pediatrics
September 2014, VOLUME134 /ISSUE Supplement 1

# Methods for the Year 6 Follow-Up Study of Children in the Infant Feeding Practices Study II

1. Sara B. Fein, PhDa,
2. Ruowei Li, MD, PhDb,
3. Jian Chen, MScb,
4. Kelley S. Scanlon, PhD, RDb, and
5. Laurence M. Grummer-Strawn, PhDb
1. aMcKing Consulting Corporation, Fairfax, Virginia; and
2. bDivision of Nutrition, Physical Activity, and Obesity, Centers for Disease Control and Prevention, Atlanta, Georgia

## Abstract

OBJECTIVE: We describe methods used in the Year 6 Follow-Up (Y6FU) of children who participated in the Infant Feeding Practices Study II (IFPSII). This study consists of a questionnaire administered 6 years after the IFPSII to characterize the health, development, and diet quality of the children.

METHODS: The Y6FU sample was a subset of those who participated in IFPSII. The IFPSII participants were drawn from a national consumer opinion panel; neither the IFPSII nor the Y6FU sample is nationally representative. The Y6FU sampling frame included all qualified participants who answered at least the first postnatal questionnaire. One questionnaire was administered by mail in 2012, and nonrespondents were contacted for a telephone interview. Survey topics included measures of health, development, diet, physical activity, screen time, and family medical history. We attempted to contact 2958 mothers and obtained completed questionnaires from 1542, a response rate of 52.1%. We conducted 2 sample evaluations, 1 comparing respondents and nonrespondents on data from IFPSII and the other comparing Y6FU respondents with 6-year-old participants in the National Survey of Children’s Health.

RESULTS: Y6FU mothers are more likely to be white, married, older, and of higher education and income than both nonresponders and nationally representative mothers. Comparisons also revealed health-related differences and similarities.

CONCLUSIONS: Although not nationally representative, the Y6FU provides a valuable database because of its wide coverage of diet and health issues and its unique ability to link early feeding patterns with outcomes at age 6 years.

• breastfeeding
• bottle feeding
• infant nutrition
• child nutrition
• cohort study
• health outcomes
• Abbreviations:
ECLS-B
Early Childhood Longitudinal Study: Birth Cohort
IFPSII
Infant Feeding Practices Study II
NHANES
National Health and Nutrition Examination Survey
NSCH
National Survey of Children’s Health
Y6FU
Year 6 Follow-Up
• The Infant Feeding Practices Study II (IFPSII) was a longitudinal study of a geographically diverse cohort of US mothers and their infants born between May 2005 and March 2006, with information about feeding and care collected 12 times in nearly monthly questionnaires from the last trimester of pregnancy through the infant’s first year of life.1 The Year 6 Follow-Up (Y6FU) study provided follow-up data on the IFPSII sample children through a single questionnaire administered in 2012, 6 years after the IFPSII, to characterize the health, development, and diet quality among children who participated in the IFPSII. In addition to these outcomes, information was collected on child care and school arrangements, cognitive stimulation at home, dental hygiene, physical activity of the child and the mother, screen time, sleep patterns, the eating environment, exposure to cigarette smoke and other inhaled contaminants, family medical history, maternal employment, health insurance coverage, and participation in government nutrition programs.

Both the IFPSII and the Y6FU were conducted collaboratively by the US Food and Drug Administration and the Centers for Disease Control and Prevention, and both studies were guided by a working group with representatives from other agencies in the Department of Health and Human Services and from the US Department of Agriculture. For the Y6FU, the Department of Health and Human Services participants in the working group included scientists from the 2 principal agencies and the Office of Women’s Health, Eunice Kennedy Shriver National Institute of Child Health and Human Development, the Maternal and Child Health Bureau in the Health Resources and Services Administration, and the Agency for Healthcare Research and Quality. Additional scientific support was provided by scientists at the Economic Research Service and the Food and Nutrition Service of the US Department of Agriculture and also by scientists who had participated in the IFPSII from universities and nonprofit organizations. The group included experts in nutrition, pediatrics, epidemiology, study design, dietary assessment, food allergy, child health, economics, and other areas.

The methods of the IFPSII have previously been published.1 This report describes the methods of the Y6FU, including a comparison of respondents and nonrespondents in the Y6FU based on data from the IFPSII and a comparison with 6-year-old children in a nationally representative study, the National Survey of Children’s Health (NSCH) 2011–2012.2

## Study Sample

The sampling frame for the Y6FU was mother–infant pairs who participated in the IFPSII. The IFPSII sample was obtained from a nationally distributed consumer opinion panel consisting of 500 000 households throughout the United States.1 Participants in the IFPSII cohort were ≥18 years old at the time of the prenatal questionnaire and had delivered a singleton infant who was ≥35 weeks’ gestation and weighed ≥5 pounds at birth. In addition, the mother and baby had to be healthy, with no condition likely to affect feeding.

To qualify for the Y6FU sample, mothers had to have completed the Neonatal (Month 1) questionnaire in the infant study (n = 3033) and not be subsequently disqualified from IFPSII (n = 75). Disqualification criteria included death of the infant or mother, being diagnosed with a condition likely to affect feeding, or living in a geographic area to which the post office stopped delivering mail because of the Gulf Coast hurricanes of 2005. In total, 2958 mothers qualified for the Y6FU.

## Data Collection and Study Approval

The data collection was conducted primarily by mail but also included telephone interviews with those who did not respond by mail. We verified the mailing address for each sample member in the National Change of Address files (records of people who have filed a change of address with the US Postal Service). If needed, we traced the current address through 4 databases (Targus, InfoUSA, Experian, and Acxiom) that compile information from various sources including the US Telecommunications Network, white page listings, public information, and proprietary sources. At each mail contact attempt, sample members with mailings returned as undeliverable were traced using the aforementioned databases to find a current address for the next mailing or to make it possible to find a telephone number for the telephone interview. Telephone interviews were attempted with nonrespondents to the mail contacts.

A notification letter announcing the study was mailed to eligible IFPSII participants in January 2012; the first survey packet was mailed in March 2012, followed by a postcard reminder. The survey packet was remailed to nonresponders in April 2012. The survey packet consisted of a letter of consent, the questionnaire and demographic form, a tape measure for measuring the child’s height, and instructions on measuring height according to standard protocol.

We attempted to conduct a telephone interview with every sample member who did not return a written questionnaire or indicate refusal and for whom a telephone number could be found (n = 1110). The telephone interview included all information asked on the written questionnaire, and questions were asked as similarly as possible to the written questionnaire.

## Sample Size and Response Rates

The final sample size for the Y6FU was 1542 (Table 2). We calculated response rates based on the recommendations of the American Association for Public Opinion Research.8 In calculating the response rates, we included both mail survey and telephone interview categories because of the mixed data collection method of the study. The main survey was sent to all IFPSII sample members for whom a mailing address could be found. The response rate was 52.1%. We also calculated other relevant measures of response, which required differentiating between sample members who were contacted but refused and those who were not contacted. However, for 374 mothers we could not distinguish between those who did not receive a questionnaire (ie, were actually not contacted) and those who received it but never responded (ie, refused). Their contact status was unknown, and we classified them as “not contacted” when calculating various rates. The cooperation rate was 95.0%, the refusal rate was 2.8%, and the contact rate was 54.9%. Because the sampling frame was not designed to collect representative data, sample weights were not calculated for these data.

TABLE 2

Disposition of the Y6FU Sample and Various AAPORa Response Rates

## Data Linking and Data Preparation

The completed Y6FU questionnaires were matched to the infant data from IFPSII by the participation number that appears in both the IFPSII data and the Y6FU data. We verified cases as matches by comparing the infant’s birthday with the birthday given for the 6-year-old child. If the birthdays did not match or the child’s birthday was left blank, we evaluated the child feeding characteristics and mother’s characteristics for consistency with the infant data. Because we recognized that some of the information could have been entered incorrectly, we did not require a match on all characteristics. We compared these variables: breastfeeding duration, infant participation in the Special Supplemental Nutrition Program for Women, Infants, and Children, whether the mother had gestational diabetes during her pregnancy with the 6-year-old, mother’s height, mother’s age, and educational attainment. We compared 71 cases that did not match on child’s birthday and rejected 15 of them because they did not match on the other characteristics.

We cleaned the Y6FU data by following a method similar to the cleaning of the IFPSII data. We consolidated variables that could have information entered into >1 field (eg, food intake could be entered as servings per day, per week, or per month). We set to missing or flagged biologically improbable or impossible values and conflicting or inconsistent responses. For example, if a mother said she returned to work when her 6-year-old child was 7 years old, the value was impossible; if the child usually slept >20 hours per night, the value was improbable. If a mother marked that the home had no pet and also marked ≥1 specific pets, the response was recorded as inconsistent.

## Sample Nonresponse Bias and Characteristics

Because the Y6FU sample is a subset of the IFPSII sample and is not nationally representative, we conducted 2 analyses to describe the potential sample bias. To describe how those who responded differed from those who did not, the first analysis compares respondents and nonrespondents to Y6FU. For this comparison, we used only information from the IFPSII for both groups because Y6FU data were not available for nonrespondents and demographic information (eg, maternal education and family income) could not be assumed to remain the same 6 years later. We compared demographic characteristics and also selected feeding and health characteristics. Income is shown as the percentage of the federal poverty level9 in 2005 because most of the infants were born in that year.

The second analysis compares Y6FU respondents with data on 6-year-old participants of the nationally representative NSCH, 2011–2012,2 using all questions that are identical or nearly so in both studies. The purpose of this analysis was to characterize the direction and extent of differences between the Y6FU sample and a nationally representative database of all 6-year-old children. Demographic characteristics of the Y6FU sample differed slightly from those in IFPSII because responses on geographic region, income, education, marital status, and other variables could have changed between the year of birth and year 6. Income is shown as percentage of the federal poverty level in 2012 for the Y6FU data.9

We used χ2 to test the significance of differences between respondents and nonrespondents to the Y6FU. For comparisons between Y6FU and NSCH, because of the complex sample design of the NSCH data, we did not have the ability to combine it with the data from Y6FU to conduct χ2 tests on the overall distributions. Therefore, we used Z tests for the differences of each variable category between the Y6FU and NSCH samples, and we applied sample weights and estimated the standard errors of the NSCH sample by using SUDAAN (RTI International, Research Triangle Park, NC) to account for its complex sampling design. We considered differences with probability of ≤.05 to be significantly different.

## Results

### Respondent and Nonrespondent Comparison

Mothers who responded to Y6FU were more likely than nonrespondents to be older, married, and white; to have higher education and higher income; to not smoke; and to have received prenatal care before 13 weeks (Table 3). In addition, respondents were more likely to be from the Northeast and less likely to be from the South. Respondents were much more likely to have returned ≥8 of the 10 postnatal questionnaires than nonrespondents. Respondents and nonrespondents did not differ by prenatal employment status, parity, having returned to work when the infant was ≤6 weeks old, or prepregnancy BMI.

TABLE 3

Distribution of Select Characteristics Among Respondents and Nonrespondents in the Y6FU, Based on Information From the IFPSII

Children of respondents were more likely than children of nonrespondents to have breastfed longer and not participated in the Special Supplemental Nutrition Program for Women, Infants, and Children. Children of respondents and of nonrespondents did not differ by infant gender, breastfeeding initiation, reporting a food allergy or intolerance during infancy, birth weight, or gestational age.

### Y6FU–NSCH Comparison

The comparison with the nationally representative NSCH sample shows that the Y6FU sample mothers were older, more likely to be married, had higher education, and were more likely to be white and less likely to be Hispanic (Table 4). Y6FU respondents were less likely to have a low household income and more likely to be from the Midwest or Northeast regions. The children in the 2 samples had a similar gender distribution and were similarly likely to have health care coverage, but the Y6FU children were more likely to have seen a dentist in the preceding year.

TABLE 4

Distribution of Select Characteristics Among Participants in the Y6FUa and Among 6-y-old Participants in the NSCH (2011–2012)b

The Y6FU children were less likely to have received food assistance in the past year. They were more likely to be homeschooled and less likely to attend public school than the NSCH children, and the Y6FU children were more likely to have an Individualized Education Program at school. The Y6FU children were more likely to have no screen time on weekdays, but they were similar to the NSCH children in the percentage who watched TV, videos, or played video games for more than an hour on weekdays. Y6FU children were less likely to currently have attention deficit disorder or attention deficit hyperactive disorder, autism or developmental delay (analyzed together), or asthma, but the percentages from the 2 samples with hearing problems were similar.

## Discussion

Linkage of the cross-sectional Y6FU data to data collected in the longitudinal IFPSII provides a unique ability to examine associations between detailed early feeding and health information and health, development, weight and height, and dietary patterns at 6 years old. The large number of potential confounders in the database enables researchers who are conducting multivariate analyses to control for many factors that might affect the associations of interest. The large sample size can support various types of multivariate analyses. The use of questions that have previously been used in national and international surveys enables researchers to compare the Y6FU results with those from other studies.

The main limitation of the study is that the sample was not nationally representative of the US population. The original IFPSII data were biased toward mothers who were older, more highly educated, of higher income, white, and less likely to smoke than mothers from the nationally representative National Survey of Family Growth.1 The sample’s loss to follow-up exacerbated these differences, as shown by the significant difference between Y6FU respondents and nonrespondents on age, education, income, race, and smoking status. The comparison of Y6FU respondents with 6-year-old children from the nationally representative NSCH confirmed the demographic and regional bias of the Y6FU sample. That comparison also indicated a similarity in some measures of health care access; however, most health outcomes indicated that the Y6FU sample was healthier than the NSCH sample. An additional limitation is that all data were reported by the child’s mother. No medical records were examined to verify weight, height, health status, or other medical characteristics.

A number of studies that have a longitudinal component and that collected information about infant feeding and subsequent health outcomes have been published in English in the past 15 years.1021 However, none of these studies included detailed measures of early feeding patterns, including breastfeeding intensity (the amount of breast milk relative to all milk), use of expressed milk, patterns of introduction of food groups, and diet quality in late infancy. In addition, nearly all these studies were conducted in a country other than the United States. Given different cultures, customary diets, exposures to contaminants, and other differences that occur between countries, it is possible that results in the United States would be different from those in other countries even when the same outcome is measured. Of the 2 studies conducted in the United States, one19 was in a specific state and concerned only diet, having no health outcomes. The second, the Early Childhood Longitudinal Study: Birth Cohort (ECLS-B)12 included only basic information about infant feeding collected retrospectively at 9 months and 2 years of age. The ECLS-B asked about duration of any and exclusive breastfeeding and the time of introduction of select foods including formula, solid food, and cow’s milk, but not fruits and vegetables or other food groups that would enable a measure of diet quality in infancy. In contrast to the other study conducted in the United States, the ECLS-B has the advantage of measuring some health outcomes at later ages, including at the year of kindergarten, which is similar to the age of the Y6FU sample.12

## Conclusions

The Y6FU database, when linked to the IFPSII database, is a valuable resource for examining associations between infant diet, care, and health status and a large number of health, diet, and development outcomes at childhood age 6 years. Its strengths include a large sample, availability of information on a large number of potential confounders, and detailed information on feeding collected nearly monthly during the infant period. The main limitation is that the sample is not representative of the US population and overrepresents children from economically and educationally advantaged families. The IFPSII data have proved to be a valuable resource for many researchers, and the Y6FU data should prove to be similarly valuable, especially when they are linked to IFPSII.

## Footnotes

• Accepted May 20, 2014.
• Address correspondence to Ruowei Li, MD, PhD, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Nutrition, Physical Activity, and Obesity, 4770 Buford Hwy, Mail Stop K25, Atlanta, GA 30341. E-mail: ril6{at}cdc.gov
• The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

• Dr Fein participated in the conceptualization and design of the study, managed the data collection, participated in data preparation and analysis, and drafted the initial manuscript; Drs Li and Grummer-Strawn participated in the conceptualization and design of the study, participated in data preparation and analysis, and reviewed and revised the manuscript; Mr Chen conducted the analysis and reviewed and revised the manuscript; Dr Scanlon reviewed and revised the manuscript; and all authors approved the final manuscript as submitted.

• FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

• FUNDING: This study was funded by the US Food and Drug Administration, Centers for Disease Control and Prevention, Office on Women's Health, National Institutes of Health, and Maternal and Child Health Bureau in the US Department of Health and Human Services.

• POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.