Adequate nutrition is essential to ensure optimal health, physical and mental growth of children . Complementary food, a transitional food is intended to meet the particular nutritional or physiologic needs of the young child. As a result, it is universally recommended that mothers should start nutritionally adequate and safe complementary food at the infant’s sixth month [2, 3]. Moreover, World Health Organization (WHO) supports the implementation of Infant and Young Child Feeding (IYCF) strategy for children aged 6–24 months as a critical element of efforts to address child malnutrition and mortality . Hence, promotion of appropriate complementary feeding prevents 6% of child deaths in countries with high child mortality rate , it is of a paramount importance for developing countries, such as Africa and South-Central Asia . A substantial number (2.1 million) of this mortality is contributed by undernutrition .
However, inappropriate complementary feeding practices have been widely documented, and stay as major public health problem in many developing countries . Studies reports from Uganda , Ghana  and South Africa  revealed that more than half of mothers initiated complementary feeding away from sixth month. Various reports identified the determinants of timely initiation of complementary feeding practices, and it mainly related to the maternal health care utilization and socio-economic characteristics. Higher maternal educational status, unemployment , being married , having antenatal care (ANC) and home birthing [11, 12] are significantly associated with initiation of complementary feeding at the right time, sixth month. Furthermore, male sex of the child and smaller family size are related with increased odds of a timely initiation of complementary feeding [12,13,14]. However, lower maternal socio-economic status is inversely associated with timely initiation of complementary feeding .
In Ethiopia child undernutrition continues as a critical public health problem  and it is an underlying cause for 53% of under five child mortality . Complementary feeding practice is also by far suboptimal in the country, according to which 51% of children are given complementary food at 6–9 months of age, and only 4% of young children aged 6–23 months are found with minimum dietary diversity . Conducting a study in an evidence dearth setting is critical to explore information on complementary feeding practices and its determinants especially; no study is conducted in the current study area. Even the previous studies are institution based [11, 13] and done in a smaller scale, districts level [12, 18], which ultimately affects the generalizability of the finding. Therefore to fill the knowledge gap, this study aimed to determine a timely initiation of complementary feeding and associated factors among children aged 6–24 months in HDSS site. The HDSS site is established to represent the rural northwest Ethiopia.
A community-based cross-sectional study was conducted in HDSS site from 01 May to 29 June 2015. The site is located in the Dabat District, northwest Ethiopia, and hosted by the University of Gondar. The district has an estimated population size of 145,458 living in 26 rural and four urban kebeles (the smallest administration unit Ethiopia). Currently, six health centers and 29 health posts are providing health services to the residents. The HDSS covers 13 randomly selected kebeles (three urban and ten rural kebeles) in different ecological zones (high land, middle land, and low land). A total of 67,385 inhabitants are living in these kebeles. The detailed data collection system, data quality control, the database, and the study setting of Dabat HDSS are described elsewhere .
The study is a part of the original survey which aimed to assess the nutritional status and feeding practice of children aged 6–59 months in Dabat HDSS. Of the total kebeles in the HDSS, eight kebeles were selected by using lottery method. All mothers with children aged 6–24 months and lived in the selected kebeles of the HDSS were included in the original survey. For households with multiple children, a child was selected randomly using lottery method. To estimate complementary feeding practice among mothers with children aged 6–24 months, sample size was calculated using Epi-info version 3.7 by considering the following assumptions; the prevalence of a timely initiation of complementary feeding in North Ethiopia was 62.8% , 95% level of confidence, 5% margin of error, 5% non-response rate, and a design effect of 1.5. Thus, a minimum sample size of 566 was obtained. However, 591 children fulfilling the inclusion criteria were found in the original survey, as a result to improve the power of the study, all the eligibles were included in the analysis.
A structured interviewer-administered questionnaire was used to collect data. The questionnaire composed of socio-demographic and economic characteristics, health care, and complementary feeding practice related variables. The questionnaire was first translated from English to Amharic, the native language of the study area, and was retranslated back to English by professional translators. Fourteen data collectors and three field supervisors were recruited for the study. Data collectors and supervisors were trained for 2 days. The tool was piloted out of the study area, thus the acceptability and applicability of the procedures and tools were evaluated. All filled questionnaires were daily checked for completeness by the supervisors and the investigators.
Complementary feeding practices were assessed according to the key indicators recommended by WHO and IYCF strategy of Ethiopia. The outcome variable, timely initiation to complementary feeding was defined as a proportion of children (6–24 months) who initiated complementary food at sixth month of age . The standardized dietary diversity score tool was used to estimate feeding practice of children. To capture the usual dietary habit of children, both a single 24-h and 7 day recall methods were employed. Minimum dietary diversity was defined as proportion of children (6–24 months) who received complementary food made from four or more food groups in the previous 24-h . Moreover, minimum meal frequency was determined as proportion of breastfed and non-breastfed children (6–24 months) who received solid, semi-solid or soft food for the minimum number of times or more in the 24-h preceding the survey .
Furthermore, the 7-day quasi food frequency, modified food group frequency was measured as the number of days the child consumed any of the seven food groups in the last 7 days preceding the date of survey. Since it shows usual dietary pattern of children, the 7-day quasi food frequency was used to overcome the limitation of 24 h recall method. A mother was asked to report for how many days her child ate any of the above seven food groups in the last 1 week. A score ranging from 1 to 7 was given depending on the number of days the child ate the listed food groups. Accordingly, if the child ate for 1–3 days it was coded as ‘1’, and 4 or more days coded as ‘2’, otherwise, the score ‘0’ was given if the child didn’t eat.
A Principal Component Analysis (PCA) was employed to estimate the mothers IYCF knowledge. Nine questions most of which are tailored from the core IYCF indicators were used to determine the mother’s knowledge. In the PCA, the knowledge item questions were summed and ranked into terciles; lowest, medium, and highest. Similarly, household wealth index was computed using a composite indicator for urban and rural residents by considering properties, like selected household assets and size of agricultural land, and PCA was performed to categorize the household wealth status into lowest, middle, and highest.
Data were entered into Epi-info version 3.5.3 and exported to Statistical Package for Social Sciences (SPSS) version 20 for analysis. Descriptive statistics, including frequencies and proportions were used to summarize variables. In order to indentify the factors associated with a timely initiation to complementary feeding, a binary logistic regression model was used. Co-linearity was checked for Household income and other independent variables like IYCF knowledge, antenatal care, and health care access and the result showed that the independent variables have no significant co-linearity. Variables with a p-values of < 0.2 in the bivariable analysis were subjected to multivariable analysis. We followed the method of Hailu et al. .