procedures
Among all health facilities reporting to Addis Ababa city health bureau those who reported at least one HIV positive child born to mothers on PMCT program of HIV from 01 January, 2014 to 31 December, 2015 were selected purposively from Addis Ababa city health bureau annual reports. Accordingly, 22 health centers and four hospitals were included into the study.
Since, number of HIV positive children born to mothers on PMTCT program in government health facilities in Addis Ababa during the period from 01 January 2014 to 31 December 2015 were only 67 children, all of them who met the inclusion criteria were recruited into the study. Consequently, all the 44 HIV positive children with their mothers were considered as cases and for each case four controls with the nearest date of birth to the cases were recruited from the same health facilities as the cases (Fig. 2).
Data collection techniques and procedures
The questionnaire was developed based on integrated PMTCT registration, exposed infants’ care follow-up records, medical record of the mother and the child. Additional potential variables were also incorporated and translated to local language (Amharic) and retranslated back to English to check for consistency and administered to the respondents with the Amharic version (Additional file 1).
Pretest of the tool was done in March, 2017 at Bishoftu Hospital, outside the study area that have similar setting with the study area, on 5 % of calculated sample size and appropriate modification was made. The data collected during pretest were not included in this report. In order to ensure confidentiality of their information, data were collected by trained and experienced clinical nurses and diploma midwifes who have at least 2 years of experience working at PMTCT clinic of the same facility.
Data related to CD4 count, viral load, WHO clinical stage, syphilis test results during pregnancy, birth weight, gestational age, duration of labor and rupture of membrane, adherence to ART and Cotrimoxazole prophylaxis therapy (CPT), sero status of the mother, the partner and the child were obtained from medical records. Whether pregnancy was planned, questions demonstrating partner involvement level and knowledge on MTCT of HIV were among variables collected by interviewing the mother. Some of these variables were cross checked against recorded values so as to reduce recall bias. Height and weight of the mothers were measured by nurses who were blinded for case or control status of the mothers.
Operational definitions
Definitive HIV test results
HIV test result identified with DNA/PCR prior to 18 months of age, or by rapid antibody test after 18 months of age and 6 weeks of cessation of breast feeding [10].
Mothers on PMTCT
Mothers who have taken ART for prevention of MTCT of HIV either during pregnancy or child birth and delivery or during breastfeeding [10].
Nutritional status of the mother
Measured using Mid Upper Arm Circumference (MUAC): If > 22 cm = Not malnourished, ≤ 22 cm = Malnourished [10].
Knowledge about MTCT of HIV
Knowledge index was built using answers to six questions: three on possible periods of MTCT of HIV and three questions on possible ways of preventing MTCT of HIV; then, those who answered 60% and above of the questions was categorized as having high knowledge and < 60% was taken as having low knowledge [22, 23].
Partner
A person with whom the woman had intimate sexual relationship and became pregnant with the index child.
Partner involvement level
The level of partner involvement in PMTCT program was measured using six questions. A total score of four to six was considered as a ‘high’ partner involvement and less than four as ‘low’ partner involvement [24].
Participated in mother to mother support group
Mother member of formally organized group formed by HIV positive mothers who pass through PMTCT services and participated in at least one regular meeting.
ART adherence
Measured based on number of missed doses within 60 days. Three or less doses, four to eight doses, nine or more doses rated as good, fair and poor respectively [10].
CPT adherence
Measured based on number of missed doses per month. Less than three doses, three to nine doses, more than nine doses rated as good, fair and poor respectively [10].
Data processing and analysis
Data were coded and double entered in to Epidata version 3.1 and exported to SPSS version 22 for cleaning and analysis. Exploration and cleaning of data were made to check for any inconsistencies, errors in coding, missing values, out of range values, unexpected data or outliers and inconsistencies were cross checked with the data in hard copy and necessary correction measures were taken.
The principal outcome variable was HIV status of the child coded as 1 for cases (HIV+) and 0 for controls (HIV-). Bivariate logistic regression analyses were performed to nominate candidate variables for multivariable analysis and those potential variables associated with the outcome at P ≤ 0.10 were included into the initial multiple logistic regression models, using backward fitting. Variables persisted to be associated with the outcome at P ≤ 0.05 were used in the final model. Adjusted Odds ratio (AOR) with its 95% CI was considered to judge for precision and decide whether independent association between outcome and independent variables exist.
Multicollinearity test was done to check whether independent variables were intercorrelated using variance inflation factor (VIF) and value of ten or more was considered for diagnosing multicollinearity and there were no seriously correlated variables. Reliability of the tools was measured using Cronbach’s alpha value and scales with value above 0.74 was accepted as reliable.
The log likelihood ratio test was used to test overall model fitness, measured based on statistical significance of the model chi-square displayed under Omnibus test of model coefficients and probability of model chi-square less than 0.05 supported the model to be good fit. In addition, Hosmer and Lemeshow test goodness of fit statistic value of more than 0.05 were used to characterize a logistic regression model as better fit.
A Wald test was used to test the statistical significance of relationship between MTCT of HIV and individual independent variables in the model. If the probability of the Wald statistic for the variable was less than the level of significance of 0.05, the null hypothesis that β coefficient for that variable equal to zero was rejected and supports the relationship.
Comparability of cases and controls on some of the continuous sociodemographic characteristics was checked using independent sample T-test and significance level greater than 0.05 and confidence interval of mean difference including zero was used to accept the null hypothesis and decide no difference. Equality of variances assumption of the independent sample T-test was checked by Levene’s test and significance level greater than 0.05 was used to decide equality of variance.
Data quality assurance
The questionnaire was designed based on integrated PMTCT register, exposed infant follow-up registration book, medical record and after extensive review of related journal articles. The questionnaire was translated to local language and administered with the Amharic version. Experienced data collectors and supervisors were recruited and training was given, including practices during the pretest. Close supervision was made and filled formats and questionnaires were checked on daily basis for completeness and consistency.
Data documentation sheet or code book was prepared and used to make data entry form in Epidata and additionally the CHK commands were applied in order to restrict out of range and illegal values. A new composite variable derived from four variables (region number, facility type, facility code and patient assigned number) was used as a unique identifier in order to prevent duplicate data entry. Data were also double entered, cross checked, validated and all the discrepancies were resolved.
Relation not assessed by this research; 
Investigated by this research
/
/
/5643/227). Support letters were issued from the IRB to relevant offices and the selected health facilities to obtain their permission for research undertakings. In addition, after elaborating the purpose of the study and importance of their participation and agreeing not to share the data to any one, written informed consent was obtained both for interview and accessing records and those who agreed to participate were included in the study. The participants were informed their full right to withdraw from the study at any time they want and their information would be discarded immediately if they decided to do so. Participants were not paid for their participation nor punished for failing to do so as all received the necessary interventions irrespective of their participation in the study. The study did not use any invasive procedure as data were extracted from records and interview of the mother. Confidentiality of the information was maintained throughout the study process and later as nurses working at PMTCT or/and ART clinics of the same health institutions to whom the mothers already disclosed their HIV status collected the data. Privacy of the participants was protected by interviewing them in a separate room.