The first 2 years of life are critical to the growth,

Transcription

1 RESEARCH Original Research The Feeding Infants and Toddlers Study 2008: Study Design and Methods RONETTE R. BRIEFEL, DrPH, RD; LAURA M. KALB; ELIZABETH CONDON, MS, RD; DENISE M. DEMING, PhD; NANCY A. CLUSEN, MS; MARY KAY FOX, MEd; LISA HARNACK, DrPH, RD; ERIN GEMMILL, PhD; MARY STEVENS; KATHLEEN C. REIDY, DrPH, RD ABSTRACT Objective Describe the study design, data collection methods, 24-hour dietary recall protocol, and sample characteristics of the Feeding Infants and Toddlers Study (FITS) Design A cross-sectional study designed to obtain information on the diets and feeding patterns of US infants, toddlers, and preschoolers ages birth to 47 months. Telephone interviews with parents and caregivers were conducted from June 2008 through January 2009 and included a household interview to recruit the household and collect information on household and child demographics and nutrition-related characteristics, and a dietary interview, including a 24-hour dietary recall collected using the 2008 Nutrition Data System for Research. A second dietary recall was collected on a random subsample to estimate usual nutrient intake distributions. Data collection instruments were built on those used in FITS 2002, with expanded survey content to address emerging issues in childhood nutrition and obesity. The dietary protocol was improved to increase reporting accuracy on portion sizes, and a bridging study was conducted to test effects of the changes in the food model booklet and protocol since FITS 2002 (n 240 cases aged 4 to 23 months). R. R. Briefel is a senior fellow, and N. A. Clusen is a senior statistician, Mathematica Policy Research, Washington, DC. L. M. Kalb is a senior survey researcher, E. Condon is a nutrition research analyst, and M. K. Fox is a senior fellow, Mathematica Policy Research, Cambridge, MA. D. M. Deming is a principal scientist, nutrition, and K. C. Reidy is head, nutrition, both with Global Meals & Drinks, Nestlé Infant Nutrition, Florham Park, NJ. L. Harnack is director and M. Stevens is service center manager, Nutrition Coordinating Center, University of Minnesota, Minneapolis. E. Gemmill is a survey researcher, Mathematica Policy Research, Princeton, NJ. Address correspondence to: Ronette R. Briefel, DrPH, RD, Mathematica Policy Research, 600 Maryland Ave SW, Suite 550, Washington, DC rbriefel@mathematica-mpr.com STATEMENT OF POTENTIAL CONFLICT OF IN- TEREST AND FUNDING/SUPPORT: See page S26. Manuscript accepted: August 11, Copyright 2010 by the American Dietetic Association /$36.00 doi: /j.jada Subjects A national random sample of 3,273 infants, toddlers, and preschoolers from birth up to age 4 years, with 2 days of dietary intake data for 701 cases. Results Among sampled households with an age-eligible child, the response rate was 60% for the recruitment interview. Of recruited households, the response rate for the dietary interview was 78%. Conclusions The FITS 2008 provides rigorous, well-tested methods and survey questions for nutrition researchers to use in other dietary studies of young children. FITS 2008 findings on the food and nutrient intakes of US children from birth up to age 4 years can inform dietetics practitioners, pediatric health practitioners, and policymakers about the dietary issues of young children. J Am Diet Assoc. 2010;110:S16-S26. The first 2 years of life are critical to the growth, development, and well-being of children (1). Feeding practices and food consumption contribute to infants and young children s nutrient adequacy or excess and also establish dietary patterns and habits that can last a lifetime. Recent research on childhood obesity has shifted from older childhood and adolescence to early nutrition in the preschool years and to examining children s food environment (2-4). Understanding infant feeding practices such as the initiation and duration of breastfeeding, the timing of introduction to solid foods, and toddlers and preschoolers meal and snack patterns is important to identify areas for improvement in establishing healthy eating patterns at an early age. The Feeding Infants and Toddlers Study (FITS) 2002 was designed to assess the food and nutrient intake patterns of infants and toddlers aged 4 to 24 months in a cross-sectional study in the United States (5). Research using data from FITS 2002 resulted in a comprehensive assessment of the diets of infants and toddlers and addressed topics such as infant feeding practices in relation to recommendations, nutrient adequacy, consumption of foods and food groups, food sources of energy, nutrients, and meal and snack patterns. Data from FITS 2002 filled a gap in information since dietary data from the National Health and Nutrition Examination Survey (NHANES) do not include breastfed infants and children in the estimates of food and nutrient intakes nor dietary findings for infants and toddlers by small age categories (ie, data are typically reported for age groups spanning 2 or more years). FITS 2002 addressed this information gap by imputing breast milk volumes and collecting information on sufficient sample sizes to allow for more precise estimates S16 Supplement to the Journal of the AMERICAN DIETETIC ASSOCIATION 2010 by the American Dietetic Association

2 of food and nutrient intakes by 3-month age groups of infants and toddlers. Overall, the FITS 2002 findings showed that the diets of infants and young children were adequate, but that by as early as age 9 to 11 months, the less-healthy eating patterns of older children and adolescents (eg, infrequent consumption of fruits and vegetables; consumption of higher-calorie, lower-nutrient items such as sweetened beverages and salty snacks) were beginning to emerge (5,6). Research findings using the FITS 2002 data were disseminated in more than 25 peer-reviewed publications, including two supplements to the Journal of the American Dietetic Association in 2004 and 2006, and more than 20 presentations at scientific and professional meetings. The FITS 2008 builds on the FITS 2002 study (7) by using a similar design and sampling plan and recommended dietary data collection and analysis methods (8,9), and expanding the age range from birth up to age 4 years. This article describes the FITS 2008 study design; survey methodology; sample characteristics; and the 24- hour dietary recall protocol, coding, and analytic techniques used to assess food and nutrient intakes. It provides detailed information on the FITS 2008 study design and methods to support FITS research articles in this and future publications. STUDY DESIGN The FITS 2008 was designed to answer two main research questions: How have food and nutrient intakes of infants and toddlers aged 4 through 23 months changed during the 6 years since FITS 2002? and, What are the food and nutrient intakes of younger infants (birth to age 3 months) and preschoolers (aged 24 to 47 months)? FITS 2008, a cross-sectional study designed to obtain information on the diets and eating patterns of US infants, toddlers, and preschoolers ages 0 to 47 months (ie, up to age 4 years), included the key design features shown in Figure 1. Sample Frame The FITS 2008 target population was all children from birth up to age 4 years living in the 50 states and the District of Columbia. Researchers designed the sampling plan so that results could be generalizable to infants, toddlers, and preschoolers from birth up to age 47 months living in the United States. Since a comprehensive and timely sample frame that includes new births does not exist, and random-digit dialing to locate a relatively rare group (young children) is costly and time-sensitive, targeted-list frames from a commercial vendor, the New Parent Database and the Consumer Database from Experian, Inc, were used as the sample frame (2008, Experian Inc, Costa Mesa, CA). The New Parent Database includes both prenatal and postnatal records and tracks names from birth to age 36 months. The Consumer Database includes records for households with children older than age 36 months. The methods used to gather names are proprietary. The sources include parenting magazines, baby store mailing lists, and baby product Household recruitment interview Screened households for eligibility. Recruited households with child between age 0-47 mo. Sampled 3,200 children to ensure adequate sample sizes for 12 different age groups of infants and toddlers aged 0-47 mo (ie, 3-mo age groups for infants and toddlers and 6-mo age groups for preschoolers) and important subgroups such as Hispanics. Collected data for eligible child (household and child demographics and nutrition-related characteristics). Used enhanced data collection instruments building on those used in FITS 2002, including expanded survey content on time spent on physical activity and television/video watching and family meals. Dietary interview Parents or caregivers of all children completed one 24-h dietary recall by telephone. A random sample of 701 parents and caregivers completed a second 24-h recall for use in estimating usual nutrient intake distributions and examining the prevalence of inadequate and excessive nutrient intakes (8,9). Used enhanced methods for collecting 24-h dietary recalls, including an updated food model booklet and an improved protocol for estimating volume of usual beverage containers. Conducted a bridging study to assess the effect of the methodologic improvements in dietary data collection between the FITS 2002 and the FITS 2008 on dietary intake estimates on a minimum target sample of 200 infants and toddlers aged 4-23 mo. Data analysis Incorporated the complex survey design and sample weights to reflect the population ages 0-4 y in data analyses. Group-level assessment of nutrient intakes used the second day of intake collected on a random subsample; usual nutrient intake distributions are then compared to the Dietary Reference Intakes using recommended analytic approaches (8,9). The proportion of the group consuming certain types of foods/food groups and mean food intake are based on 1 d of intake per person, a valid analytic approach for grouplevel estimates (8). Figure 1. Key design features of the Feeding Infants and Toddlers Study (FITS) mailing lists. The data were weighted to vital records birth data to reduce bias due to frame coverage error (10). The two databases from Experian contain records from all 50 states and the District of Columbia and, therefore, form a national sample frame. The sample design specified the selection of records from 12 age groups and included five sample draws composed of nearly 47,000 households to reach the target number of participants for the 2008 study. Cases were randomly selected for the main study. Twenty-eight percent of the main study sample was randomly assigned to have a second 24-hour dietary recall. A different subsample of children ages 4 to 23 months was selected for the bridging study. December 2010 Supplement to the Journal of the AMERICAN DIETETIC ASSOCIATION S17

3 SURVEY METHODOLOGY An important consideration in planning and designing FITS 2008 was the desire to assess changes that had occurred in the dietary patterns of 4- to 23-month-olds between 2002 and To support this research objective, it was important that the methodology used in FITS 2008 be consistent with that of FITS 2002 (7) and use recommended methods for collecting and analyzing dietary data for groups (8,9) (Figure 1). The survey instruments used in FITS 2008 were adapted from those developed for FITS 2002, and both studies used the Nutrition Data System for Research (NDSR) developed by the Nutrition Coordinating Center (NCC) at the University of Minnesota to collect and process the dietary intake data (version 2008, 2008, University of Minnesota, Minneapolis). Before conducting any data collection, new survey questions and the revised food model booklet were pretested with mothers and fathers of children in the target range. All study instruments and procedures were reviewed and approved by Mathematica s independent institutional review board (Public/Private Ventures in Philadelphia, PA). Likewise, the review board was consulted before changes were made to the survey and to the incentive amounts that were needed to improve completion rates. Recruitment Materials and Household Interview Regardless whether a sample member was selected for the main study, or a part of the bridging study or second 24-hour recall sample, the protocols for contacting sample members and conducting the recruitment interview were identical. All sample members were sent advance postcards before attempting to complete the recruiting telephone interview. In addition, an important aspect of the recruitment interview was to obtain consent to continue with the 24-hour dietary recall interview and to collect accurate contact information for sending a box of materials needed for the dietary interview. Willing respondents were mailed boxes along with a $20 incentive for participation. All study materials were available in English and Spanish. Certified translators translated and back-translated survey instruments using recommended procedures. Recruitment interviews were conducted by Mathematica s trained telephone interviewers and averaged about 15 minutes to administer using a Computerized Assisted Telephone Interviewing system. Questions were asked about demographic characteristics of the household and sampled child, the child s feeding patterns, growth and development, physical activity levels, television/video screen time, as well as other nutritionrelated topics (see Figure 2). Dietary Interview Upon completion of the household recruitment interview by telephone, respondents were mailed a box of materials needed to complete the dietary interview. The box included a brochure about the study, a cover letter explaining the next steps in the study, and a $20 incentive check to thank respondents for completing the dietary interview. For respondents who had not been reached after multiple attempts, a letter was mailed offering an additional incentive of $20 to complete the dietary interview. The box also included a food model booklet and other measurement aids and materials for the 24-hour dietary recall interview (see Figure 3). The dietary interview included two components: a 24- hour dietary recall and a short dietary questionnaire to provide information that could not be captured in the recruitment interview. Because of potential changes in children s eating behaviors from week to week, it was critical to ask questions on the introduction of foods (eg, milk, baby cereal, pureed baby foods, 100% juice, and yogurt), current breastfeeding, most recent measured height and weight, and developmental stage at the time of the 24-hour dietary recall (rather than at the recruitment interview). The developmental stage was based on physical milestones and the progression of feeding skills during the first 2 years of life (11,12). Interviewers recorded responses to the dietary interview questionnaire on hard-copy forms (that were later entered in an Excel database [2003, Microsoft Corp, Redmond WA] for ease in coding and analysis), while the 24-hour dietary recall interview was conducted entirely on computer using NDSR. Trained and certified dietary interviewers at the University of Minnesota s NCC began contacting respondents by telephone approximately 10 days after materials were mailed. In practice, there was a variable length of time depending on how long it took to reach a respondent. NCC interviewers then completed the dietary questionnaire and 24-hour dietary recall using the 2008 NDSR. Those who were selected for a second day recall were called 7 to 10 days after the first recall and sent an additional $10 incentive after completing the second interview. During the dietary interviewing period, Mathematica researchers closely monitored the number of completed interviews by age group, intake day (Day 1 and Day 2 intakes), and day-of-the-week, and worked with NCC to obtain a representative distribution of intake days while meeting study goals for sample sizes. Further, previous research indicates significant weekend/weekday differences in the percentage of energy from fat and protein among children aged 2 to 5 years (13). Approximately 26% of the main study dietary recalls were collected for weekend intakes (ie, Saturday or Sunday) and 74% reflected intake on weekdays (ie, Monday through Friday). Fifty-six dietary interviews were conducted in Spanish. 24-Hour Dietary Recall Protocol The energy intakes from FITS 2002 suggested that there may have been some degree of over-reporting of amounts of foods and beverages by some respondents (5,14). To address this concern in FITS 2008, several changes were made to the 24-hour dietary recall protocol in an effort to obtain more accurate assessments of food intake. First, in addition to the initial instructions used in FITS 2002 to remind respondents that they should only report those foods actually consumed by the child, respondents were reminded of this throughout the interview in Second, the food model booklet contained significant enhancements from the FITS 2002 version. Third, the In- S18 December 2010 Suppl 3 Volume 110 Number 12

4 Data element Recruitment/household interview Dietary interview/first 24-h dietary recall Second 24-h dietary recall Household and caregiver characteristics Demographic characteristics a Mother s height and weight Language spoken at home Family eating (dinners eaten together and fast food per week) Child characteristics Demographic characteristics b Infant feeding practices Introduction of foods Self-feeding and picky eating Food allergies Medical problem that affects eating Type of water for drinking and mixing with foods WIC c participation and foods from WIC Birth weight Height and weight Developmental milestones Physical activity and television/video watching d Child care e e Child dietary outcomes Food and beverage choices Sources of energy and nutrients Meal and snack patterns Nutrient intake Dietary supplement use e f f a Includes household composition, annual household income, parent s (main caretaker s) sex, race, Hispanicity, and education level; mother s age; parents marital status and employment. b Includes child s age, sex, race, and being Hispanic. c WIC Special Supplemental Nutrition Program for Women, Infants, and Children. d Ages 12 mo and older. e Indicates on day of recall. f Usual nutrient intake distributions use two days of intake per sampled child. Figure 2. Data elements by data collection instrment, in the Feeding Infants and Toddlers Study (FITS) FITS 2008 study letter FITS 2008 study brochure 2-c plastic measuring cup and instructions for measuring cup used most often by child Information about Foods Fed by Other Adults form to record foods fed at child care and instructions for use Cardstock ruler 14-page food model booklet List of frequently forgotten foods Two-dimensional drawings of feeding utensils (8 cups [6 sippy, 2 without lids], 6 bowls, 4 spoons (infant, toddler, teaspoon, and tablespoon) Models to measure shapes (5 x 5-in grid, 8-in concentric circles, 8-in wedge) Color pictures of mounds (8 mounds ranging from 1 tsp to 1.5 c) Developmental chart Figure 3. Feeding Infants and Toddlers Study (FITS) 2008 materials mailed to respondents before the dietary interview. All study materials were available in English and Spanish. formation about Foods Fed by Other Adults form was developed to gather information about foods the child may have consumed while in someone else s care. Fourth, respondents were asked to measure their child s sippy cups before the interview (see details below), and fifth, the collection of information on breastfeeding was simplified. The NDSR 2008 was used to collect data on all foods and beverages consumed from midnight through midnight on the previous day. The inside cover of the food model booklet included a list of frequently forgotten foods to remind respondents to report commonly omitted foods. Taken together, the multiple-pass approach used in the NDSR and interviewers probes aimed to minimize respondent error and increase accuracy in food intakes. Infant and Toddler Foods. Data entry rules provided specific guidance for entering baby foods, infant cereal, infant juice, infant formula, and breast milk. Interviewers probed for brand name information for infant and toddler products to distinguish commercial infant/toddler foods from homemade baby food or table foods, and regular December 2010 Supplement to the Journal of the AMERICAN DIETETIC ASSOCIATION S19

5 adult foods from infant/toddler foods. All infant/toddler foods and beverages are categorized together in NDSR, distinct from regular adult foods. Before data collection was begun, the database was updated to include the most recent products specifically marketed as infant and toddler milks and food products. When a reported infant/ toddler food was not available in the database, the item was entered as missing along with details (brand name, description, preparation details, package size, dimensions, weight or number consumed) and resolved after data entry was completed for the recall (see data processing text below). Infant cereal was entered as prepared or unprepared based on the form in which it was reported. Interviewers probed to determine whether the cereal was plain or mixed with other components and what fluid was used in preparation (eg, breast milk, formula, or water). Regular juice was entered if a respondent did not know if the juice was infant juice or regular juice. Respondents were probed to determine whether the juice was diluted (with water) before consumption. Infant formulas were identified by brand name (when available) and form (powder, concentrate, and ready-to-feed). The database also included generic versions of infant formula (eg, milk-based, soy-based), as well as store-brand formulas. Breast Milk. Breast milk that was breastfed to a child was not quantified during the dietary recall. Instead, a breastfed milk entry was selected and 1 serving was entered for each breastfeeding occasion. The amount of breast milk for breastfed infants was estimated during data processing (see details below). When breast milk was pumped and fed to a child in a bottle, the human milk option in NDSR was selected and the respondent was asked to quantify the amount consumed at each feeding. Portion Size Estimation. Mathematica researchers worked with a graphic artist to create a food model booklet appropriate for the diets of infants, toddlers, and preschoolers. The booklet included life-size, two-dimensional drawings of popular infant/toddler/preschooler sippy cups, cups, and glasses (eight total drawings); four spoons, including two infant/toddler feeding spoons, a teaspoon, and tablespoon; six drawings of bowls; and eight color mounds ranging from 1 tsp to 1.5 c (see Figure 3). The food model booklet was similar to that used in FITS 2002 (7) but included updated versions of infant and toddler cups and bowls to reflect 2008 products, and additional estimation tools including color mounds, a wedge model, and cardstock ruler. Some drawings were taken from the Food Model Booklet validated for the NHANES What We Eat in America survey (15). When respondents used a portion estimation model to quantify a food or beverage, interviewers used a conversion guide to determine how to enter the quantity of a reported food into the NDSR. Respondents could also quantify foods and beverages using household measuring cups and spoons, food labels or packages, or by describing standard-size foods (eg, one medium egg, two regular slices of bread). Measuring Beverage Container(s) before the Dietary Interview. The wide variety of infant/toddler/preschooler cups (including sippy cups) available on the market created potential challenges with having respondents quantify the amount of beverages consumed. Cups typically ranged in volume from 5 to 16 fl oz and were available in multiple shapes and sizes. We found that cups of similar shape often varied in volume, particularly for insulated cups that hold substantially less volume than similarly sized cups that are not insulated. To enhance the ability of respondents to provide accurate reports of the amounts of beverages their child consumed, respondents were asked to measure the volume of the cup their child used most often, using the 16-fl oz plastic measuring cup and instructions that were provided. If a child typically used more than one cup on a given day, the respondent was asked to measure the cups used most often. Foods Consumed at Child Care. If the child consumed any foods or beverages while in someone else s care on the day of the recall, interviewers used a special protocol if the respondent was unable to provide the required details. Interviewers worked with respondents to determine the most convenient way to obtain and submit information on these foods to NCC, either from the mother, father, or from a child care provider using the form developed for this purpose, the Information about Foods Fed by Other Adults form. If the respondent was unable or unwilling to contact the child care provider directly, the interviewer offered to contact the child care provider to obtain the information needed to complete the dietary recall. Approximately 2% of cases required call backs to child care providers. If the child care provider could not remember or did not participate, the recall was considered incomplete and was not included in the food and nutrient analysis. Dietary Supplements. Information on dietary supplement use on the day of the recall was collected using four of the 10 screening questions in the NDSR Dietary Supplement Assessment Module. Respondents reported the child s use of multivitamins, vitamins and minerals, prescription vitamins or minerals, fiber supplements, and over-thecounter antacids. After answering the screening questions, respondents were asked to gather product containers to provide relevant supplement details. Once the supplement was located in the database using product label details, the respondent reported the number of times taken the previous day and the amount taken each time. If the respondent did not have the container or could not provide further details, the NDSR Dietary Supplement Assessment Module database defaults or closely matched generic products were entered. When a reported product was not available in the database, the product was entered as a missing product and the respondent was asked to provide as much detail as possible, including product name, ingredients, serving size quantity, and serving size unit. A new supplement products code was created for missing dietary supplements using a similar procedure as for missing foods in the 24-hour dietary recall interview. Data Processing. Quality assurance procedures were conducted at several levels during data entry and processing. Interviewers reviewed and edited the dietary recalls immediately after administration, and NCC quality assurance scientists reviewed interviewer notes and food and nutrient outlier and error reports for 100% of recalls. In addition, a 10% random sample of recalls was selected for a 100% quality control review. This included a review of the record S20 December 2010 Suppl 3 Volume 110 Number 12

6 Table 1. Feeding Infants and Toddlers Study 2008 coding rules for assigning breast milk volumes in 24-h dietary recalls Age at recall (source) a Exclusively breastfeeding b Partially breastfeeding c Daily volume (fl oz) Infants mo (16) d e mo (17) d e Toddlers and preschoolers Volume per feeding mo (17,18) mo (18) 2 a Coding rules derived from data in references (16-18). b Only breastfeeding and/or breast milk in a bottle as the source of milk consumed as a beverage. c Any combination of breastfeeding/breast milk and infant formula, cow s milk, flavored milk, soy milk, or rice milk as a beverage. d The volume of measured breast milk fed from a bottle is subtracted from the daily volume to calculate the volume from the breastfeedings. The total volume from breastfeeding is then equally divided among the number of (non-bottle) breastfeeding occasions. e If the total daily volume of other milks exceeds the daily value listed for infants, apply the rule of3flozperfeeding. The volume of breast milk is equally divided among the number of (non-bottle) breastfeeding occasions. header and food, dietary supplement, and trailer question tabulations for completeness and accuracy. Dietary recalls that were incomplete (primarily due to unknown child care information) underwent additional review to determine the extent of missing information. This included unknown meal, food, and/or portion size information. If no information was retrieved for one or more meals or two or more snack eating occasions, the recall was assigned a final status of incomplete (n 30 cases). Recalls with one unknown snack-eating occasion were assigned a final status of complete. If some basic information was provided for a meal or food, the recall was considered complete and missing information was imputed. When basic food detail was provided by the respondent for an unknown food (eg, cereal ), specific food details and portion sizes were imputed based on the child s age group ( 2 years and 2 years). The mode, or most frequently reported, food was selected (eg, oat-ring cereal) and median portion size for age was assigned (eg, 1 2 c). Among the completed Day 1 recalls, 430 (13%) contained breastfeeding occasions among infants and children ranging in age from 0 to 42 months (some also contained measured amounts of breast milk fed from a bottle). Consulting the available literature, we assigned breast milk volumes based on the child s age in months and the total amount or volume of other milks (infant formula, cow s milk, or soy milk) reported over the course of the recall day (see Table 1 for the coding rules for breast milk volume) (16-18). The NDSR 2008 includes more than 18,000 foods (more than 7,000 brand-name products) with values for 156 nutrients, nutrient ratios, and other food components. The NDSR 2008 also links to the Dietary Supplement Assessment Module database, which includes dietary supplements reported during the NHANES and includes a comprehensive ingredient file that allows for the addition of supplements not found within the database. Before data collection was begun, the database was updated to include the most recent products currently specifically marketed as infant and toddler milks and food products. When a reported infant/toddler food was not available in the database, the item was entered as missing along with details (brand name, description, preparation details, package size, dimensions, and weight or number consumed) and resolved after data entry was completed for the recall. A total of 358 foods and 65 dietary supplements were reported and added to the FITS 2008 database. The majority (299 of 358) of the foods added were infant and toddler foods (eg, infant cereals, infant formula, juices, pureed baby foods, toddler dinners, crackers, and cookies). There were 4,897 unique foods reported in the FITS All foods and beverages reported in the 24-hour dietary recalls were assigned by Mathematica nutrition researchers to food groups in a manner consistent with those used for the food group analysis in FITS The FITS 2002 food groups were updated and expanded, as needed, to incorporate new foods and beverages reported in FITS 2008 and to address the research objectives on consumption of foods and food groups/subgroups. BRIDGING STUDY The purpose of the bridging study was to test the effect of the enhancements to the 2008 food model booklet and 24-hour dietary recall dietary protocol, and provide information to help interpret any observed differences in intake between FITS 2002 and FITS Three hundred fifty-nine cases were selected for the bridging study. The bridging study was powered to detect a 100-kcal mean difference between methods among 4- to 11-month-olds and 12- to 23-month-olds. It was not feasible to include the much larger sample sizes that would have been required to detect smaller energy differences, but we wanted to assure that the improvements made were not associated with large energy differences that would make comparisons between the two surveys difficult. Respondents first completed an interview using the 2002 dietary protocol and 2002 food model booklet, and then a second interview (about 10 to 14 days later) using the 2008 protocol and 2008 food model booklet and measurement aids. Since the 2008 method included more detailed measuring and probing which could potentially bias subsequent interviews using the 2002 method, all bridging study respondents participated in the 2002 version first. Respondents selected for the bridging study were excluded from the main study sample asked to complete a traditional second-day recall interview (since this would have been their third intake interview). After editing for 24-hour dietary recall completeness, 240 cases were in the eligible age range of 4 to 23 months (ie, 4.0 to 23.9 months) and had a complete 2002 method recall and a 2008 method recall. On average, the infants and toddlers were 0.4 months (or 2 weeks) older at the second recall using the 2008 method. There were no significant differences in demographic characteristics between the bridging study sample and the full study sample. We conducted a series of descriptive tabulations strat- December 2010 Supplement to the Journal of the AMERICAN DIETETIC ASSOCIATION S21

7 Figure 4. Mean energy intake (kcal/day) in the Feeding Infants and Toddlers Study (FITS) 2008 bridging study (n 128 infants, n 112 toddlers) and the full study (n 671 infants, n 925 toddlers). ified by age group (4 to 5, 6 to 11, and 12 to 23 months) to identify potential differences between the 2002 and 2008 methods for infants and toddlers mean intake of energy, macronutrients, micronutrients, daily intake and portion sizes of commonly consumed foods, and the proportion consuming specific foods and food groups. Because one dietary recall is not an accurate estimate of an individual s usual dietary intake, paired t tests would not be considered a useful measure. Instead, we compared the group s mean intake using the FITS 2002 method to the mean based on the FITS 2008 method ( kcal/day vs kcal/day for infants; 1, kcal/day vs 1, kcal/day for toddlers). Overall, there were no significant mean differences in energy (Figure 4), macro- or micronutrient intake, or the proportion consuming (or the percentage of energy from) specific foods or food groups (data not shown). Mean energy intakes were in a direction consistent with the cases being a few weeks older at the second (2008 method) recall. Further, mean energy intakes in the bridging study were not significantly different from the 2008 study (full sample: for infants and 1, for toddlers), providing confidence in the bridging study results (Figure 4). There were no differences in mean portion size between the 2002 and 2008 methods with one possible exception among toddlers, the mean portion size of the food group water and other beverages (including juice/juice drinks unknown if 100% and unsweetened tea and coffee) was 22 to 24 g lower using the 2008 method compared to the 2002 method (see Table 2). The changes in the food model booklet and the measurement of sippy cups used in 2008 may have led to a small ( 1 fl oz) difference in the portion size estimation of water and selected other beverages. However, differences were not observed for other types of beverages and these findings could have occurred by chance (Table 2). Among toddlers, mean differences in portion size between the 2002 and 2008 methods were very small and not significant (eg, 1/3 fl oz cow s milk, 2 tsp mashed potatoes or yogurt, 1 Tbsp oat-ring cereal, and 1 bite mashed banana). The bridging study was underpowered to detect small differences in dietary intake. Larger sample sizes may have detected smaller energy differences, but it is possible that measurement error and reporting biases inherent in 24-hour dietary recalls outweigh modest improvements in the food model booklet and/or dietary protocol. These findings suggest that the changes made to the 2008 food model booklet visual aid booklet have little overall effect on group-level estimates of energy, foods, and portion sizes, and that comparisons between 2002 and 2008 can be made with a reasonable level of confidence. RESPONSE RATES AND SAMPLE SIZES Household recruitment interviews were conducted during a 22-week period in June through vember A total of 4,339 eligible sample members completed the recruitment interview. A total of 69 interviews (1.6% of the total) were conducted in Spanish. Completion rates were calculated for both stages of the study at the recruitment interview and at the dietary interview. Of those sampled households that could be S22 December 2010 Suppl 3 Volume 110 Number 12

8 Table 2. Feeding Infants and Toddlers Study Bridging Study: Mean amount consumed (in grams) a among consumers per eating occasion by food group Child Age mo b Food/beverage 2002 method (n 112) 2008 method (n 123) 4 mean standard error 3 Beverages 100% fruit juice Apple juice Milk and milk products Milk Cow s milk Infant formula Other beverages (includes unsweetened tea) * Water * Sweetened beverages Fruit-flavored drinks Food groups Fruit (includes 100% juices) n baby food fruits Fresh or frozen Banana Grains and grain products Bread, rolls, biscuits, bagels, tortilla Cereal, non-infant Ready-to-eat Crackers, pretzels, rice cakes Meats and other protein sources Milk products (includes cheese, yogurt) Cheese Mixed dishes Sweets, sweetened beverages, and desserts c Cakes, pies, cookies, and pastries Other cookies Vegetables Cooked Starchy vegetables Potatoes a Examples of portion sizes in household measurements include: 181 g of whole cow s milk is 3 4 c; 74 g banana is 1 2 c sliced; 13 g oat-ring cereal is about 1 2 c; 57 g yogurt is about 1 4 c; and 108 g macaroni and cheese is about 1 2 c. b Age at dietary recall interview. c Food group includes sweetened beverages and fruit-flavored drinks, also shown above under Other Beverages. *P 0.05 for difference between means. located by telephone and that had a child in the eligible age range, the completion rate to the recruitment interview was 60%. Among those respondents who completed the recruitment interview, dietary recalls were completed for 78%, resulting in an analytic response rate of 47% Table 3. Feeding Infants and Toddlers Study 2008 sample sizes of infants, toddlers, and preschoolers with complete 24-h dietary recalls Sample Size Age group (mo) Day 1 Day b a Total 3, Total sample for food and nutrient analysis 3,273 a 701 b a One outlier reporting parenteral nutrition and a small amount of formula and a total of 53 kcal for the 24-h period was excluded from the food and nutrient analysis. b One outlier of a very sick child consuming a small amount of formula and rice-soaked water and a total of 59 kcal for the 24-h period was excluded from the food and nutrient analysis. ( ). Among those respondents with a Day 1 recall, 80% asked to complete a Day 2 recall did so. nrespondents to the Day 2 recall were not significantly different from those who provided a Day 2 recall. Despite using the same sample frame as in FITS 2002 and considerable effort to locate respondents, approximately 71% of sampled households could not be located; that is, the sample frame did not include a telephone number and no number could be located, the number provided was a wrong number, or no one answered after repeated call attempts. Of those households that were located, 44% did not have a child in the eligible age range. The increase in the use of cellular telephones may have contributed to the greater difficulty in locating households than in FITS 2002 because cellular telephone numbers are less likely to be listed in telephone databases. In general, other large-scale national studies have also observed decreases in telephone survey response rates during the past decade (19-21). Dietary interviews were conducted over a 30-week period from June 2008 through January Among cases that had a dietary recall interview, a small number were excluded for the following reasons: incomplete due to missing meals at child care (n 30), and child had aged-out at 54 months or older (n 17). After conducting data checks for age-at-recall and completeness of the 24-hour dietary recalls, there were 3,378 Day 1 recalls and 725 Day 2 recalls available for data analysis for the 2008 main study (Table 3). There were 104 cases that were younger than 48 months at study recruitment, but age 48 to 53.9 months at the time of their 24-hour dietary recall. These cases were retained for later analysis, but are not included in the findings for children ages birth up to 4 years (ie, 0 to 47.9 months). One Day 1 recall was excluded from the dietary analysis because December 2010 Supplement to the Journal of the AMERICAN DIETETIC ASSOCIATION S23

9 nutrients could not be assigned for the reported parenteral nutrition. Therefore, the final sample size for the FITS 2008 food and nutrient analysis reported here is 3,273 cases aged 0 to 47.9 months (Table 3). Sample Weights and Design The sample weights were calculated to reflect the probabilities of selection. Sample weights were adjusted for nonresponse to the recruitment interview and the dietary interview and coverage of the target population. A set of parent and child variables including age, marital status, Special Supplemental Nutrition Program for Women, Infants and Children participation, and employment states was used for the dietary recall nonresponse adjustments. Poststratification adjustments were calculated to reflect the US population from birth up to age 4 years and were based on the child s age in months at the time of the recruitment interview and mother s race (10). The sample design and analysis weights are incorporated into the analysis of FITS 2008 data. Sample Characteristics Table 4 shows the weighted characteristics for the dietary sample aged 0 to 47.9 months. Analyses presented here use the sample design and analysis weights that reflect the US population from birth up to age 4 years. About 21% of the sample was Hispanic, 14% non-hispanic black, 56% non-hispanic white, and 8% other race/ethnicity. About 13% lived in households with an annual income level below $20,000 and 16% above $100,000. Almost half (46%) of mothers had a college degree or higher and about half (51%) worked outside the home. About one in three children participated in the Special Supplemental Nutrition Program for Women, Infants, and Children (30%). Since it is possible that higher socioeconomic mothers/ caretakers may have been more interested in responding to FITS, we compared the weighted sample characteristics for the FITS 2008 dietary sample to available statistics from the 2008 Current Population Survey (CPS) for households with children (22). The FITS 2008 sample appears to be somewhat more likely to be married (80% vs 66%), somewhat less likely to have completed high school (20% vs 26%), and more likely to have completed post-college education (45% vs 29%) compared with CPS data for women aged 18 to 44 years. The FITS 2008 income distributions are similar to those in the CPS, except that FITS 2008 has a lower percentage with annual incomes $100,000 (17% vs 26% in CPS) (22). This finding may relate to lower income levels among households with young children, since the available CPS data are for households with children younger than age 18 years (not households with children from birth up to age 4 years as in FITS). The analysis weights in FITS adjust for age and race/ethnicity among childbearing aged women; adjusting for these two variables is often considered to make additional adjustments for other socioeconomic status variables such as education and income unnecessary. STUDY LIMITATIONS The FITS 2008 is a large-scale, unique study that uses a national sample frame, strong design features, and recommended dietary data collection and analysis methodologies. In addition to the study strengths described earlier, it is important to note study limitations that should be kept in mind when interpreting study findings. Sample Representativeness and Sample Frame In planning the FITS 2008, it was important to use the same sample frame and study design approach as was used for FITS 2002 to facilitate comparisons in dietary findings between the two studies. It was also a costeffective way to find households with infants and young children since screening by telephone to find them is very costly. One limitation of the sample frame is that Experian s methods are proprietary and could not be fully evaluated before conducting the study. Despite using the same sample frame approach for FITS 2008 as in FITS 2002, the frame quality was less complete and the contact information was less accurate and current in In addition, greater cellular telephone use contributed to a lower rate of locating respondents; in 2002, 37% of households could not be located compared to 71% in Finally, although the weights used for analysis adjust for race/ethnicity, the underrepresentation of Hispanics and non-hispanic blacks and smaller sample sizes for these subgroups may limit the ability to conduct detailed age and race/ethnicity data analysis in the future. Response Rates In addition, we had to cope with a downward trend in the US population s participation in telephone surveys in general (19,20). In 2002, 73% of eligible households contacted completed the recruitment interview compared to 60% in In the past decade, national telephone surveys have experienced reduced response rates for several reasons: difficulty in locating households due to greater cellular telephone use, the public s concern about answering questions about their children, and the public s reduced interest in participating in telephone surveys. Making it clear to respondents that FITS 2008 was a research study and not a marketing survey may have helped response rates, along with professional research and study materials, and skill in gaining cooperation from parents. In 2002, 94% of households that completed the recruitment interview responded to the dietary interview compared to 78% in In 2002, 85% of respondents completing a Day 1 recall completed the Day 2 recall compared to 80% in To reduce bias from nonresponse, the analysis weights used in FITS 2008 take into account the nonresponse at different stages, and the results appropriately reflect the age and racial/ethnic distribution of children from age birth up to 4 years in the United States. As with any data analysis of national surveys, care should be taken in interpreting differences in dietary intake between FITS 2002 and FITS 2008 since the findings may reflect a shift in population demographics, real changes in the population s (or population subgroups ) food consumption, differences in sample characteristics, and/or changes in study methods. Other population studies, including NHANES, face similar challenges in assessing dietary trends and differences in food consumption over time (23). S24 December 2010 Suppl 3 Volume 110 Number 12

10 Table 4. Characteristics of the dietary interview sample for the Feeding Infants and Toddlers Study 2008 (n 3,274) % Standard Child characteristic error Sex Male Female Ethnicity Hispanic or Latino n-hispanic or Latino Race a White Black or African American American Indian or Alaska Native Asian, Hawaiian, or Pacific Islander Other Race/ethnicity n-hispanic white n-hispanic black Hispanic Other Receives Special Supplemental Nutrition Program for Women, Infants, and Children benefits Attends day care or preschool Ever breastfed Dietary supplement user (on day of recall) Low birth weight Birth mother s first born Age of mother at birth (y) Mother s education bc 9th grade th-11th grade Completed high school Some post secondary Completed college Some graduate work/degree Father s education bd 9th grade th-11th grade (continued) Table 4. Characteristics of the dietary interview sample for the Feeding Infants and Toddlers Study 2008 (n 3,274) (continued) % Standard Child characteristic error Completed high school Some post secondary Completed college Some graduate work/degree Parents marital status Married Separated or divorced Widowed t married Partners living together Mother works Annual household income $10, $10,000-$19, $20,000-$34, $35,000-$49, $50,000-$74, $75,000-$99, $100,000-$149, $150, Missing (not reported) a More than one race may be reported. b Highest year or grade of school completed. It was reported that 0.1% of mothers and fathers completed special education classes. c Includes biological and adoptive mothers. d Includes biological and adoptive fathers. Dietary Intake The dietary information is self-reported and associated with the reporting biases and measurement error inherent in dietary reporting (8,23). Another limitation of the dietary data is that breast milk cannot be precisely measured, but the volume rules we applied are based on the available peer-reviewed articles from highly industrialized countries. Results of the bridging study suggest that the improvements to the FITS 2008 dietary protocol did not have large overall effects on the energy or food estimates, suggesting that FITS 2002 vs FITS 2008 comparisons are reliable; however, the bridging study may have had a sample size too small to estimate small effects (eg, under 100 kcal daily). Smaller differences may also be important to consider in interpreting trends in food and nutrient intake between FITS 2002 and FITS 2008, but we cannot be sure that the difference may not be attributable to changes in study methodology. CONCLUSIONS FITS 2008 was conducted using a comprehensive set of data collection survey instruments and methods that were customized for the specific needs and considerations of the target age group to collect the best possible data on dietary intakes of infants, toddlers, and preschoolers. Subsequent articles in this supplement provide descrip- December 2010 Supplement to the Journal of the AMERICAN DIETETIC ASSOCIATION S25