The study adopted a mixed-method design. The quantitative method used the Infant and Young Child Feeding Index survey to evaluate feeding practices. The qualitative method used semistructured interviews with parents to explore their feeding experiences and analyze barriers to appropriate feeding practices. The study is reported using the reporting guideline of ‘the quality of mixed methods studies in health services research’  obtained from the EQUATOR-network website.
Parents of children with chronic cholestatic liver diseases were consecutively recruited from January 2016 until March 2017 at the Children’s Hospital of Fudan University in Shanghai, China. Inclusion criteria were parents who had 6–24-month-old children with chronic cholestatic liver diseases, defined as a diagnosis established by gene sequencing or presence of at least 6 months of typical clinical and biochemical derangements . Parents were excluded if the child was a premature infant, had other serious diseases seriously influencing feeding, or could not receive normal feeding due to progressive deterioration of conditions, such as acute liver failure.
A single population proportion formula was used to determine the sample size based on the following assumptions. The proportion of appropriate feeding practices was 80% , with a 95% CI, 10% margin of error (d), and 10% nonresponse rate, resulting in a sample size of 74.
Children with chronic cholestatic liver diseases routinely visited the outpatient clinic for regular follow-up and adjustment of medical treatment. After visiting the outpatient service, parents of eligible children were invited to participate in the study.
Demographic characteristics, including sex, date of birth, gestational age, birth weight, literacy of parents, and monthly household income, were collected by asking the parents. Laboratory data were obtained from the medical records, including alanine aminotransferase, aspartate aminotransferase, total bilirubin, direct bilirubin, and gamma-glutamyl transpeptidase level.
Feeding practices were measured using the Infant and Young Child Feeding Index (ICFI), which was designed by the ‘Chinese Center for Disease Control and Prevention’ according to the WHO recommendations of child feeding practices . The ICFI has been widely used to assess the dietary intake of children aged 6–24 months and has been suggested to be efficient in assessing parents’ feeding practice in China. Numerous studies have shown that the ICFI is strongly associated with the status of children’s growth and development [18, 19]. The ICFI is a comprehensive index and includes seven variables and thirteen items: breastfeeding, bottle feeding, dietary diversity for the past 24 h, frequency of food (grains, vegetables, fruits, eggs, flesh foods, bean products, and dairy products) for the past 7 days, first formula milk feeding age, first complementary feeding age, and frequency of complementary feeding for the past 24 h. Breastfeeding and bottle feeding were assessed by a 2-point scale (yes/no), while the other items were assessed on either a 2-point or 3-point Likert scale (Electronic Supplemental Material 1). The range of the total score was 0–23, and higher scores indicated better feeding practices. Appropriate feeding practice was determined when the score was greater than 60% of the possible score, and inappropriate feeding practice was defined as ICFI < 14. In addition, given the unique nutrient requirements of children with chronic cholestatic liver diseases [8, 20], two ad-hoc questions were added to the survey to identify the intake of dairy products and medium chain triglycerides.
Descriptive data were expressed as either means and SDs, medians and interquartile ranges for continuous variables or frequencies for categorical variables. The Shapiro-Wilk test showed that the levels of total bilirubin, direct bilirubin, glutamyl transpeptidase, alanine aminotransferase, and aspartate aminotransferase had nonnormal distribution. These data were therefore expressed as the median and interquartile range (IQR).
When we analyzed the determinants of feeding practice, bivariate analysis was first conducted between the feeding index and each variable, including sex of child, age of child, birth order of child, family monthly income per capita, literacy of primary care provider, presence of cholestasis (direct bilirubin), and area of residence. Primary care providers were defined as those who took the main responsibility for disease management, feeding and daily activities. The variables marginally associated (p < 0.10) with the feeding index in the bivariate analysis were selected for multivariate analyses. Then, multivariate analysis was conducted for the feeding index with potential risk factors. In the final model, only variables significantly associated (p < 0.05) with outcome variables were retained, and the adjusted odds ratio (AOR) and 95% confidence interval (CI) were reported. Children with missing data were excluded from in the model.
All statistical analyses were performed using IBM-SPSS, version 22 (IBM Corp, Armonk, NY, USA). The significance level was set at p < 0.05.
After the survey of feeding practices, to determine the motives behind inappropriate feeding practices and promote behavioral changes, we purposefully approached participants with an ICFI score < 14. We continued data collection until thematic saturation was observed, which meant that there were no new codes or themes occurring during the ongoing analysis . In total, ten mothers and one father participated in the interviews.
Interviews with parents were conducted in a quiet meeting room in the hospital. A semistructured interview guide was developed. Example questions were as follows: ‘How did you feed your child?’, ‘What were the sources of your feeding knowledge?’, and ‘Can you explain any confusion or uncertainty about how to feed your child?’. The duration of the face-to-face interviews was 30–60 min. All interviews were audio-recorded and transcribed verbatim within 24 h after the interview.
Thematic analysis was carried out by structuring data, initial coding, categorizing experiences and perceptions, and identifying themes .
The first author (XC) read all transcripts, identified and sorted meaningful units that elucidated the study question and categorized the units into preliminary topics. Then, the interview transcripts were individually coded by two researchers (XC and YZ) and checked for similarities. Discrepancies were solved by discussion between the two researchers. Subsequently, we organized and categorized the codes into subthemes/themes, which were discussed with two other researchers (YG and JML).
The study was approved by the Research Ethics Committee of the Children’s Hospital of Fudan University (approval number 2014–143) and was performed in accordance with the ethical standards of the 1964 Declaration of Helsinki and its later amendments. Informed consent was obtained after providing information sheet and verbally explaining the study aims and procedures. Participation was voluntary and each parent signed a written consent form.