Aims
The aim of this study was to determine the factor structure of the NeoEAT - Mixed Feeding and to assess its psychometric properties, including internal consistency reliability, test-retest reliability, construct validity and known-groups validity.
Design
This was a descriptive, cross-sectional, instrument development study.
Setting
This study was conducted using online surveys through the Qualtrics survey platform. Parents were recruited for participation in this study through a variety of methods, including recruitment through Qualtrics respondent panels; a pediatric primary care clinic, pediatric feeding and swallowing clinic, and infants who had been discharged from the Neonatal Intensive Care Unit at North Carolina Children’s Hospital; ResearchMatch.com, a national health volunteer registry supported by the National Institutes of Health and the Clinical Translational Science Award (CTSA) program; Join the Conquest, a health volunteer registry through the CTSA at the University of North Carolina at Chapel Hill; a registry of parents of children with problematic feeding maintained by the investigative team; online parent support groups; and an email sent to faculty, staff, and students at the University of North Carolina at Chapel Hill.
Sample
To be eligible to participate in the study, parents had to be at least 18 years old and have an infant less than 7 months old who had been fed by both breast and bottle in the previous 7 days. Parents, for the purposes of this study, were defined as primary caregivers who were familiar with the child’s feeding and are referred to as parents throughout this manuscript. Participants had to have access to the internet in order to complete the survey and had to self-report as being able to read English. Only one parent was allowed to participate per family. If a parent had more than one infant less than 7 months old, they were asked to report on a single infant. The goal was to have parents report on a heterogeneous sample of infants, so infants were not excluded for any medical reasons, but the infant did have to be fed by mouth in the past 7 days, so exclusively tube-fed infants were excluded. The target sample size for the factor analysis was 5–10 participants per item [41]. With 89 items on the original NeoEAT – Mixed Feeding, the target sample for the factor analysis was a minimum of 445 participants.
Measures
NeoEAT - Mixed Feeding
The NeoEAT – Mixed Feeding is an 89-item parent-report measure of symptoms of problematic feeding with items that are relevant to breastfeeding, bottle-feeding, and the infant’s ability or willingness to manage changes between breast- and bottle-feeding. Items on the NeoEAT – Mixed Feeding were developed and content validated with both parents (N = 16) and clinicians (N = 9) [39]. Items on the NeoEAT – Mixed Feeding are prefaced with the phrase “My infant …” followed by a short phrase stating a behavior or symptom that would be observable by a parent with little feeding experience. Items are written at a less than 6th grade reading level [39], which is consistent with recommendations for health-related materials [42]. Response options on the NeoEAT – Mixed are on a 6-point Likert scale from Never to Always. Scores are assigned such that higher scores indicate more symptoms of problematic feeding. Positively worded items are reverse-scored to maintain consistency across the items, with higher scores indicating more problematic symptoms. The possible range of scores for the 89-item NeoEAT – Mixed Feeding was 0 to 445. The NeoEAT – Mixed Feeding is intended for infants less than 7 months who are obtaining the majority of their nutrition from liquid-based feeding (i.e., human milk and/or infant formula). The questionnaire takes approximately 5 to 10 min to complete.
Infant gastroesophageal reflux questionnaire – revised (IGERQ-R)
The IGERQ-R is a 12-item caregiver-report measure of gastroesophageal reflux-related symptoms in infants over the previous 7 days [43,44,45]. The IGERQ-R was chosen as a measure to test convergent validity because it is a parent-report assessment of a construct measured by the NeoEAT – Mixed Feeding, specifically symptoms related to gastroesophageal reflux. The tool is scored on a scale from 0 to 42, with a higher score indicating more symptoms of gastroesophageal reflux. The tool has been validated for use as both an evaluative and diagnostic instrument. Psychometric testing supports the diagnostic capability of the tool, demonstrating its ability to discriminate infants meeting the criteria for GERD diagnosis from those who do not, as well as between infants with mild, moderate and severe disease [43]. Psychometric properties also support its responsiveness to change in GERD symptoms over time, making it a valuable tool for monitoring treatment in clinical practice and evaluating outcomes in clinical trials [43]. The tool was content validated with both caregivers and physicians. Internal consistency reliability (Cronbach α = 0.86–0.87), test-retest reliability (intraclass correlation coefficient = 0.85), and construct validity were all found to be acceptable [43].
Infant gastrointestinal symptoms questionnaire (IGSQ)
The IGSQ is a 13-item parent-report questionnaire about the frequency and severity of gastrointestinal symptoms in infants in the previous 7 days [46]. The IGSQ was chosen as a parent-report measure to test convergent validity between the IGSQ and symptoms of gastrointestinal distress as measured by the NeoEAT – Mixed Feeding. The tool is scored on a scale from 13 to 65, with a higher score indicating more symptoms of gastrointestinal distress [46]. The tool is useful for clinical research on feeding tolerance and identification of infants with gastrointestinal distress. The tool has evidence of acceptable internal consistency reliability (Cronbach’s α = 0.72) and test-retest reliability (r = 0.69) [46]. Known-groups validity was supported with significant differences in scores between infants with and without parent-reported feeding problems. The tool has also shown to be sensitive to differences between human milk–fed and formula-fed infants [46].
Procedures
Parents who agreed to participate in the research study were asked to complete a survey that included the NeoEAT – Mixed Feeding, IGERQ-R, IGSQ, a series of questions about their child’s health and feeding, and questions to describe the respondent and their family. Parents were given 2 weeks to complete the survey. During this time, two reminder emails were sent to those who had not yet finished. Parents who completed the initial survey were offered a $10 gift card. The first 20% of the sample were asked whether they would be interested in completing a second survey 2 weeks later for the purpose of evaluating test-retest reliability. The second survey only included the NeoEAT – Mixed Feeding and therefore was much shorter than the initial survey. Parents were only given 1 week to complete the second survey so that the test-retest surveys were 2–3 weeks apart. Parents who completed the retest survey were offered an additional $5 gift card.
Given the potential threats to validity with online survey research, multiple strategies were employed to ensure the validity of the data used for analysis. Participants recruited through North Carolina Children’s Hospital were identified as eligible through medical record review. All other participants entered the survey through a two-step entry process, allowing for only a single response per individual. Response times to the survey were monitored and respondents who completed the survey in less than one third of the median completion time (defined by the first 10% of the sample), were removed from the survey as their responses were deemed to be too fast to reflect thoughtful and accurate data. Attention-check and verification questions were placed throughout the survey to identify careless or fraudulent respondents. Data were monitored closely and cleaned thoroughly prior to analysis.
Data analysis
Data analyses were conducted using IBM SPSS Statistics 24. Cases with > 10% missing data for the NeoEAT – Mixed Feeding were excluded from the overall analysis. Cases with > 10% data on the IGSQ, IGERQ-R or retest survey were excluded for each of those analyses separately. A missing data analysis was conducted prior to other statistical analyses. For all statistical tests, a p-value of .05 was defined as statistically significant.
Item analysis
First, inter-item correlations were calculated using Pearson’s product-moment correlation. The correlation matrix was evaluated for item-item correlations > .8 (indicating the items were measuring the same construct) and items that failed to correlate with any other item at > .3 (indicating the item may be measuring an unrelated construct) [47]. When two items were correlated at > .8, one of the items was chosen for removal. When an item failed to correlate with any other item at > .3, it was removed.
Factor analysis
Exploratory factor analysis was conducted using principal components analysis with varimax rotation. Procedures for factor analysis followed accepted guidelines for for health-related instrument development [47]. The Kaiser-Meyer-Olkin (KMO) statistic and Bartlett’s test of sphericity were evaluated as a measure of sample adequacy for factoring. Prior to further exploration of the factor analysis results, communalities of the items were reviewed and items with communalities < .5 were removed. The factor analysis was then repeated. Initially, factor extraction was based on an eigenvalue of greater than one, a method that ensures that each factor accounts for a considerable share of the total variance of the items; this method, however, can over- or under-estimate the correct number of factors [47]. The scree plot, which plots the factors against their eigenvalues in decreasing order, was then examined to determine whether a more parsimonious factor solution could be supported [47].
Using the scree plot and the number of factors represented around the bend in the curve, exploratory factor analysis was conducted forcing different factor solutions. Multiple factor solution options were explored, taking into account total variance explained, number of cross-loading items, and conceptual clarity of the factors [47]. Items that cross-loaded at > .3 on two factors were considered for movement to another factor based on the conceptual fit. Items that failed to load on any factor at > .3 were identified for removal. After final placement of items within the factors, factor names were assigned based on the concepts measured by the items within the factor; more weight was given to the most highly loaded items within each factor for naming purposes. After names were assigned to the factors, they were referred to as subscales.
Internal consistency reliability
First, internal consistency reliability was calculated within each subscale using Cronbach’s α. Acceptable Cronbach’s α is defined as greater than .7 [36]. Within each subscale, each item was evaluated for whether the subscale Cronbach’s α would increase significantly if the item were deleted. If removing an item would cause the Cronbach’s α for the subscale to change from being unacceptable to acceptable, the item was removed. Item-total correlations were evaluated as well with the target item-total correlation being greater than .3 [47]. After decisions were made about removing items within each subscale, the Cronbach’s α for the full scale was calculated.
Temporal stability
To evaluate stability of the measure over time, test-retest reliability was conducted between NeoEAT – Mixed Feeding scores that were collected from the same parent about the same infant 2–3 weeks apart. Bivariate correlations were calculated using Pearson’s product moment correlation (r, two-tailed) between the NeoEAT – Mixed Feeding scores in the initial survey with the NeoEAT – Mixed Feeding scores in the retest survey. Correlations were calculated between each subscale score as well as the total score. Because missing data would distort the subscale and/or total score and alter the test-retest reliability, cases with any missing data within each subscale were excluded from that subscale analysis and cases with any missing data at all were excluded from the NeoEAT – Mixed Feeding total test-retest reliability analysis.
Convergent validity
To evaluate convergent validity, scores on the NeoEAT – Mixed Feeding were evaluated for congruency with two other parent-report measures of related constructs: the IGERQ-R and the IGSQ. Bivariate correlations were calculated using Pearson’s product moment correlation (r, two-tailed) between the NeoEAT – Mixed Feeding scores (total and subscale scores), the IGERQ-R sum score, and the IGSQ sum score.
Known-groups validity
Known-groups validity was tested by comparing NeoEAT – Mixed Feeding total score and subscale scores between two groups that represented a subset of the sample from the factor analysis: 1) healthy infants with no feeding concerns and 2) infants with problematic feeding. To be included in the group of health infants with no feeding concerns, the parent had to report that the infant was born full-term, there were no feeding concerns, the infant did not take a prescription medication regularly, and did not have any of the following conditions: genetic disorder, congenital heart defect, developmental delay, or structural abnormality of the face, mouth, or gastrointestinal tract. To be included in the group of infants with problematic feeding, the parent had to report that either they thought the infant had a feeding problem, the infant had been diagnosed by a healthcare provider with a feeding problem, and/or the infant had a feeding tube. Data on infants who did not clearly fall into either of these categories were excluded from this analysis. Independent samples t-test was conducted comparing the two groups of infants for the NeoEAT – Mixed total score and all subscale scores.